0% found this document useful (0 votes)
219 views319 pages

Control of Nuclear Reactors and Power Plants (M.a. Schultz)

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
219 views319 pages

Control of Nuclear Reactors and Power Plants (M.a. Schultz)

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 319

Control of Nuclear Reactors

and Power Plants

M. A. SCHULTZ

Westinghouse Electric Corporation

Pittsburgh, Pennsylvania

McGRAW-HILL BOOK COMPANY, INC.

New York Toronto London

1955

FUND FOR PEACEFUL ATOMIC D£V£LOPM£HT, INft,


Generated on 2011-10-24 11:37 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Phoenix

TK

. 2L

CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Copyright © 1955 by the McGraw-Hill Book Company, Inc. Printed in the

United States of America. All rights reserved. This book, or parts thereof,

may not be reproduced in any form without permission of the publishers.

Library of Congress Catalog Card Number 55-7283

THE MAPLE PRESS COMPANY, YORK, PA.


Generated on 2011-10-24 11:37 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PREFACE

This preface is being written as President Eisenhower waves the wand

to start construction of the first commercial nuclear power plant at

Shippingport, Pennsylvania. This event, symbolizing the entrance of

the age of nuclear power, focuses attention on the background of nuclear-

power-plant control. Historically, the early nuclear-reactor-control

designers were concerned only with the similarity of the reactor to a

bomb. The problems they faced were ones of safety and of complete

distrust for refined control elements. The legend is told in the industry

that in the first reactor, constructed under the west stands of the Uni-

versity of Chicago stadium, there existed, in addition to the normal

pneumatic safety-rod mechanism, another safety rod suspended by a

rope, with a hatchet placed conveniently nearby. The progress in

nuclear control from this point to where useful power could be safely

handled has been monumental. Now, in this new age, the problem is

simply stated: Given a nuclear power plant, what is the best way of con-

trolling it? The answer is presented in engineering terms similar to those

used in any complex control problem. The newly developed techniques

of servomechanisms are quickly brought forth as a basic design method,

and now a nuclear power plant, a jet engine, or a guided missile is treated

with confident engineering certainty.

This book, one record of nuclear-control progress, is therefore largely

in elementary servo form and language. Some concessions are made to

the nuclear physicist in recognition of the essential partnership involved

between the physicist and control engineer in the design of a control


Generated on 2011-10-24 11:37 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

system for a nuclear power plant. The entire field of reactor and power-

plant control is far from covered in this book. Only one specific type of

reactor, the solid-fuel heterogeneous reactor, is used for descriptive and

illustrative purposes. More complication is generally involved in the

design of plants containing different reactors such as circulating-fuel

homogeneous reactors. However, the basic techniques for the solution

of the nuclear-control problem are presented in such a manner that the

design of control systems for other types of reactor plants may be obtained

by extension of the methods presented.

At this point it is customary to acknowledge a few of one's coworkers

in the field and to ignore the remainder as being too numerous to men-
vi PREFACE

tion. Because of the pioneering efforts of the small and somewhat closed

fraternity of engineers in this new industry, I should very much like to

acknowledge the tremendous historic labors of my associates in this field.

First, since all the present activity in this country in the field of nuclear

power plants is under the direction of the United States Atomic Energy

Commission, most of the references originally came from the basic work

accomplished under the commission. Grateful acknowledgment is made

to the AEC for permission to publish this material. Second, I should

like to thank the Westinghouse Electric Corporation for supplying me

with the necessary educational background for this project. I should

also like to thank Westinghouse and Radiation Counter Laboratories

for the use of some of the illustrations used in this book. Finally, I must

mention specifically the people at various AEC-sponsored projects who

have directly or indirectly contributed.

At Westinghouse I am particularly grateful to J. N. Grace for his basic

work, assistance, and criticism. I have borrowed liberally from my

friends and colleagues G. Anderson, W. Baer, R. T. Bayard, G. Conley,

J. C. Connor, R. C. Cunningham, R. Durnal, F. Engel, W. Esselman,

T. Fairey, J. Franz, E. F. Frisch, W. Hamilton, A. Henry, J. Kostalos,

R. Leonard, H. McCreary, W. Pagels, W. Ramage, V. Shaw, J. C. Simonds,

C. Single, G. Stubbs, O. Swift, S. Wallach, and J. Wolff.

At the General Electric Company's Knolls Atomic Power Laboratories

the initial servomechanism concept of the reactor transfer function was

achieved by J. Owens and J. Piggott. E. Wade of this laboratory also


Generated on 2011-10-24 11:37 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

made many contributions to reactor-control instrumentation.

I am indebted to my friend W. Pease, formerly of the Massachusetts

Institute of Technology, for my initial education in the control of nuclear

reactors. He, of course, was responsible for the automatic control design

of the Brookhaven reactor.

At the Argonne National Laboratory J. M. Harrer, J. Dietrich,

J. Deshong, and D. Krukoff, among others, were responsible for the

tremendous effort to make an engineering science of reactor control. It

was their initial work on the oscillation of a reactor that gave the servo

engineer respectability in the nuclear field.

At the Oak Ridge National Laboratory T. Cole and W. Jordan were

always of assistance to me on power-plant-control problems while they

pioneered with their colleagues on the Materials Testing Reactor control

system.

My thanks go to J. Newgard, R. Longini, and W. Brazeale for reading

this book in manuscript form.

M. A. SCHULTZ
CONTENTS

Preface v

CHAPTER 1. INTRODUCTION 1

1-1. Introduction and Purpose. 1-2. Analogy of a Nuclear Power Plant to

a Direct-current-generator System. 1-3. Example of Analogy of a Nuclear

Power Plant to a Direct-current-generator System. 1-4. Philosophy of

Reactor and Plant Control. 1-5. Control-system Specification. 1-6.

Scope of Text.

CHAPTER 2. ELEMENTARY PHYSICS OF REACTOR CONTROL 10

2-1. Description of a Reactor. 2-2. Fission Process. 2-3. Neutron Level.

2-4. Reactor Period. 2-5. Reactor State. 2-6. Prompt Critical. 2-7.

Subcritical Level Operation. 2-8. Subcritical Period. 2-9. Critical Opera-

tion. 2-10. Supercritical Operation. 2-11. Elementary Reactor Operation.

2-12. Depletion. 2-13. Elementary Reactor Operation with Negative Tem-

perature Coefficient. 2-14. Fission-product Poisoning. 2-15. Inventory of

Items Affecting Reactivity. 2-16. Control-rod Effectiveness.

CHAPTER 3. REACTOR KINETICS 29

3-1. Introduction. 3-2. Solution of Kinetic Equations for Step-function

Input in Sk. 3-3. Solution of Kinetic Equations for Ramp Function. 3-4.

Approximate Solution of Kinetic Equations for a Ramp Function for a

Critical Reactor. 3-5. Solution of Kinetic Equations for Sinusoidal Input

Sk.

CHAPTER 4. ^AUTOMATIC REACTOR CONTROL 48

4-1. Elementary Reactor as a Control Device. 4-2. Reactor Representa-

tion with Temperature Coefficient and Poisoning Feedback Loops. 4-3.


Generated on 2011-10-24 11:37 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Negative Temperature Coefficient Feedback. 4-4. Poisoning Feedback.

4-5. General Requirements for Automatic Power-level Control. 4-6. Gen-

eral Description of a Reactor Automatic-control System. 4-7. Control-

loop Response. 4-8. On-Off-type Reactor-control-system Operation. 4-9.

Transient Response of Control Loop. 4-10. Determination of Control-loop

Performance by Simulation Technique. 4-11. Peak Limiting by Negative

Temperature Coefficient. 4-12. Evaluation of Transient Response. 4-13.

Procedure for the Selection of Control-system Constants.

CHAPTER 5. REACTOR CONTROL MECHANISMS 98

5-1. General Requirements of Control-rod Mechanisms. 5-2. Motors and

Mechanisms for Control Rods: Nonpressurized Systems. 5-3. Pressurized

Control-rod Drive Systems. 5-4. Scramming Mechanisms. 5-5. Energy

Storage Devices. 5-6. Buffers. 5-7. Rod Position Indication. 5-8.

Horsepower Requirements.

vii
viii CONTENTS

CHAPTER 6. NUCLEAR POWER PLANT CONTROL 124

6-1. Introduction. 6-2. Description of Basic Elements of a Nuclear Power

Plant. 6-3. Steady-state Programming. 6-4. Elementary Thermo-

dynamics of the Basic Loop. 6-5. Transfer Function Representation of

Basic Plant Components. 6-6. General Recapitulation of the Dynamic

Performance of the Basic Plant. 6-7. Temperature Feedback Loop Analy-

sis. 6-8. Coolant Mixing. 6-9. Flow Changes. 6-10. Analysis for Multi-

ple-section Reactor. 6-11. Application and Limitations of Temperature

Feedback Loop Transfer Functions. 6-12. Temperature Coefficient Reac-

tivity Feedback Loop. 6-13. Transient Analysis. 6-14. External Reactor

Control System. 6-15. Automatic Plant Control. 6-16. Stability Analysis

for Demand Loop.

CHAPTER 7. REACTOR CONTROL RADIATION DETECTORS 187

7-1. Measurement Problem. 7-2. Ranges of Measurements. 7-3. Descrip-

tion of Instruments. 7-4. Effect of Gamma Radiation on Instrument

Responses. 7-5. Effects of Reactor Operation and Rod Shadowing on

Neutron Measurements. 7-6. Temperature Effects on Neutron-measuring

Instruments. 7-7. Instrument Calibration and Intercalibration. 7-8.

Instrument Circuits.

CHAPTER 8. OPERATIONAL CONTROL PROBLEMS: STARTUP 213

8-1. Neutron Sources. 8-2. Initial Reactor Startup. 8-3. Subsequent

Reactor Startups. 8-4. Operational Startup Requirements. 8-5. Safety

Startup Considerations. 8-6. Poison Considerations. 8-7. Startup Con-

trol Systems.
Generated on 2011-10-24 11:37 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

CHAPTER 9. OPERATIONAL CONTROL PROBLEMS: POWER OPERATION . . . 245

9-1. Requirements for Reactivity Changes at Power Level. 9-2. Automatic

Control.

CHAPTER 10. OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 256

10-1. Shutdown Philosophy. 10-2. Fundamentals of Scram Protection.

10-3. Accidents. 10-4. Alarms and Cutbacks. 10-5. Last-ditch Emer-

gency Shutoff Measures. 10-6. Scramming Circuits.

CHAPTER 11. SIMULATORS 282

11-1. Elementary Analogue Computing Techniques. 11-2. Reactor

Kinetic Simulators. 11-3. Subcritical Reactor Simulator. 11-4. Xenon

Simulator. 11-5. Power-plant Simulators.

Problems 303

Index 309
CHAPTER 1

I NTRODU CTION

1-1. Introduction and Purpose. At present the state of nuclear power

plants in this country is a fluid one, with many technical ramifications

being entwined with political considerations. Nevertheless, in the fields

of reactor and nuclear-plant control several ideas have been crystallized

and are already regarded in terms of long-standing theory. As there is

no universal agreement regarding the best type of power plant, there

obviously can be no agreement as to the best type of control system.

Each reactor plant that has been built thus far contains a different con-

trol system. These control systems differ radically in mechanical design,

but many common theoretical problems and basic design concepts have

arisen. An attempt will be made in this text to present these common

points.

Another aim of this book is to present an elementary picture of reactor

and nuclear-plant control for the new group of control engineers now

entering this field. Historically, nuclear power plants grew from nuclear

reactors, which in turn grew from basic nuclear physics. The detailed

understanding of the design and synthesis of a nuclear reactor is a com-

plex subject steeped in intricate mathematics and clothed in security.

It is fortunate that the control problems of nuclear reactors can be han-

dled by simplified conventional methods which are now familiar to those

in the servomechanisms field. However, it is often necessary for the

control designer to make certain assumptions and simplifications con-

cerning nuclear reactors, which in some cases may create concern on the
Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

part of the nuclear physicists that they and the control designers are not

talking about the same terms.

It is now generally recognized that the nuclear-power business is in a

transition stage from the physicists to the engineers. The plants that

have been constructed are as complex in their own way as are the basic

physical equations upon which the reactors are founded. The engineer

therefore tends to regard the reactor only as a component in a much

larger system, and consequently he deals with it in conventional engineer-

ing terms which are compatible with the rest of the system. The phys-

1
2 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

icist in turn has been more concerned with the intricate details of the

internal reactor structure and tends to regard the plant as an auxiliary

device which is a necessary evil.

The problem of reactor control has existed since the first reactor and

has been the subject of extensive study for many years. The problem

of nuclear plant control is a newer one, and the answers are not as well

known.

1-2. Analogy of a Nuclear Power Plant to a Direct-current-generator

System. Let us consider a reactor operating by itself serving no function

other than perpetuating a chain reaction. This type of operation might

be compared with the open-circuit no-load opefation of a d-c generator.

Tying a load onto the reactor and extracting power from it would corre-

COOLANT

PUMP CONDENSATE

PUMP

FIG. 1-1. Block diagram of elementary nuclear power plant containing pressurized water

reactor and conventional steam system.

spond to tying a load onto the d-c generator. In the case of the generator

it would easily be anticipated that the load would affect the generator

characteristics. Historically, it was not quite so apparent that the power

plant would affect the characteristics of the reactor. This reaction could

occur in a complex plant from many sources. Even minor auxiliary

devices could find their performances reflected back on the basic reactor

performance. In order to gain a better appreciation of this problem


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

from an over-all point of view, the reactor plant, d-c generator analogy

can be pursued further by an illustrative example.

1-3. Example of Analogy of a Nuclear Power Plant to a Direct-

current-generator System.1! Let us assume that our nuclear power

plant consists of a pressurized water-cooled reactor system and a conven-

t Superior numerals in the text correspond to the numbered References at the end

of each chapter.
INTRODUCTION 3

tional steam-turbine system as shown in Fig. 1-1. In this plant high-

pressure water is used to cool the reactor and extract heat from it. This

heat is transferred to the secondary loop in a steam-generator system

consisting of a boiler and a steam separator. The output loop of the

plant contains a steam turbine, condenser, and all the necessary auxil-

iaries. The turbine is directly coupled to a load, in this case presumably

an electric generator. Both the primary coolant and steam systems are

closed loops.

Control Program. Many types of programs of primary and secondary

parameters can be set up for a plant of this sort, depending upon the

components and local specifications. As a direct relationship exists

between the water temperatures of the primary loop and the steam tem-

perature and pressure of the secondary loop, a control program may be

specified from either loop. For the purpose of this discussion the plant

operation will be specified from the primary loop in terms of the coolant

temperatures at the reactor inlet and at the reactor outlet as functions

of power level. The following symbols are used:

Q = total reactor power output

Th = coolant temperature at reactor outlet

Tc = coolant temperature at boiler outlet

Tw = average coolant temperature = (Th + Tc)/2

Ts = steam temperature at outlet of steam generator

ps — absolute steam pressure at outlet of steam generator

Hs = enthalpy of steam at outlet of steam generator


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Hx = enthalpy of exhaust steam at turbine outlet, for isentropic

expansion

Hw = enthalpy of feed water

F, = rate of steam flow

Let us assume that our control program is such that the average tempera-

ture of the primary loop coolant is held constant regardless of the load

requirements of the secondary portion of the plant. This so-called con-

stant-T.v program causes no change in primary coolant volume as the

power output is changed, and a small simple water pressurizer may be

used. The flow of water created by the pump is at a fixed rate and does

not change as a function of the power level. The specific relationships

between the primary and secondary temperatures of this type of plant

control are shown in Fig. 1-2, for the arbitrary condition of T.v = 500°F.

It can be seen in this plant that the steam temperatures fall off very

rapidly as the power output is increased. This fall in steam temperature

calls for a corresponding drop in steam pressure.

Thermodynamic Analysis. In analyzing this plant it can be seen that

the power output of the reactor is proportional to Th — Tc. The con-
4 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

stant primary coolant flow is assumed such that, again using arbitrary

numbers, at full output of the reactor Th — Tc = 50°F. The power

transferred from the primary-coolant water to the secondary loop is

proportional to T,v — Ts. The proportionality constant depends on the

power rating and on the boiler dimensions. For illustrative purposes

let us again assume that at unity power T.v — T, = 60°F, which will be

designated as rated full power. The numerical values of all these tem-

700

800 r

600 700

500 600

£ 400 iu~ 500

o o:

8I

g2

Q. UJ

300 400

<

UJ

uj

200 300

100

200

100
Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

23456

REACTOR POWER

FIG. 1-2. Temperature and pressure control conditions for plant having constant-Tat,

program.

peratures are the ones given in Fig. 1-2. In this elementary plant the

steam leaving the steam generator is of high quality, but it is not super-

heated. However, we can assume that the steam generator furnishes

dry and saturated steam at all power levels. The steam pressure then

depends only on the steam temperature and may be obtained from steam

tables.

If we assume that the changes in potential energy and kinetic energy

of the steam are negligible compared with changes in enthalpy throughout

the steam loop, the power delivered to any component becomes simply

F, &H, where F, (Ib/hr) is the steam-flow rate and &H (Btu/lb) is the

enthalpy drop across the component. Then F,, Hw, and Hx can be cal-
INTRODUCTION 5

culated, and the turbine output power may be plotted as a function of

reactor power. The normalized output power and efficiencies are plotted

in Fig. 1-3. The curves represent fully the over-all steady-state perform-

ance of this type of nuclear power plant.

