the muthor • • • • :

AUTOMATIC PROCESS CONTROL

INDUSTRIAL INSTRUMENTATION
principles OF INDUSTRIAL
PROCESS CONTROL

Automatic
Process Control
Donald P. Eckman ....
Professor of Irntrurnmlaiion
Engineering
Case Institute of Technology

WJLEY EASTERN PRIVATE LIMITED^

NEW DELHI
 First U.S. Edition, 1958
Fourth Printing, 1967
First Wiley Eastern Reprint, 1972

Published by Anand R. KundajiJor Wiley Eastern Private Limited,

J 41 South Extension 1, New Delhi 110049 and printed by
Raj Bandhu Industiial Company, C 61 Maya Pun f[. New Delhi 27
Printed in India
 Preface

This book affords a study of automatic control with particular

emphasis, by way of example, on process control. The techniques of
analysis are used to the fullest extent, and enough detail is carefully
presented so that some of the more difficult problems in automatic
process control may be inspected rather closely.
The purpose book is to present to the beginning engineer the
of this
im porta nt principles of automatic control, beginning with process
analysis and carrying on into the generalized behavior of closed-loop
systems. System problems are given a great deal of study.
The book is intended primarily for the undergraduate in engineering,
and a knowledge of the elements of calculus, difTerential equations,
mechanics, thermodynamics, and fluid"' mechanics is assumed. LaPlacc
transform is not used except in the fast chapter, although its use may
be helpful from the beginning.
The teacher may wish to peruse the subjects in the topical order of the
book. However, other orders may be used. For a study beginning with
measurement and control, the following order of chapters is suggested:

1, 5, 3, and/or 0, 7, 2, and 4.

For a study having emphasis on the mathematical approach the follow-

ing order might be used:

1, Appendix, 4, 9, 10, 5, 7, 8.

An attempt has been made to keep the material as independent as

possible,but without i*epetition. Thus, the order may be made flexible

and the logic developed to individual requirements.

 VI Preface

The material appearing herein that was first published in PrincipleB

of Automatic Process Control has been completely edited and rewritten
and the material up-dated to include all recent %lvancements and de-
velopments in the field of process control.
I wish to acknowledge the many contributions made to the writing
of this text by colleagues at Case Institute of Technology and the many
students, both graduate and undergraduate, who have suggested topics
and problems and criticized not a few. In particular. Dr. Irving
Lefkowitz, the author’s associate, is to be thanked for many valuable
suggestions. Professor Lawrence G. Seigel made many contributions
to this work. Professor G. L. Tuve also, through his vast experience, is

responsible for improvements in presentation of the laboratory. Mrs.

D. P. Eckman is to receive much credit for manuscript and proof
preparation.

Donald P. Eckman
Clet'eland, Ohio
April, 1958

AUTOMATIC PHOCERS CONTROL is based in part on PRIN-

CIPLES OF INDUSTRIAL PROCESS CONTROL by Donald #V*IP
P.
Eckman, a book that was written and published while the author was
employed by the Brown Instrument Company, now^ operated as the
Brown Instruments Division of the Minneapolis-Honey well Regulator
Company.
 Contents

chapter one The Science of Automatic Control . . 1

two Process Characteristics . . . . 12

three Controller Characteristics . . 59

four Closed Loop in Automatic Control . 78

five Measuring (Feedback) Elements . 123

six Controlling Elements 155

seven Final Control Elements 195

eight Process Instrumentation 238

nine . Sinusoidal Analysis 289

ten Stability Analysis 299

appendix 325

index . . . 364

vii
 Variables Elements
Set point A Input elements
Reference input Gx Control elements
Actuating signal G2 System elements
Manipulated variable H Feedback elements
Controlled variable N Load elements
Feedback variable
Load variable
 chapter

The Science
of Automatic l.ontrol

Automatic process control is a principal part of the industrial progress

during what is now termed the second industrial revolution. The
increased use of the science of automatic control has come about through
evolution rather than revolution, an evolution that results from the
widespread use of the techniques of measurement and control. The uni-
versal recognition of the advantages of automatic control has been high-
d through increased study of automation.
Automatic process control is used primarily because it results in
economy of operation of industrial processes that more than pays for the
expense of the control equipment.. In addition, however, there are many
intangible gains such as the elimination of work not requiring a worth-
while mental efTort and a corresponding demand for work with a much
higher skill. Elimination of human error is another positive contribution
of use of automatic control.
The principle of automatic control, the employment, of a feedback or
measurement to actuate a controlling mechanism, is very simple. This
ma.y be seen by an inspection of the diagram on the facing page. The
same principle of automatic control is used in so many dilTerent fields
that it would be impossible to name them all a few of them are: —
(jhemical and petroleum process control, control of furnaces in steel
manufacture, machine-tool control, and missile guidance.
Understanding of the principle of automatic control is necessary in
modern, analytical engineering because its use is as common as the use
of the principles of electricity or of thermodynamics and should therefore
a most important part of the span of knowledge of the engine/*r The
use of analog and digital (computing machines has made possible applica-
1
2 Automatic Process Control

tion of automatic control ideas to physical fiystems that only a few years
ago were impossible to analyze or control
A clear concept of the principle of automatic control benefits the
engineer because, as pointed out by J. Von Neuman^, ‘There is no point
in using exact methods where there is no clarity in concepts and the
issues towhich they are to be applied.*^ Furthermore a need for auto-
matic control has often resulted in a careful and systematic analysis of
the physical part of the process.
Automatic control devices employ the principle of feedback within
themselves in order to better their performance and are therefore a
subject of study. Although this book will discuss automatic control
from a general standpoint, most of the examples are selected from process
control. Some discussion of the principle of operation of control mech-
anisms is presented in Chapters 6 and 7.

What Is Automatic Control?

Automatic control is the maintenance of a desired value of a quantity
or condition by measuring the existing value, comparing it to the desired
value, and employing the difference to initiate action for reducing this dif--
ference. Thus automatic control requires a closed loop of action and
reaction operating without human aid.
To illustrate the closed-loop action, consider the control of a home-
heating system as illustrated by Fig. 1-1. Suppose that it is desiuML^o

maintain the temperature of the home at 72 F. This temperature is

the desired value or set point} A
thermometer is installed on an inside
wall of the home and measures an existing room temperature. This
temperature is to be the controlled variable. A person watching the
* A glossary of automatic control terminology is included in the Appendix.
The Science of Automatic Control 3

thermometer notes that the temperature is 69 F and is therefore less

than the desired value. The actuating signal is 3 deg. An action is
then taken to reducg this difference by throwing a switch that turns on
the fuel gas to the furnace burners. The flow of fuel gas is the manipu-
lated variable. As the furnace heats, warm air is delivered to the room,
and the temperature will presently increase. In a short time the temper-
ature becomes too high, and the whole sequence must be repeated in
the opposite direction. The control action is characterized by the
closed loop from cdfctrolled variable to deviation to manipulated vari-
able to controlled variable as illustrated in Fig. 1-2.

Automatic control is accomplished by employing a machine to per-

form such dull and time-consuming tasks as watching a thermometer
and operating switches. The great sociological- economic advantage of
automatic control is that it frees human beings from menial tasks. The
technological advantage of automatic control is that a machine can
perform tasks more rapidly and consistently than a human being can.
The automatic controller is the machine employed to maintain the
controlled variable at the desired value. In the home-heating control
discussed above, a thermostat or thermostatic controller, as shown
schematically in Fig. 1-3, would be installed. The desired value of
temperature is set on the adjusting knob. The existing temperature is
measured by the bimetallic strip, which winds and unwinds slightly
with changes in temperature. When the temperature increases above
the set point, the bimetallic strip moves its contact to the left and
4 Automatic Process Control

breaks the electric circuit. This action de-energizes the solenoid valve
which turns off th5 gas to the furnace.
The most general type of automatic control system is shown in the

frontispiece. system under control, represent^ by G 2 and N, is

A
under the action of the controller, represented hy A, H, and Qy, The
value of the controlled variable c is measured, and the feedback or meas-
uring element transmits a signal to the control elements. The feedback
variable b is compared to the reference variable r, and the difference is

thr actuating signal The reference variable is provided by the input

^dement and is directly related to the set point v. The controller opei-
ates on the system through the manipulated variable m.
"J'h(;r(‘ are two general problems in automatic (jontrol. The first of
these is tli(‘ control to n*duce the effect of load disturbances Referring
again to the frontispiece, the set point is fixed at a desired value and is

not a function of time;

V = V 0“1)

The load variable, however, is changing either continuously or sporadi-

cally with .some function of time
-2
( 1 )

The automatic controller must be designed to prevtmt the load disturb-

anees u from alfeeting the value of the controlled variable. The control
of load disturbaiieos is generally a problem of automatic process control.
Load-disturbance control is best illustrated by a process water heater
for wliicli the operational diagram is shown in Fig. 1-4. The outlet
temperature of the heater c is measured and operates an automatic
controller which adjusts the flow of heat m into the heater The purpose
of the automatic controller is to maintain a constant outlet temperature
 The Science of Automatic Control 5

in spite of variations in the inlet water teini)erature u. This kind of

automatic control is characterized mainly by a fixed set point but a
variable load. •

Fro 1-4. Automatic control emplo\ r*<l lor load van.ifion.

The second problem in automatic control is the control of the system

variable to follow a desired set point. In this problem the load variable
is usually fixed and is not a function of time;

u ^ U (1-3)

The set point, on the other hand, is changinj^ in a continuous or dis-

continuous manner,
V = 9{t) (1-4)

The automatic controller must be designed to force the controlled

variable to follow the changes in set point as closely as possible. F ollow-
ing control of this kind is generally a problem of servomechanisms and
is not as often encountered in automatic process control.

Fin. 1-5. Automatic control employed for following.

Following control is best illustrated by the ship-steering mechanism

of Fig. 1-5. For a very large ship it would be impossible to operate the
6 Automatic Process Control

rudder by human hands alone. Consequently, a small wheel at the

steering station, the position of which is the desired value v of rudder
position, transmits a signal r to the steering m^hanism. The latter
often consists of hydraulic power devices which move the rudder to the
desired position. The actual position of the rudder c is compared to
the reference value and any difference operates the hydraulic control
r,

mechanism. This kind of automatic control is characterized mainly by

a variable desired value but a fixed load. Following control is employed
in automatic process control, for example, when heat-treating furnaces
are required to have a temperature raised and lowered according to
a certain time .s(!hedule.

Automatic Control in Industry

The first use of automatic control seems to have been the flyball
governor on Watts’s steam engine in about 1775. This device was
employed to regulate the speed of the engine by manipulating the steam
flow by means of a valve. Thus all the elements of feedback are present.
lOarlier in history there does not appear any known reference to the

use of automatic control, and it does seem incredible that in the ancient
water works of Home, (J recce, or Egypt no recorded use of closed-loop
water regulating devices was made. Even if there was, there is no dis-
cussion of the philosophy of automatic control. j
The first analysis of automatic control is the mathematical discussion
of the flyball governor by James Clerk Ma.xwell in 1868. A set of
historical references is provided at the end of the chapter. Further
application of the governor techni<iue.s to other engines and turbines
was made, and in the early 1900’s the application to process control
began. At the same time regulators and servomechanisms were being
studied for their application to steam-power regulators and ship
steering
The first general theory of automatic control was founded by Nyquist
in the famous article on “Hegeneration Theory.” This study provided
the basis for the determination of stability of systems without having to
solve com()lot('ly the differential equations. Further developments in
electric mnplifiers and electric servomechanisms provided the many
locus and fre(|iiency techniques in use today.
The general applications of process control did not begin until the
193()’s. The usefulness of control techniques quickly established their
value, so that by the 1940’s rather complex control networks were in
common use.
Automatic control devices are used in almost every phase of indus-
trial operations. They are commonly employed in
The Science of Automatic Control 7

1. Processing industries such as petroleum, chemical, steel, power,

and food for the control of temperature, pressure, flow, and similar
variables. •
2. Goods manufactures such as automobile parts, refrigerators, and
radios for the control of assembly operations, work flow, heat treating,
and similar operations.
3. Transportation systems such as railway, airplanes, free missiles,
and ships.
^
4. Power machines such as machine tools, compressors and pumps,
prime movers, and electric power-supply units for the control of posi-
tion, speed, and power.

Automatic control devices are used becau.se their application results

in economical behavior of the system under control or because they are
required for humanitarian purposes. There are many advantages to
the widespread use of automatic control. Some of these are:

() Increase in quantity or number of products.

() Improvement in quality of products.
(c) Improvement in uniformity of products.
(d) Savings in processing materials.
(e) Savings in energy or power requirement.
in plant equipment.
4. ID Savings
(g) Decrease of human drudgery.

These factors generally lead to an increase in productivity.

The widespread application of automatic control in industry has made
necessary the upgrading and education of a large class of semiskilhHl
workers to a higher skill, that of operating and maintaining instrumenta-
tion and control equipment.

Cybernetics und Instrumentation

The science of cybernetics or instrumentation is concerned with the
phenomena of communication and control whether in nature, in man,
or in machines. As shown in Fig. M), there an‘ two division.^ of
work in the field of instrumentation
1. The study of communication and information theory.
2. The study of control and feedback theory.

The important laws of communication and control deal with informa-

tion concerning the state and behavior of systems and are not primarily
* N. Weiner, Cybernetics, John Wiley and Sons, New York, 1948.
S. Tsien, Engineering Cybernetics, McGraw-Hill Book Company, New York,
1954.
 8 Automatic Process Control

concerned with energy or its transfer to or within a system. The use

of energy is only secondary to the main purpose of control or
communication. •
Communication and information theory*’^ is based upon the concept
that all ideas can be expressed in messages translatable into a common
language. The amount of information can be defined and therefore

Instrumentation
cybernetics

Communication Control
and and
information feedback
theory theory

Fig, Mi Science of instrumeD tat ion

ineaHured, and, a« a consequence, laws governing the transmission of

information can be stated. The te(*hnology of measurement, telemeter-
ing, television, language structuie, number systems, and automatic
computation employ the basic ideas of information and data handling
and processing. ' ^

Control of processes Control of manufacturing Control of power

involving change of involving change of above the level
material state physical shape ofa few watts

Fig. 1-7. Technology of instrumentation.

The field of automatic control from the subjective viewpoint can be

divided into three divisions as shown in Fig. 1-7. These subjects are
not mutually exclusive but are found often in industry.

^ C. £. Shannon Mathematical Theory of Communication/' Bell System Tech.

./., Vol. 27, 1948, pp. 379, 623.
*S. Goldman InfarnuUaJon Theory, Prentice-Hall, New York, 1966.
The Science of Automatic Control 9

Information and Energy

Communications #nd control is primarily concerned with the manipu-
lation of information concerning physical systems. Consider the system
shown in Fig. 1-8. The physical diagram is that kind of diagram used
in engineering to depict the important fea-
tures of the system. In thermodynamics
it is system diagram, in mechanics
called a
it termed a fr^-body diagram, and m
is

other sciences it is termed differently. The

main purpose of suidi diagrams is to show
the flow of material and energy (mass and
energy transfer). The physical diagram in
Fig. 1-8 shows that there is an inflow 7n,

an outflow q, and fluid is stored in the

vessel.
The block diagi am illustrates the behavior
of the system by depicting the action of the
variables of the system. The circle* repre-
sents an algebraic function of addition as is k
seen in Fig. 1-9 The rectangular box repre- T

sents a dynamic function such that the vari- Physical diaKram

stBie^X, called the output, is a function of for system analysis,
time and is also a function of the variable y,
called the input. Notice that the terms input and output refer to sig-

nals and not necessarily to mass or energy flow.

Algebraic function Dynamic function

xstz^y XBf(y,t)

Fin 1-9. Block-diagram elements.

Fig 1-10 Block diagrams for proce8.s of Fig 1-8.

The block diagrams of Fig. 1-10 correspond to the system shown in

Fig. 1 -8 . k not unique and its arrangement depends

block diagram is

upon the viewpoint of the analysis. The diagram on the left is drawn
10 Automatic Process Control
as though the outflow is the important variable being influenced by
inflow. The feedback occurs because an increase in outflow feeds back
to decrease head h. At the same time an increase iH inflow will increase
the outflow. Thus, the relationship of variables may be written as

q = /(A, t) (1-5)

h = g(m - q, t) (1-6)

Therefore, the overall relationship is

g - q, 0], (1-7)

The important point is that the block diagram illustrates the relation of
these variables.
The block diagram on the right in Fig. 1-10 is for the same process
but is shown to head in the vessel
illustrate that inflow influences the
and that the outflow is the secondary variable. This diagram does not
relate one to one to the mass flow because tlie input is flow and the
output is head. Recombining equations 1-5 and 1-6, the overall
relationship is

h^gWm -f(hj)lt\ (1
~8 )

There are only a few rules for the use of block diagrams:

1 Blocks and circles may be rearranged without destroying vaJitljlv

if the system can be described by linear differential equations with
constant coefficients.
2. Only one line may enter, and only one line may leave a rectangular
block.
3. Only two lines may enter, and only one line may leave a circle.

The block diagram is used as an aid in visualizing the behavior of a

system and in obtaining a solution of the problems.

HISTORICAL REFERENCES
Tht* followiriK refereiK’es are selected for the contribution each makes to the
knowledpje of automatic control. Prior to 1900 there are many articles on governors,
but most of them provide consitleration of iiidi\ idual detailed problems.

1808 J. C. Maxwell, “On Cov^crriors/’ Froc Hoyal Soc. London, \ol. 16, p. 270.
1893 A. Stodola, “Uber die Itegulierung von I'uibineii," Schweiz. Bauztg., Vol. 22,
p. 117.
1907 A. M. Liapounov, “Probl^me g^n^ral de lu stabihU* du inouvement,” Ann.
fac. set. univ. Toulouse
1911 It von Misea, “Itegulierung dea Maachinemganges," Encyl. der Math. ITiss.,

hand IV, 2 Teilband, p. 254.

1919 W. Trinka, Governors and the Governing of Prime Movers, D. Van Noatrand
Co., New York.
The Science of Automatic Control 11

1921 M. Tolle, Regdung der KTaftmaschinen, J. Springer, Berlin.

1022 N. Minorsky, “Directional Stability of Automatically Steered Bodies/’ J.
Am Soc. Naval^Engrat.y Vol. 34, p. 280.
1026 T. Stein, “Regelung und Auegleich in Dampfanlagen/’ J. Springer, Berlin.
1030 G. Wiinsrh, “Regler fur Druck und Afenge,” R. Oldenbourg, Munich.
1932 H. Nyquist, “Regeneration Theory,” Bell System Tech. J., Vol. 13, p. 1.
1932 M. F. Behar, Fundamentals of Instrumentation, Instruments Publishing Co.,
1934
Pittsburgh.'
1033 J. J. Grebe, R.II. Boundy, R. W. Cermak, “The Control of Chemical Proc-
esses.” Tran9^Am. Soc. Chem. Engrs., Vol. 29, p. 211.
H. S. Black, “Stabilized Feedback Amplifiers,” Bell System Tech. J., Vol. 13,
p. 1.

1934 H. L. Ilazen, “Theory of Servomechanisms,” J. Franklin Inst., Vol. 218, p. 279.

1934 A. Ivanofl, “Tlieoretical Foundations of the Automatic Regulation of Tem-

perature.” J Inst Fuel,\o\ 7, p. 117. .

1935 S. D. MiterelT, “Principles Lbiderlying the Rational Solution of Automatic

Control Problems,” Am. Sor. Mech Engrs., Trans., Vol. 57, p. 159.
1936 A. Calleiidar, D. R. llartree, A. Porter, “Time Lag in a Control System —I/*
Phil. Trans., Royal Soc. London, Vol. A235, p. 415.

PROBLEMS

1-1. Give examples of the use of automatic control devices for (a) reducing
the disturbances due to load variables in a system; (6) human safety.
» W, Name and describe briefly several automatic control devices found
in the home.
1-3. Name and
doKcnbe briefly several instruments found in the home.
1-4. Sketch a block diagram for a man in a boat with an outboard motor
performing the operation of steering for a point ashore.
1-5. Describe a device that could be employed to maintain the constant
speed of an automobile.
 Process *•

Characteristics

A process, in the science of automatic control, denotes an operation

or series of operations on fluid or solid materials during which the mate-
rials arc placed in a more useful state. The physical or chemical state
of the materials is not necessarily altered. In chemical engineering,
a unit process involves a change of chemical state. In mechanical
engineering, a process usually produces a change in physical state. ^ ••

Many external and internal conditions affect the performance of

a process. These (conditions may be expressed in terms of process
variables sucli as temperature, pressure, flow, Ikiuid level, dimension,
weight, volume, etc. The process may be controlled by measuring a
variable repre.senting the desired state of the product and automatically
adjusting one of the other variables of the process. Ambient conditions
must always be inelud(*d in the list of process variables.

Process Variables

The controlled variable of the process should be that variable which

most form or state of the product. Direct
directly indicates the desired
controlfrom product quality is most likely to insure proper performance
of the process and to produce and maintain the desired quality of
product. Consider the water heater in Fig. 2-1. The purpose of the
heater is to maintain a supply* of heated water. The variable most
indicative of this purpose is the temperature of water at the heater out-
let, and this is selected as the <x)n trolled variable.
Indirect control from a secondary variable of the process may be
necessary w^hen direct control is difficult to accomplish. For example,
an annealing furnace is designed to produce properly annealed metal
12
Process Characteristics 13

parts, and therefore the controlled variable should be the annealed

condition of the metal. However, this measurement is very difficult to
make with simple^ equipment, and it is necessary to select furnace
temperature as the controlled variable. It is assumed that the annealed
condition of the metal is directly related to furnace temperature.
Indirect control of a process is usually not as effective as direct control,
because a definite and fixed relationship does not always exist between
the secondar}'^ controlled variable and the form or state of the product.

I’kj. 2-1, A watcr-heatJfiK proct^ss.

tm ••

The manipulated variable of the process is that variable which is

selected for adjustment by the automatic controller so as fo maintain

the controlled varial)le at the desired value. The manipulated variabhi
may be any of the process variables that causes a fast response of the
controlled variable and is rela(iv(*ly easy to manipulate. For the water
heater of Fig. 2-1, the h(;at flow* m should be manipulated by the con-
troller. It is possible, but not as praefieal, to manipulate the water
throughput rate w or the incoming \\ater temperature ii.

The load variables of the process are all other independent \'ariables
except the controlled variable and the manipulated variable. For the
water heater of Fig. 2-1, the incoming water temperature v is the load
variable. It is expected that the automatic controller will (ujrrect for

fluctuations in load variables and maintain the controlled variable at

the desired value.

Process Degree of Freedom

The state of a process or the configuration of a sy stein is determined
when each of its degrees of freedom is specified. Consider, fur example,
a ball placed on a billiard table. In order to specify its position, we
14 Automatic Process CJontrol
would require three coordinates: one north-south coordinate, one east-
west coordinate, and the height. However, the height is not arbitrary
because it is given by the height of the table surfa(jp above a reference
plane. Consequently the ball has two degrees of freedom. This result
is derived from the following relation:

n = Hr - Ue (2-1)

where n — number of degrees of freedom

- number of variables of th<» sy.stem ••
riv

rif number of defining equations of the system

In the example of the billiard ball there are three variables of position,
one defining equation (height = constant), and therefore two degrees
of freedom.
Similarly, a process has a finite number of degrees of freedom. Con-
sider the heat exchanger of Fig. 2-1, There are four variables:

w = inlet temperature
c = outlet temperature
w = water flow rate
rn — heat input rate

and There is one defining equation obtained from conservation

71^ -= 4,

of energy law of thermodynamics).

(first Therefore, Tie = 1.
number of degrees of freedom are
n - — Tie ~ ^
— 3

System variables and system parameters must be carefully distinguished.

For example, the weight of water contained in the heater of Fig. 2-1
And the specific heat of water are parameters not variables.
The number of independently acting automatic controllers on a system
or process may not exceed the jinmbcr of degrees of freedom. To illustrate
this rule, suppose that tlirec automatic controllers are installed on the
heater of Fig. 2-1 so that inlet temperature u is maintained constant, the
water flow rate w is maintained constant, and the heat input rate m is

maintained constant. Then all three degrees of freedom are specified

and outlet temperature c must be constant. More than three automatic
controllers cannot be employed because redundancy would exist (only
four variables and five equations). However, fewer than three auto-
matic controllers are adequate, and, in fact, only one controller is usually
employed.
For chemical processes involving separation, distillation, or fraction-

ation where heterogeneous equilibrium exists and where each component

 Process Characteristics 15

ispresent in each phase, a modification of the rule of equation 2~1 may

be derived. It is known as Gibbs’s phase rule,

• n = Ue - rip +2 (3-2)

where n = number of chemical degrees of freedom

rie = number of components
Up = number of phases

This applies only to vhe chemical states of the process, and the number
2 in the equation atJove represents temperature and pressure. For an
isothermal process
n = Tic - rip + I (2-3)

and for a constant-pressure process

w = w, - Tip + I (2-4)

For example, consider a steam boiler producing saturated steam. The

number of components is one (water), and the number of phases are
two (liquid and gas). Therefore, the number of degrees of freedom are'
w = 1 - 2 + 2 = 1

and either temperature or pressure (but not both) may be selected as

the independent variable. For a boiler producing superheated steam
tlffe trumber of degrees of freedom are two, and both temperature and
pressure must be controlled.

Example 2-1. P\)r the liquid-to-liquid heat exchanger shown in the figure,
the following variables are given below the diagram:

c = heated fluid outlet temperature

m= heating fluid flow rate
ui - heating fluid inlet temperature
U2 = heating fluid outlet temperature
U3 = heated fluid flow rate
U4 = heated fluid inlet temperature ^

Calculate the maximum number of independent controllers that could be used.

 :

16 Automatic Process Control

There are six variables:
n, =* 6

There is one equation (conservation of energy)

He = 1

The number of degrees of freedom are. assuming no boiling of liquids,

?i = () — ] — 5

and no more than live automatic controllers may be used.

Example 2-2 . A binary mixture (benzene and toluene) is to be distilled

at atinosjiherie pressure (figure at left). The variables aie-
•"
-— //
== overhead flow
To condenser =
r. overhead temperature
m = heat input rate

('aleulate the degrees of freedom.

m. heat input For the distillation at constant
pressuH'

V rir
~ 1

The number of components me two. and the number of phases are two. There-
fore
?? - 2 — 2 d- 1 - 1 (temperature v)

For the process, the number of variables are 1hre(‘ (7, r. and ?«), and the defin-
ing relations.'in* two (conservation of eiKTgy and mass), therefore the ni*niW‘r

of degrees of freedom is

n - n,. — rt, -- — 2 --
1

Thus, no more than one automatic controller mav be employed.

Example 2 -H. A binaiy mixture is to be distilled by the continuous

process (fractionation) which is shown in the figure below Tlie variables
are related to the supply of mateiial and energy an<l are listed below:
 :

Process Cliaracteristios 17

Cl = overhead temperature Ui
~ bottom flow rate
cj = overhead pressure Ui feed temperature
C| = overhead coi^osition u« = feed pressure
a « overhead flow rate uj * feed composition
ui * bottom temperature u% =* feed per cent vapor
ui - bottom pressure Ui - feed flow rate
uz - bottom composition m = steam flow rate (heat input)

Note that the process may contain internal heat exchange or any other
arrangement of equipment. Also the pressures 12, and Ur are generally
the same. Calculate iFie number of degrees of freedom. What are the maxi-
mum number of automatic controllers that may l>e used?
Employing Gibbs’s phase rule at the overhead there are two com})onents
and two phases'
n - 2 - 2 -h 2 = 2

so that any two of the three variables (ci, r^, c^) are independent. The same
IS true at the bottom where any two of the three variables (ui, Ui, uz) are

independent. For the feed also any two of the three variables (ws, ur, uj)
are independent.
The degrees of freedom may now be calculated. The number of variables
are eleven
Ovei head— any two of ci, a, c^; and
—
Hottoin any two of wi, uz, and w.4
Heat input -"W
Feed --any two of 1/5, ut, uj; and Ug, uv

The number of equations are three (conservation of each mass or continuity,

and conservation of energy). Then
- 1 1 — 3 = 8 deg of freedom

No more than eight automatic controllers may he used.

(^huraclerialicH of Physical Syaleins

The basic concepts for system analysis in automatic control are llie
fundamental laws relating the behavior of the system. Statical or
steady-state behavior is important, but even more important in auto-
matic control is the dynamic behavior of mechanical, fluid, thermal,
and electric .‘systems. The important characteristics of fluid and thermal
processes are their resistance and capacitance.
Electric systems composed of pure resistance elements and jiure
capacitAiice elements in various combinations have many character-
istics in common with fluid and thermal processes Consequently a
study of electric systems serves to clarify many similar ideas in fluid
and thermal systems
The flow law for electric conductors (Fig. 2-2 ) having resistance but
no inductance or capacitance is Ohm's Law,

fi — to = iR ( 2- 5 )
 Automatic Process Control

where v = electric charge, coulombs. Electric capacitance in the usual

range of voltage and charge does not depend upon the voltage or charge.
Electric systems of resistance and capacitance may be analyzed by
employing these two laws.
Liquid systems composed of liquid-filled tanks or vessels and con-
nected through pipes, tubes, orifices, valves, and other flow-restricting
devices may be analyzed by using the fundamental laws governing the
flow of fluids. Liquid tanks are considered to have a free surface of
liquid. Connecting pipes are assumed to be full of liquid. In liquid-
filled systems it is also assumed that fluid accelerations are small, that
is, steady flow or nearly steady flow persists.
The flow laws for liquid-conducting elements as in Fig. 2-3 are of two
different types: turbulent flow for which the Reynolds number is greater
 Process Characteristics 19

than about 4000, and laminar flow for which the Reynolds number is
lessthan about 20^. fn turbulent flow through pipes, orifices, valves,

•. I

R, resistance C, capacitance

Fiq. 2-3. Fluid (liquid) resistance and capacitance.

and other flow-restricting devices in general, the flow is found from

Bernoulli’s law and may be reduced to

q = KAV2g(hi - h2 ) (^9)

where q = liquid flow rate, ftVsec

K = & flow coeflficient (usually about 0.6)^
A = area of restriction, ft^
g = acceleration due to gravity, ft/sec^
h = head of liquid, ft
TV

This law may be compared to Ohm’s law for electric conductors equa-
tion 2-5, the basic difference being that it involves the square root of
the head or potential.
The turbulent resistance is found from

if the flow coefficient

®
7?
= ^=
K
^
is considered constant.
^ =
^

Therefore the turbulent

^2) /g,j_2 /r» -i/w

resistance not constant but depends upon the flow rate and head
is

differential existing at any time. Consequently it is necessary to define

a turbulent resistance at a particular value of flow and head and to
employ this value of resistance over a narrow operating range. For
every new operating range, a new value of resistance is required. In
many instances the values of the flow coefficient or area are not known.
The resistance may be determined graphically by plotting head against
flow and calculating the slope.

* For pipe flow the pipe-friction formula is usually employed where K

=» l/\^FL/D,

F ia the friction factor, L is the equivalent pipe length, and D is the internal diameter
of the pipe.

II
20 Automatic Process Control
Example 2^4. The flow verBUs differential head for a 1-in. orifice is

plotted in the adjacent figure. The orifice coefficient is assumed to be constant

at 0.6 except for very low flow where the Reynolds nunfoer is small.

Flow, q, ft^/sec

The resistanro may be calculated at any particular head and flow and will
vary directly with flow. The resistance at any head may be calculated from

The resistance for different heads is shown m the following table:

Head, ft Resistance, soc/ft"

16 306
18 324
20 342
22 358
24 374

head averages about 20 ft, the resistance changes only plus

If the differential
10 to minus 10 per cent for a 4-ft variation in head.

For laminar flow in general, the flow in circular tubes or pipes is

found from the Poiseuillc-Hagen law:

^ ^
128mL mvL
.tB-
where h — head, ft

V = kinematic viscosity, ftVsec

L = length of tube or pipe, ft
(211)
D— inside diameter of. pipe, ft
^

q = liquid flow rate, ft^/sec

p = absolute viscosity, Ih sec/lt^ = yv/g
y - fluid density, Ib/ft/^

The laminar flow' law is directly comparable to Ohm’s law because

the flow (current) is directly proportional to tlie head (potential).
Process Characteristics 21

Unfortunately, laminar flow is not often encountered in industrial

practice.
The laminar resistance is found by

sec/ft^ (2-12)
dq

Laminar resistance is constant and is directly analogous to electric

resistance.
The conlimiity ftiac or law of conservation of mass for a tank contain-
ing liquid, as in Fig. 2-3, is

4^ = 9
at
(2-13)

where A = cross-section area of tank at liquid surface, ft^

h = head, ft

t = time, sec

q == flow rate into tank, ft^/sec

'Fhis law should be compared to the charging law for electric capacitors

(equation 2 7).
Liquid capacitance is defined bv

C = (2-14)
. dh

where v = volume of liquid in tank, ft'^ Liquid capacitance of a ta-nk

is therefore equal to the cross-section area of the tank A taken at the
liquid surface. If the tank has a constant cross-section area, the liquid
capacitance is constant for any head.
Gas systems consisting of pressure vessels or chambers and various
connecting pipes, vah es, etc. may be analyzed by using the fundamental
laws for the flow of compressible gases. For systems in which the pres-
sure differentials are less than about 5 per cent of the static pressure, the
compressibility is not usually important and they maj’^ be treated as
liquid systems. Ventilating and other air-transport systems where
changes of air density are small may be treated as incompressible flow
systems. Cienerally, however, the pressure differentials are substantial
in most industrial problems.
The Jtuw laws for gas in conducting elements (Fig. 2-4) must also be
given for turbulent and laminar flow. For turbulent flow through
and valves, the steady-flow energy equation (first law of
pipes, orifices,
thermodynamics) lor adiabatic flow of ideal gases is

w -- KA Y V2g(i>i - P2)7 (2-15)

22 Automatic Process Control

where »
tr gas flow rate, Ib/sec
=
/iC a flow coefficient
A = area of restriction, ft*
Y - rational expansion factor
7 = gas density, Ib/ft*
p = pressure Ib/ft*

Turbulent gas-flow resistance is, therefore,

R = ^
dw
sec/ft*
**
(2-16)

This resistance is not easily calculated because the expansion factor Y

depends considerably upon pressure. In this case it is easier to determine

R, resistance C, capacitance

Fio. 2-4. Fluid (Raa) resistance and capacitance.

resistancefrom a plot of pressure against flow for any particular device.

Laminar gas-flow resistance of tubes and pipes can be calculated fron>
the Poiseui II e-Hagen law (equation 2-11).
A special case of gas flow occurs when the pressure differential is
larger than a critical value. In such instances, critical or “sonic*' flow
is obtained when the velocity in the orifice is acoustic velocity and down-
stream pressure waves cannot propagate upstream. '^The flow rate,
therefore, depends only upon upstream pressure and temperature

w = (KaA/\^Ti)pj where the coefficient Ka is different for each gas

and Temperature Ti is the upstream temperature and pressure p
orifice.

is the upstream pressure.

The continuity law for a pressure vessel (Fig. 2-4) may be written

where C= gas capacitance, ft*

p = pressure, Ib/ft*
t = time, sec
=
11? flow rate, Ib/sec

It is assumed that the volume of the vessel is constant.

 :

Process Characteristics 23

Ooi capacitance is defined by

- 18 )
(2

where v = weight of gas in vessel, lb. The capacitance of a pressure

vessel must be calculated from thermodynamic relations because the
gas expands from a region of high pressure into the vessel at lower
pressure, or expands from the vessel into a region of lower pressure.
A polytropio expansii^p process is assumed so that the chai.^e of state
of the gas lies along some path between an isothermal and adiabatic path,

—
V
= constant ( 2 - 19 )

where p = gas density, Ib/ft^

n = poly tropic exponent (n = 1.0 for isothermal expansion and
n = ratio of specific heats for adiabatic expansion)

The gas capacitance can now be calculated by employing the ideal

gas law
dv
C = ( 2~ 20 )
dp

where V = volume of vessel, ft’^

'
R = gas constant for a specific gas, ft/deg
T = temperature of gas, deg

Numerous tests show that the poly tropic exponent n is approximately

1.0 to 1.2 for uninsulated metal ve.ssels at common pressures and tem-
peratures. Therefore the gas capacitance of a vessel is constant and
is analogous to electric capacitance.

Example 2-5. Calculate the gas capacitance of a 20-gal pressure vessel

containing air at 200 F.

C =
V 20 X 0.134
= 0.000076 ft*
fiRT 1 X 63.3 X 660
Employing moles and the universal gas constant R,

vVi

where p = pressure, Ib/ft*

V\ - volume, ft*
V = weight, lb
M = molecular weight of gas, Ib/mole
R - universal gas constant, Ib-ft/mole deg R”‘ = 1546
T temperature, deg R
 )

24 Automatic Process Control

Then

dp"" RT
29 X 20 X 0.134 =
C = 0.000076 ft*
1646 X 660
Thermal systems involving heat transfer from one substance to another
are also characterized by resistance and capacitance. In thermal
systems it will be assumed that substances characterized by resistance
to heat flow have negligible storage of heat (as a very thin air film),
and substances characterized by heat storage have negligible resistance
to heat flow (as a block of copper or aluminum).

R, resistance C, capacitance

P’lci. 2-5. Thermal re.sistaiice and (iapacitance

The fljoxv laws for heat conductors (Fig. 2-5) are of two diflerent types
corresponding to conduction or convection, and radiation. For conduc-
tion of heat through a specific conductor, the heat flow is given by the
Fourier law which reduces to

KA
TZiOi -h) (2-2\)

where q = heat flow, Btu/.sec

K- thermal conductivity, Btu/ft sec~* deg~^
A = area normal to heat flow, ft^

AX = thickness of conductor, ft

6 = temperature, deg

For convection heat transfer,

q^^HAiSi-do) ( 2- 22 )

where H= convection coefficient, Btu/ft^ sec”^ deg“^

The thermal resistance is, therefore,

„ dS AX
—
« = -- = ,

- deg sec/lHu (conduction

^
(2-23)
dq KA
Process Characteristics 25

or

R= —
Jiq
= —
HA
deg sec/Btu (convection) ( 2* 24 )

Thus, the thermal resistance is constant if thermal conductivity K is

constant or if the convection coefficient H is constant. Generally these

factors arc nearly constant and thermal resistance is directly analogous
to electric, resistance.

Example A*!;tpady heat flow occurs through a J-in. thick carbon

2-6.
an area of 2.0 sq ft. The temperature drop is from 330 F to
steel plate with
180 F. Cfllculatc the resistance. The thermal conrluftivity is 29. Btu/ft
deg“' hr”’ For steady-state heat conduction.

dim ^
(AA’)
^
~
0.125 3600
0.65 deg sec/Btu
,lq
~ KA 12 ^ 29 X 2

For radiation heat transfer, the flow is given by the Stefan-Boltzmann

law for a surface receiving radiation from a tlack body:

7 = KAEidi^ - $2*) ( 2 - 25 )

where q = heat flow, Btu/sec

/C = a constant = 47.4 X 10”*^ Btu/ft^ sec~* deg""^

E= emissivity
A - suiface area, ft^
0 = temperature, deg R
This is a fourth-power law as compared to a square-root relation for
turbulent fluid flow, and a linear law for electric current.
The radiation resisinnee is

(je
R ( 2 -26 )
dq

where 6a = average of radiator and receiver temperatures. Radiation

resistance varies inversely as the cube of temperature and must be
employed for a small range of temperature. Actually, the resistance
calculated above is not seriously in error even when the source and
receiver temperatures differ by a factor of two.

Example 2-7. A steady flow of lieat occurs in an electrically heating

furnace with walls at 1800 F to a large steel casting at 1400 F. The surface
area is 1.0 sq ft. Calculate the resistance. Assume emissivity is one.

d6 1
^
dq 4KAEej * 4 X 47.4 X 1.0 X 20J10®

fl - 60 deg sec/Btu
26 Automatic Process Control

The temperature change law is a consequence of the first law of thermo-

dynamics relating heat and internal energy. For the metal block of
Fig. 2''5, heat into the system raises the internal energy and for a given
system,

C^
at
= q (2-27)

where C = thermal capacitance, Btu/deg

t = time, sec
••

The thermal capacitance is, therefore,

C= TFCp (2-28)

where W ^ weight of block, Ib

Cp — specific heat at constant pressure, Btu/deg lb”*

Thermal capacitance is directly analogous to electric capacitance.

Example 2-S. Calculate the thermal capacitance of a 30-gal water tank.

From equation 2- 28
C = WCp = 30 X 8.3 X 1 - 249 Btu/deg
Note: It is sometimes necessary to include the capacitance of the vessel
walls.

Thermal systems do not always have isolated resistance and capaci-

tance. A thick stone wall, for example, has resistance to flow of he6^,
by virtue of finite thermal conductivityand capacitance because of the
specific heat of the material.Such systems arc^ said to have distributed
capacitance and resistance and the ccpiations above do not strictly
apply. Fortunately there are many physical problems in which thermal
resistances and capacitors are sufficiently isolated so that some calcula-
tions of characteristics can be made.
Electrical, thermal, liquid, and gas characteristics are summarized
in Table 2-1.

Table 2-1 . Dimensions of Process Characteristics

P^lectrical Thermal Liquid Gas
Quantity coulomb Btu ft* lb
Potential volt deg ft lb/ft»
Time sec sec sec sec
Coul.
Flow = amp Btu/sec ft’/sec Ib/sec
sec

Capacitance —
Coul.
volt
TT * fstrad Btu/deg ft*

volt
Resistance = ohm ,
deg-sec/Btu sec/ft* sec/ft*
amp
Process Characteristics 27
Mechanical systems may be composed of various masses connected
by members providing damping effects and spring effects. Although
some systems are analogous to electric systems, the analog comparison
may be made in many ways. Consequently, it is better to consider
mechanical systems on their own merits.

Fio. 2-6 Characteristics of mechanical systems.

Newton^s second law of ipotion defines the relation between a mass,

and an applied force. As in Fig. 2-6,
its acceleration,

dt\ di)
= M dt^
(2-29)

where / = force, lb
i = time, sec
M = mass, lb«secVft
X = displacement, ft

The mass is considered constant.

Damping is provided by piston dashpots as in Fig. 2-6 and by well-
lubricated surfaces moving over one another. For an oil-filled dashpot
the force resisting a velocity of motion arises from the pressure differen-
tial at the piston recjuired to cause laminar flow through the capillary

fluid resistance. A damping law is given by

/-«! (MO)

where B= damping coefficient, lb sec/ft. Damping is also provided by

the shear forces of viscous fluids acting against a moving surface.
Spring action as in the coil spring of Fig. 2-0 and for cantilever and
pinned-end beams may be described by a form of Hooke's law,

/ = /iTx (2-31)

where K= spring gradient or modulus, Ib/ft. Basically, a spring

action results from compressive, tensile, or shear stresses below the
elastic limit.

Elements of Process Dynamics ^

Dynamic analysis is necessary in automatic control because automatic

control is baaed upon dynamic action. Whereas dynamic or transient
 m I
D C

Block diagram

I' lG 2-7 The proportional element.

Analysis of processes is made easier through a study ot the elements

of process dynamics Tljese elements are

1 Proportional element
2. Capacitance element.
3 Time-constant element
4 Oscillatory element
Process Characteristics 29

These are the '^building blocks’" of processes and almost all industrial
processes are described by arrangements of numbers of these elements
in various series and parallel arrangements.
The 'proportional element is described in Fig. 2-7. Consider, for
example, the capillary shown in the physical diagram. The flow rate
of liquid m through the capillary is regarded as the variable being
(‘hanged and thus is termed the input variable. The head c is the
!-csult variabl^Jow rate and is regarded as the output variable.
of the
A capillary constitutes a laminar resistance and the flow-head equation
may be written
c = Rm (2-32)

where c = output variable (head)

R = resistance
m — input variable (flow)

This relation may be considered in the following form:

Output = (system function) (input) (2-33)

rhus, the output is determined from the product of the system function
;ind the input. In the example of the capillary the system function
is the resistance of the capillary. Note that the system function is a
characteristic of the system and is given in terms of the physical
characteristics
Other proportional elements are electrical resistance, gas flow resist-

cance, thermal resistance, and the mechanical spring.

The response of a proportional element is also shown in Fig. 2-7.
For a unit step change of input flow w, the output repeats this step
at the same time but with magnitude R.
The capacitance element is illustrated in Fig. 2-8. As one example of
a capacitance element, consider the liquid tank shown in the physical
diagram. The flow into or out of the tank m is considered the input
variable and the head of liquid in the tank c is the output variable.
For a capacitance element,

dc
C— = m (2-34)

where C - capacitance
c = output variable (head)
I -- time
m= input variable (flow)
30 Automatic Process Control

In order to obtain the system function the operational notation of differ-

ential equations must be used:

d
(3-86)

c, head
,
flow

Physical diagram

1 c
i

Cs

Block diagram

Fig. 2-8. The capacitaDce element.

where s = differential operator.^ Then eciuation 2-34 may be written

(Cs)c = m (2-36)
or

(2-37)

* the symbol D, At this particular time

The diflFerential operator is sometimes given
there no relation other than intentional coincidence between the differential
is

operator « and the LaPlace transform complex variable a.

Process Characteristic. 31

The system function is therefore l/(7« and describes symbolically the

capacitance element. Again the product of system function and input
variable gives the Output variable. Other capacitance elements are

Time, t —

Block diagram

Fig. 2-9. The time-constant element

illustratedby electric capacitance, gas capacitance, and thermal capac-

itanceand are characterized by storage. The response of a capacitance
element is found by integrating equation 2-34. For a constant inflow
M and with the tank initially empty,

This equation is plotted in Fig. 2-8. For a step change of input flow M,
30 Automatic Process Control
In order to obtain the Bystem function the operational notation of differ-

ential equations must be used :

d
{2-35 )

Physical diagram

1 c
Ca

Block diagram

Fig. 2-8. The capacitance element.

where 8 = differential operator.^ Then equation 2-34 may be written

(Cs)c = m (2-36)
or

(2-37)

^ Q{>erat.or is sometimes given the symbol D, At this particular time

The difTerential
there no relation other than intentional coincidence between the differential
is

operator a and the La Place transform complex variable a.

Process Characteristic. 31

The system function therefore 1/C« and describes s3rmbolically the

is

capacitance element. Again the product of system function and input

variable gives the Output variable. Other capacitance elements are

fit, flow

Physical diagram

R c ^

Ts+l

Block diagram

Fig. 2-9. Ttie time-constant element.

illustratedby electric capacitance, gas capacitance, and thermal capac-

itanceand are characterized by storage The response of a capacitance
element is found by integrating equation 2-34. For a constant inflow
M and with the tank initially empty,

'-si'"'"-©'
This equation is plotted in Fig. 2-8. For a step change of input flow M,
 32 Automatic Process Control

the output variable increases linearly with time with a rate of change
inversely proportional to capacitance.
The time-constant element is shown in Fig. 2-9 fiftid is exemplified by
the liquid tank and resistance. The input variable is the inflow rate m
and the output variable is the tank head c. For the tank capacitance C,

C- fit
= m-q (2-39)

where q is the outflow. For the fluid resistance /?,

(i
= (2-40)
R
Combining equations 2-39 and 2-40 to eliminate outflow q,

lie
KC~->tc=Rm (2-^1)
fit

The product of resistance and capacitance RC is a time which will be

designated as 7’. 'J'hen the operational e(] nation is

This sys'em function is cliaractenstic of the time-constant element

Any series arrangement and resistance in which the rate
of capacitance
of change of potential at the capacitance is decreased by increasing
outflow through a resistance produces a time-constant clement. Com-
binations of electrical, Tniuid, gas, and thermal resistance and capacitance
may produce a time constant element.
1’hc response of the time-constant element is found by solving equa-
tion 2-41 for a step change of inflow M
with the tank initially empty.*

c = RM {I - (2-^3)

This equation is j)lotted in Fig. 2-9.

* I’he Holution t.o any difTeronlial equation of the form

IS given by the sum of the particular mlegral and complementary function. The
particular integral im

X =: By
if y 16 a constant. The complementary function is the solution of
 Process Characteristics 33
The time (T = RC) is the time constant of the element and is found
experimentally when I = T,

c = = RM(l - 0.368) = O.mRM

TherofoK* tiio time constant is the time required to reach 63.2 per cent
of the final value. Times for other magnitudes of change are found
from tables of

Per cent response Units of T

50.0 0.69
G3.2 1.00
95.0 3.00
98.0 3.90
99.0 4.60

There are other important characteristics of the exponential response of

the time constant element As shown in Fig. 2-9, the final value for
a unit step change is proportional to the resistance R. The time con-
stant is the intersection time of the initial slope at the final value and
the initial slope is inversely pioportional to the time constant.
The oscillatory clement is shown in Fig. 2-10. Although it is not
eueJuntered in ordinary liquid, gas, and thermal processes, it is typical
ofmany measuring instruments such as the Bourdon-tube pressure gage.
Consider the mass spring and damping system of Fig. 2-10. Newton’s
second law of motion gives

M -• -B ~ - Kc + rn (2-44)

where the force m. is regarded as the input variable and the displacement
of themass is eonsidered the output variable. Rewriting equation 2-44

and in

r = Ce

Substituting in the equation imniedi.ately above,

0
and solving for X,
X = 1/A
The complete solution is therefore

T ^ By + Ce-"^
The arbitrary constant C is found from initial conditions.
34 Automatic Process Control
in operationnl form,

(3 -«)

Time, I

u. e, dispiacement
1
m, force-^

Physical diagram

m l/JT e
TV + 2rTs+l

Block diagram

Fig. 2-10. The oscillatory element.

For convenience the following definitions are often employed:

T^ characteristic time (^)

f damping ratio (^7)

Prooees Characteristics 35
Combining Uie last three equations, there results

m (2-48)
+ 2[Tt + 1 /
The system function is typical of oscillatory elements when the damp-
ing ratio { is less than one (underdamped).
The response of an oscillatory element is shown in Fig. 2~10 for a
unit step change of mput variable. Solving equation 2-48, under the
conditions that the system is initiaUy at rest,^

c-^[l--^L^8invT^l-e-f‘/»’coeVr^i;j (iM9)

* The solution to any differential equation of the form

d^x dx

is given by the sum of the particular integral and complementary function. The
particular integral is

X ^ Cy
if V is constant. The complementary function is the solution of

d*j dx

anh is found by substituting

X -
in the equation above,
AKkh-^^ - «0
Solving for X from
ilX* - H- 1 =0
there results from the quadratic formula

-f B d= \/jg» - 44
Xi,Xj
24
The form of the complementary function depends upon whether > 44, B* * 44,
or < 44 In the latter case the roots X are complex and the complete solution is
.

I = Cy +
Employing Euler’s relation,

c * e cos Xf t sin Xf

the solution may be placed in the form

where K% and Ka are arbitrary constants which must be evaluated from initial con-
ditions.
 :

36 Automatic Process Control

The period of oscillation may be determined from the characteristic
time and damping ratio.
27rT
P = (2-50)
Vi
For damping ratios ^ of one hall and less, the square-root term approaches
unity. The frequency of oscillation is the inverse of the period in cycles
per second
When the damping 2-48 is e(|ual to or greater than
ratio in equation
one or, if > 4KM in system is no longer an oscil-
cejuation 2 45, the
latory element. That is, for a unit stop change of the input variable, the
response does not over.shoot the final \'alue but changes smoothly from
initial to final value Under this condition the system function of
equation 2-45 may be factored

1 l/K
(2-51)
Ms- 4- Bs -f K (Tis H- 1){T2S + 1)

where the tune constants arc given by

J_
B
(2-52)
7'., 2M
'fherefore a second-order system with overdamped action > \KM)
may be factoied into two time-constant elements If the system is
critically damped these two time-constants are et|ual
The response* of an oscillatory elemi'iit always overshoots the final
value and comes to Hist practically at a time dependent upon the
amount of damping and the characteristic lime. The amount of over-
shoot is also greatly dependent upon the damping The frequency of
oscillation depends primarily upon the characteristic time T

Liquid Processes
Proc(‘sses involving the How ot liquuls through connected pipes and
\cssels are common cwamples ol industrial processes and are often
employed for demonstrating the operation of many different kinds of
processes.
The single-vessel process of Fig’ 2-1 1 has a constant outflow u but
a variable inflow ui The inflow m assumed to he the manipulated
is

variable, and the vessel head is assumed to be the variable to be con-

trolled. c The continuity relation for the vessel is

Cc = m — u (2-53)
Process Characteristics 37

where C is the vessel capacitance. The system equation is

The system function is l/Cs and the process is identified as a capaci-

tance element. The outflow u is the load variable. The operational

equation indicates that the action of the load variable u is identical but
opposite in sign to the action of the manipulated variable. A step
increase of the manipulated variable m made while holding outflow u

Fkj 2-1 ] (\Hpm*itan< »*-type liquid-level process.

fi.xed prodiU'cs a linear increase oi head r with time a.s identified in

Fig 2 8 A stop increase of load // made while holding inflow 7ti fixed
produce^; a linear decrease of head c with time

Fk; 2-12 Kquivalenl block dianiam.s

The block diagram foi tlie piocess of Fig. 2 11 may he drawni in two
ways as shown in Fig. 2 12 These two diagrams produce identical
relationships hetw'een the variables of the system.

Fio. 2-13 Timft-conatant-type liquid-level process

The single vessel of Fig. 2-13 in which the outflow is through a liquid -

flow resistance is identified as a time-constant element. The inflow m

38 Automatic Process Control

is considered the manipulaled variable, the vessel head c is assumed to

be the variable to be controlled, and the downstream head u is the load
variable. The continuity relation for the vessel isf

Ct m^ (2‘-55)

where C= vessel capacitance

qo = outflow

The outflow resistance is assumed to be laminar so that outflow depends

upon head differential:

9o = ^(c-u) (2-56)

where R = valve resistance. Combining these two equations to elimi-

nate outflow qo there results

Tc + c = Rm -f u (2-57)

where T => RC is the time constant of the process. The system equa-
tion is

Thus the process is identified as a time-constant element. Its response

to a step change of inflow m (head u fixed) is the typical exponential
response of Fig. 2-9. Note also that a steady-state change of inflow m
results ultimately in R units of change in head. A step increase of
head u results also in the typical exponential response of head but the
magnitude of steady-state change is the same for both head u and vessel
head c (unit sensitivity or gain).
The process of Fig. 2-14 consists of two vessels in series or “cascade”
and, although it is not a common industrial process, the general arrange-
ment is often found in thermal processes. The inflow m will be con-
sidered the manipulated variable and the head c in the lower tank will
be considered the variable to be controlled. Inflows ui and U2 will be
considered load variables. Downstream head at the lower vessel is

considered constant.
The continuity relation for the upper vessel is

C 2A 2 = m -f- W2 ~ (2-59)

where C2 = capacitance (area) of upper vessel

A2 = head in upper vessel
qr — outflow at upper vessel
 :

Process Characteristics

The outflow is assumed to be directly proportional to head

resistance)
/l2 “ Qrtt2 (2“fl0)

where is the outlet valve resistance. Combining these equations,

T^2 “4" ^2 “ H“ -^2^2 (2^0

Fig. 2-14. Two time-coDstants in eeries.

where T2 = R 2 C2 and is the time constant of the upper vessel. The

latter equation may be placed in operational form

(2-62)

The block diagram is shown by the lower left-hand rectangular blocks

and circle of Fig. 2-14.
The continuity relation for the lower vessel is

C\c = + Ml - qo (2-63)

where Ci = capacitance (area) of lower vessel

= outflow, ft^/sec

The outflow is assumed to be directly proportion^ to head (laminar

flow),
c « qjt, (2-64)
 :

40 Automatic Process Control

where Ri is the outlet valve resistance. Combining these last two

equations,

7’ic +c= ~/l2

^2 + RiUi
*
(2-C5)

where Ti = RiCi and is the time constant of the lower vessel. The
latter equation may be placed in operational form

The block diagram representing this eijuation is shown in P'ig. 2-14 by

the right-hand set of two blocks and circle. The process is thus seen

Fig. 2-16. Equivalent block diagraniH.

to be two time-consiant elements in serie.s, Combining equations 2-62

and 2-66 to eliminate head hz gives

R, .
Ri
(2-C7)
{TiS+\)iTzs+\) (7’,.s-+1)(7’,a+1) (7VTi)
and the o[)erationul equation lor the uxeiall .system j.s obtained. The
block diagram for the overall system is shown in Fig, 2-15 where com-
parison may be made to the previous block diagram.
For two time-constant elements in series, the response is found by
solving the differential equation for the overall .system. If the load
Process Characteristics 41

variables u\ and U 2 are zero (flows turned off), equation 2-67 becomes

T 1 T2 C + (Ti + T2)6 +c = Him (2-68)

This is a linear second-order differential equation and is typical of many

industrial processes. The process response can be obtained by allowing
a step change of inflow m. The solution is

Fio. 2-16, Response of processes with several equal time couslante in series.

where c/ = Him is the final steady value of head in the lower vessel
and Co is the initial value of head in the lower tank. Another solution is

LZ-Ii. = 1 _ e-tiT, _ ± if 7.J = (2.70)

Cf - Ce
'
Ti

This equation is plotted in Fig. 2-16, together with the response of a

single time-constant process. Notice that the addition of the second
time-constant produces the typical S-shaped curve and serves to retard
the initial response.
Liquid processes, and in fact many other kinds of processes, may
involve more than two time constants in series. Sometimes three, four,
42 Automatic Process Control
or many more capacitances are connected in series. The response
such processes is shown in Fig. 2-16. The response is always of the
typical S-shape but is more pronounced as the numbir of time constants
becomes larger. It is much easier to define such processes by their
individual time constants Ti, T2 Tz Tn than by the response that
,
.

is produced by the overall system.

Example 2-9. A single capacitance process like that of Fig. 2-13 has
a normal operating head of 4 ft and a normal value of outflow of 0.125 cu ft
per sec. The cross-section area of the vessel is 6 sq ft. •Calculate the time
constant.
If the resistance to flow is parabolic, the resistance is given by

C
(-)
\ Q /normal
eft*
“
^ = 64 sec/ft*

T flC * 64 X 6 » 384 sec

Example 2-10. Show that the time constant of a single vessel as shown
in Fig. 2-13 is proportional to the time required to change the fluid in the
vessel.
The capacitance is the area of the vessel and the resistance of the outlet
valve depends upon head and flow. For turbulent resistance,

X A - 2 2
9o outflow

For laminar resistance

T RC -
•V.
XA , = hA— Qo
volume of fluid
outflow

Example 2-11. The two-capacitance process shown in the figure consists

of two vessels in series (not cascaded). Derive the process differential equa-
tion and define the two system time constants.

For the left-hand vessel

C2A2 m- qr

and for linear resistance R%

A| - C « QrRf

For the right-hand vessel

Cid » 5r - 9#
 ;

Process Characteristics 43

and for linear resistanoe Bh

e - g^Ri

where Ct « capacitanA of left-hand vessel

Rs « resistance between vessels
Cl » capacitance of right-hand vessel

Ck)nibining the four equations to eliminate head h 2 and flow gr,

TiTzS -i- (Ti + T2 4* RiC 2 )d c = H\m

where Ti — and Tt « R 2C 2 , Comparing this result to equation 2-^,
l^iCi
the difference that the factor RtC 2 is added to the second term of the equa-
is
tion. Employing the quadratic solution we may define two new time constants
and Tt so that

1 (Ti + T + RiCi)
2 =b V{Ti ^Ti + RiC2y - iTiTt
2T1T2

and write the above equation in the new form

TaTi/i + (Ta -h Tt)6 +c = film

Thus the two process time constants are Ta and

Gas Processes
Processes involving the flow of gas through connected pipe lines and
pressure vessels are conunon in industrial practice.

|m

Fio. 2-17. Single time-oonstant-type gas process.