The Electrical Analogue of the Constant-T^ Program. The most direct

d-c generator analogue for the situation just described is to permit

enthalpy drop AH to be represented by a voltage drop AF and the steam

flow F, by a current 7. The product F, AH represents power, as does the

product / AV. The d-c circuit that is roughly analogous to the nuclear

power plant is shown as Fig. 1-4. This circuit has a source of power, a

power-consuming section representing condenser loss, and a load repre-

senting shaft power. These three components are in series since the

components of the steam loop are in series. The reactor and steam gen-

erator are simulated by the d-c generator. The generated voltage corre-

sponding to Hs — Hw drops about 10 percent from no load to maximum

load. The condenser is represented by an opposing battery and resistor.

The battery is used because the power lost in the condenser is more nearly

proportional to Fs than to Fs2. The turbine is represented by a variable

resistance. The output characteristics of this analogous plant are given

in Fig. 1-5, and they roughly approximate the plant characteristics shown

in Fig. 1-3.

From the above analogy it can be seen that the steady-state over-all

plant performance of a nuclear power plant is not too dissimilar from that

of conventional systems. It can also be seen that in the particular type


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

of plant control presented, the control is strongly related to steam-plant

control and the role of reactor control is apt to be a subservient one. The

nuclear-power-plant-control engineer thus finds that his education must

be a mixture of nuclear physics and servomechanisms with a strong over-

tone of conventional thermodynamics.

1-4. Philosophy of Reactor and Plant Control. Before setting out to

design a specific reactor plant, the designer of the control system must have

a complete philosophy of operation in mind. Up to the present this philos-

ophy has been the one in which the control system as well as all other

auxiliary components must be supersaf e. The peculiar position of nuclear

power plants has been such that if one were to blow up inadvertently,

the resulting publicity would severely harm the entire program of nuclear

power for several years. For this reason it is to be anticipated that

control-system design philosophy in the future will also be of the super-

safe variety.

It should be pointed out that technically possible accidents which might

occur in a nuclear power plant are not so severe as might first be imagined.

It is popularly thought that the principal difference between an "atomic

bomb" and a nuclear power plant is one of control—in the first case the
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

energy is given off instantly, in the second case it is given off slowly. In

the event of a failure of the control system the obvious thought is that the

power plant might become a bomb. Hurwitz2 has indicated that a con-

1200

1000

90

40

30 y

20 i

<

10

23456

REACTOR POWER

FIG. 1-3. Performance of constant-Tao plant as a function of reactor power output.

siderable matter of degree is involved in a nuclear-reactor accident.

Roughly, the damage caused in a reactor accident would approximate the

damage caused by an amount of TNT equivalent in weight to the amount

of uranium in the nuclear reactor.

This accident obviously creates


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

less damage than the destruction

wrought by an atomic bomb. But

from a philosophical point of view,

R ,_b it will undoubtedly be many years

-AAA/ |l|l ' before nuclear power plants are

regarded by the public in the same

TURBINE

POWER

CONDENSER LOSS

FIG. 1-4. Analogous d-c-generator system.

classification as other industrial

plants. Consequently, every effort

must still be expended to obtain supersafe control systems.

Conventional Power-station and Airplane Philosophies. Under this

given supersafe over-all philosophy of operation, two subphilosophies of


INTRODUCTION

reactor shutdown are available to the control designer. These sub-

philosophies may be called the conventional power-station philosophy

and the airplane philosophy. In the conventional power-station phi-

losophy, the generator and other expensive items in the circuit must, in

the event of a significant component failure in any portion of the plant,

be disconnected and shut down at once in order to protect the large

investment involved. Other machines are easily available to take up

the load, and no great harm is done by taking any one machine off the

120

100

>

Q 60

Ul

a:

UJ

§40

20

50

40

30

o
Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

UJ

20|

10

1 33456

GENERATED POWER

FIG. 1-5. Performance of d-c-generator system as a function of generated power.

line. In the case of a hypothetical nuclear power-generating station,

under this philosophy the reactor would be shut down as fast as possible,

not only for the above reason but also because if it were not shut down,

it would be likely to aggravate any failure situation by continuing to

pour out power.

In the airplane type of philosophy, component failures must protect

themselves, and components, including reactors, must fail in such a

manner as not to endanger the system. In other words, the airplane

engine cannot be shut off for any external cause or the entire structure

may be lost. Both these philosophies are available to present-day

nuclear-power-plant designers, and the one that should be used obviously

depends upon the specified operating situations of the plant. That these
8 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

philosophies affect basic design is quite apparent. In a given plant, for

example, the desire to protect the reactor may be so strong that all of the

reactor instrumentation might be provided in triplicate, while in the

remainder of the plant only single-channel instrumentation would be

used.

Automatic-control Philosophy. Another philosophical point that must

be determined quite early in the design of a reactor-power-plant control

system is the degree of automatic control desired. Here again the ulti-

mate usage of the plant greatly affects the decision. For example, a

military plant might conceivably have less automation than a central

station nuclear power plant, on the grounds that operating costs for labor

are not particularly important in the military plant. In considering

over-all safety, the proposition has been advanced for extensive use of

automatic control rather than manual control, on the basis that, although

automatic systems have been known to fail, they can be made to fail

safely. When an automatic system is operating properly, it never makes

a mistake. The same cannot be said for the human being as a control

element.

1-5. Control-system Specification. It is interesting to note at this

point that the nuclear-plant-control designer does not specify many of

the basic conditions that he has to meet. The reactor designer will

specify the amount of heat that can be taken from the reactor and how

much overload for how long a period of time will be permitted. The

pump designer will specify how much cooling fluid can be circulated.
Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

The boiler designer will specify given steam temperatures and pressure

ranges. The metallurgist will specify maximum temperatures through-

out the plant, which must not be exceeded if excessive corrosions and

strains are not to exist.

The control-system designer must tie all these factors together in indi-

vidual and over-all control loops in such a manner that everyone is sat-

isfied. Then he conceivably has time to ask himself the inevitable ques-

tion: Is it stable? Fortunately, as will be shown later, most nuclear

power plants are inherently stable, but it is well known that even the

most stable basic system can be upset by improper detailed control design.

1-6. Scope of Text. The ultimate scope of this book is to lead new

nuclear-reactor-control designers to the final stage of obtaining an over-

all picture of the requirements of nuclear-power-plant-control design.

It is assumed that the reader has a basic knowledge of the elementary

processes of reactor physics.3 This text will first present a review of the

elementary physics of reactor control. Engineering symbols are used in

this review, and very little hint as to the complex basic physics problems

involved is presented. Rather, the reactor is regarded as a "black box"

and its performance described by simple external measurements. Once


INTRODUCTION 9

this physical picture is established, a mathematical presentation of the

response of a basic reactor to various types of driving functions is pre-

sented, and the transfer function of the reactor is derived for future use

as an element in a control system. A reactor control loop is then pre-

sented, and its response to transient disturbances studied. Since a partic-

ularly interesting and critical element of the reactor control loop is the

actual output-control mechanism, a diversion is taken at this point to

describe the reactor control requirements and to present some examples

of present-day practices. The stage has now been set for a short glance

at over-all plant control, and some of the basic problems of plant response

and programming will be looked at.

The details of some of the special reactor components that are not

familiar to the control designer will then be studied. Nuclear instru-

mentation is pursued and its problems are presented. The operation of

a nuclear power plant is then investigated because it can be shown that

certain types of operations affect the individual control-element design.

Startup problems, power range operational problems, and shutdown prob-

lems are given. Finally, because it is realized that an inescapable urge

often exists on the part of any designer to try out his device, Chap. 11

deals with electronic plant simulators. By means of these computing

machines the hazards existing with attempts to try out a new device on an

actual reactor plant may be eliminated.

REFERENCES

1. Schultz, M. A., and J. N. Grace: A Simple Analogy to a Nuclear Power Plant,


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

WAPD-T-38, Westinghouse Atomic Power Division, Pittsburgh, Pa. Declassified,

June, 1953.

2. Hurwitz, H., Jr.: Safeguard Considerations for Nuclear Power Plants, "Proceed-

ings of the 1953 Conference on Nuclear Engineering," University of California

Press, Berkeley, Calif., 1953.

3. Glasstone, S., and M. C. Edlund: "The Elements of Nuclear Reactor Theory,"

D. Van Nostrand Company, Inc., New York, 1952.


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
ELEMENTARY PHYSICS OF REACTOR CONTROL 11

to be in a fixed solid form. A description of the control of circulating-fuel

reactors or others in nonsolid form will not be attempted. The fuel is in

close proximity to, and intermingled with, a moderating material such

as hydrogen, beryllium, or any suitable light element. The moderating

material is used to slow down fission neutrons to where they possess the

desired energy spectrum. The combination of fuel, moderator, and asso-

ciated structural components will be called the core. Passing through

the core and in intimate contact with it is a heat-transfer material.

Gases, water, or liquid metals may be used as the heat-transfer material.

The moderator may also be circulated through the core and serve as the

heat-transfer material. Outside the core proper is a reflector which is

used to conserve neutrons and "bounce" them back into the core in an

optical reflector sense. Surrounding the reflector is a biological shield

which serves the purpose of attenuating the radiations emanating from

the core. This shield is usually a combination shield which attenuates

both neutrons and gamma rays. Inside the core or the reflector are

located the control rods, the basic purpose of which is to regulate the

power level of the core by controlling the number of neutrons in it.

2-2. Fission Process. Inside the core there exists an initial source of

neutrons from some radioactive decay process. When a neutron of a

given energy is absorbed by a uranium nucleus, there is a finite probabil-

ity of splitting this nucleus into two or more fragments. This process is

called fission, and a large amount of energy is produced in this fission

process (approximately 200 Mev per fission). In the act of fission two
Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

to three neutrons are released from the fragmentation, and these neutrons

are capable of creating more fissions in other uranium nuclei under the

proper conditions. The two to three neutrons that are produced in fis-

sion may take part in several reactions, all of which are competitive. A

neutron may be absorbed in core material other than uranium. It may

be lost to the core by leaking out of the system. In any event, for a chain

reaction to take place, for each uranium nucleus capturing a neutron and

undergoing fission, a minimum of one neutron on the average must be

produced, which in turn creates the fission of another nucleus.

Reactor Types. A nuclear reactor, then, is a system usually consisting

of a moderator, a fuel containing fissionable material, heat-removing

means, and a geometric structure in which a chain reaction can be main-

tained. In the fission process so-called fast neutrons are produced.

These fast neutrons have high energies. Inside the reactor these neu-

trons may suffer scattering collisions, mainly elastic, as a result of which

their energy is decreased. And as mentioned, they may also be absorbed

by the various materials in the system or be lost through escape from it.

Depending upon the relative amounts and nature of the moderator, fuel,

other substances, geometrical arrangement, and the dimensions of the


12 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

system, the main portion of the neutron absorptions by uranium leading

to fission will take place within a certain energy range.

If most of the fissions result from the capture of neutrons which have

been slowed down to thermal energies by collisions with the moderating

material, the so-called thermal neutrons, the system is referred to as a

thermal reactor. When most of the fission processes are caused by the

absorption of neutrons of higher energy, sometimes called intermediate

neutrons, the term intermediate reactor is used. The usual range of

neutron energies in an intermediate reactor is from thermal energy up to

about 1,000 ev. If the main source of fissions is the capture of fast

neutrons directly by the fuel without the neutrons having suffered any

energy losses, the system is called a fast reactor. Power reactors in gen-

eral are of thermal and intermediate types; fast reactors are usually used

in weapons and will not be discussed further.

Multiplication Factor. The chain reaction condition that each uranium

nucleus capturing a neutron and undergoing fission must ultimately yield

a minimum of one neutron, which in turn also causes fission, leads to a

definition of a multiplication factor k. The multiplication factor k may

be defined as the ratio of the number of neutrons in any one generation

to the number of corresponding neutrons of the immediately preceding

generation. If k is equal to or slightly greater than unity, a chain reac-

tion can take place. If k is less than unity, the chain reaction cannot

persist and will ultimately die down.

For the chain reaction in the core to keep going, the production of
Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

neutrons must equal the leakage plus the absorption of the neutrons.

Therefore k = production/ (leakage + absorption). In this text the

symbol k will be used to refer to the multiplication factor of the so-called

infinite pile, in some texts referred to as &,-n/. The term k — 1 is defined

as kex (k excess) and again is usually rigorously used in conjunction with

an infinite pile. The term kei represents the amount the multiplication

differs from unity, and as most reactors operate around unity, kex gives

a more accurate picture as to the state of the reactor.

The more common multiplication factor which is used in conjunction

with specific reactors is fce// (k effective), which is the effective multiplica-

tion factor for a given finite-sized reactor. Reactivity is defined for a

finite specific reactor in a similar manner to kex as

P=

In this text the symbol Sk will be used for reactivity, meaning the amount

the multiplication factor of a specific reactor differs from unity or

Sk = p = keff, ~ 1 (2-2)

Keff
ELEMENTARY PHYSICS OF REACTOR CONTROL 13

From an engineering point of view these terms are generally used in the

neighborhood of a multiplication factor of unity. For control problems

of a general nature k and keff have been used in the past interchangeably,

and k,x and Sk have similarly been loosely interchanged. The cause of

this confusion stems from the fact that for control purposes the internal

structure of the reactor and the details of the multiplication in the core

are not important. The control designer obtains his information from

external measurements of an over-all type. To him the reactor has a

given multiplication factor. He usually does not concern himself about

the size, shape, and composition of the reactor.

Neutron Lifetime. The average time between successive neutron gen-

erations in an infinite reactor is defined as the neutron lifetime I. The

symbol I* is used for the mean effective lifetime of a neutron in a finite

reactor containing uranium 235. In other words, I* is the mean time

which elapses from when neutrons are produced in fission until they return

again to fission or are lost to the reaction. The term /* may be con-

sidered as

I* = I (2-3)

Again, strictly speaking, this equation is the result of a "one-group"

theory calculation, or it applies rigorously only to large reactors.1

2-3. Neutron Level. The excess of neutrons in a finite reactor from

one generation over the preceding generation is then Sk. If there are

initially n neutrons per cubic centimeter present in the core, the rate of

increase in each generation is nSk. If I* is the effective time between


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

succeeding generations,

dn Sk ,0 .,

di = I* n (2-4)

and integrating this equation yields

n = noe(«*"*" (2-5)

where no is the number of neutrons per cubic centimeter initially and n

is the number after a lapse of time t. On this basis the number of neu-

trons rises exponentially with time if the effective multiplication factor

is greater than unity.

The number of neutrons in the core is proportional to the number of

fissions occurring, and for 3 X 1010 fissions per second 1 watt of power is

produced. The power output of a reactor then is proportional to the

number of neutrons in the core in any given time interval, and the symbol

n is used to designate neutron level, with the implication that a power

level is involved.
2-4. Reactor Period.

defined as

CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The period of a reactor can arbitrarily be

Period =

(l/n)(dn/dt)

(2-6)

with the inverse reactor period (l/n)(dn/dt) being the quantity usually

measured. We can solve for the period of the reactor of Eq. (2-5),

where n = noe(«*/i>:", and find

Period = - =

OK

sec

(2-7)

Therefore in terms of period, Eq. (2-5) becomes n = noe'/T. An alter-

nate definition of reactor period may now be given. The period of a

reactor is that amount of time which the reactor would take to change

its level by a factor of e = 2.716. It will be observed that the period of

a reactor is a dynamic quantity. That is, when the reactor is in opera-

tion at a fixed power level, the period is infinite. Only when the reactor

is changing its level is there a finite measurable period.

2-5. Reactor State. The state of a reactor at any given instant is

defined by the use of the multiplication factor. When k = 1, the reactor

is said to be critical, k < 1 subcritical, and k > 1 supercritical. It will


Generated on 2011-10-24 11:38 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

be noted that no power level is involved in the definition of criticality.

A reactor may be critical at a level of 1 watt or a megawatt.

Delayed Neutrons. In the above equations it has been assumed that

all the neutrons created in the fission process were given off instantly

and had a lifetime of I. Actually a small fraction of the neutrons created

in fission are given off at discrete amounts of time after the actual fission

process occurs. These neutrons which are produced after fissioning rep-

resent approximately 0.75 percent of the total neutrons produced and are

called delayed neutrons. The heart of reactor control depends upon

delayed neutrons. Table 2-1 indicates the properties of the delayed neu-

TABLE 2-1. Properties of Delayed Neutrons Given Off in

(j236 Fission Process by Thermal Neutrons

Mean life

ti, sec

Decay constant

Xi, sec"1

Fraction of total neutrons

0.071

0.62

2.19

6.50

31.7

80.2

14.0

1.61

0.456

0.151

0.0315

0.0124

0.00025

0.00084

0.0024

0.0021

0.0017

0.00026

14
ELEMENTARY PHYSICS OF REACTOR CONTROL

irons which are given off in the uranium 235 fission process by thermal

neutrons.4'6'7 The delayed neutrons are given off in six distinct groups

at different times and in different quantities. The symbol /3 is used to

denote the total fraction of the delayed neutrons with ft being the frac-

tion of the delayed neutrons in the iih group of delayed neutrons. Sim-

ilarly, Xj represents the decay constant of the ith group of delayed neu-

trons. For certain problems it is convenient to treat all of the delayed

neutrons as a single delayed group having a total fraction /3 = 0.0075 and

an average decay constant X = 0.1 sec~1.

2-6. Prompt Critical. When the effective multiplication factor of a

reactor is 1.0075, the reactor is said to be prompt critical. This state-

ment means that the reactor would be capable of sustaining a chain

reaction without the use of the delayed neutrons. If fc is greater than

1.0075, extremely rapid exponential multiplication of reactor power level

results. For this reason most control systems attempt to prevent k

from ever becoming greater than 1.0075.

2-7. Subcritical Level Operation. Let

us assume that we have a neutron-

multiplying medium which has a multi-

plication factor fc < 1 and a given neu-

tron lifetime I, as shown in Fig. 2-2. A

control rod which may be considered

merely a device for absorbing neutrons,

and thus ensuring that our multiplying


Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

medium is subcritical, may also be added.

Let us insert in this medium a source of

neutrons. Such sources exist in nature

naturally from cosmic rays, or neutrons

may be artificially provided from radio-

active isotopic mixtures such as radium beryllium or polonium beryllium.

Under the condition of subcriticality it can be shown simply that the

number of neutrons which exist in this multiplying medium at the end

of a sufficiently long interval of time is

MULTIPLYING MEDIUM

NEUTRON SOURCE

FIG. 2-2. Black-box representation

of a subcritical reactor having a

multiplication factor k and a mean

neutron lifetime /.

n = n0(l + k + fc2 +

or in closed form

1 - k"

(2-8)

(2-9)

At the end of a sufficiently long interval of time f or k < 1, this equation

degenerates into

n_ _ 1

(2-10)

no 1 — fc
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

This ratio is known as the subcritical multiplication factor, and all reac-

tors exhibit this effect. The result of this equation is shown graphically

in Fig. 2-3 for a source suddenly inserted into a multiplying medium.

With a k of 0.5 in the multiplying medium, the number of neutrons con-

sequently levels off at the end of several lifetimes I to a value of n/no = 2.

The subcritical multiplication factor of this reactor then is 2. If one

were to change k to 0.9 by removing part of the control rod from the

medium, the subcritical multiplication would ultimately be 10, and so

on. As k approaches 1, the subcriti-

cal multiplication factor approaches

infinity and the number of neutrons

in the medium rises in a straight line

4 8 12 16 20

TIME, NEUTRON LIFETIMES

FIG. 2-3. Subcritical multiplication.

24

-4 -2 0

PER CENT REACTIVITY

FIG. 2-4. Relative power level as a

function of reactivity remaining in the

reactor for infinitely slow reactivity

change.

with time. It will be observed that as the subcritical multiplication

factor becomes higher and higher, more time is always taken for the
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

medium to settle out at a given level. And finally it does not settle out

at all but continues to rise. This situation holds only if there is a source

present. Without an actual source, the neutron level in any subcritical

medium must ultimately die down to zero.

16

Let us now examine how the power level changes in this subcriticar

reactor as we increase the k of the medium slowly toward unity by

removing the control rods. The simplest case is to consider pulling the

rods at an infinitely slow rate, and in this case the total multiplication is

always the subcritical multiplication. Because the withdrawal rate is


ELEMENTARY PHYSICS OF REACTOR CONTROL 17

so slow, the decay times of even the longest-lived delayed emitters are

short in comparison with the time for a noticeable reactivity change.

Therefore all the delayed neutrons have ample time to be emitted before

the power changes appreciably. The number of neutrons present then

conforms at each instant to the subcritical multiplication formula of

Eq. (2-10). Figure 2-4 shows the neutron-level build-up under this

condition. The curve is a true hyperbola and approaches criticality

asymptotically. The curve is plotted so that when k = 0.9 the power

level is 1.

2-8. Subcritical Period. The period that results from this very slow

pulling rate can be obtained from the definition of period of Eq. (2-6).

Performing the suggested differentiation on the subcritical multiplication

formula results in the expression

Period =

From this expression it can be seen that if the rate of change of k is con-

stant with time, then the period decreases directly as the reactivity

remaining in the reactor becomes smaller. For our example of extremely

slow rod pulling at a constant time rate of change, the period approaches

zero as the medium approaches criticality.

2-9. Critical Operation. It will be recalled from the definition of

criticality that k = 1 for this condition to exist. No mention was made

of sources in this definition; consequently, although it may have been

inferred that the power level of the medium was constant with k = 1, it

is obvious from Sec. 2-7 that the source neutrons continue to add in and
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

create a rising power level. From a practical point of view this phenom-

enon is noticeable only at extremely low operating levels and is useful

in determining the operation of reactors in a low power condition. A

reactor operating at a power level high enough to produce useful power

represents at criticality a steady multiplication, by one, of billions of

neutrons. The usual reactor source strength may vary from a few neu-

trons per second to possibly a few million neutrons per second. The

number of neutrons involved in the source emission then is only a minute

percentage of the number of neutrons involved in a power operation.

Consequently, for all practical purposes at power operation, k = 1 rep-

resents a state of constant power level.

2-10. Supercritical Operation. Equation (2-5) indicates how a

reactor behaves when k > I if all the neutrons were prompt. An illus-

trative example of the change of level under this condition may be given.

Let us assume that we have a critical multiplying medium with an I* of

10~3 sec. This is roughly the value of I* for a large graphite moderated

reactor. Now let us suddenly insert a reactivity change of + Sk = 0.003


18 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

into the reactor. Equation (2-5) then indicates that at the end of 3 sec

the power level will have risen by a factor of 8,000.

Let us now examine the situation when delayed neutrons are present.

Our neutron-level equation becomes of the form6

dn Sk ft .

=n~n

where (7,- is the concentration of the delayed neutrons emitted of group i

and the other symbols have the same meanings as previously described.

d is defined by

f -f»-^ <2-13)

The rate of change of n has the contribution of the delayed neutrons

subtracted from the prompt neutrons, but of course the concentration of

delayed neutrons coming in from the past must be added to make up the

total rate of change.

Complete solutions to the above equations will be given in Chap. 3 for

various types of Sk disturbances. For comparison purposes and in order

to obtain a feeling for the effects of the delayed neutrons, it is interesting

to solve the approximate equations that result when the delayed neutrons

are assumed all to be bunched in one group which has an average value

of X of 0.1 sec~1 (see Ref. 5). ft then becomes /3 and d becomes C.

Let us also assume that Sk is small so that k and /ce// can be used inter-

changeably. Under these conditions, Eqs. (2-12) and (2-13) become

dn Sk — 0 . . „ ,0 .

^ = -^-n + XC (2-14)
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

f-£»-XC (2-15)

The solution of these equations is a summation of two exponential terms

of the forms

- c-a

—

n(0) o — o b — a

where a - Ap ^f _ gfc ' (2-17)

c = S (2-19)
ELEMENTARY PHYSICS OF REACTOR CONTROL

If the assumption is further made that, when practical numbers are

used, product terms containing A can be neglected, Eq. (2-16) becomes

simply

â„¢ (t\ a nl-

(2-20)

n(Q)

- Sk

- Sk

In order to compare actual numbers with the prompt neutron example

previously given, let us again insert a Sk of +0.003 into a critical reactor

having an I* of 10~3 sec. Then

n(0)

= 1.67e°-067' - 0.67e-4-6'

(2-21)

A plot of this result is shown in Fig. 2-5. The effect of the delayed neu-

trons is at once apparent. It will be recalled from Eq. (2-5) that with

only prompt neutrons, the power level soared to 8,000 times the original

level in 3 sec when +0.0035& was inserted into the multiplying medium.

3.5

3.0

2.5

2.0
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

1.5

1-0

0.1

0.2 0.3

10

0.5 1.0 235

TIME, SECONDS

FIG. 2-5. Relative neutron level as a function of time for a step reactivity change of 0.003.

From Fig. 2-5 it can be seen that at the end of 3 sec the power level has

risen by a factor of only 2.1. Thus the effect of a mere 0.75 percent of

delayed neutrons is such as to make the entire problem of reactor control

a simple feasible one rather than a most difficult if not impossible one.

An examination of the approximate equation (2-20) shows that the

first term ultimately predominates, and after a few tenths of a second,

the second term may be neglected. The second term contributes to

what is called a transient period, whereas the first term creates a so-called

stable reactor period. The stable period of Eq. (2-20) would then be

T=

= 15 sec

(2-22)

- Sk

\Sk
20 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

for the example just presented as against the prompt neutron period of

Eq. (2-7)

3P - L = 0.33 sec (2-23)

o/c

for the corresponding prompt neutron example.

2-11. Elementary Reactor Operation. The role of the control rods

must now be examined in more detail. It will be recalled that the pro-

duction of neutrons equals the leakage plus absorption for critical opera-

tion. If it is desired to change the production of neutrons, there are the

two choices of manipulating either the leakage or the absorption. To

change the leakage one might mechanically put a hole or a window in the

reflector. Changing the multiplication by absorption is the more com-

monly used method, particularly for thermal reactors; and control rods

of cadmium, boron steel, and other high-thermal-neutron-cross-section

materials may be used to absorb neutrons from the reaction. It will be

assumed that the control rods mentioned in this book are of the absorptive

type and are located in the core unless otherwise specified.

Control rods may be moved in or out of the reactor singly or in banks.

They are given various names according to their functions. Certain

groups of rods may be designated as safety rods or shutoff rods. Other

rods may be shim rods, whose function is to affect the power level in a

coarse manner. A regulator rod is often used to cause fine changes in

power level.

It will be recognized that a multiplication factor k > 1 must inherejitly


Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

beTSuiIt into the reactor and then some reactivity removed by inserting

control rods partially into the reactor in order to obtain critical operation.

The total multiplication the multiplying medium possesses when the

control rods are completely extracted minus one will be termed excess

reactivity. One minus the total amount of reactivity in the medium

when the control rods are completely inserted will be called the shutdown

reactivity of the reactor.

Let us now examine the method of changing the power level in a simple

reactor. Figure 2-6 illustrates the process. Ajssume that the reactor

is initially critical at a low power level and it is desired to increase jhis

level. The first step_is__tp extract a control rod a small amount and

change the multiplication factor k from 1 to a value slightly greater than

I^_ The neutron level Lheu starts to risej_jx>ughly in accordance with

Eq. (2-20)1 As the power level rises and approaches the desired ultimate

level, it is obvious that the control rod must then~be~inserted back to

wKenTTi; =TT IF no anticipation is provided in the system, it can be

seen from Mg. 2-6 that the control rod must oscillate about the k = 1

position, but ultimately it will settle down at the original position from
ELEMENTARY PHYSICS OF REACTOR CONTROL

21

which it started. We have then a system in which the power level is

independent of rod position, and in order to change power level one moves

a control rod temporarily in or out of the medium and then returns it

right back to its original position.

2-12. Depletion. The preceding reactor concept is the classic one of

reactor control. It is recognized that control rods may have to be moved

during the lifetime of a reactor because of fuel depletion. That is, as the

uranium in the reactor is used up, the number of fissions occurring will

III)

CRITICAL POSITION

FINAL LEVEL

TIME

FIG. 2-6. Control-rod positions for a manual-level change.

decrease, consequently k will be reduced and control rods will have to be

moved out to compensate for the reduction in k.

2-13. Elementary Reactor Operation with Negative Temperature

Coefficient. Other causes exist for moving control rods, one of them

being the temperature of the reactor. Most reactors have what is termed

a negative temperature coefficient. This term means that as the reactor

heats up, its reactivity is reduced. Reactors which have water or gas

as moderators usually have large negative temperature coefficients. This

temperature coefficient will later be shown to be a most important control-

system parameter, but for the present let us examine the basic operation
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

of a reactor with a negative temperature coefficient.

Assume that a reactor is critical at a low power level and consequently

is effectively at room temperature. The control rods are in a given


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

fixed position. Now, through some external means such as that of heat-

ing the normal coolant of the reactor, let us raise the average tempera-

ture of the reactor to where it might actually run as a power reactor.

This process of heating will reduce the reactivity of the reactor. Con-

sequently, the reactor is no longer critical, but probably greatly sub-

critical. Control rods must then be extracted to make up for this loss

of reactivity. Actually it does not matter whether the heat is applied

from an external source or whether the reactor power level is changed.

Cold critical then refers to the position of the control rods when the reac-

tor is critical at room temperature, and hot critical refers to the position

COOLANT OUT

CONTROL

ROD

'out

ToV

Tin

POWER LEVEL

POWER LEVEL

COOLANT IN

CONSTANT

TEMPERATURE

FIG. 2-7. Elementary operation of a reactor having negative temperature coefficient and

constant coolant inlet temperature.

of the control rods when the reactor is critical at its normal operating
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

temperature.

An interesting type of control-rod operation results when the inlet

temperature to this type of reactor remains constant. Figure 2-7 illus-

trates this condition. For a simple example assume that the coolant

into the reactor is supplied from a faucet or equivalent at constant tem-

perature and the coolant out of the reactor is used in heating radiators

and then dumped. If the flow from the faucet is constant, the tempera-

ture versus power level is shown in Fig. 2-7. Here, as the power output

of the reactor rises, the average temperature rises, and to compensate

for this rise in average temperature, control rods must be extracted from

the reactor in order to keep the multiplication factor unity. Assuming

then that the control rods are equally effective at all positions in the

22
ELEMENTARY PHYSICS OF REACTOR CONTROL 23

reactor, it can be seen that under this condition the position of the control

rods is directly proportional to the power output of the reactor. In a prac-

tical situation many other types of programming can exist. The position

of the control rods will rarely be either of the two simple functions of

power level just described.

2-14. Fission-product Poisoning. It has been indicated that control

rods, depletion, and temperature affect the multiplication factor of a

reactor. Another quantity must also be introduced which plays a vital

role in the operation of large thermal power-producing reactors. This

quantity is fission-product poisoning. As the reactor continues to oper-

ate, fission products are created from the uranium. Many direct fission

products exist, and in addition, a host of daughter nuclides are created by

decaying emissions from these fission products. Some of these direct

and indirect nuclides may have large cross sections for the absorption of

neutrons, and therefore they can act as poisons. If these poisons are

produced in appreciable amounts, they can affect the over-all multiplica-

tion of the reactor. Because some of these poisons may continue to be

formed by radioactive decay even after the reactor is shut down, the con-

centration of the poison may increase to a maximum after reactor shut-

down. It is apparent then that additional excess reactivity must be

designed into a thermal reactor to take care of these poisons.

Because of their large thermal neutron absorption cross sections, two

nuclides are of particular interest, xenon 135 and samarium 149. Xenon

135 is formed as a result of the decay of the direct fission product tel-
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

lurium 135. Tellurium 135 actually consists of over 5 percent of the

direct fission products and decays rapidly by ft emission in the following

manner:

1 min 6.7 hr 9.2 hr 2.1X106 yr

> Cs136 - -* Ba136

Barium 135 is stable. Xenon 135 has an absorption cross section for

thermal neutrons of approximately 3.5 X 106 barns and decays to cesium

with a half life of approximately 9.2 hr.

Sm149, on the other hand, is the stable end product of the chain

1.7 hr 47 hr

Nd149 - » Pm149 - » Sm149 (stable)

This reaction occurs in roughly 1.5 percent of the fissions, and the Sm149

has a somewhat lower cross section for thermal neutrons of approximately

5.3 X 104 barns. For these reasons it does not contribute as much to

the poisoning of a reactor as does the Xe136 and consequently the Sm149

effect will be ignored in future control discussions on poisoning.

In order to introduce terminology that we shall use later in discussing

reactor control, the above-described relationships can be put into mathe-

matical form as follows.


24 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The equation for the concentration of Xe136 in a reactor at any time

becomes

rlX'

^- = (7,2/ - ffJC')* + \J' - \*X' (2-24)

where X' = number of atoms of Xe136 present per cubic centimeter at

any time t

yx = fractional yield of xenon as direct fission product

trx = microscopic thermal-neutron absorption cross section of

Xe136 (3.5 X 106 barns)

<£ = thermal-neutron flux

Xi = decay constant of I136

/' = number of atoms of I136 present per cubic centimeter at any

time t

\x = decay constant of Xe136

£/ = macroscopic fission cross section of fuel in reactor

A similar equation can be written for the concentration of I136 at any

time. Because the half life of Te136 is very short compared with that of

I136 (1 min against 6.7 hr), we can make the simplifying assumption that

the direct fission product is I136. Then

^-' = T,S/0 - XJ' (2-25)

where Ti is the fractional yield of I136 from the direct fission process and

the other terms have the meaning just defined.

It can be seen that two kinetic effects occur which concern reactor

control. The first effect is the so-called equilibrium poisoning, and the
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

second the peak poisoning. The equilibrium poisoning occurs during

reactor power operation, and the peak poisoning after shutdown of the

reactor from a high power level. From Eq. (2-24), the amount of xenon

present in the reactor builds up from the concentration of the iodine 135

and dies off both by the radioactive decay into cesium and from the

destruction of xenon 135 by thermal-neutron absorption. It is obvious

that, after long operation at a given fixed power level, an equilibrium can

exist between the build-up and the two decays, so that at the end of a long

period of time a steady amount of poison can exist in the reactor. Fig-

ure 2-8 shows the time scale to build up to equilibrium for such a process.