As an example consider the pressure vessel of Fig. 2-17. Pressure c

in the vessel is the variable to be controlled. The stem position or
opening m of the inlet valve, not the inflow Wi must be taken as the
44 Automatic Process Control

manipulated variable, because the characteristics of the throttling

valve enter into the response of the process. The continuity relation
for the vessel is

Cc — u\ — Wo (2-71)

whore C == gas capacitance of vessel

= inflow
Wo = outflow

The vessel outflow Wo is a function of two variables the pressure differ-

ential is not large, so that acoustic velo(;ity is not obtained:

“ /(r, (2-72)

where uq is the downstream pressure. The total differential for a func-

tion of two variables is

dw„ - dc 4 (- ') dll,, (2-73)

This situation is illustrated by Fig. 2 18 in which the flow rate is

plotted for an actual valv(‘, with do\\’nstreain pressure Uu constant,

first

and second, with upstream pressure c constant The partial derivatives

of equation 2-73 are shown as slopes of the pressure-flow curve. Because
the right-hand partial derivative is negative, a negative sign is used in
equation 2-73. If the pressure ratio is less than critical (0.53 critical
pressure ratio for air), the flow rate does not depend upon downstream
pressure and the second term of equation 2-73 We will assume
is zero.
that the outlet valve is operated near the horizontal and vertical dotted
lines of Fig. 2-18. In a small operating region, the slopes may be con-
sidered constant and equation 2-73 may be integrated:

"•-©'Hi)
where Mo is a constant of integration.
The inflow (ic,) is a function of three variables, the stem position m
of the inlet valve, the upstream pressure wi, and the vessel pressure c,

so that
tCa = f{m, Ui, c) (2-75)

The total differential is

The partial derivatives may be evaluated as for the outlet \'ul\'e witli

the addition of one more plot of flow iat(' against valve stem position
46 Automatic Process Control

Combining the equations 2-71, 2-74, and 2-77 the Tproce^s eqiwJtUm is

Tt c ^ K\fn -f- K^.\ (2-78)

or in operational form

“ “+
'
(t^) ““ + (?^)
where

y
C X
\/Ra + 1/^4 *
^/Ra + V®4
1
,j. ^ i
* So/fta + ®
fil/fl. + fil/Kk

The block diagram is shown in Fig. 2-17. The process response is first

order with a time constant T. Comp«.re the response of this process to

the response of the liquid-level process in Fig. 2-13.

Example 2-12. A pressure vessel connected as in Fig. 2-17 has a supply

pressure ui of 45 psig, a vessel pressure of 15 psig, and an air exhaust to atmos-
phere. The average flow rate is 1.0 lb per minute and the vessel volume is
10 cu ft. Calculate the system function.
First, the critical pressure ratios give:

14.7 297
== 27.8, 56.0
0.53 0.53

Since 27.8 — 14.7 < 15 psig vessel pressure the outlet valve has
== 13.1 psi
critical (acoustic) Also because 56.0 — 14.7 = 41.3 < 45 psig, the
flow.
inlet valve has critical flow. Therefore /2o ® and and thereby “
Aj =* 0 in equati<|n 2-79. The time constant is
15 X 144 10
= min
RaC ^ 0.76
1.0 1.0 X 53.3 X 530

Ki RoK^
Ha
Kz - 2.0
Ri

The system function is, therefore.

c = Ui

Example 2-13. The single capacitance pressure process shown in the

figure supplied with a constant flow u lb per sec. The controlled variable
is

is the vessel pressure c psi gage. The stem position of the outlet valve is m
the manipulated variable. The outlet valve downstream pressure is less than
critical. Derive the process equation.
Prooess Characteristics 47
The continuity relation is
Ct »
The outflow does notsdepend upon downstream pressure, so

or

Wa * Kftn -h c + Mq
The process equatioi^ is

Td -h c = -RK^m + RMo
and the operational equation is

where T = RC, The resistance R of the outlet valve is the slope of the
upstream pressure versus flow curve at constant valve stem position and
is usually obtained by test. In drawing the block diagram, constants (such
as RMo) are usually omitted because they do not directly influence process
dynamics.

Flow Processes
The automatic control of fluid flow in pipelines would at first appear
to be a simple problem of manipulating a throttling valve so as to
maintain the flow rate at the desired value. In practice, the analysis
involves a large number of factors related to flow characteristics, and
a separate consideration of flow-rate processes is necessary. The analysis
of flow-rate processes will be made by assuming no effect due to inertia
of flowing fluid. This assumption is quite close to reality, because in
practical applications the changes in variables occur relatively slowly.
In addition it is assumed that the capacitance of pipelines is negligible.
Liquid-flow problems will be discussed first, and then gas-flow problems

will be reviewed.
Liquid flow rate is assumed to be measured by some type of metering
device such as an orifice, nozzle, or Venturi as shown in Fig. 2-19. The
flow rate through the metering device is given by

c JiAiV2g{uo - wi)
'
(2-BO)
48 Automatic Process Control

where c liquid flow rate

J 1
flow coefficient (including velocity-of-approach factor)
Ai area of restriction
Uo upstream head
W| downstream head
<7 gravitational constant, ft/sec^

Fig. 2-19, A flow-rate process.

The coefficients Ji and area Ai may be considered constant in some

operating region of flow and head differential. Then, since flow is a
function of two variables (wo and Wi), the total differential is defined as

(2-81)

The partial derivatives are equal, so

(2-82)
^
dU{) dui 2 \Uf) — U\l 2(wn -

Equation 2-81 may be integrated provided the flow is approximately

constant:
c-«:i(Mo-ai) + 3/i (2-83)

where Mi is the constant of integration. The flow also passes through

the throttling valve and the flow rate is given by

c = J2A2ni^2g{ui - uz) (2-84)

where J 2 = flow coeflBcient

A 2 = maximum valve port area
m = fraction of valve port, opening
wj = downstream head
Process Characteristics 49
The coefficient Ji and the area Aj may be considered constant in some
operating region of flow rate, head, and \ alve opening. Then, since
flow is a function of "hree variables (m, Wi, 112 ), the total differential is

defined as

dc = (£-) dm + + (^) du, (2-85)

(^)
The partial derivatives may be calculated from ecjuation 2-84:

dc c
K„ ^ = «/2A2V2j7(t/i
/
- U2 ) - — (2-86)
dm ?7i

where is the valve sensitivity. Also,

^^ ^
dvj <9?/2
J2A2 m /
2 \//i
2ij

— ^/2/ 2(i/i
c

— U2 )
(2-87)

The throttling-valve flow equation is now written

c = K^m -b A' 2 (wi - H 2 ) -f ^f 2 ( 2 88 )

-

by integrating equation 2 -85 with M 2 as a constant of integration.

The “process equation^' can be found by ediminating variable U\
between equations 2 -83 and 2-88:

— U2 ) (2-89)

where M.s is a constant. Simplifying,

c = iKa)fn {Kb)u^ — {Kb)u 2 + M^ (2-90)

where Ka
Ki -f K2
KIK 2 c
A6 =
K, -h k2 2(?<o - WlO

Thus the “process'’ is a simple algebraic one and constitute.-, a propor-

tional element as shown in Fig. 2-19. The response of the process is

direct; that is, any change ui the jiosition of the throttling valve m
results immediately in the same change in flow rate modified by the
process sensitivity Ka. Changes in upstream head (uo) and down-
stream head ( 1^ 2 ) act similarly.
Gas flow-rate processes may
be analyzed by the same procetlure as
that employed for liquid flow-rate processes. However, the fl«)w rate
depends upon three variables rather than two because the weight flow
of gas through a restriction depends upon upstream temperature, up-
stream pressure, and downstream pressure. Also, the flow through the
50 Automatic Process Control

throttling valve depends upon similar variables. In order to eliminate

two of the variables, the energy equation (first law of thermodynamics)
must be employed. Thus the analysis becomea rather complex even
when simplifying assumptions are made. Consequently, we rely upon
the previous analysis to write
c = + {Kq)uq + {K\)ui — (K2 )u2 + M (2-91)

where c = gas flow rate — Ib/sec

Ka = constant, Ib/sec
m= position of throttling valve
Ko == constant, Ib/sec (deg R)~^
uo = upstream temperature, deg R
Ki = constant, ftVsec
Ui = upstream pressure, Ib/ft^
K2 = constant, ft*/sec
U2 = valve downstream pressure, \h/i\?
M = constant of integration
The constant must be evaluated from graphs of flow
coefficients (A's)
rate c versus the four variables m, Uq, wi, and U 2 If the pressure ratio
.

at the throttling valve is greater than critical, the flow does not depend
upon downstream pressure U 2 and the factor /ir 2 is zero.

Thermal Processes i

All thermal processes possess characteristics attributed to capacitance

(specific heat and weight) and resistance. Thermal processes, however,
are clearly distinguished from fluid processes, because the resistance and
capacitance are usually distributed throughout heat transfer paths.
Fortunately there are a number of thermal-process elements that are
characterized mainly by capacitance (agitated fluids) and by resistance
(insulation).
The air-heating process of Fig. 2-20 is an example of a simple thermal
process. Heat flow m into the system is provided by an electrical heater
(usually rated in watts electrical input). Since all of the electrical
energy must be transferred to the air,

Cc = w -h QPv - QPc (2-92)

where C = thermal capacitance = IFP, Btu/deg

W= weight of air contained in heater, lb
c = heater outlet temperature, deg
m= heat input, Btu/sec
Q= weight flow of air through heater, Ib/sec
P »= specific heat of air, Btu/lb deg“^
u s inlet air temperature, deg
Process Characteristics 51

The heat losses and the capacitance of metal parts in the heater are
assumed to be negligible. Rewriting equation 2-92, the process equa-
tion is obtained:

where T = C/QP = W/Q is the time constant of the process. The

block diagram is shown in Fig. 2-20. The process represents a time-

|m. Btu/sec

Fio. 2-20. Time-cons tan t>tvpe thermal process.

constant element with time constant T. It is interesting to note that

the dynamic characteristics of this process are the same as for the
liquid-vessel process of Fig. 2-13 and the gas process of Fig. 2-17,
because they are all time-constant elements characterized by a single
time constant.
The tliermal time constant of most simple thermal processes is given
by the storage and throughput rate. Thus, the time constant is cal-

culated from

^
W weight of fluid contained

Q weight rate of throughput

This time is the inverse of the rate at which the fluid substance is

changed.
The thermal process of Fig. 2-21 consists of
a heating furnace such
as might be used in continuous heat treating. A gas-air mixture is
burned in the furnace and represents a specified rate of heat addition, m.
We shall suppose that the main heating load is the metal boxes con-
52 Automatic Process Control

taiiiing metallic parts. These boxes are carried continuously through

the furnace by the conveyor. Heat losses occur at the furnace walls

Fin 2 -21. Time-constant-type thermal process.

and at the stack. The controlled variable c is to bo the temperature ^f

the boxes and parts. P'or conservation of energy

(PKL)c = m - {PKV)c - lJA{c - v) - q, (2-94)

where P = specific heat of boxes and parts, Btu/lb deg“^

K == material on conveyor, Jb/ft
L — length of furnace, ft
V= velocity of conveyor, ft/sec
V= heat loss coefficient Btu/ft^ deg”' sec”^
A = heat loss area, ft^
u = ambient temperature, deg
= stack heat loss, Btu/sec

The stack losses are assumed to be a function of the heat input and the
furnace temperature.

9, =/(m,c) (2-95)

Then, the relation of dilferential changes is

Process Characteristics 53
For reasonably small fluctuation in heat input and furnace temperature,
the above partial derivatives ore nearly constant, and

g. = Kitn + K 2Q + M (2-97)

Substituting equation 2-97 in equation 2-94, there results after some

rearranging of terms

(2-98)

PKL
where T =
PKV + K2
1 - Ai
K, =
PKV + UA -f K 2
UA
K,
PKV +UAA-K 2
The process thus turns out to be the single time-constant type. Notice
that the amount of material on the conveyor K alters the time constant
of the process. Control of thermal processes often involves variable
parameters in this manner.
^Heat-exchanger processes of the fluid-to-fluid type require a depth
of analysis beyond the scope of this text. These processes are usually
not describable by linear differential equations with constant coefficients,
although any given heat exchanger may be tested experimentally and
its dynamic characteristics determined. Sometimes, simple approxima-
tions of time constants may be made from experimental data.

Example 2-14. For the air-heating process of Fig. 2-20, the normal air
inlet flow is 300 cu per min at atmospheric pressure. The inlet temperature
ft

is 80 F. The volume of the heater is 500 cu ft. Calculate the time constant
of the process.

T
W ^ 500 X 0.072 X 60
100 sec
g 300 X 0.072

Dead Time
Dead time is defined as any definite delay between two related
actions. Dead time is illustrated in Fig. 2 -22 When the temperature-
measuring element is installed at a distance downstream of a heat
exchanger a pure time delay occurs liefore the heated fluid passes from
the heater outlet to the pmint of location of the temperature measuring
element. For example, if the fluid flows at 10 ft per see and tHe distance
is 10 ft, the dead time is one second.
 54 Automatic Process Control
Dead time may be encountered in processes involving pressure. For
example if pressure is measured at the end of a long pipe or tube a dead
time occurs because of the time for transmission*^ of the first pressure
wave down the length of the pipe. For air, the wave velocity is roughly
1000 ft per sec and appreciable dead time would occur for a pipe more
than a few hundred feet long. For liquids, the wave velocity is much
faster, and the dead time is usually negligible. Dead time may also be
found in processes involving chemical reactions when a finite time
must elapse before a reaction begins to occur.

iU it/, temperature
i IV V IV. fluid flow rate

Negligible dead time Larger dead time

Ficj. 2-22 Wator-heating proce&s with and without dead time.

The response of process elements with dead time is shown in Fig.

2-23. The response is simply delayed by a time equal to the dead time L.
Therefore if the input to the element is f{i), the output is given by,

Rt - L) (2-99)

The latter function may be expanded in a Taylor series:

~
f(i-L)^m -f'(0L+r(i) +/''"(«)
^+ 2- 100 )
• •
•
(

where the primes denote the derivative with respect to time. Employ-
ing the operator .s*

fit

The power
- L) = fit)

series in the
[l - Ls + ^
above equation is
+ ^ + •]

that for the exponential.

(2-101)

Therefore
/(f-L) = c”^*/(0 (2-102)

The equation is in the form of output equals system function times

Process Characteristics 55

input. Consequently the system function for dead time is

G = (2-103)

This function will be dealt with more later on.

Time, t —

Time, t
^
Fig. 2-23. Response of process elements with dead time, L.

PROBLEMS

3-1 . Name the variables of a liquid-heating tank heated by electric cur-

rent in an electric heater.

56 Automatic Process Control
2-2* A dam in a small stream stores water and the outlet is adjustable by
a weir. Name the variables.
2-3. A room is heated by a gas heater. Name some of the system variables.
*
2-4. A triangular weir has the equation

q = 2gh^

Calculate the resistance.

2-5. A flow-head device has the equation

fj — h^ •

Calculate the resistan<*e.

2-6. A liquid storage vessel is spherical in shape. Calculate the capaci-
tance as a function of head.
2-7. If the outflow at a vessel is proportional to the square root of head,
what shape vessel results in a steady change of head?
2-6. If the outflow at a vessel proportional to the square root of head,
is
what shape vessel results in a rate of change of head proportional to the head?
2-<>. Njime five flow laws relating to fluids, heat, and electricity.
2-10. Calculate the gas capacitance of a lO-cu ft vessel containing air at
room temperature.
2-11. Calculate the resistance of 1-sq ft area of a brick wall, 2.5 in.
thicknejss.
2-12. C/alculate the thermal capacitance of
1 gal of water.

2-13. Calculate the thermal capacitance of 1 cu in. of copper.

2-11. Which of the following five gases causes the greatest and whieb
the least capacitance in the same vessel at the same temperature: air, ammonia,
carbon dioxide, hydrogen, steam.
2-15. A system known to have a time-constant response requires 5 min-
utes to indicate 98 per cent of response. What is the time constant?
2-16. A tank operating at 10 ft head and 5 gpm outflow through a valve
has a cross-section area of 10 sq ft. Calculate the time constant.
2-17, Show that tlie time constant of a liquid vessel with turbulent
resistance T = 2H A /Q.
is

2-18. the greatest degree of slowness of action is required of a liquid-

If
storage vessel, would a long cylinder be placed vertical or horizontal.
2-19. For a rectangulai shaped vessel of dimensions A\,B\, and L ft,
calculate a number of capacitances.
2-20. An oscillatory element has a mass of 2000 lb, damping ratio of D
one half, and a natural frequency oscillation of one cycle per minute. Cal-
culate the spring constant K and damping rate D.
2-21. Heat conduction in a cylindrical element is governed by the
Fourier Law:
du

Calculate the thermal resistance of a tube of length L, inside diameter Z)<,

and outside diameter Do.

For the following problems, derive the system function relating the vari-
ables of the system:
 0*haplvr

Controller
Characteristics

The automatic controller, includinp its measuring means, determines

the value of the controlled variable, compares the actual veiue to the
desired value, determines the deviation, and produces the counteraction
necessary to maintain the smallest possible deviation. The method by
which the automatic controller produces the counteraction is called the
mode of contiol or control action.
In analyzing a specific control problem a choice based on economic
factors must be made among the various control actions. Generally
speaking, the more difficult the control problem, the moi complicated ft

the controlling means become. This does not at all mean tnat a compli-
cated automatic controller is necessary to produce good i.utomatic
control: on the contrary, the simplest control devices are often capable
of providing a high quality of control.
The control actions discussed in this chapter may operate through
either mechanical, pneumatic, hydraulic, or electric means. The various
mechanisms of automatic controllers are discussed in Chapter 6.

The Automatic Controller

An automatic controller. Fig. 3-1, consists of a measuring means

(c — 6), an input means (v —^r), an actuating signal means (r, ft, —+ c),

and a controlling means (e — m), where

c == controlled variable

ft = feedback variable
« = set point
r == reference input
e= actuating signal
m — manipulated variable
59
60 Automatic Process Control

The measuring means converts the controlled variable (temperature,

pressure, etc.) into an indicated variable (usuallyTi displacement, pres-
sure, or electrical signal). The input means converts the set point

Fig. 3-1. Automatic controller diagrams.

(temperature, pressure, etc.) into a reference input of the same units as

the feedback variable. The acluoixng means is simply a subtracting
device following the law
e ~ r - b (3-1)

The controlling means by amplifying, differ-

alters the actuating signal
produce a controller output which operates
entiating, integrating, etc. to
a final control element for changing the magnitude of the manipulated
variable m.
The remote-set pressure regulator of Fig. 3-1 contains all of these
means. The measuring means is the diaphragm and lower chamber
which converts the downstream pressure (the controlled variable c) into
an upward force. The upward force is the feedback variable b. The
input means is the diaphragm and upper chamber which converts the de-
sired value of pressure (set point u) into a downward force The
 b

Controller Characteristics r»i

downward force is the reference input r. The actuating means is the

diaphragm which subtracts the reference input r and the feedback
variable h to obtain the difference as the actuating signal e. The con-
trolling means is the stem and plug which convert the actuating signal
into a variation of flow of fluid through the regulator. The flow rate is
the manipulated variable m.
The study of controlling means and the modes of control is made
easier if some simplifications of these details are made. First, industrial
automatic controller# usually involve a steady value of the set point.
Consequently, the input means is usually a simple mechanical device
and the set point is identical to the reference input expressed in the
same units as the controlled variable. Thus, throughout this chapter
we shall take

V « r

Second, the measuring lag will be assumed negligible, so

c = 6

Therefore, the actuating signal e is termed the deviation, and

e = v~c-r—
Deviation is the difference between controlled variable and set point
expressed iu units of the controlled variable c. The arrangement in

Fig. 3-2 indicates these simplifications.

The controlling means includes the Control
/ means
final control element which is usually
some type of fluid-control valve.
Throughout this chapter it is assumed 6 me
that the final control element is a
Fig. 3-2. Block diagram for an
linear element such that the controller ideal automatic controller.
output (often a pneumatic pressure or
an electrical signal) produces a proportional change in fluid flow which
is the manipulated variable m. Final control elements are discussed in
more detail in Chapter 7.

Proportional Control
Proportional action is a mode of controller action in which there is a
continuous linear relation between values of the deviation and manip-
ulated variable. Thus the action of the controlled variable is repeated
and amplified in the action of the final control element. For purposes
of flexibility, an adjustment of the control action is provide^ and is

termed proportional sensitivity.

 62 Automatic Process Control

Proportional control is illustrated by the liquid-level control of Fig.

3-3. The connected to the stem of the control
float lever is directly
valve. A set-point mechanism has been omitted'?rom the figure. The
controlled variable is the vessel head c and the manipulated variable is

Fio. 3-3, Proportional control action.

the vessel inflow m. The action of the controller is such that a rise of
level c in the vessel rotates the valve operating lever clockwise thereby
closing the valve a proportionate amount. This reduces the inflow to
the vessel and tends to prevent the level from rising. If the distance
from float bearing to connecting link is large, a small change in level
produces a large change of flow into the vessel. If this distance is small,
ControUer Characteristics 63
a large change in level results in a small change of flow into the vessel.
Thus, proportional control follows the law

+M (3-2)

where m = manipulated variable

Kc = proportional sensitivity
ilf = a constant
e = deviation
•

The proportional sensitivity

Ke is the change of manipulated variable
caused by unit change of deviation. In the proportional controller of
Fig. 3-3 the units of proportional sensitivity are cubic feet per foot-
second. The proportional band is equivalent to the inverse of propor-
tional sensitivity and is defined as the change in level necessary to
operate the valve through full stroke. If the controller has a scale
showing values of the controlled variable, the proportional band is the
percentage of full scale change of the controlled variable required to
operate the valve through fuU stroke.
The constant M
in equation 3-2 may be termed the manual-reset
constant because the selection of a value for determines the normal M
(zero deviation) value of the manipulated variable. On most propor-
tional controllers there is an adjusting knob or other mechanism for
selecting the value of the constant.
The operation of proportional control action is illustrated in Fig. 3-3.
For a step (sudden) change in deviation

c = 0 ^ < 0

e = E (3-3)

where E is & constant. Substituting in equation 3-2,

m - M KcE (3-4)

The change manipulated variable corresponds exactly to the change

in
in deviation with a degree of amplification depending upon the setting
of proportional sensitivity Kc. Thus a proportional controller is simply
an amplifier with adjustable gain.

Example 3-1.For the proportional controller of Fig. 3-3, the float arm
is 14 in. long,the link to bearing distance is 2 in. and the valve stroke is 1 in.
The valve passes 10 gal per minute when full open. What is the controller
sensitivity?

Am Ag, 0.134 (10 - 0)

Ac
“ Ah 60 ^ (14/2 X 1 - 0)1/12
64 Automatic Process Control
Euiinple A proportional controller haa an output m changing linearly
from 0 to 15 psi when the deviation e changes from —100 to 0 to 100 degrees.
Calculate the controller sensitivity.
Am 15-0
Ac* “ 0 075 psi /deg
Ac 100 - (- 100 )

Integral Control
Integral action is a mode of control action in which the value of the
manipulated variable m is changed at a rate propctional to the devia-
tion. Thus if the deviation is doubled over a previous value, the final

control element is moved twice as fast. When the controlled variable

is at the set point (zero deviation), the final control element remains
stationary.
Integral control is illustrated in Fig. 3-4. A set-point mechanism is

omitted in the drawing. A variable ratio speed reducer, consisting of

two parallel disks with a friction drive roller between, operates a control
valve through gears. The left-hand disk is driven at constant speed by
an electric or other motor. The position of the friction drive roller is
set by the float and ann. The controller action is as follows: A rise in
level in the tank causes the drive roller to move up from the neutral
(zero speed) point. The speed of motion of the valve stem is proportional
to the change in head. A fall of level moves the drive roller below the
neutral point and the vah^e is moved at a p?oportionate speed in the
opposite direction.
Integra] control follows the law

(3-5)

or, in integrated form

(3-6)

where m= manipulated variable

T, = integral time
e = deviation
M ~ constant of integration

The operational form of the equation is

m = ( 3- 7 )

and is indicated in Fig. 3-4.

The integral time T, is defined as the time of change of manipulated
variable caused by a unit change of deviation In the example of Fig.
Controller Characteristics 65

3-4, the units of integral time are seconds squared per square foot. A
connection of the roller link in Fig. 3-4 closer to the float produces a
small integral time, and a connection closer to the float bearing produces
a large integral time

Integral control action is shown in Fig. 3-4. For a step change of

deviation
e = 0 ^ < 0
(3-8)
e = K / ^ 0
 ;

66 Automatic Process Control

where E\b 9k constant. Substituting in equation 3-6 and integrating,

Ei
m-M^- * (M)
Thus, the manipulated variable changes linearly with time and **inte-
grates” the area under the deviation function.For a unit step change
of deviation (E ^ 1.0), the slope of the line is inverse of the integral time.

Example 3-3. A hydraulic integral controller has an integral time of

^
a
sec in. per in.
20-in. stroke.
When the deviation e is zero, the piston is at the middle of
Calculate the rate of piston motion if the deviation changes
suddenly by (a) 2 in., (b) —4 in.
Employing equation 3-5, (a) ih = 10 X 2 = 20 in. per sec; (b) *= 10 X

(—4) s —40 in. per sec.

A negative sign indicates that the piston moves in an arbitrarily negative
direction.

Example In a certain integral controller the deviation changes

3-4.
sinusoidally with time. Show that the phase of the manipulated variable m
is always ^degrees behind the deviation.
The deviation is taken to be
e = sinci)<

Then, from equation 3-6

m« sin (at dt + M
and
^”
(aJ t
(at M
Since the manipulated variable m
is given by a negative cosine function, the

phase lag is seen to be ir/2 or 90 deg.

Proportional-Integral Control
Integral control action is often combined additively with proportional
control action. The combination is termed proportional-integral action
and is used for the purpose of obtaining certain advantages of both
control actions.
Proportional-integral control action is defined by the following
differential equation

m= ~«
I
+ KcC (3-10)
t

iQt^rel proportional

or, in integrated form,

m= M
y jedt+ Kce+ (3-11)

inteipal proportional
Controller Characteristics 67
where m» manipulated variable
Kt =“ proportional sensitivity
Ti * integral time
e ^ deviation
t =* time
M = constant of integration
These equations illustrate the simple addition of proportional and
integral control actioifs. In operational form

m= Ke (3-12)

where the system function Kc/(l\s) identifies the integral action and the
system function Kc identifies the proportional action.
Proportional-integral control action has two adjustment parameters,
Kc and as indicated in equations 3-10, 3-11, or 3-12. The propor-
tional sensitivity is defined the same as for the proportional control
action: With the integral response turned off (T* —» « ), the proportional
sensitivity is the number of units change in manipulated variable m per

unit change of deviation e. Note in equation 3-12 that the proportional

sensitivity Kc affects both the proportional and integral parts of the
action.
The integral action adjustment is the integral time. For a step
change of deviation e, the integral time is the time required to add an
increment of response equal to the original step change of response as
indicated in Fig. 3-5. Reset rate is defined as the number of times per
minute that the proportional part of the response is duplicated. Reset
rate is therefore called ‘Repeats per minute” and is the inverse of integral
time.
Proportional-integral control action is shown in Fig. 3-5. For a step
change of deviation
e = 0 i <0
(3-13)
e = E t^{)

where E is a constant. Substituting in equation 3-11

(3-14)

This is the equation for a straight line. The first term, i/Tt, is the in-
tegral response, and the second term is the proportional response. The
latter is indicated by the dot ted line of Fig. 3-5. Notice that the response
68 Automatic Process Control

of proportional-integral action in Fig. 3-5 is the simple addition of the

proportional control action of Fig. 3-3 and the integral control action of
Fig. 3-A.

Time, t

e
^
Fid. 3-5 rropcjrtional-reset control action.

Example 3-5. A liquid-level controller has an output (proportional to

the manipulated variable) given in psi pressure. A test of its action is made
as follows: zero deviation is found by holding the float lever in a position to
give zero rate of change of output. The float lever is then moved suddenly
by 2.0 in. from the zero position. The output changes rapidly by 4 psi and
then changes at the rate of 6 psi per minute. What is the controller sensitivity,
and integral time?
Sensitivity = 2 psi /in

Integinl time = = 0.07 min

6 psi/iiiin

Example 3-6. The deviation v of a pioportional-integral controller is

sinusoidal. Calculate tlie phase lag of the output m and prove that the
phase lag depends upon integral time.
Assume
= sincof

Substituting in eouation 3-11,

~
K f
~ j
siti u)t dt Kf sin wf -h Af
Controller Characteristics 69
Integrating,
m- Af -1
K, 0)^
COB + sin o)/

^
Manipulating into a function of a double angle,

m—

The phase is the second term of the angle:

Phase == tan“*

The phase lag depends inversely upon integral time 'A: for large integral time,
the phase lag is small; for small integral time, the phase lag is large.

Proportional-Derivative Control
A derivative control action may be added to proportional control
action and the combination termed a proportional-derivative control
action. Other terms for derivative response are rate response and lead
component. Derivative control action may be defined as a control
action in which the magnitude of the manipulated variable is propor-
tional to the rate of change of deviation.
A proportional-derivative control action is defined by

m= A> + KcTae + M (3-15)

proportional derivative

where m= manipulated variable

Kc = proportional sensitivity
e = deviation
Ta = derivative time
M~ 2l constant

and is the simple addition of proportional controller action and rate

controller action as shown by the operational equation

m- M = A%(1 -f T^ff)e (3-16)

Proportional-derivative action is not adequately described by employ-

ing a step change of deviation because the time derivative of a step
change is infinite at the time of the change. Consequently, a linear
change of deviation must be used:

c = Et (3-17)
where E= sl constant
t = time
70 Automatic Process Control

Substituting in equation 3~16

m-M KcE(t + Ta) (3-18)

The deviation is defined at time

whereas the manipulated variable is
defined at time t + and the net
effect is to shift the manipulated
variable ahmd by a time Td, the derivative time. In other words as
shown in Fig. 3-6, the controller response now leads the time change of

+ Trff)

Fio. 3-6. Proportional-derivative control action.

deviation. To this extent a derivative response ^^anticipates.’’ Of

course, a derivative response can never anticipate an action that has
not yet taken place.
Derivative time is defined as the amount of lead, expressed in units
of time, that the control action is given. In other words, derivative
time is the time interval by which the rate action advances the effect
of the proportional control action.

Example 3-7. A proportional-derivative controller has the following rela-

tion for the proportional action only:

deviatione, —5 to 5 in.

manipulated variable m, 0 to 1 volt

Controller Characteristics 71
If the derivatisre time Ta is 5 seconds and the deviation e changes at a rate
of +2 in. per minute, how much voltage output is added by the derivative
action?
^
The proportional sensitivity is

From equation 3-18

m increment = E(0A)Td
*
= A(0.1)5 = 0 017 volt

Example 3-8. A proportional-derivative controller has a sensitivity Ke

of 1.0, and a derivative time (Td) of 1.0 min. If the deviation is sinusoidal,
calculate the phase of oscillation of the manipulated variable.
The deviation may be given by

e » sinciif

Substituting in equation 3-15,

m^ Kc sin wt -|- KeTau cos <A

and, writing as a function of a double angle,

m = KcVl H- {wTdVain ((/d -f uTi)

The phase leads;

Phase = t8LTr^u)Td

which depends upon derivative time.

Proportional-Integral Derivative Action

Hie additive combination of proportional action, integral action, and
derivative action is termed proportional-integral-derivative action.
It is defined by the difTerential equation

m = ^e-\-K^ + KcT^ (3-19)

or

m=^^"jedl + Kce + K,Tje + M (3-20)

where m= manipulated variable

Kc = proportional sensitivity
Tt = integral time
€= deviation
Td = derivative time
Af s a constant
72 Automatic Process Control

The operational equation ie

(3
-21 )

Fia. ^7. ProportionaJ-reset-cierivativc control action

Proportioiial-integral-derivative control action is illusti-ated in Fig. 3-7

by plotting the change manipulated variable for a linoar time change
in
of deviation. Assuming the deviation to be given by

c = Et (3-22)

where E ~ a constant
t = time

and substituting in equation 3-20

m-M = KcE Jldl + i+ (3 23)•

Integrating the first term

m-M = + +
/ Td] (3-24)
 Controller Characteristics 73

The proportional part of the control action repeats the change of

deviation (lower straight line), ^he derivative part of the control action
adds an increment o^ manipulated variable so that the proportional plus
derivative action is shifted ahead in time (middle straight line). The
integral part of the control action adds a further increment of manipu-
lated variable proportional to the area under the deviation line and, as
Fig. 3-7 shows, the increment increases because the area increases at
an increasing rate. The combination of proportional, integral, and
derivative actions lyay be made in any sequence, because these actions
are described by linear differential equations.
The three adjustment parameters, proportional sensitivity A'c, in-
tegral time Tj, and derivative time Td are defined the same as for
proportional-integral and proportional-derivative control actions.

Two^Position Control
The two-position control action, or on-off control, is undoubtedly the
most widely used type of control for both industrial and domestic service.
It is the kind of control generally employed on home-heating systems
and domestic water heaters.
Two-position control is a position type of controller action in which
the manipulated variable is quickly changed to either a maximum or
minimum value depending upon whether the controlled variable is
greater or less than the set point. The minimum value of the manipu-
lated variable is usually zero (off) This mode of control is illustrated
by the electric-level control in Fig. 3 8. A float in the vessel operates

Set
1
point
1

L_ c
h J

m ^

—
1

L_ c •-

Fig. 3-8. Two-poMlion coutiul.

an electric switch which controls power to a solenoid valve. When the

liquid level rises, the switch contacts are closed, the solenoid valve closes,
and the inflow is cut off. When the liquid level falls, the switch contacts
are opened, the solenoid valve opens, and the inflow resumes. If the
 74 Automatic Process Control

float lever has no bearing friction and the electrical contacts draw no
arc, the action is sharp or “knife-edge” as shown in Fig. 3-9.
The equations for two-position control are t

m = Ml when c > 0
(3-25)
m = Mo when e < 0

where m = manipulated variable

Ml = maximum value of manipulated variable (on)
Mo = minimum value of manipulated variablfj (off)
e = deviation

Thus two-position control must be described by two equations, each

appl3ring in a certain region of deviation.
A differential gap in two-position control causes the manipulated
variable to maintain its previous value until the controlled variable has
moved slightly beyond the set point. In actual operation it is the same
as hysteresis, as may be seen from Fig. 3-9.

Two- position action Two- position action

with differentiai-gap

Fig. ^9. Two-position control with and without differential gap.

A differential gap is caused in the two-position controller of Fig. 3-8

if The liquid
small static friction exists at the bearing on the float arm.
level must then rise slightly above the desired value to create sufiicient
buoyant force to overcome friction when the level is rising. Also, the
liquid level must fall slightly below the desired value when the level is
falling so that the weight force may overcome the friction. This kind
of differential gap may be caused by unintentional friction and lost
motion.
A differential gap may be intentional, as when a magnet is installed
on the float arm in Fig. 3-8, causing a hysteresis in float-aim action.
 Controller Characteristics 75
Similar arrangements are common in domestic thermostats and are
employed for the purpose of preventing rapid operation of switches and
solenoid valves and for reducing arcing of electrical contacts.
The three-position mode of control is one in which the manipulated
variable takes one of three values: high, medium, or low, depending
upon whether the deviation is large positively, close to zero, or large
negatively. Similarly, four- and five-position control may be used.

Single-Speed Floating Control

In the single-speed floating mode of control the manipulated variable
changes at a constant rate one direction when the deviation is positive
in
and in the opposite direction at a constant rate when the deviation is
negative. This action may be visualized from Fig. 3-8 if the solenoid
is replaced by a reversible motor with gear reducer
to move the control

-AT 1
m t

Fio. 3-10. Single-flpeed floating control.

valve stem. When the level rises, the switch contact is made,^nd the
motor-reducer slowly closes the control valve. As soon as the level
falls, the switch contact is broken and the motor-reducer reverses its di-

rection of rotation and opens the control valve. (A double-throw elec-

trical relay may be required.)
 76 Automatic Process Control

A neutral zone is used in single-speed floating control so that the motor

remains stationary when the deviation is small, as indicated in Fig. 3^10.
The equations for single-speed floating control with neutral zone are

m = M\ when fi > +N
m = 0 when +N > e > — JV (3-26)

rh = —Mo when e < —N

where N is the neutral zone in units of deviation.

Multispeed floating control is sometimes used in which several rates

of change of the manipulated variable correspond to several magnitudes
of deviation.

PROBLEMS

.VI . Prove tliat the sum of a sine wave and a cosine wave is another sine
wave of different amplitude and phase.

A sin wi + B cos cot =- V -f- sm^wt -f tair*

Use Euler’s relation c*"" = cos x -f isin x or employ vectors in the proof.
V2. A controller has the function m= (1 Tis -f- T 2 h^)e which in-
-f-

cludes an effect of the second derivative. Plot the steady-state response m

if e ^ sin (at. Use Ti = T2 = 1/ca.

V3. A controller has the function

= - / fedtdt + M
Prove that the phase of the output lags 180 degrees behind the input.
3-4. A controller has the following functions in series; {TiS d- 1),
(T 2 S -f- 1), X/Ts. What parameters determine derivative time, propor-
tional sensitivity, and integral time?
For the problems below, determine what control actions are present and
state the parameters derivative time, proportional sensitivity, and integral
time if they exist.

V5. 3-6.
 vhaptpr

Closed Loop
in Automatic Control

The problem of process control originates with the necessity for

minimizing the effect of changes in load variables. The process and the
automatic controller, acting together, comprise the controlled system,
and the characteristics of the process as well as the characteristics of the
controller affect the performance of the complete system.

Qoeing the Loop

The closed-loop system is obtained by connecting the process and the
automatic controller as shown by the operational diagram of Fig. 4-1.
Feedback is denoted in the diagram by the lower line acting to the left.

Controller Process

u I I

Fiq. 4-1. Closing the loop.

This feedback is negative, so that an increase in the controlled variable

c effects a reduction of the manipulated variable ?a, which in turn acts
to reduce the amount of change of the controlled variable. Thus the
alternate name for automatic control is negative feedback control, or
78
Closed Loop in Automatic Control 79
just feedback control. Feedback is physically embodied in the measuring
element of the automatic controller.
The study of autohiatic control throughout this chapter is made under
the following assumptions:

1. The measuring lag of the controller is zero so that the controlled

variable is the feedback variable: 6 = c.

2. The controller lag is zero so that the control actions are those
described in the pre^nous chapter.

The analysis of the action of the controlled system requires the use
of the differential equations describing the action of each element of the
system: the process equation as developed in Chapter 2, the controller
equation, and the deviation equation as developed in Chapter 3.

The process equation in general operational form is, from Fig. 4-1

n
c = G2m + 5Z (4-1)

where G2 = process system function

Ni = process load function

It is assumed that there are n load variables at the process, whereas only
one load is indicated in Fig. 4-1.
The controller equation in general operational form is

m= G\e (4-2)

where Gi = controller system function

The deviation equation is

e = V — c (4-3)

The study of controlled system behavior is usually made in steps begin-

ning with the calculation of open-loop performance. That is, if there
were no feedback in Fig. 4-1, the response of the controlled variable to
changes in deviation or to changes in load variable would be given by
the combination of the controller equation 4-2 and the process equation
4-1:

c = GiGje + z
• -1
(4-4)

This equation represents the open-loop response and is often found by

testing the particular system under study.
80 Automatic Process Control

The closed-loop performance is found by combining equations 4-1,

4-2, and 4-3 to eliminate variables m anti e\
t

This equation relates the magnitude of the controlled variable to the

magnitude of the set point v and magnitude of load variables w,-.
The closed-loop performance is also found from the same equations

by eliminating variables m and r:

The deviation is the same as the actuating signal e because the measuring
lag is zero (thus h = c), and because the set point v is the same as refer-
ence input r.

'Fhe closed-loop equations 4-5 and 4-6 illustrate that there are two
direct sources of deviation:

(a) ^'ariation of load variables, Ut

(h) variation of set point, v

‘
These sources are regarded as the 'disturbances,” the effect of which the
automatic controller is expected to minimize
From the standpoint of industrial process control we are generally
interested in that part of the deviation due to changes in load variable.
For a constant set point, the deviation caused by any one of the load
variables u is

=
(t —I
-f GiGoJ
(4-7)

From the standpoint of “following” changes of set-point v, but with

all load variables u constant, the deviation is

Cr = (4-8)
(1 + W,)'
The deviation resulting from a variation of set point is sometimes
employed to establish stability 'Of performance of the system.

Proportional Control
One of the simplest closed loops in automatic control is the propor-
tional control of the single-capacitance process shown in Fig. 4-2, On
Closed Loop in Automatic Control 81

the {riiyncal diagram a eitcle symbd indicateB that an autmnatic con*

titdler controls a variable of the process (LC indicates a level control in
this example). The controller operates a control (throttling) valve at
the inflow line.

Fig. 4r~2. Proportional control of a single-capacitance process.

The process equation is written under the assumptions that the up-
stream head at the control valve is constant:

-(7iTTr)"' + (wVi)'
where c = controlled variable = vessel head
R = resistance of outlet valve
T = RA = vessel time constant
A = vessel capacitance = area
m “ manipulated variable = inflow
u - load variable downstream head
82 Automatic Process Control

The control equation for proportional control is*

m= Kcfi = Ke(v - c) (4-10)

t
where Kc = proportional sensitivity
The system operational equation is obtained by combining equations
4-9 and 4-10 to eliminate variables m and c:

where Ti = T/{RKe 1). +

For a constant value of set point v and for
a step change of the load variable w, the deviation is

* " "
U/fc + + i)

where U = magnitude of step change in load. The differential equation

is found by inverting the operational equation 4-12:

A particular solution of equation 4-12 is plotted in Fig. 4-3 for several

values of the parameter RKc. A value of the parameter RKe = 0
represents the response of the process without automatic control. There
can be no doubt that the use of proportional control has considerably
reduce ! the effect of variation of load variable u in two ways:

1. Smaller offset: Offset Eo is a steady deviation resulting from a step

change in load variable and is characteristic of proportional control.
Without automatic control the offset is U and is equal to the change in

^ Ordinarily the proportional control equation is written

m* Ktfi H- M
where M is the value of inflow when deviation e is zero. The constant M is suppressed
by writing
(m - Af ) * Kce

and measuring the value of inflow m from the zero-deviation value M so that
. m » /C«e
This procedure is followed in equations in this chapter by measuring values of
all

the variables c, e, m, and u from the ‘^normal” value existing when the deviation is
zero This procedure makes it unnecessary to carry all the constants through each
step of manipulation of the equations.
 Closed Loop in Automatic Control 83

outlet'-valve downstream head. M^th proportional contnd the offset is

- U
B. - 14
(4 )
RKc + 1

Thus as the proportional sensitivity Kc is increased, the offset is reduced.

The offset characteristic of proportional control is caused by the definite
proportional relationship that must exist between deviation and manipu-
lated variable.

Fig. 4-3. Response of proportional control of a single-capacitance process.

2. Faster stabilization: With proportional control the time to reach

^ per cent of the final value is about four time constants or

Stabilizing time (4-15)

ItK-c I
1

Thus, as the proportional sensitivity is increased, the stabilizing

time is considerably reduced.
A high value of proportional sensitivity Kc reduces offset and decreases
an extremely high proportional sensi-
stabilizing time, and, in this case,
tivity could theoretically be used. In an actual application, there is
an upper limit of proportional sensitivity that cannot be exceeded
because excessive oscillation would result. The oscillation would be
caused by any existing measuring lag and controlling lag that were not
included in this analysis.

Example 4-1. Calculate the offset and stabilization time for tbe level
control system shown in the figure.
 I

Automatic Process Control

Given u = 2.0 cfs

* 0.1 cfs step increase
Ke^ 2 cfs/ft
A = 10 sq ft

The process equation is (if outflow u does not depend on head)

The controller equation is, for proportional control,

m = {Kc)e {Kc){v - c)

The system equation is

where T - system time constant = A/K^

For a step increase in the load variable only

+ /
= ^
'Phe offset is

Eo - — 0.05 ft decrease of head

Kc 2

The .stahiliziiiK tunc is about toui times the time conctant-

A 40
Stiibilizini; time = 4 77- ~ ~ 20 sec
Kc 2

Proportioriiil con hoi of a two-capaintance process, as in Fig. 4-4, is a

more practical example of automatic control. The process equation is

/?!
^ " (4-16)
(7’,s + 1) (7’vs + 1) T-.s + I

where c = controlled variable = head in lowei- vessel

R = resistance of lower outlet valve

7’i ^ time constant of lower vessel = RiA\

To = time constant of upper vessel = R 2 A 2

m = manipulated variable = inflow to upper vessel
V = load variable = inflow to lower vessel

Combining equations 4-10 and the proportional control equation to

eliminate the controlled variable c, the relation between deviati<m, aet
 1 1

Closed Loop in Automatic Control 85

p<^t, and load variable is

(Tys + 1) CTiS +1) .

R^ (r** + 1)
~
(r,* + 1) (Tjs + 1) + "
(Tis + 1) (TaS + 1) + ftiA.
(4-17)

Fio. 4-4. Proportional control of a two-capacitanoe process (see Fig 2-14)

This equation can be placed in a moie convenient form;

^ /_L_\ (T,s -f 10(^2^' “b i)

+ 1/ 7’V + 2f7’s + 1

( ) -J(Tgs + l) 4-18)
V/f,/f.+ 1/ rv + 2r7’« +
T1T2
where T characteristic time (4-19)
RiK, 4- I

f I = damping ratio (4-20)

ylAl\T2{R,K,--b 1)

The response of the controlled system to a change of load variable u

and with the set point v fixed may be found by returning equation 4-18
to difTerential form:

rj + 2fT* + .. + (4-21
86 Automatic Process Control

For a step change of load variable u, equation 4~21 becomes by sub-

stituting u ^ U

TH + 2m + (4-22)
and

Equation 4-22 is a second-order linear difPerentlul equation and its

solution depends upon the characteristic time T and the damping ratio
i. The solutions may be classified according to damping.

(o) Underdamped or oscillatory, f < 1.0.

(6) Critically damped f = 1.0.

(c) Overdamped f > 1.0.

Several solutions are plotted in Fig. 4-5. A control parameter R\Kc of

zero represents the response of the process without automatic control.
The effect of proportional control is to reduce considerably the deviation.
The choice of a value of proportional sensitivity Kc determines the
damping ratio f of the system. As proportional sensitivity Kc is made
smaller, the response is made more stable, but the offset is larger. A
compromise value of proportional sensitivity giving a damping ratio f
of about 0.33 is generally considered satisfactory. Then from equation
4-20 above

The “best'' proportional sensitivity may be calculated by equation 4-24

for this particular process.
The frequency of oscillation in cycles per unit time is given by

It is interesting to note that from the standpoint of stability there is

no difference whether the first vessel or the second vessel is the larger of
the two, because the damping ratio f and frequency of oscillation have
the same value when the time constants are interchanged.
The effect of dead time in proportional control is shown by the test
results of P^ig. 4-5, The proportional sensitivity must be greatly reduced
(proportional band increased) in order to maintain stability. Thus, the
offset is large, and the stabilization is slow. It is readily observed that
a small dead time causes serious consequences in proportional control.
Closed Loop in Automatic Control 87

The characteristics of proportional control are summarized in the

magnitude of offset accompanying a change in one of the load variables
of the process. If changes in load variables are small, proportional con-

2 3
Time, tlTi

Fio. 4-5. Proportional control of a two-capacitance procfeaa.

trol is almost always effective. Furthermore, if all the time constants

(lags) of the controlled system except one are small, the proportional
sensitivity may be made large and the offset is small under any conditions.
 88 Automatic Process Control
Example 4-2. A process (Fig. 4-4) has time constants of Ti » 10 sec,
and T% » 20 sec. The outlet resistance (Ri) is 10 sec per sq ft. Calculate
the proportional sensitivity for a damping ratio of one third. Calculate the
a change of downstream head of
offset for 1 ft.

From equation 4-24

Kc = *[(2 + i + 2)f - 1) = 0.91 sq ft/sec

From equation 4-2.3

10 X 1
0.99 ft
9.1 + I

Example 4-3. A process can be designed so that two vessels as in Fig. 4-4
have the characteri.stics:
(a) Ti = Tz = 00 sec, and = 1/8 sec per sq ft or (b) =* 30 sec,

T2 - 120 sec, and R - 1/8 sec per sq ft. Which design provides the least
offset? Wliich design provides the fastest response (frequency)?

(ft) A, = 8 X 1(1 + 1+ 2)| - 1] = 64.0 rf.s/ft

“ U
- ^ ^ (ri 1 )
It per unit of

(b) K. = 8((4 +}+ 2)1 - 1) = 104 cfs/ft

~ V ^ \ 0 009 ft per unit of V

[ ]3 /+ 1 ]

This example illustrates the rule that all lags in the system except one should
be small. The process above with the small tank and the large tank provides
less offset and faster response.

Integral Control

Integral control is often employed for control of fluid flow, liquid level,
and pressure. Consider integral control of the single-capacitance process
of Fig. 4-6. The process equation is

R
c VI + “ (4-26)
7V + 1 + 1

where c * controlled variable =' vessel head

T = vessel time constant == RC
R = resistance of outlet valve
C = vessel capacitance * A =*= area
m» manipulated variable » inflow
u = load variable = doumstream head
Closed Loop in Automatic Control 80

The control equation is

Fig. 4-6. Integral control of a single-capacitance process.

where 1\ = integral time of integral controller. The system equation

is ol)tained hy eliminating m and c between equations 4-20 and 4-27.

*
^ ^It
/ Ts+\
+ 2^T,k + 1
T^/_
R \t: +
1

2f7’c6' H" t)"

(4-28)

wheie T, = chai-acteristic time

r damping ratio

Reverting to differential form for a step change in tlie load v ariable

(ti - U\
-f 2iTri +€= 0 ( 4 -29 )
90 Automatic Process Control

This is a second-order linear differential equation and its solution depends

upon the damping ratio.

Several solutions are plotted in Fig. 4-7. The offset is zero as pre-
dictedby the system equation, because the integral control action forces
a complete return to the set point when there is a change in load.

Fia. 4-7. Response of integral control of a single-capacitance process.

The integral time T, of the controller must be selected to provide

proper damping. no offset, a moderate damping ratio may
Since there is

be chosen and a value of about one third is usually satisfactory. Then,

the integral time can be computed from

T, = (4^0)
With this value of integral time a minimum deviation is obtained without
encountering excessive oscillation.
Instability is po-ssible with integral control when the process does not
have self-regulation. The arrangement of Fig. 4-8 illustrates this. The
process equation is (if the outflow does not depend upon head)

c (4-31)
Cs ,
C«

where c = controlled variable = head in vessel

C = capacitance of vessel = area
tn = manipulated variable = inflow
u = load variable =* outflow
Closed Loop in Automatic Control 91

Employing the control equation and the deviation equation as in the

example above, the system equation becomo*?

CTiC +e = CTfi + r.ti (4-32)

The transient solution to this second-order differential equation is a

sinusoidal function without damping. Therefore, this system arrange-

Fio. 4-8. Integral control of a single-capacitance process (unstable).

ment produces a continuous oscillation of the controlled variable.

Fortunately most industrial processes have self-regulation and this
difficulty is usually avoided.
Integral control may be employed for control of fluid-flow rate as
shown in Fig. 4-9. The flow controller FC adjusts a throttling valve in
series with the metering device so as to maintain a constant differential
head — h\ across the metering device. The “process” equation is

(see Chapter 2)
c = Kam + (wo - W2 ) (4-33)
92 Automatic Process Control

where c « controlled variable liquid flow rate

Ka - ^ constant
Kb = Si constant
m = position of throttling valve

Equation 4-33 is valid only in the operating region of flow rate whei*e
the partial derivative factors involved in the /C's are nearly constant.

Kio. 4-9 InleKrai coritn)! of lujuid-How

Employing the same control equation and eiror equation as in the

examples above, the system equation is

Is KiTsii\) KbTs
^ ~
Ts +1 Ts + I 7V+ ‘I

where T = time constant = Tr/Ku- For a step change of load variable

U2 ,
with upstream head ?/y and set point v constant, the differential
equation becomes
Te -t (• = 0 (4-35)

The solution to this equation is plotted in Fig. 4-lU.

is no offset
For a step change in load variable there llie response
to changes in upstream head n© same as for changes in downstream
is the
head but in the opposite direction. The maximum initial deviation
depends upon the magnitude of change in load variable and the ratio of
Closed Loop in Automatic Control 93
partial derivatives (/Cs). The controller cannot reduce the maximum
initial deviation because the change in upstream or downstream head
immediately affects the flow rate before the controller is able to act.
The time constant of the controlled system response T is reduced by
decreasing the controller integral time. However, the integral time
cannot be made extremely small because of other lags (time constants)
that usually exist in the system.

Fig 4-10. Response of integral control of liquid-flow rate.

The characteristics of integral control contribute to the reduction or

elimination of offset. Furthermore, integral control is effective only
when there are few energy-storage elements in the process. Although it

has not been demonstrated here, a few actual tests will show that a
small dead time makes a system with integral control uncontrollable
(steady or increasing oscillation).

Example 4-4. A process has a capacitance of 60 Btu per deg F and a

resistance of 1.0 deg F per Btu per second. Calculate the integral time and
the period of oscillation for integral control.
From equation 4-30, and with a damping ratio of

The period of oscillation is found from the characteristic time (Tc),

2rTc _ J60 X 26.7

- 84.0 sec
 (

94 Automatic Process Control

This example illustrates the point that integral control of processes with laige
time constants is slow that is, the period of oscillation is long.
;

Proportional-Integral Control
Proportional control and integral control are often combined in order
to obtain the advantages of inherent stability of proportional control
and of elimination of offset by integral control. Proportional-integral
control is the most generally useful of all types of control. We will
consider examples of this type of control, first, for processes without
storage elements, and second, for single-capacitance processes.
Proportional-integral control of fluid-flow rate is shown in P'ig. 4-9.
The process equation is,

c = Knm + Ki,(wo - U2 ) (4-36)

where the quantities are defined as for equation 4-33. For proportional-
integral control,

w = A-, 1 + e (4-37)
(
where Kc = proportional sensitivity (inverse of proportional band)
T, = integral time (inverse of reset rate)

Employing the deviation equation to eliminate c and m from the above

two equations, the system equation is

— — - i. \ '
_ (4-38)
KaKATs+iJ ^ k„kAtx+
uo)

KqKc d" 1
where T - Ti
KaKc
The response of the system to a step change in either load variable may
be found from equation 4-38. Foi a step change in downstream head
(« 2 ),
Te +e= 0 (4-39)

The solution to this equation is plotted in Fig. 4-11 for representative

values of Ka and Kb. Comparing this response to that of Fig. 4-10 for
integral control only, it is seen that the effect of proportional control is

to reduce the initial deviation. Naturally the offset is zero because of

the integral action of the controller. Although the curves of Fig. 4-11
would indicate that the proportional sensitivity could be increased to a
very high value, and the integral time decreased, in a practical applica-
tion the existence of other lags in the controlled system prevents using
high controller sensitivity and small integral time.
96 Automatic Process Control

Another example of proportional-integral control is illustrated in

Fig. 4-12. The level in the tank is controlled by adjusting the throttling
valve at the tank inlet. The process equation is

R 1
c “ '
m "T“
.

u (4-40)
Ts -h 1 Ts+l
where c = controlled variable = A = vessel head
H = resistance of outlet valve
T = vessel time constant= RC
C = vessel capacitance= A = vessel area
m = manipulated variable = inflow
u = load variable = downstream head

Time, iJT

Fio. 4-13. Response of proportional-integral control of a single-capacitance process.

Combining the process equation 4-40 with the control equation 4-37 to
eliminate w, and the deviation equation to eliminate c, there results

T,s
’
" ~ TMTs -b RK,(l\s + 1)
'
TMTs H- + RKr{T,s -b 1)
-f 1 ) I
)

(4-41)

This equation may be placed in the' more convenient form

Closed Loop in Autonmtic Contn^ 97

where Te VT{T/RKt “ characteristic time

, j{RKc + I)*r.- , .

Changing proportional sensitivity

Time, mm
Changing integral time

Fig 4-14. Adjustment of controller sensitivity and integral time for proportional-
integral control.
98 Automatic Process Control

The response of the system to a step change in load variable u but

with constant set point v is found from equation 4-42. Rewriting the
equation to differential form

Tch + 2fTcC + e = 0 (4-43)

The particular solution for given initial conditions is sho\\Ti in Fig. 4-13.
The damping ratio in I'ig. 4-13 is maintained constant at 0.707, a

value that produces modeiate oscillation. The chaiacter of the response

depends upon the ratio of controller sensitivity and integral time. The
u‘.-.ults show that better response is obtained if large controller sensitivity

and short integral time are used.

Proper values for controller sensitivity Kc and integral time Tt may
be selected for any controlled system, but their calcuLation is complicated
except for simple processes. The effect of these adjustments is illustrated
in Fig. 4-14. The overall effect is relatively small in view of the wide
range of variation of the control oarameters. A proportional sensitivity
between 20 and 40 Ls approximately optimum for this example. A higher
proportional sensitivity produces greater oscillation and might actually
result inan ever-increasing oscillation until some limit is reached. Refer-
ring to the bottom figure, infinite integral time (zero reset rate) allows
the usual offset. An integral time of approximately 5 min is optimum.
A further decrease of integral time usually results in excessive oscilla-
tion and may actually produce an increasing iimplitude of cycle. The
frequency of oscillation due to excessive integral action is much smaller
than the frequency of oscillation caused by excessive proportional
action.
Proportional-integral controlmay be employed for any process having
characteristicsamenable to the use of either response used separately.
The limitations of proportional-reset control arc due only to the excessive
stabilization time required when the process has many energy storage
elements or dead time.

Example 4-5. For proportional-integral control of process shown in the

figure, calculate the relation between proportional sensitivity and integral
time for a damping ratio of onc-third. The process equation is

1 1

where C = process capacitance. The

controller equation is
Closed Loop in Automatic Control 99
Combining to eliminate m and adding the deviation equation to eliminate e,

and with outSow u » ilf we have ,

V
CT,8^ + KciTiS + 1)
or

V Tc^8^ + 2f + 1

fCTi\^
where Tc
kkJ , . 4C J
For a damping ratio of one-third
4C
KcTi
9
Use of Derivative Control
Derivative control action is useful for control of systems having a
very large number of storage elements and for control of systems having
dead time. As an example, consider the control of head in the lower
vessel of Fig. 4-15. The process equation is

R _i_
(4-44)
i',s(r2s + 1) c,«“
where c controlled variable head in lower vessel
Cl capacitance = = area of lower vessel
Ti time constant = RCi
'
T2 upper vessel time constant = RC 2
C2 capacitance = A2 = area of upper vessel
R resistance of upper outlet valve
m manipulated variable = upper vessel inflow
u load variable = lower vessel outflow

For proportional derivative control the equation is

m= K,{TdS + \)e (4-45)

where Kf = proportional sensitivity

Ta = derivative time

Employing the usual deviation equation, the system equation is obtained

by eliminating variables m and c,
Tis{T2S 1)

Tis(T2S + 1) + KKciT^s + 1)

RiT2S +1)
TMT2S + 1 + ) + 1)
100 Automatic ProoeBS 'Control

Thk equatim may be placed in more convenient form

1 r,«(r,« + i) (Tcc + i)
V +
.