The amount of poisoning involved depends upon the steady power level

of the reactor and the design of the particular reactor involved. The

steady-state xenon concentration may be obtained simply by setting

dX'/dt = 0 in Eq. (2-24) and dl'/dt = 0 in Eq. (2-25). Then


ELEMENTARY PHYSICS OF REACTOR CONTROL 25

Let us now shut down a thermal reactor as fast as possible from a high

power level operating condition. In this instance the thermal-neutron

level is reduced to effectively zero, and consequently the decay term due

to thermal-neutron absorption of the xenon 135 no longer exists. Conse-

quently, the xenon 135 concentration builds up to a maximum from the

iodine 135 which has been previously formed. Ultimately the radio-

RELATIVE EQUILIBRIUM

XENON POISONING REACTIVITY

O 0 o O r4

^O K> t> i" <a O

) 10 20 30 40 50 60 70 80 9

TIME, HOURS

FIG. 2-8. Equilibrium xenon poisoning build-up for an enriched thermal reactor
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

10

40

20 30

TIME AFTER SHUTDOWN, HOURS

FIG. 2-9. Relative peak xenon poisoning reactivity as a function of time after shutdown for

a thermal reactor.

active decay of the xenon 135 to cesium 135 takes over and the total

xenon 135 concentration drops off. The time involved in this process is

shown in Fig. 2-9, with a peak in xenon concentration appearing in

approximately 11 hr after shutdown. The magnitude of this peak again

depends upon the initial steady power level of the reactor and its specific

design. However, the peak xenon poisoning may be many times the

value of the equilibrium xenon 135 poisoning. It will be recognized that

an entire range of poisoning conditions can exist, depending upon the


2« CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

magnitude of the initial power level and the extent to which the power

level is shut down.

Other variables of operation can enter. After a reactor has been shut

down and the xenon poison concentration partially built up, the reactor

may be turned on again and subsequently burn out the xenon back to

the equilibrium condition. This operation can call for a fast rate of

change of reactivity from the control rods to keep the reactor critical

and is one determining factor in the speed of reactor control.

From the curve on Fig. 2-9 a very appreciable amount of reactivity

may be involved in the peak xenon poisoning. It is conceivable that this

amount of reactivity may be so much that the reactor does not contain

sufficient uranium to completely "override" this peak even when the

control rods are pulled out all the way. Under these conditions, where

only a fixed amount of reactivity is available, unless the reactor is started

up quickly after a shutdown, a large period of time will exist in which it

will be impossible to start the reactor until the xenon decays down. As

an example from Fig. 2-9, let us assume that sufficient reactivity exists

in a reactor that it can still be made critical up to J^ hr after shutdown

from a given power level. The figure indicates that unless a startup is

made within this Y% hr, it may be 40 hr before the reactor can be started

up again.

2-15. Inventory of Items Affecting Reactivity. It has been shown that

control rods must be moved to compensate for depletion, temperature,

and poisoning. If these effects are defined in reactivity worth, control


Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

rods may also be so rated. Sufficient extra reactivity must be designed

into the control rods to hold the reactor safely shut down when all rods are

inserted. The amount of negative reactivity needed to keep a reactor

safely shut down differs from reactor to reactor. Approximately — 2 per-

cent in reactivity is probably the minimum value used. A range of 5

to 10 percent in negative reactivity is more usual for reactor shutdown.

To form a reactivity inventory, let us again take an example. Assume

that a given reactor changes its reactivity by 5 percent going from cold

critical to hot critical, it has a 3 percent reactivity allowance for depletion

and low cross-section poisons, and 10 percent in reactivity is needed to

completely override xenon poisoning after shutdown. An excess reactiv-

ity of 17 percent is designed into the machine, and it is desired that the

shutdown reactivity be 9 percent. Shutdown reactivity is defined here

as the amount of negative reactivity in a new cold reactor when all the

rods are fully inserted. For this particular set of conditions it can be

seen that rod motion will be required to overcome 18 percent in reactivity

for the depletion, temperature, and peak poisoning effects. With an

excess reactivity of 17 percent, the reactor will not fully override the

xenon poisoning on the last day of its rated life provided that it is started
ELEMENTARY PHYSICS OF REACTOR CONTROL

27

up hot. Rather, it will override the xenon completely for two-thirds of

its life and in the last third it will be necessary to start up the reactor

cold, to operate the reactor in such a manner that the poisons do not build

up fully, or to provide for a possible waiting period. The 9 percent shut-

down reactivity requires that the reactivity value or rod worth of the

rods fully inserted into the cold clean reactor must be at least 26 percent.

The inventory just completed is crude, as rod worths usually change

with temperature and poisoning. These effects would be taken into

account in an actual reactor plant design.

Many control schemes can be conceived so that poisons are not per-

mitted to build up beyond the capabilities of the control system. It is

an interesting coincidence, however, that with normal one-shift operation

of a reactor plant on an 8- to 12-hr-day basis, starting up the next morn-

ing is usually a startup made quite close to the peak of the xenon poison-

ing. Therefore, on occasion, even the type of working day must be con-

sidered in the design of how much reactivity shall be built into the reactor.

2-16. Control-rod Effectiveness. The above discussions have assumed

that the effects of control-rod position in reactivity would be linear with

<v

CONTROL

>

%
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

J_

ALL

IN

ALL

OUT

ALL

OUT

FIG. 2-10. Control-rod effectiveness as a function of rod position in a reactor.

rod position in the reactor. Actually this is not the case. Moving a

control rod or a bank of control rods in a simple geometric reactor such as

a cylinder normally produces a change in reactivity which varies approxi-

mately as the sine squared of the rod position. To make the matter more

complex, the position of each individual rod usually can affect the reac-

tivity worth of other nearby rods. The sine-squared approximation

may generally be used for control purposes even though it is found to be

incorrect at the end points. It would not be expected that the worth of

a rod would have a zero slope at its end positions. Figure 2-10 indicates

the differential reactivity and the total reactivity of an elementary reactor

having this type of rod worth. It can be shown that the peak-to-average

ratio of rod effectiveness is approximately 3:1. It will also be shown

later that for safety considerations it is quite important to know at what

rate of reactivity change the reactor goes through critical during a startup
28 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

problem. Consequently, the factor of 3:1 must be taken into account

for safety considerations as well as the obvious nonlinear change in loop

gain servo considerations.

The minimum rod effectiveness is also important in problems such as

are encountered in following transient xenon poisoning by means of con-

trol rods. An attempt should be made to design the reactor control-rod

positions so that minimum rod worth does not occur when the xenon

transient is decaying down at its maximum rate and consequently is

inserting reactivity into the reactor at its maximum rate. For this con-

dition the control rods must put in reactivity fast enough to override the

xenon poisoning.

REFERENCES

1. Glasstone, S., and M. C. Edlund: "The Elements of Nuclear Reactor Theory,"

D. Van Nostrand Company, Inc., New York, 1952.

2. Stephenson, R.: "Introduction to Nuclear Engineering," McGraw-Hill Book

Company, Inc., New York, 1954.

3. Murray, R. L.: "Introduction to Nuclear Engineering," Prentice-Hall, Inc., New

York, 1954.

4. Hughes, D. J., J. Dabbs, A. Cahn, and D. Hall: Delayed Neutrons from Fission

of U236, Phys. Rev., vol. 73, no. 2, p. Ill, 1948.

5. Blizard, E. P., and F. S. Maienschein: Sources of Radiation, ORNL-56-6-12Q, 1952.

6. Soodak, H., and E. C. Campbell: "Elementary Pile Theory," John Wiley & Sons,

Inc., New York, 1950.

7. Sun, K. H., et al.: Delayed Neutrons from U23s and Th232 Fission, Phys. Rev., vol.
Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

79, no. 1, p. 3, July 1, 1950.

8. Phillips, H. B.: "Differential Equations," John Wiley & Sons, Inc., New York,

1939.
CHAPTER 3

REACTOR KINETICS

The first two chapters have provided background from which we can

start to examine design problems and operating problems of reactor con-

trol. Later the complex problems of controlling reactors associated with

power plants will be discussed. Before the control of reactors and plants

can be investigated, more must be known about the kinetic characteristics

of the reactor. In servo language, before synthesizing the control of the

reactor as a servo component, it is necessary to know its open-loop

responses. In later operating problems, responses to step functions,

ramp functions, and sine waves all fit into specific parts of reactor opera-

tion. It is the analytic solutions of the reactor response to these driving

functions which will be examined in this chapter.

3-1. Introduction. These responses will be studied from a black-box

type of approach; i.e., the only significant reactor constants that are

present are Sk, I*, and n in a multiplying medium. The temperature

coefficient will be presumed to be zero. Measurements of these quanti-

ties can be made external to the reactor, and its performance calculated.

It is well known, of course, that spatial effects inside the reactor can

cause changes in reactivity and output. However, the one-point black-

box type of treatment usually yields kinetic answers that are consistent

with experimental accuracy.1

The response of the reactor to step changes in Sk will first be developed.

Then solutions will be provided for ramp functions in Sk. Finally the

reactor will be treated by methods of servo theory, whereby a sinusoidal


Generated on 2011-10-24 11:39 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

input in Sk is compared with a sinusoidal output in power level. In

this way the transfer function will be derived for future use in control

circuits.

3-2. Solution of Kinetic Equations for Step-function Input in Sk. The

kinetic equations of a chain-reacting pile have been derived in the litera-

ture many times.2~4 We shall use the form previously developed in

Chap. 2.

29
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

•

dn Sk —

Hi

• i

(3-D

(3-2)

where the symbols have the meanings previously denned. The nature

of the solutions of Eqs. (3-1) and (3-2) for step function inputs is also

well known.6'6 These kinetic equations can be combined to form a

8k

0.010

0.008

0.006

0.004

0.002

t"=10"3 SEC

.If =10"* SEC"

. l*=10"5 SEC

0.01

AI

0.1

P,, SECONDS-'

n—rm
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

10

FIG. 3-1. Chart for response of reactors to a step change in reactivity, positive step only.

single differential equation of the seventh order in n. For a step input

in Sk the solution will take the form

n(0 =

(3-3)

where the first exponent PI has the same sign as Sk, the input disturbance,

and where the other six exponents are negative.

30

Equations (3-1) and (3-2) are a family of linear differential equations

with constant coefficients, and there is a definite relationship among the

values of Aj, Pji and Sk. This relationship can be shown graphically
REACTOR KINETICS

j=7 j=6 .

31

0.006

0.004

0.002

8k

-0.002

-0.004

-0.006

5 j=4 j«3 j=2

-10,000 -1000 -100 -10 -1 -0.1 -0.01 -0.001

Pj, SECONDS"1

0'8[7TTTTT~I I in ! i M i T Mill i I I

-1.2

-1.6

-2.0

FIG. 3-2. Chart for response of reactors to a step change in reactivity.

and is presented for reactors of I* = 10~3 sec, 10~4 sec, and 10~6 sec in

Figs. 3-1 and 3-2. Figures 3-1 and 3-2 are used when a positive 5/c step

is involved, and Fig. 3-2 only is needed for negative Sk steps.

These curves are used in the following manner: The size of the Sk step

is selected and the value of the exponent Pi is read from the top graph of

Fig. 3-1 or the farthest curve to the right top of Fig. 3-2. Then reading
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

vertically down from that value of PI to the bottom graph, the corre-
32 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

spending coefficient Ai may be read. In a similar manner, one continues

across the top curves of Fig. 3-2, from right to left, locates the value of

Pj opposite the ordinate of the selected Sk, and then reads down to the

corresponding coefficient Aj.

Using these charts in an example, one might derive the equation for

the response of a reactor having an I* = 10~4 sec when a Sk step of

0 Z 4 6 8 10 12 14

TIME, SECONDS

FIG. 3-3. Relative neutron level versus time for positive step function reactivity changes,

/* = 10-4sec.

— 0.0035fc is inserted into the reactor. Reading across and down for each

exponent and coefficient, the equation for n(f) can be found as

n(f) = tt0(0.19e-°-012< + 0.24e-°-022' + O.lOe-"-11' + 0.01e-°-8s'

+ 0.026-1-6' + 0.005e-131 + 0.28e-110') (3-4)

It can be seen from Figs. 3-1 and 3-2 that the positive exponent PI

increases rapidly with increasing Sk and the coefficient AI also increases

very rapidly. These rapid rises take place at values close to prompt

critical.

Figure 3-3 shows the neutron-level response of a reactor having an

I* of 10~4 sec for various step Sk inputs. The value of I*, as noted from
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
REACTOR KINETICS

33

Figs. 3-1 and 3-2, principally affects the front edge of the rise. After

an initial rise caused mostly by the prompt neutrons, the reactor settles

down to a steady period created by the delayed neutrons. Figure 3-4

indicates the response of this reactor for longer periods of time, and it

can be seen that for step reactivities in the neighborhood of Sk = 0.005 or

greater, extremely rapid and long rises occur.

20

40

100 120

140

60 80

TIME, SECONDS

FIG. 3-4. Relative neutron level versus time for positive step reactivity changes, /* = 10~4

sec, large time scale.

Figures 3-5 and 3-6 illustrate the performance of a reactor having an

I* of 10~4 sec, for negative 8k step inputs. Here it will again be noted

that after an initial drop, the rate of drop is determined by the delayed

emitters and the ultimate rate by the longest-lived delayed emitter, which,

from Table 2-1, has a time constant of approximately 80 sec. These

effects can be seen mathematically by examination of Eq. (3-4). It can

be seen that the last term of this equation is the predominant one for very

short periods of time. However, the two terms preceding the last term

in Eq. (3-4) have comparatively small coefficients. The first four terms
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

have small exponents. Therefore, after the large initial drop, the neu-

tron concentration changes slowly and behaves ultimately as indicated

by the first term of Eq. (3-4).


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

4 6 8 10

TIME, SECONDS

FIG. 3-5. Relative neutron level versus time for negative step function reactivity changes,

/* = 1(T4 sec

20

120 140

40 60 80 100

TIME, SECONDS

34

FIG. 3-6. Relative neutron level versus time for negative step function reactivity changes,

/* = 10~4 sec, large time scale.


Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
REACTOR KINETICS

35

KT

ur1

After the first dropoff, the reactor level cannot fall off at a rate faster

than an 80-sec period. Figure 3-7 illustrates this condition. It will be

noted that this period determines

the maximum rate of neutron shut-

down of a reactor. Some reactors

are capable of being shut down

by 10 decades. To reach such a

shutdown level would require a

minimum of 30 min.

Approximate Solution for Initial

Response.7 The shape of the ini-

tial response of a reactor to a

step change in 5/c can usually be

analyzed by an approximation

method. During an initial time

interval of the order of J^o sec,

the delayed-neutron emitters can

be considered a constant source

of neutrons. If the reactor is in

equilibrium, the delayed-neutron

emitters yield /3 neutrons for each


Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

neutron produced. That is, they

are acting as a constant source of strength d3/Z*)n0. Hence the initial

io*

KT

10 15 2O

TIME. MINUTES

25

30

FIG. 3-7. Relative

shutdown.

neutron level after

1.0.

0.2

0.4 0.6 0.8 1.0

TIME, SECONDS

FIG. 3-8. Front edge shape for a step function change in reactivity showing results of using

approximate formula.
36 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

response of the reactor is described by

dn 6k - ft 8

dt I* P ° ( )

The solution of Eq. (3-5) is

n(0) Sk -ft ' Sk -ft"

The neutron density in this expression approaches the asymptotic value

of — 8/(Sk — 8). How close the approximate expression matches the

actual case can be seen in Fig. 3-8. Here the initial rise created by the

insertion of a 0.002SA; step into a reactor having an I* of 10~3 sec is com-

pared with the approximate expression. The front edges are alike for

the first few tenths of a second.

3-3. Solution of Kinetic Equations for Ramp Function. Injecting a

ramp function of the form Sk = a + yt into the black box produces use-

ful information, as this is an approximation of what happens when a

control rod is pulled out of a reactor. It will be recalled that control-

rod effectiveness is usually such that the reactivity does not change

linearly as a function of rod position. Nevertheless, useful information

can be obtained by considering linear rates of reactivity change and mod-

ifying the slope of these linear reactivity rates of change in discrete

intervals if desired.

The exact solution of the reactor kinetic equations for this type of

ramp input is a complex one, tedious to calculate, and will only be outlined

below, f Where many solutions are required, the use of a reactor kinetic

simulator, as described in Chap. 11, is usually preferred to the direct


Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

analytical method.

The analytical method follows these steps: the basic kinetic equations

(3-1) and (3-2) are first presumed to consist of only one group of delayed

neutrons having a single value of X and 8. It can be shown that the

solution is not particularly affected when several groups of delayed neu-

trons are used. Then Eqs. (3-1) and (3-2) become

dn Sk — 8 i \n \ g

~dt = I* " ' ' ' '

^ = p n - \C (3-8)

These equations are then combined to form a second-order differential

equation in n whereby

t This solution was originally obtained by Sylvan Wallach, and the presentation

here closely follows his method with his permission.


REACTOR KINETICS 37

d*n , ft + \I* - Sk dn Sk\ ^ „ ., _.

w + ------ p ---- ^ - - -JT n = \S (3-9)

and as Sk = a + 7^, Eq. (3-9) can be written in the form

/72« •/«

H + (At + B) ^ + (« + Z))n = \S (3-10)

The solution of this equation is represented as an integral expression

similar to a Laplace transform, whereby

n = ex'R(x) dx (3-11)

£

R (x) is denned as

and P(x) and Q(x) are the polynomials

P(x) = x2 + Bx + D (3-13)

and Q(x) = Ax + C (3-14)

Equation (3-11) is a proper solution of Eq. (3-10), provided that a

suitable integration path £ is selected in the complex x plane. We can

now substitute Eq. (3-11) into Eq. (3-10) and find that

/ e«R(x)[t(Ax + C) + (x2 + Bx + D)] dx = \S (3-15)

£

Substituting the definitions of Eqs. (3-13) and (3-14) into Eq. (3-15)

/ ex'R(x-)[Q(x)t + P(x)] dx = AS (3-16)

£

The suggested integration of Eq. (3-16) is then performed, first inte-

grating by parts as follows:

/ ex'R(x)Q(x)tdx = f UdV (3-17)


Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

_p

where U = R(x}Q(x) (3-18)

and dV = te* dx (3-19)

The expression for \S is then found to reduce to

\S = [R(x)Q(x}et]fl (3-20)
38 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

where Xi and x2 are the end points of the path of integration £. R(x)

may be found by substituting Eqs. (3-13) and (3-14) into Eq. (3-12)

= o (x + 0" '

R(x) = o x + exp x + B - <fc (3-21)

where , - D ~ ~ (3-22)

and <T is an arbitrary constant of integration depending upon the limits

of integration imposed by each problem.

The final solution in terms of the original parameters becomes

n(t) = ff ex(Jt-«/7-x"V27(x + X)0x/T-i dx (3_23)

£

The constant a and the path £ remain to be determined. The complete

solution of Eq. (3-10) is given by two linearly independent solutions of the

homogeneous equation (S — 0) and by any solution of the inhomogeneous

equation. Accordingly, three paths of integration are required. Along

these paths Eq. (3-11) must converge and Eq. (3-20) must be satisfied

with appropriate values for S.

For the case encountered in reactor startup problems, the path of

integration £ can be taken as the positive real axis from zero to infinity.

As Sk is continually increasing, 7 is positive. Then the x2 term in the

exponential of Eq. (3-23) is always negative. The x term is negative

when Sk is below prompt critical, zero at prompt critical, and positive

when the reactivity exceeds prompt critical. Hence it is clear that as the

reactivity increases with time, the neutron density begins to grow very

rapidly at prompt critical, but below prompt critical the growth is com-
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

paratively small.

For simple computational purposes a may be evaluated as that mul-

tiplication factor which brings n(0) to a value of 5/1 — k at t = 0.

Figure 3-9 presents a solution to the ramp function input problem for

Sk = -0.333 + 7 X 10~4Z (3-24)

and I* = 10~4 sec.s

As can be seen from Eq. (3-23), the neutron-level response is not partic-

ularly sensitive to I*. It will be noted that below critical the form of the

response is that of the subcritical multiplication curve (Fig. 2-4) and

above prompt critical the reactor literally "takes off."

The effect of different linear reactivity change rates may be seen from

Fig. 3-10. s The curves are plotted as a function of the reactivity remain-

ing in the reactor. Here the subcritical multiplication curve is furnished


REACTOR KINETICS

39

as a reference curve, being the case of infinitely slow reactivity change.

It can be seen that as reactivity is inserted into the reactor at higher and

higher rates, the critical point comes at lower and lower neutron levels.

Of interest also is the reactor period as a function of linear reactivity

rates of change. Figure 3-11 indicates the periods that result as the

multiplication factor of the reactor comes closer and closer to unity

at linear rates of change. It is of interest to note that at high reactivity

change rates, a short period, easily detectable by measuring instruments,

470 480

TIME AFTER REACTIVITY CHANGE

STARTS, SECONDS

FIG. 3-9. Relative neutron level versus time

for ramp function input.

-6 -5 -4 -3 -2 -1

PER CENT REACTIVITY

FIG. 3-10. Relative neutron levels versus

reactivity remaining in a reactor for

various ramp function reactivity changes.

is available at quite low multiplication factors. Whereas if the rate of

change of reactivity is small, the multiplication factor must be very nearly

unity before a period in the neighborhood of 20 sec results. Figure 3-11

is somewhat deceptive in that it might possibly be interpreted as meaning

that one should, in an actual startup operation of a reactor, extract the


Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

rods rapidly in. order to see quickly a measurable period. Figure 3-12

is a plot of the same information as a function of time, and here it can be

seen that one actually has a somewhat more gradual operating approach

to a given period if the reactivity is inserted at a slower rate. Startup

rates of reactivity change are discussed in detail in Chap. 8.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

3-4. Approximate Solution of Kinetic Equations for a Ramp Function

for a Critical Reactor. A problem that arises quite frequently in theo-

retical reactor operation is one in which the reactor is critical and reac-

tivity is steadily increased because of malfunctioning of a control rod.

The following assumptions are made: First, the reactor is operating at

1.25

6K CHANGE AT 0.0076/SEC

6k CHANGE AT 0.0038/SEC

6k CHANGE AT 0.0016/SEC

0.90 0.92 0.94 0.96 O.98 1.0

MULTIPLICATION FACTOR k

FIG. 3-11. Reactor period versus multiplication factor for various ramp function reactivity

change rates.

a steady state before the application of the disturbance. Second, the

disturbance starting at time zero has the form of Sk = At.

On the basis of these assumptions, for short-time intervals we can

develop this solution.

Equation (3-1) takes the same form as approximate equation (3-5)

dn

dt

Sk —

now, however,

Therefore

I*
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

n+

(3-25)

(3-26)

(3-27)

40

Sk = At

dn _ At - ft

dt I* H
REACTOR KINETICS

41

2.5

*-6l

fik

CHANGE

CHANGE

CHANGE

6k CHAN

AT 0.00

AT 0.00

AT O.OC

GE AT 0

re/SEC

38/SEC

H6/SEC

00045/S

EC-*

10

//

'/
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

20

40

-sS

40

240

80 120 160 200

TIME, SECONDS

FIG. 3-12. Reactor period versus time for various ramp function reactivity change rates.

The solution for this equation is of the form

n = [exp(J - P dO]Jexp(/P dt)]Q dt + C\ (3-28)

where P = ?<[~l" Q = ~j^ (3-29)

II

Then

a' C\

e-u^,d< + _)- (3.30)

— 7!k

n0 r

Since at t = 0, n/n0 = 1, C/n0 must equal I*/ft, the integral term in

Eq. (3-30) may be rewritten in the form

2Al*

where

IA

Jo

and

7 = —

(3-31)

a ~ \2Z*
42 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Substituting at + y = n, we get

rt

Jo

- dt = -

2"

r> /*o;(-T-'y

— /

VTT Jo

(3-32)

This form is that of the probability integral whose values are available

in mathematical tables.9 The complete solution then becomes

(3-33)

Plotted from this equation, Fig. 3-13 shows the results of suddenly apply-

ing rates of change of Sk to a critical reactor. It is easily seen that put-

8k CHANGE

'AT 0.06/SEC

100

8k CHANGE

0.012/SEC

400

500
Generated on 2011-10-24 11:40 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

200 300

TIME, MILLISECONDS

FIG. 3-13. Relative neutron level versus time for various ramp function reactivity change

rates, using approximate formula.

ing in reactivity at finite rates into a reactor, instead of in step function

fashion, results in considerably slower rates of initial rise of neutron level.

This type of calculation has been presented in some detail, as it is useful

in operational problems involving study of accidents.

n0

3-5. Solution of Kinetic Equations for Sinusoidal Input in 8k.u In the

dynamic analysis of any system it is often necessary to determine the

frequency response or transfer function10'11 of each of the elements of

the system. The concept of a transfer function for a reactor implies a

sinusoidal variation of Sk. Consequently, the set of linear differential

equations (3-1) and (3-2) is no longer a set with constant coefficients, a


REACTOR KINETICS 43

necessary condition for developing a transfer function. However, if

we assume sufficiently small excursions of n, these equations can be

approximated by a set with constant coefficients.

Let n be composed of two parts: no, a steady-state value, and Sn, a

small excursion. Similarly, let d be composed of a steady-state Cio and

an excursion Sd. Equations (3-1) and (3-2) can be rewritten first by

combining Eq. (3-2) with Eq. (3-1)

dn Sk ,„ ,

(3-34)

r\

because / fa: = ft. Then

dn dSn Sk , SkSn v vwt ,„ „_,

\o-oo)

8k8n/l* may be neglected in comparison with (Sk/l*)na and

= ft (no + sn) _ Xi(C<0 + id) (3-36)

as in the steady state

OL/i'O n Pi -i x-y /o 0*7^

—37- = w = 7* ^o ~~ AiOio t^o-o/;

Equation (3-36) becomes

(3-38)

Reducing Eqs. (3-35) and (3-38) to Laplace-transform operational form

gives
Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

sSn(s) = ^ Sk(s) - s £ <5C,-(s) (3-39)

»=i

s5C,-(s) = ^ «»(«) - Xi6C,-(s) (3-40)

where the initial condition transforms have been dropped. This is per-

missible since we define a transfer function in terms of the steady-state

response. Combining Eqs. (3-39) and (3-40) gives


or

CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

n (* + A<)

Sn(s) _ no i = i

I* <.,

s [] (s + n)

(3-42)

At this point it is necessary to evaluate the roots of the sixth-degree

equation formed by expanding the denominator in Eq. (3-42) and equat-

-8

0.001

100

0.01 0.1 1.0 10

FREQUENCY, CYCLES PER SECOND

FIG. 3-14. Derivation of reactor transfer function amplitude using break frequency method

ing it to zero. The method is outlined in Refs. 11 and 12. For an

example of I* = 10~4 sec, the roots are found to be

ri = 77.0 r4 = 0.336

r2 = 13.38 r6 = 0.080

r-3 = 1.43 r6 = 0.0147

The resulting complete transfer function for I* = 10~4 sec is

sl f \ T*

Sk(s) I*

(s + 14) (s + 1.61)(s + 0.456)(s + 0.151)(s + 0.0315) (s + 0.0124)


Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

0.0136s(s + 77) (s + 13.38) (s + 1.43) (s + 0.336) (s + 0.0805) (s + 0.0147)

(3-43)

44

The complex fractional term is now the normalized frequency response

plotted in Fig. 3-14.


REACTOR KINETICS

45

From this data it is now possible to plot the frequency response char-

acteristic (Bode diagram) 13 of the pile, replacing s by j<a. This is shown

in Fig. 3-14, using the break frequency method.10

-12

I*=5x10~5 SEC

l*=1.25x1CTs SEC

-V*=1Cr4SEC

-16

0.001 0.01 0.1 1.0 10

FREQUENCY, CYCLES PER SECOND

FIG. 3-15. Amplitude of reactor transfer function.

100

20

g-40

uj

< -60

UJ

<n

a -80
Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

-100

-120

I I I I I Ml 1 1 I I I

/-V

-r=10-« SEC

l*-5x!0"*SEC

l*=1.25x10'3SEC-

0.001

0.01

III

10

0.1 1

FREQUENCY. CYCLES PER SECOND

FIG. 3-16. Phase shift of reactor transfer function.

100

It can be seen from the foregoing analysis that a change in I* changes

the gain and the roots. Figure 3-15 gives the amplitude response of the

reactor_for_varipus values of Z*7 ~Tigure7:WiEows the phase "shiiT under

The I* = 10~4-sec curve "Is" arrjitrafiTy" hormalfzed

so that 0 db falls at 1 cycle/sec. This requires the multiplication of


46 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the gain factor no/l* by the gain of the frequency dependent portion,

•

n («+ ^)

—^j— — at 1 cycle

s [I (s + r4)

t=i

For many engineering purposes the expression of Eq. (3-43) can be

simplified by expressing the parentheses in the numerator as a single

lead term and those of the denominator as a single lag term. Thus

Sn(s) _ n0 . (s + X)

Sk(s) ~ ^ Al A,s(s + f)

The choice of values of X and f, the average delayed-neutron-decay time

and average root, is a matter of engineering judgment. One choice for

Eq. (3-43) might be X = 0.075 sec~1, and f = 50 sec~l. The correspond-

ing value of AI would then be 0.021.

The principal feature to be noticed from the transfer function derived

above as Eq. (3-41) is the nonlinearity of this transfer function. That

is, the small signal sinusoidal gain of the reactor as a circuit element

depends upon the level at which it is operating. This situation is an

intolerable one for a reactor control loop, and the dependence upon level

must be removed before satisfactory automatic control of a reactor can

be achieved. This problem is discussed in Chap. 4.

The use of this type of transfer function to describe a reactor as a

control-loop element has been justified on the basis of a classic experi-


Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ment performed by Harrer, Boyer, and Krucoff on the CP-2 reactor.1

disturbance in Sk was produced by oscillating a control^

rod sinusoidally over a small amplitude, and fHe neutron output of the

_rpa(^.Qr measured as a function of the input disturbance frequency. The

results obtained were so convincing tfiaT"reactor transfer functions are

used with as much confidence in present-day control design as is the

transfer function of a conventional amplifier.

REFERENCES

1. Harrer, J. M., R. E. Boyer, and D. Krucoff: Transfer Function of Argonne CP-2

Reactor, Nucleonics, vol. 10, no. 8, p. 32, 1952.

2. Glasstone, S., and M. C. Edlund: "The Elements of Nuclear Reactor Theory,"

D. Van Nostrand Company, Inc., New York, 1952.

3. Soodak, H., and E. C. Campbell: "Elementary Pile Theory," John Wiley & Sons,

Inc., New York, 1950.

4. Goodman, C.: "The Science and Engineering of Nuclear Power," vol. 142,

Addison-VVesley Publishing Company, Cambridge, Mass., 1947.


REACTOR KINETICS 47

5. Isbin, H. S., and J. W. Gorman: Applications of Pile-kinetic Equations, Nucleonics,

vol. 10, no. 11, p. 68, November, 1952.

6. Hurwitz, H.: Derivation and Integration of the Pile-kinetic Equations, Nucleonics,

vol. 5, no. 1, p. 61, July, 1949.

7. Lundby, Arne: Kinetic Behavior of a Thermal Heavy Water Reactor, "Pro-

ceedings of the 1953 Conference on Nuclear Engineering," University of California

Press, Berkeley, Calif., 1953.

8. Schultz, M. A.: "Automatic Control of Power Reactors," AECD-3163, 1950.

9. Lowan, A. N.: "Tables of Probability Functions," Federal Works Agency of the

WPA, National Bureau of Standards, Washington, 1941.—-

10. Brown, G. S., and D. P. Campbell: "Principles of Servomechanisms," John Wiley

& Sons, Inc., New York, 1951.

11. Chestnut, H., and R. W. Mayer: "Servomechanisms and Regulating System

Design," John Wiley & Sons, Inc., New York, 1951.

12. Franz, J. P.: Pile Transfer Functions, AECD-3260, 1949.

13. Bode, H. W.: "Network Analysis and Feedback Amplifier Design," D. Van

Nostrand Company, Inc., New York, 1945.

14. Bowen, J. H.: Automatic Control Characteristics of Thermal Neutron Reactors,

Proc. Inst. Elec. Engrs. (London), vol. 100, pt. 1, p. 102, 1953.
Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CHAPTER 4

AUTOMATIC REACTOR CONTROL

The response of the reactor to various types of driving functions having

been determined, it is necessary to review again basic reactor operation

before attempting to use the reactor as an element in a control loop. In

this chapter we shall first examine the reactor more closely as a control

component. How the elementary reactor can be considered using servo-

mechanisms technique and how it is modified by temperature coefficient

and poisoning will be shown. When the various forms of the reactor

transfer function have been indicated, control loops will be tied around

the reactor and the system examined for stability and transient response.

In this chapter only automatic reactor control in the power-level range

will be discussed. Startup and shutdown problems will be presented

later.

4-1. Elementary Reactor As a Control Device. Let us now examine

the reactor as a control element. Figures 3-15 and 3-16, in conjunction

with Eq. (3-41) defining the reactor transfer function, completely describe

an elementary reactor for control purposes. It will be noted that from

ji transfer function point of view, Eq. (3-41), the break point of highest

frequency is determined by the largest root Si, which^in turn, depends

upon the value of fi/l*. It is the value of this root which determines the

time behavior that distinguishes a thermal reactorfrom an intermediate

or fasTTreactor.In a fast reactor the value of I* is so small that the last

break Si occurs at a very high frequency, usually above 100 cycles. If

an attempt is made to devise a control in this frequency range, difficult


Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

component problems arise. Fortunately it is not necessary to devise a

control system operating clear out through the last break in the frequency

response. From a physical point of view, controlling out to this last

break represents control on prompt neutrons.

From another point of view, one can examine the open-loop transient

response of a reactor to a step function such as is given in Figs. 3-3 and

3-4. It will be noted that an attempt can be made to contain the initial

steep front edge of the rise by means of a control system. This initial

front, edge is also dependent upon ft/I*. However, if slower control can

48
AUTOMATIC REACTOR CONTROL

49

be tolerated, one can design a control system that ignores this front edge

and operates on the flatter portion of the open-loop response which occurs

a second or two following the front edge.

In a thermal reactor, however, the break Si occurs within the frequency

range of many conventional servo systems. It appears from this simple

viewpoint that a thermal reactor in which the break point is less than

10 cycles/sec can be controlled, if desired, on prompt neutrons. Stabil-

ity problems exist, and they will be discussed later in this chapter.