(4-»7)
RK^ r.V + 2iTcS + 1 (k) + 2f Trf + 1

Time, f/Tg

Fig. 4-15. Proportional-derivative control of a two-capacitance process.

where Tc = ^T^T^/RKc = characteristic time

(T, + RK,Tif damping ratio

A LiTi^ rr rr

The response of the system to a step change in load variable u may

be found using the system equation 4-47 so that

T‘e + 2fT-eC + e - U/Kc (4-48)

Closed Loop in Automatic Control 101

whm V is the change of load vai-iable u. The particular solutioii is

plotted in Fig. 4-15 for a damping ratio of one-third. Then, as derivative

Fio. 4-16. Um of derivative control.

action is added to the controller {Td ~ O. 5 T2 ), the increased damping

becomes readily apparent. The advantage of derivative action is that

the proportional sensitivity may now be made larger without producing
 :

102 Automatic Process Control

excessive oscillation. This, in turn, reduces offset. It is sometiines

possible through the use of proportional derivative control to reduce
offset to such a small value that a integral control action would not be
required.
Proportional-integral-derivative control is more effective for control of
processes with many energy-storage elements than the proportional-
integral control action used alone. The response of controlled systems
with many energy-storage elements is difficult to calculate, so we rely
upon the results of tests of an actual system as shdwn in Fig. 4-10. For
controlled systems with many energy storage elements (Fig. 4-16 (A)),
the addition of derivative control action considerably reduces the maxi-
mum deviation and the time to stabilize at the set point. With deriva-
tive control, the proportional sensitivity may be increased and the
integral time reduced. For controlled systems with dead time (Fig.
4-16(5)) similar results are obtained.

Comparison of Proportional, Integral, and Derivative Control

Each of the modes of control is applicable to processes having certain
characteristics, and the importance of designing the controlled system
must not be overlooked. If all processes consisted only of one capaci-
tance without dead time, there would be no necessit}’^ for any complex
control actions. The difficulty is that few if any industrial processes are
80 simple in dynamic structure.
The effectiveness of the various modes of control may be illustrated
by comparing the responses to a load change on a given process. The
process used for comparison is that of Fig. 4-4

* “
“ + \Tis+
(Tis + DCrj* + 1) 1)

where the vessel time constants are equal, T = 20 sec. As shown in

a change of inflow to the lower vessel may result in appreciable
Fig. 4-17,
deviation for several minutes. The following comments apply to each
type of control. The numbers correspond to the numbered curves of
Fig. 4-17.

1. Proportional-derivative control provides the smallest maximum

error because the derivative part qf the response allow.s the proportional
sensitivity to be increased to a high value. The stabilization time is

the smallest because of the derivative action. Offset is allowed but is

only half that experienced without derivative action.

2. Proportional-integral-derivative control has the next smallest
maximum deviation and offset is eliminated because of the integral
Closed Loop in Automatic Control 103

action. Notice, however, that the addition of integral action markedly

increases the stabilization time.

Curve
Mode
Control
of
Prop
Sensitivity.

RK,
Denv
Time.

I’d.
sec
Period
of
Cycle.

sec
Damping
e-i/r
T,
sec
Max
Error

units
B
1 Proportional^denvative 16
- 0.9 32 15 018 0.06

4
Prop int

Proportional
denv

Proportional integral
10

4
22

-
3
a
H
2.0 44

45

66
20

20

30
021

0j»

0.37
0

0.11

5 Integral - 35 210 100 069 0

3. Proportional control has a larger maximum deviation than con-

trollers with derivative action because of the absence of this stabilizing
influence. Offset is also larger.
4. Proportional-integral control has no offset because of the integral
action. The unstabilizing influence of integral response is reflected in

the large maximum deviation and the persisting deviation.

104 Automatic Process Control

5 . Integral control is best suited for the control of processes having

little or no energy storage and the results of the comparison are not
representative of all integral control. However, on this process, the
results indicate a large maximum deviation and a long stabilization time.

With the results of this comparison in mind, it is logical to ask why

proportional-integral-derivative control action not universally em- is

ployed. The answer is generally based on economic reasons, because

each additional control action usually requires an additional piece of
equipment that must be purchased, installed, a'nd maintained. In
addition, each control action may require adjustment of a parameter
such as proportional sensitivity, integral time, or derivative time. This
often requires considerable installation and maintenance time in order
to obtain the proper adjustment of parameters. Consequently, the
simplest mode providing adequate control is usually the most desirable.

Static Error, OfTset, and Velocity Error

Deviation that results from a changing load or set point may be cal-
culated directly without recourse to the complete solution of the system
equations. For steady or steadily changing disturbances the particular
integral of the differential equation provides the steady-state solution.
Static error results if, when the set point is changed from one value to
another value, the controlled variable does not follow exactly. In other
words a steady deviation results. A step change of set point is given by

v{t) =0 ^ < 0
(4-60)
v{t) = F / ^ 0

Thus V is the magnitude of change of the set point. Employing equation

4-8 for the set point deviation

_ V
(4-51)
1 -f- G1G2

This equation represents the deviation as a function of time since

Gi and G2 generally contain the operator s. For steady-state behavior
all derivatives are zero, thereby implying that the operator 5 can be
replaced by zero. The static error can therefore be calculated from the
relation

(4-52)
f -^0

In order to evaluate the static error it is only necessary to know the

individual system functions (?] and G2 .
Closed Loop in Automatic Control 105

Offset may result when the load on the system changes, thereby causing
a steady deviation. For a step change in load u,

u(t) =0 / < 0
(4-63)
u(i) ^ U < ^ 0

where U is the magnitude of change of the load. Employing equation

4-7 for the load deviation, there results

-NU
(4-64)
1 + G1G2
This equation represents the deviation as a f unction of time. For steady-
state behavior we again replace s by zero and the offset is given by

NU 1
(4-65)
1 -h G1G2]

The offset or load-static error as it is sometimes called may be calculated

when the system functions N, Gi and G2 are known.
Velocity error results when the set point is a steadily changing quantity,
and the controlled varial)le, in attempting to follow the steadily changing
input, ^‘lags’^ behind with a steady deviation. For a set-point change
described by

J=A, t^O (4-66)

at

where A = rate of change of the set point. In operational form,

sf; = A (4-57)

Substituting the latter equation in the deviation equation 4-8, there

results

fit,
= (4-68)

This equation represents the deviation as a function of time. For steady-

state behavior all derivatives are zero so that the steady-state velocity
error is given by'

E. - lim
*—*0
--^1
.5 1 -h G1G2J
(4-59)

In order to evaluate the velocity error it is only necessary to know the

system functions Gi and G2 .

* The proof of this relationship is best found by employing the final value theorem
of the La Flare transform.
106 Automatic Process Control
Eiample 4-6. Calculate the etatio error for the eyatem in Fig. 4-4

_V
E, lim
“
=
•-»o RiK^
"
^ ifc + l

(Ti* + 1)^

Example 4-7. For the proportional-control system in Fig. 4-4, calculate

the offset.
RiU
E, lim

(7’i. + i)(r2« + i)

Compare this result to that of equation 4-23.

Example 4r^. Determine the offset for the integral-control 83rstem of

Fig. 4-6.
___U
I 'T* 1

E, lim „ -0
1 4 ^
^ ^8(^8 + 1)J
Thus integral control eliminates offset.

Example 4-9. Determine the velocity error for the system in Fig. 4-6.

i + A.
^ T’.AT’i + l/J
Therefore this type of system may have a steady velocity error.

Example 4-10. Determine the velocity error for the system in Fig. 4-4
with proportional-derivative control:

jF, = lim - -
•-*0 s
,
RiKc(l + Tds)
\Ti8 + 1){T2S 1)

The limit is infinite, meaning that the deviation increases without limit and
that the controlled variable will gradually fall further and further behind
the set point.

Two-Position Control
Two-position control has the widest industrial and domestic use on
processes having notmore than two energy-storage elements. In opera-
tion, two-position control is very simple, but in theory the action is
difficult to analyze because of the discontinuous nature of changes in the
manipulated variable. Nevertheless, problems of two-position control
can be solved by quantitative consideration. In the discussion that
Closed Loop in Automatic Control 107

follows, two-position control* will be considered as having a differential

gap and dead time, because few, if any, industrial controUed B3r8tem8 are
without them.
The most elementary case of two-position control is illustrated in
Fig. 4-18.The level control LC turns on the inflow when the level ip

Fig. 4-18. Two-position control of a Bingle>capacitanne process.

too low, and turns it off when the level is too high. For an increase in

level (valve open),

Cc —Jn - u when c<i;-fG, c>0 (4-60)

and for a decrease in level (valve closed),

Ct = —u when c>v — Cr, c<0 (4-61)

where C= vessel capacitance= vessel area

= c controlled variable= vessel head
m= manipulated variable = inflow
u = load variable = outflow
V = set point
G= differential gap
108 Automatic Process Control

For constant values of manipulated variable, set point, and load, the
above two equations may be integrated. For an increase in level

e, = when «, < + <7.^, > 0 (4-62)

For a decrease in level

/ -f Kb when ei > - G, e, < 0 (4-63)

where ei — —c = (c — t^) = deviation

U = value of load variable n
M= maximum value of manipulated variable m
K = constant of integration

Equations 4-G2 and 4-03 describe the action of the system and are
As might be expected, the oscillation consists of a
plotted in Fig. 4--19.

Fig. 4-19. Two-position control of a single-capacitance process

series of connected straight lines. The controlled variable ^^overshoots''

the differential gap by a time equal to the dead time, L. Thus

A (4-64)

where A = amplitude of oscillation. The double amplitude or total

amplitude The amplitude of oscillation is not dependent upon
is 2i4.

the magnitude of the load variable. From Fig. 4-19 it is noted that a
change of load variable may cause an unsymmetrical action, but what-
ever reduction of amplitude is gained on one side is lost on the other side.
The amplitude also depends upon the size of the inlet valve Af, the dead
Closed Loop in Automatic Control 100
time L and the process capacitance C. For small amplitude the process
capacitance should be large, the dead time should be small, and the inlet
valve size should be just larger than that necessary to handle the largest
load.
The period of cycle is calculated from the amplitude divided by the
rate of change of the controlled variable:

where P= period of oscillation. Large, but not excessive, p>eriod is

usually considered desirable because this reduces wear on the controlling

mechanisms. Large period and small amplitude requires large process
capacitance C and small inlet valve size Af.

Example 4-11. The process of Fig. 4-18 has a steady outflow of 0.5 cu ft
per sec, a vessel area of 1.0 sq ft and a dead time of 0.5 sec. The differential
gap is 1 in. Calculate the amplitude and period of oscillation.
First calculate the valve size M. This should be about 50 per cent greater
than the maximum outflow;

Af « 1.5 X 0.5 = 0.75 cfs

The amplitude is

= 14-
0.75 X 0.5 X 12
.4 3.25 in.

The period is

The second example of two-position control is shown in Fig. 4-20,

differingfrom the previous example only in having a time-constant

response. For an increase in level (valve open),

Tc 4- c = Hm 4“ u when c<i;4-G, c>0 (4-00)

and for a decrease in level (valve closed),

Tc r — u when c > v - c <0 (4-07)

where T = vessel time constant =- flC

R = outlet-valve resistance
C = vessel capacitance = vessel area
c = controlled variable = vessel head
m= manipulated variable = inflow
u = load variable = downstream head
V= set point (at center of differential gap)
C = differential gap
no Automatic Process Control

For constant values of manipulated variable, set point, and load, the
above two equations can be integrated. For an increase in level

Cl = + KjCl - when ei < +G,ei > 0 (4-68)

Fig 4-20 Two-jioHition control of a single-capac’itance process.

For a decrease in level

€] = -h Cff when ei > — G, ei < 0 (4-69)

where ti - -e = - V) — deviation
(c

Ea = /fA/ -f- F ~ maximum potential

(/
- {U - V) - minimum potential
et,

M - maximum value of manipulated variable

U ^ value of load variable
V = value of set point
Kb = constant of integration

These two equations describe the response of the controlled variable and
are plotted in Fig. 4-21. The oscillations are composed of segments of
exponential curves Without the operation of the controller, the con-
 Closed Loop in Automatic Control 111

would have followed the dotted curves to the potential

trolled variable
values The 'overshoot” beyond the differential gap is due to
(e®, c^).
*

the dead time L. The period of oscillation depends directly upon the
process time constant and upon the amplitude of oscillation. The
amplitude is reduced by a small value of dead time L and a large value
of process time constant T.
Two-position control of processes having more than one capacitance
is similar in general respects to the control of single-capacitance pro-
ceases. The additioi^l capacitances of the system act to “round-off”
the peaks of the controlled variable oscillation.

Fig 4-21. Two-position control of a single-capacitance process.

The differential gap is often intentionally introduced in two-position

control in order to increase the period of oscillation and thereby reduce
wear on the controlling mechanisms. As shown by equation 4-65 it is
effective for this purpose. A negative differential gap is sometimes
employed to reduce the amplitude of cycle, and, according to equation
4-64, this would be effective. If used carefully, and if the system param-
eters such as resistances and capacitances are constant, the amplitude
of oscillation may be reduced somewhat.
In conclusion, two-position control is best suited for the control of
single-capacitance processes in which the process capacitance is very
large and dead time (or equivalent dead time) very small. There is
little effect due to changes of load variables other than to alter the

symmetry of the cycle.

112 Automatic Process Control

Single-Speed Floating Control

Single-speed floating control is often useful for control of prooesses
or systems having little or no capacitance and self-regulation. The con-
assumed to have a neutral zone; otherwise the controlled vari-
troller is
able would oscillate continuously, because the manipulated variable
would be changing continually in one direction or the other. A single-
speed floating controller without a neutral zone acts in a manner similar
to two-position control.
The operation of single-speed floating control is illustrated by the
diagrams of Fig. 4-22. For a process such as liquid-flow rate which has
no capacitance, a change in load variable (no or U 2 ) causes an immediate
but proportional change in controlled variable c. Then, as the controlled

Fiq. 4-22. Single-speed floating control of liquid-flow rate

variable moves out of the neutral zone, the throttling valve begins to
close at a constant rate and the controlled variable returns to the neutral
zone as shown in Fig. 4-23. (0)hapare this result to the integral control
in Fig. 4-10 for the same process.) Therefore, single-speed floating
control is stable for processes without capacitance or energy-storage
elements as long as dead time is small. With dead time, the overshoot
into the neutral zone requires increasing of the width of neutral zone
 ^

Closed Loop in Automatic Control 113

N and decreasing the floating rate so that the overshoot is less than the
width of the neutral zone.

0
Time —
Single -capacitance process with self-regulation

Fio 4-23. Response of single-speed floating control.

The operation ol single-speed floating control of processes witlumt

self-regulation is unstable. When a change in load variable occMirs, the
controlled variable moves out of the neutral zone, and the iillet valve
liegins to move at a constant rate. The controlled variable oscillates
 114 Automatic Process Control

continuously across the neutral zone because the controlled variable, on

each swing, is returned to the neutral zone at a rate equal to that rate
at which it left the neutral zone on a previous swing. If dead time is
present the amplitude of oscillation increases on each swing because each
succeeding slope through the neutral zone becomes greater.
Single-speed floating control of a process with a single time constant
is shown in Fig. 4-23. The response of the controlled variable to a
change in load variable is indicated in Fig. 4-23 for various values of
floating rate (rate of change of manipulated variable). A fast floating
rate causes continuous oscillation of the controlled variable because the
process has not had time to reduce the slope in the neutral zone on each
successive swing. The proper floating rate combined with a neutral zone
of proper width will produce a well-damped response of the controlled
variable.
Single-speed floating control for processes having more than one
capacitance is not generally successful, because in order to achieve
stability the floating rate must be made so small that excessive over-
shoot and amplitude of oscillation results.
Summarizing, single-speed floating control is most effective for control
of processes having little or no energy storage. Dead time or equivalent
dead time should be small. Offset is restricted to the neutral zone.

Ziegler-Nidhols Method
The calculation of the response of the controlled system is usually
difficult because^ of the mathematical complexity of solving the high-
order differential equations for each new control problem. In addition,
one solution of these equations is generally not sufficient because it is

desired to explore the effects of several parameters of the system. This

situation often forces the use of computing machines (differential
analyzers) in finding and making compromises in design of various por-
tions of the controlled system.
The Ziegler-Nichols method affords a systematic way of determining
controller settings for best performance of a controlled system and, in
*
addition, provides a simple quantitative pi(^ture of control system be-
havior. This method is based on the approximation of the behavior of
a complex dynamic system by the behavior of a system composed only
of a single dead-time element and a time-constant element. The appar-
ent dead time and apparent time constant are then used to calculate
performance of the system.
The determination of the system apparent dead time and apparent
time constant may be made either by calculation or by test. Generally
if a system is complex dynamically, it is also difficult to calculate a
Closed Loop in Automatic Control 115

response. Consequently the Zie^er-Nichols method is primarily an

experimental one.
The open-loop transient test is used to determine the magnitude of
the apparent dead time and apparent time constant. As shown in Fig.
4-24, the closed loop opened at a point immediately following the con-
is

troller, and the entire system is maintained in a steady state at the

normal operating values of all variables. (Generally the loop is cut just

Fio. 4-24. Open-loop transient test.

after the controller incchanism by employing the manual control

apparatus A step change of the variable th is introduced, and the
result of this change is recorded at point b just before the controller.
Thus, a step-change response is obtained for the open-loop system
excluding the controller itself. The following precautions must be
employed during this test:

1. The step change must be as small in magnitude as can be recorded

and interpreted.
2. The step change should begin with the controlled variable hist

below the set point and end just above the set point.
3. The step changes should be made in both directions if it is su^

peeled that the system is not linear.

116 Automatic Process Control

4. None of the variables should be allowed to attain maximum or

minimum values or in any way become limited during the test.
There should be no appreciable friction or backlash causing hys-
5.

teresis orno appreciable dead zone in any elements of the system.

6. All elements having appreciable lag must be included in the open

loop test. If the controller has a lag, the controller should be set for
proportional control action only and the response recorded at m* in
Fig. 4-24 instead.

The recorded open-loop transient response is almost always an S-shape

curve as indicated in Fig. 4-24. 1'he zero point of the step change must
be carefully marked.
The open-loop response may be approximated by making the meas-
urements shown in Fig. 4-24. First, a line is drawn tangent to the point
of inflection of the response curve. Second, a horizontal line is drawn
tangent to the final value of the response curve. The measurements
made are

K = Magnitude of change in units of re<;orde<.l variable

N = Reaction rate in^ units per minute
L = Apparent dead time in minutes
M = magnitude of change in units of variable being changed
A simple check of the approximation may be made as follows: A hori-
zontal line at 63.2 per cent of the total change should give a value of T
in minutes approximately equal to the value of A/A. If this value
does not check within 15 per cent, either the slope line is drawn incor-
rectly or there may be a serious nonlinear behavior of the system. If

the apparent dead time and the apparent time constant T taken in

each direction are within about 10 per cent, a simple average may be
used. If they are not within 10 per cent, there is a serious nonlinear
behavior in the system.
The lag ratio should next he calculated;

NL = L
R - — (dimensionless) (4-70)
A I

If the process has no self-regulation, the value of K is extremely large

and the lag ratio R is nearly zero. If, on the other hand, the p^oces^-
reaction rate ;V is extremely large (rapid), then the lag ratio R is also
extremely large
I’he following equations may be used for calculating the desired
factors regarding performance of the controlled system. The Cohen
 Closed Loop in Automatic Control 117

Coon equations* are given for proportional, proportional-integral,

proportional and proportional-integral-derivative control.
derivative,
The equations for integral control and two-position control are pre-
sented by the author.^
Proportional Control, The control action is described by

The setting of propoftional sensitivity for one-quarter ratio of succeed-

ing amplitudes of transient response,

M / /2\ units of m
(4-71)
AL \ 3/ units of e

For this setting, the following transient response characteristics result:

P= (4-72)

where P period of oscillation. Also,

NL ( ^ \ units of c
Eo (4-73)
M \3 -f 4/?/ units of rn disturbance

where Eo = offset. The step disturbance may be located at any point

in the loop but it is expressed in terms of the equivalent value of manipu-
lated variable m, usually a valve-position change, giving the same
steady-state offset.
D= l.bEo (4-74)

where D maximum deviation.

integral Control, The control action is described by

m ^ 1
(4-75)
c TiS

The setting of integral time for one-quarter ratio of succeeding ampli-

tudes of transient response is

NL^ /I 4- unit of c
(4-76)
4M \ / units of m per min

*
“Theoretical Consideration of Retarded Control" by G. H. Cohen and G. A.
Coon, ASMS Trans., Vol. 76, p. 827, July, 1963.
•“Phase Plane Analysis" by D. P. Eckman, Mech. Engrg., Vol p 682, July
1953.
118 Automatic Process Control

For this setting the following tranaientrresponse characteristics result:

/> = 4.8L(l + i) (4-77)

where P = period of oscillation.

PropoTtionaUDerivative Control The control action is described
by

= KAl + Tis) (4-78)

J ,

The settings for one-quarter ratio of amplitudes and minimum offset

are
units of m
(4-79)
units of c

(4-80)

For Utese settings the following transient characteristics result:

(^ 1 )

units of c
E" "I (4-^2)
M VlO
<10 H- 9r)
9R/ units of
1 m disturbance
D - l.5E^

If independent adjustments are employed at the controller

(Ke -h Kd8)j set the derivative adjustment Kp to the value given by
the product KcT^. Notice that settings of derivative time cannot be
obtained for processes having an apparent dead time L greater than
three times the apparent time constant K/N.
Proportional^ Integral ControL The control action is described by

+ (4-»3)

The settings for one-quarter ratio of amplitudes and a compromise

between minimum area and period are

\ units of m
(4-84)
*
NL \10 ^ 12,) units of c

min (4-85)
*
\9 + 20R/
Closed Lwp in Automatic Conti-oi 1 19

If independent adjustments are employed at the controller (A’e A|/ s), +

set integral adjustment (K,) to the value given by the ratio KJTi.
PropartionaUIntegraUDerivative Control. The control action is

described by

(4-86)

The settings for one-quarter ratio of amplitudes, asymptotic return,

and minimum area art

(4-87)
:(7f)
" (4-88)
)
,11 + 2R/

32 + 6R\ (4-89)
,13 + Br)
Note: 6.
Ti

If the control action is described by

(4-90)

le settings must be calculated by trial and error from

K.' = (4-91)
1 + 11
T,'

Ti' = K,Ti
Kc'

K,'Ti
T,' = (4-92)
K,

f the control action is given by

- =
e
—+
s
/C. + ( 4-«8 )
120 Automatic Process Control

the settings must be calculated from

(4-94)

(3 + 2R)M
" GATL*
9

Two- Position Control. The open-loop test for two-position con-

trol is conducted in much the same manner as for any other control
except that sometimes necessary to employ larger changes in the
it is

manipulated variable m. This is because the final element very often

has only two positions, a relatively large value and a relatively small
value.
The amplitude and period of resulting oscillation for the closed-loop
svstem may be calculated from

A = (1 - «-") + (4-96)
2M
and

where A = ainjilitudc of cycle, units of c

infix
— maximum value of manipulated variable, units of m
^min = minimum value of manipulated variable, units of m (This
is often zero.)
G = amplitude of differential gap, units of r

P = period of cycle, minutes

V - set points, units of c

Note that the amplitude A and the differential gap G are expressed in
half-amplitude —that is, twice these values give the total cycle and total
gap. The equation for amplitude is adequate for any two-position
control, but the equation for period is valid only when the two-position
cycle is relatively symmetrical. When the lag ratio R is less than
about 0.2, equation 4-95 may be written

[K{M max min)

.4 = (? -f -o] R (4-97)
2M
Example 4-12. .\ii open-loop transient-response tent shows a temperature
.system to have an apparent dead time L of 0.8 rain, a reaction rate -V of 0.5
Closed Loop in Automatic Control 121

degrees per minute, and a process ultimate temperature change /T of 10 degrees

for a change in manipulated variable Af of 10 per cent of valve setting. Calcu-
late the proportional sensitivity for proportional control.
From equation 4-70
0.5 X 0.8
10
From equation 4-71

cent/degree

Example 4-13. For ii process with R == 0.5, what is the decrease in period
of oscillation if derivative control is employed with proportional control'^
From equations 4-72 and 4-81

^ (5 + 6ie)(10-f 7P)
^ 8_
Pp (6+ l0P)(12-h3/?) 11

or a decrease of about 27 per cent.

Example 4-14. For a process with R - 0, what is the decrease in offset

allowed when derivative control is employed?

or a decrease of about 20 per cent.

Example 4-15. An
open-loop transient test shows a pressure system to
have an apparent dead time L of 0.04 minutes, a reaction rate of 2.5 psi N
per minute, and a process ultimate change K
of 10 psi for a change of valve
setting M of 5 per cent. If the set point is 100 psi, and the differential gap
is 0.5 [)si, calculate the amplitude and period of on-off control.

NL 2.5X0.04

For small P, use equation 4-07

.1 = 0.5
+ [— -2^^^ - 05
.

]
0.01 = l£j

For smiill H equation 4-90 is

P„
4Z,^
Z„
4L/1, 1 +Z \
^ tanh g (2 " 1 - z)
or

^ ~
R R {KM„„ - VM)
O.IC 1.5X5
0.01 10 X 100 - 100 X 6
0.24 min « 14.4 sec
 122 Automatic Process Control

PROBLEMS

4-1. Explain physically why offset occurs, using the mechanical float-

level control of Fig. 3-3 in the previous chapter.

4-2. Explain physically why offset is zero, using the mechanical integral
control of Fig. 3-4 in the previous chapter.
4-3. If a process-control system contains a capacitance element l/Cs,
prove that integral control is not necessary to eliminate static error if only
time-constant elements are in the remainder of the system.
4^. Prove that velocity error is not zero if proportional-integral control is

used and if the remainder of the system contains only time-constant elements.
4-5. Sketch the resulting oscillations for on-off control when (a) differ-
ential gap G is zero; {b) dead time L is zero; (c) capacitance C is made small;
(d) control-element size M
only 10 per cent larger than load flow U,
is

4-6. For processes with small dead time, how much should the propor-
when derivative action is added?
tional sensitivity be increased
4-7. For processes with small lag, how much should the proportional
sensitivity be decreased when integral action is added?

4-8. In Fig. 4-1, Oi = /Cc, (7i = ~ ,

iV = ~ . Show that the time con-
es es
stant of the system response is C/Kc. Would this system have offset?

4-9. In Fig. 4- 1, G, = X, (l + ,
G, = ,
iV = . Calculate the

condition for critically damped response. . .

4-10. In Fig. i-lGi’- K,{T + 1), N = Calculate the

system time constant.
4-11. InFig.4-l,G, - A^,G, = JV = 7 =r^. Calculate the
offset.
^* + 1
,

+\
M2. InFig.4-l,G, = /Ce + -,Gj = ^=r^,JV = ;=r-^. Calculate
s Is -t I i « 1 -h
the relation necessary for a damping ratio of 0.707.

4-1.. I. Fig. 4-1. ft - ft . 14 -

,

^
Calculate the proportional sensitivity for critical damping. Calculate the
static error. n o
4-14. l.P«.4-l,ft - lt,(r.. + l),ft . N. .

Calculate the proportional sensitivity for a damping ratio of one third. Calcu-
late the period of oscillation.
4-15. In Fig. 4-1, Gy * G, - — T7
n
A.; . . N
Tii
1

+ l

Calculate the period of oscillation and the offset.

 ••haptor

Measuring
(Feedback) Elements

Measurement of the value of the controlled variable is necessary

in order to determine the magnitude of deviation. The performance of
the measuring means is therefore an important factor in the operation
of an automatic controller.
The characteristics of a measuring means, such as a thermometer,
pressure gage, or flowmeter, are important in three ways:

1. The principle of operation of the measuring means often determines

the method by which the automatic controller mechanism operates.
2. The characteristics of the measuring means affect the indication
of the value of the controlled variable.
3. The dynamic characteristics of the measuring means directly
influence the operation of the automatic controller.

The meap
will be reviewed from
principle of operation of measuring
the standpoint of automatic control; that no emphasis is placed upon
is,

the display (the indication or recording) provided by the measuring

means. For a broad study of the subject of measurement, the reader
should consult the i dated textbook, Indwtrud Instrummlalion}

Types of Measuring Means

The basic operation performed by the measuring means is to convert
or transduce the controlled variable into another variable thatmay be
employed for operation of the automatic controller. As shown in Fig.
5-1, the measuring means H
converts the controlled variable into the

^ D. P. Eckman, John Wiley and Scui, Inc., New York, 1960.

123
 . )

124 Automatic Process Control

feedback variable b in order that the controller may determine the

actuating signal e:

e = r — b

where r = reference input. The actuating signal may be different than

the deviation {v - c) because the feedback variable b may have a dif-
ferent value than the controlled variable c.

Gi Controiiing element

G2 Final control element

Gs Process
/f Measuring element

Fia. 5-1. The measuring element.

The feedback variable b may be a mechanical displacement, a force,

a pressure, or an electrical signtil

1. A displacement element converts the controlled variable into a

mechanical displacement. (Example: a mercury thermometer converts
temperature into a height of mercury column by thermal expansion.)
2. A Jorcc cl(?ment converts the controlled variable into a force or
torque. (Example: a displacement float-level gage converts liquid level
into a buoyant force on the float.) Pressure elements may be con-
sidered force elements when the pressuic is applied to the area of an
elastic clement such as a bellows or piston. Almost any force element
may produce a displacement by applying the force to a spring. A mutu-
ally (‘\clu.sive classifica^on of displacement, force, and pressure elements
could probably’ not be made
3. An elednc element converts the controlled variable into some form
of electrical (juantity such as voltage or current. (Example: a thermo-
couple converts temperature into a direct current voltage.

Whether a system is force balance, displacement balance, or electrical

balance depends upon its design as a pneumatic or hydraulic mechanism
or as an electrical mechanism.

Temperature Elements
Temperature ma>’ be nieasui’ed by mechanical, fluid or electrical
means.
 Measuring (Feedback) Elements 125

Expansion thermometers employ the thermal expansion of an elastic

member to produce a motion of an indicating pointer proportional to
the temperature. Expansion thermometers are displacement type be-
cause they convert temperature into a proportional displacement.

1. Mercury-in-glass themiometers are the simplest and most widely

used temperature-measuring element m both laboratory and industry.
The temperature range of the industrial mcrcury-in-glass thermometer
is about —38 to 950 F. The mercury thermometer is employed for
two-position control T)y utilizing an electrical contact at the surface of
the mercury.
2. Bimetallic thermometers are also widely used in industry The
bimetal is usually Invar for the low-expansion metal and either brass or
a nickel alloy for the high-expansion metal. The temperature range of
the industrial bimetallic thermonn'ter is about —40 to 800 F. The
bimetallic element is often employed for two-position control, as in
domestic and industrial thermostats.
3. Liquid-expansion thermometers utilize the thermal expansion of a
fluid to provide an indication of the temperature. A cylindrical bulb
filled with a liquid is subjected to the temperature to be measured. A
.small-diameter tube or capillary connects the bulb with the receiving
element in the controller. This receiving element consists of a bourdon
tube, a helix, a spiral, or a bellows. Expansion of the fluid inside the
system causes the free end of the bourdon tube or bellows to move, and
this motion is utilized to operate tho controlling means. The range of
the liquid-expansion thermometer is about —35 to 1000 F.

Pressure thermometers employ the principle that the pressure of a

gas or vapor is proportional to temperature. The pressure is then
measured by a bourdon tube or a bellows, and the force produced by the
system pressure is used to operate the (controlling means. Pressure
thermometers are therefore force elements. When deflection of the
bourdon tube or bellows is allowed, a displacement element results,
1. Gas-expansion thermo meters operate at essentially constant volume,
the gas expanding according to its temperature. The range of the gas
thermometer is about —200 to 800 F. Nitrogen is the most commonly
used gas.
2. Vapor-actuated LhcrmomeUis operate fiom the vapor pressure (jf

a liquid that partially fills the system. Since the vapor prcssui c (k'pends
solely upon the temperature at the free surface of the licjuid, th(^ vapor-
actuated thermometer indicates only the temperature existing at the
free surface. Vapor thermometers may be used for tcunppratures as
low as -50 F and as high as 600 F.
126 Automatic Process Control

A thermocouple is composed of two dissimilar wires welded together

at one end. When the temperature at the welded junction changes, an
electrical potential is generated and appears at the free ends of the wire.
This potential is the result of two distinct functions known as the
Thomson effect and the Peltier effect.
The Thomson effect is that portion of the total potential caused by
the temperature gradient over a single section of homogeneous wire.
The Peltier effect is that portion of the total potential produced by
the contact of two dissimilar wires. Both potentials vary with temper-
ature. Thus, the thermocouple is an electric element.
The five most commonly used industrial thermocouples are listed in
Table 5-1. Data on thermocouple emf versus temperature relationships
are given in the related book. Industrial Instrumentation.

Table 5-1. Characteristics of Thermocouples

Thermocouple (Positive Useful Range, Millivolts

Metal First) -F per °F*
Copper-Constantan -300 to 600 0.0330
Irou-Constantan 0 to 1300t 0.0350
Chromel-Alumel 600 to 18001 0.0215
Pt 10% Rh-platinum 1300 to 2900 0.00661
Pt 13% Rh-platinum 1300 to 2900 0.00754
* At maximum temperature,
listed
t Add 200 F in reducing atmosphere,
t Add 200 F in oxidizing atmosphere.

The thermocouple measuring junction is usually located some dis-

tance from the controller, and thermocouple lead wires are used to
extend the reference junction to the controller. There are two methods
of selecting lead wires: (1) the thermocouple wires themselves may be
extended to the controller if the thermocouple wire is not expensive,
and (2) the lead wires may be made of less expensive thermocouple
metals having the same thermoelectric properties as the thermocouple.
A thermal well, protects the thermocouple from corrosion and con-
tamination. The iron-Constantan thermocouple may have the iron
wire in the form of a closed-end tube with the insulated Constantan wire
running through the center and swaged into the end of the tube to form
the hot junction. Such a design reduces mass and speeds the response.
The resistance thermometer is T^idely used because of its recognized
accuracy and simplicity in industrial operations. The resistance
thermometer is an electric element and operates on the principle that
the electrical resistance of a wire changes with temperature. The indus-
trial resistance thermometer bulb consists essentially of a coil of fine wire
Measuring (Feedback) Elements 127

wound on or in a frame of insulating material, often mica or glass.

Platinum, nickel, and copper resistance bulbs are used, and some of
their characteristics are shown in Table 5-2.

Table 5-2. Characteristics of Resistance Thermometer Bulhe

Wire Material Bulb Resistance, ohms Useful Range,
Platinum 10-35 - 300 to 1200
Nickel 100-300 - 300 to 600
Copper 10 —40 to 260

Resistance thermometer bulbs are generally used with a thermal well,

except when temperature measurements are made in dry air.

The thermistor is another electric element employed as a temperature

measuring means and is similar to a resistance thermometer. Ther-
mistors are made from a mixture of metallic oxides and provide a specific
resistance versus temperature relation. The resistance may either
decrease or increase with increasing temperature.
The radiation element operates by measuring tlie thermal radiation
from a heated body. Radiation in a band from approximately 0.1 to 8.0
microns in the visible and infrared portion of the spectrum is directed
by a lens or mirror to the thermopile or vacuum thermocouple. A
thermopile a number of thermocouples connected in series, the meas-
is

uring junctions having been flattened and blackened to provide high

absorption of radiation. The reference junction is thermally connected
to the housing of the radiation unH. The radiation unit is therefore
an electric clement.
The radiation unit is directed (sighted) on the surface, the temper-
ature of which is being measured. The temperatures between 800 and
3200 F are best suited to the radiation unit, but temperatures as low
as 200 F may be measured.

Pressure Elements
The measurement of pressure or vacuum for purposes of control is
generally accomplished by applying the unknown pressure to a movable,
elastic member of constant area. The force thus created is balanced by

a spring for(*e or other balancing force and the resulting movement

operates the controlling means. Mechanical pressure elements may
therefore be either displacement- or force-type depending upon the
physical arrangement.

1. Bourdon tube, spiral, and helix elements are the most common
actuating elements for the pressure gage. These types of elements do
not require a spring since the elastic modulus of the metal is utilized to
128 Automatic Proceiss Control

graduate the deflection due to changes in pressure. Bourdon tube gages

are almost always employed for measuring static pressure, not dififer-
ential pressure.
2. The bellows pressure gage is used for the measurement of diflFeren-

tial pressure by applying the two unknown pressures to the two bellows.
The force of the bellows is balanced by a spring, and the resulting dis-
placement is proportional to the applied differential pressure. Absolute
pressure may be measui’ed by evacuating one bellows to a very low ab-
solute pressure, and applying the unknown absolute pressure to the
opposite bellows.
3. The hell pressure gage is useful for the measurement of low pies-
sures near atmospheric. Thesuspended on a beam in a sealing
bell is
liquid, and the unknown pressure i.s applied underneath the bell. "Jlic

force of the pressure on the bell is balanced either by weights or by a

spring. Differential pressures are measured by applying the second pi os-
Bure over the bell.

4. The manometer differential pressure gage operates from a metal

float inan enlarged leg of a mercury U-tube. The metal float follows
the changes in level of ^-he mercury in the chamber as the differential
pressure changes. The motion of the float is transmitted through a
-

pressure- tight shaft and is utilized for operating the controlling means.
5. In the ring-balance differential pressure gage the ring is balanced
on a rotating shaft or knife-edge bearings. The differential piessiire
displaces the sealing liquid in the ring and the rotation of the ring
results from the net torque due to the different pressures acting against
the clo.sed end.s of the ring.
(). The electrical pressure gage employs an unbonded or bonded strain
gage for indicating the pressure acting against the bellows area.

Liquid-Level Measurement
Measurement of li(iuid level is accomplished by two different methods.
Head measurement by a float following the liquid-level surface i)rovide.s
.‘iccurate measurement of liquid volume independent of liquid density

Pressure measurement by a pressure gage connected to the bottom of

a tank provides accurate measure of liquid weight independent of
liquid density.

1. The float and tape method is employed for open \eshels. Other
mechanisms employ a free float on a pi\'oted arm to operate an indicat-
ing means.
2. The diaphragm-box system is used for pressure measurement in
open vessels. The metal box contains a flexible diaphragm which acts
 Measuring (Feedback) Elements 129

to compress the gas in the pressure receiver line when the liquid level
increases. The pressure receiver is then calibrated in terms of liquid
head. The bubbler system and the air- trap system are similar means of
liquid-level inoasiireinent.
3. The (lisplacrmenl-floal metfr may be used in either open or closed
pressure vessels The buoyant force caused by the displaced liquid is

used to operate the indicating means.

Many electrical types of lifjuid-4evel control devices do not indicate

or record the level. ^An electrode is sometimes introduced to make
electric contact through the liquid and operate a relay for controlling
the level. Another type utilizes a column of the liquid to break a light
heam and actuate a photocell relay. Tn some devices a (‘olumn of liquid
serves as a portion of a variable electric* (capacitor to operate an elec-
tronic control circuit.

Fluid Flow Rate Measuring Elements

The control of fluid flow rate in closed pipes is accomplished generally

by one two means; the head flowmeter, or the area flowmeter.
of
The head flowmeter operates by measuring the pressure differential
across a suitable restriction to flow. 3'he most common restrictions are
the orifice, flow-nozzle, \'enturi, Pitot tube, and Venturi-Pitot Ring.
The pressure-differential is measured by any of the pressure-differential
meters (le*scribed in a previous seetion of this chapter. For further
details on the theory and ase of head flowmeters the reader is referred to
the textbook Indmiriat Instrumentation.

1. An orifice may be installed with flange taps or with pipe taps in

which the upstream tap is 2.5 pipe diameters and the downstream tap
is 8.0 pipe diameters from the orifice, or with vena-conf racta taps which
locate the upstream tap 1.0 pipe diameter from the orifice and the down-
stream tap at the vena coiitracta. Flange taps are the simplest and most
economical, pipe taps provide the smallest pressure differential, and
vena-con tracta taps provide the largest pressure differential.

2. The flow-nozzle is arranged with a well-rounded entrance section

so that the permanent pressure loss is small. It is therefore useful in
measuring large quantity flow rates of steam in power plants.
3. The Venturi tube has both a smooth entrance and exit so that the
pennanent pressure loss is very small. It may be used to measure the
flow of fluids containing solid particles but is employed principally for
measuring large floAv rates of liquids.
4. The Pitot tube operates by converting velocity head into static head.
The differential head is thus proportional to velocity at the streamline
130 Automatic Process Control
centering on the forward opening. The pitot tube requires calibration
ina particular application. It is most useful in measuring flow rate in
very large pipes and ducts.
6. The Venturi-Pitot ring operates on the principle of the pitot tube,

but two factors of design serve to increase the differential head for a
given flow rate. First, the venturi section increases velocity so that
velocity head is and second, the pitot has both forward and
increased,
rearward openings to increase differential head.

The pressure differential across the restriction's given by a form of

Bernoulli’s law for incompressible fluirls so that the flow rate varies as
the square-root of pressure differential. Thus a pressure-differential
meter used with an orifice is calibrated for flow rate with a parabolic
scale having wider graduations at high flow than at low flow. The flow
coefficients must be taken from tabulated data.
The area flowmeter operates on the principle that the variation of area
of flow-stream retjuired to produce a constant pressure-differential is
proportional to flow rate. The rotameter is one form of the area meter.

The fluid passes up the tube through the annular space around the float.

The float is supported in the fluid stream by the pressure differential,

buoyancy, and drag acting upward against the weight acting down-
ward. A scale etched on the glass metering tube is calibrated in flow
rate and is read by noting the position of the float against the scale.
The electromagnetic flowmeter element is an clement operating
electric
on the principle that the liquid flowing through a nonmagnetic tube in
a magnetic field induces a voltage that is proportional to flow rate and
field intensity. The field is supplied by an electromagnet. Electrodes
sealed in the tube detect a direct-current or alternating-current- voltage.
The flowing fluid should have some electric conductivity. The flow
rate is measured independently of viscosity, velocity disturbance, etc.
The by passing the fluid out one
Coriolis flowmeter element, operates
radial tube and back another radial tube. The force difference between
the two paths is caused by the Coriolis acceleration. 'J'he force is

measured by an elastic member, strain gages, and slip rings. The flow
rate is measured independently of viscosity, velocity disturbance, etc.

Pneumatic Transmission
It is often required to measure the value of a controlled variable
where the location of the point of measurement is at a considerable
distance from the location of the controller. Most measuring devices
such as a mercury thermometer, pressure gage, or flow-rate meter would
then require fluid-line connections of great length. This cannot be done
Measuring (Feedback) Elements 131

because excessive measuring lag would result. Therefore some trans-

mission or telemetering means must be employed. Pneumatic transmis-
sion is one method of transmitting the value of the controlled variable.
The pneumatic transmission system shown in Fig. 5-2 may be used

Fig. 5-2. Pneumatic transnussiou.

for distances up to many hundreds of feet. The controlled variable is

converted to an air pressure at the transmitter T. The air pressure is

then conducted through a single tube to the receiver R where it is

transduced to a position or force for operation of the controller.

The details of a pneumatic transmission system are illustrated in
Fig. 5-3 for a particular system transmitting a measured pressure c
of range 0 to 100 psi gage. The transmitter consists of a double-bellows

c, measured
pressure,

Receiver Transmitter

Fig. 5-3 A pneumatic transmisaion system.

system with the measured pressure in the smaller bellows balanced by

a transmitted pressure acting against the larger bellows. The trans-
mitter acts as follows; For a measured pressure c of 50 psi, the trans-
mitted pressure bo is 9 psi. When the measured pressure c increases
suddenly to 51 psi, the bellows assembly moves to the left and covers
the exhaust nozzle. This increases the transmitted pressure bo because
air flows into the system through the supply restriction. The trans-
mitted pressure bo increases to 9.12 psi and balances the ne^ measured
pressure c. Tt is apparent that the transmitted pressure is proportional
 132 Automatic Process Control
to the measured pressure. Usually an amplif3ring pilot is employed at
the transmitter in order to increase the air-flow capacity of the trans-
mitter. Other controlled variables such as temperature or flow rate
may be transmitted by suitably designed transmitters.
The connecting tube carries the transmitted pressure to the i-eceiver.
This tube is almost always one-quarter in. O.D. standard copper,
aluminum, or plastic tubing.
The receiver is simply a pressure-gage element calibrated for 3-15 psi
input pressure range.
Practically the only limitation of distance of pneumatic transmission
is the lag of transmitting the signal through a long connecting tube.
7'he capacitance of the tube is caused by the volume of the tube and it

100 200 400 600 800 1000

Distance, feel

Fig 5-4 Appn>xiination of iransrruasion lag

IS distnbuted along the length of the tube The resistance of the tube
IS due to fluid friction aiul is likewise distributed along the tube. Because
of these distributed paraini'lets, the calculation of the lag is extremely
difficult and it is necessary to te\y upon experimental tests to determine
the lag.^ The results of such a test, shown in Fig. 5-4, are given in
^ J C. Moise, A and Analytical InvesUyatum of Long Pneumatic TranS'
Theoretical
mission Iaucs, Doctoral thesis. Case Institute of Technology, Cleveland. Ohio, 1962.
Measuring (Feedback) Elements 133

terms of the approximate first-order time constaut T that most nearly

represents the test data.The dead time L due to the time for transmis-
sion of the pressure wave down the tube is also indicated in Fig. 5-4.
Thus pneumatic transmission may be employed up to 350 ft distance if
a time constant of about 1 second and a dead time of one-third second
are not important in automatic control, and up to about 1000 ft dis-
tance if a time constant of about 7 seconds and a dead time of one
second are not important.
Two methods are used to improve the speed of pneumatic transmis-
sion. First, as mentioned previously, an amplifying pilot shown in

Fig. 5-5 may be used.

Lag compensated
Temperature transmitter

Fig 5-5. Pneumatic transmission pilots.

The amplifying pilot provides a gain of one so that the output (trans-
mitted) pressure is the same as the input pressure. The pilot is simply
inserted in the transmission line. Its advantagi; derives from the large
ports of the pilot valve so that a large flow of air is allowed when a
change of pressure takes place. The operation is as follows: An increase
in input pressure acts against the diaphragm and pushes the pilot stem
to the left. This opens the supply port and allows air to flow into the
output volume. The increased output pressure acts to the right against
the second diaphragm and exactly balances the input pressure. On a
decrease in input pressure, air is exhausted from bet ween the diaphragms.
The lag-compensated transmitter provides a boosting action that
tends to overcome transmission lag. The operation is as follows: An
increase in temperature at the thermometer bulb increases the pressure
in the right inner bellows. This move.s the bellows assembly to the left,
134 Automatic Process Control

covers the exhaust nozzle, and increases the output pressure to a large
magnitude. Then, as air flows through the restriction to the outer
bellows chamber, the bellows moves gradually to the right and reduces
the output pressure to its new but larger value. The action is similar
to a derivative effect and is described more precisely by

^ “
"b 1
(5-1)
c Tlj. + 1

where Tl = time constant lag of transmitter

6o = transmitted pressure
Td = compensating time constant
c = controlled variable

If the upper restriction were not present in Fig. 5-5, the compensation
Td would be zero. Thus, the right-hand (derivative) term of equa-
tion 5-1 indicates a pure derivative effect in the presence of a time-
constant lag Tl. The net effect is that of reducing the time-constant
lag of the system. The dead-time lag is also partially compensated.

Electric Transmission

Electric transmission or telemetering systems operate by transducing

the controlled variable into an electric signal, usually voltage or a
current. Electric systems have a great advantage of nearly unlimited

Open loop

Closed transmission loop

Fig. 5~6. Transmission methods.

Measuring (Feedback) Elements 135

distances of transmission either by wire-carried signal or by radio

linkage. Electric transmission systems employed in automatic control
are of several types as shown in Fig. 5-6. The transducer is indicated
at T, The open-loop system employs a transducer such as a thermo-
couple in which the feedback signal 6 is a voltage. The series closed-loop
system is employed in three forms; closed-loop amplifiers at both
transmitting and receiving end, or at either end only. The cloeed-
transmission-loop system is sometimes advantageous for particular
•
transducers.

0 to 0.5 V, a-c
c, displacement

=/= Open-loop transmission

(T^ Swartwout Company)
i

Closed-loop transmitter
(Manning, Maxwell, & Moore)

Fjg 5-7. Electrical transmitters

The open-loop transmitter employing a differential transformer is

shown in Fig. 5-7. The differential transformer consists of two to four
coils with a movable core centered in the coils. The primary coils are
connected to a source of alternating voltage. The voltage induced in
the secondary coils will depend upon the position of the core. The
transmission signal is a 60-cycle voltage of 0 to 0.50 volts. The advan-
tage of this system is that the arrangement is simple and the speed of
response is limited only by the transducer itself. The disadvantage,
as with any open-loop system, is that careful ambient temperature and
pressure compensation may be required and supply voltages must be
constant.
The closed-loop transmitter in Fig. 5-7 employs a null-balance stage
in the transmitter. The transducer provides a force C acting on the
balance arm. The arm is detected by the oscil-
position of the balance
136 Automatic Process Control
lator coil and amplifier A is energized. The direct-current output of
the amplifier is fed back to the voice coil-magnet drive on the balance

arm. Thus the output direct current of 0.5 to 5.0 milliamperes is pro-
portional to input force. The advantages of this system are that the
direct-current transmission signal is not easily interfered with and tlie
signal may be manipulated easily, giving great flexibility.

b, displacement
c, displacement

(Automatic Temperature ConUol Company)

Fig 5-H Elerlncal transmitters.

Closed-transinission-loop systems are shown in Fig. 5-8 The self-

balancing bridge transmitter employs the Wheatstone bridge. A
change
jn temperature c at the resistance element (or thermistor) causes an
unbalance of the bridge which is detected and amplilied to drive the
reversible motor. The motor rebalances the bridge The output vari-
able b is the displacement of the output arm, which may be used to
operate a control means.
The sell-balancing differential transformer system of Fig. 5-8 employs
an input transformer and a feedback transformer interconnected so that
the differential voltage of the secondaries is proportional to relative
unbalance. The two diffcTcntial tran.sformers must be operated from
 Measuring (Feedback) Elements 137

the same line source so that the primaries are in phase. The closed-
transmission-loop systems have the advantage of flexibility.
The impulse-duration transmission system is shown in Fig. 5-9. The
transmitter consists of a synchronous motor driving an electric contact
through a cam in such a way that the duration of make of the contact is
proportional to the variable being measured. The receiver is a clutch

b, displacement

e. pressure

Fig 5-9 ImpulHe-duiiition transiriiHsion

controlled synchronous motor driving in such a way that, from a given

starting point of drive, the impulse signal actuates a magnet clutch to
drive an output arm in one direction. When the signal ends the oppas-
ing clutch is actuated to drive the output arm in the opposite direction.
Thus the output-arm position b is determined once each cycle. The
advantage of this two-wire system is that no synchronization i.s required,
and the signal requires only twm wires. The speed is generally slow,
with a cycle of from 5 to 60 seconds.
There are many other forms of electrics transmission such us synchro
systems, and potentiometer systems. These are too nuineious to
review here.

Firsl-Order Response
Speed primary measuring element and the measur-
of response of the
ing means of a controller the most important single factor affecting
is

the operation ol an automatic controller. Since automatic control is

a continuous, dynamic function,* the rate of detection and the time
element in response of the measuring means form an essential part of
automatic control analysis.
If we wwe suddenly to immerse the bare bulb of an expansion ther-
mometer an agitated salt bath maintained at a constant temperature,
in
the thermometer pen would rise as indicated by the curve of Fig. 5-10.
136 Automatic Process Control
lator coil and amplifier A is energized. The direct-current output of
the amplifier is fed back to the voice coil-magnet drive on the balance

arm. Thus the output direct current of 0.5 to 5.0 milliamperes is pro-
portional to input force. The advantages of this system are that the
direct-current transmission signal is not easily interfered with and tlie

signal may be manipulated easily, giving great flexibility.

b, displacement

lb. displacement
c, displacement

Self-balanring differential transformer system

(Automatic Temperature Control Company)

Fig 5~H Electrical traiismitterf,

Closed-transmission-loop systems are shown in Fig. 5 8 The self-

employs the Wheatstone bridge A change
balnnc.iug bridge transmitter
in temperature c at the resistance element (or thermistor) cau.ses an

unbalance of the bridge which is detected and amiilified to drive the

reversible motor. The motor rebalances the bridge. I’ho output vari-
able b is the displacement of the output arm, which may be used to
operate a control means.
The self-balancing dilTorential transformer system of Fig. 5-8 employs
an input trnnsfornier and a feedback transformer interconnected so that
the differential voltage of the secondaries is proportional to relative
unbalance. The two differential transformers must be operated from
 Measuring (Feedback) Elements 137
the same line source so that the primaries are in phase. The closed-
transmission-loop systems have the advantage of flexibility.
The impulse-duration transmission system is shown in Fig. 5-9. The
ti*ansmitter consists of a synchronous motor driving an electric contact
through a cam in such a way that the duration of make of the contact is
proportional to the variable being measured. The receiver is a clutch

Synchronous
motor
Fig. 5--9. InipulHe-dumtion tniusmiHsion.

controlled synchronous motor driving in such a way that,from a given

starting point of drive, the impulse signal actuates a magnet clutch to
drive an output arm in one direction. When the signal ends the oppos-
ing clutch is actuated to drive the output arm in the opposite direction.
Thus the output-arm position b is determined once each cycle. The
advantage of this two-wire system is that no synchronization is re(4 uired,

and the signal requires only two wires. The speed is generally slow,
with a cycle of from 5 to GO seconds.
There are many other forms of electric transmission such as synchro
systems, and potentiometer systems. These are too numerous to
review here.

First-Order Response
Speed primary measuring element and the measur-
of response of the
ing means most important single factor affecting
of a controller is the
the operation of an automatic controller. Since automatic control is
a continuous, d 3 'iiainic function,* the rate of detection and the time
element in response of the measuring means form an essential part of
automatic control analysis.
If we were suddenly to immerse the bare bulb of an expansion ther-
mometer in an agitated salt bath maintamed at a constant /temperature,
the thermometer pen would rise as indicated by the curve of Fig. 5-10.
138 Automatic Process Control

llie curve appears to be exponential, and it approaches the bath temper-

ature gradually. This curve shows that if it requires 12 seconds for the
thermometer to indicate 90 per cent of the change, it takes another

Fig. 5-10. Respoiise of thermometer element.

12 seconds to reach 99 per cent, and another 12 seconds to reach 99.9 per
cent of the final value. The difiBculty of expressing lag in terms of time
for indicating a complete change is that theoretically the final
temper-
ature will never be reached.
Measuring (Feedback) Elements 139

The response curve for a ^ple thermal element is derived from the
laws of conservation of energy and heat transfer. As indicated in
Fig. 1^10, the increase in energy is equal to the inflow of energy, or

(WP)h - q (5-2)

where W = weight of element, lb

P= specific heat of element = Btu Ib""^ deg”‘
6 = temperature of element, deg
q = heat flow\o element, Btu/sec

For convection transfer of heat

q = HA{c-h) (5-3)

where H= convection coefficient of heat transfer, Btu/ft^ deg“^ sec”*

A = surface area of element, ft^
c = temperature of surrounding fluid, deg

Combining equations 5-2 and 5-3 to eliminate heat flow 9 a ,

first order
differential equation results:

Tb +b= c (5-4)

where T= time constant =* WP/HA. The operational form of the

equation found by setting s — d/dt. Solving for the relation between
is

output variable b and input variable c, the system function is obtained

b 1
__
~ (5-6)
c Ts -h 1

The operational diagram for the thermometer is shown in Fig. 5-11.

The response to a sudden (step) change of temperature is found by
solving equation 5-4 for particular conditions:

= 1 - (5-6)
c - bo/

w’herc 5o = initial temperature of element. This response is plotted in

Fig. 5-11. The element time constant T may be evaluated experi-
mentally from the response curve. When

the value of the temperature ratio is

6-60
= (1 - 0.368) = 0.632
c — bo

Therefore, the time constant of a simple thermal element is the time

 140 Automatic Process Control

required to reach 63.2 per cent of the final value. Note also that the
initial slope intersects the final value at a time T.
The general approach to the response of a measuring means is indi-

cated by the block diagram in Fig. 5-11. The controlled or measured

variable c is the input to some measuring device, and the output is the

0 12 Time, tfT
3 4

Fio. 5-11. First-order response.

indicated variable b. The indicated variable h may or may not actually

be indicated on a calibrated scale.
The time response of many simple physical systems in engineering
may be characterized by the general system function

K
(^7 )
1

where T= time constant

K = system steady-state gain

Systems exhibiting this system function are termed time constant

elements.

Example 5-1. A test of a system shows a response similar to that in

Fig. 5-11 which indicates a first-order response, and 7.0 sec is required to
indicate one third of the total change. What is the element time constant?
 K

Measuring (Feedback) Elements 141

TMm for show

1 - e— * 0.333 fori = 0.406
Then
t
0.406 or T *
f

Thermal-Element Lag
Temperature-measuring elements such as the expansion thermometer,
thermocouple, and resistance thermometer possess appreciable measur-
and
ing lag because of the teat capacity of the materials of construction
the heat transfer characteristics ot the surrounding media. Heat is
transferred to the thermal element by conduction, convection, and
radiation but the conduction heat transfer is usually small compared
to the convection and radiation transfer of heat.
The bare thermocouple and bare thermometer bulb installed in
rapidly moving fluid have almost first-order response as may be seen
by the actual test curves of Fig. 5-12. The element time constant can
be calculated if all the material and fluid characteristics were known.
From equation 5-4, the time constant is

WP
T
HA
Therefore a small time constant requires small weight but large surface
area, low specific heat, and high heat transfer coefficient.
The heat transfer coefficient can be calculated for forced convection
transfer of heat from a fluid flowing at right angles to a cylindrical
thermal element from
Nu = n(Nj,f\Nrr
where Nu = Nusselt number = HD/
H— convection coefficient, Btu/ft^ F~^ sec~^
D= diameter of tube or rod, ft
K - thermal conductivity of fluid, Btu/ft F~^ sec
m = a constant given below
n = a constant given below
Np = Prandtl number = gCn/K
C = specific heat of fluid, Btu/lb deg""*
=
fi absolute viscosity, lb sec/ft^ = vy/g
v = kinematic viscosity, ftVsec
7 *= fluid density Ib/ft^
Nr = Reynolds number * VD/v
V - velocity of fluid, ft/sec
142 Automatic IVooeas Control
The oonstante of the equation depoid upon R^olda numbw as follows

Nr n m
4(M000 0.68 0.47
4000-40,000 0.19 0.62
40,000-400,000 0.26 0,80

Time, sec

Fic. 5-12. Response of thermal elements. (Not necessarily typical of each type.)

* E. R. G. Eckert, Introduction to The Transfer of Heat and Mass, McGraw-Hill

Rook Company, New York, 1950
Measuring (Feedback) Elements 143

The velocity ol fluid flow V is a most important factor in response.

Equation &-9 shows that the thermal-element time constant should vary

Mean temperature, deg R

Fio 5-13. Effect of fluid velocity and mean temperature on thermal-element

time constant.

inversely as the 0.6 power of velocity. The data of Fig. 5-13 show
reasonable agreement.
The characteristics of the fluid surrounding the element are also
 Automatic Process Control
important. Equation 5 -9 indicates

so that high thermal conductivity, high specific heat, high density, and
low viscosity are necessary for fast response.

Example 5-2. A mercury thermometer bulb is i-in. long by ^in. diameter.

The glass envelope is very thin. Estimate the time constant in water flowing
at 10 ft per sec at a temperature of 140 F.