4-2. Reactor Representation with Temperature Coefficient and

Poisoning Feedback Loops. The representation we have been using for

a reactor transfer function is a somewhat na'ive one in that it ignores

some of the fundamental processes that must go on within a reactor

and a plant. These processes can be considered in the form of feedback

loops which modify the elementary reactor transfer function. When

feedback of any type is present, the problem of stability exists.

There are two distinct types of feedback loops which must be con-

sidered. These are external loops and internal loops. The external and

internal feedback loops can be considered to

behave independently and are represented block-

wise in Fig. 4-1. The external feedback loop is

a loop which is associated with an external reac-

tor plant. It is related to the circulation of the

primary coolant through the reactor. In a nu-


Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

clear power plant an external coolant is circu-

lated through a reactor, some heat is removed

from this coolant, and the same coolant at a

new temperature is reinjected into the reactor.

A discrete amount of time is involved in the

process of circulating the coolant from the reac-

tor through the heat exchanger and back into

the reactor. A given amount of attenuation is

also involved in this path, and the combination

or^this gain and phase shift in conjunction with

the transfer function of the reactor determines

"We~shall

FIG. 4-1. Block diagram

of nuclear reactor show-

ing normal feedback paths.

the stabiljty"of a nuclear power plant.

examine this external feedback loop in detail in

Chap. 6 and for the present shall confine ourselves to two internal or

local feedback loops of the reactor.

The internal loops are caused by temperature coefficient and by poison-

ing. The internal loops appear as indicated in Fig. 4-2, and each effect

can be assigned a transfer function. We shall call the local temperature

feedback transfer function KTcGTc(s) and the poisoning feedback trans-

fer function KxGx(s).


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

>

1 —

LOCAL

TEMPERATURE

COEFFICIENT

FEEDBACK

KTCGTC(s)

POISONING

FEEDBACK

KXGX(S)

REACTOR

KRGR(S)

FIG. 4-2. Reactor internal feed-

back paths.

As mentioned in Sec. 2-13, the negative temperature coefficient oper-

ates primarily because of a decrease in density of the moderator and

reflector as the temperature within a reactor is raised. Other items such

as changes of cross section and changes in

leakage are involved within the internal

geometry of the reactor. It is difficult to

consider the reactor as a single lumped net-

work"and properly take care of all the

detailed spatial changes that occur inside a~


Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

re^1roT~a"s~a'~ruiictipn of temperature. In

?>fder to handle the problem completely it

is necessary to treat the reactor as a dis-

tributed network and determine the effects

of temperature upon reactivity in each por-

tion of the network. This method has been

developed in the classified literature but is

too complex to be useful here. Instead, we

shall continue the simplified black-box analy-

sis and propose a simple point-type mecha-

nism for a temperature feedback transfer

function. The reactor transfer function, as

it is modified by the local temperature

coefficient feedback loop, will then be examined.

Poisoning acts in a similar manner as indicated in Sec. 2-14. Xenon

135 atoms are created both directly and indirectly in the fission process,

and these atoms change the reactivity of the reactor in a complex manner

with time. The effect of the local temperature coefficient feedback loop

upon the reactor transfer function will be developed first. Then the

effect of the poisoning feedback loop upon the performance of this com-

bination will be shown.

4-3. Negative Temperature Coefficient Feedback. From a servo

point of view we can combine parallel transfer functions and examine

the over-all stability of the combination. First, considering the tem-

perature coefficient situation, the elementary transfer function we have

been using for the reactor indicates that the reactor by itself is unstable.

That is, at zero frequency the reactor has infinite jjam. Physically, when

a small am.ount~oT positive reaciiyity. is ^inserted into the j^actor, its

"power leVeTrTses indefinitely. It can be felt instinctively that the neg-

ative "temperature Coefficient effect "is a stabilizing one in that, as the

reactor power rises,~tEe temperature rises, the reactivity is reduced,_and

conceivably thlTpower-level rise is jialted at some point. To examine

this~process weTshTrll consider tKeblock diagram of Fig. 4-3 and shall use

the well-known relationship for parallel transfer functions.1-2 This rela--

50
AUTOMATIC REACTOR CONTROL

tionship says that

KBGR(s)KTCGTc(s)

51

(4-1)

where KRTcGRTc(s) is the new combined over-all transfer function and

KTcGTc(s) is the transfer function of the local temperature coefficient

effect. It is now necessary to examine the form of KTcGTc(s). The

process to be used is as follows: It is first assumed that the reactor is

generating enough power so that it can change the temperature of the

LOCAL TEMPERATURE

COEFFICIENT EFFECT

KRTCGRTC(S)

COMBINED REACTOR

AND

LOCAL TEMPERATURE

COEFFICIENT EFFECT

TRANSFER FUNCTION

(a) (b)

FIG. 4-3. Combination of reactor transfer function with local temperature coefficient feed-

back transfer function, (a) Individual transfer functions, (b) Combination transfer

function.

moderator in a short space of time. No plant is attached to the reactor

to remove this power. However, if it is desirable to suppose that a

reactor coolant is also present in the reactor, this coolant would be fixed
Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

and not circulating.

Then small variations in neutron level are presumed to occur in a

sinusoidal manner. The neutron-level change affects the reactor fuel

temperature level. The heat from the fuel then causes a change in the

moderator-temperature level and this temperature change in turn may

be represented as reactivity change. If, over a small range, these effects

are assumed to be linear, an elementary form of transfer function may

then be derived. Let

Tm = ASn (4-2)

where Tm = fuel temperature change

Sn = neutron-level change

A = a constant depending upon power level, moderator, and

coolant.

We can now relate the fuel temperature to the moderator temperature

Tw, in the form of a simple exponential lag. This approximation will

describe the elementary heat flow from the fuel to the moderator. In
52 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Laplace notation

r.(«) = ^TT (4-3)

To ~T~ i

where r is the time constant of the exponential heat transfer lag.12

Sk(s), however, is proportional to Tw via the temperature coefficient.

That is,

Sk(s) = (TC)Tw(s) (4-4)

where (TC) is the temperature coefficient. The transfer function

KTcGrc(s) then becomes

The gain term of the transfer function may be lumped into a new gain

constant KTC and the transfer function becomes

We now perform the suggested combination of the transfer functions

given in Eq. (4-1) and

/(I + rs)] (4-7)

The temperature coefficient transfer function may be plotted for given

values of r and KTc, and the combination of this transfer function with the

reactor transfer function can be accomplished graphically quite easily by

use of Nichols charts3 or the equivalent. Such a combination is shown in

Figs. 4-4 and 4-5 for a reactor having an Z* of 10~4 sec, T = 0.159 sec, and

several values of the temperature coefficient gain factor KTC- The

amplitude curves of Fig. 4-4 indicate that at very low frequencies the

characteristic of the over-all combined transfer function is determined

solely by the temperature coefficient effect feedback. The gain is no

longer infinite at zero frequency, but a finite value of gain equal to 1 /KTC
Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

results from the combination. At very high frequencies the combined

transfer function takes on the shape of the reactor curve which depends

principally on I*. We can generalize to the extent that if the break fre-

quency caused by the time delay r is low compared with the highest

natural break frequency of the reactor set by ft /I*, the high-frequency

response of KRTcGsTc(s) is not affected by the temperature coefficient.

Figure 4-5 indicates that at low frequencies the phase shift approaches

zero for reactors with negative temperature coefficients, instead of — 90°

as in the case of the elementary reactor alone. At high frequencies,


AUTOMATIC REACTOR CONTROL

53

30

26

0.001

0.01

100

0.1 1

FREQUENCY, CYCLES PER SECOND

FIG. 4-4. Amplitude response of combination reactor and temperature coefficient feed-

back transfer functions for several temperature coefficient gains.

150

-100

0.001

0.01

100

0.1 1

FREQUENCY, CYCLES PER SECOND

FIG. 4-5. Phase shift response of combination reactor and temperature coefficient feed-

back transfer functions for several temperature coefficient gains.

again, the phase shift of the combination approaches the phase shift of

the reactor. Interestingly enough, some phase lead may result from the

combination, and the position and magnitude of this lead depend upon T.

It will be recalled that the reactor gain is a function of level and con-

sequently the decibel scale for the amplitude in Fig. 4-4 is relative to some
Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

power level. The gain of the temperature feedback loop actually is

proportional to power level through the constant A. As power level is

reduced, the feedback effect gradually vanishes. A reactor that is nat-

urally stable at high power levels is more difficult to control at low levels.

For this reason one might limit the range of automatic control to 1 or 2

decades of power level. For automatic control, then, the reactor should

not be shut down below 1 percent of full power. For analytical purposes

KRGR(S] may be defined in terms of (Sn/n)/Sk. In this form it is inde-

pendent of level.

From the example of Figs. 4-4 and 4-5, it is evident that this simple

representation of the combination of the two transfer functions is quite

stable. Other methods can be used for determining the stability of more

complex reactor negative temperature coefficient feedback systems.

Weinberg and Ergen16 have presented a method for homogeneous reactors,

and Lipkin16 has extended their method to treat heterogeneous reactors.

However, the servo engineer will prefer to obtain equivalent results by

refining the above analysis and adding more terms to the feedback transfer

function. These more complex system representations have a real possi-

bility of being unstable, but fortunately

in practical reactor design and opera-

, tion the oscillatory condition is not a

common one.

4-4. Poisoning Feedback. Having

established the method of determining


Generated on 2011-10-24 11:41 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the stability of a reactor combined with

a local negative temperature coefficient

effect, we can continue by a similar

process and examine the poison feed-

back loop. As all reactors have a tem-

perature coefficient of some sort, we can

start with the representation of the

reactor transfer function as KRTC&RTC(S) rather than KRGB(s'), and the

poisoning feedback loop can then be tied around this combined system.

Figure 4-6 shows the block diagram of the network that will be analyzed.

It is now necessary to develop the form for KxGx(s).

Repeating Eqs. (2-24) and (2-25) for convenience we have

COMBINED

REACTOR AND

TEMPERATURE COEFFICIENT

TRANSFER FUNCTION

FIG. 4-6. Elementary block diagram

poisoning feedback.

dX'

dt

dt

= \,r +

= -\,r

- \xxr

(4-8)

(4-9)

54

In order to keep the solutions of these equations general, we now can

divide Eqs. (4-8) and (4-9) by Z/ since the value of 2/ is not the same for
AUTOMATIC REACTOR CONTROL 55

all reactors. Thus, in generalized form the basic poisoning equations

become

^ = X,7 + (7x - vxX)<t, - \*X (4-10)

^ = - X,/ + 7/0 (4-11)

Using a similar technique to the method of deriving the reactor transfer

function (Sec. 3-4), let us divide the variables into steady-state parts and

incremental variations about this steady state such that

X = X0 + SX <t> = 0o + 50 7 = 70 + 57 (4-12)

then Eq. (4-10) becomes

jt (X0 + SX) = X,/0 + (7, - «r^0)*o - XxXo + \,SI

+ (7x - <rxA'0)50 - (<rx00 + \,)SX (4-13)

where (505-XVx) has been neglected. In the steady state

^? = 0 = X,/0 + (y* - «r^T0)*o - \xX0 (4-14)

^ = \,Sl + (y, - axX0)d<t> - (ff^o + \,)SX (4-15)

Similarly, Eq. (4-11) becomes

jt (I0 + 57) = -X,/0 + 7/0o - X/57 + 7/S<*> (4-16)

In the steady state

*L° = 0 = - X//o + 7/0o (4-17)

and = -X/S7 + y,&4> (4-18)

Transforming Eqs. (4-15) and (4-18) to Laplace notation

sSX(s) = X,57(s) + (yx - <rxXo)8<t.(s) - («r*00 + \I)BX(s) (4-19)

s57(s) = -X/57(s) + y,8<t>(s) (4-20)

«/(.) = (4-21)

o -p A;
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Substituting Eq. (4-21) into Eq. (4-19)

s8X(s) = X/7^.(S) + (7x - <r,X0)50(S) - («r,*o + Xx)5X(s) (4-22)


56 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

SX(s) _ X,7/ + (s -I-X/)(7x — <rxX0} .

-,-, \ = , I ^ \/ I T—_|_ \ 1\9ffO)

i \.

(yx — ffxXo) ( S -| -i - y- + X; J

= - , , . ;, . 7" °^ .\ '- (4-24)

\S ~T~ \i)\S -p ffx<po ~T~ nx)

Equation (4-24), then, is the desired small signal transfer function. It

can be seen from this equation that 5A"(s)/50(s) is a peculiar transfer

function with its amplitude and phase shift very dependent on A~o which,

in turn, depends on 0o. The symbols 0 and n are being used inter-

changeably in this section. The physicist generally uses 0; the servo

engineer uses n.

We may examine the transfer function of Eq. (4-24) numerically by

setting in the following constants from Stephenson:4

X, = 2.9 X 10~6

\x = 2.1 X 10-6

y, = 0.056

yx = 0.003

ffx = 3.5 X 10-1s

First, the relationship between Xo and 00 is determined from Eq. (2-26)

in generalized form

v (7x -

r—-j- — .-

AI T 0x90
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Fig. 4-7a shows this relationship of the xenon concentration as a func-

tion of neutron flux level. It can be seen that the poisoning rises linearly

with flux until a flux level of approximately 1012 is reached. At higher

flux levels the xenon concentration rises more slowly until at flux levels

of approximately 1014 and higher there is no further increase in poison

concentration.

From a transfer function point of view there are two flux levels which

should be examined carefully. From Eq. (4-24) it can be seen that when

(yx — ffxX0) = 0, corresponding to a flux level of 00 = 3 X 10", the

phase of the transfer function shifts its ultimate end point with frequency.

That is, at high frequencies when 0o < 3 X 1011, there is a total phase

shift in the xenon feedback path of —90°. When 0o > 3 X 1011, there

is an ultimate —270° phase shift in the transfer function. This condition

lasts until the point where \iyi/(yx — <rxX0) + X/ = 0. The transfer

function gain at zero frequency when yx — <rxXo — 0 can easily be deter-

mined to be yi/(o-x<l>o + Xi), and there is no discontinuity in gain as a

function of 0o-
AUTOMATIC REACTOR CONTROL 57

It would appear that another point of interest might be when

.-r + X, - 0

as here the sign in the parentheses in the numerator of Eq. (4-24) would

again be positive. This condition stipulates that

7/ + 7x

(4-26)

But from Eq. (4-24) this cannot happen without infinite flux. Therefore,

at high flux levels above <j>0 = 3 X 1011, \ifi/(yx — <rxX0) + X; will always

1014

101

10s

108

10" I05 106

10T

101'

1015 1016

10s 109 10'° 10" 1012 1013

NEUTRON FLUX - CM~2SECT1

FIG. 4-7a. Equilibrium xenon concentration as a function of flux level.

be negative. Let us now examine numerically the transfer function at

flux levels above and below 3 X 1011.

Figures 4-7 and 4-8 show the relative response of the xenon concen-

tration to small oscillations in flux normalized in gain about the <p0 = 1014

case. The absolute level for this curve is +19.1 db. It will be noted
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

from the phase curves that for <£0 below 3 X 1011, the total phase lag at

high frequencies is 90° whereas at higher flux levels the total phase lag is

270°, as previously indicated.

The gain at zero frequency is a constant for fluxes roughly below

0o = 1010. At higher levels the gain steadily decreases with flux. This

fact could have been presumed from the saturation effect of Fig. 4-7a.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

-40

10'

FREQUENCY, CYCLES PER SECOND

FIG. 4-7. Amplitude of transfer function of xenon feedback factor SX(s}/8<j>(s).

-50

-100

-150

-200

-250

-300

FREQUENCY, CYCLES PER SECOND

FIG. 4-8. Phase shift of transfer function of xenon feedback factor 5X(s) S<p(s).

58

Stability with Xenon Feedback. It has been shown that any elementary

reactor containing a negative temperature coefficient of reactivity is

stable no matter how small the negative temperature coefficient. A con-

sideration of the xenon feedback loop now will require modification of

this statement. Let us tie a xenon feedback loop around a reactor con-
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
AUTOMATIC REACTOR CONTROL 59

taining a negative temperature coefficient. The block diagram will be

that of Fig. 4-9a. From Fig. 4-8 the xenon transfer function shows

180° phase lag in the neighborhood of 10~6 cycle/sec, and it is this fre-

quency range which is of interest. For frequencies below 10~8 cycle/sec

the reactor transfer function with negative temperature coefficient

(b)

(a)

KTC

(C)

FREQUENCY IN CPS

FIG. 4-9. Elementary block diagram of xenon feedback loop stability considerations,

(a) Individual transfer functions, (b) Combined transfer functions, (c) Approximation

to the response of K.RTcGRTcd) for frequencies below 10~3 cps.

KBTCGRTC(S) becomes a straight line having a gain of 1/KTc and a phase

shift of 0°. This condition comes about because the time constant of the

negative temperature coefficient is presumed to be fast (see Figs. 4-4 and

4-5). This situation is shown in Fig. 4-96 and c, and for this analysis we

shall assume that none of the frequencies above 10~3 cycles is important.

The transfer functions for the xenon concentration shown in Figs. 4-7

and 4-8 represent only &X"(s)/50(s). It is now necessary to multiply

this transfer function by a gain term which will, in general, be a constant.


Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
AUTOMATIC REACTOR CONTROL 61

back loops, we are now in a position to attempt to control a reactor with

an external control loop. We shall first examine broad operational

requirements. To change the power level in a reactor requires the move-

ment of a control rod or rods in such a manner that the multiplication

factor c)F the reactor is first caused to deviate from unity. The rods are

la'ter lnoved to return the multiplication factor to unity. In changing

power level two conditions must ultimately be satisfied by any control

system: The power output level of the reactor must be the demanded

power level within a given error, and the multiplication factor must

ultimately be 1.