A"p (water) = = 3.02

A

V o.'i X 10"' X 12

Nu = 0.19(3)» »(20,000)» ‘* = 122

„ K .. 0.38 X 122 X 12 , „
3600 X 0.125

WP ^
ir P^Jjym P -- = 0.073
A iirDL

^ WP 0.073
^ M = IT = ^

For radicUion transfer of heat to a thermal element from a black-body

source, the heat flow g in Btu per sec is given by:

where K == radiation constant = 47.4 X 10“*^ Btu ftVsec deg“*

E “ einissivity of thermal-element material
A ^ surface area of thermal element, ft^
f = temperature of source, deg 11

b = temperature of receiving element, deg R

This is a nonlinear equation and may be replaced by

q = HrA{c-b) (5-H)
if the coefficient of radiation transfer Hr is defined by

„ 1 dq 4KEfi

where 0 is the average of the source and element temperatures in

deg R/100. This approximation is in error by only 10 per cent for a
source temperature twice the element temperature.
Measuring (Feedback) Elements 145

The time constant for a thermal element responding to radiation

transfer of heat depends greatly upon the average operating temper-
ature $. As equation 5-12 indicates, the time constant should vary
inversely as the cube of temperature. The data in Fig. 5-13 illustrates
this effect. The emissivity of the thermal-element material should be
as large as possible and thermal response is often improved by roughen-
ing and blackening the surface of the thermal element. In some cases
it is necessary to sacrifice response by using a thermal element of low
emissivity in order to reduce the radiation error.

Example 5-3. A thermal well made of standard J-inch iron pipe is used
furnace (reducing atmosphere) at 1400 F. (O.D.
in a = 0.84 in., 0.85 lb per ft).

The well is 2 ft long. Estimate its time constant.

The element probably receives most heat by radiation. Assume emissivity
IS 0.8.

//r - 4 X 48 X 10- U.Ui

100

1 70 X 0.1 1

r
X 12 ^ 42
0.0 1 X X X 0.84 X 2

Prcssurc-Eli'iucul

Fluid capacitance and resistance are the sources of m(‘asuriiig lag in

pressure elements when long fluid-filled lines are necessary to connect

Displacement element
measurlnp, liquTd pressure

Fig 5-14. Pressure elementF.

a pressure gage to the point of measurement. Inertia of moving fluids

and moving pressure-element parts also cause measuring lag, but this
will be neglected. The volume of connecting lines is assumed small
compared to pressure- element volume.
The bellows pressure gage of Fig. 5-14 is a deflectional system in
which the displacement of the bellows and spring is proportional to the
 146 Automatic Process Control
pressure in the bellows. For a change of pressure e, liquid must flow
from the vessel to the element. Employing the continuity relation, then
(Ay)b = w (5-13)

where A = bellows effective area, ft*

7 - fluid density, Ib/ft*
b = bellov/s displacement, ft
w= fluid flow rate, Ib/sec

The displacement of the spring is proportional to fjressure:

pA - Kb (S-14)

where p = bellows pressure, Ib/ft*

K= spring gradient, Ib/ft

The flow of liquid w into the bellows through the long connecting line
depends upon the pressure difference and resistance of the line and is
given by

W= (C - p) (5-15)
I
where R= resistance of line, sec/ft*
c = pressure to be measured, Ib/ft*

Combining these equations to eliminate pressure p, the differential

equation relating output deflection to input pressure is

Tb ^c
+b=K (5-16)

where T = element time constant.

For incompressible fluids operating into an elastic element the time

constant is

T-K
(y )
The capacitance of the system is given by the parameters enclosed by
the parenthesis. Naturally small capacitance results in small element
time constant and requires a small bellows area and large spring gradient.
Therefore, small displacement leads to small element time constant.
The resistance R of the connecting line should also be small. Line
resistance is discussed later in this section.
The diaphragm pressure gage of Fig. 5-14 is considered to be a null
system which a balancing force 6, supplied by some external mech-
in
anism, maintains the diaphragm in a fixed position. Then for a fixed*
 Measuring (Feedback) Elements 147
volume pressure chamber the rate of change of pressure ie given by

Cf^w (5-18)

where C = capacitance = V/{nRTi), ft*

V= volume,
n = polytropic constant
R = g£is constant, ft/dog
T\ ~ temperature, deg R
w ~ flow rate, ^|)/see

The balancing force b is assumed to balance the chamber pressure;

6 = pA (5-19)
where h — balancing force, lb
p = chamber pressure, Ib/ft*
A = diaphragm ofToctive area, ft*
The flow of gas through the connecting line is given by equation 5-16,
Combining these equations, the differential equation is

RCh +h= Ac (5-20)

For compressible fluids operating into elements of more or less fixed

volume, the time constant is therefore

T= SC^e( L-) (fi-21)

Therefore small volume and small line rcosistance are conducive to fast
response.
The resistance of a long connecting line having bends, fittings, and
valves depends upon whether laminar or turbulent flow exists. Actu-
ally under transient conditions the flow increases from zero to some
high velocity, usually turbulent, then decreases to zero if there are no
leaks in the line. Thus the calculation of line resistance is made very
difficult and for approximation purposes it is usually assumed that
laminar flow exists. Acjcording to the Poiseuille Jlagen law, the
laminar resistance is

dp __
128 yL _ 128 /iL
(5-22)
dw irD^y

where R — resistance, see/ft*

V — kinematic viscosity-ftV«^*c
L — line length, ft
D“ line inside diameter, ft

g = absolute viscosity-lb sec/ft*

y = fluid density, Ib/ft*
 2

148 Automatic Process Control

The laminar resistance is usually less than the actual resistance by as
much as 10 to 100 per cent. The best method of determining resistance
is by actual test of the element time constant. The capacitance may be
calculated with reasonable accuracy.
Summarizing, the pressure-element time c‘onstant may be calculated
from
128 mL/A^
(liquid) (5-23)
,D* \k)
and r

(Raa) (5-24)
tD* \npj
where To volume of element at operating pressure, ft^
n polytropic constant (n = 1 = isothermal)
Pa average absolute pressure.
A element area, ft^

K element spring rate, Ib/ft

absolute viscosity of ga.s, lb sec/f
L length of line, ft

D inside diameter of line, ft

Example 5-4. One

pressure element (null type) has no elasticity but
1.0-cu in, volume. Another pressure element (dcflectional type) has an area
of 2.5 sq in. and a spring gradient of 100 lb per in. but negligible end volume.
Which one is likely to have the smallest lag when measuring an air pressure
of 10 lb per aq in. at normal temperature? Compare the capacitance of
these elements.
For null type,

Yl 1 X 0.123
=
C =• 2.0 X 10“» ft^
nPa Tra XIX 24.7 X 144

For dcflectional tvpc,

A'l (2.5)^ X 0.123
=
C - 3.0 X 10-^ ft
K (144)2 X 100 X 12

Example 5-5. Calculate the laminar resistance of 100 ft of

tubing for air and water at 68 F.
For water,
VISvL _
~ 128 X 1.1 X I0-‘ X 1 00 X 12^
^ “
girD* 32.2 X 3.14 X (0.187)^

R = 23,000 8ec7'ft2

For air.

^ 12M ^ 128 X 16.8 X 10~»^ X 100 X 12^

'girD* 32.2 X 3.14 X (0.187)^

R = 370,000 sec /ft2

 )

Measuring (Feedback) Elements 149

Second-Order Response
The response of many temperature and pressure elements cannot be
adequately described by a first-order differential equation. Pressure
elements, for example, may have inertia effects in addition to fluid
effects. Temperature elements often possess more than one energy
storage or capacitance. Therefore, a second-order differential equation
must be used to describe adequately many pressure- and temperature-
measuring elements. ^
The response manometers or U-tubes is sf^cond order because
of fluid
inertia of themanometer fluid is involved. Applying Newton’s second
law to the manometer of Fig. 5-15,

-{RA“y)b — (Ay)h -}- cA (5-25)

or

© b+ (RA)b-¥’b= -c
7
(5-26)

where L = length through manometer liquid, ft

- displacement difference, ft
b
R - fluid resistance of manomet(‘r tube, sec/ft^
A = area of manometer tube, ft^
c = measured pressure, Ib/ft^

7 = manometer fluid density, Ib/ft^

This second-order differential e(|uation describes the response of the

manometer. Solutions to this equation are expressed in simpler form by
combining coefficients of the equation in a certain manner to obtain
combinations having particular significance. Equation 5-26 may be
written
^
^
(5 27)
7 7’V 4- 2r7\v -f- 1

where T ~ dial acteristic time inverse is the natural

V^L/fl
u ! idai n ped f req uei icy

_RA jg
~ damping ratio =
L

2 \ (

These quantities are characteristics of the system. When the damping

ratio is less than one, (f < 1), the system is underdamped as shown
by the response to a step change in measured pressure in Fig. 6-15.
When the damping ratio is one (f = 1), the system is critically damped,
150 Automatic ProoeBS Control

and when the damping ratio is greater than one (f > 1 ), the ^jratem
18 overdamped.

Fio. 5- 15 . Step response of a typical second-order system.

For best response a manometer should have a short length so that

the characteristic time T is small. However, the length of a manometer
is specified by the pressure to be measured and the density of the ma-

nometer fluid. Consequently, fast response is obtained by selecting

a dense manometer fluid. The damping ratio should be approximately
 1

Measuring (Feedback) Elements 151

0.6 to 0.8 (see Fig. 5~15) so that fast response is obtained without
excessive oscillation.
Many dynamic systems in engineering are described by a second-
order linear differential equation with constant coefficients. Such equa-
tions can always be expressed in the following form where the coefficient
of the zero order term is one:

Ab + Bh-\-b^ Kc (1^28)

When the system is imderdamped, equation 5-28 may be written

c
“ rV + 2rr« +
where T= VZ = characteristic time

f == = damping ratio
When the system is overdamped, equation 6-28 may be written

b
^
K (5-80)
C iTi8+lKT2S+l)

where Ti, T2 = - [1 ± Vl - 4A/^] are the two time constants of

2
the system. For u critically damped system the two time constants are
equal. The.se system functions are indicated in Fig. 5-16.

Underdamped Criticaiiy damped Overdamped

Fig. 5-16. System fuDctione for second-order systeme.

Example 5-6. For the bellows gage of Fig 5-J4 with a mass of the M
spring and bellows, derive an expression for the characteristic time and damp-
ing ratio.
When measuring liquid pressure, Newton^s second law yields

MS pA - Kb
Also, the motion of the bellows depends upon the flow rate in the line:

(Ay)S = w = ~(c p)

Combining,
 152 Automatic Process Control

Therefore T •

ly/UK
The natural frequency ifi

^ ^ /I.
2tT 2fr\M
Example 5^7. A pressure of 15 psi maximum is to be measured with a
mercury U-tube. What is tlic smallest characteristic time?

m /l5 X 144

The jesponse of many thermal elements are second order when a

thermal well surrounds the element. Tn a number of applications it is
not possible to expose the bare thermocouple, resistance thermometer
bulb, or expansion thermometer bulb to the fluid whose temperature is
being measured. The thermal well provides physical protection as well
as protection from corrosion and contamination.

Fin 5-17. Sccoml-order responHe of thermocouple and well.

The cross section of a thermocouple and well in Fig. 5-17 indicates

the temperatures and flow of heat. The e( 4 uations for the system are
(neglecting thermal conductivities of materials)

C262 = ^2 — (5-31)

9* 1
t\2
(c-w (5-82)

^*
1^1 r» (^2 “ ^>l) (M3)
Ki
Measuring (Feedback) Elements 153

where Ct * well capacitanoe » W2Cp2i Btu/dcg

Cl == thermocouple capacitance = WiCpi, Btu/deg
W = weight, lb
Cp = specific heat, Btu/deg lb”"*

Rx = heat-transfer resistance, well to T/c = l/HxA\

= he^t-transfer resistance, fluid to well = I/H2A2
H = heat-transfer coefficient, Btu/ft" sec”' deg”'
A = area, ft^
bi = temperature at thermocouple, deg
62 = temperature of well, deg
f = measured temperature, deg

due to conduction is neglected

In this analysis the heat transfer resistance
because it is compared to fluid film (convection) resistance.
usually small
Eliminating heat flows and temperature 62 from the above equations

h 1
(5-34)
(R2C2R^C\)s^ + + R2Ci + /?2C2).^ -f I

This system function can be factored into two time constants:

c (Tis+ 1)(7V-|- 1)
where
1
„ R\(^\ ~t~ R 2^ 1 ~i~ ^ 2 (^2 r. 4R2^2Ri^I
^
L —
Fx'tI" m,'c,R2C7'~
[ \ {RxC\ 4 R 2 CX H- R2C2y\

A close inspection of the damping ratio will show that it cannot be less

than one Therefore, the response of the system can be expressed by

two time constants, and T2.

Example 5-8, A thermal well has a time constant of 60 sec when installed
without a thermocoujile in it. The thermocouple shows a time constant of
3 3ec when suddenly inserted into the well. The weight of the thermocouple
!

junction is 1 oz and the weight of the thermal well is 0.5 lb. Assume the
specific heats to be the same. Calculate the time constants.

RiC} 4 RiCi
2
4RC =2 2 R\Ci
“T +
,
C'l E>
+
I
~
R\Ci 4RC 4RC _3
2 1 2 2 A CO j- —
'^2
2 “2 2X0.6
RiCi 4 R 2C 4 R 2 C 2
1
35.2
2

^R2C2RiCi 60 XJ
(R^Ci 4 ^22^1 4 i22Cj)* (35.2)2
154 Automatic Process Control

Ti, r, - 35.2fi ± V\ - (0.16)1

Th Ti « 68 Bcc 2.6 sec

,

PROBLEMS

5-1 . A thermometer having a time constant of 5 sec is placed in a bath the

temperature of wliich is changing linearly at a rate of 1 deg per sec. How
much does the thermometer read in error? ^
5-2. The time constant of a thermometer in flowing liquid is 10 sec. If
the fluid velocity is doubled, estimate the new time constant.
5-3. The time constant of a thermocouple in a furnace at 1 000 F is 2 sec.
What would the time constant be at 1800 F?
5-4. A bellows pressure gage (area 2 sq in., spring 200 lb per in.) is
connected by a ^inch I.D. tubing 50 ft long. Measuring water pressure,
what is its time constant?
5-5. A mercury U-tube has 2 ft total length of mercury in a J-in. I.D.
diameter tube. What is the period of oscillation and the damping ratio?
5-6. A thermal well and thermocouple each have a simple time constant.
For these elements in series show that there is little difference in response
by neglecting the smaller time constant if it is one tenth or less of the largei
number. Assume a linearly changing input and solve differential equations.
 rlutplvr

Controlling Elements

The automatic controller may operate through purely mechanical

means, but more often an auxiliary source offluid or electric power is
employed to actuate the control mechanism and to drive a final control
element. The controlling means is usually classified according to the
kind of power employed directly in operation: self-operated, pneumatic,
hydraulic, and electric. The combination of electrically operated
control elements and fluid-power-operated final control element is

particularly used.
The automatic controller consists of the measuring element H of
Fig. 6-1, the actuating signal element (the circle), the control element

G. O,

Gi Controlling element

G2 Final control element

G3 Process
H Measuring element

Fio. 6-1. The controlling means.

(?i, and the final control element (/ 2 . In self-operated controllers, these

elements are contained usually in one contiguous assembly. Power-
operated controllers, on the other hand, and particularly those involving
transmission of signals employ separate devices for the measuring ele-
ment H and the final element
155
156 Automatic Process (biitrol

Self-Operated Controllers
Many automatic controllers operate by employing power developed
by the measuring means. These controllers are usually simple and in-
expensive, and are widely used in both industrial and domestic service.
The pressure regulator or pressure-reducing valve of Fig. 6~2 may be
either a “floAv" type or a “dead-end” type. The only difference between
these two types is that the “dead-end” type incorporates a relief valve,

“Flow" type "Dead end" type

Fig, 6-2. Pressure regu!Htx)rf(

indicated by the opening through the diaphragm in Fig. 6-2. The set
point determined by the adjustment of the spring compression. The
is

diaphragm “measures” the outlet pressure, and the actuating signal is

the net force acting on the valve plug. The manipulated A'^ariable is the
flow rate past the valve plug. The operation is as follows: When the
outlet pressure is too low, the diaphragm moves downward because
downward spring force is greater than the upward force, due to pressure
acting against the area of the diaphragm. This increases the flow into
the outlet and raises the outlet pressure so that the upward pressure
force equals the downward spring force. Conversely, if the outlet pres-
sure is too high, the valve plug reduces the flow at the outlet.
The “dead-end” type legulator releases fluid from the outlet through
the relief vah e wlien the outlet pressure is too high. Therefore it may
be used only on fluids such as air that may be released to the atmos-
phere.
The “flow” type regulator can only reduce the flow to zero; it cannot
allow flow out of tlie system. *
Thus, it is useful for water and fuel-gas
pressure control.
A pressure regulator has a proportional-control action because the
flow rate is firoportional to the actuating signal. The proportional
sensitiA’ity is tlie change in flow rate per unit change of valve-plug dis-
Controlling Elements 157

placement and isgoverned by the spring gradient and the valve-flow

characteristic. The proportional sensitivity is usually not adjustable.
The float-level controller of Fig. 6-3 is widely used for control of liquid
level in open or closed vessels. For such applications the valve stem is
connected to an arm on the shaft, and the shaft runs through sealed
bearings to the interior of the tank. The the arm on
float is fastened to
the inside end of the shaft, and the valve must be operated by direct
mechanical power from the float. Consequently, the valve plug is almost
always designed for^iinimum fluid forces. The control action is, of
course, proportional, and the proportional sensitivity may sometimes be
made adjustable through adjustable linkage.

Fig. 5-3. Direct-connected level Fio. 6-4. Thermostatic valve,

contndler.

A “radiator thermostat valve” iu constructed along similar lines (Fig.

6-4). Temperature measured by the thermometer bulb and the liquid
is

expansion operates the bellows and valve plug. Sometimes a powerful

bimetallic element is used in place of the liquid-expansion element. A
set-point mechanism, omitted from the figure, usually consists of a
threaded portion on the bellows container so arranged as to raise or
lower the bellows position. The operation is as follows: When the tem-
perature increases, the liquid expands, and the bellows forces the valve
plug downward. This reduces the flow of steam or hot water into the
controlled space. The control action is proportional and the proportional
sensitivity is governed by the size of the thermal element and the flow
characteristic of the valve.
These are but a few examples of the hundreds of self-operated con-
trollers employed in industry for pressure, temperature, and liquid-level
control. Nearly all self-operated controllers employ proportional-
control action in which the value of manipulated variable is proportional
to actuating signal. Generally, the proportional sensitivity is not
adjustable.
168 Automatic Process Control

Pkieumadc Proportional Contn^erB (Displaoement Type)

Pneumatically operated controllers for industrial process control are
simple and easily maintained and provide high-power amplification
using compressed air as a source of auxiliary power.
The pneumatic controller consists of two parts: a controller mechanism
incorporating the control action and providing an air-pressure output and
a final control element which positions a control valve in accordance with
the air-pressure output of the controller. Conseqiiently, the output air
pressure of the controller will be considered proportional to the manip-
ulated variable. Final control elements are considered in Chapter 7.

The displacement-type pneumatic controller operates by converting

the controlled variable and set point into corresponding displacements
and manipulating these displacements in order to produce a pressure
output. The (uUuating signal means in Fig. 6-5 consists of a differential
link, one end of which is positioned mechanically by the measuring means
according to the magnitude of the controlled variable c, and the other
end is positioned to correspond to the set point v. The difference is the
actuating signal and is the displacement of the center of the link.

A proportional controller of very simple type is the nozzle-flapper

arrangement of Fig. 6-6. Air at about 20 psig is supplied through a
small orifice or restriction (about 0.010 in. I.D.) to the oj>en nozzle. The
nozzle orifice about 0.015 iq. I.D, A flapper is positioned against
is

the nozzle opening in accordance with the actuating signal e so that the
nozzle back pressure mi is inversely proportional to the distance be-
tween nozzle opening and flapper. A
motion of the flapper of about
0.0015 in. is sufficient for nearly full This device may pro-
output range.
duce proportional control if the flapper is accurately positioned by a
Controlling Elements 159

measuring means. Over a reasonably linear range

mi = + Ml (6-1)

where mi = output pressure (nozzle back pressure)

Kc = proportional sensitivity
e = actuating signal
Ml = normal output pressure

The proportional sensitivity is the change in output pressure per unit of

actuating signal e and rftay be adjusted by moving the flapper pivot point.
The proportional controller of Fig. 6-7 employs a power-amplifying
pilot for providing a larger quantity of air flow than can be provided

Fiq. 6-7. Proportional controller with pilot.

through the small restriction of Fig. 6-6. The nozzle back-pressure,

instead of operating a final control element directly, is carried to the
bellows chamber and acts against the area of the bellows. Because the
bellows has some stiffness (spring gradient) the pilot valve plug is posi-
tioned between the inlet and outlet ports in accordance with the nozzle
back-pressure. A low nozzle back-pressure positions the valve plug to
the left, throttles the exhaust port, and causes a high output pressure
mi. A high nozzle back-pressure positions the valve plug to the right,
throttles the supply port, and causes a low output pressure mi. The
advantages of adding the pilot are: (1) the actuating signal e versus
output pressure mi relation may be made linear; and (2) the capacity for
air flow can be considerably increased.
A proportional controller incorporating a negative feeaoack circuit is

shown in Fig. 6-8. The system is similar to the mechanism of Fig. 6-7
except that the output nii is connected to a feedback bellows which
operates to repo.sition the flapper. The nozzle restriction is usually in-
corporated into the bcxly of the pilot and is omitted from the figure.

This type of controller incorporates a closed loop, whereas the controllers

discussed above are the open-loop type. The operation is as follows:
When the actuating signal increases, the flapper covers the nozzle, the
160 Automatic Process Control
nozzle back-pressure is increasedi and the output mi increases. (A
direct-acting pilot is assumed.) The increased output pressure acts
against the bellows and repositions the flapper to the left away from the
nozzle. If the operation is made stable by proper design, there results a

Fi<3. 6-8. Proportional oonlrolh;r with pilot and fredhack.

slight increase in output nii corresponding to the increase in actuating

signal c. Wlien th<‘ feedback link is in the upper position, the feedback
is small, the resultant oulput pressure change is large, and the propor-
tional sensitivity is large (proportional band small). When the feedback
link is in the lower po.sitioii, the proportional sensitivity is small (pro-
portional band large). The advantages of adding the negative feedback
are the same as for its use on process control: (1) I’he effect of variations
of supply pressu’'c, temperature, and output leakage is reduced to a
minimum (usually made negligible); (2) The proportional sensitivity or
proportional band may be varied over a wide range, usually as much as
1200 to 1 With this system displacement amplification up to 10,000
to 1 is easily obtained.

Proportional-derivative control is obtained by adding a feedback re-

striction as shown in Fig. G-9. This results in delayed negative feedback.

The operation cannot be described adequately by employing a sudden
(step) change in actual ing signal, so we will assume that the actuating
Controlling Elements IGl

signal changes linearly with time. The flapper then moves slowly to
the right, covers the nozzle, and causes a slow increase in output pressure
nil. There is a flow of air through the feedback restriction resulting
from the fact that the pressure in the feedback bellows is considerably
less than it would have been without the restriction in the line. This
delays and reduces the feedback, and, since the feedback is negative, the
output pressure is higher and leads instead of lags the actuating signal e.
Thus the delayed negative feedback produces derivative response. The
operational equation •describing the proportional-derivative action is

ideally

— = Kr(\ + Tas) ( 6-2 )

e

where = proportional sensitivity (inverse of proportional band)

Td = derivative time

The derivative time is the product of the resistance of the feedback

restriction and the capacitance of the feedback bellows. Therefore,
adjustment of the restriction 7? adjusts the derivative time Tj.
A proportional-integral controller is illustrated in Fig. 6~10. It differs
from the proportional controller of Fig. 6-8 only in the addition of the

positive feedback bellows and restriction.

5K3-c,alled The free ends of
two bellows are rigidly connected, and the flapper is attached to the
bellows rod by the adjustable link indicated in the figure. The operation
is as follows: When the actuating signal e suddenly increases (a step
change), the flapper covers the nozzle and increases the output pressure
mi The output pressure is transmitted to the negative feedback bellows,
.

repositions the flapper,and stabilizes the output pressure at the new

value. But there is now a pressure drop across the restriction because
the pressure in the positive feedback bellows remains at its former valqe.
162 Automatic Process Control
Thus there is a small flow of air into the positive feedback bellows, the
flapper is moved slowly toward the nozzle and the output pressure
steadily increases. The rise in output pressure continues until the supply
pressure limit is reached. The output pressure therefore has an
added portion representing the integral of the step change in actuating
signal. Ideally the operational equation for the system would be

(6-3)

where = integral time (inverse of reset rate)

Kc ~ proportional sensitivity (inverse of proportional band)

The proportional sensitivity is adjusted by ^electing a position of the

flapper connecting link. The integral time is adjusted by setting the
restriction R.
Proportional-integral-derivative by adding a
control is achieved
restriction in the negative feedback line in Fig. 6-10 as was done in
Fig. 6- 9. This combination provides an addition of the reset and deriva-
tive responses and the explanation of its operation proceeds as outlined
above for each response. If the restriction is added in the feedback line
at point 2 in Fig. (>-10, the response may he represented by

nil

e LJ’aS + (> + 0+ r..] (6-4;

where - time constant of negative feedback bellows and resistance

Ri (derivative time)
73 = time constant of positive feedback bellows and resistance R
(integral time)

This arrangement thus provides the proper control action as long as the
two resistances an* not set at the same value Such controllers are
arranged so that the derivative time T 2 cannot be set to a value equal
to or greater than the integral time
If the restriction is added in the feedback line at point 1, Fig. 0-10,
the response* may be represented by

(o-r>)
c

where T\ = time constant of negative feedback bellows and resistance*

(derivative time)
T3 = time constant of positive feedback bellows and resistance R
(integral time)
Controlling Elements 163

This arrangement provides the deared control action and avoids any
Loiitations of setting derivative time.

Pneumatic Proportional Controllers (Force Type)

The force-type pneumatic controller operates by converting the con
trolled variable and set point into corresponding forces and manipulating
these forces in order to produce a pressure output. The advantages of
this t3q>e of pneumatic controller is that greater flexibility is provided
for achieving various^kinds of control action. Industrial controllers of
this type are often termed ''stack controllers.’^
The actuating signal mechanism is shown in Fig. f>-l 1 The set-point .

means is a "dead-end” type pressure regulator which provides a pressure

Fio. 0-11. Actuating-signal means for a force-type pneumatic controller.

corresponding to the set point v. The measuring means consists of a

suitable pneumatic transmitter arranged to provide a pressure cor-
responding to the controlled variable. These two pressures are intro-
duced to the set of three diaphragms which subtracts the controlled
variable c from the set point v and produces a force corresponding to the
actuating signal.

Fig. 6-12. Force-type proportional controller.

The proportional controller of Fig. 6-12 employ." the actuating-signal

means of Fig. (i-ll as the left-hand set of three diaphragms' The right-
164 Automatic Process Control
hatid diaphraipn operates a pilot to convert the actuating-sisnal force
into a pressure output mi. Tlie restrictions Ri and R2 allow adjustment
of the proportional sensitivity Kc of the controller.
The operation is as follows: When the controlled variable exceeds the
set point, the actuating signal force acts to the rightand pushes the pilot
stem to open the supply port. The output pressure mi increases and
this pressure causes a force to the left in chamber mi (negative feedback),

thus balancing the actuating signal. Now, if restriction Ri is closed

there a fixed proportional sensitivity of one (proportional band 100
is

per cent). On the other hand, if restriction Ri were wide open, restric-
tion R 2 to atmosphere being fixed, the pressure in chamber / would
always be the same as that in chamber mi. In the latter case there is
no negative feedback and the proportional sensitivity is very high. For
settings of restrictionRi between full open and closed, the pressure in
chamber / is governed by the flow through the two orifices in series (f^i
and 7^ 2 ) so that positive feedback cancels a varying portion of the
negative feedback. Therefore, the restriction Ri adjusts the proportional
sensitivity of the controller.
Integral response can be added to a proportional controller by the
addition of delayed positive feedback as shown in Fig. 6-13. The pro-

Fig. 6-13. Proportional-integral controller (fixed band).

portional-band adjusting restrictions and chamber have been omitted for

simplicity so that this controller has a fixed proportional sensitivity of
one and an adjustable integral time. The operation is as follows:
Suppose that the controlled variable increases suddenly to a new steady
value. This causes a force to 'the right which is balanced by a higher
output pressure mi. There is now a pressure drop across the resistance
in the line leading to chamber p. Air flows into the chamber causing a
force to the right which further increases the output pressure mi. This
action continues with the output pressure gradually increasing and
integral action results. The mechanisms of Fig. 6-12 and G-13 can be
 )

Controlling Elements 165

combined to provide adjustable proportional sensitivity (restriction Ri

and adjustable integral time (restriction R^),
A derivative action can be added to the proportional controller by
providing delayed negative feedback as shown in Fig. 6-14. The pro-
portional-senBiti\'ity adjusting restrictions and chamber and the integral
restrictions and chamber have been omitted for simplicity, so that this
controller has a fixed proportional sensitivity of one and an adjustable
derivative time. The operation is as follows: When the controlled
variable increases gradually, a steadily increasing actuating signal
causes an increasing force to the right and an increasing output pressure

Fio. 6-14. Proportional-derivative controller (fixed hand).

nil. The negative feedback is negligible in chamber mj, because the

area of the diaphragm is small. The feedback is therefore delayed
through restriction into chamber d. The delayed negative feedback
causes a higher output pressure, and therefore the output pressure leads
the input pressure by a time equivalent to the time constant of the
resistance R4
and the capacitance of chamber d.
The mechanisms of Fig. 6-12, 6-13, and 6-14 are combined by adding
proper combinations of diaphragms and restrictions to produce propor-
tional, proportional-derivative, proportional-integral and proportional-
integral-derivative controllers. However, this combination may be
made in two ways. First, if the controller has both delayed negative
feedback and delayed positive feedback it is of a noncascade type, and
the ideal response may be represented by

[i + + r..]
1 (6^)

where /Cc = proportional sensitivity

1\
- integral time
Td = derivative time
160 Automatic Process Control

Second, if a proportional-derivative control unit output is connected as

the input to a proportional-integral control unit, it is a cascade type
and the ideal response may be represented by

Where the symbols are the same as above. This equation is easily de-
rived by combining equations 6-2 and 6-3 to eliminate the intermediate
\^ariable. This type of eontroller has certain advantuges with respect to
start-up of batch processes.

Modified proportional -integral control Modified proportional -derivative control

Fia. 6-15. Passive- type pneumatic control (direct and inverse derivative of the
Moore Products Co.)

The force-type control mechanism of Fig. 6-- 15 employs a somewhat

different principle of operation to provide a very useful combination
control action. With the device in the left-hand figure, a modified or
limited proportional-reset action is obtained as follows: an increase in
actuating-signal pressure e acts against the small diaphragm and causes
the pilot to increase slightly the output pressure nii. There is now a
pressure drop across the restriction R and avS air flows slowly into the
second chamber, the pressure there slowly increases. The output pres-
sure therefore slowly increases until it ecjuals the actuating-signal
pressure. The equation for this response can be shown to be

where m\ = output pressure

K= proportional sensitivity - An/A-^
a - gain ratio = A /A 2
1

T= time constant of restriction aiul chamber

e = actuating-signal pressure
 )

Controlling Elements 1(>7

When the gain ratio or is small (it is usually set at about one tenth
the effect of the control action is somewhat like a proportional-integral
action in which the time constant T adjusts the integral-like time con-
stant.
With the device on the right in Fig. 6-15, the first and second dia-
phragms are reversed, w'ith the result that the gain ratio a is also inverted
and a modified proportional-derivative action is obtained. The action
is described by etiuation (fi-S), but with the gain ratio a equal to approxi-

mately ten. Th(! time constant T tlien becomes a derivative-like

adjustment.

Air Supply for Pneumatic Systems'

The greate.st problem connected with pneumatic controllers is the
maintenjm(‘e of a clean, dry air supply at constant pressure. Moisture,
oil, corrosive liquids, or foreign particle.s carried into the pnciunatic*
system from the air supply will eventually cau.sc troubl(‘

150-200

Kio. &-lt) C'oinplotp :iir-siipply system lor largr iriHtHllali(»rj of pniMiiimtir

An air .supply arraugeiuent, as lu Fig. 6 Iti, may be used tor a largr

amount of (*ontrol etiuipment. An intake filter is located outside* tin*

compressor room through whicli the air passes to the conipressoi

A steam- or electric-driven compressor of desired capacity at a delivery
pressure between 150 and 200 lb per stj in. gage is common. Tbti air
passes into a storage tank, through a pressure regulator holding the
pr(‘.'>.‘'Ure to 100 lb per sq in., and then to an aftereooler. The atteruooler

*
UrproducecI by permission from Ptincipla vj Indnulttal Process CorUrol, John
Wiley and Sons, Inc., New York, 1049
 108 Automatic Process C.'oiitrol

or condenser requires a water supply. A porous stone filter removes

oil,and a dehydrator completes the removal of moisture. A second
pressure regulator is sometimes added to provide constant reduced
pressure in the supply lines leading directly to the control rooms for a
group of pneumatic control equipment.
In the majority of small installations an arrangement as shown in
Fig. G -17 is satisfactory, especially where excessively low or high temper-
atures are not encountered and where the intake air is relatively dry.
An intake filter relno^'es dust and solids from the intake air. A coin-

t'KJ. 6-17. Air-bupply .system for small installation of pneumatic control equipment

[)ressor with attached storage tank contains a pressure-actuated switch

to maintain relatively constant pressure in the tank. An aftercooler
and large-capacity filter removes excess oil, moisture, and solids.

Solids are removed most easily by a suitable intake filter at tlie

compressor. The intake filter should be located at a point where it may

gather the cleanest air possible; this point is generally outside the build-
ing housing the compressor. Edge-type, gauze, or waste filters may be
placed at various points in the system to separate solid particles or
foreign matter. Clean connecting lines of brass or copper are desirable
to eliminate rust.
Moisture is usually adequately removed with a storage tank of proper
size and a compressor aftercooler. Where moisture conditions are
severe, chemical or electric driers in addition to aftercooling may be
required to remove moisture more completely. Chemical driers con-
taining silica gel, activated alumina, calcium chloride, alcohol, glycol, or
other drying agents are common equipment. When the drying agent
requires replacement or regeneration, a two-section dehydrator should
be installed, only one section being used at a time while the other sec-
tion is being regenerated. If the drying agent is a liquid, an additional
unit may be needed for removing the drying agent carried over in the
Coutrolling Elements 169

dry air. Lines should be trapped and installed on a slope to drain

properly. inserted if there is a possibility that
Check valves should be
liquids or other gasesmight back into the lines from other sources.
Oil should be prevented from entering into the system. The air
compressor should never be overloaded, since it will pump more oil when
running at high loads. A water-sealed rotary compressor where no oil
comes in contact with the air is sometimes preferred. The air com-
pix'ssor should receive careful maintenance in order to obtain high effi-

ciency at all times, ^n aftercooler aids in removing oil vapor as well

a.s moisture. Filters of the porous stone, fuller’s earth, centrifugal, or

waste type effectively remove larger suspensions of oil. Solvent-type

filters may be used to remove oil, although it may then be necessary to

remove the solvent dissolved in the air.

An individualshould be placed at each controller to remove any
filter

remaining moisture, oil, or solids just before the air enters the con-
troller. The waste-type filter is most common and is quite satisfactory
under ordinaiy conditions. An individual small-capacity pressure
regulator should accompany each controller or group of controllers in
order to provide proper pressure regulation.
Moisture-free compressed air does not freeze even at subzero temper-
atures. Therefore, methods of effectively removing moisture also aid in
avoiding trouble due to freezing of air lines. Antifreeze injection sys-
tems arc not adequate, since the total amount of entrained fluid in the

air may be increased.

Locating the air lines so as to avoid low-tempcj ature areas as much
as possible is of considerable aid in eliminating either condensation or

freezing. It may be possible to maintain the temperature of a group of

air lines above freezing by means of a steam tracer. Burying the lines
in the ground below the frost line, with suitable traps to remove con-
densate may be effective.
sometimes possible to substitute other compressed gases than
It is
air for tlu* operation of pneumatic controllers where an adequate air
supply cannot be provided. Such gases must not bring about any cor-
rosive action or deterioration of the parts of the control system. For
strictly emergency short-time operation of a single control system
bottled nitrogen may be feasible.

Hydraulic Controllers
Hydraulically-operated controllers for industrial process control pro-
vide great power and positiveness of action since they employ a source
of high-pressure oil as auxiliary power. In certain instances high-
pressure water is used as a power source
170 Automatic Process Control

'I'hr hvdraulic controller usually consists of two operating units, a

f>ilof oil pressure and flow output and a power (\ylinder
vaK'C to control
and piston to provide the required displacement of a valve or other final
control element. The output displacement will be considered propor-
tional to the manipulated variable.
'Fhe operation of hydraulic controllers may he illustrated by Fig. 6-18.
The measuring means is the diaphragm which produces a force propor-
tional to the magnitude of the controlled variable
r. The set -point
means is the spring and knob by which a force proportioned to the value
of the set point is imposed. The actuating .signal means is the dif-

ferential link connecting the diaphragm and spring. Oil at high pressure
is supplied to the pilot valve. The operation is as follows: When the
actuating signal increases, the pilot valve stem is moved to the left

This supplies oil to the right side of the power (‘ylinder and exhausts oil
from tb(‘ left side. Consequently the power piston moves to the left.
Its rate of motion is proportional to the actuating signal. The result-
ing integral action is described by

m, =-J-e (6-9)
7 t’S

w here nii
~ position of power piston
T, ^ integral time
e actuating signal

The integral timt‘ is sometimes* made adjustable at the linkage betw^een

iiijiut arm and pilot stem
I’oucF* is supplied to hydraulic systems by means of an electric motor-
pump assembly as shown in Fig. (C19. The pump, usually a gear type,
draws oil from the sump through the filter The .•sump colle<as the
 —
Controlling Elements 171

return from the system and serves to dissipate heat. This latter
oil

function is very important in hydraulic systems, and it may infre-

quently be necessary to employ an oil cooler. The pump and motor
run continuously.
An accumulator sometimes connected to the pump outlet to serve
is

as a reservoir of high pressure oil. The accumulator contains a dia-

phragm with the gas side filled to several hundred pounds per s(|uare

Compressed

Fig. 6-19. Hydraulic power supply

inch of gas, usually nitrogen. With suitable check valves, the accumu-
lator may run the control system for several minutes in the event of
electrical failure. The main sources of difficulty in hydraulic systems
are dirt and Extreme care must be taken to insure that the sys-
heat.
tem is initially clean. After that, particles of metal and oil deposits
are bound to occur, and the system must be periodically cleaned. Heat
causes great variations in viscosity of the oil, and cooling must be
arranged. Usually a large sump may dissipate sufficient heat by
radiation.
Feedback link

(\
— DoooU
Tx
'

0 ooo
y q
From pump f | To sump Power cylinder-^

Fig, 6-liO. Hydraulic proiM^rtional i ruitnd

The hydraulic proportional controller in Fig. (>-20 illustrates the one-

tube hydraulic system. It is similar in many respects to a pneumatic
controller. The operation is as follows: When the actuating signal
increases, the pilot valve stem is moved to the left. This action par-
tially opens the (ionnecting tube to the sump, the oil pressure at the
power cylinder is reduced, and the spring pushes the power piston to the
left. The feedback link then moves to the right, tends to close the
 172 Automatic Process Control

pilot valve, and balances the power piston at its new position. Thus,
the position of the power piston mi is proportional to the actuating
signal e,

Ttii = KcC ( 6— 10 )

where Kg = proportional sensitivity

The proportional sensitivity may be adjusted by moving the connection

of the feedback link. The one-tube hydraulic controller is often em-
ployed for speed control of engines.

Fig, r>-21, llyilniulic pi-oportional-inR'gral conti'ol (Askania Regulator Company).

A hydraulic proportional-integral controller of a somewhat ditterent

type is shown in Fig. h-21. The diaphragm causes a force proportional
to the value of the controlled variable to act against the balance arm.
The set-point spring also causes a force on the balance arm, as does the
feedback spring. The motion of the balance arm displaces a free-
swinging nozzle Avhich directs a jet of oil at high pressure into a dis-
tributor block connected to both sides of the power cylinder. The
oiwration is With an increase in controlled variable, the jet
as follows:
]}ipe swings to the right and increases the pressure in the right side of
the distributor block. This increases the pressure on the left side of
the power cylinder and decreases the pressure on the left side of the
feedback cylinder. The power-cylinder piston moves to the right and
causes oil to flow into the right side of the feedback cylinder. The
feedback piston moves to the left, increases the force on the balance
arm, and returns the jet pipe to a balanced position. This is propor-
tional action. However, the pressure drop across the feedback piston
causes a flow of oil through the integral needle valve and allows the
feedback piston to move to the right. This action moves the jet pipe
Controlling Elements 173
to the right and causes continuing motion of the power piston. This is
the integral action. The combination is proportional-integral action:

where Kc = proportional sensitivity

Ti = integral time

The proportional sensitivity is adjusted by suitable levers between the

feedback spring and balance arm. The integral time is adjusted at the
needle valve on the feedback piston.
Many other types of hydraulic systems are in use, but the three
described here are the most common for process control. Integral
action, and any combination of proportional, integral, and derivative
actions may be obtained.

Electrical Proportional Controllers (Feedback Type)

Electrically operated controllers for industrial process control are veiy
and are particularly useful when long distances arc
flexible in operation
necessary between various portions of the closed loop. Electrical pro-
portional controllers fall naturally into two classes:

1. Those employing direct feedback transmission from the power

actuator.
2. Those providing a direct current or voltage output to which a

power actuator responds.

— Feedback

Propo.iional
fWvtwVi
sensibvity — 1|“

Detector- amplifier-relay Power motor

Fic. 6-22. Electrical proportional control (Leeds <fe Northrup Co.).

The d~c bridge 'proportional controller is illustrated in Fig. G-22. The

slider or contact arm displacement on the left slidewire of Fig. 6~22 is

proportional to the actuating signal. Usually this contact arm and

potentiometer are mounted in a self-balancing potentiometer or resist-
ance thermometer so that sufficient power is available to proyide accu-
rate positioning. The slidewire is located in a fixed bridge so that the
174 Automatic Process Control

d-c voltage to the amplifier is proportional to the actuating signal. The

amplifier is a combined detector, rectifier, power amplifier, and relay
which closes a forward drive circuit or reverse drive circuit to the
electricpower motor w'hen the actuating signal bridge voltage is not
balanced by the feedback voltage. The power motor also operates a
contact arm on an identical fixed bridge to provide a d-c feedback volt-
age acroas resistor R, When the feedback voltage equals the input
voltage the two contact arms are in correspondence so that; the motor
output arm displacement mi is directly proporticual to the actuating
signal e. The power motor may be used to position a throttling fluid
valve or other final control element. The control action is given by

m, = KrP (6-12)

where ?wi = position of power-motor output arm

Kc proportional sensitivity
e = actuating signal

Proportional sensitivity adjustment is provided by the variable resistor

in the actuating signal (left) bridge.
The proportional-derivative circuit is shown in Fig. 6-23. This cir-
cuit provides a delayed negative feedback and produces a response
proportional to the rate of change of the actuating signal. (Compare

Proportional -derivative circuit Proportional -integral circuit

Fic; (’>-2.'^ PY'iicJliack circuits for the electric controller of Fig. 6-22.

(he (*l('ctnciil system to the pneumatic vsystem of Figs. (Ml and 6-14.)
'flu* derivative time is proportional to the electrical time constant RC

and is .selected by adjusting resistor R.

The proportional-integral circuit is also shown in Fig. 6-23 and is

inserted at the points indicated in Fig. 6-22. This circuit provides an

“advanced feedback” and results an integral action added to the
in
proportional action. as follows: Upon a motion of the
The operation is

(contactor arm to the left in Fig. 6-22, away from the set point, a d-c
voltage is passed to the amplifier and drives the power motor in one
direction. The feedback voltage provided by the power-motor siidewire
Controlling Elements 175

causes a current to flow into the capacitor C, charging the capacitor and
resulting in a voltage drop across the resistor R, This is proportional
action. Now, however, in order to maintain a steady current through
must be continually charged. This can only
resistor R, the capacitor
be done by a slow and continuous motion of the power-motor arm to
the left. This is integral action. The integral time is proportional to
the electrical time constant RC and is selected by adjusting resistor R,
Proportional-integral derivative action is provided by inserting a suit-
able combination of the feedback circuits of Fig. 6-23 in the circuit
of Fig. 6-22.
The fl-c bridge proportional controller is shown in Fig. 6-24. As in the
previous electric controller the actuating-signal slide wire (at the left)

has a sliding contactor positioned according to the actuating signal,

usually by means of a self-balancing potentiometer or resistance ther-
mometer. The power-motor slidewire on the right has its contactor

Integral amplifier

Fia. 6-24. Electric projiortional-integral control (Minneayjolis-Honeywell

R<‘gulutor Co ).

positioned by the output arm of the power motor. The feedback bridge
provides integral action as will be explained later. The operation is as
follows: When an increase in actuating signal moves the slidewire con-
tactor to the left, a-c voltage and resulting current is carried by r(‘-

sistor RaThis causes a voltage at the input of the amplifier-relay. Tin'

a-c amplifier operates an electric relay to close a forward drive circuit
or a reverse drive circuit when the current through resistor Ra is not
 176 Automatic Process Control

balanced by a feedback current through resistor The feedback

voltage from the power-motor slidewire
is applied to the integral bridge.

Neglecting the operation of this bridge, the net efTect of the feedback
voltage is to cause a current through resistor Rh which must balance the
current through resistor Ra. This is proportional action and the power-
motor arm must ‘‘follow’* the actuating signal contactor. Integral
action is as follows: Whenever a current through resistor Ra,
there is

caused by the actuating-signal voltage, the controlled variable is not at

the set point. The integral bridge, consisting of,»the four thermistors,
is not balanced because of the necessary current through resistor /?&•

Therefore, a voltage results at points 1 and 2 of the bridge and the

integral amplifier is energized. The integral amplifier operates heaters
which heat the thermistors of the reset bridge. The action of the
integral amplifier is to change the current through resistor R^,, again
energizing the amplifier-relay and driving the power-motor arm further
to the left. This in turn causes further unbalance at the integral bridge
which results in a continuous motion of the power-motor arm. The pro-
portional sensitivity is adjusted by setting the ratio of resistors and
The integral time is adjusted by setting the integral amplifier
input resistance.

Kia 5-25. Electric proportional-reset controller (The Hays Corp.).

The pneumatic-feedback electric controller is shown in Fig. b 25. The

controlled variable c is a pressure and acts against the diaphragm causing
a force to the right against the balance arm. The set point is a spring
foree acting on the balance arm. If the controlled variable increases,

the balance arm moves to the right and causes the contact to close a
circuit to the power-motor. The output arm of the power-motor moves
counter clockwise and moves the piston of the feedback cylinder to
the left. This action compresses the air in the feedback system and the
feedback diaphragm provides a feedback force balancing the actuating
signal force.This is proportional action. The pressure in the feedback
system is now greater than atmospheric pressure and air flows out the
Controlling Elements 177

integral restriction to atmosphere. This decreases the force on the

balance arm, again operates the power motor to continue the counter-
clockwise motion. This is the integral action. For an opposite change
of the controlled variable, a vacuum is caused in the feedback system.
The proportional sensitivity may be adjusted through suitable levers at
the feedback diaphragm or by a change in volume of the feedback
system. The integral time is adjusted by setting the restriction to
atmosphere.
The principal characteristic of the electrical controllers des(^ribed in
this section is the use of the reversible electric motor as a final control

element with the position of the output-arm fed back in a proper manner
to provide the desired control action. The advantages of this system
arc that the actuation is electrical and therefore distance between
elements is theoretically unlimited. Another advantage is that the
actuator is enclosed within the controller loop.

Electronic Proportional Controllers

The electronic proportional controller is distinguished by the use of

an electronic d-c amplifier with appropriate feedback for providing the
desired control action. Generally too, the output is simply a direct
current which is transmitted to any type of final control element arranged
for operation from a direct current.

Fjc. 6-26. Electroiiii; proportional-integral derivative ccnitroller (general scheme

similar to that provided by the Swartwout Ojmpany).

The a-c voltage-input proportional arnlroller is illustrated in Fig. G -20

and differs from the preceding electric controllers in that the actuating
signal is an a-c voltage rather than a mechanical dispiacement. The
measuring means consists of a resistance thermometer element or an
electrical pressure gage providing a 60-cycle voltage proportional to the
controlled variable. The set point means is a potentiometer providing
a 60-cycle voltage proportional to the set point. "I’lie difference in these
voltages is the actuating signal w^hich is passed to the a-c amplifier.
A rectifiei- follows to convert the a-c signal to a proportional d-c voltage.
 178 Automatic Process Control

The first d-c amplifier has a ‘^delayed*’ negative feedback and provides
proportional-derivative action as has been described previously for other
electric controllers as well as for pneumatic controllers. Proportional-
integral action is provided by the second d-c amplifier with ''advanced^'
negative feedback as has been described for the electric controller of
Fig. 6- 23 . Afljustable proportional sensitivity is provided by an adjust-
able gain of the a-c amplifier. Adjustable derivative time is provided
by the setting of the feedbacik resistor of the first d-c amplifier. Adjust-
able integral time is provided by the setting of thf? feedback resistor of
the second d-c amplifier. The output of the controller is a d-c current
which is carried through wires to a final control element. The com-
bination of proportional-derivative action and proportional-integral
action in cascade has an advantage with respect to start-up of control
systems. For the proportional action of the a-c amplifier and rectifier,

(6
- 13 )

For the proportional-derivative amplifier circuit,

^ = K2(1 + T^)
W3
(fi- 14 )

For the proportional-integral amplifier circuit

nil
(^ 15 )
m2
where e
— actuating signal = i; — c
K, = gain of a-c amplifier and rectifier
K2 =- gain of first d-c amplifier = (R2 + Rz)/Ri
Ki = gain of second d-c amplifier = R2/R\
mi = output current (or voltage) of second d-c amplifier
m2 = output voltage of first d-c amplifier
m3 --= output voltage of a-c amplifier
derivative time = R2RzC/{R2 R3)
- integral time = R2R3C/(R2 + R3)

Combining to eliminate the intermediate voltages m2 and 7/13,

The combination provides proportional-integral-derivative action as

shown by equation 6-16 wherein the derivative time and integral time

also affect the overall proportional sensitivity. The latter is also affected
Controlling Elements 179

by the gain of each amplifier. However the gains of the d-c amplifiers
are generally fixed and the gain of the a-c amplifier made adjustable.
The **voice-coiV* motor shown in Fig. 6-27 is the primary transducing
device employed with electronic controllers in order to convert direct
current into a mechanical force. The operation is based upon the force
caused in a conductor carrying a current through a steady magnetic
field. The force is linearly proportional to the product of current and
field strength.

The motor is constructed by winding a coil of fine wire in a cylindrical

form. The coil is maintained in an air gap between pole pieces con-
nected to a permanent magnet. The convStruction is therefore the same
as a permanent magnet loud-speaker. A current through the coil causes
a force to jiush the coil up or down through the gap. "I'he spring con-
verts the motor to a deflectional system so that displacement is propor-
tional to current.
The d-c current input proportional controller is shown in Fig. 6-28.
The difference between this controller and the controller of Fig. 6-26
lies in the arrangement of the actuating-signal means. The transmitter
produces a direct current of a few milliamperes proportional to the
controlled variable c. This current passes through the “voice-coir’ on
the balance detector and causes a force to act on the balance beam
opposing the action of the set-point spring. The resulting deflection of
the balance beam “couples” the oscillator c’oil The signal from the
oscillator coil is amplified to a d-c voltage by the preamplifier, and the
voltage is negatively fed back to a second “voice-coil” on the balance
beam thus restoring force balance at the beam. The output' signal of
the preamplifier is a d-c voltage proportional to the difference between
 180 Automatic Process Control

set point v and controlled variable c and is the actuating signal voltage.
This voltage is acted upon by the proportional-derivative amplifier and

network (first d-c amplifier). This amplifier has a delayed negative

feedback and the derivative time is adjusted by the resistor. The output
voltage of the first d-c amplifier is passed to the second d-c amplifier for
proportional-integral action. Integral time is adjusted by the resistor Ri
and the proportional sensitivity is adjusted by resistor Ke. The output
of the second amplifier is a direct current of a few milliamperes.

Fia. e-28. Electronic proportional-integral derivative controller (general Bcheme

similar to Manning, Maxwell & Moore, Tnc ).

The principal advantage of the electric controller of the electronic

d-c output current t3T3e is that it may be connected by simple two-wire
cables, and it provides great flexibility in obtaining combinations of
‘controller action.

Theory of Automatic Controller Circuits

All automatic controllers employ the basic principles of amplifiers
and lag elements produce proportional, integral, and deriva-
in order to
tive action or whatever action is desired. Every automatic controller
mechanism employing feedback is a miniature controlled system within
itself and is subject to all the analysis techniejues employed for automatic
control. Consequently, operational diagrams may be drawn to illus-

trate analogous forms of pneumatic, hydraulic, and electric controllers.

Several amplifiers employed in automatic controllers are shown

schematically in Fig. 6'29. The mechanical and fluid amplifiers are used

for either displacement or force amplification by suitable modification.

The electronic amplifier is employed for either voltage or current ampli-
 Controlling Elements 181

ficatioD. Magnetic and transistor amplifiers are also employed as well

as many ingenious combinations and modifications of all these amplifiers.

Hydraulic amplifier Electronic amplifier

Fio. 6-29. Amplifiers for control elements.

Proportional controllers usually employ feedback as shown in Fig. 6-30.

The pilot gain is usually large and the equation for the operation is

easily shown to be

(6-17)

Fig 6-30. Analog circuits for proportional control.

 Automatic Process Control

where Kj, » pilot amplifier gain

Ke = feedback gain

Since the pilot gain Kp is usually large (10^ to 10^) then

6- 18 )
(

The feedback gain becomes the proportional sensitivity of the controller.

The following figures show the use of these principles: Fig. 6-8, pneu-
matic; Fig. 6-12, pneumatic; Fig. 6-20, hydra*ulic; and Fig. 6-28,
The electrical d-c voltage analog of Fig. 6-30 may be analyzed
electrical.

by assuming a current flow i into the input resistor Ri '.

e — ei = iR\ (6-19)

The current flow into the amplifier is negligible (grid current only) so
all current ^ passes through the feedback resistor 7^2-

ei — mi = iR2 (6-20)

The amplifier has a fixed gain K and in addition changes the sign of the
voltage:
wij = Kci (6—21)

Combining the above three equations to eliminate current i and volt-

age ci,

mi Ro 1
~ ~
T Tf,

KRi
The amplifier gain K is usually large (10^ to 10®) so that the above
equation reduces to

- = - R2
Till

^ (6-23)
€ Hi

The negative sign shows that the output voltage (reference to ground)
is negative for a positive input voltage. As a proportional controller
the proportional sensitivity is adjusted by selecting the ratio of re-
sistances (R 2 /Ri)-
Proportional-derivatwe controllers one of two basic
employ either
principles: (1) delayed negative feedback, or (2) advanced input, as
shown in Fig. 6-31. The delayed negative feedback employs a feedback
capacitance to ground. Thus the output voltage mi must continually
change in order to charge the capacitor, 3delding a derivative effect.
The operational equation may be derived by the same procedure as
 Controlling Elements 183

before. Since the amplifier gain is very large (10^ to 10®), the opera-
tional equation for the left circuit of Fig. 6-31 is

-ji:e(r<,s+ 1) (&-24)

where Kc = 2R/Ri
Td = RC/2
The derivative time,T<f is adjusted by setting either or both resistors;
the proportional sensitivity is adjusted by setting the ratio of resistors.

Ri R2
r-V\Ar-r-W\rn

i>
Advanced input

Fig. 6-31. Analog circuits for proportional-derivative control.

The delayed negative feedback principle has been illustrated previously

in Figs. 6-9 and 6-14 (pneumatic) and Figs. 6-23, 6-26, and 6-28
(electrical).
The advanced input method is not employed as often but is very
easy to accomplish with fluid amplifiers. The operational equation
may be shown to be

—= -KciTiS + 1) (6-26)
e

where Ke = R 2 /R 1

Td = RiC

This method is illustrated by the pneumatic circuit of Fig. 6-15 although

there the gain is not large.
Proportional4ntegral controllers may be one of three types: (1) delayed
positive feedback (Fig. 6-32); (2) advanced negative feedback (Fig.

6-33); (3) delayed input (Fig. 6-33).

The delayed positive feedback method employs a negative feedback
through lower resistor R4 ,
thereby producing a proportional action.
Two feedback amplifiers are involved. The delayed feedback through
upper resistors /2, i?i, and R^ is positive because of the additional
feedback amplifier. The positive feedback produces a regeneration of
amount controlled by capacitor C and resistors R. The operational
184 Automatic Process Control

equation may be shown to be, if the resistor Ri * 2R,

7 +
where Kc = R4/R2
= RC/2
The integral time is set by adjusting either one or both resistors R.
The proportional sensitivity is set by ratios of feedback resistors. This

Fig. 6-32. Analog circuit for proportional-integral control —delayed positive

feedback.

method employing mechanical feedback amplifiers and a pneumatic

amplifier is exemplified by the pneumatic controllers of Figs. 6-10

and 6-13.
The advanced negative feedback method of Fig. 6-33 is commonly
employed in electric controllers. By inspection of Fig. 6-33 it is seen

VW 11“

Advanced feedback Delayed input

Fio. 6-33. Analog circuits for proportional-integral control.

that unless the input voltage is zero there must exist a rate of change

of output voltage in order to m^ntain a current into the condenser and

thus its charge. The operational equation is

mi
(6-27)
e
where Ke R/R\
Ti RC
Controlling Elements 185

The integral time is adjustable by resistor R and the proportional

sensitivity by the ratio of resistors. The
systems of Figs.
electrical
6-26 and 6-28 use this method.
The delayed input method of Fig. 6-33 is sometimes employed as a
lag network in order to achieve proportional-integral action. The oper-
ational equation is

where Kc = R^/2Ri*
a = 2/^2/ (2/^2 “h ^1)
T = (2/?2 -f Ri)C/2

If it is arranged that resistor R2 is much smaller than resistor Ri and

it capacitor C is large, then the coefficient a of equation 6-28 is negligible
and the response approaches proportional-integral action. This method
is that employed in the pneumatic controller of Fig. 6-15.

Other control actions such as proportional-integral-derivative action

may be obtained via a suitable combination of the five circuits of
Figs. 6-30, 31, 32, and 33. These combinations and many others that
are in infrecjuent use are too numerous to review here. Many of these
combinations perform according to the equation

— = — + K, + Kis (6-29)
€ S

where Ki is some integral-like adjustment, Kc is some gain adjustment,

and Kd is some derivative-like adjustment. The three adjustable coeffi-
cients are not always individually adjustable by physical knobs at the
controller. It is usually the case that the knobs actually adjust a com-
bination of coefficients (Kij Kct Kd). The manufacturer of the con-
troller should be consulted as to the arrangement of the adjustable
parameters of the controller.

Electrical Proportional Controllers for Electric Heating

The by amplitude modula-
throttling or modulating of electrical heat
tion of electricpower proves to be expensive because of the large amount
of electrical equipment involved. A simpler expedient is to turn on and
off the electric power by means of contactors or relays. The power
can be modulated as an average over a period of time by turning on the
electric power for var3ring time fractions of a fixed time pycle. Thus if
a relay were turned off every P minutes but turned on for / minutes in
each P period, the average electric power could be manipulated. The
186 Automatic Process Control

manipulated variable is then

/
m= (6-30 )
P
where m= fraction of time on = average power
/ = time on
P= period of cycle (usually fixed)

The period of cycle P is about 10 to 120 sec for most furnaces.

A proportional controller for electric heating ^ shown in Fig. 6-34.
The actuating a position given by a self-balancing poten-
signal e is

tiometer, for example, and determines the position of the upper contact

' 7^
j©
r Furnace heater
Synchronous motor

Fig. 6-34. Proportional controller ff)r electric heating.

of a single-pole switch. The lower contact is carried on a movable arm,

the end of which is oscillated up and down by a cam and synchronous
motor. The time of ‘‘make'' of the contact is proportional to the height
of the upper contact; the higher the contact, the less time on. There-
fore, the manipulated variable m is proportional to the actuating signal e,
and proportional control results. The proportional sensitivity is set by
adjusting the levers. The period is adjustable by setting the cam speed.

Floating
motor

ij--
Furnace
Synchronous heater
motor

Fio. 6-35. Single-fipeed floating controller for electric heating.

The single-speed floating controller of Fig. 6-35 is commonly used for

electric furnaces. A milli voltmeter pyrometer, potentiometer pyrometer,
or expansion thermometer contains Hl-LOW contacts operated in such
a manner that when the controlled variable is below the set point, the
low contact is made; above the set point, the HI contact is made.
Controlling Elements 187

In a narrow region near the set point, neither contact is made. This
latter region is termed the neutral zone.
The HI-LOW contacts supply power to drive a slow-speed motor
in one direction or another, and to increase “time on“ when the LOW
contact is made and to decrease time on when the HI contact is made.
The floating motor usually requires several minutes to change from zero
“time on“ to full “time on.“ Single-speed floating control action re-
sults. In equation form,

1) m
when c> p H- Ar/2

1m=0 when V+ N/2 < c> V - N/2 ( 6 ^ 1 )

\m = -K when c<v - N/2

where m= fraction of “time on“
K — floating rate
c — controlled variable
V = set point
N= neutral zone

The floating rate K is adjusted by selecting a gear ratio for the floating
motor. The neutral zone N is adjusted by setting the position of the
HI-LOW contacts. The period of cycle P is adjusted by selecting a gear
ratio for the synchronous motor.
A ’proportional-integral controller for electric heating is shown in
Fig. 6-36. This arrangement provides a control action in which the

Fio. 6-^. Proportional-integral control for electric heating (Leeds and

Northrup Co.).
188 Automatic Process Control

fraction of ‘Hime-oii” of the load relay m is proportional to the actu*

ating signal e. In addition, integral and derivative actions are also
provided. The actuating signal is generally a mechanical displacement
provided by a self-balancing potentiometer or resistance thermometer.
The operation of the circuit is described briefly as follows: Suppose that
the actuating signal is positive so that the contactor at e is above the
center position. This causes a voltage to appear at the amplifier which
closes the load relay and feedback relay. This connects the battery to
the resistor R, and capacitors Ci and C2 begin slowly to charge. Soon
the voltage across capacitor C2 exceeds the previous input voltage and
the amplifier drops out both relays. Tn actual practice the relays must
be interlocked. The feedback relay is then connected so as to discharge
the capacitors. If the capacitor charge and discharge rates are the
same, then the load relay on for one-half time and off for one-half
is

time. than they charge, the load relay

If the capacitors discharge faster
is on for greater than one-half time and if the capacitors discharge more

slowly than they charge, the load relay is on for less than one-half
time. The fraction of time on mi is therefore proportional to the actu-
ating signal because the rates of (charge and discharge are determined
by the net voltages from the input and feedback.

Fig. 6-37. ProportioDal-integral control for electric heating (Minneapolis-Hon^y-

well Regulator Ck).).

The total period of cycling is governed by the gain of the amplifier

and the interlocking of the relays. The proportional sensitivity Kc is
adjusted by the potentiometer at the input circuit. Integral time Ti
and derivative time Td are adjusted by the resistors in the feedback
circuit.
Controlling Elements 189

The proportional-integral controller ^own

6-37 operates in a
in Fig.
very similar manner except that the cycling action is governed by the
thermal bridge. When the feedback relay closes the upper contact,
the lower heater and thermistor rise in temperature and unbalance the
center-tap feedback voltage. When the feedback voltage exceeds the
input voltage, then the amplifier reverses the feedback relay. The total
period of cycling is adjusted at potentiometer P by selecting the over-
swing voltage of the thermistor bridge. Integral action is provided
by the integral amplifier, heaters, and thermistors, the latter being
thermally lagged. Whenever the actuating signal is not zero, the in-
tegral amplifier turnson the appropriate integral heater thus providing
a feedback voltage to actuate the proportional amplifier. Integral time
is adjusted by setting the gain of the integral amplifier.

Two-Position and Floating Controllers

Two-position controllers are generally electrical in operation although
pneumatic proportional controllers with the proportional sensitivity set
very high are sometimes used. The thermostats in Fig. 6-38 illustrate
the various differential gaps in two-position control. The thermostat
on the left operates load contacts directly from the bimetallic element.
The set point is adjusted by setting one contact or by repositioning the
whole bimetal element. The gap of this arrangement is theoretically
zero although in actual practice there is a very small gap due to con-
tact “stickiness.”

Positive gap Negative gap

Two-poaition (thermostat) controllers.

A positive differential gap is often necessary in two-position control

in order to save wear on the control apparatus. The mechanism of
Fig. 6-38 employs a toggle action to produce a positi\'e gap. The gap is
adjustable by the tension of the spring because this determines the sup-
pressive (hysteresis) force apting on the element.
190 Automatic Process Control

A negative differential gap is often used on domestic thermostats in

order to reduce an abnormally long period of oscillation. An auxiliary
heater operated by the load contacts or by an auxiliary relay turns on
the heater when the contact is made. Thus the bimetal element opens
the load contact before the actual temperature attains the same point.
The load contact make point is normal. This is sometimes termed
'^anticipation.^’
There are many forms of two-position controllers of the "thermostat”
or "pressurestat” type for both domestic and industrial application.
These offer many arrangements of load contacts, set point, and differen-

tial gap. Generally, a solenoid valve of one type or another or an

electrical heating element, motor, or other power device is operated by
the controller.
For industrial service the controller of Fig. 6-39 is often used because
itonly requires moving a light aluminum vane in or out of two coils.
The vane may be carried directly by the galvanometer pointer of a

milli voltmeter. The coils may bo positioned by a set-point mechanism

or may be fixed. The left-hand tube circuit is an oscillator with the
vane out of the coils. When the vane is between the coils, the oscillation
is The right-hand tube circuit picks up the oscillatioji
greatly reduced.
and blocks the relay current. Thus the relay coil is energized when the
vane is between the coils and is de-energized when the vane is out of
the coils.

For floating controller action a neutral zone is usually necessary.

This is obtained by a second contact position in the controller and is
often obtained with two independently set on-off control mechanisms
Single-speed floating control may be achieved by means of the same
equipment as for two-position control with an electric-motor-operated
valve. The difference lies only in the speed of operation of the motor.
In tw'o-position control the motor valve timing is 120 sec or less. In
single-speed floating control the timing for valve stroke is usually
Controlling Elements 191

120 sec or more. An electrical interrupter is sometimes employed in

conjunction with the motor to decrease the speed of opening and closing
the valve.

PROBLEMS

(^1. Find e = /(v, c) for a motion of the differential lever.

&^2. Show that the pneumatic amplifier has proportional actions described
by pressure m - Ke.

6--3. Show that the pneumatic amplifier has proportional actions described
by m = Ke.
192 Automatic Process Control
6-4. The flow through the open nozzle when the velocity at the restriction

is acoustic velocity is given by 7 = KjrDep/y/ T, where K

is the orifice co-

efficient, and T is upstream temperature. Linearize the equation.

-ir

6-5. Assuming a linear relation for the nozzle, derive the output-input
relations for the hydraulic controller {A
- piston area, K — spring constant,
K 2 *= dq/dp nozzle,
= r)q/dp supply, K\ = dg/de nozzle).

6-6. For the gas-pressure regulator calculate the offset when acoustic
velocity exists at the throat.

6-7. CalculaU> the input-output relationship, and state the kind of control
action.

cl
Controlling Elements 193
M. Calculate the input-output relationship, and state the kind of oontrol
action.

6-9. Calculate the input-output relationship, and state the kind of control
*
action,

6-10. The device shown in the figure has proportional-integral action.

The position of the input shaft t is transmitted directly to the output m
through the dilTerential gears. The input c is also integrated by the disk and
is added into the output by rotation of the outside gear of the differential.

Assuming nominal gear ratios write a description of the operation.

1!>4 Automatic Process Control

6-12. A new controller action can be synthesized by deriving system

functions G and H. If G represents a high-gain amplifier, derive H to give

proportional-integral action; c = A+ r.
 Final Control Flements

The final control element is the mechanism which alters the value of
the manipulated variable in response to the output signal from the
automatic control device. The position of the final control element in
the automatic control loop is shown in Fig. 7-1. The final control

Gj La*!" C,"*
1 I 1 I 1
^ 1
I

'

b
L-^-l
!
H '

t*
!
1

I I

Gi Controlling element

G2 Final control element

G3 Process
H Measuring element

Fia. 7-1. The final control element.

clement often consists of two parts: first, an actuator which is used to

I ranslate the output signal of the automatic controller into a position of
a member exerting large power; and second, a device to adjust the value
of the manipulated variable (usually a flow rate of a fluid).
The actuator must provide an accurate output position proportional
to the input signal in spite of various forces acting on the output member.
The most important forces are

1. Inertia forces caused by the mass of moving parts

2. Static friction forces during impending motion of two adjacent
surfaces
3. Thrust forces caused by weight and unbalanced fluid pressure.
195
 :

196 Automatic Prooess Control

Thus, the actuator is often required to employ a power-amplifying

mechanism. As with automatic controllers, the actuator may operate
by pneumatic, hydraulic, electrical, or a combination of these means.

Pneumatic Actuators
Pneumatic actuators may operate directly from the pneumatic output
signal from a pneumatic controller, or they may employ a separate
source of compressed air. There are five common methods of operation
of pneumatic actuators. These are called

1. Spring actuator
2. Spring actuator with positioner
3. Springless actuator
4. Piston actuator
5. Motor actuator.

A spring achmior shown in Fig. 7-2 operates directly from the air-
pressure output of a pneumatic controller in order to provide an output
position proportional to the input air pressure. The diaphragm is

3-15 psi 0-2 in.

Fig. 7-2. Spring and diapliragm actuator.

usually made of fabric-base rubber,molded to form, and supported by

a backing plate. The input air pressure mi acts against the diaphragm
and causes a downward force which compresses the spring. At static
balance, the force of the air pressiu'e against the diaphragm equals the
spring compression force

(wj - mo) A = (7-1)

Final Control Elements 197

wtiere my » input air pressure} Ib/in.^

mo » input air pressure at zero stroke, Ib/in.^
A = effective area of diaphragm, in.*

K= spring gradient, Ib/in.

X = output (stem) displacement, in.

The standard input operating range of spring actuators is 3 to 15 psi

gage. The output displacement or stroke is generally between J and
3 in. and is limited by the allowable stroke of the diaphragm. For
longer strokes a piston-spring combination is employed.
The performance of a spring actuator is generally satistactory pro-
viding it is not used under conditions of excessive force on the stem,
Inertia forces due to the mass of moving parts must be limited by the
natural frequency of the system,

^ VK/M (7-2)

where /„ = natural frequency, cycles/sec

K - spring gradient, Ib/in.
M — total mass of moving parts, lb sec*/in.

The natural frequency should be at least 25 cycles per second, other-

wise the actuator stem may oscillate continuously when damping is

negligible.
Static friction forces must be limited to a low enough value that
excessive hysteresis does not result. For hysteresis less than one per
cent of full travel,

F/ ( 7^)

where F/ = static friction force, lb

Mr input operating range, Ib/in.*

This may be a serious limitation. For example, a spring actuator with

an effective area of 100 sq in. and an operating range of 3 to 15 psi cannot
.support more than 12 lb friction if the hysteresis is to remain less than
one per cent of full range. Thrust forces are also limited by the ability
of the actuator to provide full operating stroke,

Ft < moA (7-4)

where Ft is the total thrust force acting in one direction. This may also
be a serious limitation. For example, a spring actuator with an effective
area of 100 sq in. and an initial air pressure setting of 3 psi cannot sup-
port more than 300 lb unbalanced force. In addition, thrust forces
198 Automatic Process Control
must be relatively constant otherwise the stem position will not be
directly related to the input air pressure. The performance of a spring
actuator is also influenced by the characteristics of the spring and
diaphragm. A well-designed actuator has a linear static relation be-
tween input air pressure and output stroke if the effective area of the
diaphragm and the spring gradient are constant throughout the stroke.
Hysteresis due to the stresses in the spring and diaphragm are usually
less than one or two per cent of full stroke. .

The spring actuator often requires a positioner as shown in Fig. 7-^

when static friction forces are large or when the response of the motor is
too slow. The positioner consists of an input bellows, a nozzle and
amplifying pilot, and the feedback levers and spring. An air supply of

Fio. 7-3. Spring: and diaphragm motor with positioner.

from 20 to 100 psig must be provided. The operation is as follows:

When the input air pressure mi increases, the input bellows moves to the
rightand causes the baffle to cover the nozzle. The nozzle back-pressure
change is amplified by the pilot and is transmitted to the diaphragm.
The diaphragm moves down and the feedback lever compresses the
spring to return the baffle to a balanced position. Thus the actuator
Final Control Elements 199

i«tein assumes a position dictated by the input air pressure. The spring
actuator becomes a power means and the characteristics of the spring
and diaphragm are relatively less important. The use of the positioner
results in several improvements in performance.

1. Hysteresis is reduced and linearity

is usually improved because the

static operation isgoverned by the feedback spring and input bellows.

2. The actuator can handle much higher static friction forces because

of the amplifying pilot.

3. Variable thrust forces on the motor stem do not disturb the stem
position to any great extent.
4. Speed of response is generally improved because the pneumatic
controller must supply sufficient air to fill the small input bellows rather
than the large actuator chamber.

The use of a positioner with a spring actuator does not improve the
ability of the actuator to handle larger inertia or thrust forces unless
special adjustments of motor operating range are made. The only dis-
advantage in the use of a positioner is that it may require maintenance.
The 8pringle88 actuator shown in Fig. 7-4 is useful for large thrust
forces. The only difference between the springless actuator and the

spring actuator with positioner of Fig. 7-3, is that the spring of the
actuator is replaced by a pressure regulator which maintains a constant
pressure on the under side of the diaphragm. An air supply at a pres-
sure of 20 to 100 psig is required. The operation of the springless
actuator is as follows: Assume that the cushion regulator is set to pro-
vide 9 psig pressure on the under side of the diaphragm. At static
balance and with no thrust force on the actuator stem, the upper side
pressure must be 9 psig. Then if the input pressure increases, the
nozzle back pressure increases, and the upper side pressure is raised to
198 Automatic Process Control
must be relatively constant otherwise the stem position will not be
directly related to the input air pressure. The performance of a spring
actuator is also influenced by the characteristics of the spring and
diaphragm. A well-designed actuator has a linear static relation be-
tween input air pressure and output stroke if the effective area of the
diaphragm and the spring gradient are constant throughout the stroke.
Hysteresis due to the stresses in the spring and diaphragm are usually
less than one or two per cent of full stroke. ,

The spring actuator often requires a positioner as showm in Fig. 7-3

when static friction forces are large or when the response of the motor is
too slow. The positioner consists of an input bellows, a nozzle and
amplifying pilot, and the feedback levers and spring. An air supply of

Fro, 7-3. Spring and diaphragm motor with positioner.

from 20 to 100 psig must be provided. The operation is as follows:

When the input air pressure mi increases, the input bellows moves to the
right and causes the baffle to cover the nozzle. The nozzle back-pressure
change is amplified by the pilot and is transmitted to the diaphragm.
The diaphragm moves down and the feedback lever compresses the
spring to return the baffle tO a balanced position. Thus the actuator
Final Control Elements 199

stem assumes a position dictated by the input air pressure. The spring
actuator becomes a power means and the characteristics of the spring
and diaphragm are relatively less important. The use of the positioner
results in several improvements in performance.

1. Hysteresis is reduced and linearity

is usually improved because the

static operation isgoverned by the feedback spring and input bellows.

2. The actuator can handle much higher static friction forces because

of the amplifying pilot.

3. Variable thrust forces on the motor stem do not disturb the stem
position to any great extent.
4. Speed of response is generally improved because the pneumatic
controllermust supply sufficient air to fill the small input bellows rather
than the large actuator chamber.

The use of a positioner with a spring actuator does not improve the
ability of the actuator tohandle larger inertia or thrust forces unless
special adjustments of motor operating range are made. The only dis-
advantage in the use of a positioner is that it may require maintenance.
The Bpringless actuaior shown in Fig. 7-4 is useful for large thrust
forces. The only difference between the springless actuator and the

spring actuator with positioner of Fig. 7-3, is that the spring of the
actuator is replaced by a pressure regulator which maintains a constant
pressure on the under side of the diaphragm. An air supply at a pres-
sure of 20 to 100 psig is required. The operation of the springleas
actuator is as follows: Assume that the cushion regulator is set to pro-
vide 9 psig pressure on the under side of the diaphragm. At static
balance and with no thrust force on the actuator stem, the upper side
pressure must be 9 psig. Then if the input pressure increases, the
nozzle back pressure increases, and the upper side pressure is raised to
200 Automatic Control

a high value. The actuator stem then moves downward and, as the
actuator stem attains the new position, the upper side pressure is
returned to 9 psig. If there is an upward thrust force on the actuator
stem, the under side pressure remains at 9 psig but the positioner
raises the upper side pressure until static balance is achieved. For a
downward thrust force the upper side pressure is reduced below 9 psig.
Thus, the springless actuator can counteract a thrust force equal to
approximately the underside pressure times the area of the diaphragm.
This is generally from three to ten times the tfirust force handled by
a spring actuator with or without a positioner.

Fin. 7-5. Doubicj-ucimg pist<>ii-}u;tuat<>r.

The d()vble~acti7ig 'piston actuator of Fig. 7-5 is employed for larger

thrust forces than can be handled by the single-acting actuator, and the
piston is used in order to obtain long stroke. The pilot is gc>ierally
a spool-type diverting valve and requires an air supply of 30 to 100 psig
pressure. The operation is as follows: When the input pressure r«i
increases, the bellows moves to the right and pushes the pilot spool
upward. This action opens the upper side of the cylinder to the air
sujjply and opens the lower side to atmosphere; thus the action is to
return the piston to the neutral position. Thus the position of the
piston is proportional to the input pressure. A double-acting piston
actuator can handle a thrust .force equal to about 80 per cent of the
supply pressure times the area of the piston.
The rotary actuator in Fig. 7-0 is used for very large tl^rust force or
torque. The air motor is a reversible vane-type or positi ve-displacemen t-
type motor operating from 80 to 100 psig air pressure. The operation
is as follows: When the input pressure nii increases, the pilot piston
moves upward and supplies high pressure to one side of the air motor.
The other side of the air motor is exhausted. The motor drives the rack
 Final Control Elements 201

downward, compresses the feedback spring, and returns the pilot piston
to the neutral position. Actuators of this kind are employed in sizes
from 1 to 15 hp and will handle thrusts as high as 100,000 lb.

Table 7-1 summarizes the application of pneumatic actuators.

Table 7-1. Range of Application of Pneumatic Motor Operator*

Friction
F’orce
Thrust for 1%
Approx. Stroke, Force, Hysteresis,
hp in. lb lb
Spring actuator (3-15 psi range) 0,05 to 3 to 800 to 30
Spring actuator and positioner 0.15 to 3 to 800 to 300
Cushion actuator 0.15 to 3 to 2000 to 300
Double-acting piston actuator 0,1-1 to 36 to 5000 to 600
Rotary motor actuator 1-15 to 60 to 100,000

All pneumatic actuators require a source of clean, dry air The

arrangement of pi oper air supply has been discussed in Chapter 6 (see
Figs. 6-16 and 6-17). For pneumatic actuators it is necessary also
that the requisite quantity of air be available, because pneumatic
actuators require large power input to produce large power output.
It is particularly important that supply lines be made large enough

so that line losses are small.

Example 7-1. A spring actuator is to be used for positioning a dead weight

of 300 lb on its stem. The pressure input is 3 to 15 psig and the stroke
is 2 in. What is the minimum diaphragm area and resulting natural frequency?
From equation 7-4,
Ft
^ 300 lb
100 sq in.
mo 3 psi
202 Automatic Process Control
From equations 7-2 and 7-1,

“ _L ± iMid
2ir \M 2ir V Mx
^ Jl^ /l2 X 100 X 32.2 X 12
4.4 cps
2ir \ 300 X2
Example 7-2. A springiess actuator has a diaphragm of 100 sq in. area.
Its positioner operates from 3 to 15 psig. The cushion (under-side) pres-
sure is set to 5 psig. What range of thrust load can ke accommodated?

Maximum upward thrust = (15 — 5) X 100 = 10001b

Maximum downward thrust = (5 - 3) X 100 = 200 lb

Electro-Pneumatic Actuators
When electric control systems are employed, it is often advantageous
to use a pneumatic actuator. a suitable air supply is available, a
If
pneumatic actuator can provide very large power output and may be
operated directly from an electric control system. This requires trans-
ducing the electrical output of the controller into an input variable for
the actuator.
The eleclro-pneurnatic pilot of Fig. 7-7 is arranged to convert an
electrical signal input to a proportional air pressure output. An elec-
tric control system employing this pilot may operate with any of the

Air supply
I

pneumatic actuators previously discussed. The input electric signal

(usually a direct current) enters the “voice-coil” motor. The “voice-
coil” motor was described by Fig. 6-27 in Chapter 6. The input coil is
supported in the field of a permanent magnet so that the coil afifords a
force proportional to the magnitude of the input d-c current. The force
causes a deflection of the balance beam, covers the nozzle, and results
 Final Control Elements 203
in an increase of output pressure. The output pressure acts on the
feedback bellows to cause a torque on the balance beam equal but
opposite to that of the voice coil. The output pressure therefore is
proportional to the input d-c current.

The eUctro-pneMmatic actuator of Fig. 7-8 combines the voice coil

and the pilot in the positioner of a pneumatic actuator. The motion of
the output of the actuator is related to the balance beam through the
feedback lever. The output position of the actuator is therefore pro-
portional to input d-c current.

Hydraulic Actuators
Hydraulic actuators, as used for industrial process control, accept a
signal from a pneumatic controller or an electric controller and employ
hydraulic pressure to drive an output member. The hydraulic actuator
is used where high speed and large forces (or large power) are required.
The hydraulic piston actuator in Fig. 7-9 has as its input mi the
position of the vertical lever. For a pneumatic-hydraulic actuator the
input would be the position of a bellows. The balance lever pivots at
the bottom so that an increase of input (to the left) pushes the pilot
piston to the left. This action opens the left end of the piston to supply
pressure and opens the right end of the piston to drain. The large
power piston, therefore, moves to the right until, as the balance lever
rotates about the topmost end, the pilot piston is returned to center,
the motion of the output xi is therefore proportional/ to the input
motion nii. The hydraulic actuator requires a continuously running
electric motor and pump to provide a source of high-pressure oil, and a
drain or sump to collect the return. Hydraulic supply was discussed
in detail in Chapter 6.
204 Automatic Process Control

The motor actuator uses a variable-delivery pump shown in Fig. 7-10

in order to produce a proportional output position or speed. The pomp
consists of a disk driven by a constant-speed electric motor. A number

Oil
Oram supply Drain

Fig. 7-0, Hydraulic piston actuator.

of pistons are attached to the disk and reciprocate in a cylinder body.

The cylinder body rotates with the disk. When the disk and cylinder
body are axial, the pistons remain stationary and there is no oil delivery.

Maximum delivery

Fig. 7-10. Hydraulic variable-delivery pump

When the cylinder body is tilted, the pistons reciprocate and the oil

delivery is proportional to the amount of tilt. The actuator is shown in

Fig. 7-11. The input signal mj is the position of the tilt control lever.
Final Control Elements 205

TTie pump connected to a positive-displacement motor. The

is directly

motor is ph 3r8ically the same as the pump but has a fixed ^^tilt” and thus
a fixed displacement. The same arrangement may be employed for

Suction t lorain
Fia. 7-1 1 . Hydraulic motor operator with rotative output.

for
position control (output position proportional to input signal) or
speed control (output speed proportional to input signal). Hydraulic
motor actuators are used in sizes from three to three hundred horsepower.
The electro-hydraulic piston actuator is shown in Fig. 7-12. The input
device is the voice-coil motor which positions a three-land pilot spool.
The pilot controls the flow of oil to the cylinder, and the piston
motor

Fig. 7-12. Electro-hydraulic piston actuator.

actuates a spring feedback to the pilot spool. The operation is as

follows; An increase in cui’rent to the voice coil causes the arm to swing

downward thereby pushing down the pilot spool. This action connects
side to
the lower side of the cylinder to drain and connects the upper
The piston then moves downward, and the feedback
supply pressure.
Tlie position
spring pushes back to rebalance the system at equilibrium.
current.
of tlie piston rod is therefore proportional to input direct
206 Automatic Process Control
Hydraulic actuators require a hydraulic power supply. This subject
has been discussed in Chapter 6, see Fig. 6-19. Generally, individual
power supplies are required at each actuator unless several actuators are
grouped close enough together to use a single large unit. Hydraulic
power lines should not be run for appreciable distances. The pressure
employed in most hydraulic systems usually is from 400 to 1000 psi gage.
Example 7-3. A hydraulic piston actuator with a 4-in. bore and 3-in.
stroke used with a pump having a maximum delivery of 5 gpm. Neglecting
is

acceleration and output load, calculate the time to travel the complete stroke.

_ ir4* X 3 X 60 ‘
^
Traverse t;me = ^ ^ ^ ^ ^3,
= 1-96 sec

Electric Motor Actuators

Electric motor actuators are employed to provide a position output
corresponding to an input electric signal. These are of three basic types:

1. The bridge type systems discussed in Chapter 6, see Figs. 6-22, 6-24
2. The relay-type reversible motor drive (see also Fig. 6-25)
3. The amplifier-t3rpe reversible motor drive
The relay-type electric actuator is shown in Fig. 7-13. A low-inertia
reversible a-c motor drives the output through gears. A mechanical
feedback is employed such that a rotation at the output rotates a lever
Feedback

Fia. 7-13. Relay-type electric actuator.

acting in a feedback spring. The operation is as follows: An increase

in direct current at the input coil causes a force to act in the balance
beam and pulls the beam down thus ‘‘making"
the upper contact.
This drives the motor one direction so that the output and feedback
in
rotate, thereby pulling on the spring and rebalancing the voice-coil
armature. Output position is therefore proportional to input current.
Many variations of this general scheme are employed. The power ampli-
fiermay be inserted between the balance beam and motor so that the
contacts do not have to carry large current. The amplifier may be
actuated by a differential transformer or an oscillator coil on the balance
beam. Relays may be eliminated if saturable-core reactors are used.
Final Control Elements 207

An amplifier-type electric actuator is shown in Fig. 7-14. Ilie input

signal is applied directly to the input stage of the d-c amplifier. The
modulated to an a-c signal of varying phase, amplified,
difference signal is
and passed to a two-phase motor. The motor, usually termed a servo-
motor, has a low-inertia rotor connected to a low-inertia output gear
train. One motor winding is energized from a-c line. When the con-
trolled phase leads the line phase, the motor is driven in one direction,
and when the phase lags the line phase, the motor is driven in the
opposite direction. The output drives the load and a feedback poten-
tiometer. When th^ input and output positions match, the difference

Pio, 7-14. Amplifier-drive electric motor actuator (The Conoflow Corporation)

signal is both motor windings are energized with the rame phase,
zero,
and the motor is stalled. There are many variations of this general
arrangement. For example, a pair of selsyns (synchros) may be used
inan a-c system. The amplifier may employ a grid-controlled thyratron,
or magnetic amplifiers may be used.
The power of the electric motor actuator is often J hp (about 200
watts) or less. One-tenth horsepower (75 watts) or less is more common.
Thus the electric motor actuator is generally employed for low power,
whereas pneumatic or hydraulic actuators are used for large power.

Example 7-4. Anmotor actuator incorporates a 25-watt motor.

electric
Full travel of 2 in. is Assume a gear train efficiency of 50
desired in 10 sec.
per cent and calculate the maximum output force.

r.
Force = ^
0.5 X — _X
650 25 X II
j
X 10
5501b

Two-Poeition Motor Actuators

For two-position control an actuator need only assume ^two positions
so as to open and close a fluid valve or other element.The pneumatic
diaphragm actuator of Fig. 7-15 provides two positions of the stem
corresponding to large input pre^ure and zero (gage) input pressure.
208 Autoimitic l^'ocjesa C'ontrol

The input pressure is placed over the diaphragm for upward-acting

thrust force or under the diaphragm for downward-acting thrust foroe.
If the thrust forces cannot be depended upon, a light spring is added

to the actuator stem. For larger strokes and forces a piston-cylinder

combination may be employed.
Input
I

Input Forward Reverse

^ I
Output
Electrical motor
Clectrii dl solenoid

Fi(5 7-15. Two-position operators.

The electric solenoid of Fig. 7-15 is widely employed for two-position

control with either direct or alternating current Spring-closing or
spring-opening types may be selected depending upon which is desired
for safe operation.
The electric motor of Fig. 7-13 may be used in several ways. A revers-
ible electricmotor geared to an output arm may be run either forward
or reverse and the output motion limited by the position of limit switch
on the output arm. A nonreversible motor with spring return may also
be used so long as the gear ratio is low enough that the spring will
drive the armature backward through the gear train.
Final Control Elements

Final Element Power Failure

In the event of power failure the final control element, whether it is

pneumatic, hydraulic, or electric, may have one of three possible

behaviors:

1. The element may fail open, forward, or upward (closed if a valve,

forward if a rotary motor, and upward if a sliding stem). This will be
referred to hereafter as “open.’*
2. The element may fail and hold the position which it was in last.
This will be termed “last position.”
3. The element may fail revei*se or down. This will be termed
“closed.”

It is possible that any one of these may be desirable in a given instance.

For example, it is usually necessary for a boiler draft damper to fail
“open” so that gases may pass out of the system. In a continuous blend-
ing stream, it is generally desirable for a valve to fail in its “last position”
because a minimum amount of product is spoiled while the situation is
being remedied. On the other hand, the fuel-oil valve to a furnace
should fail “closed” for greatest safety.
Pneumatic or hydraulic piston or diaphragm actuators are easily
arranged to fail in the open or closed position by means of a spring force
or by the force due to pressure locked on one side. Spring actuators
almost always fail either open or closed because of the spring action.
Failure in last position is usually more difficult to accomplish but may
be done with suitable check valves to lock system pressures. The seals
may be expected to last for a number of hours. Various mechanical
toggle and pin latches are infrequently used.
Rotary pneumatic, hydraulic, and electric actuators almost always
employ some form of gear reduction. The efficiency of these gear trains
is usually less than 50 per cent and they are therefore self-locking.

That is, the input cannot be driven backward by a toixiue applied at the
output. Such actuators therefore fail in the last position by inherent
action. Failure at the open or closed position can only be a(icompli!^hed
by declutching mechanisms and like.
The thrust force on single-seat valves and butteifly valves may either
assist or prevent closing or opening on power failure. Generally it is
better to arrange that the forces due to pressure differential assist the
action desired. It is well to remember, however, that thrust forces are
not always in the direction of pressure differential.
210 Automatic Process Control

Methods of Fluid Control

A source of fluid head is required to provide the transport of fluids
through processing equipment. In a tubular heat exchanger, for
example, a pump is generally used to move the fluid through the fluid

Fia. 7-16. Methods of fluid control, (o) Series throttling. (6) Bypass, (c) Variable
delivery.

resistance caused by the tubes of the heat exchanger. Several arrange-

ments of the fluid source may be used depending upon the type of fluid
and processing equipment. A pump, compressor, blower, or fan is
ordinarily used to provide the source of fluid head or pressure. Six
arrangements of fluid sources are shown in Fig. 7-16. A control valve
Final Control Elements 211

or motor operator is used in each arrangement in order to provide a

proportioning of the flow of fluid through the process equipment.
Series throttling is probably the most common arrangement of a fluid
source. Referring to Fig. 7-16, a head tank provides a source of liquid
head or gas pressure. The head in the tank requires separate automatic
control in order to maintain a constant head. A control valve in series
with the head tank throttles the flow to the process equipment. Alter-
nately a constant-speed motor, turbine, or engine drives a pump. (A
turbine or engine requVes a speed governor in order to maintain its speed
within desired limits). Series throttling from a pump source may be
used with any type of axial or radial pump, compressor, blower, or fan,
but the head-flow characteristic should be as flat as possible in order to
provide constant head. Axial gas compressors do not have a smooth
head-flow characteristic and cannot operate at low flow rates. Series
throttling cannot be used with a positive-displacement pump or blower
because the outlet head becomes excessive at low flow rates.
The bypass is a simple arrangement used for positive-displacement
pumps, compressors, and blowers. The driving motor, turbine, or
engine operates at constant speed. Flow through the process equipment
is proportioned by passing a portion of the flow either to the pump
inlet or to the downstream side of the process equipment. The economy
of the bypass arrangement is not always high, because a considerable
portion of the energy stored in the fluid is wasted in the bypass. A better
arrangement for a positive-displacement pump is to throttle the delivery.
Variable delivery is a very useful source arrangement for either very
low flow rates or for very large flow rates. Referring again to Fig. 7-16,
a variable-delivery pump is driven by a constant-speed motor. A motor-
operator is connected to the stroke- varying arm of the
in such pump
a way that the volume delivery of the pump is changed. A relief valve
is sometimes necessary at the pump outlet in order to prevent excessive

pressure. A variable-speed drive connected to a pump, compressor, or

blower is also a variable-delivery source. Two methods are in common
use; first, the speed of an electric motor, steam turbine, or steam engine
may be changed by throttling the energy supplied to the motor, and
second, a variable-speed drive may be interposed between the constant-
soecd motor and the pump. The variable-speed method has an advan-
tage for verj^ large pumps and compressors because it is less expensive
to throttle the steam-flow to a turbine or engine than it is to throttle
or bypass the pump outlet.
The choice of any one arrangement of a fluid source depends upon a
great number of factors such as the type of fluid, the size of the installa-
tion, and the efficiency. The arrangement of the fluid source is very
212 Automatic Process Control
important in automatic control, and it is probably safe to say that most
difficulties in automatic control arise because of poorly installed or
maintained fluid sources.

Fluid Flow Through Control Vaivcis

A control valve in a pipeline acts as a variable restriction. Referring

to Fig. 7-17, the vertical movement of the plug and stem of the control
^'ah'p changes the area of opening of the port. The flow rate of the

Fio. 7-17. A singip-seat control valve and pneumatic actuator.

fluid passing through the port is therefore proportioned or throttled by

positioning the valve stem. The stem is in turn positioned by the
actuator.
The equation governing the flow of fluid through a restriction such
as a valve may be derived from the laws of fluid mechanics. For a con-
tiol A'alve, the flow rate of liquid is assumed to be given by

= 'KiaV2g(hi (7-5)

when^ /// flow rate, ft^/sec

K =
1 a flow coorticient
a = area of control valve port, ft^

g = acceleration due to gravity, ft/sec^

hi ~ upstream static head of flowing fluid, ft
/i 2 = downstream static head of flowing fluid, ft
Final Control Elements 213
The aiea of port opening of the control valve will be assumed propor-
tional to stem position so that

Kia = Kx (7-6)

where K= an overall coefficient, ftVft

X = stem position or lift, ft

Combining equations 7-5 and 7-6 to eliminate a,

* m= \K^/2g(hi - /i 2 )la; (7-7)

This equation illustrates that the flow rate m through the control valve
is directly proportional to lift;c if (1) the differential head (h\ - /12 ) is

constant and (2) the overall coefficient K is constant. Unfortunately,

these two conditions rarely prevail and it is necessary to study these
effects in detail.
A. control valve is installed in a pipeline in conjunction with the other

equipment necessary to employ the fluid flow for process control.

Heat exchangers, pumps, fans, pipe fittings, and hand valves are such
equipment.
A aeries resistance may result from the pipeline, orifices, hand valves,
heat exchangers, or other equipment installed in series with the valve.
Consider, as an example, the seriet resistance caused by pipe-friction.
Fig. 7-18. The head loss in a pipeline can be calculated from the fric-

tion formula

L
= (7-«)

where AHl ~ head loss in feet of flowing fluid

F = friction coefficient (from Moody diagram)
L = equivalent length of pipe including fittings, ft

D= inside diameter of pipeline, ft

V = velocity of flow, ft/sec

The head loss therefore depends upon flow rate. Equation 7-8 may be
written in terms of flow rate:

f-Vw
nhere m= Bow rate, ftVaec. Including head loss (iJii) in equation 7-7,

m= [KV2g{H„ - - Mi)lx (7-10)

214 Automatic Prooetts Control

Curve Valve Valve Line Line

Size, in Diff. Size, in Diff.

A 1 ?0 psi 2 1
JS 1 18 IH 3
c 1 16 IH 5
D 1 9 1 12

Fio. 7-18. Effect of line size for a given control valve.

Substituting equation 7-9 into equation 7-10 to eliminate (AAl), there

results

2g(Ho - H 2)
(7-11)
I 1 -f aix^

16FLK^ ^ line (AAl) at maximum flow

where ai
valve (AA) at maximum flow

Thus the flow is not directly proportional to the valve stem position
and the extent of nonlinearity depends upon line pressure difTerential.
 Final Control Elements 215
This equation is plotted in Fig. 7-18 for a 1-in. sise control valve installed
in dififerent size pipelines. The pressure source Ho and pressure tenninsp
tion H2 are considered constant, and the equivalent line length L is about
100 ft. The friction factor F and overall valve coefficient K are con-
sidered constant. The effect of the larger head loss caused by smaller
pipeline size is very noticeable at high flow rates. The smaller lines do
not allow the control valve to pass the desired maximum flow of 35 gpm.
In order to obtain the desired maximum flow rate, four steps may
be taken :

1. Decrease head loss so that differential head at the control valve

is greater.
2. Increase pressure source Hq.
3. Decrease pressure termination H2 .

4. Increase control- valve size.

In many cases the pipeline installation may be made so as to reduce

head losses by selecting larger sizes of lines or equipment. The pressure
source may sometimes be increased by selecting larger pumping equip-
ment. Usually it is not possible to change the pressure termination.
Lacking any of these alternatives, the control valve size must be in-
creased.
The effect of increasing valve size in order to obtain any desired
maximum flow may be calculated as follows: Rewriting equation 7-10
by dividing through by the maximum flow rate,

m ^ Ho — H2 — X
(7-12)
M“\ Hi - H2 X
where m= flow rate at any lift

M = maximum flow rate (at full open)

Ho — constant upstream pressure source
H2 = constant downstream pressure termination
Ml = total head loss for all series flow equipment
a; = lift

X = maximum lift (at full open)

Hi = valve upstream pressure (at full open)
Total head loss may be expressed in a similar manner. Rewriting
equation 7-9 by dividing through by the maximum flow r^te,

(7-13)
Ho -Hi
216 Automatic Process Control

Combining equations 7-12 and 7-13 to eliminate the head loss term and
solving for the flow rate,

Hi -H2 valve head differential at maximum flow

where a ==
Ho -//2 valve head differential at zero flow

This equation is plotted in Fig. 7-19. A control valve of sufficient size

is used so that the desired maximum flow is obtained in each case. Any
flow equipment in series with a control valve results in a change of the

ox

TIM
—a B fraction of total pressure
diff. across control valve
I
when wide open

Fraction of lift

^ Fixed resistance

Fig. 7-19, Control-valve operatirni with Hcnes resistance.

flow-lift characteristic of the control valve. It is quite apparent from

a study of Fig. 7-19 that a differential head at the control valve of less
than 25 per cent of total differential head causes an inability of the
control valve to throttle the fluid flow; that is, the valve effectively acts
as if it were only open or closed. The consequences of the “loss of
control” are discussed later in this section.
Final Control Elements 217

A cerUriJugdl pump or fan aowrce is often employed m fluid-flow sys-

tems and has a characteristic that the outlet head decreases with increas-

ijc r

Centrifugal
pump or fan

Fig. 7-20. ('Jontrol valve with a pump or fan source.

ing flow rate. This is termed the pump or fan head characteristic and
is illustrated in Fig. 7-20. 1'he head is approximately related to flow
at constant speed by

= 1-/3 (7-15)
Ho

where hi outlet head

Ho outlet head with outlet closed (zero flow)
a capacity constant
m outlet flow rate
M desired maximum flow rate

In general, the capacity coefficient p is about 0.2 to 0.4 for many centrif-
ugal pumps and fans.
218 Automatic Process Control
The flow through the control valve is

m jhi - H2 X
M “ \Ihi --H2X
(7-16)

where Hi = upstream head at maximum flow. Combining equations

7-15 and 7-16 to eliminate head hi and if downstream head
2 is H
constant,
m 1 X
M (7-17)

4a-h (1

where a
H -H21 valve head differential at maximum flo Av
Equa-
Ho - /fa valve head differential at zero flow

tion 7-17 exactly the same as equation 7-14 for the case of series
is

resistance. Therefore the graph of Fig. 7- 19 is also valid for a centrifugal

pump or fan source. If wide throttling of the valve is desired, the

E
3
E
X
19
E

0.2 0.4 0.6 0.8 1.0

Fraction of maximum lift, x/X

:3><:
'C Bypass hand valve
or orifice

Fio. 7-21. Control- valve operation with parallel resistance.

 Final Control Elements 219
number a above must be as close to one as possible. This requires that
thepump or fan characteristic should be as flat as possible or else that
the pump capacity should be as large as possible. Generally the proper
pump size is selected by means of an economic study. Many (tmes,
however an economic study indicates a pump too smdU to provide an
j

adequate differential head for the control vcdue.

The bypass or parallel resistance is shown in Fig. 7-21. Assuming
the differential head constant, the flow through the bypass is given by
t M 2 — K2'\/2g(Hi — H2 ) (7—18)

where M = flow rate through b3rpa8S

2 (assumed constant) ft^/sec
K2
= bypass flow coefficient
Hi = upstream head, ft
H2 = downstream head, ft
The flow through the control valve is given by

mi = [KV2g{Hi - H2)]x (7-19)

where mi = flow through control valve, ft®/sec

X = lift of control valve, fraction

Adding equations 7-18 and 7-19 for total flow

2= (7-20)

flow through bypass with control valve closed

where e =
M total flow with valve open

The effect of the bypass is simply to reduce the amount of control of

flow rate, and to alter the minimum flow as shown in Fig. 7-21.
The overall effect of either a series or parallel resistance, or 0 centrif-,

ugal pump or fan source, is to alter the flow-lift characteristic of a con-

trol valve. It is recalled that the proportional sensitivity of a propor-
tional controller is given by

(7-21)

where Ke = proportional sensitivity

m= manipulated variable (fluid-flow rate)
c = deviation
X = control-valve position or lift

Therefore, changes in proportional sensitivity result directly from

changes in slope of the control valve flow-lift characteristic. Thus, it is
220 Automatic Process Control

possible to alter the overall proporticmal sensitivity of the controller by

as much as 50 to by different arrangements of the control valve installa-
1

tion. 'I'his change of propoilional sensitivity may or may not be desir-

able in automatic control.
Summarizing control valve operation it is possible to state two oper-
ating principles for automatic control valves:

1. A fluid throttling control valve cannot operate with minimum differ-

ential head.
2. The differential head at a fluid-throttling control valve is never
arbitrary. It is a result of specifying flow rate, series and parallel
resistances, and pressure source and termination.

As a consequence of these principles it is imperative that the control

valve arrangement be selected with a great deal of care. The pressure
source should provide sufficiently high head so that adequate throttling
of the fluid flow is obtained in spite of head losses caused by series
resistances. A bypass resistance should be carefully and not indis-
criminately employed.

Example 7*5. A control valve is installed in the exit line of a heat e.\-

changer.The pressure at the inlet of the exchanger ffo is constant. The

pressure drop across the heat exchanger at maximum flow is 40 psi. The
control-valve downstream pressure is 20 psig and is constant. What control
valve pressure drop must be used and what upstream pressure at the heat
exchanger is required?
If the throughput at the heat exchanger changes over wide limits, good
throttling at the control valve i.s necessary. Consequently the control- valve
pressure drop should not be less than 50 per cent of total. Thus, 40 psi
differential pressure at the valve is desired and the upstream pressure at the
heat exchanger must be 100 psig at maximum flow.

Example 7*6. A centrifugal pump is available as a fluid source for a heat

exchanger. At maximum flow the head is 100 psig, and at shut-off the head
is 140 psig. At maximum flow, line losses are 15 psi and the heat exchanger
pressure drop is 75 psi. The control-valve back pressure is atmospheric.
Can this pump be used?
Referring to equation 7-14,

100 ~ 15 - 75 10
a = 0.071
lio 140

Thus the is only 7 per cent at maximum

control-valve pressuit; differential
flow. This pump
should not be used, because wuth such a small pressure
differential the valve will not throttle the flow' adequately.

Sliding-Stem Control Valves

Control valves in which the plug is operated by means of reciprocating
motion are termed sliding-stem valves and are of the following types:
Final Control Elements 221

1. Single-seat plug valves.

2. Double-seat plug valves.
3. Lifting-gate valves.

The single-seat and double-seat control valves are shown in Fig. 7-22.
The single-seat valve has only one port opening between seat and plug
and the entire flow passes through this port. It has the following

Double -seat Single-seat

control valve control valve

Fio. 7-22. Sliding-plug control valve.

features: (a) It is (6) It can be shut off to pro-

simple in construction.
vide zero flow, There is a large force acting on the valve stem
(c)

because of the differential head acting across the port and seat area.
The double-seat valve has two port openings and two seats and two
plugs. The port openings are not usually identical in size. This type has
the following features: (a) Net force acting on the valve stem is gener-
ally small (therefore ^‘pressure balanced”), (b) It cannot be shut off
tightly because of differential temperature expansion of valve plug and
valve body.
A few types of plugs for single-seat and double-seat valves are shown
in Figs. 7-23 and 7-24. The piston type plug has one or more grooves

Piston type V-port type Parabolic, ratio, Poppat type

or throttle type

Fig. 7-23. A few types of single-seat valve plugs.

222 Automatic Process Control
along its length and the flow passes vertically in the grooves between the
plug and seat ring. The V-port t3rpe plug is open on the inside and the
flow passes horizontally through the triangular shaped area over the

V-port type Parabolic, ratio,

or throttle type

Fig. 7-24. Two typce of double-seat valve plugs.

seat ring. The parabolic plug presents an annular area to flow between
the plug and seat ring. The poppet-type plug offers a cylindrical-shaped
flow area and
used with small total lift.
is

The gate valve in Fig, 7-25 is often used for fluids containing solid
matter, because it presents an open area directly to the flow of fluid

and does not involve a change of direction of flow stream. A gate

valve can usually be shut-off tightly by wedging into the seat. The chop-
ping action at shut-off is very useful for stringy materials such as
paper pulp.
 :

Final Control Elements 223

The weir valve in Fig. 7-25 is particularly suited to certain chemical

fluids, because has a smooth contour inside the body with no ''pockets”
it

for solid matter, and because it has no packing gland around the stem.
The flexible diaphragm of rubber or other nonmetallic material is
positioned by the plunger and stem. Fluid pressure inside the valve
body holds the diaphragm smoothly against the plunger.

A Decreasing sensitivity type

B Linear type
C Equal-percentage type

Fig. 7-26. Control-valve “characteristics.”

The flow-iift characteristics of sliding-stem valves generally fall into

three approximate categories as shown in Fig. 7-26. These are

A. Decreasing sensitivity type. The valve sensitivity {Am/ Ax) at

any flow decreases with increasing flow. The maximum port area is
sufficiently large that pressure losses elsewhere than at the valve port
may restrict the maximum flow.
B. Linear type. The valve sensitivity {Am/Ax) is more or less
constant at any flow.
C. Increasing sensitivity type. This is termed the equal-percentage,
logarithmic, parabolic, or “characterized” type of valve because the
flow-lift curve plotted on semilogarithmic coordinates is approximately
a straight line. The valve sensitivity {Am/ Ax) increases with increasing
flow rate. The valve sensitivity at any given flow rate is a constant
percentage of the given flow rate, thus the term equal-percentage.
224 Automatic ProoesaJ^ontrol
Actual flow characteristics of a number of control valves, all of
2^in. size, are shown in Fig. 7-27. These are given for a constant

Fraction of maximum lift, xjX

1.0

2)^- in Size
double- seat valve

0.8

t 0.4

0 02 04 06 "08 10
Fraction of maximum lift, xjX

Fig. J-27. Actual flow-lift characteristics of sliding-stem control valves.

A. Weir (Fig 7-25) E. V-port or throttle plug (Figs. 7-23, 7-24)
B. Poppet (Fig. 7-23) F. Hatio plug (Fig. 7-23)
C. V-port (Figs. 7-23, 7-24)

differentialhead across the control valve. The flow characteristic is

further altered inany actual installation by any series resistance, pump
source, or bypass resistance, as discussed previously.
Final Control Elements 225

The rangeability of a control valve is the ratio of maximum control-

lable flow to minimum controllable flow;

m (maximum controllable)
(7-22)
m (minimum controllable)
where R = rangeability number. Turndown is a similar concept based
on the ratio of normal maximum flow to minimum controllable flow.
m (normal maximum)
T = (7-23)
m (minimum controllable)

Normal maximum flow is generally taken as 70 per cent of maximum

flow so that
(7-24)

The minimum controllable flow of a control valve depends upon its con-
struction. Clearances must be allowed in order to prevent binding and
sticking, and the flow through these clearances constitutes the minimum
controllable flow. The minimum controllable flow for a single-seat
valve is not zero unless the throttling seat and shut-off seat are identical
and have perfect alignment. The rangeability of a sliding-stem control
valve is usually between 20 and 70.
The importance of rangeability and turndown lies in the application
of the control valve. For example, if the design of an oil burner and
furnace requires a 30 to 1 range of oil flow to accommodate various loads
on the furnace, the turndown must be at least 30 and the rangeability
must be at least 43.
Example 7-7. Flow through a linear valve (constant sensitivity type
characteristic) is given by
m
M
where m is the flow at any lift x, M
is the maximum flow at maximum lift

X, and R is the rangeability. If the valve passes 10 gpm at a maximum lift

of 2.0 in. and the rangeability is 20, compute the valve sensitivity.
From the above equation
dm /R - \\M 19 10 gpm
4.75 gpm/in.
dx~\ R /X"20^ 2in.

Example 7-8. Flow through an equal-percentage valve (increasing sensi-

tivity type characteristic) is given by

^ = RMxy-u
M
where the symbols are the same as in Example 7-7. Derive an expression for
valve sensitivity.
226 Automatic Procees Control

Butterfly valve Louver

Fig. 7-28. Several rotating-ehaft valves.

Rotating-Shaft G>ntrol Valves

Control valves in which the restriction is accomplished by the rota-
tion of a plug or vane may be called rotating-shaft type. These are:

1. Rotating-plug valves
2. Butterfly valves
3. Louvers
Final Control Elements 227

The raUUing^iig vake is illustrated in Fig. 7-28. The plug is a

cylindrical or conical element with a transverse opening. It is rotated
in the valve body by an external lever so that the opening on one side

Fraction of maximum rotation, x/X

Fio. 7-29. Actual flow-angle characteristica of rotary-shaft control valves.

of the plug is The shape of the opening

gradually covered or uncovered.
or port may be circular, V-shape, rectangular, or any form that is
desired to produce a given flow-angle characteristic. The flow-angle
characteristic is shown in Fig. 7-29 for a near-rectangular port shape.
228 Automatic Process Control

A rotatinf-plug valve having a conical plug can generally be closed

tightly and has high rangeability. This type of valve is often employed
for throttling the flow of oil to burner systems.
The huUerfly valves shown in Fig. 7-28, consists of a single vane rotat-
ing inside a circular or rectangular pipe or casing. The shaft projects
through the casing and may be operated externally. rotation The total
of the vane is usually restricted to about 60 degrees, because the addi-
tional 30 deg does not produce much further increase in flow. The
V-port butterfly valve incorporates a V-slot in the body so that rotation
of the vane opens a portion of the V-slot. The flow-angle characteristic
is shown in Fig. 7-29 for a 60 deg butterfly valve. The rangeability
may vary from 5 to 50 and tight shut-off may be obtained with special
design. The butterfly valve is most often employed in sizes from 4 to
60 in. for the control of air and gas. It is also used for liquid flow if
the pressure differential is not large.
The UniveTj Fig. 7-28, consists of two or more rectangular vanes
mounted on above the other and interconnected so as to
shafts one
rotate together. The vanes are operated by an external lever. In the
unirotational louver the vanes remain parallel at all positions. In a
counterrotational louver alternate vanes rotate in an opposite direction.
Flow guides are sometimes installed between adjacent vanes in order to
improve the effectiveness of throttling. The flow-angle characteristic is
shown in Fig. 7-29 for a 90 deg unirotational louver. It may be seen
that the sensitivity is very high at raidflow and that the last 30 deg of
rotation is relatively ineffective. A louver cannot provide tight shut-off
because of the long length of seating surfaces. Louvers are used exclu-
sively for control of air flow (draft) at low pressure.

Control* Valve Sizing

The proper size of control valve is important because of the effect on

the operation of the automatic controller. If the control valve is over-
size, for example, the valve must operate at low lift and the minimum
controllable flow is too large. In addition, the lower part of the flow-lift
characteristic is most likely to be nonuniform in shape. On the other
hand if the control valve is undersize, the maximum flow desired for
operation of a process may not be provided.
The flow of a liquid through a fully open control valve is assumed
to be given by

m - Kayj^^ (7-26)
 Final Control Elements 229

where m » flow rate, ft^/sec

IT B a flow coeflScient
a B area of port opening, ft’
g B acceleration due to gravity, ft/sec’
AP = pressure differential, Ib/ft’
y B fluid density, Ib/ft’

The flow coefficient K and port area a are different for every style or
size of control valve. Consequently, it is standard practice to combine
certain terms of the above equation into a single number Cv, termed the
size coefficient:

Wl (7-26)

where niL = liquid flow rate in gallons per minute at the conditions for
which specific gravity G is taken
Cv = size coefficient
AP = pressure drop, Ib/in.’
GB specific gravity of liquid (referred to water) at either flow-
ing or standard conditions

The size coefficient Cv is defined as the flow rate of water in gallons

per minute provided by a pressure differential of 1.0 lb per sq fn. through
a fully open control valve. The size coefficient for any control valve
must be determined by actual test. The size coefficient for a control
valve of the sliding stem and plug type is very approximately equal to
the square of the nominal valve size multiplied by ten.
For the flow of gas or steam, the size coefficient with suitable con-
version factors is also employed. The form of equation is an approxima-
tion to the complete isentropic flow equation, known as Fliegner’s
formula when used for air'

»«. = 760C,n^^ (7-27)

where =nig flow rate of gas, Ib/hr

BCp size coefficient (obtained for liquids)

Pi = upstream absolute pressure, Ib/in.’

AP = pressure differential, Ib/in.’
P B gas constant for flowing gas
Ti = upstream absolute temperature

‘Kiefer, x^ J., and M. C. Stuart, Principles of Engineering Thermodynamics,

2iid ed., John Wiley and Som Inc., New York, 1264.
230 Automatic Process Control

The caiculatioD of flow rate of fluid throu^ a control valve therefore

depends directly upon the experimentally determined size coeflicient C„.
As might be expected, the size coefficient is not a constant and depends
upon both flow rate and fluid viscosity. In other words, the size coeffi-
cient is some function of a number corresponding to Re3molds number.
The extent of this relation is not fully known.
The calculation procedure for flow of gas through a control valve
must be modified when the pressure differential is greater than a critical
value. When the downstream pressure at a control valve is sufficiently
small, the flow rate does not depend upon downstream pressure. At the
critical pressure ratio, or belcv, the gas attains the velocity of sound

through the valve ports. The critical pressure ratio for air and steam are

Pt/Pi
Air (70 F) 0.628
Steam (212 F) 0.539

The critical pressure ratio does not vary greatly for all common gases
and a single flow formula may be employed with reasonable accuracy^

to = 158,000 (7-28)
V K I

where w= flow rate, Ib/hr

Cv — size coefficient
*
Pi = upstream absolute pressure, Ib/in
Vi = upstream specific volume, ft®/lb

Determination of control-valve size, flow rate, or pressure differential

is made in most industrial applications by the use of a nomograph or
a slide rule. A typical nomograph is illustrated in Fig. 7-30. Generally,
these nomographs or slide rules are based upon experimentally deter-
mined size coefficients and the equations given above, or upon experi-
mentally determined empirical relations. It is important to note that
control valves are not manufactured in an infinite number of sizes but
are generally available in standard pipe sizes only.
The procedure for selection of control valve size may be as follows:

1. For the particular process under control, determine,

(a) The maximum valucf. of flow rate required to sustain the con-
trolled variable under any condition of process operation. This is the
normal maximum flow rate.

(h) The value of flow rate that will be required most of the time.
This is the normal flow raie.
*
See Kent, R. T., Mechanical Engtneera’ Handbook^ 12th ed., John Wil^ and Sou
Inc., New York, 19W.
Final Control Elements 231

from

data

actual

Obtain

only.

estimation

(For

manufacturer.).

valvn*.

control

fixing

for

Nomogr^h

7-30.

Fic.
232 Automatic Process Control
(c) The minimum value of flow rate required to sustain the controlled
A’ariable under any conditions of process operation. This is the normal
minimum flow rate,

2. maximum flow rate which the control valve is to provide.

Select the
This is maximum flow rate of about 70 per
generally based on the normal
cent of maximum flow rate. The additional flow is a factor of safety
which allows for low estimation of pressure losses and high estima-
tion of valve flow rate. The maximum flow rate usually selected is

about normal maximum flow.

1.4 times the •

3. Select the style and type of control valve to provide best operation

for the fluid to be handled. Check the rangeability to insure that the
minimum controllable flow is generously smaller than the normal
minimum flow rate desired.
4. Calculate the pressure differential at the control valve at the
maximum flow. This requires calculating line pressure losses, flow
equipment pressure losses, and determining upstream head from pump
or fan characteristic curves.
5. Determine the control-valve size from manufacturers nomographs,
charts, or slide-rules. If the fluid viscosity is high, or line velocities are
exceptionally low (Reynolds number in the pipe line less than about
10,000) the size coefficient C, will be low, and the manufacturer should
be asked to determine the valve size.

6. Calculate the characteristic coefficient:

Differential pressure with control valve wide open

(7-29)
Differential pressure with control valve closed

If the number is relatively small (less than 0.10) or if

1
“
/ Y (7-30)
2500 Wmi J
where M = maximum flow rate
Mnku = normal minimum flow rate

then the control valve will, at times, be required to operate at less than
2 per cent of total lift. If this occurs, the control valve would require
accurate positioning by a powerful actuator, and the valve ports and
plug must be precisely manufactured. Otherwise it is best to inspect the
control-valve installation to determine whether the differential head can
be increased by selecting larger piunps, larger line sizes, or minimizing
losses in series flow equipment.