A given reactor may contain many control rods. These control rods

may be operated individually or in groups or banks. Large groups of

rods are moved quite slowly and usually function in control work as a

coarse shimming system. These ganged rods are therefore sometimes

called shim rods and may be worth several percent in reactivity. Because

of the obvious hazard involved in making large reactivity changes quickly.

the purpose of shim rods is usually to take care of slow changes such as

depletion jmd^jjoisonirig.Shim-rod motion is therefore quite slow, and

rates of change of reactivity between 10~6Sfc/sec and 10~35fc/sec are in

"common usage.

To take care of the routine power changes required in a reactor plant

and to handle intermittent transients which may occur, one rod may be

designated as a regulator rod. This regulator rod can be positioned

automatically. Regulator rods may move quite quickly, but to ensure


Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

safety, the maximum worth of a regulator rod is usually kept below 0.006

in reactivity. In this manner the regulator rod can never inject as much

as prompt critical reactivity into the reactor. In the event that more

reactivity than 0.006 is needed for a particular operation, the shim rods

may be connected to follow up the regulator rod position. That is,

after the regulator rod has run out of reactivity, the shim rods can be

called upon to move in the same direction as the regulator rod and thus

slowly provide the necessary reactivity.

We have seen from Sec. 2-11 that manual control of a reactor is apt to

be a tedious process requiring great alertness and training on the part of

the operator. In changing power level, skill and anticipation will be

required to prevent power-level overshoot. Poisoning and temperature

fluctuations create the need for periodic control-rod manipulation. It

is somewhat obvious that an automatic-control system can do a better

job than a human being.

The above discussion presupposes the need for an external reactor

control system. Some reactors having large negative temperature coeffi-

cients may _not_jrequire an external regulating system. The" criteria

involved are presented later in this chapter.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

An automatic-control system usually consists of a control loop around

the regulator rod or a shim-rod group and can be operated as either a

proportional regulating system or a discontinuous regulating system. A

proportional regulating system is one in which the position of the control

rod or rods is changed in proportion to and in phase opposition with any

error created either by a powerdemand change or byan internalsystem

transient. A~3iscontlhuous regulating system is one in which no control

is attained unless an error, which is some fixed percentage away from a

demanded set of conditions, is set up in the control loop. When suffi-

cient deviation occurs from the demanded conditions, the control-rod

position is usually changed at a fixed velocity. Discontinuous regulating

systems can be designed to hold reactor power level to within 0.5 percent.

Better accuracy can be obtained with a proportional type of control.

A discontinuous system offers the advantage that it is less sensitive to

noise, which may originate in detecting elements. In high-fluxj-eac_tar£L_

hflWfWPr, nftJSft is limially not a aoriniia prnhlpm^ nnrl fnr mn.ny piirprvaog

the jncreased accuracy warrants the use of the proportional_sy_stem.

4-6. General Description of a Reactor Automatic-control System.

Figure 4-11 shows a broad block diagram of the type of control system

POWER

DEMAND

COMPARATOR

ERROR

SIGNAL
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

AMPLIFIER

ACTUATOR

NEUTRON

DETECTOR

REACTOR

CONTROL

RODS

FIG. 4-11. Block diagram of the essential elements of a reactor automatic-control loop.

generally used to control a reactor in the power-level range. The reactor

multiplication is changed by movement of the control rods. For this

discussion we may first consider that we are using a regulator rod in a

proportional system. The output of the reactor, in terms of power out-

put or neutron level, is measured by a neutron detector, usually an ioniza-

tion chamber. This reactor output is then compared with the desired

power demand in the comparator, and any error between the actual out-

put and the demanded output is noted and amplified in the error-signal

amplifier. This amplifier finally controls an actuator which moves the

rods the proper amount and in the proper direction to eliminate the

error. The comparator, error-signal amplifier, and actuator may be of

any suitable type. Pneumatic, hydraulic, electrical, and mechanical

62
63

AUTOMATIC REACTOR CONTROL

devices have all been used, and some will be described in Chap. 5. A

brief description now follows of each of the components in the loop from

a control-system point of view.

Reactor. The principal feature of the reactor to be noted in this con-

trol loop is its nonlinearity. As mentioned in Sec. 3-4, the reactor gain

is proportional to the level at which the reactor is operating. This is a

completely intolerable situation, as most control loops can operate in a

stable manner only over a limited range in gain. Means to eliminate

gain dependence upon level must be added to the control loop.

Comparator. The function of the comparator is twofold. First, it

provides that the error signal is essentially the subtraction between the

output of the neutron-detector sig-

nal and a power demand signal.

Second, it is also used to compensate

for the embarrassing nonlinearity of

the reactor. The form of the out-

put signal from the comparator

should be error/level.

A simple comparator circuit is

shown in Fig. 4-12. Here Vn is the

signal from the neutron detector pro-

portional to the power output of the

reactor, Ve is the output error sig-


Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

nal, and the power demand reference voltage V0 is denned arbitrarily as

7o = Eb/K. Then

FIG. 4-12. Elementary

parator circuit.

form of com-

Ve = KVn -

- - Et

Letting an error signal Sn = n — no

V F (n

V e = .C/b I

n0

Sn

n0

(4-27)

(4-28)

(4-29)

(4-30)

which is of the proper form so that the output of the comparator is

inversely proportional to the neutron-demand level. Hence, the effect

on the control loop of the reactor gain dependence on level is essentially

canceled when the comparator is connected to the reactor. Actually, a

signal inversely proportional to n rather than n0 would be more desirable.

That is, if the reactor power level changes, for complete cancellation of

the nonlinearity during the transient, the comparator should divide by


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the actual level existing at any moment in the reactor. The difference

between the actual level n and the demanded steady-state level no is

usually quite small, as the control

system rapidly acts to make the two

quantities the same. In practice it

is much easier to obtain the term no

than n for use in a practical compa-

rator circuit. It will be noted that

the symbol no is used here to denote

demand level in the steady-state con-

dition. This symbol has also been

used in Chaps. 2 and 3 to denote the

actual operating level of the reactor.

Comparators having perfect can-

cellation over an infinite range can-

not be constructed practically, but

good cancellation can take place

over a limited range of about 100 to

1. Figure 4-13 shows a usable com-

parator circuit in which the demand

signal is linearly proportional to the displacement D of the potentiometer

P. When no current is taken from V,

FIG. 4-13. Elementary form of com-

parator circuit in which the demand signal


Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

is directly proportional to the potentiom-

eter arm rotation.

ve =

R + DP

Vn - Eb

(4-31)

where 0 < D < 1.

Assume that when

Vn = V0 V< = 0 (4-32)

where Fo is the demand voltage proportional to no. At Ve = 0

Eb =

R + DP °

Fo = Eb l + —

(4-33)

From these expressions Fo is linear with D. The error signal then

becomes

V, =

but

R + DP

Eb

n ~ Fo)

R + DP Fo

(4-34)

(4-35)

(4-36)

64

which again is of the desired form.


AUTOMATIC REACTOR CONTROL 65

To ensure that no current is taken from Ve, computing techniques may

be used and an operational amplifier connected, as indicated in Fig. 4-14.

Here R' and R", in parallel, have the same resistance as the original R

of Fig. 4-13. The bias voltage of the amplifier also becomes the equiv-

alent of the battery Eb.

Magnetic-amplifier Comparator. For many applications it is undesir-

able to use vacuum tubes in a comparator circuit and, consequently,

1'HH JL

FIG. 4-14. Nonloading comparator circuit.

FEEDBACK

MAGNETIC

AMPLIFIER

FIG. 4-15. Magnetic-amplifier comparator.

magnetic amplifiers have been used. In a magnetic amplifier an output

current proportional to the difference between two control signals is

generally obtained by using the control windings in opposition. Thus,

one control signal tends to saturate the core while the other tends to

unsaturate it. In this manner good linearity can also be obtained.

Figure 4-15 indicates a schematic block diagram of a magnetic-amplifier

comparator system. The output current IL of this sort of device can

be of the form IL = A(n — n0) where A is the effective gain of the mag-

netic amplifier and n and no are input currents proportional to the neutron
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
66 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

level and the neutron-demand level. Magnetic amplifiers are usually

operated using feedback, and feedback windings can easily be provided.

Negative feedback may be used in the amplifier as well as the more

customary positive feedback, and the familiar expression for the gain of

an amplifier having negative feedback may be stated as

_ A(n- no) . .

where B is the feedback factor. If AB is now made large compared

with 1,

/. = -° (4-38)

It now becomes necessary only to make B proportional to n or no. Again,

in practical circuits, it is easier to obtain n0 than n, and a shaft rotation

might be used to generate n0 and also to connect a potentiometer into the

feedback circuit. Therefore, the final output load current becomes

= n-n0

n0

Error-signal Amplifier. The error-signal amplifier may be a conven-

tional vacuum tube or magnetic amplifier. It usually amplifies the

error-signal level from a few milliwatts to a few watts in order to control

the actuating device. If an attempt is made to control the actuator

motor directly, such as would be done if the control motor were a two-

phase a-c servomotor, a few kilowatts might even be required from the

amplifier. The frequency response of the amplifier usually presents no

problem in comparison with that of the rest of the system. The amplifier
Generated on 2011-10-24 11:42 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

may therefore be regarded as pure gain in a servo concept.

Rod Drive Mechanism and Actuator. As mentioned previously, almost

any physical form of actuator and mechanism may be used. The actua-

tor mechanism usually is called upon to perform a dual function. That

is, in the case of shutdown of a nuclear reactor, a rapid action may be

required, whereas for startup and power-level control, only a compara-

tively slow motion is needed. The actuator frequency response may

range from a few cycles per minute to a few cycles per second. The

actuator usually contains a power amplifier, and the output from the

error-signal amplifier may be used to supply such conventional devices

as hydraulic spool valves or torque motors. In some instances, if a d-c

motor is the output actuating prime mover, control may be obtained by

feeding the field voltage on an amplidyne or a Ward Leonard system

motor generator set. The actuator is then coupled to a control rod

whose weight, in present-day reactors, may be up to several hundred


AUTOMATIC REACTOR CONTROL

VOLTAGE

FIG. 4-16. Characteristic saturation curves

of an ionization chamber.

pounds. A detailed discussion of actuator requirements is given in

Chap. 5.

Neutron Detectors. The neutron detector in a control system is usually

an ionization chamber. This device will be treated more thoroughly in

Chap. 7. For present purposes it may be stated that the ionization

chamber is a device which gives out a current proportional to the number

of neutrons per second which enter it. No frequency-response problems

in conventional control circuits exist, as the frequency response of the

chamber is not affected by opera-

tions up to a few hundred cycles

per second. For analysis purposes

the ionization chamber may be con-

sidered to behave as a current

source similar to a pentode, as in-

dicated in Fig. 4-16. The amount

of current which is put out at satu-

ration is a function of the amount

of radiation present. The param-

eter which corresponds to grid

voltage in a vacuum tube is radia-

tion level in an ionization chamber.


Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

4-7. Control-loop Response. Having examined the basic character-

istics of the components of a reactor control loop, we may fittingly ask

the question: What do we expect from this loop? The first requirement

is that the loop be unconditionally stable. That is, under no condition

of conceivable practical reactor operation are sustained oscillations to be

permitted. The second requirement is that the loop shall respond sat-

isfactorily under transient conditions. The transients to be considered

are generally of two types: a change in the power demand level or a change

in the reactivity of the reactor.

Control-loop Stability. The absolute stability of a reactor control sys-

tem can be determined by servo synthesis techniques once the transfer

functions of all the components in a loop are known. Elementary stabil-

ity considerations can be accomplished by studying simplified servo sys-

tems involving the reactor. Both proportional and on-off type systems

will be examined. We shall first consider proportional systems.

In an example we can combine the transfer functions of the error-

signal amplifier, the actuator, and the control-rod mechanism into a

single transfer function KBGB(s), as indicated in Fig. 4-17. The reactor

transfer function KRGR(S) may be considered as given in Eq. 3-23 with-

out any temperature coefficient feedback. If a temperature coefficient

were present, KRTCGRTC(S) would be used instead of KRGR(s). We can

now examine the over-all loop response as a function of some of the char-
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

acteristics of KaOB(s), as it is quite obvious that the stability of the loop

shown in Fig. 4-17 depends upon the characteristic's of ATB(7/)(s). Let us

first consider that we can represent JCB(7B(s) by an elementary quadratic

FIG. 4-17. Block diagram elementary re-

actor control loop, with lumped external

transfer functions.

FIG. 4-18. Elementary reactor control-loop

transfer functions for simplified control

system.

0.001

2000

0.01 0.1 1 10

FREQUENCY, CYCLES PER SECOND

FIG. 4-19. Normalized transfer functions of external reactor control circuit for three values

of external system natural frequency.

form of transfer function. Physically this might correspond to having

a motor driving a control rod.

GB(s) =

(4-40)

68

would be the form of the frequency variant portion of the transfer func-

tion. We shall examine the insertion of various amounts of gain later.

It is then desired to determine the stability of a control loop containing

this form of control transfer function over a range of natural frequencies


Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

/« = wn/2?r. For simplification let us assume that the damping coeffi-

cient f is fixed at f = 0.5 and that we are using a reactor having the trans-
AUTOMATIC REACTOR CONTROL

69

-1,0

fer function given for an I* of 10~4 sec. Our simple block diagram

becomes as shown in Fig. 4-18. The normalized transfer function of

KBGB(s) is shown in Fig. 4-19 for three values of /n. These values of

/n are approximately 0.1 cycle/sec, 1 cycle/sec, and 10 cycles/sec. For

stability purposes we may now examine the open-loop gain KBGR(s)

KBGB(s). The indicated multiplication can be accomplished graphically,

and the Nyquist plots14 giving the loci of the open-loop gain points may

be examined in Fig. 4-20. It is

obvious from the form of the reac-

tor transfer function that at zero

frequency the gain is infinite at an

angle of —90°. As the possibility

also exists th^,t a total of 270° phase

shift can be attained by the open-

loop combination, oscillations can

occur in this loop as a function of

gain. The Nyquist plot of Fig.

4-20 for an arbitrary fixed gain

shows that as the natural frequency

is made higher, the possibility of

oscillation becomes better, and, in-

deed, for the case of /n = 10 cycles,


Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the (—1,0) point is encircled and

oscillations would occur. In other-

words, in the design of an elemen-

tary reactor control system of this

type, the placement of the natural frequency of the control system

is important if some arbitrary gain must be furnished for transient

performance.

Phase shift compensation, of course, can be provided to permit higher

gain, and most reactor control systems contain compensating networks.

The common types of network that have been used in reactor control

systems are given in Fig. 4-21. It must be emphasized again that this

analysis was for a reactor with zero temperature coefficient. The method

of analysis would be the same for a reactor with a negative temperature

coefficient.

fn=1.0 CYCLES

PER SECOND

fn=0.1 CYCLES

PER SECOND

FIG. 4-20. Nyquist plots of elementary re-

actor control system for arbitrary gain.

4-8. On-Off-type Reactor Control System. Until now we have been

discussing proportional types of control systems. The more common

system is the simple on-off type of control. Here, when the difference

in signal between n and n0 reaches a preset amount, a motor is started

and caused to run at constant velocity until the error signal is reduced to

another predetermined value. This discontinuous-type system has a

dead zone in its control and is inherently incapable of the same accuracy
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

as the proportional system. Other advantages exist, however. The

dead-zone principle succeeds in making control-rod positioning simple in

that mechanical brakes, gearing, and other forms of friction may be used

to keep a control rod in a given position within the dead zone. In the

(c)

FIG. 4-21. Types of compensation networks commonly used in reactor automatic control

loops, (a) Tachometer feedback, (b) Position feedback, (c) Derivative network.

70

proportional system a control rod is held in place through the torque

developed by the control-drive motor and thus the complete system power

must be available to keep the proportional system control rod in a fixed

position. As previously mentioned, another advantage of the on-off-

type system is that it is less susceptible to noise because of its dead zone.

The on-off-type control system is at present quite common for large

experimental reactors, and the control system for the CP-3 reactor has
Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
AUTOMATIC REACTOR CONTROL

71

been described by Harrer.6 The method of analysis to be presented here

is a general one developed by Grace,6 and a fictitious reactor system will

be presented as an illustration of the method.

The stability of an on-off system can be analyzed in the following

manner: It can be felt intuitively that in order to use the on-off-type

reactor control system, the reactor should have a negative temperature

coefficient. If the reactor system by itself is inherently unstable, the

presence of the dead zone in the on-off system may well cause the control

system to be in continuous oscillation. Although it is quite possible to

keep the reactor under control with a continuously oscillating system,

prudence dictates that the wear on the control-system components should

be minimized. Consequently, we shall assume that the reactors under

discussion in this section have negative temperature coefficients so that

we can use KRTcGRTc(s) as the transfer function. After the control-

system analysis is developed, we shall examine the effect of the magnitude

of the temperature coefficient.

The reactors using discontinuous-type control systems, such as the

CP-3, are reactors that in general merely maintain a fixed power level.