Example 7-9. A fully open valve passes 100 gpm of water at

2-in. sise
a pressure differential of 7.0 psi. Calculate the C, coefficient.
Final Control Elements 233
From equation 7-26

Note that the size squared (4) multiplied by 10 equals 40.

Example
1.
7-10. A heating furnace requires a control valve passing 10 gpm
preheated light fuel oil (sp. gr. » 0.8) at full load and only 0.2 gpm at the
emallest heating load. The pressure differential at wide o))en is 20 psi. The
source
2. pressure is constant at 50 psi gage, but there is 10 psi drop in the oil
preheater and 20 psi drop at the furnace burner nozzles. Detenninc the
valve size.

Select a maximum flow rate of

14 gpm

The rangeability must be

6.
or the turndown is 50.

3. A rangeability of 70 is quite largo and requires careful selection of valve

plug style. Probably a rotating plug valve or possibly a single-seat plug
valve would be satisfactory.
4. The pressure differential is 20 psi.
5. From Fig. 7-30,

g m. size passes 6 gpm (too small)

i in. size passes 12 gpm
f in. size passes 31 gpm (too large)

The ’}-iu. valve is selected because it is the nearest size passing the desired
14 gpm.
The characteristic coefficient is (equation 7-29)

20
“ 50

so that the control-valve differential is 40 per cent of total differential and the
characteristic is satisfactory.

Throllling Electrical Energy

Many heating processes employ electric energy instead of energy
contained in fluids. In order to control such processes, electrical energy
must be modulated in some way. There are two somewhat different
methods of modulating electric energy:

1. Regulate the power on or off with varying periods of on and off

times. This is accomplished by a two-position controller or by an
average-position controller. The final control element is an electric
power relay.
234 Automatic Process Control

2. Regulate the voltage to the load by automatically adjusting the

voltage. The final-control element may be an adjustable rheostat,
variable transformer, controlled thyratron circuit, or a saturable-core
reactor.

Electric relays for power loads must have appropriate contact life

because they are required to be turned on and off frequently, especially

in two-position control.

Series rheostat

Variable transformer Saturable reactor

Fio. 7-31. Control of electric heating.

A series rheostat as in Fig. 7-31 may be used for heating elements.

If the voltage supply E is constant, then the load power is given by

*1 - (7-31)
Af Rl +r
where m« power to load, watts
M s maximum power watts
« resistance of load, ohms
r « series control resistance, ohms
 ^

Filial Control Elements 235

The control resistance or rheostat is adjusted by a motor actuator,
usually pneumatic or electric,

Re - r ^ Rctni (7-32)

where Re — total resistance of series rheostat, ohms

mi s input signal to motor operator

Combining equations 7-31 and 7-32 to eliminate r, the relationship of

m aq^ operator input mi is
heating energy

m
(7-33)
M ,^Rc Rc
1
+ icl Kl
This equation is plotted in Fig. 7-32. The rangeability of this method
depends upon the value of resistance of the control rheostat,

R= 1 + (7-34)

Where R — rangeability. However, for rangeability greater than 6, the

power characteristic becomes too steep. Consequently series rheostat
control should be used only for low rangeability. In addition, a control
rheostat of high resistance and high wattage is likely to be very bulky
and expensive.
A variable transformer as in Fig. 7-31 is widely employed for control
of electric heating. The power characteristic can be shown to be

(7-35)

if the output voltage of the transformer is directly proportional to

motor operator position. The characteristic is plotted in Fig. 7-32 and
is parabolic in form. A parabolic characteristic is almost identical to an
equal-percentage or logarithmic characteristic.
The saturable reactor of Fig. 7-31 is another
form of control for electric
power. Alternating voltage E is supplied to the reactor,
and the alter-
nating voltage across the load depends upon the direct-current voltage on
the control winding. The control voltage may be supplied or may be
obtained directly as the output signal of an electric automatic controller.
The reactor operates by regulation of the flux passing from the primary
(input) winding to the secondary (output) winding. "
The flux regula-

tion is accomplished by a controlled degree of saturation provided by

the direct-current control winding.
236 Automatic Process Control

Example 7-11. A 1500-watt resistance heater for 110 volts is to be regu-

lated by a series rheostat. Foi* a rangeability of 5, calculate the resistance
and power rating of the series rheostat.
The resistance of the heater is

12,100
Rl = 8.1 ohms
M 1500
From equation 7~34

Rc * Rl(R - 1) 8.1 X (5 - 1) = 32.4 ohms

Final Control Elements 387

It will probably be DeceBsary to use a atandard 3&-ohm rheoitat giviiig a raii|e-

ability of 5.32. At minimum setting the power abeorbed in the control
rheostat is

E% 12,100 X 36
2300 watts
+ (36 + 8.1)*

PROBLEMS

7-1. Calculate th^instantaneous horsepower required to oscillate a weight

of one pound over l-in. amplitude sine wave at 100 cps. Develop the gen*
eral formula.
7-2. A pneumatic actuator requires Q ctm of free air at a pressure P psig.
Calculate the horsepower required to run a motor-compressor of efficiency £.
7-3. A hydraulic actuator requires Q gpm at pressure P psig continuously.
Calculate the horsepower of the pump-motor of efficiency E.
7-4. A valve actuator is to stroke once each minute, 24 hours per day,
and must run five years. How many cycles represent minimum life?
7-5. A reversible electric motor has a stall torque of 30 os-in. and a no-load
speed of 3200 rpm. The torque-speed curve is a straight line. The output
drives through a gear train. Calculate the maximum output force at a speed
of 2 in. per sec if the gear train is 40 per cent efficient.

7-6. Calculate the time constant of a spring and diaphragm actuator

operated through a pneumatic line of resistance R. {C = capacitance of top
works, A = diaphragm area, K^ spring constant).
7-7. If a control valve has a parabolic plug and has 60 per cent of the
pressure differential across it at fall flow, show that the flow versus position

is nearly linear from 20 per cent flow up to maximum. (Hint; plot

1.41g ^ \

7-8. The are^ of opening of a valve versus lift is given by A = a + bxK

Derive the flow versus lift characteristic for this parabolic valve.
7-9. Liquid glycerin (density 78 Ib/ft’) is required at a maximum flow of

18 gpm. The line drop is 40 psi and the valve pressure drop is 10 psi. What
size plug valve is required?

7-10. Water flow is controlled by a 3.0-in. butterfly valve with a 10 psi

differencial. What is the maximum flow rate?
 vhapU^r

Process
Instrumentation

The results of automatic control must always be evaluated in terms of

the quality of the finished product rather than in terms of accuracy or
deviation of the controlled variable. The general purpose of automatic
control is to obtain maximum efficiency of process operation.
Many processes are comparatively simple and possess a certain degree
of inherent stability. On the other hand, some processes possess a con-
fusing array of capacities, lags, and load changes. Between these two
extremes lie a large number of processes, complex to a varying degree,
and requiring careful analysis.
As a preliminary step, this analysis should include a separation of
the variables to be controlled. Each controlled process can then be
studied individually from the standpoint of capacity, lags, and load
changes. Until this analysis has been made, there is little likelihood
that the process can be successfully controUed.
The control system should be properly engineered, special care being
taken in the selection of primaiy and final elements. In too many
applications the controller is handicapped by being required to overcome
such factors as a large measuring lag or a large dead zone in the con-
trol valve.

The general block diagram is seen in Fig. 8-1 which includes a feed-
back element as well as an input element. The operational equations
for the system are

11 input element (8-1)

il 1 signal element (8-2)

m= G\t controller (8-3)

238
Process Instrumentation 239

c =» G2TO + Nu process or qrstem (8-*)

He feedback element (8-«)

d « 1; — c deviation (8-6)

A new variable (d = deviation) must be defined in order to distinguish

the actual difference d between controlled variable and set point from
the apparent difference or actuating signal e.

Fio. 8^1. Automatic control with input and feedback elements.

The feedback variable b is the quantity indicated or recorded by the

measuring element H. The controlled variable is never actually recorded

anywhere and its magnitude must be inferred from the feedback vari-
able b. Therefore it is of interest to compare the controlled variable
in the general equations:

AG G2
1 N
(8-7)
1 + 0 ,(72^ 1 +G G H'*
1 2

to the feedback or measured variable given by the equation

AG1G2H NH
8-8 )
, ,
5 -S ff -1- u (
I -h G G2 H
1 1 + GiG2ff
The actual deviation is given by

1 -h GiG2(ff - A) N (8-9)
1 + G1G2H 1 G1G2//

For load changes, the change of set point is taken as zero and the part
of the deviation due to load may be calculated. For set point changes,
the change of load is taken as zero and the part of the deviation due to the
changing set point may be detennined. Notice that t^ese equations
reduce to those of Chapter 4 when the input and feedback elements
contain no time elements {A — H = 1.0).
240 Automatic Process Control
Example 8-1. Derive the general equation for actuating signal.
From equations S-1 through 8-9
A NH
”
* 1 + GiOtH 1 + GiOtH “
The actuating signal differs from deviation (see equation 8-9).

Effect of Measurement (Feedback) Lag

The effect of a lag in the measuring or feedback means is insidious;
that is, the effect of the lag tends to become obscured by its own exist-
ence. The effect of a large measuring lag is to almost always cause large
amplitude oscillations and slow return or stabilization. On the other
hand a large measuring lag causes great attenuation, so that the large
amplitude oscillations are not seen.
The process of Fig. 8-1 is typical of many industrial processes such
as tank heaters, heat exchangers, and furnaces. For example, consider
proportional control of a time-constant process:

Cl = Kc proportional control with negligible lag

R
U2 - process time constant Tp
Tp-d- 1

R
N = effect of load
T^+l
1
H= measuring time-element constant Tm
1

The actual response of the controlled variable to a load change is given by

^ RiTmS + 1)
“ (8-10)
+ (fp + T^)s + (RKc + 1)

The indicated response to a load change is given by

^ R - 11 )
.
“ (8
+ (Tp + T„)b + {RK, + 1)

The damping ratio will be set to approximately one third by selecting

a value of proportional sensitivity Kc. The solutions to each of these
equations to a sudden change of load u are plotted in Fig. 8-2 for a pro-
cess time constant (Tp) of 20 sec and two different measuring element
time constants.
First compare the values of the feedback variable b and the controlled
variable c. As the curves of Fig. 8-2 indicate, both have the same gen-
eral characteristics, but the feedback variable shows less change and is
Process Instrumeiitation 241

reUrded from the actual controlled variable. In fact, the feedback

variable lags the controlled variaUe by about 45 degrees phase. It is
well to remember that the controlled variable curve c is not usually
indicated or recorded anywhere in the system. Therefore, changes of
the actual controlled variable are always larger in magnitude than those
indicated by the feedback variable.

Time Time

Fig. S-2. Effect of measuring lag.

Second, compare the changes in controlled variable when the measur-

ing lag is 2 sec to the case when the measuring lag is 10 sec. In the
latter case the automatic controller is much less effective. This is
indicated by the smaller value of proportional sensitivity when the
measuring lag is larger. In consequence the offset is increased by about
170 per cent by the larger measuring lag, and in addition the control is
made much slower acting. Therefore, an increased measuring lag ma}
result in larger offset and slower recovery from a load change.
In temperature control, the measuring-element lag is likely to be the
largest lag except for the process lag and it is desirable that the measuring
lag be made as small as possible compared to the largest lag, a ratio
of one tenth is a good figure to maintain.
In liquid-level control, the process is often of a single energy storage
type, and the magnitude of measuring lag is generally not important.
If any choice is possible, a measuring means should be selected that has

the smallest measuring lag; that is, a measuring lag that is as small as
possible compared to the largest time constant of the process.
In pressure control, the process is often of a single energy storage
type. The magnitude of measuring lag is generally small except when
long pressure connecting lines must be used. In this ca&e also, the
measuring lag should be maintained less than a small fraction of the
largest time constant of the process
242 Automatic Process Control

In flow control, the lags of the measuring means and the controlling
means are generally the only important lags in the ^stem. For the
flow-control system of Fig. it is assumed that the measuring means

Fio. Flow control with measuring lag.

such as orifice head-meter, has a single time-constant measuring lag.

The lag of the controlling means is neglected in this first example. The
system functions are
II control valve sensitivity

II downstream pressure sensitivity

II upstream pressure sensitivity

integral control

H=— measuring-element time constant

TnS -f 1

The system equation for a change of load U2 with load uo and set ]K>int v

fixed is
K2T8(T^8 -f 1)
(8-12)
rr*8* + Ts + X,
The feedback variable is given by

-KjTs
(8-13)
TT„^ + r* + X,
For a sudden (step) change of load variable U 2 the solutions ,
of each of
these equations are plotted in Fig. 8-4. In each case the damping ratio
is selected as one third (therefore ^KiT^ = 9T).
Comparing the response with large measuring lag to the response
with small measuring lag it is seen that the effect is to delay and attenu-
ate the indication of changes in the controlled variable. However, a
most serious consequence of the measuring lag in flow control is that
 Process Instrumentation 243
an inspection of the indicated value of the controlled variable b might
lead to the conclusion that the measuring lag of 20 sec produces a better
control than if the measuring lag Is 5 sec (compare the two solid curves
of Fig. 8-4), whereas actually the measuring lag has made the control
much less effective (compare the two dotted curves).

Time Tima

Fio. 8-4. Effect of measuring lag in flow control.

In flow control it often happens that the measuring lag and controller
lag are of approximately the same magnitude, and both are larger than
any other lags of the system. In this case, the system functions of
Fig. 8-3 are

O2 = Ki valve-flow sensitivitv

N2 = K 2 downstream pressure sensitivity

Nq = Kq upstream pressure sensitivity

Kc
Oi proportional control with controller (or valve)
TcS-\- 1
time constant Te

measuring element time constant

The system operational equations, relating change of controlled vari-

able to a change in load U2 but with
j
set point v and load t^o fixed, is

- = + Dcr,* + 1)
+ iT„ + T,)s + (KiK, + 1)

For the feedback variable

-Ki(T^ + 1)
(8-16)
TcT„s^ + (r« + r.)« + (KtK, + 1)
 1

244 Automatic Process Control

The solutions for a step change in load Uf are given in Fig. S-li for
three cases; measuring lag greater than Tm » 10, equal to Tm^ 5.5,
and less than Tm = 3, the controller lag. The damping ratio is main-
tained at one third, and in addition a ratio of lags has been selected that
results in the same period of oscillation (T^Tc = 30). Thus in each of

Fig. 8-5. MeasuriDg lag and controller lag in flow control.

the three cases the actual controlled variable c has precisely the same
behavior, i.e., the same damping ratio and same period of oscillation.
The upper curve in the right-hand diagram has been displaced upward
for better comparison.
It is apparent that, whereas the actual change of controlled variable
is the same in all three cases, the indicated variable shows less change as

the measuring lag Tm becomes larger. On the other hand, it is not valid
to conclude that the controller lag should be large because if the con-
troller lag is not the largest in the system, increasing it does not mate-
rially improve the control of the actual controlled variable.

Example 8-2. A flow-control system has a valve lag of 5 sec and a meas-
uring time constant of 7 sec, and proportional control is used. Would doubling
the measuring lag improve the flow control?
From equation 8-14, the damping ratio will be made 1.0 so
(T^ + T^r = miKc + DTcTm
Old KiKc - 0.029

New KiKc = 0.29

The characteristic time is

r.r«
KiK, +
Old T = 5.8 sec

New T = 7.4 sec

 Process iDSirumentation 245
No. The deviation ie the same and the atabilixation ia alower. The record
would look better, however.

ProceM Duturbancee
A number of factors affect the quality of control, the most important
of which is the behavior of the process load. The process load may be
defined in terms of the setting of the final control element to
the controlled variable at the set point.
The location and source of process loads are illustrated in Fig. 8-6 by
the gas-fired continuofis heating furnace. The heat losses are made up
by the stack U4, radiation us, and work c. The heat supply is made up
by work U3 and fuel m.

A load change is caused by variations in the rate of energy supply or

loss due to causes other than controller action. For example, if the gas
pressure Ui changes, the flow of fuel will be altered. Another example
is the clogging of the burners which decreases the flow of fuel for any

one valve setting. For another example, the heat may be changed by
varying the temperature of the work as it enters the furnace. The most
conunon load change, however, is caused by variations in the rate at
which the work flows through the process. Other load changes are
those due to stack loss 1/4, caused by changing draft pressure, and radi-
ation or leakage loss U5, due to varying ambient conditions. The effects
of all these load changes must also be corrected by the controller.
The location of the load change has a great influence in the magnitude
of the resulting change in the controlled variable. This is illustrated
by the curves of Fig. 8-7 The curve on the left represents the result
.

of a unit change of variable U2 (temperature of the incoming work) in

Fig. 8-6. When the load change occurs “close* to the controlled vari-
^

able, the magnitude is large compared to the magnitude when the load
change is located far from the controlled variable. The curve on the
right shows the result of a unit change in variable uy (pressure of the
246 Automatic Process Control
fuel gas) in fig. 8-6. Notice that the magnitude of the change is about
two-thirds less.

The offset due to a step change of process load is easily calculated

from the operational equation. Under steady-state conditions, all
derivatives are zero, and the differential equation reduces to an algebraic

Fxa. 8~7. Location of load change.

equation. Derivatives are zero when the operator s is zero, and the
offset for a unit step change of load is given by

Dq = lim -1 (8-16)
•-»o [1 + GiGiH]
Substituting Gi = XeCl + T8)/Ta for proportional-integral control in
equation 8-16 there results

-
Do^liin (8-17)
® ^oT$ + G»H(Ta+l)
where T= integral time. The limit is zero and the offset is therefore

zero if the function G2 H/N does not contain any factors where n > 0.

Elements of the capacitance type in the process function (73 may

make the use of integral control unnecessary. Rewriting equation 8-16
for proportional control Gi = Kc, and with G2 = 1/Us, the result is

Do = lim (^18)
•-*0

Thus there is no offset when the function H/N does not contain any
8^ factors.

Eiample 8-^. For a process with two time lags,

5 ^ ^
" (Tib DTfi' Tis +l
 I

Process Instrumentation 247

and a controller time constant la^ Te, calculate the offset for proportional
control.
From equation 8-16 and a change of load u,

Do » lim
t-HO KcR
+ IHTia l)T^
Do = 0
A disturbance of the set point or input may be required in many
process-control systeips as when a process variable is required to follow
a program, or when the system is a part of another system. The devia-
tion resulting from a disturbance of the set point; i.e., when the set
point is a function of time, is given by equation 8-9,

1 + GiGjjH - A)
(8-19)
1 O1G2H
The character of the input element A affects the performance of the
system. must be noted that the input element does not influence
First it

the stability, because this stability is governed by the closed-loop func-

tion G 1 G2 H. Element A is not in this loop. The input element does,
however, affect the overall response.
The effect of the input element may be seen more clearly by equar
tion &-7,
O G2
8-20 )
1
c {A)v (
1 -|" G G2H
1

so that the system be considered as two systems in series. When

may
the input element a simple lag like a time-constant element, the system
is

response is appropriately delayed and attenuated in the normal fashion.

If the input element contains a derivative term for example (T$ + 1),

it is possible to improve the performance of the system. Suppose

A ^T8 +
G\ “ 1/T|8

Gz = l/iTi8 + 1)

H= 1.0

Af = 0

The ^stem function becomes

T8+ I
’ (8-211
fiTi^ -f Ti8 -f 1
 1 1

248 Automatic Process Control

Setting the integral time Ti to give critioal damping {Ti 4ri) there
lesulte
r« +
(g-22)
(Tae+l)*"

where T2 = 2Ti. Without input element lead {T * 0), the system

has second-order response. If the input element lead can be made equal
to the process time constant {T = 27i), a first order system results and
the overall response could be made much faster.
Static error may result from a changing input dr set point. For a unit
step change of the set point, the static error is given by

1 -h - A)
Ds = lim (^23)
•-«o 1 + G1G2H
This error is usually zero in process control.

Example 8-4« The static error caused by a unit step change of input may
be written
— G162A 1
Ds = lim Tl
•-»o L 1 + OiOtHJi
For the following functions calculate the static error:

Ke
QxGt ^ = +
WfW’

Ds = lim r.1
•
— +
•-*0
I ^ (r« + i)« 1

Kc TffS

Dfi = 1 -
^
1-4-1

Magnitude of Process Load

Often a controller is hajidicapped by being called upon to operate
under conditions of large variations in process load. The resultant effect
upon the quality of control may
be such that a controller satisfactorily
operating with high process load may produce entirely undesirable
results at low process load. This effect is brought about by chang-
ing values of system parameters when the various parts of the system
operate at different steady-state values of the pertinent variaUeB. If
Process Instnimentation 249
these parameters are inside the loop, it may be possible to compensate
the effect of the variations.
The steady-ertate sensitivities of the qrstem are usually:

Controller K, = —
Ax
Ae

Am
“ “7“
Control element
Ax
•
Ac
Process
Am

II
Feedback

If it is assumed that under operating conditions the loop gain KgKJCpKm

is to be so proportioned that the control action has the same behavior
at all loads, then it is possible to derive a control element function from

dm _ g(m)
(^24 )
lb KcKpKn,
or
KcKpK^
X dm (8-25)
g{m)

where g{rn) is the loop sensitivity as a function of load.

The effect of magnitude of process load will be illustrated by several

examples of typical processes. Consider the liquid-l^vel control of
Fig. 8-8. Generally the measurement and controller lags are very
miLlI compared to the process li^. The steady-state sensitivities
 -

Automatic Process C!ontrol

(gains) are

Controller

Control valve Kp

Process jR = —
Am
as shown in the operational diagram of Fig. 8-8. The system opera-
tional equation is

KcKJt(Ti8+l)
''
TMTa + \) + KcKMTi8 + l)^
RTi8
“ ^ '
TMT» + 1) + KJCMTiS + 1)

where Ti » integral time

T— process time constant « RC
C= process capacitance

For a damping ratio of one third, the relationship among the parameters

(1 + KcKJi)^ 4T
(8-27)
KcKJl “ 9Ti

Since the loop static gain KcKJi is usually at least ten, we may write

KcKJt^^

KcK,^- (8-29)

The control-valve relation can now be derived:

„ dm 4C
(8-30)
dx 9KcTi

where ICe is a constant, providing the automatic controller has linear

proportional action. Integrating gives

adx *= ax (8-31)
 Process Instrumentation 251
The a linear relation between flow and position at the
latter requires
control valve, as shown in Fig. 8-10. If equation 8-27 is solved exactly,
the resulting flow-position characteristic does not vaiy more a few
per cent from a linear relation. It may be shown that controlling outlet
flow instead of inlet flow, and, the use of proportional control in stead of
proportional-integral control provides the same result.

Fio. 8-0. Temperature-control Bystem.

Consider the outlet temperatiu^ control of Fig. 8-9. Generally the

measuring lag and the process lag are the largest lags in such systems.
The steady-state sensitivities (gains) are:

Ax
Controller
Ae

Control element iVv = —

Am
Ax

Ac
Process
Am

as shown in the operational diagram, The system operational equar

tion is

KcKJC^{T2S + 1)
(Ts + \){T28 + 1) + KcK,K^
+
^ ^ ( ^2
8 )
(T8+l){T2S+l) + KcKJC^
252 Automatic Process Control
IjO

oa
Liquid-level controIxJ
^
0.6
-V
I
a4
/
./ Temperature control
0.2

0.2 0.4 0.6 0.8 1C

Position

Position

Fig. 8-10. Desired characteristics of cootrol elements.

where = measuring element lag

T = process lag W/Q
W = weight of fluid in tank
Q = flow rate of liquid through tank
u = inlet temperature
= ^/QP
= Specific heat of fluid
“a,
Prooeas Instrumentation 253
For a given dam|»ng ratio {, the following relatirmBhip of the parameter
is desired:
1 (T+Tj)* -
1
4f* Tfi

Substituting for the “variable” parameters, Kf, T

dm WP
dx

In thifi equation T2 ,
P, Wj and Kc are constants. The throughput
rate Q, however, affects the heat input because steady-state con-
servation of energy requires
m = QPD (8-3fi)

where D « temperature difference inlet to outlet. Combining equa-

tioiiB ^-34 and 8-35 to eliminate Q, and considering a damping ratio

erf 0.707,
dm
^
WP r/ fzm V ,
(8-36)
dx 2KcTil\WPD/
The equation is now integrable and may he written,

dmi
xi^K (8-37)
fmi* H- 1

T'2m
where mi
WPD
X
Xl
2KcD
K = constant

Integrating,
tan ^
mi = Xi K (8-38)
or
mi = tan Xi (8-39)

This relation is plotted in Fig. 8-10. The particular temperature-control

iqrstem of Fig. 8-9 would therefore require a control element of the
increasing sensitivity type. Notice that the tangent function, parabolic
function (irxiVZ), and the exponential function (e® — 1) are all
nearly identical. On the other hand, for small values these functions
are all nearly linear.
254 Automatic Process Control

Consider the flow-control system of Fig. 8-11. The measuring lag

(at H) and the controller lag (at Gi) are usually the only lags of any

magnitude. These lags are also constants. The flow equation for the
orifice or other restriction is

q = KoVmi - p (8-40)

or regarding p = /(ui, q) and linearizing,

p - Ui - Khq (8-41)

where

(8-42)
\dq/ui

The flow equation for the valve is

q = Ml — M2 (8-43)

where it is assumed that the permanent pressure loss of the measuring

element is negligible (as for a Venturi). Regarding q = /(x, ui, U 2 ) and
linearizing,

q = K2X + Kn{u\ — U2 ) (8-44)

where

K2 = r~l = KvVui - U2 (8-46)

LdxJ

These relations are shown in the operational diagram of Fig. 8-11.

The must be constant (equal to 4), regardless
loop sensitivity (gain)
of the particular values of lags and loads (ui, U 2 ), because the system
loop gain depends only upon the time constants, and these do not
Prooess Instrumentatioii 255
depend upon pressure or flow. The loop gain is therefore

KJC^Ki = A (8
-47 )

L_. L_
'(S’!
1 11
(»-48 )
*
UIMI Ui

KeKp \/Ui — Uj A 8-49 )

(
It©

Eliminating the pressure differential by employing equation 8-43,

where AKq^/2Kc = a. This relation is plotted in Fig. 8-10. Notice

that the system calls for a control element of the decreasing, sensi-
tivity type when the set point is changed but pressure differential is
constant.
Summarizing the effects of variable loop sensitivities, it can be shown
for many process control systems that the effect of changing sensitivity
caused by vaiying process load can be adequately compensated by
'^shaping'' the final element characteristic.
Difficulty in analyzing every control application is caused by three

important factors: first, changes in process system function with load

are different for each disturbance; second, within each process the sys-
tem function is different for each disturbance; and third, valve flow
and area characteristics are not always identical. Consequently, from a
theoretical standpoint, different valve characteristics are required for
different process arrangements. It must be emphasized, however, that
process loads do not often vary to any great extent, so that effects of
process reaction and load are minimized.

Cascade Control Loops

Many processes are controlled by regulating the flow of a heating
medium such as steam, gas, oil, or fuel for supplying heat to a process.
Variations in flow not dictated by the by changes
controller are caused
from changes
in pressure differential at the valve, which, in turn, result
in pressure of the supply, changes in downstream pressure, and so on.
These changes are difficult to counteract since they must carry through
the process before they are detected in the controller. Supply changes
sometimes occur suddenly or over a wide range, and deviation may
become excessive before a new balance of conditions can be established.
256 Automatic Process Control

Consider the control system shown in Fig. 8-12. The temperature

controller TC measures exit temperature of the process c and transmits
a signal V 2 which becomes the set point of the flow controller FC. The
flow controller measures and controls flow rate C2 by adjusting a control-

Fxg. S-12. Control with metered flow.

valve setting m. Load (ua) at the inner loop is a variable upstream

pressure at the control valve. Jjoad U4 at the outer loop causes varia-
tion in controlled variable. The operational equations are:

C2 — G^ni + NsUg c = G4 C2 "h N 4 U4

M = G262 V2 == Oie
(8^51)
^2 = t>2
“ e = V — b

62 ~ H2C2 h - HiC
The controlled variable is given by

(GiG2G3G4)v H- {G4N2)u3 ^ + (?2g3^2A^4)u4

(N4
(8-52)
1 + G 2 G3 H 2 + G O 2 G3 G 4 H
1 1

The independent action of the inner loop is described by

G 2 G3 ^ N 3
(8-53)
1 + G2G3H2 1 + G2G3H2
The inner loop is generally a flow-control system and is considerably
faster than the outer loop which is generally liquid-level or temperature
control. Consequently, the type of control employed in the inner loop
(at G2 ) is not as important as the type of control in the outer loop (Gi).
The following observations are made, based upon a detailed study of
the loop equations:

1. The control action of the inner loop should be the simplest. Omit
integral or derivative responses unless speciflcally desired.
2. Set the inner loop for tight control; i.e., high loop gain.
 ;

Process Instrumentation 257

3. H
Measuring lag of inner loop (at 2 ) provides a derivative action
for outer loop. Consequently this measuring lag is not always harmful
as long as it is not excessive.

Example 8-5. Control systems sometimes employ a pressure controller

(regulator) to hold valve upstream pressure constant and it is often used in
place of metered control. State an advantage and a disadvantage for this
method as shown in the accompanying figure

1. Advantage: Usually less expensive because a simple fluid regulator

replaces expensive flow controller.
2. Disadvantage: Requires two valves in series with attendant pressure

Cascade control is extensively applied in control of liquid level when

a liquid-level controller regulates the set point of a flow controller.
Pressure changes in the vessel and downstream from the control valve
frequently make the control of liquid level in the tank difficult. By
means of cascade control the effect of these pressure changes can be
eliminated. Both liquid level and outflow can thereby be made more
stable and consistent.
A pressure-balanced valve is another form of cascade control. The
valve is so constructed that the output pressure of the controller is

applied to one side of the diaphragm and the downstream gas pressure
is applied to the other side of the diaphragm. The pressure-balanced
valve is actually a self-operated pressure regulator whose set point is

regulated by the master controller. Any change in gas pressure i.s

counteracted by the pressure-balanced valve and is not allowed to pass

into the process.
Ratio-flow control systems are arranged as shown in Fig. 8-13. The
primary instrument here is not a controller but a transmitter. The set
point of the controller is set in direct relation to the magnitude of the
primary flow. As the magnitude of the primary variable changes, the
258 Automatic Process Control

set point of tlie eontroller is automatically moved to a new value so

that an exact ratio is maintained between primary and secondary
variables.
The controlled variable is given by

G,G2AK N
(8-54)
1 1 -hGiG2H
Several important charat^teristics of ratio control may now be shown.
Suppose, by way of example,

(8-55)

K= k

where Ta = primary measuring element lag

k ~ desired ratio
Tr = integral time of controller
Tn = controller measuring clement lag

Substituting in equation 8-54 and considering load u fixed,

^ / TnS + l \ TtS -h 1

(8-56)
kv ” \TaS+ 1 / T.'l

Kc
In order to obtain the best “following^’' of the controlled variable to the
desired variable, the lags* Ta and the integral time T* should all be
as small as possible. The Kc should be as large
proportional sensitivity
as possible and the control system requires careful adjustment

Batch-Process Control
A process in which the materials or work are stationary at one physical
location while being treated is termed a batch process. This type of
Process Instrumentation 259

process is distinguished from the continuous process in which the mate-

rials flow more or less continuously. Batch processes are most often of
the thermal type where materials are placed in a vessel or furnace, and
the system is brought up to temperature and pressure and controlled
for a period of time. If the process reaction rate is slow, and lags and
dead time are small, a two-position (on-off) controller is generally
satisfactory.
If the process reaction rate is large or if lags are not small, it may
be necessary to employ proportional control in order to avoid excessive
cycling of the temperature. Suppose that w'e consider the action of
the controller through one complete batch. Figure 8-1* shows the

Fig 8 -14. Overnhoot iu batch-proceas ('ontrol

temperature and the corre.sponding valve action with proportional

control. As the teiiiperatiirc rises when the processing begins, the valve
is open. When the timiperature reaches the lower edge of th(‘ propor-
tional hand the valve begins to close, 'fhe temperature then eyide.s
about the ^et point iiefore bei'oming stable.
Witii most ])roces.Nes a gradually decreasing valve setting is reijuired

in Older to balance energy (offset)

The giadual clo. nig of the valve can be aceom|)lished only by a
corresponding deviation of the variable from the set point. As long as
proportional semsitivity is high, offset will also be small.
Offset may be eliminated by means of integral response. The action
of the proportional-integral controller is shown in P'ig. 8-14. The offset
of temperature is nearly eliminated, and the temperature is maintained
close to the .set point after it has stabilized. Notice that tlie initial

overshoot of temperature, when approaching the set point for the first

time, is mui h larger than with only proportional control

 260 Automatic Process Control
1'he large initial overshoot is due to the action of integral response.
During the heating-up period the temperature is, of course, below the
set point. The normal action of integral response is to shift the pro-
portional band to urge the temperature toward the set point. Since
the valve is already full open, integral response simply shifts the pro-
portional band all the way.
Overshoot is prevented in one of several ways. First, many electric
controllers with integral response incorporate a “rate of approach
setter. “ This device limits the integral action until the first approach
to the set point has been made. Second, a controller which incorporates
a proportional-derivative unit followed by a proportional integral unit
also may be adjusted to prevent overshoot. In the latter case the deriv-
ative response provides “anticipation** by making the integral con-
troller think that the controlled variable has approached the set point

e fni fll2
K^(l + T^8) ^ 4. JLA

Fig. 8-16 Series-cuDtrol elements.

as illustrated in Fig. 8-15. The sensitivity of the first controller Ki

may be adjusted so that with a given setting of derivative time the over-
shoot is eliminated.
In the processing of metals and chemicals it is sometimes necessary
to var\' the controlled variable over a definite time schedule. For exam-
ple, ill annealing steel, a schedule may be required such that the furnace
Process Instrumentation 261
temperature is raised to 1500 F
4 hr, held for 8 hr, and lowered to
in
500 F in 6 hr. This time schedule must be incorporated into the auto-
matic control system.
Time-variable control is accomplished similarly to ratio control. All
systems operate on the principle that the set point is moved through the
desired time schedule. The control system functions to maintain the
controlled variable, temperature in the example above, as close to the
moving set point as required.
The time-variable control system is also illustrated by the general
diagram of Fig. 8-13, where v = f{t) is the desired time schedule. Very
often it is required to change the controlled variable at a steady rate in
such batch processes involving controlled rates of heating or cooling.
In these applications the controlled variable lags behind the desired
value. This may be calculated by employing the same technique as
previously:
K 1 + GiG2(H - A) "!
D„ == Urn (8-57)
j -»0 .8 1 “h J

Usually, but not always, the velocity error is small.

Batch processes are nearly always defined by temperature, pressure,
or associated conditions such as composition of environment. The
degrees of freedom are usually well-defined. The purpose of such
processes is to produce one or more products at (a) a given composition,
(6) a maximum amount, and (c) best economy (employing least mate-

rials, energy, and time).

The product composition desired is that at the end of the processing
period and thus cannot be measured. Consequently it is necessary to
manipulate the variables of the process in such a manner that the
behavior of the process insures obtaining the desired composition.
Maximum product and best economy result when the variables of the
process are properly manipulated. Generally this is achieved by a time
schedule for all variables of the system. A computing machine may be
used to insure a relationship among variables providing best operation.

Continuou8-Pro<^8 Control
A process in which the materials or work flows more or less continu-
ously through a plant apparatus while being treated is termed a con-
tinuous process. The problems of continuous-process control are caused
by load changes.
Flow rate of the materials is almost always important in continuous
processing, particularly where quantitative reactions are involved as
in blending. On the other hand the supply of material is sometimes not
262 Automatic Process Control
coiistaut and a flow control is necessary as shown in Fig. 8^16. The
resulting flow of materials is nearly constant in spite of large changes of
head in the tank. It is necessary to size the tank only large enough so
that the longest charging period will not run the tank over, and the
longest off-tinie of the inlet will not allow the tank to run dry. In order
to keep load changes at a minimum, a short tank of large cross-section
area (large capacitance) should be employed.

1 1
n 1 1 1
1 1

Time-

Kic} 8 10 Storage- vessol control.

In many applications, particularly in continuous processing, the out-

flow of one unit becomes the feed to a succeeding unit. In order to
obtain stability of operation in the plant, sometimes important that
it is

lluctuatioij of outflow and inflow be reduced to a minimum, thereby

maintaining all feeds relatively constant. The storage capacity of the
vessel may be utilized to proportion outflow against changes of inflow.
I'hc vessel thereby serves as a surge tank for absorbing fluctuations in
flow rales. Averaging control gets its name since the outflow is "aver-
aged” against level, and the level is controlled between upper and lower
limits rather than at a single point.
For the vessel of Fig. 8-17 without, control of outflow (outlet valve
at a fixed setting),

Ac = — q (8-58)

where A = cross-section area of tank. When the vessel is under con-

trolled pressure, as is generally the case, the outflow is thei) constant if

the pressure differential at the valve is constant. The inflow, head, and
outflow are shown by the dotted curves in Fig. 8-17. Obviously if

there is no control, the tank would ultimately fill or run dry. If control

of level is added,
-q = Kci^v -- c) (8-59)
 Process Instrumentation 263
where Ke is the proportional sensitivity of the proportional controller.
Combining equations 8-58 and 8-59,
1
9 ^ (8-60)
u 1

and
c 1
_ 1
(8-61)
Ic, Ts-f 1

where T = A/Kc^ The inflow, head, and outflow are also shown in
Fig. 8-17. The time constant T may be selected by adjusting the pro-
portional sensitivity of the controller so as to ''spread out" the inflow

Fig. 8--17, burge-vessel control.

change over a period of about four time constants. In order to absorb

changes of inflow the time constant T should be made as large as possible
(proportional sensitivity Kr as small as possible) without completely
filling or emptying the tank. Sometimes a proportional-integral con-
troller is used with integral time set to a large value so that the outflow
is further increased and the head in the tank maintained near the
set point.
Continuous processes possess a number of degrees of freedom given
by the numbers of variables and defining relations for the system (refer
to the first sections of Chapter 2). These variables are generally the
temperature, pressure, flow-rate, and composition of each of the entering
and leaving materials. Usually the purpose of the process is to produce
one or more products at (a) a given composition, (6) a given or maximum
flow-rate, and (c) best economy (employing least materials, energy,
personnel time, and equipment).
Product composition is best insured by measuring product composi-
tion and controlling it by manipulating one of the degrees of freedom
of the process.
 :

264 Automatic Process Control

Fixed product flow rate usually requires flow controllers at several

points of entering and leaving materials. If the product flow rate is

to be a maximum, the flow controllers shoifld not be employed, or else

they should be manipulated to achieve one of the purposes of the process.
Best economy is accomplished if at all by maintaining all process
variables in a predetermined relation such that the highest efliciency,
least waste, or some other criteria are satisfied. This practically pre-
cludes the complete use of automatic control because holding all vari-

ables constant does not allow variable inlet condiiiions or variable flow
rate. Economy can be achieved by employing a computing machine
and relating all variables, even in a transient state, so that the desired
criteria are met.

Optimizing or Computing Machine (Control

The objective of any process control system is to insure that the
system under control produces the best product at maximum quantity
and least cost. In order to achieve this objective the present approach
is to use automatic controllers on each of the degrees of freedom of

the system, thereby holding the system in a fixed state. A process

operator (human) then manipulates each of the controllers until the
objective stated above is accomplished.
Operation by a human overseer is rarely as effective as it might be
for several reasons:

1. The processing relations are often so complex that a working grasp

of the interrelations are beyond human abilities.

2. The process may operate in such a fast time scale that human
action is always too late.

3. The process may operate in such a slow time scale that human
attention is ineffective.

4. may be of utmost concern.

Safety
5. Human operation is not possible in locations where human beings
cannot go.

Optimizing (sometimes called optimalizing) control may be arranged

by employing a computing machine to perform the operations normally
provided by a human overseer. There are two general methods by
which computer control can be accomplished
A. A perturbation method whereby the various degrees of freedom
of the system are changed incrementally and periodically until a selected
function of performance is observed to be a maximum.
B. A continuous computing method wherein the relations defining
optimum performance of the system are derived, and the control system
Process Instrumentation 265

is designed to manipulate the process variable in such a way that the

optimum conditions are achiex ed.

The 'perturbation method} has recently been employed in the control

of heat-engine performance, and a simple version of the method has
been employed on heating furnaces. Suppose that it is desired to main-
tain best fuel-air ratio for maximum heat release in combustion in a
furnace subject to whatever limitations may apply. This may be acc.oni-
plished by changing the air flow at constant fuel flow by incremental
amounts and on a fijaed time period. I'he resulting change of heat
release can be measured by combinations of flame temperature and
stack losses, and, if the change of heat release is positive, another incre-
mental step in air flow is made in the same direction as previously. All
the while, fuel flow is controlled directly from work temperature. This
incremental stepping is continued until a negative change in heat release
results. The direction isthen reversed and, stepping is continued until
again a negative change in heat release results. Thus the system will
continue to oscillate about the maximum point.
The perturbation method is particularly applicable to systems having
relatively simple optimum functions, and its great advantage is that
little if any information about sys-

tem behavior need be known. The

disadvantages of this method are
that it is difficult to apply to sys-

tems having more than one degree

of freedom and to systems with long
time lags. A further disadvantage
is that this method will not locate
an absolute maximum as shown in
Fig,8-18 but may stop at the lower
maximum unless special features are
built into the system.
The conimuous computing method^ System variable, m
has been developed for use on either g-lS. Optimum .ystem function,
batch or coiitinuous processing.
The general method is illustrated in Fig. 8-19 for a batch process. As-
sume that the behavior of the process reejuires constant temperature,
but, by properly manipulating process pressure, the batch may be taken
from an initial point to the proper final point in least time.
6. Draper, Y. T. Li, “Principles of Optimalizing Control ^Systems and an
Application to an Internal Combustion Engine," ASME Publication, 1051.
* I. Lefkowiti, Process AuUmation — Report /, 1954-1956, Case Institute of Tech-
nology, Cleveland, Ohio.
 266 Automatic Process Control

The optimum behavior of th^ system is derived in general form.

Euler- La Grange technitiues and the calculus of variation are useful in
this deri\'ation. These equations are built into computer Ci. The
final desired product (luality c/ is set into servo Si, The output of the
servo is the pressure as a function of time. This function and the
existing pressure Pi, temperature Ti, and coefficient K are fed into
computer Ci which then computes the predicted final value of product
quality c/c- This value is compared to the desired value, and the pres-
sure is manipulated so as to “steer” the procqfjs to the proper final
point, but always under the limitations imposed by minimum time.

Fio 8-19. G*n»[)UtinK-mafhinp oontrol

There are two difficulties t-o be overcome with this method of control;
first, the behavior of the system may 'ot be known in general analytical
form, and sen-ond, the sy.st{*m jna> have long time delays. In order to use
the continuous computing method the system behavior must be known
to some extent The and em])iri-
behavioi' can be tested experimentally
cal reliilions found either by transient or sinusoidal response methods
or l)y correlation methods.
Often it is sufficient to know only the general form of the behavior
(as the geiicial ditTerential e<iuations with unknown coefficients). In
this cose a self-checking method can be ased I\eferring to Fig. 8-19,
system variable Cm is nieasurcfl and is fed directly to servo S 2 which sets
the unknown coefficient K of the reaction. Computer C 2 takes data on
temperature, pressure, and coefficient and computes the \alue of the
system variable that should exist, C^. T1k‘ computed \alue is fed back
to the servo, ami a new coefficient determined until computed and
measured values match, d'he form of the system reaction must be
assumed and fitted into computer d'he “self-cheeking” elements,
S 2 and Co, insure that the assumed equations fit process behavior in the
1

region of actual operation. In a sense, the self-cliccking elements con-

struct an analog of the process within the com)>uter
]Voce«s liii^truinentaliuu 2«7

long time delays (lags) are encountered, the continuous-

computing method is also able to ‘^predict’^ behavior during the period
of the time delays and operate the process on schedule as though the
time delay does not exist. This method works quite satisfactorily so
long as the system behavior is not completely erratic.

PROBLEMS

8-1. 'I'he system block diagram ran be redrawn with GiGiH in the for-
w'ard loop when a feedback element // is present. What function is required
at the input element?
8-2. A system (see Fig. S-1 ) has the following elements.
A - 1 A - I

6’i = 7’s + 1 // - l/(7Vs -b 1)

G, - (1/T,.9)*

Calculate the values of derivative time T and measuring lag Tn to give a

transient response consisting of three equal tune constants.
8-3. F or the SNstein of problem 8 2, calculate the values of (a) static
error, {h) offset, (c) vidocity error, id) offset due to rate ol change ol load.
8-4. A
furnace has a heating rate of 50 deg per min for full valce opemng
and a mc.asining element time constant of 0(1 sec. The measuring element
time constant can be reduced to 20 see Would this change be worth while if
proportional control is used?
8-5. A gas pressure vessel has pressure control by throttling the outlet
which operates at acoustic velocity. Tlie inlet is uncontrolled. Proportional
control is used and the valv'e time com.tant is Tv. Prove that a linear valve
provides constant loof) gain at all loads and set points.
8-6 . Considering ca.scade control :i.s in Fig, 8-12 w ith
1

G =
2 Kr, C3 = A', Hz =
7’.s- -h 1

show- that the inner loop i.s a lea<l-lag network.

8-7. Considering cascade control a.s in Fig S-12 with

= = A (fA - -
Gi K], G‘2 2,
-b 1
'

7Wl
I

N, - A',, Gi A 4 = 0
+
’
"1
r,.v

show that the off.set with load W3 is greater when the inner loop is broken
(Ao - 0) ,

8-8. For tlie storage vessel m Fig S Hi, sketch the outflow as a function
of time with the, inflow as shown if (a) integral control i.s used an<l (/>) propor-

tional control is used.

268 Automatic Process Control

For the storage vemel in Fig. 6-17 with a step inflow of U units,
vessel area A,and proportional-integral control, calculate the offset in flow rate.
6-10. For the surge vessel in Fig. 8-17 with integral control T,-, and no valve
or measuring lags, calculate the integral time for critically damped performanoe.
6-11. For the surge vessel in Fig. 8-17 with proportional control Fe, a
valve lag Tv but no measuring lag, calculate the proportional sensitivity for
critically damped performance.
 cha/titT

Sinusoidal Analysis

1'he use of sinusoidal methods of analysis and test of dynamic systems

has gained widespread popularity because of the simplicity in treating
complex control systems. The transient analysis of the previous chap-
ters is largely limited to second-order systems unless a specific system is

considered. Sinusoidal analysis aids in qualitative considerations

regarding stability of control systems and therefore is a convenient
method for the engineer.

Sinusoidal Disturbance and Response

The response of a dynamic system may be found by imposing a dis-

turbance at one point in the system, holding all variables but one con-
stant, and noting the result upon the desired variable as indicated in

Fig. 9-1. A sinusoidal change made in variable e while holding variable

u constant results usually in the typical periodic curve at variable h.

Fio 9-1 HidusokIh) re^ipoude

26f)
270 Automatic Proc6(» Control

It is not necessary that the response & of a system to a sinusoidal

input be sinusoidal If the performance of the system is described by

a linear differential equation with constant coefficients then the response

must be sinusoidal and, furthermore, the response will be of the same
frequency as the disturbing function.
A nonsinusoidal response will result if the system is not linear; that
is, if the system performance is not described by a linear differential
equation with constant coefficients. In this case the principle of super-
position does not apply, and there may be a great deal of difference if

the order of elements in the system is changed. In Uddition, the response

need not be of the same frequency as the input disturbance.
The steady-state sinusoidal response of linear sy.stems is found by
imposing a steady sinusoidal signal at variable m as indicated in Fig. 9-2,

m = sin uii (9-1)

The resulting periodic wave is recorded at variable c. The output c

usually lags behind the input m and is usually smaller in amplitude.

1 c
2^ + 1

Fia. 9- 2. A time-constant element.

It is important to note that the response or output

is taken after the

transient response has disappeared. For the purposes of illustration we

will find the response of a time constant element to a sinusoidal input

first, by classical differential equation solution, second, by complex

solution, and third, by operational solution.

The classical method is to substitute the disturbing function:

do
di
+c= sin (9-2)

Next we construct the particular integral by the usual methods. Try

the following solution'

c = i4 sin ctf/ -h cos od (^)

 —

Sinusoidal .Analysis 271

where A and B are, as yet, unknown constants. Substituting in eciua-

tion 9-2, there results

AcjT cos ut - BcjT sin col A- A sin cot + B cos cot = sin cot (9-4)

This must produce an identity so like-terms are compared to yield the

following equations:
AaT H- B= 0
(9-5)
A - BuiT = 1

Solving for unknowns A and B there results

A -—1
1 +
-6
(9 )

n
1 +
The particular integral is, therefore,

c = -
—“T ;

2 ^^ ‘*^0 (9-7)
1 W*'i

Rewriting this result as a single function

c = —" T— T- sin
'-
(cot
— tan“* coT) (^8)
V\ +
Comparing the output c to the input m (equation 9-1), it is seen that
the output sine wave has lesser amplitude. The ratio of the amplitudes is
designated by the absolute value signs thus-

I
(9-9)
fn V\ +
It is also seen that the output sine wave is delayed and the phase is

indicated by the angle sign thus:

—tan ^
uT (9-10)

The magnitude and phase of the system can therefore be calculated.

The complex solution is found by writing the input of equation 9-1
in complex form:
TW = Im (9-11)
272 Automatic Process Control

wbere Im means *^the imaginary part of." The function above is de-
fined by Euler^s relation,

* COB + i sin id (9-12)

where i « The imaginary part of Euler's relation is the sine

function.
Imposing the disturbing function on the time constant element,

T^
at
+ c = Im e*“‘ (9-13)
I

In order to find a particular integral for this equation, we rely upon a

theorem of mathematics which, stated without proof, insures that for
any linear differential equation with constant coefficients the real part
of a solution is the solution to the real part of the disturbing function,
and the imaginary part of a solution is the solution to the imaginary part
of the disturbing function. Consequently the Im may be dropped,

r^
at
+c = e"‘ (9-14)

The solution to this equation is given by

c = (g~15)

Substituting in equation 9-14 gives

Simplifying with the usual precaution that we do not divide by zero,

+ l)Be'* = 1 (9-17)
Solving for B,
1
B= (9-18)
(i«T + 1) c'*

The solution is then, combining equations 9-1 6 and 18,

Next solve for the real part in the following series of manipulations:

« - wt -f t ain ftrf) (9-20)

Clearing the denominator of t, we find

/ 1 - iaT\
(9-21)
Sinusoidal Analysis 273

Breaking into real and imaginary parta, the solution is

cos <at uT mi id ,mnU — aT cos od

(^»)

It is recalled that we desire the imaginary part, therefore the desired

solution is,

c = (sin wt - u)T COB <at) (9-23)

1 “T W I

or

c = —7==== sin - tan“‘ uT) (9-24)

VVT^ (u)t

In this manner the response ol systems described by linear differential

equations with constant coefficients can be found by employing the
procedures of the complex variable. This procedure is, however, lengthy
and short-cut methods may be employed.
The operational solution for the response of the system may be found
with operational mathematics. Stating the system equation in oper-
ational form for constant u,

f
£. - (9-25)
m “ Ta
+ 1

Replacing s by tea,

c 1
(9-26)
in iwT -f- 1

Compare this equation to equation 9-19. Clearing the denominator of

equation 9-26 by multiplying both numerator and denominator by the
complex conjugate (1 - iwT)
c 1 — iwT
(9-27)
“
7n 1 “1

Compare this equation to equation 9-21. The magnitude of this com-

plexnumber is the square root of the sum of the squares of real and
imaginary parts,
'

1
'

- (9-28)
- \ 1 +
The phase of a complex number is the arctangent of the imaginary part
divided by the real part,

— = tan”^ ( -w'jT) (9-29)

7W
 274 Automatic Process Control
Compare these results to equations 9-9 and 9-10. EquationB 9-28 and
9-29 show the amplitude and phase of sinusoidal response of the system
to a sinusoidal disturbance of unit amplitude.

Example 9-1. Calculate the amplitude ratio and phase of the sinusoidal
response of the system described by

The system function is

n+ re - m

c 1

m " Ts(Ts -h 1)
Substituting « = iu,
1
^
m iuiT{l + iuT)
Clearing,

^ __
^‘(1 — _ —<j)T ~ i
m“ "
o)T(l + wTd -f (^f^)
The amplitude is

1 .
1
/
\ (1+ + ^2^2)2
or

f
w I
I wT V 1 -f

The phase is

The sign of the phase must be considered carefully because the inverse func-
tions are multivalued.

Operations with Complex Numbers

Calculations of the steady-slate sinusoidal behavior of systems is

generally made by employing complex n urn tiers. As described in the

previous section, these calculations are considerably simplified if the
rules of manipulation of complex numbers arc usoil to the fullest
A complex number can be written in rectangular form

z = a + t6 (9-30)

and sketched in Fig. 0-3 by marking off length u t)ii he real number
is t

axis,and marking of! the length 6 on the imaginary number axis. When
a complex number is written in polar form, only the magnitude and the
angle need be given.
The magnitude is calculated from Fig. 9-3.

= M (9-31)
 Sinusoidal Analysis 275
and the angle is given by

/S = » = tan-* “ (9-32)

Addition of complex numbers is made by adding real parts and

imaginary parts respectively,

-h S 2 = a -h + c 4- id = (o + c) 4- i(b + d) (9-33)

Fig. 9-3 Vector complex Dumbere.

Addition is vector addition as shown in Fig. 9-3. Multiplication of

complex numbers is performed by algebraic manipulation:

(a 4" tb)(r 4' id) *= (ac — 6(f) 4- i(6c 4" ad) (9-34()

Multiplication is the most useful manipulation in dynamic systems

studies. By algebraic manipulation,

2i’Z2 - (a 4- tb)(c 4- id) = (oc - bd) 4- i(6c 4- ad) (9-35)

However, it may also be shown that

is, -lol = V(^+ 6*)(c* + d*) (9-36)

SO that magnitudes are multiplied. The phase is

/5r22 = tan""^ - 4- tan“' - (9-37)

a c

so that the phases are added. Therefore the product of two complex
numbers is obtained when the magnitudes are multiplied and the phases
are added.
The conjugate of a complex number is obtained by changing the sign
of the imaginary pari,

f = a 4- t6, z* = a — ib
276 Automatic Process Control

as illustrated in Fig. 9-4. The conjugate is often used in clearing com-

plex fractions because
h h-ii* h-li*
li
” o* + 6*
(9-38)

Multiplication of a complex number by t is equivalent to a rotation of the

vector through 90 deg, because the real and imaginary parts are inter-
changed as indicated in Fig. 9-4.

Fig. 0-4. Vector complex aumbers.

Example 9-2. State the magnitude and phaae of the following numben:
Phase
Number Magnitude Degrees Radians
4 -i3 5.0 -36.8 -0.643
+ v^ + v^ 2.0 +45.0 +0.785
i 1.0 +90.0 +1.67
-10 10.0 180.0 3.14

Example 9-3. Multiply the number 3 + 4i by t and add its conjugate

to 7 4- 7i.
(3 -h 4t) X i * -4 4- 3i

(7 -f 7i) 4- (-4 - 3i) = 3 4- 4i

Plotting Sinusoidal or Frequency Response

The response of systems to a sinusoidal disturbance is» in general
terms,
c * 4 sin (wf 4- (0-39)

where A = amplitude and 4> = phase. This function may be plotted

as in Fig. 9-1, but the result is relatively uninformative. In order to
visualize the response characteristics, it is much more descriptive to
SnuBoicUl Analyau 277

plot the amplitude and phase, since it is known that the function is
wuaoidal. Furthermore the amplitude and phase arc a function of
frequency so that sinusoidal response may be considered in the frequency
domain rather than the time domain.
Sinusoidal response may be plotted in three different ways: (1) The
rectangular plot with amplitude ratio versus frequency, and phase versus
frequency; (2) a polar plot with magnitude and phase shown in vector
form with frequency as a parameter; (3) a phaie margin plot with mag-
nitude shown versus a function of phase with frequency as a parameter.

Fr^iuency, wT Frequency, «T
Fio. 9-5. Magnitude- and phaae-frequency plot.

The rectangular plots of magnitude vs frequency and phase (angle)

versus frequency are by far the most common. This type of plot, often
termed a Bode plot, is shown in Fig. 9-5 for the time-<Jonstant element:

(9-40)

The magnitude ratio is

(9-41)

The magnitude ratio is plotted on log-log paper in order to gain the

advantage of simple multiplication and in order to employ straight line

asymptotes:
Log|G| . -ilogd (9-42)

At low frequencies, wT « 1.0, the magnitude ratio is a line horizontal

at one since as wT —* 0,
Log |(?|
=0 or ^
|G’| «= 1.0 (9-43)
278 Automatic Process Control

At high frequencies, tuT » 1.0, the magnitude ratio is a line from upper
left to lower right with a slope of minus 45 deg because as wT —> •

Log |Cr| -log^r or |g1 * (9-44)

The phase is,

[G = tan“‘ - wT (^6)
and is also plotted in Fig. 9-5. This graph is made on semi-log paper.
Because of the arctangent function there is no linear approximation of
the curve but, however, the function is symmetrical about the wT = 1.0
point.
The polar plot of real versus imaginary part is also used in stability
studies. From the complex form of the system function,

1 - iu)T
G (9-46)
1 + iu>T 1 -h

The real and imaginary parts may be calculated for different values of
uiT and the locus of the end points of the vectors form a curve known as
the system locus. The plot is as easily made by plotting magnitude
and phase. The system locus is shown in Fig. 9-6 for a time-constant
elen\ent. Note that the frequency wT becomes a parameter marking
out different points along the locus. The magnitude is the distance
from a given point to the origin, and the phase is the angle taken counter-
clockwise from the positive real axis.

The magnitude vs phase plot, Nichols chart or phase margin plot, is

often used to define system performance. The magnitude is plotted
vertically, and the phase is plotted horizontally, as shown in Fig. 9-7.
Sinusoidal Analysis 279
Instead of phase, the phase margin is plotted. The phase margin is 180
deg plus the phase. For a time-constant element, the plot for various
frequencies begins at the right-hand center of the chart and proceeds
downward and to the left. Frequently it is necessary to use extensions of
the chart so as to provide for a gain greater than one and for negative
phase margin.

Sinusoidal Response of Systems

The frequency response of systems containing a number of different

dynamic elements may be calculated by suitably employing the magni-
tude and phase characteristics of each element separately. The calcula-
tion of magnitude is made from

IGiOjI = lC,| X IG2 I

(9-47)

On the rectangular and phase margin plots, the magnitudes arc set
against a logarithmic scale, so that magnitudes are multiplied by adding
vertical distances. The calculation of phase is made from

(9-48)

This amounts to the addition of phase. These points will be illustrated

by examples at the end of this section.
Combinations of various dynamic elements may be investigated wiui-
out conversion to frequency form. Any such combination can usually
be described by the op)erational function

r( ) = K
^
-f
• - •
-f ais + Op) (9-49)
•
H- -h •
biS bo)

where fc, w, and n are positive integers. The total phase at sufficiently
high frequencies approaches

= 90(m -n -k) deg (9-50)

Therefore the phase at infinite frequency corresponds to the net order of

the denominator in units of ir/2. The phase at low frequencies is given
by
/G = -90A:deg (9-51)
»o

or the phase at zero frequency corresponds to the net order of factorable

3 in the denominator.
 '

28() Automatic Process Control

The magnitude of system functions can similarly be investigated.
The magnitude at infinite frequency is found from

\G\ =0 if n -1- fc >

«—» •

\6\ if n -1- A: = (9^2)

W-* « On
and
“ * if n -1- fc <
i?L
The magnitude at zero frequency is found from

|g| = « if ik > 0

lel-^
••-*0Oo
if - 0

|Gl=o if A: < 0
«r-^

The rate of change of magnitude may also be determined since the rate
corresponds to the units of phase lag. In terms of the slope when the sys-
tem response is plotted on log-log coordinates,

Slope of IG'I
—m -n - k (9“54)

and
Slopeof 1^1 = -ib (9-56)
4#-*0

By way of example suppose that it is desired to compute the frequency

response of the system function:

1 1
Q= (9-66)
TsiTs -h 1) Ts -f 1

The first element, l/Ts, is sketched on the magnitude-frequency plot

of Fig. 9-8. The second element, l/fTs -|-
1), is next sketched. The
magnitudes are multiplied (distances added by using dividers if neces-
sary). I’he phases are added. From the form of the operational func-
tion 9-56 it is known that:

(a) Infinite frequency phase lag approaches 180 deg

(m— n — = 0 — 1 — 1*= —2).

(b) Zero frequency phase lag approaches 90 deg (fc * -1).

(c) The magnitude at infinite frequency is zero.
Sinusoidal Analysis 281

(d) The magnitude at zero frequency is infinite.

(e) .
The infinite frequency rate of change of slope is -2.
if) The zero frequency rate of change of dope is ~1.

As3rmptotic plots of magnitude versus frequency can be made because

the value of the magnitude becomes asymptotic to straight-line functions
at low and high frequency.Thus, asymptotic or straight-line plots may
be made many
systems by combining the asymptotic plots for each
for
individual component.

Fig. 9-8. Sketching frequency responee.

Plotting of magnitude of functions is very often done in terms of

decibels, abbreviated db. This unit of magnitude is defined as follows:

1.0 decibel = 20 logic 1.1122 (^67)

The magnitude of system functions are then calculated as follows

= 20 logic |(7| (^)

The magnitude of functions may be plotted on semilog coordinates with
frequency as abscissa. Phase margin diagrams are then plotted with
rectangular paper.
Several significant magnitudes may be noted in the table :

Magnitude Decibels
0.10 -20
0.50 -6.02
0.707 -3.01
1.0 0
2.0 6.02
10.0 20
282 Automatic Process (Control

Frequencies are often referred to as octaves (a factor of 2) or decades

(a factor of 10). Slopes of magnitude functions are then given in the
several terms

-1, -6d6/octave, -20d6/decade

and these are all equivalent.

Example !M>. Sketch the phase margin of

(s ^
The result is plotted in the figure shown below:

Phase margin, deg Frequency, cps

Example 9-5. Sketch the magnitude-frequency function for

215* 4-105-1-1

First, always factor the function if possible:

(75 4- 1)(35 + 1)

The response may now be plotted as two time-constant elements. The result
is plotted in the above figure.

Sinusoidal Response of Dynamic Elements

Frequency response of various process and controller elements may be
calculated most easily by employing the notation of the complex vari-
able. Frequency responses of various elements will be calculated and
their magnitudes plotted in the magnitude-frequency phase, polar, and
magnitude-phase diagrams.
Sinusoidal Analysis 283

The capacitance element has the operational function

The complex vector possesses an imaginary part only. The magnitude

ratio, and phase are

These are plotted in i'ig. 0-9. The magnitude versus frcfiuency plot
shows the magnitude to decrease inversely Avith fre(|uency. Note that

Fig. 9-9 Frequency response of pupacitanre element.

the slope is -15 deg (1:1). The phase lag, PL is constant at 00 deg
forany frequency. The polar plot is a vertical line along the negative
end of the imaginary number axis. Th(* magnitude-phase margin dia-
gram is also a vertical line of infinite length.
The time-instant element has the operational function

«=
r/+i i+lr
The magnitude ratio and phase are,

!(?l = ’ /6' = -tair' 0)7’ {9-G2)

These are plotted in Fig. 9-10. The magnitude versus frequency rela-
tion approximates l.O at low frequencies and falls off at high frequencies.
Again the slope is - 15 deg (1:1) at high frequency. The phase lag is
small at low frc(iuencics and approaches 90 deg for high frequencies.
At the frequency wT = 1,

1
<? 1 = 0.707, Z:G=_45deB
284 Autootatic Process Control

Fio. 9-10. Frequency response of time-constant element.

The polar plot is a semicircle as shown. The phase margin plot begins
on the right and passes downward.
The oscillatory element has the operational equation

(9-63)
TV + 2fTs + 1 (1 - «*T*) + »(2f«T)

The magnitude ratio and phase are

Va -«*F)5 +'(2^r? ^ (i -
(9-64)

These are plotted in Fig. 9-11 for several values of damping ratio {.
The magnitude is greater than one for moderate frequencies and low

Fio. 9-11. Frequency response of oscillalory clement.

damping. The phase lag is always small at low frequeneies aud

approaches 180 deg at high frequencies. The damping ratio affects the
phase slightly at frequencies just above and below resonance. The polar
plot is a much distorted semicircle and is tangent from below to the
negative real axis at high frequencies, because the phase lag is 180 deg
in that region. The phase margin plot is tangent to the zero margin line
 ,

Sinusoidal Analysis

at high frequencies. At the frequency wT > 1

/^--90deg ^

Prirporiional^niegral control has the operational equation

0-K.(,+±)^K. (1 -h uoT)
^ + ( 9 ^)
ia>T (cT

The magnitude ratio and phase are

/G = -tan-^

These are plotted in Fig. 9-12 for a value of proportional sensitivity of

one. The magnitude is greater than one at low frequencies, and tiie

Fraquency, caT Phaie margbi

Fio. 9-12. Frequency response of proportional-integral controller.

magnitude at sero frequency is vital to the

characteristic of infinite
integral action of the controller. A change of value of proportional
sensitivity Ke would simply move the plot bodily up and down on the
left-hand and right-hand graphs. A change of proportional sensitivity
expands the scale radially in the polar plot. Thus the change of pro-
portional sensitivity does not alter the phase.
ProportionoMerivalwe control has the operational equation:

G= Kc(l + Ts) Kc{\ -f (9-67)

The magnitude ratio and phase are

ICl = vTTiTF, (G - Un-‘ uT (9-68)

These are plotted in Fig. 9-13 for a value of proportional sensitivity of

one. The magnitude is greater than one at high frequencies, and the
characteristic of infinite gain at infinite frequencies is important in
the derivative action. A change in proportional sensitivity moves the
286 Automatic Process Control

Frequency, <aT Phase margin

Fig. 9-13. Frequency response of proportionaUderivativc contrf)Uer.

magnitude-phase and phase margin plots vertically. If we compare

these results to those of Fig. 9-12 for proportional-integral control, the
“opposite” characteristics of these two control actions are readily seen.
Proporlional-inte^al-chrivative control has the operational e(|uation:

G= K. ^1 + (1 + T^s) K, (1 + i^Ta) <9-69)

The magnitude ratio and phase are

"(1 +ic’‘T/)
|G| = (9-70)

—“ tan ^
w7'j “f” tan ^
(9-71)

These are plotted in Fig. 9-14 for a proportional sensitivity of one.

The magnitude is always greater than one and the phase passes from a
lag to a lead at high frequency. The system has infinite gain at both
zero and infinite frequencies. A change in proportional sensitivity
moves the magnitude-phase curve and phase margin curve vertically

Frequency. Phase margin

Fig. 9-14. Frequency response of pn)p<»rtional-integral-(lerivfttive cfudrol with

5T^ = r,.
 Sinusoidal Analysis 287
Example !M». Sketch the polar plot for a system with the operational
function Ts,
The vector complex number is calculated

G= T8-*xoiT
The amplitude and phase are

1G| = ojT, /G « -f90deg

The polar plot is a vertical line out along the positive end of the imaginary axis.

Example 9-7. A thermocouple has a time constant of 5 sec. At what

frefjuency will the element attenuate 50 |)er cent?
B rom ecjuation 9- 62,

^ - 1 = Oj

Example 9-8. The a pressure gage shows a frequency response as

test ol
shown in the diagram. What are
tlie characteristics of the system?
The test points are clearly asymptotic to a line horizontal at one. At high
frequencies the data lies on a line at negative 45 deg. Drawing these lines it
is seen that the intersection is at 0.08 cps. As a check, the amplitude at this
frequency is about 0.7 (sliould Ik; 0.707). All factors point to a first-order
system, and if there are other time constants they must be very small and
can be ignored. The intersection point is always at wT ~ 1, so

Frequency, cps Frequency, cps

Example 9-9. A thermometer and thermal \Nell have two time constants
where Ti - 100 sec, T-i — 10 sec. Plot the ainplitu<le-frequency response.
The plot of magnitude may be determined by using the approximate
(asymptotic) method: making the graplis on log- log paper, the response due
to each time constant may be added separately. The two response.** are then
combined additively to obtain the asymptotir resjionse The mterseciiuu
 Automatic Prooees Control
frequencies are found from

1
0.0016 opa
2ir!r,

1
h 2rT2
0.016 ope

Example 9->10« A two-position controller maintains a temperature of a

furnace oscillatory over 80 deg total amplitude with a period of one minute.
A thermocouple and well having a first-order time constant of 96 sec is used
to measure the same temperature. What amplitude jvould it indicate?
Referring to Fig. 9-10,

^ X 96 10

The amplitude is

« X 80 = 8 deg

Dead Time, Diatributed-Parameter, and Exothermic Elements

The operational equation for dead time may be written

G (9-72)

where L is the dead time and e is the base of natural logarithms. This
relation Chapter 2, by classical methods.
was derived in
The frequency response of a dead-time element is calculated in the
usual way. Replacing s by ia>,
-L$ .
^ (9-73)

The magnitude ratio and phase are

|(j1 = 1.0, iG = -wL (radians) (9-74)

These are plotted in Fig. 9-15. Dead time has the characteristic of a
pure phase shift since the magnitude is always one. It is easily seen
Sinusoidal Analysis 289
from the graphs that no other dynamic element builds up phase lag so
much as dead time. In combination with other dynamic elements the
phase lag of dead time is simply added to obtain the total phase lag.
A useful approximation for dead time L is the function from the
Fade list,^
2 -
G
U (9-^75)
2 4" IjS

The magnitude of this function is eveiywhere 1.0 and the phase lag
corresponds very closely to dead time up to frequencies given by
wL — 0.5. In addition, this function may be easier to incorporate into
complex system analysis.
Thermal process control may involve elements having distributed
parameters; that is, elements with distributed thermal resistance and
distributed thermal capacitance. Pneumatic systems having long trans-
mission lines may also be similar. As a typical example, consider the
heat transfer through a solid material such as a thick wall or tube.
The temperature distribution is governed by the equation,

d^c
(9-76)

where c == temperature at point x, time deg

X = distance into material, ft
= inverse of thermal diffusivity, sec/ft^
i = time, sec

Employing operational notation, a total differential equation results:

=
^+ (a^s)c 0 (9-77)

A solution to this equation is

r = (9-78)

Substituting in equation 9-77 and sohdiig for a,

a = ± (fr-79)

The solution is therefore, with two independent constants,

c = Ae^ + (9-80)

'
J. G. Truxal, Auio?naiic Feedback Control System Syntkme, McGraw-Hill Book
Company, Inc., New York, 1966, p. 560.
290 Autoinatie Process C'ontrol

From boundary oonditionB,

c « C|, X » 0

ac
0 ,
X =» L
dx

The second condition results because it is assumed that no heat loss

occurs at the inside (insulated on one face). The unknown coefiEicients

A and B may be evaluated from the boundary conditions, and the result-
ing solution is o
+_e^\
c = Cl (9-81)
+ 1 /

The temperature at the inside face can be calculated from

<=1 ( 9-82 )

The system function is therefore

Ci
_ ( 9-83 )

or in hyperbolic form
C2 1
(^84)
Cl cosh V —o?L^s
This is the system function for one particular distributed parameter
system. Other boundary conditions yield a similar function, usually
one multiplied by a complicated ratio of hyperbolic functions.
In order to expand the system function it is necessary to write the
complex number a in a different form, employing the square root of t.

V V -a^LHuf = Va^Z?w V -I
The operational function is therefore

5 1
( 9-86 )
Cl cosh /3(1 — t)

where 2/5* * a*L*«. This function may be expanded and the magnitude
and phase derived. The magnitude is

C2 I 1
(f)-87)
Cl (cosh* P cos* p + sinh* p sin* p) ^
Sinusoidal Analysis 291

The phase may be calculated from

= tan“* ( - tanh tan /}) (9^)

These are plotted in Fig. 9-16 where wT = 2/8*. The magnitude drops
off exponentially with the square root of frequency, and the phase lag

builds up indefinitely with the square root of frequency. Notice the

great similarity of the phase lag for dead time and the phase lag for
distributed systems. This leads to the Ziegler-Nichols approximation
for distributed systems

(9-89)
Tos -f 1

In this approximation a dead time and a time-constant element are used

to simulate the effect of distributed parameters. In most control prob-
lems this approximation is quite good.
Exothermic reactions are often encountered in process control. Such
systems have a heat-generating source, generally chemical in nature, in
which the amount of heat liberated is directly proportional to the
temperature. This situation is described by

c%=at
-OT +? (»-90)

where C = capacitance of system

c = temperature
m— heat extracted by cooling
q = exothermic heat = K^c

Kt is an exothermic coefficient. Combining these equations to eliminate

 ;

202 Automatic Process Control

the variable q, the system function results.

( 9-01 )

where T = C/K^ is the time constant. The transient solution to this

equation is, of course, an exponential with positive exponent. Thus,
the system temperature grows without bound on a sudden increase in
heat input.
The amplitude and phase of this function are easily computed
Kr'
|G| = (9-92)
Vl +
/tf = tan * uT
The relations are plotted in Fig. 9-17. It is important to note that the
phase begins at zero and increases positively. It must be remembered

Frequency, Phase margin

Fiq. 9-17. Exothe^raic reactions

that in exothermic systems it is generally necessary to add and extract

heat at different times. Consequently proper attention must be paid

to the sign of the function for heat addition \/{Ts — 1) and for heat
extraction 1/(1 - Ts). The magnitude ratios are the same but the
phase differs by 180 degrees.

Example 9-11. A pneumatic transmission line probably has a rl«ad time

proportional to the length of line due to the time for transmission of pressure
waves along the tube. Assuming a wave velocity of 1100 ft per sec, calculate
the phase lag per unit length of line.

PL
2irfi
^0
1100 ^ 2ir'

where I length of line in feet. Therefore

The result is easy to remember: one-third deg per ft per cps.

Sinusoidal Analysis 293
Euunplc 9-12. Calculate the characteristic time for copper, alu-
minum, lead, steel, and glass walls of 0.1-in. thickness.

Diffuuvity ft*/lir a*L* set

Copper 4.35 0.057

Aluminum 3.66 0.068
I^ead 0.92 0.27
Steel 0.57 0.44
Glass 0.013 22

Limiting, Dead-Zone* and Hysteresis Element!^

Many functions involved in automatic control are not describable

by elementary techniques, and they must be described by graphical or
other methods. This section is devoted to a description of three of
these elements.
The output a nonlinear element will not necessarily be sinusoidal
of
even if a pure sinusoid, but it will generally contain funda-
the input is

mental and harmonic waves. The frequency of the fundamental of the

output will often be the same as the frequency of the input. In this
case a describing function can be used and the element may be treated
as a linear element.
The describing function is a transfer function for an element having
slight nonlinearity and is expressed in terms of the amplitude and phase
of the fundamental of the output signal related to the amplitude and
phase of the input signal. The describing function can be obtained
analytically or experimentally.
Describing functions are generally useful in automatic control analysis
if the neglect of the harmonic in the output signal does not cause diffi-

culties. They may be used with adequate results when the element in

question is followed by one or more elements having appreciable

attenuation.
Limiting (sometimes called saturation) occurs wherever one of the
variables attains and cannot exceed maximum or minimum values. For
example, a control valve cannot do more than go wide open or tight
closed. This situation can be described by the frequency response of the
element. Referring to Fig. 9-18, the output of the element y plotted
against input x providing a statical calibration. The limits are given
at values -Xu The response of the system to a sinusoidal input
is also shown in Fig. 9-18. The output y follows input x to the point
at which limiting occurs This point is given by

Bi = sin
‘ — if Xa ^ Xi (9-93)
2d4 Automatic Process Control

wlicre 6i « angle at which limiting occur», radianb

Xa - amplitude of input signal x
Xi s amplitude at which limiting occurs

The frequency response of such an element is defined somewhat differ-

ently; the magnitude ratio is the ratio of the magnitude of the funda-
mental of the output to the magnitude of the sinusoidal input, and the

Fig. Sinusoidal re'^pon.se during limiting

phase is the angle of the fundamental of the output referred to the angle
of the input. It is not very difficult to write the Fourier series for the

output y and to compute the amplitude and phase of the fundamental

of this wave.* The phase is zero; that is, there is no phase lag between
input and output. The magnitude ratio is

\G\ = 2 (29/ + sin 29,) (9-94)

The magnitude ratio is not a function of frequency but depends upon

the amplitude of the input signal. Essentially, the gain of the element
decreases as the amplitude increases. Limiting in automatic control
therefore has the following effects:

1. If loop sensitivity is adjusted properly for large amplitude signals,

the loop sensitivity increases when the amplitude signal decreases and
insufficient stability may result.

2. If loop sensitivity is adjusted properly for small amplitude signals,

the loop sensitivity decreases when the amplitude of signal increases

and excessive stability may result.

* J. G. Truxal, AiitomcUic Feedback Conirol System Synthesis, McGraw-Hill Book

Company, New York, 1955, p. 568.
SinuBoidal Analysis

Dead tone, illustrated in Fig. 9-19, occurs in systems whenever there

isa range erf values of input signal to which the system does not respond.
An example of **pure” dead zone is difficult to cite; however, some auU^
matic controllers intentionally include a dead zone through mechanical
or electronic means. In a certain sense, dead zone is the opposite
of

limiting as may be seen in Fig. 9-19 where dead zone cuts off the

Fig. 0-19. Sinusoidal response and dead zone.

“bottom’* of the sinusoidal signal. Limiting cuts off the top of the
signal. In dead-zone action, the output signal begins at an angle given by

= 8in-’^‘' ifXa^X, (&-96)

where = angle at which output signal begins,

Xa = amplitude of input signal x,

Xd ~ amplitude at which dead zone begins.

As l>efore, the magnitude and phase may be calculated from the Fourier

series. The phase is zero. The magnitude ratio is

joj = - (t - %- Bin 2tfd) (9-fl6)

The magnitude ratio is not a function of frequency but depends upon

the amplitude of input signal. Dead zone in automatic control therefore
has the following effects:

1. If loop sensitivity is adjusted properly

for large amplitude signals,
decreases,
the loop sensitivity decreases when the amplitude of signal
and excessive stability may result.

2. If loop sensitivity is adjusted properly for small amplitude signals,

the loop sensitivity increases when the amplitude of signal increases
and insufficient stability may result.
 a

296 Automatic Process Control

Hysteresis occurs most often because of dry (ideal) friction in mechan-
ical systems, backlash and subsequent friction in gearing, electro-
magnetic effects, and so on. The action is described in Fig. 9~20. The

Kui U-20 Simaioidal resi)onse with hysteresib.

static calibration produces the characteristic hysteresis loop. The out-

put signal begins at an angle given by

Sh - sin“ (9-97)

wliere (9/* ~ angle at which liysteresis icvrrirMl occurs

X = amplitude of input signal x
Xk ~ amplitude of hysteresis reversal

Idle output signal continues until reversal of the velocity of input occurs.
The output signal then holds its last value while crossing the hysteresis
loop The inagiiitudo latio is

lG| = ^ \ \l + cos 29*)* + (it 4- 29/, + sin 29*)® (9-98)

27r

The phase lag is

/ 1 -f- cos 26ff \

/G = tan”' (9-99)
VtT -( 20,, 4- MD 20 J
I'luis hysteresis not only causes variable gain but aisu a phase lag.
'Fhe gain decreases with smaller input signals. The phase lag increases
for smaller signals. In automatic control these effects tend to cancel
each other on changing amplitude of signal, but the overall effect of

hysteresis is always to cause less effective control action.

Example 9-13. A control valve has 2 per cent dead zone. The amplitude
(>1 signal generally oc(‘urriTig is o per cent. Estimate the gain.
Sinusoidal Analysis 297
From equation 9-95,
Bd * 8m“‘ I = 0.41 radian

'fhe magnitude is

0.82 sin 0.82

1
^ -
1
1 0.51
IT IT

Testing Sinusoidal Response

Obtaining a frequency response requires special equipment and careful
testing. However, it is generally true that much information (!an be
obtained from even a»few very rough test values. The input sinusoidal
signal should be reasonably free from harmonics (distortion), but it is

not at all necessary that the input signal be purely sinusoidal. In fact,
a rectangular or triangular wave can also be emiiloyed.
The sinusoidal signal may have to have mechanical, pneumatic, or
electrical form. Thus a sine-wave generator is required. The frequency
(cycles per second) of the signal may be very approximately as follows:

Thermal processes 0.001 to 1 0 cps

Fluid processes 0.005 to 10 cps
Process control mechanisms 0.01 to 100 cps
Fast control mechanisms 0.1 to 1000 cps

It is probable that a signal generator cannot cover the entire langc of

frec^uencies (10®) required. Be^ow about 0.01 cps, a mechanical g .'ci-

at or is almost always employed. This consists of a variable speed drive,

a sine-cosine mechanism (cam-push rod usually), and a suitable
pneunuitic or electrical amplifier if necessary. The variable speed dri^ e
may be mechanical or electrical. Above 0.01 cps a suitable electric
generator such as an audio oscillator together with a suitable pneuiimlic
amplifier is satisfactory.
The recording of input and output signals is best done on a two-pen

electrical recording mechanism. For frequencies up to about 0.1 cps

ordinary process-type fluid or electric potentiometric jecorders may
be used. Up to 100 cps frequenc.y the recording oscillograph provides
better accuracy. Abo\ e 100 cps frequency the oscilloscope or photo-
electric must be used. It is sometimes convenient to
oscillograph
actuate the recording mecthaiiism paper drive from the sine-wave
generator in order that all periods of cycling occupy the same physical
length on the chait. This makes calculation of phase lag much easier

There are two methods of determining phasf' lag. First, simply record
input and output and scale the difference. Second, provide a variable
jjhasing of the recording of the input so that shifting phase of the input,
to match the input and output records enables reading phase lag directly.
 :

298 Automatic Process Control

In actually performing a frequency response test, it is always best to

explore the response at a fixed frequency giving about 20 to 50 per cent
attenuation but with several amplitudes of input signal. This is to
insure that limiting, dead zone, or hysteresis will be recognized. It is
best to record and plot test data as it is taken in order to save time
by not taking unnecessary’ frequencies.
All physical systems have some nonlinearities. Therefore it is best
to consider carefully the amplitude of input signal. A large signal will
undoubtedly saturate a part of the equipment whereas a small signal
may fall below frictional or dead-zone thresholds. It may be neces-
sary to try several amplitude signals in order to select a representa-
tive result.

PROBLEMS

9-1. Calculate and plot the amplitude and phase of the following func-
tions
1

(o) (6) (Ts + iy (c) Ta - 1

Tb
Tb 7’« + 1 1
(/)
Ts+ I 2Ts + 1 (r« + l)(27’s + l)

9-2. Sketch the phase of proportional-derivative-integral control for

Td * 0.1 Td = Tt, and Td — 10 T,. What effect does varying this ratio
have in control behavior?
9-3. Plot l/(Ps -f 1)* in polar form.
9-4. Calculate the amplitude and phase of l/(Th^ + !)• What kind of
system does this represent?
9-5. By plotting, compare the phase of e~^* and (2 — Ls)/{2 -h Ls).
9-6. A system with exothermic reaction and heat losses is described by
the following equations
Ccs = m q — ql

q = KdC

«X,= |c

Describe the type of system for all values of RK^.

9-7. A thermometer has ah apparent time constant of 5 see. The bulb
has a T*B-in. thick bronze wall. Does the characteristic time of the wall
seriously affect the response?
9-8. A system limits (saturates) at signal of 10 volts amplitude. A signa
of 30 volts amplitude is passed to the system. Calculate the attenuation.
9-9. Because the angle and the sine function are nearly the same up to

45 deg, show that the magnitude with limiting is given by iCj = -

T A| ^ •
 rh n/firr

Stability Analysis

Stability of operation of the process and automatic control system is

achieved when the deviation is maintained within predetermined limits.

Stability is necessary because without it the system may execute wild

oscillation leading to wear, failure, or destruction of parts of the system.
Good stability is difficult to define. Recovery with minimum area
applies to the control of processes in which time of departure from
the set point is of as great importance as magnitude of deviation. In
order to stabilize the controlled variable in the shortest time, optimum
stability must be obtained. In this way the smallest (juantity of product
is the least improperly processed.
Minimum deviation is a requirement for recovery of many controlled
systems where ill effects are caused in the process equipment or in the
product by momentarily excessive deviation. Here the magnitude of
deviation is of greater importance than time of deviation. The limits

of control are therefore generally narrow, and it is desired to remain

as close as possible to the set point.
Minimum oscillation is sometimes required where periodic disturb-
ances in the processing unit and plant must be avoided. If the output of

one process is carried to a second process, as often occurs in continuous

chemical processing, any cycling of the controlled variable becomes a
cycling load change to the second process. Greater stability of control
may be necessary in order to keep cycling at a minimum.
All the criteria for judging stability of control are derived from the
purpose and requirement for which automatic control is used. The
quality of control is, therefore, always relative in the sense that it must
necessarily depend upon factors of application. A satisfactory recovery
299
300 Automatic Process ('ontrol

and stability in one controlled system may be entirely unsatisfactory

in another.
The study of stability is greatly enhanced by the use of the La Place
transform. The reader is referred to a number of texts on the subject
and to a brief review in the appendix.

The Transfer Function

Thci analysis of a dynamic system provides a set of differential equa-
tions describing the performance of the system. For a simple system
described by a single differeniial e(|uation, for example,

( ^
+ -1 /
' fill \)
* 5,-'
the La Place transform of the equation is

As^^y - Asya ~ Aya -f Bs£v Bya -h C£y - (10-2)

where ya — value of y B.t 0

ya = value of dy/dt at ^ > 0

This relation may be solved symbolically in order to find the response of

the system,

-f (Ayivs -h .4i/n + Bya) (10-3^

As- A- BsA'C As^ + BsA-C
The re>i)onse of the system, oi rather the transform of system response,
is characterized by the system fui^ction,

/l.s'" 4- Bs + C
regardless of any initial conditions (yo, yo)* Suppose that the initial
conditions are zero {ya - 0, ya ~ 0), and tliat the system is disturbed
starting from “rest” condition.
1
£.r ( 10- 4 )
As^ 4- Bs + (

'I'he response of the system is therefore given by tlie product of system

function and disturbing funetion.
The transfer function is therefore defined as the ratio of the La Place
transform of the responding variable diviiled by the La Place transform
of the disturbance variable, the transforms taken witJi zero initial

conditions. From eipiation 1(L4, the transfer function is

Sy 1
Stability Analysis 301
Thus there no essential diffei'ence between the system operational func-
is

tion of Chapters 1 through 9 and the transfer function. An alternate

definition for the transfer functionis the La Place transform of the

response of a system to a unit impulse disturbance starting with zero

initial conditions. From equation 10-4, with a unit impulse distui'bance,

= 1
( 10 -6 )

the transfer function becomes

1

&is) = £( /) =
2 (10-7)
As- -h 5s 4- C

The transfer function concept is extremely important in dynamic sys-

tems analysis. In spite of seeming limitations imposed by initial condi-
tions it is nevertheless true that the transfer functions for a system com-
pletely define ike performance of (he system. In any case, it is always
possible to define the response and disturbances variables so that initial
conditions are zero.
One serious limitation of the use of transfer functions is that for some
kinds of systems the La Place transforms of the equations do not e/ist
because the difTerential equations contain essential nonlinearities. How-
ever, even in such cases approximations may be made, and the transfer
function ideas may
be employed. Furthermore, a transfer function
may often be determined experimentally even when they are didiculf
to deduce by analysis
Fluid and thermal process systems exhibit many difierent dynamic
characteristics, but all such systems may be de8cril)ed by employing
combinations of the following transfer functions:

1
("apacitance element (10-8)
Ts

1
Time-constant element (1(^9)
Ts -f 1

Oscillatory element, i
^ 1.0 (10- 10)
Th- f + 1

1
Dead-time elemeni (10- U 1

The time-constant element (equation 10-9) and the oscillatory element

(equation 10-10) are not necessarily fundamental but may be derived
through closed-loop action as shown in P’ig 10-1.
 1

302 Automatic Ptooen Control

p
“ r V + 2fn +
Fio. 10-1 . Generation circuite for time-couetaut and uacillatory elements.

The transfer function for dead time (equation 10-11) may be derived
by use of La Place transform. Consider a dynamic action such that the
disturbance (input) is

The transform is
^ = m ( 10
- 12 )

£x^£f(t) (10-13)

The only action by the element is to delay the function in time without
otherwise altering it. Thus the response (output) can be given by

»-=/(<- L) (10-14)

The transform may be written

t
ed II 1 «
(10-15)

Changing the variable under the integral sign, replace (/

— L) by t

£y = [ d{t + L) (10-16)
Jq
or
r•
me-'dt (10-17)
Jo

By definition of the La Place transform.

£y = (10-18)

The transfer function for dead-time action is, therefore,

G(8) (10-19)

Note that this function was derived in Chapter 2 by expansion in

Taylor’s series. Keferring to Fig. 10-2, it is common to designate the
 :

Stability Analysis 303

ttansfonns of a function of time by a capital letter, thus
C- C(») - £c(<) (lO-aO)

and similaily for all other vaiiablea

Fio. 10-2. The closed loop.

The Tnmsfer Function and Stability

The performance of the control system can be illustrated by the trans-

fer functions of the system (see Fig. 10-2)
Open loop B = {GiG^)E {NH)U (10-21)

Closed loop C= v+ ^ V (10-22)

1 + G\G2H 1 •+ GiG^H

D l-OiG2A+GiG2n N
Closed loop U (10-23)
1 + G G2 H 1 1 + G G2H
1

The controlled variable or deviation may be calculated once the form

of set-point function V or load function U are known.
The closed-loop transfer function for a system responding to a step
change of load with set point fixed is, from equation 10-22,
N
C U (10-24)
1 4" G G2 H1

In general, each term of the transfer function is a function of the com-

plex variable «. Substituting for these functions, whatever they may
be, would yield
+ Om-l^
OwS" OiS + Op
^

(10-25)
6„a" + bis + bo •

where 0^ and &n constants and m and n are integers. In any real
physical system, furthermore, n ^ m. The denominator of the transfer
function may be factored.

„ 0**’" + • •
•
aiS + Oo (10-26)
{» - ri)(« -Ti) (« - r.)
3f)4 Automatic Process Control
whi le are the roots of the denominator polynomial. Whether or not
any roots are zero, or alike in value, the inversion of this transform can
only yield terms like

C(t) = /(A, (10-27)

where r,i are the roots of the denominator polynomial and may be real,

imaginary, or complex numbers. The exponent k occurs wherever

there are equal roots. The number A is a constant and may occur for
zero roots. The time dependent terms, therefore, have the form

where rn = a + f/3, a complex number. Now, if the real part of the

root a is negative, the response of the controlled variable will be bounded
because only exponentials with negative exponents will appear. On the
other hand, if the real part of the root a is positive, the response of
the controlled variable will ultimately increase to very large values and
the system is said to be unstable.
The problem of stability in automatic control then reduces to the
problem of detennining whether the denominator of the closed-loop
transfer function has any roots with positive real parts.

Example 10-1. As an illustration, consider the transfer function

G =
" + ^

.S'‘
-f - 4
Factoring the denominator,
1
^
-f - 1)

Since one of the iactorH luis h pi>Ritiv(; root — 1), the system represented (.s-

by G IS unstable This is made evident by carrying out the following steps:

Expanding into partial 1‘ractions,

(\
a - d-
(.^ d- ,S- - f

The unknown cocthcicnt.s arc determined and substituted into the partial
fractions’

r ^ i {
9\(« +
^
2)*
2
5-h2^.s—l/
M
Using the La Place transform table to get the inverse transformations for a
unit impulse input,
c - - 2c“*' 4- 2en

Since the third term has a i>ositive exponent, c increases exponentially with
time, i.e., the output has no steady-state value, and the system is unstable.
Stability Analysis 305

Routh-Hurwiiz Condition for Stability

The Houth-Hurwitz condition for stability may be used to find

whether the characteristic diflerential equation for the system has any
roots with positive real pails.
The transient response to a change in load or in set point (see equa-
tion 10-25) is determined by the characteristic equation

-f
• •
-f k, - 0 (10-28)

Note that the Icft-halid side of this equation is determined by the

denominator of the transfer function.
The response of the system is determined by the roots of the idiarac-
tcristic equation 10-28 becaii.^e these roots determine the functions in

the solution to the differential equation 'fhe miinlxu* of roots eiiuals

the degree of the characteristic equation Instability can be deter-
mined at once if

1. All terms of the chariieteristic equation are not of the same sign.
2. All terms of the characteristie equation are not present in descend-
ing powers.

Either of these conditions may b(* detected by inspection and may l>e

used to deter mine stability

The procedure of Houth is a^ l(»llow.s Arrange the coef^lcl(‘^t.^ of the
differential equation in row and - i »hinin^

> 1 '^ ^

extending the rows until .di ^ im- j'I ^ •

d 1 In- (‘()(*tlir-i«*n(.s of the third
row are found by cross-mult jpIh'mi loi-

Using rous two and three, tlie coedii‘i« il- ul liu‘ louitii low are tound
by cross-multiplying in the same inannei-, and now rows are formed in

this way untilno terms remain. In tin* course of development, the coeffi-

cients in any row may be multiplied by a po.sitive number without alteir-

ing the result.

The system will be stable and non-osciilaU»rv m the steady state if

(1) all the coefficients of the characteristic erjuatiun.- are of the same
sign, and (2) all ihe derived c<x*friei(Mil'; of the first column of the anav
aie of oiu >ign
306 Automatic Process Control
There are two exceptions to the process:

(a) If the first-column term in any row is zero, but the remaining
terms are not zero, replace the zero by an arbitrarily small number e
and proceed.
(b) If all the coefficients in the second or any derived row are zero,
there are roots of equal magnitude lying radially opposite on the real-
imaginary plane. In this case form a polynomial with the coefficients
of the last row, take the coefficients of the derivative of this polynomial
for the next row and proceed.

The Routh-Hurwitz condition for stability is of limited usefulness in

automatic control because no indication is given of the amount of insta-
bility and the direction to go to avoid it. On the other hand, it is often
possible to explore the influence of one or two parameters of a system
by considering values that may produce instability.

Example 10-2. Employ the Routh-Hurwitz condition for stability for

(«* +s+

4X7-1X8 .5
1

4
2)(.< -h 2)(« -f 1)

,
«

4X4-1X0
SO
-f 4«* -f 7«*

7
+

4,
8s +4
4

0
= 0

5 X 8 - 4 X 4
^
^
5 5

The coefficients of the equation are all positive, and the left column is entirely
positive, so the system is not unstable.

Example 10-3. Examine the stability and determine any limitations of

parameter k for stability:

1 + - s + k){s + 1)2 = 2s* -f -1 4 i2k - l)s 4- 4* 1) -0

It is immediately apparent that k must be greater than +} to insure that all

terms are present and non-negative. Next construct the array:

2 k ib4- 1

3 2^' - 1

2 - * A: + 1

2ifc* 4" 4A: 4- 11

- 2

The third term of the first column requires that k < 2, and the last term
requires that k > Since these conditions cannot be fulfilled simultaneously,
2.
there is no value of k allowing stability.
Stability Analysis 307

The IVansfer Locus

The stability of control systems may be determined from the closed-
loop transfer function,

C* y ^ TJ
1 4 G1G2H
“ 1 + G1G2H
As stated before, a control system is unstable when the denominator of
the closed-loop transfer function has roots with positive real parts.
Study of this condition is made easier by considering the actuating
signal rather than the controlled variable,

£= ^
R- U (10-29)
1 4" G 1 G2 H 1 4" G G2 H
1

There is no loss of generality, because the stability of the input element A

or the load-element N may be determined separately. The condition
for stability now reduces to deter-
mining whether
1 -h G1G2H
has any roots with positive real
parts.
The transfer locus must be em-
ployed in order to determine wheth-
er the function 1 4- G G2 H
1 has roots
with positive real parts. In general
the function 1 -f G G2 H
1 is a func-
tion of the La Place transform com-
plex variable s:

I + Gi{s)G2{s)H(8) (10-30)

A complex variable is defined by

8 = a 4- iw (10-31)
Fig. 10-3. Roots in the complex
where a is the real part and o) is plane.
the imaginary part of the complex
number. Therefore, the function 1 + H
G 1 G2 is also a complex number
and may be represented on the complex plane.
The roots of (1 -f G 1 G2 H) are specific values of the complex variable a
and may be located on the s plane in Fig. 10-3. The roots may be ri
(a negative real number) and r 2 a positive real number by way of
example. Suppose that we define a vector (« — r 2 ) as in Fig. 10-3 by
drawing a line from the root a point on the imaginary axis. Suppose
308 Automatic Process Coutrol

now that the tip of the vector follows the dotted line in a counterclock*
wise manner where the path of the end of the vector travels from posi-
tive infinity down the imaginary axis to negative infinity and around
a circle of infinite radius back to the starting point. Thus, the value of
complex variable s changes from
-fioi as w— 4- 00

to
- Ki) as w — — 00

and back again. In this case it is desired that th*e real part of s be zero
(o = 0) so as to make this complete encirclement of the positive half
of the 8 plane.
For a single positive root (as r2 ) the vector s - r2 makes one counter-
clockwise rotation, but for a negative root (as ri) the vector s - ri
makes no net revolutions. Therefore it is concluded that a root can
be located by the simple scheme of determining the number oa'’ rotations
of a test vector when .s == lu in the particular path encircling the right
half plane. If there happens to be a root lying squarely in the imaginary
axis or at zero, it is possible to alter the path of 5 ever so slightly around
these points and proceed in the normal manner.
It is desired that s take on values given by tw, so

1 H- Gi(s)G2(ii)ff(s) 1 + G,(iu))G2(?:a))//(/w) (10-32)

and tkt resulting function can be plotted on the function ()lane as shown
in Fig. 10^-4. The problem may be transferred from the a plane of
Fig. 10-3 to the G(s) plane in Fig. 10-4 b, conformal mapping.
Suppose that the dotted ^ine in Fig. 10-4 represents G G2H
1
for dif-

ferent values of s — fw. A drawn from the origin 0 to the point P

line

on the locus r present-s the vector G 1 G 2 H. The line from the origin to
the (-1,0) j)(jint Q is the vector ( —1 ). By .subtraction the line joining
QP is the vector 1 4- G G2 G
1
Thus it is concluded that we ne(‘d only
plot the locus of the open-loop function,

Gj (lCo)G2 (fw)f^

and a vector from the point ( -1, 0) repre.sents the denominator of the
transfer function. The transfer locus is a plot of a function of variable a

for values of s' = iu from positive infinity to negative infinity. Foi'

convenu'm only the values from positive infinity to zero need be plotted
o,

because values for negative $ = iw are a mirror reflection about the

horizontal (real) axis.
The transfer locus is a closed curve and must be drawn with care if
the function G G2 H
1 contains an s” in the denominator (n may be cither
a positive or a negative integer) If n is positive, then the transfer locus
Stability Analysis 309
may pass to infinity as frequency w approaches aero, ^'hc positive fre-
quency end of the locus must then be connected by n/2 counterclockwise
rotations at an infinite radius in order to connect to the negative fre-
quency end of the locus and thereby form a closed curve.

Example 10-4. Plot the transfer locus for

s*(r« + 1)
310 Automatic Process Control
As shown in the figure, it is necessary to connect the ends of the locus by
making

complete counterclockwise rotation.

Example 10-5. Plot the transfer locus for

1
Q1O2H
»(Ta -h 1)

Im

As shown in the figure, it is necessary to connect the ends of the locus by

m^lring
n 1
2*2
complete counterclockwise rotation.

Nyquist Stability Criterion

The Nyquist stability criterion follows directly from a consideration
of the “poles'^ and ‘‘zeros** of the denominator of the closed-loop transfer
function (1 + G1G2H) and an inspection of the transfer locus of the
open-loop transfer function {G1G2H).
A “pole** of 1 -b G1G2H is said to occur at a value of s » Sp when the
value of 1 + G\G2H is infinite and if Sp has a nonzero positive real part.
Thus the function
1
1 -h G1G2H *
(3s - i)(«-h n
 1

Stability Analysis 31

has a pole at « 1/3. It should be noted that the value of 8 » ~ 1

+
is not counted as a pole because it has a negative real part. A pole of
1 + GiG^H is a pole of GiG2 Hy so that in counting poles only the open-
loop transfer function G G2 H
1 need be considered.
A "zero" of 1 -f G\G2 H is said to occur at a value of « « <0 when the
value of 1 + G\G2H is zero and if so has a nonzero positive real part.
For example, the function

t (3.

has a zero at «o = +1- Values of s = « are not considered as being

zeros.
The Nyquist criterion may be stated as follows: A control system
is stable only if the number of zeros of the function 1 -f G G2 H
1 is zero,

and if the sum of the net number of counterclockwise rotations of the vector
1 G G2H and of the
4- 1 number of poles of the function G 1 G 2 is zero. H The
vector 1 -h G G2 H is considered
1 to follow the transfer locus G G2H from
1

4^5= -foo toOto - <». Thus

AT +P= 0 (10-33)

where N = number of net counterclockwise revolutions of vector QP in

Fig. lQ-4
P= number of poles of G G2 H
1 (number of values of s with
positive real parts that make this function infinite).

then there are no zeros of 1 + G G2 H and therefore no

1 roots with posi-
tive real parts in the denominator of
1

1 + G1G2H
In many problems in automatic control it is interesting to note that the
number of poles and zeros is zero so that the Nyquist criterion reduces
to requiring no net rotations of the vector. The criterion then may
simplify to the statement that the transfer locus does not encircle the
(- 1, 0) point.

Example 10-6. Determine the stability of

- rfi
GiGfH does not have a pole so P = 0. The transfer locus in the figure shows
H- 1.0 counterclockwise rotations. Thus ^

* 41.0
AT

«
P 0
312 Automatic Process Control

and
iV -h P 0

Therefore, the system is unstable.

The Nyquist criterion may be applied whether G1O2H is analytically

determined or plotted from experimental results. The power of the
method lies in the fact that the analytical form of G1G2H need not be
known and that it is not necessary to factor high order polynomials.
There are two limitations which we have implied in the development
of the theory behind the Nyquist criterion:

1 The system is linear or can be represented by linear equations

.

so that theLa Place* ransronns may be applied.

t

2 The open-loop tiaii^l’c: function G\G2H approaches zero as the

.

frequency approaches infinity I'liis must be true in any real system

with a finite source of power. If the transfer function does not go to

zero for infinite frequencies, then means only that we have left out
it

some terms which are important at very high frequencies. It is correct

in this situation to merely draw in the extension of G1G2H so that it
ends up at the origin.

The procedure for determining stability by application of the Nyquist

criterion may be summarized as follows:

A. Plot the open-loop transfer locus GiG^H on the complex plane for
frequency varying from zero to infinity.
(a) If experimental data is iised^ plot the locus directly from the
phase and magnitude data.
Stability Analysis 3ia

(6) If the transfer function GiG^H is given, make sure that the
denominator is of order equal to or greater than that of the numer-
ator. If this is not true it is necessary to go back to the analysis and
pick up some of the higher order terms that have been neglected.
(c) If the locus does not terminate at the origin for infinite fre-
quency, then draw an arbitrary line extending the locus to the origin.
(d) Plot the locus for positive frequencies. The plot for negative
frequencies will always be the mirror image of the plot for positive
frequencies, Reflected about the real axis.
(e) If there are rf factors of s in the denominator of GiG^Hy then join
the open ends of the locus of G G2 H with
1 nr radians of arc (at infinite
radius), going counterclockwise.
B. Draw the vector from — 1 to a point on the locus. Let the point
vary from «=-!-«> to <*>
= — and count the net number of counter-
clockwise rotations of the vector. This is N.
C. Determine the number of poles in the function Gi(j 2 H. 1'hese
are the number of values of a having positive real parts (counting each
repeated value) that make 1 + G G2H infinite.
1 This is P.
D. The system is stable if and only if

AT +P- 0

The complex plane diagram can be used for more complete investiga-
tion of system performance short of solving for transient response For
example, it is often required to find the closed-loop performance know-
ing the open-loop performance. For any control system,

(open loop) (
10- 34 )

and
C
(closed loop) ( 10 -36 )
V
The analysis is made a great deal easier it the inverse functions are used
The inverse of a vector is obtained by inverting the amplitude and chang-
ing the sign of the phase. Let

(10-36)

Then
V
(10-37)
C
Ck>nBequently, as shown in Fig. 10-5, all that is necessary is to plot the
function Y and the function H/A The output-input response C/V is
.
 . Automatic Process Control
rfl4
this meUi^
the final locus. Unfortunately
eaaly obtuned by inverting
measuring-element transfer function
is useful if the feedback or

Fxo. 10-5. Inverae polar plot.

process control only when

the
is 1.0.This situation occurs in automatic
is zero.
measuring element (feedback) lag
Determine the stability of
Example 10-7.
B K
B “ (108 + !)•
Stability Analysis 315
with iT » 1.0 for all frequencies. The function for negative frequencies (the
dotted portion of the curve) is the reflection about the real axis of the portion
of the curve for positive frequencies. Thus only the portion for positive fre-
quencies requires calculation.
The system is stable because for any value of K
the net rotation of the
test vector is zero and there are no poles. Notice that the only effect of an
increase in the number /fT is to expand the plot radially, or conversely, to
contract the radial coordinate markings. Thus, if the point ( — 1, 0) is located
at the first circle outward, the gain Kis 10 and the system is still stable.

Example 10-8. Determine the stability of

*
B 10
E" (lOs + D*
This function may consist of proportional control (sensitivity 10) of a system
with three equal time constants of 16 sec. The function is plotted in Ex.
Fig. 10-8. The locus passes just beyond the point (— 1, 0). The net rotation

of the test vector is two (iV = 2) revolutions, and since there are no poles
the system is unstable. If the proportional sensitivity is reduced to about 2.8,
the system becomes marginally stable.

Example 10-^. Determine a maximum value of proportional sensitivity

for stability of the function

B iCCeOs -f 1)
E* 60«(120s + l){30s -h 1)

This open-loop function may represent proportional plus integral control

(integral time = 60 sec) for a system composed of two time, constants (120 sec,
30 sec). The function is plotted in Ex. Fig. 10-9. For positive frequencies
the locus begins at 90 deg lag and infinite magmtude. Th^ at zero in-
quency the locus is tangent to the n^ative imaginary axis (this is due to the
integral response 60s in the denominator). For negative frequencies
the locus
316 Automatic Process Control
becomes tangent to the positive imaginary axis. The problem of joining the
two loci at sero frequency is resolved as follows: For each s term in the derum-
intdoTj eonJtinue the locw one half revolution in the eounterdoekwiee directum to

form the complete locus.(«*'•

requires one complete circle, s* one and one-half
circles, etc.)As shown in Fig. 10-8 the locus makes a half circle of infinite
magnitude at o) « 0. The test vector however makes no net rotations and
the system is stable for any value of proportional sensitivity K.

Example 10-10. Determine the maximum value of proportional sensitivity

for
B K
E c‘*(20» -h 1)
Stability Anulysis :{17

This fuDction may represent proportional control of a system having a dead

time of 5 sec and a time constant of 20 sec. The open-loop function is plotted
in Ex. Fig. 10-10 for a proportional sensitivity of iC *= 1.0. Notice that the
dead time causes unlimited phase lag and as a result the locus spirals counter-
clockwise and outward as frequency decreases. The limit of proportional
sensitivity is about 7.0 at a frequency of 0.007 radians per second as the locus
passes through the Nyquist point ( — 1,0).

Example lO-ll. Determine the stability of

B ^ -2
^
E (K- 1)(4.+ 1)

This function is plotted in Ex. Fig. 10--11. The net counterclockwise rotation
of the test vector is zero and N= 0. Since there is a pole in the right half
plane at s = 1.0, therefore P= -fl O. Thus
N -{-P 9^0

The system is therefore unstable. In addition it should be noticed that the

function 1 + O G 2 H has a zero at
1
= (3 ± V^)I2 and this fact alone makes
the system unstable.

Bode Mapnitude-Phase Method of Analysis

The magnitude-phase method of analysis may be employed to deter-
mine the performance and stability of an automatic control system.
It is undoubtedly the most widely used method of aiialysi.s arul syn-
thesis of control systems. It is also one of the simplest methods.
The Bode magnitude-phase method of analysis may be used for

a wide range of problems, including the analysis of systems containing

some nonlinear elements. Generally, however, it more useful for
i.**

sy.‘^tems having linear response.

318 Automatic Process Control
In order to demonstrate the selection of stability by the magnitude-
phase method consider the simple second-order closed-loop system for
which the open-loop response is

Tb{T9 + 2{)
The open-loop magnitude is

i
lei
«7’V(4f* +
and the open-loop phase is

/G = tan ^

Employing these relations it is possible to solve for the closed-loop

damping ratio for any selected open-loop phase at a frequency o) for which
the magnitude is one. The results are:

Open-Loop Phase, Closed-Loop Closed-Loop

deg Damping Ratio, Period of Oscillation
-105 0.95 20.0T
-120 0.61 7.9T
-135 0.42 6.9T
-150 0.27 6.6T
-165 0.13 6.2T

Therefore, a selection of open-loop phase will determine the closed-loop

behavior for a second-order system. Generally it is desirable to select
a damping ratio of about 0.3 to 0.5 and thus cm open-loop phase lag of
135 deg seems appropriate.
The response of higher order systems may be investigated in a similar
manner. It is usually found that here too the phase should be about
136 deg for a reasonable system response. If the system contains some
nonlinearities, it is advisable to check stability by more than one method.
For closed-loop systems the analysis need only employ the open-loop
response
'
B = GiG2HE~\-NU (ia-38)

For the action within the loop, the influence of the load variable is not
important, so

= (10-39)
I
Stability Analysis 319
The maisnitude ratio and phase,

are plotted on log-log coordinates.

The amplitude plot is made on log-log coordinate paper, with fre-
quency (q> radians per second or / cycles per second) as the abscissa
and magnitude ratio as the ordinate. The amplitude function is ob-
tained directly from the transfer function,
‘
!(?,(«) G»(*)ff(«)|. by (1(M0)

or, since amplitudes must be multiplied,

X l(?a(to)l X lH(t«)l (1(M1)

On log coordinates, however, multiplying magnitudes means adding

ordinates. Thus, the magnitude of several elements in series is obtained
most easily by plotting each element and adding.
The phase function is plotted separately on semi-log coordinate paper
with frequency on the log coordinate abscissa and phase in degrees as
the ordinate. The phase function is obtained directly from the transfer
function:

or, since phases must be added,

/Gi(iw) -f /G2(io>) + /H(iw) (10-43)

The stability and performance may be determined from the graphs

by the following rules:

1. If the phase lag is less than 180 deg at the highest frequency for

which the magnitude is one, the closed loop is stable. Conversely if

the phase lag is greater than 180 deg, then the system is unstable.
2. If the phase lag is approximately 135 deg at the highest frequency

for which the magnitude is one, the closed-loop system is stable.

Furthermore, the closed-loop response will be damped and slightly
oscillatory in a ^^satisfactory” manner. Sometimes the analysis may
be made by employing the magnitude-frequency diagram without
reference to the phase lag. It is best however to use role 3 as a check
on the method listed above.
3. If the magnitude-frequency curve does not change slope appreci-

ably over two octaves of frequency near the highest frequency at which
the magnitude is one, and if the negative slope of the magnitude fre-
320 Automatic Process Control
queiicy curve is less than 2.0 (plotted on log coordinates), the system
is probably stable. Conversely if the negative slope is greater than 2.0

then the system is probably unstable.

4. If the magnitude-frequency curve does not change slope appreci-
ably over two octaves of frequency near the highest frequency at which
the magnitude is one, and if the negative slope is 1.5, the closed-loop
system is probably stable. Furthermore, the closed-loop response will
be damped and slightly oscillatory in a ‘‘satisfactory” manner.

As an example consider the following system function:

C _ lU
(10-44)
E “ (16«+ 1)^

This function could represent proportional control of a three-time-con-

stant process. The proportional sensitivity of the controller is 10 and
the process time constants are each 16.0 sec. The magnitude is

C\ 10
(10-45)

The phase is

-3tan~'a>7’ (10-46)

The magnitude and phase are plotted in Fig. lQ-6. Stability checks

1, 2, 3, 4 above are as follow.^:

1. The system is just unstable because the pha.se is slightly more than
180 deg when the magnitude i.s oiie (190 deg at 0.0191 cps)
 : :

Stability Analysis 321

2. Satisfactory stability is achieved if the proportiimai sensitivity is
made equal to 2.83. This is easily checked by sliding the magnitude
curve vertically downward until it passes through the MR » 1.0,
/ « 0.01 point. The magnitude is then 1.0 when the phase lag is 135 deg.
3. The magnitude-frequency curve is reasonably flat near a frequency
of 0.01 cps and the slope is 2.36 at that point consequently the system
is probably unstable.
4. Satisfactory stability is obtained if the proportional sensitivity is

made about 2.6. This

found by sliding the magnitude curve down-
is

ward until the slope i^ 1.5 at the point at which the magnitude is 1.0.

For a proportional sensitivity of 10 the transient response has an

exponential factor and a slowly growing oscillatory factor. The solution
of the differential equation for clojsed-loop behavior is

c = -f 35e®
o®®'
sin (0.1 16f - 31®) (10-47)

A unit impulse disturbance was assumed and the roots of the character-
istic equation found by trial division

(5« + l)(820s2 ^ 10« -f 11) (10-48)

As the solution in equation 10-47 shows, the response is definitely

unstable because of the positive exponent of the second exponential
term. This produces an ever increasing amplitude of oscillation
For a proportional sensitivity K of 2.83 the solution of the differential
equation for the same conditions becomes

c = 170c-® + 170c-® ®*®‘

sin (0.077< - 90®) (10-49)

The roots of the characteristic equation were found by trial division

(G.7c 4- 1)(6105^ 4 23c -h 3.83) (10-50)

yiQ, 10-7. Traneient responac of s\ stem of Fig. 10-6,

322 Automatic Process Control

This solution is a stable one, as shown in Fig. 10-7. The damping ratio
of the oscillatoiy component is 0 25 and is generally conddeied
satisfactory.

Example 10-12. Determine the integral time for integral control with
two time constants and 0.625 sec. The open-loop function is
of 2.5 sec

C 1

E !r«(2.5« + 1)(0.626« -h D
The magnitude and phase are

—
wT \ (1+
/
6.25a)*) (1
^

+ 0.39a)*)
—90* — tan“^ 2.5a) — tan“^ 0.625a)

First, the phase angle does not depend upon the integral time T, The fre-
quency for 135 deg phase lag is easily calculated;

0.28
For a magnitude of 1.0,
1
ojT
v'(l + 0.49M1 + 0.0306)
and
1
2.9 sec
0.28 X 1.24

The integral time should therefore be about 2.9 sec.

Example 10-13. Determine the proportional sensitivity for a control sys-

tem having a dead time of 5 sec and a single time constant of 20 sec.
The open-loop function is

C K
""
E 6»*(20s -h 1)

The magnitude and phase are

A'

Vl +400u*

^ - (tan-' 20«) - (Su, X

For 135 deg phase lag, the frequency o) must be about 0.20 radians f)er sec.
The amplitude must be one, so

If = Vl + 400 X 0.2’ = V17 = 402

Stability Analysis 323

PROBLEMS

IM. Check the following Bystems for stability. The function given is

1 + GiGiH.
(o) 6«» + n<* + 6« + l (c) lOOsH 210»‘ -I- 121s* + 22s + 1

(6) s» + s* + s + 21 (d) rs» + rs* + s

10-2. The Ziegler-Nichols approximation for a process is

The proportionaMerivative control equation is,

A’,(r^ + 1)

Show that the best settings are nearly

as given in Chapter 4.

10-3. A proportional controller is used for a system having n equal time

constants, Calculate the open-loop gain Kc for 135 deg phase lag with

(?.+ !)•

and the loop closed. Does the open-loop gain depend upon the value of the
time constant?
10-4. A controller and process function are

Aj
c+^'TsfT's -I- 1)

Calculate the open-loop gain Kc for best estimated res(x>nBe.

 f

apppndix

Automatic Control Definitions

Actuating signal^ is the reference input inii^us the primary feedback.
Automatic controller is a mechanism which measures the \'alue of a
variable quantity or condition and operates to correct or limit deviation
of the measured value from a selected reference.
Capacitance C, is the amount of energy or material which must be
added to a closed system to cause unit change in potential.
Control elementfi, Gi, comprise the portion of the control sysir;. ^
hi(‘li

includes control action relating the manipulated variable to the actuating

signal.

Control ratio, c/r, is the response of the controlled variable to a change

'

of reference input.
Controlled system, G2, is the body, process, or machine, a particular
quantity or Condition of which is to be controlled.
Controlled variable, c, is that quantity or condition of the controlled
system which is directly measured and controlled.
CofUrolling means are the elements of an automatic controller which
are involved in producing a corrective action.
Corner frequency of a transfer function is the frequency at which hne>
asymptotic to its log-magnitude vs. log-frequency curve intersect.
Dead time, L, is a fixed inter\^al of time between the change of an input
to an element and the beginning of response to the input.
Dead zone, Xj, is the largest range of values of the input variable to
which an element does not respond.
Derivative action is a controller action in which there is a continuou-s
linear relation between the derivative of the actuating signal and output
signal of the controller.
325
326 Automatic Process Control

Derivative time is the time difference by which the output of a propor-

tional-derivative controller leads (is ahead of) the input when the input
changes linearly with time.
Deviation, d, is the difference at any instant between the value of the
controlled variable and the set point.
Differential gap is the two-position controller adjustment: the smallest
range of values through which the controlled variable must pass in order
to change the output signal of the controller from maximum to minimum.
Disturhancr, w, is a signal (other than the reference input) which tends
to affect the value of the controlled variable.
Error ratio, e/r, is the response of the closed-loop system actuating
signal to a change of the reference input.
Feedback elements, H, comprise the portion of the controller which
establishes the relationship between the primary feedback and the con-
trolled variable.
Final control element is a portion of controlling means which directly
determines the value of manipulated variable.
Frequency response of a system or element is the steady-state magnitude
ratio and the difference in phase of the output with respect to a sinus-
oidal input.
Gain of a system or element is the ratio of magnitude of the output
with respect to the magnitude of sinusoidal input.
Input elements, A, comprise the portion of the controller which pro-
vides the reference input in response to the set point.
Integral action is a controller action in which there is a continuous

linear relation between value of controlled variable and rate of change

of controller output signal.
Integral time, is the time required for the output of a proportional-
integral controller tochange an amount equal to the amount of propor-
tional re.sponse provided by a step change of actuating signal.
Lag is the retardation or delay of one physical condition with respect
to some other condition to which it is closely related.
Load change is the change in process conditions which requires a change
in the average value of manipulated variable to maintain the controlled
variable at the desired value.
Load error. See Offset. ,

Manipulated variable, m, is the quantity or condition which is varied

by the automatic controller so as to affect the value of controlled variable.
Offset, Ec, is the steady-state deviation resulting from a change in
v«luc of the load variable
On-off action. See Two-position action.
Automatic Control Definitions 327
Open^-Ufop gain, h/e, is the ratio of the change in the feedback variable
and the change in actuating signal.
Phase margin
is the angle by which the phase of the open-loop response

of an element or system differs from 180 deg.

Primary feedback, is a signal which is a function of the controlled

variable and which is compared with the reference input to obtain the
actuating signal.
Proportional action is a controller action in which there is a continuous

linear relation between value of the controlled variable and the value of
the output signal of the controller.
Proportional sensitivity, Ke, is a proportional action adjustment. It
isthe steady-state ratio of the change of controller output variable and
the change in actuating signal.
Rangeability is the ratio of majdmum flow to minimum controllable
flow through a final-control element.
Raie time. See Derivative time.
Reset rate is the inverse of integral time.
Reference input, r, is a signal established as a standard of comparison
for a feedback control system by virtue of its relation to the command.
Response time of a system or element is the time required for the out-
put to first reach a specified value after the application of a step input
or disturbance.
Rise time of a system or element is the time required for the oui^mt to
increase from one specified percentage of the final value to another,
following the application of a step input.
Servomechanism is a feedback control system in which the controlled
variable is mechanical position.
Set point is the selected reference value of controlled variable which
it is desired to maintain.
Settling time is the time required for the absolute value of the difference
between the output and its final value to become and remain less than a

specified amoiuit, following the application of a step input or dis-

turbance.
Static error. Eg, is the steady-state deviation resulting from a change
in value of the set point.
Time constant is the time required for the output of a first-order
system to change from a given value to within 36.8 per cent of the final
value (or 63.2 per cent of the amount of total change) when a step change
of input is made.
Transfer locus is a plot of magnitude and phase angle of sinusoidal
response of a system.
 FtG.

A-.

Transient

+
rntimnw

2rr..c

?r
>
do
system

-
0
to

a
step

input.

(oj^noQ ffiaoaij dv^mio^ny

Transient Response 329

input.

impulse

unit

a do-VT’c

to

system

co»0.

second-order
with

0
=
h
of c
-f

response

2f
+
rc*c
Tranrient

A-2.

Fig.

» 'ap.-i^iuSe^
Sinusoidal Response 331
332 Automatic Process Control

La Place Transformation
The solution of linear integro-differential equations by ordinaiy
methods requires obtaining (a) a characteristic equation, (&) the roots
of the characteristic equation, (c) the complimentary solution, (d) par-

ticular integral, and (c) the known initial or boundary conditions in

order to evaluate the undetermined coefficients of the final equation.

The La Place transformation permits a change of variable from the
time domain to the complex variable or “s” domain. The resulting
functional expression can then be manipulated by the laws governing
ordinary algebra even though is a complex variable. After the
desired manipulations are performed an inverse transformation returns
the original function expressed in the s domain back to its original

domain in a more useful form.

The “direct” transformation is made using the integral

.Y(s)=j|| x(t)e-“di (A-1)

where the function z(t) is in the time domain before transformation and
.Y (s) is in the s domain after transformation. S 5mibolically this trana-
formation is written

X{s) = £ 2(0 (A-2)

Equation A-2 is read “Y(«) equals the La Place transform of x(t)"

The La Place transform of a function x(t) may be obtained by sub-
stituting in equation A-1 and integrating. For example, if

. 2 . (0 = t

then the La Place transform is obtained from

Y(,v) = /
(U
Jo

Performing the integration by parts,

X{») = f-
p
Thus, it is possible to evaluate the transform by direct integration. In
order to save time a table of transforms, such as Table A-1 or Table A-2
may be used. The table of direct transforms is a set of functions that
may be transformed by direct integration as may be seen in Table A-1.
The inverse transform table is simply a set of solutions with arbitrary
coefficients for a corresponding set of simple differential equations.
 )

LaPIace Transformation 333

The traiisfomis for all the terms appearing in the difTerential equation

02^(0 4- aii(<) 4- aox(t) = yoit(t)

can be found using the theorems of Table A-1. The disturbance u(l)
ife the step function

u(t) * 0 , / < 0
ft(i) = 1, / ^ 0

The transformed result is an equation in the variable which may be

solved algebraically foj X(s).

-
(a2«’ -f njji -f ao)X{s) - //o -f )
4* ni.r(0-f

Two quantities appear in the transformed equation which are \'alues

of the variable x and its derivatives at t = Od*. These quantities are
the initial conditions of the problem and should be evaluated immediately
in terms of given information before a solution is found by an inverse
transformation. This is one of the impoHant propertias of the La Place
transformation in that all of the terminal conditions of the problem are
accounted for during subsequent manipulation.
By solving for the ^'ariable X (.s) there results

+
A («)
= yo 4- (a2S^
7“tt:
«(a2«4- ai» —T—
ais)x(()+) 4- O2 «j:(0+)
4- uo)
^

'riiib function repre.>ents the La Place transfonn of the variable x(,l).

dlie La Place transform may now be inverted by evaluating the

integral,

X(sK'<ys (A-3)

This requires integration with respect to the complex variable s. This

step is usually somewhat more difficult than direct transformation and

can be made unnecessary by employing a different procedure. If the

function X(5) can be expanded in a series, the inverse transforms may
be found fiom the relatively simple polynomial of each term of the
expansion. The transform
bmS"'* 4-
^4- • •
4- fee
X(s) = 4-
(A-4)
4- •
ais 4“ ao

can be expanded by factoring the denominator

mfi'” 4- 4- • ' •

X{8) = - -
(A-6)
(s - ri)(5 - r2)(« r3j • •
•
(6 r„)
334 Automatic Process Control

Tiililc A-1. Operatioo and Function IVanaforms

Operation or
Function Function fit) Transform F(«)

Derivative -m
d*
^/(O «*F(») - ^(0) -/'(O)

-.«y(o) - »/'(o) -rm

Integral im dt ;?(»)
8
+ ;/-*(0)
8

Translation f(t - a) e-^‘F(s)

Unit impulse (Dirac or delta function) 1

1
Unit step u(t)
8

1
Ramp t

n/
r gfi+i

OJ
Sinusoid sin u)t
«* -f

8
cos 0)1

1
g-df
Exponential
« -h a

Initial value lim fit) lim 8F(s)

<-*o »- •
1

Final value lim /«) lim sF(8)

r-*0

where the roots (rn) of the denominator are found, including zero and
repeated roots. The transform can now be expanded in a series of
terms,

fxi») /t(») /n(«)

Xi») (A-6)
(s - ri) (s - rj) (s - »„)

Cionsidering the terms one at a time, the inversion may be made by

recognizing a series of relatively simple inversions. These are compiled
in a table of transforms as in Tables A-1 and A-2. ^
LaPlace Transformation 335

Table A-2. Inverse Transforms

f'M /«)

1 Unit impulse, A(0 « Urn

o-*o a
—
1
Unit step, u(0
8

(Ts + 1)" • r"(n ~ 1)1

1
/ sin \/r^«/7')
rv+2f?’s+i T’Vi-r*
«-f"’'sin(V'l-rH«/7’)-*l
1

«(T*s*+2frs+l)
where ^ - tan *
^

(7’i/7’*)e-‘"''

l-2f(7’,/7’)+(r,/7’)*
1 (7’)-‘e-'"'8in |vT^«/r)-*l
(r,«+l)(7’V+2fT<i+l) + (l-mi-2f(7’/7’,)+(7'/ri)‘l^
. . .
(T'./T’Ov/i-r’
wher«*=ten'
,

1 rj-2f(r2/7’)+(i’,/r)qx._„„
tL 1-C J
T 2 S-I-1
8in(Vl-f*(</7’)+*)
(rv+2r7’«+i)
,
(7’t/r)Vr=r
where
,
<^-tan ,

|l-2f(r2/7') + (7’2/r)»)We-f'’’
TasH-l
sin l\/rT*(«/7’)+'*’l
<i(TV+2f7'»+l)
. . ,
.(Tt/T)Vi-r ....yi-f’
where-l-tan
-f

The partial expansion fraction may be made as follows.

1. For a number of district roots (no repeated roots), factor the

polynomial and write

iifl
B(8)
_di_.
S - Si
+ -A_
S - S2
+
S
^
“
(A-7)
33G Automatic Process Control

Note that a /oro root may be treated exactly like a nonsero root. In
order to solve for the A's, multiply the transform by the denominator
factor and set s equal to that particular root. Thus,

AW
® (s - Si)(s - 82)(8 - «4)
• *
•
(« - «n) «

Example A-1. Find the inversion for

A(s) a-l- 1
B{8) (.s + 4)(i» + 3)(aH-2)

_ Aj Az A4
.s + 2 fi -h .3 s + 4

.s-
+ 1 -1 I

(A- -h 4)(.S- -h 3) 2 X 1 2

«+ 1 -2
Az ^ -1-2
(8 -h 2)(s + 4) -1 X 1

5-h 1 -3 3
Aa
(« + 3)(s + 2) -^l X -2 2

The desired expansion

A{&)
fi(«)
is

2 (« +
12 2) -f 3)
^
3
2
1

(8 -f 4)

The inversion is (Table A- 1 or Table A-2)

—
m = -2- + 2*
‘ip-

2. For m repeated roots with (n - rn) distinct roots, write the re-
peated root to each power:

A (s) ^ Ai An—m Bfn

B{8) 8 - 8i
^

8 - Sn_m (« “ Sm)*"

Bj
-H 4. ...
(« - s^r-^ is - an.)

(A-8)

In order to solve for the A^s, multiply by the denominator factor and
set 8 equal to that particular root as in the previous case. In order to
solve for the B% a derivative is necessary,

A (a)
Bm = - -
(a ai)--*(s

^
ds (fi — ai)
• •
• (a - 8nr-^) Mm
Differential Analysers 337
Example A-t. Find the inversion for
A(s) *4- 3
B(8) a(«-|-2)(«-|-
D*

'

+ O2 ^-
«
« /
(i
• N*
+ 1)*^
'

(s-l- 1)

Ai = U + 3X I
3
(.^-h2)C.-h 1)*U 2

A, =
a + 3
I
1

.V 4- 3
- - -2
4 2)

B. =
d &“
43 -1
d« s(fi 4 2)
The desired expansion is

A is) 3 1 I

Bis) 28 2(s4 2) («4 1)2 (« 4 1)

The inversion is

/«) = - - 2te-‘' - c- Mn£.)

^

list* of Differential Analyzers

The differential analyzer in one of its forms, mechanical, electronic
analog, or electronic digital, has become a common engineering tool for
the analysis of automatic control systems. By employing the differential
analyzer the various portions of a control system may be studied and
the results obtained in either the sinusoidal or transient response form.
Discussion in this text will be directed mainly toward use of electronic
analog computers, but the general discussion and results apply equally
to forms of computer.
all

The
electronic analog computer employs feedback d-c amplifiers and
capacitors and resistor networks to accomplish mathematical manipula-
tion. These elements are shown in Fig. A-5. The d-c amplifier consists

Megohrnt Microfarads

Fio. A-6. ADalog computei elements

338 Automatic Process Control

usually of ordinary triode tubes with resistance coupling. The input

direct voltage is applied to the grid of the first stage, and the output
voltage is generally taken from a cathode-follower output section. The
operation of the amplifier is such that

6o
= -Keg (A-9)

where e© = output direct voltage with respect to ground

K= amplifier gain (usually from 10® to 10®)
eg
— input voltage with respect to ground

The o\itput voltage must be negative relative to the positive input

voltage. The capacitors and resistors arc high quality components
having small h*alvagc and small temperature coefficient.

The coefficient circuit is shown in Fig. A-6. All direct current passing

Coefficient Integration Addition

Fkj. A-(>. Bhkic computer circuits.

through resistor /i\ passes through resistor R; because the grid current
to the input of the amplifier is negligible. Then
ex - eg = (A- 10)
and
eg eg %R f (A>11)

Combining erjuations A-9, -10, -11 to eliminate current i and grid

voltage Cg,

(A-12)

If amplifier gain K is greater than about 1000, and if the resistance ratio
R//Ri is not greater than about 10, then the equation above reduces
to, with small error,

(A-I3)

Actually the amplifier gain is usually inuc\h higher than 1000, so that
the error is less than a small fi action of one per cent. As the last equation
show&, this circuit may be used for multiplying by a fixed coefficient.

The coefficient is st^t by selecting the ratio of resistances.

Differential Analyzers 339

The inUgraiing circuit is also shown in Fig. A-6. For the resistor and
capacitor
1 11 (A-14)

- «•) - * {A-16)
5
Eliminating current i and grid voltage €g by employing equation A-9,
there results

Again if amplifier gain is high enough we may write with little error,

e.= -/~* + K' (A-17)

Thus this circuit provides integration. The initial value of the output
is the constant A"' and is represented by the initial voltage charge across
the capacitor. T ntegration is always accompanied by a “scale coefficient”
of l/RC. By selecting values of the time constant RC a change of time
scale may be made. Scaling is discussed in a later section.
The adding circmi of Fig. A-'6 may be analyzed in a manner similar
to the coefficient and integrating circuits. With current ii in the
resistor Ri and current ?2 through resistor the equation becomes

Again if
K' I
the amplifier gain
i;)]
"
©
K is large and if all resistances are equal,

Co = -Cl - C2 (A-19)

Thus the summing circuit simply adds voltages while changing their
algebraic signs.
The combinations of coefficient, addition, and integration now allow
solution of ordinary linear differential equations with constant co-
efficients. A voltage input function corresponding to the desired dis-
turbance function is required. A means of indicating or recording direct
voltage is also necessary in order to display the solution. The solution
of equation proceeds as follows:

1. Solve for the highest derivative term of the equation,

2. Set up the analog cir^it for all terms on the right of the ecjuation.
3. Close the loop to compel agreement of left and right sides of the
equation.
 340 Automatic Process Control
4. Check to see that the number of amplifiers in each and every closed
loop 18 odd.
5 Check to see if circuits can be simplified.
Eiample A-3. Given the equation

sf + *-»«)
Solve for the highest derivative:

The circuit is displayed in Fig. A-7 in sequence of drawing. First start

with a line in which the voltage represents the velocity (highest derivative).
Passing through an integrating circuit provides minus x (ignore initial con-
ditions for the present). Next add yif) and minux x by using the adding

first step

circuit. Then multiply by ^ using the coefficient circuit. This yields the
right-hand side of the equation. Close the loop by connecting the output
back to the input, thereby comjselling agreement. Note that there are three
amplifiers in the loop.
The disturbance voltage is introduced at y(t) o^nd may be a 20-volt change
in voltage representing a step change. The cutput (minus z) is recorded or
indicated and is the solution to the equation. As will be shown later this
circuit may be simplified.
Differential Analyzers 341
Example A-4. Given the simultaneous equations
* * *(0 +y
22/ -f 2y = a: +i
The equations are solved for the highest derivatives:

y = *(0 — X
—i » X — 2y — 2y
Next draw the each equation and interconnect the circuits. The
circuit!^ for
result is shownin Fig. A-8. Notice that the amplifier 6 in the lower row is
used merely to change signs. This is often necessary in computei problems.
Note that each of the three loops contains an odd number of amplifiers.

When the actual analog circuits are drawn out, a short study often
indicates that certain combination functions could be employed and
some amplifiers eliminated. For example, the circuit in tlieupper row
in Fig. A-9 could be simplified to one circuit as shown. The circuit
emplojdng fewer components is always more accurate.

Ftg A-9. Example of circuit simplification.

342 Automatic Process Control

Time scaling is the alteration of the time scale of the actual problem.
Consider a differential equation,

d^x dx
M-^ + B- + Kx^y(l) (A-20)

Select a change of time variable so that

T = at (A-21)

where r = ‘^machine'* time

a = time coefficient
t = “real” time

Substituting equation A-21 in equation A-20,

„ d^x dx / t\

Thus a time scaling requires alteration of the coefficients of the differen-

tial equation and is employed for several reasons:

1. The actual problem solution may require so much time that speed-
ing up is necessary to save time.
2. The actual problem solution occurs so quickly that a slowing down
is necessary to ^^see^’ what is going on.
3. Better accuracy is obtained by operating a computer in the best
ranges of time events, neither too fast nor too slow.

The coefficients computer {a^My aB) must take into account

set into the
the fact that every integration is accompanied by the multiplication of

the solution by the inverse of the integrator time constant RC. This is
handled most easily by labelling every point on the computer diagram
with the proper values of variables, including the scaling.

Example A-5. Set up the computer diagram for

10
J+I = j/(t)

where t is in minutes. Scaling,

i
iy - x)
dr 10a
II

The circuit shown in Fig. A-10 is simplified from that of Fig. A-7. If the
computer is to speed up the problem by a factor of 60, then

T machine time !_
7 real time 60
Differential Analyzers 343
The capacitor and resistance values that may be selected are shown in the
tabl^ below.

Rt Ri c
1 megohm 1 megohm 10 mfd
5 megohms 5 megohms 2 mfd
10 megohms 10 megohms 1 mfd
All these combinations provide the same scaling and the selections of one or
the other is dictated by the particular kind of computer employed.

Fig. A- 10. Example of time scaling.

Magnitude scaling may be necessary in automatic control problems

when a specific physical problem is being solved or when nonlinear
effects are present. Let
X^ px (A-22)
where X = machine volts
P — magnitude coefficient
j - real vai'iablo

Substituting in equation as used for time scaling,

aHl~^ + aB'f-+KX=tiy{^
Thus the only effect of magnitude scaling is to alter the magnitude of the
disturbance signal y{t) when solving linear equations.

Example A-6. Scale the equation

dh
X
dT*

The scaled equation is

dr a
Suppose that a time-scale factor of ^ is employed so that a = in order
so that
to speed up the problem, fjpppose also that 50 volts equals 10 in.

P - b volts per in. The disturbing function originally proceeded at the rate
of 1 in. per sec. The disturbing function must be a direct voltage changing
at the rate of

P I in . /OOX 5 volts
= SOOvolts/sec
a sec \1 / in.
344 Automatic Process Control
If this rate of change of voltage is too high for the available function generating
equipment then new values of time and magnitude scaling must be selected
until all parts of the problem satisfactorily fit the computing equipment.

Analog computers may be used with a repetitive or nonrepetitive

method of operation. In nonrepetitive operation the problem is scaled
in time to take place in approximately 1-10 min, the disturbance or input
signal actuated once, and the solution plotted on paper by an oscillo-
graphic recorder. In repetitive operation the problem is scaled in time
to take place in approximately 0.0 10 to 1.0 sec qnd the disturbance or
input is inserted as a repetitive wave of frequency 100 cps to 1 cps.
The solution is, in this case, usually displayed on an oscilloscope, some-
times photographically recorded. In repetitive operation, the integrators
must usually be returned to zero voltage and initial conditions inserted
in the form of repetitive pulses. This action is termed “clamping.”
Initial conditions correspond in every case to the initial charge on the
capacitors. Since there are a number of integrators equal to the order
of the differential equation, there will be a like number of initial condi-
tions to be set. For such purposes a d-c power supply with a number of
outlets must be available. Each capacitor is connected by a switch to
the voltage to which it is to be initially charged. After all capacitors are
charged the switch is turned to the proper connection. The solution
may then begin by introducing the disturbance function. Fortunately
in automatic control zero initial conditions are very often used.
The input or disturbance must be introduced as a voltage at some
function of time after time scaling,

»(«-»(;)
The input signals most commonly used in automatic control ar^:

1. Step change which can be provided by a simple switch with a

battery or a square-wave generator set for long period.
2. Impulse can be simulated by a rectangular wave or pulse generator.
3. Hamp function y = t can l)e generated with a triangular wave
generator or by integrating a step funciioii.
4. Parabolic function y == tr can be generated by a twice-integrated
step function or by squaring techniques.
5. Sinusoidal functions are best generatejJ by an oscillator but are
sometimes generated in the computer using two integrators in series.

The output of the computer must be displayed in order to obtain the

solution to the problem. This is usually done on an oscillograph recorder
but the solution may also be displayed on an oscilloscope. Either method
Differential Analyzei-s 345
is satisfactory when it is recalled that specific and accurate solutions to
automatic control problems is usually not necessary.The computer
is more often employed for system studies than for particular problem
solution.
Nonlinear operations are often encountered in process control and
must be accommodated in the analog computer. In fact, some non-
linear problems can only be solved by a computer. Multiplication and
division can be accomplished with most analog computers by means of
specialized electroniccircuits, the details of which may be found in man}^
reference books.'

1. Multiplication: Multipliers nan be used to multiply two functions

of time:
= z(f)

These have two input points (accepting voltages) and one output
voltage.Whereas the product of two large numbers is even larger, it
is common to reduce the output by a factor, as

most multipliers will multiply either positive or negative values.

2 . Division may sometimes be obtained by the feedback circuit given
by

X
0 -x)^X + V

as shown in Fig. A-ll. In this circuit extreme caution should be used

in handling negative values so that the feedback loop is stable. Other

nonlinear effects such as Jiysteresis, dead zone, and limiting may be

accurately simulated by employing diode circuits. A function generator
may also be employed to introduce nonlinear functions.

^G. A. Korn and T. M. Korn, Ei^onic Analog CompiUer/i, McGraw-Hill Book

Company, New York, 1962.
346 Automatic Process Control

EXPERIMENT 1. PROCESS TIME CONSTANT

Preparation I Chapter 2
The object of this experiment is to determine the transient response to a
step change of a firBtK)rder fluid process.
The process, shown in Fig. A-12, consists of a fluid storage vessel with an
inlet flow set by a control valve and with the outflow through a valve, orilioe,

Valve Vessel

A Line -pressure regulator

(set to give sonic velocity
through control valve)
B Control valve
R Resistance (nozzle or valve)
T Pressure transmitter

Gas flow process

Fig. A-12. Time-constant study.

noscle or other restricting device. With the inflow set to a midvalue, and
after waiting a short time for steady-state conditions to be achieved, the
pressure or head in the vessel is constant, and owtflow equals inflow. A quick
charge of inflow of about 10 per cent magnitude will cause a pressure or head
change in the vessel. This change should be recorded on an instrument. This
is the transient response of the process. The time constant can be determiner!
from the experiment and compared to the calculated value.
 :

Experiment 347
The following procedure is suggested:

1. Measure the outflow by plotting pressure or head versus flow

resistance
for several values near the operating point. The resistance is calculated froin
the slope of the curve:

^ ^
aq
(liquid) sec/ft*

dv
^ ^ sec/ft*
dq

2. Measure the capacitance of the vessel

C = A (liquid) ft^

V
C = (isothermal gas) ft*
Kq I

where Rg = ideal gas constant (53.3 ft/R, air).

3. Calculate the time constant T = RC.

4. Record the process transient response on a suitable pressure or head

recorder. Use the same operating points as in previous steps.

5. Measure the time constant as the time for 63.2 per cent of the total
change. This value should check the calculated value within 10 per cent.
6. Sketch the block diagram of the apparatus and show the system function
in terms of its actual numbers, R and T:

R
TsA' 1

EXPERIMENT 2. PROPORTIONAL-INTEGRAL CONTROLLERS

Preparation Chapter 3 required. Chapter 6 suggested
:

The object of this experiment is to investigate the operation of a pneumatic

or electronic controller with proportional-integral action and to determine
the calibration of the proportional and integral adjustments.
The controller to be tested should be inspected carefully and its general
operation understood through the manufacturers' catalogs and instruction
l)ook8. In particular the controller should be connected to supply air of the
correct pressure or a line of proper voltage and frequency. The output of the
controller should he connected to a large indicating instrument and to a
control element in accordance with tlie usual method of installation.
Either the measuring element c or the set point v of the controller should
be fixed, preferably the measuring element, so that the set point can be varied
manually to test the action. The measuring clement input (or pointer) should
be positioned at a midvalue. The integral action should be turned off (integral
time to infinity) or as near 4o off as possible. The proportional sensitivity
should be set to a middle value; 50 per cent band or 0.5 gain is satisfactory.
The set pioint may then be varied in small steps and the output of the
controller noted for each value of the input, thereby calibrating the proper*
tional action.
348 Automatic Process Control
The integi al action may be tested by measuring the time required for the
output signal to change a specified amount. With the proportional sensitivity
set to midvalue and the integral time adjustment set to the smallest value
(fastest reset rate), move the set point up or down until the output remains
constant and is at a mid value. The integral adjustment should now be turned
to the largest value (off) and the set point moved to give a change in output
of about 10 per cent.
The time required to repeat the original change (10 per cent) may be
measured with a stop or wrist watch and is the integral time of the controller.
The following procedure is suggested:
1. Fix measuring element or input at midvalue. Turn integral adjustment
to as near off as possibleand set proportional sensitivity to desired value.
2. Calibrate output vaiiahle against values of set point and plot on linear
coordinate paper. Repeat for a number of values of proportional sensitivity.
The results should be a .set of straight lines. Calculate the slope

^=
dv
proportional sensitivity

3. Turn adjustment full on and set proportional sensitivity to mid-

integral
value. Move output is stationary at a midvalue. This is sero
set point until
actuating signal. Turn integral adjustment off.
4. Move set point to give a change of output of some definite amount, such
as 1 psi or 10 per cent or 0.2 milliamperes. Turn integral adjustment quickly
to the desired value.
5. Measure the time required to repeat this definite change of output set

above. This is the integral time or inverse of reset rate.

6. Sketch the block diagram of the apparatus and show the system function
in terms of its actual numbers (adjustments at midvalue).

EXPERIMENT 3. INTEGRAL CONTROLLERS

Preparalion : Chapter 3 required, Chapter 6 suggested
The object of this experiment is to determine the action of an integral or
pro|)ortional“Sjieed floating controller.
'Fhc controller to be tested should be insjxjcted carefully and its operation
understood through the manufacturers’ catalogs and instruction books. In
particular, the controller should be connected in accordance uith the usual
method of installation. The output of the controller should be connected to
a large indicating instrument. Either the measuring element c or the set
point V of the controller should be fixed, preferably the measuring element,
so that the set point can be- varied manually. A given change of set point
will result in the controller output having a rajp of change {dm/dt) directly
proportional to the magnitude and sense of the change in set point.
The following procedure is sugge.sted:
1. Fix measuring element or input at midvalue.
2. Change set-point value until output is stationaiy and at a midv'aluc
This is zero actuating signal.
Experiment 349
3. Change set-point value a small but definite amount, and measure the
rate of change of the controller output. Repeat for several values of set point
change.
4. Plot rate of change of output versus values of set point on linear co-
ordinate paper. Calpulate the integral time from

5.Explain why the upper and lower limits of velocity of output exist.
6.Sketch the block diagram of the apparatus and show the system function
in terms of its actual nu gibers.

EXPERIMENT 4. PROPORTIONAL-DERIVATIVE CONTROLLERS

Preparation ; Chapter 8, Chapter 6, and La Place transform
The object of this experiment is to determine the effect on the controller
output of the derivative action provided by a proportional-derivative con-
troller.
The controller to be tested should be inspected carefully and its general
operation understood through the manufacturers’ catalogs and instruction
books. In particular, the controller should be connected to supply air of the
correct pressure or a line of proper voltage and frequency. The output of
the controller should be connected to a recording instrument and to a control
element in accordance with the usual method of iDstallation.
The measuring-element input t of the controller should be fixed so that the
set point can be manually varied in order to test the action. The input should
be set to a midvalue. The proportional sensitivity should be set to a large
value: 100 per cent band or 1.0 gain.
The set point may then be changed quickly from one value to another
value about 5 per cent higher or lower, thereby simulating a step change.
The change of output signal should be recorded and from this record it is
possible to obtain the derivative time and the controller lag as will be explained
in the following section.
The transfer function of the controller is given by

" Tr.8 -h 1

where m = controller output

To - derivative time
Tl =* time constant
e. » input signal
Kr = proportional sensitivity

The response to a step change Sf magnitude E is given by

TD»+lk
350 Automatic Process Control
This trsDsform esn he written as
TdEKc EKe
M(8)
T l8 1 8{T l8 -f- 1
)

and inverted

"•(0 - [i +
This curve is plotted in Fig. A-13 for Td = STl and Kc = 0.5. The ratio
of the maximum to the final value of m is the time constant ratio Td/Tl-
The 63.2 per cent time of the decay is Tl-

The following procedure is suggested:

1. Fix measuring element at midvalue. Set proportional sensitivity to

100 per cent or 1.0 gain. Set derivative time to moderate value (approxi-
mately 30 sec). Move set point slowly until controller output is at mid value.
2. Make a step change of set point of some definite amount, say 5 per cent,
and record the resulting change in controller output.
3. From a record of the controller output, measure the highest output
obtained (see Fig. A-13). The ratio of time constants is the ratio of the
highest value to the final value of output:

Td _ ynma» ~ m,
Tl
where wi, = the initial output before set point was changed.

mmax ~ maximum value at time of change (it may be necessary to

extrapolate the curve backward to find this point).
m« = final value of output.

4. From the record measure the lag time constant Tl and calculate Tji.
5. Repeat for several magnitudes of changes and in both directions.
Experiment 35I
6.Sketch the block diagram of the apparatuB and show the syitem function
in termB of its actual numbers (adjustments at mid value).

KcjX -h To 8 )
X -X- T 1,8

EXPERIMENT 5. PROPORTIONAL CONTROL OF A PROCESS

Preparation: Chapters 2 and 3, or 6 and 4
The object of this experiment is to measure and calculate the open-loop
gain for a proportional automatic process control system.
The control system is shown in block diagram form in Fig. A-14. A
recording instrument should be connected at feedback variable b in order to

z Manual It

Process
control

h'
Input
element
Controller

kH
Valve Process
hO
Measuring
element

Fio. A~14. Process control loop.

record the transient response of the process under control. The automatic
controller should have manual control means by which manual control of the
process (y~^x connection) can be obtained. Automatic control is achieved
by switching in the controller {p—* x connection).
The open-loop response, b output to x input, should be second-order, other-
wise loop gain is not easily calculated. 'J'he response is generally
R
- (7\s+l)(Tts+l)''

where R = process or system sensitivity (units of 6/unit8 of x) and the two

time constants may be (a) valve and process, {b) valve and measuring element,
(c) process (two), (d) process and measuring element. Other combinations
are not likely to occur. It is assumed that all other lags are negligible.
The two time constants (Ti and I'i) may be either calculated or measured.
The loop gain (RKc) can then be calculated for proportional control:

RRc + Tj
if* \T2 Tj

where f = damping ratio of closed-loop response.

The process can be placed on automatic control and the proportional

densitivity of the controller adjusted until a transient response to a set-point
 :

352 Automatic Process Contiol

change (or change) exhibits the desired damping ratio (compare to curves
loarl
in the appendix).The loop gain is measured by switching to manual control
and making a steady-state test of the ratio p/.T. The test value should be
compared to the calculated value.
Two difficulties are often encountered in this experiment. First, make sure
that all variables are operating in the same range of values during all tests.
Fluid processes are usually nonlinear and values of resistance change with
operating point. Second, use a sufficiently small amplitude of signal so that
none of the apparatus is caused to go to limits.
The following procedure is suggested
1 Select operating set point v for system and check to see that all variables
.

p, m, c, and b are in middle of respective ranges.

2. Determine the system time constants (use previous ctata if i)oasible).

3. Calculate the loop gain RKc for one-third damping ratio.
Experimentally determine the value of proportional sensitivit}" Kc to
4.
give one-thirddamping ratio on control record.
5. Use manual control to check loop gain. Set Az to give a Ax and measure
Ap divided by Ax. Check with calculated value.
Calculate the static error and offset.
6.
Test the static error due to a change in set point, and comp.are U> the
7.
calculated value.
8. Test the offset due to a change in load (if a load change can be made),
and compare to the calculated value.
9. Sketch the block diagram of the apparatus and show the system function
in terms of .actual numbers.

EXPERIMENT 6. PROPORTIONAl^lN1'EGR.\L-DERIVATlVE
CONTROL OF A PROCESS
Preparation: Chapters 2 and 3, or 6 and 4
The object to determine the “best" setting of a
of this experiment is

when controlling an actual process.

proportional-integial-derivative c.mtrollej-
The control system is shown in the block diagram of Fig. A-14. A recording
instrument should be connecte<l at feedback variable b in order to record the
response of the process to various changes of parameter. The open-loop
response, b output to r input, may be of any order, but the number and size
of all lags should be known. It is preferable to have two time constants so
that the response is
a :
K =
(r.«+ l)(7’,s+ !)•'

The time constants may be determined by calculation or measurement and

the open-loop gain (RKr) for a given damping ratio may thereby be calculated.
There are three methods of determining the proper value for proportional-
sensitivity;

1. Analytical method wherein the proi)er damping is determined from

solution of the differential equation. For processes with only two time
constants, Tu T^,

HR. - open-loop gain “ ^

Experiment 353
where f - damping ratio (0.33 is a good value). This value of loop gain may
be tried, and the period of oscillation noted (P^). The apparent lag of the
prooessis
L - 0.2Pp
The settings of integral and derivative time may now be determined by the
Ziegler-Nichols equations.
2. Reaction curve method wherein the response of the open-loop system
(minus the controller) is obtained from a step disturbance, as detailed in
Chapter 4. The parameters R, L, N, K
are then scaled from the curve and
the desired settings comjjuted.
3. Stability limit method wherein the resj)Oii6e of the closed-loop system is
observed for gradually increased settings of proportional sensitivity. Note
the value of proportional sensitivity /Cu which just causes continuous oscilla-
tion. Note also the period of oscillation P„. Then

A'c * 0.6K„
and
L » 0.25/"«

The settings of integral and derivative time may now he determined by tlie
Ziegler-Nichols equations
The procedure is as follows •

1. delect operating set point v for system and check to see that all x anablcs
p, fit, c,and b are in middle of respective ranges.
2. Turn oflF integral effect and derivative effect.
Gradually increase proportional sensitivity so tlu»i a stead^> (>' '^’.dioD
3.
begins. Make sure that none of the variables p, m, c, or h is allowed to limit
during this operation. Move set point to see that marginal stability is definite
and sustained. Note proportional sensitivity and period of oscillation Pu.
Set proportional sensitivity to 60 per cent of the value causing marginal
4.
stability. Check now by moving the set point slightly to see that the transient
response is satisfactory one-ninth ratio of succeeding amplitudes is suggested.
;

Trim setting if necessary. This setting will be denoted as A',.,.

o. Calculate the “apparent lag’’ L of the process as one (juaiter of the
observed jjeriod of oscillation.
6. The settings of the proportional, integial and (leri\ati\e effects should
now Ix: calculated for a nonserics-type controller'

T, = 2.fiL

For a series-type controller;

K,' = ; A*
3

A/ = 0.4L

T,' = 2L
 354 Automatic Process Control
With the selected settings check the stability by moving the set point slightly
to see that the transient repsonse is satisfactory. Trim the settings if neces-
sary» maintaining a one-ninth ratio of amplitudes, good return, and small
period of oscillation.
7. Check the settings obtained above by emplo3rmg the reaction-curve
method.
8. Compare the various values of parameters /Ce, T,, and Td obtained in
the tests and discuss any significant differences.

EXPERIMENT 7, ANALOG COMPUTER ANALYSIS OF CONTROL

Preparation: Chapters 2, 3, and 4, and Appendix
I'he object of this experiment is to set up a process with proportional-
integfel-derivative control on an analog computer and determine the transient,
performance with proper controller adjustments.
For illustration a system consisting of two time constant elements in series
under proportional-integral-derivative control, with the control element
having a significant time constant of its own, will be considered. The analog
computer circuit is shown in Fig. A-15. The control-element (valve) time
constant is represented by amplifier 5.
The transfer function of this part of the analog is
m _ _J
I R^CiS -f- 1

where RiCi is the valve time constant (usually 1 to 10 sec). The first process
time constant is represented at amplifier 6,

m R%C%s -f- 1

so that the (iroccss time (constant is set by RtCt. The second process time
constant is represented at amplifier 7,

c 1

h R^C^8 + 1

and this process time constant is set by RjCi-

The controller is represented by amplifiers 1, 2, 3, and 4. The set-point
voltage V and the controlled variable c feed into the amplifier 1 The ampli^er .

is set for a maximum gain of 10 with the potentiometer R^ at maximum. The

proportional sensitivity is therefore set by and so is the loop gain.
The open-loop gain may therefore be read directly at potentiometer Rz. The
second amplifier provides proportional-derivative action, and the third ampli-
fier is necessary for sign qhanging. The transfer function of amplifiers 2 and
Sis

“ = (52(1 — 52)/?2C'2K + 1

SO that an adjustment ^2 of potentiometer Rz will vary the derivative time

from 0 to 0.25 R 2 C 2 The input resistor R 2 to amplifier 2 is fixed but has the
.

same resistance value as potentiometer Rz. Only potentiometer Rt in tlie

 :

35G Automatic Process Control

feedback is adjusted. Amplifier 4 proxndes proportional-integral action and
its transfer function is

^
- - +
1
^ SiRiCiS
80 that an adjustment S 4 of potentiometer
1 R4 will vary the integral time from
0 to R 4C 4 .

Load changes may he introduced at ue and u? if their effects are to be

observed.
The and amplitude scaled thereby select-
entire circuit sliould next be time
and capacitors, Rj through Rj, and C2 through C7.
ing values of all resistors
When the computer is set up and ready for operation the open-loop gain
may be determined. The derivative and integral time can then be set and the
effect of load changes observed.
The following procedure is suggested

1 . T 4 Tt, and T? for the process to be investi-

Measure the time constants ,

gated. one of the time constants is zero, simply remove the corresponding
If
capacitor in the computer circuit.
2 Time scale the computer circuit. The time scaling will depend upon the
.

make of computer in use. In general, since this is an exploratory problem

and accuracy is not important, the fastest scaling should be used. For the
Philbrick computer a scale factor of a - 4 X 10”^ is suggested. Set the com-
puted values of resistors and capacitors in the circuit.
3. Set the derivative time to zero and the integral time to infinity (as large
as possible). Try a loop gain of 5,0 (set potentiometer Rs to 50 per cent) and
test the stability by moving the set point. Select a value of gain that gives
a desirable transient response.
4. Turn in derivative time (^lotentiometer R 2 ) and increase loop gain a)K>ut

25 per cent. Try several related \ alues of derivative time and loop gain until
a transient response with good stability and minimum ])eriod of oscillation is
obtained.
5. Turn off the derivative action and decrease loop gain by 35 per cent from
the last step (90 per cent of original value). Set the integral time to give a
good return on a load change.
6 Turn in derivative time to 30 per cent more than the value set in step 4.
.

Set integral time to double that value found in step 5. Set proportional
sensitivity to about 20 per cent higher than in step 5. Try several values near
these settings to see if a better response can be obtained. Use the highest
derivative time possible because this will improve loop gain and provide
smaller period of oscillation.
7. Leave all settings at their l>est values and turn out derivative action to
show its effect. Next turn out integral action and with a load change at Mi or
u? determine whether the offset is

1
"
1 -I- loop gain
as a fraction.
S. Eliminate the control-element lag by removing capacitor Ct. Find new
control settings to det(‘rmine whether there is a significant improvement when
the control-element lag is removed.
Experiment 357

EXPERIMENT 8. TWO-POSITION CONTROL

Preparation: Chaptem 2, 3, and 4
The object of this experiment ie to observe the action of a two-position
controller with differential gap.
The process should consist of one large capacitance (or time constant)
together with a minimum number of small time constants. A small dead time
may also be present, llie two-position controller should have a differential
gap, preferably adjustable. The controlled variable should be recorded so
that various features of the behavior can be studied.
The following procedure is suggested:

1. Move the set point with the measured (feedback) variable maintained

at a fixed value and determine the value of the differential gap G.

2. Start the system in operation and maintain the control element open.
The maximum value attained by the controlled variable may be called K
5.
above that obtained when the control element is off.
3. Set the set point at the value K/2 and let the system control about this
point for a short time. Note the amplitude of oscillation A and the period of
oscillation P. (Note: always measure amplitude as double amplitudes and
divide
6. by 2 to obtain G and A.)
4. From the Ziegler-Nichols equations, calculate the lag ratio (see Chapter
4)
L . K-^G
R

where L - apparent dead time and T is an apparent process turn' .iistant.

Calculate the apparent dead time from

2A
tanh^*
R K
The apparent process time constant T may now be calculated.
Estimate the amplitude and period of oscillation if differential gap 0
is zero. The following equations are used.

A = 1(1

P = 4Litonh->^

where V * value of set point. Verify by actual test.

7. Move new value not equal to K/2, With a given

the set point to a
differential gap, theamplitude and period of oscillation should not change
significantly when an unsymmetrical cycle is observed.
8. Move the set point to the opposite direction as in step 7 and observe
^
that the saw-tooth-like wave reverses its symmetry.
9. Explain how the magnitude of *'load'' on a system can be determined
by inspection of the shape of the saw-tooth wave.
358 Automatic Procese Control

EXPERIMENT 9. OONTMN. VALVE CBARACTERISTICS

Preparation: Chaptm 2, S, 4, 7, and S
The object of this experiment is to S 3mthe8ise a flow-position characterktie
for a fluid control valve used for process control.
There are sometimes encountered problems in automatic process control in
which the maintenance of controller stability under all conditions is difficult
even if the pressure differential across the control valve is constant. In that
case a specified control characteristic may be selected to keep loop gain at its
best level. However, nothing is usually done about maintaining integral
time or derivative time at their best values. ,

The process should be a fluid process (as in Fig. A-12) with a nonlinear
resistance at the outlet, the usual case. The controller should be set for two
responses such as proportional-derivative. Although the industrial problem
usually involves control for various load changes,it is easier in the laboratory

to move the Consequently, a valve characteristic will

controller set point.
be synthesized for various set points. A different valve characteristic would
be required for variable load but it would be obtained through the same
procedure.
The suggested procedure is

1 . Set the controller in operation and check all operating values of variables
to insure correct operation.
2. Beginning with the set point at a midvalue (50 per cent) determine the

best loop gain and integral or derivative time to give a certain transient
response. Record the value of flow through the control valve required to
4.
maintain the set point, and record the value of controller output x at each of
these points, l^te these values.
3. Repeat o()eration 2 for as many set-point values as are desired, at least
5. The loop gain
20, 35, 60, 65, 80 per cent. is

^
de dx
dm ^
dm
where ^ controller sensitivity = Kc
at
dm —
valve sensitivity
dx
dc
process sensitivity == Kp
dm
The set point (or controlled variable) should be plotted against valve flow.
The slope of this curve is Kp.
The required valve characteristic may now be synthesised by calculating
dm K
dx dc
^
dm (f?

The controller proportional sensitivity is to remain constant, so

dm K
Experiment 359

Uae the values of K and K„ and calculate the slope of the curve K, at each
value of flow m.
6. By cut and try, parabolic curve fitting and integration, or other methoda,
find the curve m » fix). Compare this curve to the actual plot of m veraua x
for the valve in use, and indicate whether it has the desired charaeteriatic.
7. If the prorcFS is of second order and the constants are known, uae the
analytical method given in Chapter 8 to verify these results.
 Answers to Problems

1-2. Furnace thermostat, hot water heater control, water-preseure regulator, toilet

tank water-level control, electric iron thermostat, refrigeration roiitrol.

1-3. Wattmeter, water meter, gas meter, thennometer, pressure gage.

2- 1 . Inflow, outflow, inlet temperature, outlet temperature, electric current.

2-2 . Inflow behind dam, outflow, head, weir setting.
Air temperature, stack loss, air infiltration, air inlet temperature, heat loM,
gas flow.
2-4 . 2H/5Q
2-5 . HInQ
2-4». C « )r(DA- A*)forO^ D
2-7 . Area proportional to square root of head.
2-6 . Area proportional to inverse of square root of head.
2-9 . e * Ri electrical Ohm's law
Ri fluid Poisseuille-Hagen law
Ri fluid Bernoulli's law
e* * Ri heat Stefan'Roltzmann law
e ^ Ri heat Fourier law
2-10 . 0.05 in.* isothermal 2-11. 2900.0 sec F/Btu
2-12 Btu per F
. 8.3 2-13. 0.028 Btu/F
2- 14 Hydrogen smallest
. (least molecular weight), carbon dioxide largest
2- 15 . 77.0 sec
2-16 . 300 min
2-18 Horizontal
. if 4L > rP
2-19 AB, AL, BL, y/A*
. + B*L, y/A* + L*B, y/B* + L^A
2-20 . K - 0.067 Ib/in.; D - 6.6 lb iw/ft
1 2a
2-21 . R m ,

2wKL A
2-22. [T’jjrjjs* -H (I'll ri- — (rijs -f l)Asut
2-23. [RCRiCiS^ + (i2C -b RiC + ffiCOa -}- Ijc = Rim
2-24. {RCn -f l)c - -RK^m -H i?u
200
Aaswcrs to Problems 3«1
a-ts. (Ms* + Bs + Jf)i -/ + K'v + Bs»
a-27. f-Btx
a-28. mt(R + fit + JUiiCt) m mi(fii 4- RHtCt)
M9. [(Bi + A)Cs 4- l]mt » (RCt + l)mi
a-W. ZTi 4- Dm, - m,
(T*s* 4-
(B/Ay)» 4- 11c “ m/y, when y
ICB/ff)** 4- fluid deniity, K “ tube »m
»-aa. -(Ms* 4- Bs 4- K)e ~ Am, K - WRA^T/V^
V - volume under piston, R > gns constant, T ~ trmperatun
a-aa. (Ts 4-
3- D® - m, t - pDC/nr
p “ density, B =diameter, C = up. hi., B => oonv. coeff.
4-
J-i. riTj/Ti 4- 7 ’j: ?V(7’i 4- 7’,); (Ti 4- Tt)
3-5. Prop. K\ integral KT
3-5. Prop. T\IT\, integral T2
3-7. Prop. (Ti + rO/Ta; integral (Ti + T’l); derivative T^T^fiTi + P*)
3-8. Prop. JiC; derivative (Pi + Ti)/K
3-9. Prop. T\IT 2 \
derivative P2
10. Prop. 1; derivative 2P; eecond derivative P*.

6. 25':o 4-7. 10%

4-8. Yob 4-9. 4C - KJ,
4-10. Ta + CjKc 4-11. t//(l -I- RK^)
4-12. 2K,RT = (1 + RK^Y 4-13. «* T IAH T 2
1 1 )
fsero

4-14. (fi 4- fiiK .rs)* 9 - ARiK^T iTf, P - 2v V(T^W(R^)

4-15. P - 2t VTiTt/(RiK, 4- D; B, - (//(BiK, 4- D
S-1. 5 deg 5-2. At>oul 7 8€w
5^3. 0.54 sec 5-4. R - 12,000 eec/ft*; P - 0.06 lec
5-5. 1.6 sec 5^. Lag * Pi + P»

6- 1 . he ^ Kv -\r (/>
—

6-2 . m r where #Cd = diaphragm etifTnees

»
1 -f-

A, * supply orifice coeff. 9u>ldm, » control nozaie coeff. Aw/dx^ A * dia-

phragm area

6-.1 .

r, ,

AKx
6-5. (Pfi + Dm - A>, T
K{Ki -HR,)’ “
RCR, + Ki)
6-6. * R/AC where R spring gradient, A - diaphragm area, C
nozzle coefficient 9w/dx

6-7. VI » RjC, prop, integral

6 -8. m * RC, prop, integral

362 Automatic Procees Control
6-9. m - - (1 + Ti)€, T“ RC, prop, darivittive
6-
6-H. Ration W “f ^
(l - |)j.
7- where E- radiue of circle.
U. a- Ta/{Tt + 1)
-
W. ««
BP.
“ sin 2ui: 48 bp 7-S. hp - PQ/1720S

4 . 2,628,000 7-5. lOrlb

7-^. A(C + AVJC)

7-9. 1.0 ID. 7-10. 420 gpm

.. ^ = OiGiH \
a-l. from C

a-2 . T ^ vh'i, J'l ^ Ti/V27

9r3, (tt) Zero (b) Zero (c) Zero (d) Zero
•-4. Probably yes. Period is reduced by 67%.

»<6. Cl - cKK, * + ir*

aTs “b 1

K,
M. Withi;,
1 + K^Ki + KiKiKi
- Kz
Without
1 4 KxKzKi
ft-6. (a)Square wave as shown. (&) Triangular wave.
9r9. Zero
MO. Ti * 4KiA/Ki^, A vessel area, Ki *= dq/dx valve, Kz * dqfdc valve
^
-11 . K-K-
0.11
8

where A = vessel area,

A’l= dq/dx valve,
^2 “ dq/dc valve
M. (a) l/uT, -90°, {b) (1 + 2 tan-'wP
(c) Vl 4- tan“^ — wT (2nd quadrant)
id) uT/Vl 4- co*r* 90” - tan-‘«r

1 1 4 -

~ tan-‘ 26,r

(/) 1/ \/(l +w*7’®)(l 4 4«-T*); -tan-^wT -tan”‘2«r

1
tan~H) (2zfd quadrant)
1 - ,

— 4wZ«
0^. —uL and tan *

4 —« L*
M. <1 time constant element
RK^ a: 1 capacitance element
RKt > 1 exothermic
Answers to Problems 363
w. •V. 0.29 sec; Probably not.
60%.

IM. (o) « » -1, -0.5, -0.33 stable

(fc) * -3, 1 ± unstable

(c) « - -1, -1, 0.05 (-l±»V^) .table

(d) . -0, (-1 +i\/3)/27’ unstable

10-3. /Cf *

atuT « 0.40, ^ » 135® (open loop), Ke * 0.42

 Index

Accumulator, 171 Batch process control, 258

^
AcouBtic flow, 22, 230 Bernoulli’s law, 19
Actuating signal, 59 Block diagram, 9
See aUo Deviation Bode, analysis, 317
Actuator, dead zone, 296 diagram, 277
definition, 195 Boiler, 15
electric, 206 Bourdon Tube, 33
electro-hydraulic, 205
electro-pneumatic, 202 Capacitance, distributed, 26
failure,209 electric, 18
hydraulic, 203 fluid, 19
pneumatic, 196 gas, 23, 147
two-position, 207 liquid, 19, 21, 146
Adjustment, see Control adjustment thermal, 26
Air heater, 50, 53 Capacitance element, 29, 283, 301
Air supply, 167 Cascade control, 254
Amplitude ratio, 271 Cascade elements, 38, 42
Analog computer, 337, 355 Characteristic time, 34
Answers to problems, 360 Closed loop, 78
Anticipation, 70 Cohen, G. H., 117
Apparent dead time, 116 Communication, 7
Askania regulator company, 172 Complex variable, 271, 274
Asymptotic plots, 281 Computing machine, 264
Automatic control, advantages, 7 Conoflow corporation, 207
benefits, 1, 2 Continuity law, 21, 22, 36
definition, 2, 223 Control, action, 59
purpose, 238, 261 direct, 12
quality, 299 equation, 79
Averaging control, 202 indirect, 13
law of, 14
Backlash, 296 Control adjustment, 104
Index 365
Control valve, bypaes, 218 Eleotric controllers, integral, 174
characteristics, 216, 223, 249, 358 proportional, 173
dead zone, 296 sensitivity, 174
equal percentage, 223 Electric heating, controllers, 185
dow, 48, 212 modulation, 233
flow coefllcient, 48 processes, 50
head losses, 213 Eleotric transmission, 134
pressure balance, 257 Electro-hydraulic actuator, 205
pressure drop, 212 Electro-pneumatic actuator, 202
rangeability, 225 Error, see Actuating signal
rotating shaft, 226 Exothermic element, 288
^
sising, 228 Experiments, 346
sliding stem, 221
use, 61 Feedback, control, 78
Controlling means, 50 lag, 240
Convection, see Heat transfer means, see Mensiiring means
Coon, G. A., 117 variable, 239
Coulomb’s law, 17 Final control element, definition, 61, 195
Cybernetics, 7 failure, 209
First-order system, see Time-constant
Damping, 27 element
Damping ratio, 34 Float-level controller, 62, 63, 65, 167
Dead time, 53, 86, 112, 116, 133, 288, 301 Floating rate, 75, 112
Dead zone, 293 Flow coefficient (control valve), 212, 228
Decibels, ^1 Flow control, 91, 94, 112, 211, 242, 254,
Degree of freedon, 13 257, 262
Derivative time, adjustment, 100, 118, Flow measurement, 129
119, 161, 165, 178, 182 Flow process, 49
definition, 69 Fourier law, 24
phase, 70 Fractionation, see Distillation
Describing function, 293 Frequency of oscillation, see Period of

Desired value, see Set-point variation oscillation

Deviation, 61, 79, 239 Frequency response, 282
Diagram, block, 9 Friction, 296
physical, 9 Furnace control, 6, 12, 51, 240, 245, 260,
Diaphragm valve, 196 265
Differential analyzers, 337
Differential gap, 74, 107, 189 Gas process, 43
Differential operator, 30 Gibbs’s phase rule, 15, 17
Direct control, 12 Goldman, S., 8
Distillation, 16
Distributed parameter, 289 Hays Corporation, 176
Draper, C. S., 266 Heat exchanger, 14, 15, 53, 240
D3rnamic elements, nee Process elements Heat transfer, conduction, 24
convection, 24, 139, 141
Eckert, E. R. G., 142 • radiation, 24, 149
Eckman, D. P., 116, 123 Heat treating, 6, 12
Electric actuators, 206 Hooke’s law, 27
Electric controllers, derivative, 174 Hydraulic actuators, 203
for heating, 186 Hydraulic controllers, integral, 66, 170
366 Index
Hydraulic controllers, oil supply, 170 Neutral sone, 70, 112
proportional, 171 Newton's law, 27, 33
Hysteresis, 293 Nichols plot, 278
Nyquist criterion, 310
Inertia, 27, 195 Nyquist, H., 6
Information theory, 7
Input, see Set-point variation Offset, 82, 90, 92, 94, 103, 104, 117, 246,
Input disturbance, 247 259
Input means, 60 On-Off control, 73
Instrumentation, 7 Open loop, 79, 115
Integral control, action, 64 Operator, differential, 30
adjustment, 64, 90, 117 LaPlace transform, 30, 332
effect of measuring lag, 242, 247 Optimizing control, 264
experiment on, 348 Oriflce, 47, 242
hydraulic, 170 Oscillatory element, 33, 149, 284, 301
performance, 88, 104 Overshoot, 259
phase, 66
Integral time, adjustment, 90, 98, 117- Period of oscillation, 36, 86, 103, 108
119, 162, 164, 170, 173, 178, 182 117, 118, 120
definition, 64, 67 Phase lag, 271
Inverse derivative, 166 Phase-margin plot, 278
Physical diagram, 9
IiaPlace transform, 300, 332 Pipe friction, 19, 213
Lead component, 68 Piston actuator, 196
Leeds and Northrup Company, 173, 187 Pneumatic actuators, 196
liofkowitx, I., 265 Pneumatic controllers, air supply, 167
Li, Y. T„.266 derivative, 160, 165
Limitiili) 29B integral, 161, 164
line losses, 213 proportional, 158, 163
Liquid-level, control, 241, 249, 257, 263 sensitivity, 159, 164
gap. 128 Pneumatic transmission, 131, 292
Liquid process, 36 Poiseuille-Hagen laws, 20, 22, 147
Load disturbance, 4, 13, 80, 105, 239, Polar plot, 278
245 Pole, 310
Locus, 278, 307 Positioner, 198
Pressure balanced valve, 257
Magnitude ratio, 271 Pressure control, 241
Magnitude scaling, 343 Pressure-element lag, 145, 287
Manometer, 128, 149 Pit*5^uic gage, 127, 287
Manual reset, 63 Pressure regulator, 60, 156, 257
Maxwell, J. C., 6 Process, batch, 258
Measuring lag, 240 continuous, 261
Measuring means, 59, 60 defmition, 12
Minneapolis-lloneywell llegu'lator Com- disturbance, 245
pany, 175, 188 elements, 28
Mode of control, see Control action equatl;;^, 78
Moore Products Company, 166 experiment on, 346
Motor actuator, 196 flow, 47
Multiple time constants, 41 gas, 43
Multispeed floating control, 76 liquid, 36
Index 367
Procen, thfimud, 50 Proportional-ieset action, 66
Proportional control, action, 61, 181 Proportional sensitivity, adjustment, 82,
adjuBtment, 63, 82, 117 86, 94, 117-119, 158. 164, 172-174,
batch prooesaeB, 259 178, 181
damping 86
ratio, definition, 63
effect of measuring lag, 240, 243 Proportional-speed-floating action, 64
electric, 173, 177
electric heating, 185,233 Quality of control, 299
experiment on, 351
frequency of oscillation, 86 Radiation element, 127
hydraulic, 171 Rangeability, 225
*
offset, 82 Rate, action, 69
performance, 80, 103 time, 70
pneumatic, 158, 163 Ratio-flow control, 257
sensitivity, 63 Reaction rate, 116
stabilization, 83 Repeats per minute, 67
Proportional band, 63 Reset action, 64
Proportional-derivative control, action, Reset rate, 67
69, 166, 182 Resistance, acoustic flow, 22
adjustment, 99, 118 conduction, 24
batch processes, 260 convection, 24
electric, 174, 178 distributed, 26
experiment on, 349 electric, 18
frequency response, 285 fluid, 19
performance, 99, 102 gas, 22
phase, 70 laminar, 20, 29, 147
pneumatic, 160, 165 liquid, 19
Proportional-integral control, action, 66, orifice, 20
183 pipe, 19
adjustment, 67, 118 radiation, 24
batch processes, 259 thermal, 24
electric, 174, 178 turbulent, 19, 22
electric heating, 187 Resistance thermometers, 127
experiment on, 347 Re 3moldB number,18, 20
frequency response, 285 Routh-Hurwitz condition, 305
hydraulic, 172
offset, 246 Saturation, 293
performance, 94, 103 Second-order system, see Oscillatory
phase, 68 element
pneumatic, 161, 164 Self-regulation, 91, 116
Proportional clement, 29 Servomechanisms, 5
ProportionaJ-integi'ul-derivative control, Set-point variation, 80, 104, 239, 247
acUon, 71, 185 Shannon, C. E., 8
adjustment, 119 Ship steering, 5
electric, 175, 178 Single-speed floating control, 75, 112,
experiment on, 352 186, 190
frequency response, 286 • Sinusoidal, analysis, 269
hydraulic, 178 testing, 297
performance, 102 Sonic flow, see Acoustic flow
pneumatic, 162, 165 Spring actuator, 106
368 Index
SpringlefiH B<‘tuator, 1P6 Transfer locus, 307
Staliiiity analyRU, 299 Trunsmissiou, bleed time, 54, 188
Static error, 104, 248 electric, 184
Stefan-Boltzmann law, 25, 144 lag compensation, 183
Storago-veseel control, 282 pneumatic, 130
Surgeveaselfl, 282 signal, 8
Swartwout Compan\% 177 Transfer lag, tee Multiple time constants
System function, 29, 301 Tnixal, J. G., 289, 294
«ee af«o Transfer function Tsicn, H. S., 7
Turndown, 225
Telemetering, tee TraiiHmission Two-position actuators, 207
Temperature control, 241, 251, 256, Two-position control, 73, lOG, 120, 189,
289 288, 357
Thermocouples; 126, 152, 287
Thermometer, 125, 137, 287 Valve characteristicr,251
Tltermostat, 3, 75, 157, 190 Valve ix>6itioner,198
Three-position control, 74 Velocit}’ error, 104, 261
Time-constant element, definition, 32, Voice-coil motor, 179, 202
139 Von Neumann, J., 2
experiment on, 346
frequency response, 283 Water heater, 5, 13, 26, 240
liquid, 38, 42 Watt, James, 6
multiple, 41 Weiner, X., 7
response, 136 V
transfer function, 301 Zero, 310
Time scaling, 342 Ziegler-Nichols method, 114, 201, 353,
Transfer function, 300 357