These reactors usually have large values of I*, and reactivity is changed

only at very slow rates. The design of all auxiliary systems about the

reactor is such that nothing is capable of moving fast and consequently

no fast transients can occur. Therefore, in the study of the design of the

control system for such reactors, the important factor becomes stability
Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

rather than fast transient response.

PHASE

~j COMPENSATION [

ERROR

SIGNAL

AMPLIFIER

CONTACTOR

AMPLIFIER

REACTOR

CONTROL ROD

FIG. 4-22. Block diagram of discontinuous-type servo reactor control loop.

Figure 4-22 is a block diagram of the type of system under considera-

tion. The reactor of this block diagram may be the type of reactor

described in Sec. 4-3. The reactor and its temperature coefficient will

be treated as having a combined transfer function KRTCGRTC(S), as also

given in Sec. 4-3. A compensating network is shown in the loop to pro-

vide phase shift compensation. The analysis will first be shown without

this compensating network, and the network later added to increase the
72 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

dynamic stability of the system. The contactor amplifier indicated in

the block diagram may be considered as a simple relay or relays, which

close a set of contacts when the error-signal level reaches a fixed amount,

then open these contacts when the error-signal level drops below another

fixed value. A corresponding set of contacts is opened and closed in a

similar manner when the sign of the error signal reverses. These con-

tacts cause a motor to rotate in one direction or the other and to be

stopped when the signal level is too small to keep either set of contacts

closed. The drive motor will be considered as running at constant veloc-

ity after an initial time lag in getting started. A control rod is assumed

to be geared to the motor, and the control-rod motion will therefore be

characterized by being able to change reactivity at a fixed rate in Sk/sec.

Analysis Procedure. The type of system presented in Fig. 4-22 is a

nonlinear system, the nonlinearity being caused by the contactor. This

nonlinearity may be expressed simply in that the output of the contactor

is not a linear function of its input but rather a discontinuous function.

The gain of a closed loop consisting of linear components is a function

only of frequency. Any nonlinearity in a system, such as the on-off

feature in this reactor control loop, causes the loop gain to be a function

of both frequency and amplitude. When the amplitude-dependent

transfer functions in the system can be separated from those which

depend upon frequency, the loop gain can be expressed as a product.

This product is loop gain = BKAG, where A is a complex function of

amplitude independent of frequency, G is a complex function of frequency


Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

independent of amplitude, K is a constant gain factor, and B is the feed-

back factor which equals — 1 in this reactor control loop. The over-all

transfer function of a system with B = — 1 is the familiar expression

Output = KAG KA

Input "" 1 + KAG ~ G~l + KA ( '

Expressing the over-all transfer function in the form given in Eq. (4-41)

indicates that the stability of the system can be obtained by comparing

G~l and —KA. The conventional method of performing this comparison

is from a polar plot of G~l and — KA. The value of G~l can be plotted

for all values of frequency, and the value of — K A can be plotted for all

values of amplitude. If the two loci intersect, that is, if G~l = — KA, the

system is capable of sustaining an oscillation. The frequency of the

oscillation would be the frequency corresponding to the point of inter-

section on the G~1 locus, and the amplitude can be found from the — KA

locus.

In the analysis of the block diagram of Fig. 4-22, the loop gain of the

reactor control loop can be divided into an over-all frequency function G

and an amplitude varying function A in the manner just described. The


AUTOMATIC REACTOR CONTROL 73

function G is the product of the transfer functions of the reactor, the error-

signal amplifier, and the drive motor. The function A comes from the

contactor amplifier alone and describes the effects of the relays. The

method of determining the transfer function of the relays will be based

on the development of Kochenburger7 and essentially depends upon the

assumption that only the fundamental component of the square wave

signal of rod velocity coming out of the relay is significant. Higher

harmonics are more attenuated by the system, particularly by the motor,

and consequently may be ignored.

-b „ -a

it

1 no~n

ERROR SIGNAL

W.

a.

FIG. 4-23. Relay program.

Figure 4-23 illustrates the relay cycle and the terminology that will

be used. The error signal into the relays is proportional to (n0 — n)/n0.

As the error signal increases, it reaches the point b, which closes the relay
Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

starting the drive motor and creating the reactivity rate demand signal

V. Once the control rods are started moving in the direction to reduce

the error signal, the hysteresis of the relay causes the contacts in the relay

to remain closed until the point a is reached and the drive motor is turned

off. It is assumed that the contactor amplifier is symmetrical for nega-

tive signals, with — b being the negative start limit and —a being the

negative stop limit. Let us now assume that the output of the error-

signal amplifier is x = \x\ sin otf. Figure 4-24 indicates the phase rela-

tionships between the input signal x and the fundamental component v

of the output signal. The terms which are of consequence in the analysis

are the ratios b/a, V/b, and \x\/b. From this figure it can be seen that

if b/a = 1, the output of v is in phase with the input x. When b/a is

greater than 1, phase shift occurs between x and v, with the output phase

lagging the input phase. The worst case is when a = 0, and at this point
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

a 45° phase shift occurs for \x\/b = 1. When \x\/b is very large, it can

also be seen that the phase shift between output and input approaches zero.

The magnitude of the fundamental component of the contactor square

wave output is also a function of b/a and x\/b. It is obvious that the

T-CONTACTOR

\ OUTPUT

V-FUNDAMENTAL COMPONENT

\ OF CONTACTOR OUTPUT, V

INPUT

x-lxl Sin cot

b=Q

b>a

WHEN b=a, NO PHASE SHIFT

WHEN t»a, FUNDAMENTAL OF SQUARE WAVE OUTPUT LAGS INPUT

WORST CASE WHEN 0=0, 45° PHASE SHIFT FOR -^=1

WHEN igl IS LARGE, PHASE SHIFT APPROACHES ZERO

FIG. 4-24. Contactor input and output phase considerations.

AMPLITUDE RATIO, .!£!

*D

FIG. 4-25. Relationships of A&/a(|x|/b) as functions of x|/b.

74

gain of the contactor, that is, its output/input ratio, falls off as |a;| is

increased. The ratio V/b appears directly as a gain factor in the con-

tactor amplifier, making the complete transfer function of the nonlinear


Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

section (V/b)Ab/a(\x\/b). The transfer relationship Ab/a(\x\/b) without

the fixed gain multiplier V/b is plotted in Fig. 4-25 for various values of
AUTOMATIC REACTOR CONTROL

75

b/a. It can be seen from this curve that the locus terminates abruptly at

\x\/b = 1. Obviously, for all amplitudes of x less than the start limit b,

the system does not respond. Figure 4-25 also indicates clearly that the

maximum value of phase shift can be only 45°.

In anticipation of our comparing the nonlinear complex amplitude

function — Ab/a(\x\/b) with the frequency-dependent function G~l, we

may replot the results of Fig. 4-25 in polar coordinates as Fig. 4-26. In

this figure we have also reversed the sign to obtain —Ab/a(\x\/b'). The

120'

150°

180°-4

1.1 10

FIG. 4-26. Relationships of — Aj/a(|x|/b) as functions of x|/b in polar coordinates.

1*1=1.4

abrupt stop at x\/b = 1 and the maximum phase shift of 45° are again

emphasized. These curves are presented again for various values of b/a

without the constant multiplying factor V/b. It is evident that as the

fixed gain portion of the contactor amplifier is increased in magnitude, the

curves of Fig. 4-26 will all be moved out radially.

Example of On-Off-type Reactor-control-system Operation. We are now

in a position to try the above approach. First, we can select a specific

control-loop configuration specifying transfer functions and gains for all

the components of the loop. Figure 4-27 indicates a block diagram of a


Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

specific loop. The transfer function form of the contactor has just been

derived. The actual reactivity rate into the reactor is the contactor out-

put v delayed by the lag in the drive motor, which has the form 1/(1 + rs).
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Since the motor delay depends on frequency, it may be conveniently

lumped in with the reactor transfer function. The demanded reactivity

rate v would then be input to the combined transfer functions of the

reactor and motor. For this example we may choose a specific value for

T = 0.5 sec. This is a reasonable value which many actual motors can

exceed.

For the transfer function of the reactor we shall select the combined

reactor temperature coefficient transfer functions KRTcGsTc(s) of Figs.

4-14 and 4-15 and carry through the analysis for different values of KTc.

It will be noted that our input to the reactor is now not a change in reac-

tivity but a change in rate of change of reactivity. Consequently, our

=x=lxl sin

Sn/n

sSk

FIG. 4-27. Servo block diagram of discontinuous reactor control loop indicating transfer

functions of principal components.

output will be a change in level as a function of a change in reactivity

input rate. That is, the transfer function is of the form (Sn/n)/sSk.

The 1/s term will modify the transfer function representation of Fig. 4-4

and effectively make the amplitude changes more pronounced as a func-

tion of frequency.

We may now combine all of the frequency dependent portions of the

transfer function G(s), invert this function, and plot it on a polar diagram

as shown in Fig. 4-28. It can be seen from this that the larger the tem-
Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

perature coefficient, the more open is the locus. Consequently, there is a

greater area for stable operation. With no temperature coefficient, the

smaller enclosed area drastically limits the range of V/b which can be

permitted for a stable system. The shape of these curves will, of course,

be modified by any specific motor time delay constant T.

76

Let us now go back to the comparison of the amplitude function and the

frequency function by combining the curves of Fig. 4-26 and Fig. 4-28 to

the same scale. This combination is shown in Fig. 4-29. Under these

conditions it will be recalled that V/b = 1 and it is obvious that the sys-

tem will oscillate for all but the very largest of the temperature coeffi-
AUTOMATIC REACTOR CONTROL

77

cient gain time constants. Even for the case where KTc = 0.019, the

phase margin is so small that the system is apt to be unsatisfactory.

This situation can be corrected by reducing V/b, and it is fortunate

that safety considerations usually require this ratio to be considerably

less than 1. Current practice in reactor operations limits V to a range of

10~3 to 10~6 5fc/sec. The start limit b depends upon the accuracy to

which it is desired to hold the power level, as b effectively represents an

100°

90°

110°

120'

130

140'

170'

180

20

-20

GAIN, db

FIG. 4-28. Polar transfer function G~1(s).

indeterminate dead zone. A value of b of the order of 0.1 might be a

reasonable number. Consequently, V/b might range from 10~2 to 10~6.

It will be noted that as V/b is reduced, the control system containing no

temperature coefficient becomes conditionally stable; that is, a range of


Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

V/b exists whereby, if its value were larger than a given amount or smaller

than another amount, the system could oscillate. For reactors having

negative temperature coefficients of sufficient magnitude this condition

does not exist. The only problem is to maintain V/b below a given

amount. Gain margins and phase margins here are determined by expe-

rience. In linear servo practice a phase margin of 35° and a gain margin

of 8 db are typical values for satisfactory stability and transient response.

In the nonlinear case it is likely that smaller phase margins might be


78 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

used, as the amplitude of any oscillation would be limited by the output

of the contactor.

Phase Shift Compensation. The safety considerations in limiting V/b

may provide a control system that does not oscillate but nevertheless

does not have suitable phase margin for a desired transient response.

The situation may be corrected by providing proper phase compensation

as indicated in the block diagram of Fig. 4-22. As an example, let us

100°

110°

120

130

140'

150'

170

180

160'

20

-20 -40

GAIN, db

-60

FIG. 4-29. Comparison at unity gain of amplitude function — A»/a(|x|/b) with frequency

function G~1(s).

consider the case just presented where b/a = 4, V = 10~35/c/sec, b = 0.1,

and KTC = 0.0047. This situation is shown in Fig. 4-30. It can be seen
Generated on 2011-10-24 11:43 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

that although adequate gain margin exists, the phase margin is only 22°,

which may be small when one considers that this elementary analysis did

not take into account any of the second-order effects which must exist

in a practical physical system. Phase compensation may then be used

and a G~l curve modified, as shown by the dotted line in Fig. 4-30. Here

adequate phase margin is provided. Some type of phase compensation

is generally used in practical systems.

4-9. Transient Response of Control Loop. The assumption is usually

made that an optimum steady-state response leads to a reasonable tran-


AUTOMATIC REACTOR CONTROL

79

sient response, but these qualifications depend upon the reactor usage

and its safety factors. A transient overshoot of 2:1 probably is not

damaging to most reactors providing it does not last too long. However,

under certain high-power operating conditions, transient overshoots of

this magnitude could conceivably cause damage to a specific reactor.

Rather than attempting to set down absolute permissible magnitudes of

transient response, it is more fitting to examine now the form of the

100°

110°

120'

130

140f

170'

180'

150°

160°

20

-60

0 -20 -40

GAIN, db

FIG. 4-30. Stability of discontinuous control-loop example with and without phase com-

pensation.

response to given transients and to let the suitability of this response be


Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

determined from an over-all picture.

The transient response to a change in power-level demand is usually

unimportant compared with the response to a change in reactivity. The

power demand change rate can be limited externally to any desired value,

but the internal rate of change of reactivity may, on occasion, be beyond

the control of the design engineer. Consequently, this is the type of

transient that is apt to be more disturbing to the system. We shall

consider the system response to a severe disturbance which will be a

sudden step change in reactivity. Physically, the problem becomes:

Given a specific control loop, how does the reactor output power level

change if the reactor is suddenly subjected to a step change in reactivity


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

from any cause? Controlwise, the problem resolves into: Given the

transfer functions of all the control-system components, what is the time

variation of the output power signal Vn(t) for a step-function input of

reactivity?

In order to approach the form of the output response, let us again

simplify the block diagram of Fig. 4-11 into the servo block diagram of

Fig. 4-31. Here again the reactor transfer function has been indicated

as KRGR(s) and the error-signal amplifier, the actuator, and control-rod

mechanism have been lumped together into the common transfer func-

tion KBGB(s). The comparator is indicated as the summation device

between F0 and Fn, and the neutron detector is indicated as having only

FIG. 4-31. Servo block diagram of reactor control loop used for transient analysis.

pure gain K0 independent of frequency. In order to inject steps of reac-

tivity conveniently, means have been provided to insert the disturbance

5fce. It will be noted from this block diagram that no phase compensation

has been added and, for simplicity, the temperature coefficient of reactiv-

ity has been ignored. From the block diagram of Fig. 4-31 we can solve

for the Laplace transform Vn(s) in terms of the transfer functions of the

system components and the reactivity. We can then take the inverse

transform to obtain finally Vn(t). From Fig. 4-31 we can see the follow-

ing equivalents using Laplace notation:

Ve(s) = F0(s) - Fn(s)

Sk(s) = Ski(s) + Ske(s)

Vn(s) = K0n(s)
Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

(4-42)

(4-43)

(4-44)

(4-45)

80

It will be recalled from the determination of the transfer function of

the reactor given in Sec. 3-4 that only oscillations about a steady-state

level were considered in deriving the transfer function. This simplifica-

tion enabled us to remove the steady-state power-level term from the

analysis. To consider the effect of transient changes of reactivity on the


AUTOMATIC REACTOR CONTROL 81

sj'stem as a whole, however, it is necessary to reintroduce the steady-

state term. From this statement it follows that

n(s) = n0(s) + 8n(s) (4-46)

«n(«) = K«GB(s)Sfc(s) (4-47)

Now combining Eqs. (4-42) to (4-47) we have

Vn(s) = K0n0(s) + K0Sn(s) (4-48)

= 70(«) + K0KBGR(s)Sk(s) (4-49)

-V.(s) = Vn(s) - V0(s) = KoKnGn^Sk^s) + Ske(s)} (4-50)

K^GBVe(s) + Ske(s)} (4-51)

K0KBGR(s)Ske(s)

K0KBKBGR(s)GB(s)

(4-52)

I •".||*».If-'Itv«v>"v«v«V i A ro\

h 1 + KoKBKBGR(s)GB(s) l

We can assume that during the reactivity change, the power-level demand

does not change, so that VQ(f) is a constant.

In order to determine the inverse transform of Fn(s) we must know the

actual transfer functions of the components in the control loop. For

the components indicated in Fig. 4-31 we must stipulate a transfer func-

tion form for KRGR(s) and KBGB(s). Elementary approximations will

again be used to indicate the form of solution. Let us assume that we

are going to attempt to control our reactor with a fast control system of

such a speed that we do not have to consider delayed neutrons but rather

are effectively controlling on prompt neutrons. We shall also assume


Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

that the response of the control system is so fast that we can ignore the

integrating effect of the reactor. For a solution of this type we can

approximate the reactor transfer function by

(4-54)

We can similarly approximate the transfer function of the driving sys-

tem, which consists of the amplifier and the actuator mechanism. An

illustrative form of transfer function for a hydraulic mechanism might

be given by

We are now ready to substitute the expressions for the transfer functions

into Eq. (4-53), which yields

*2

k^ (4-56)
82 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The Laplace transform of a unit step function is 1/s, so that

«*.(«) = —

(4-57)

where Afc represents the magnitude of the step change. Equation (4-56)

then reduces to the form

where

S2 + diS + OO

T s4 + t

flo = <*>2

ai = 2fo>

..

(4-58)

The inverse transform of this equation can be found to be of the forms

Vn(t) = V0 + AV0Sk[e-«l - Be~?' + Ce~* cos (xt + 0)] (4-59)

The form of these results, of course, depends on the specific values of

the constants used. For illustration, Eq. (4-59) can be solved for the case

Ul

s^-6k=0.02 STEP

6k=0.015 STEP-

8k=0.003 STEP

r6k=0.001 STEP

0.2

0.4
Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

0.12

0.14

0.16

0.6 0.8 0.10

TIME, SECONDS

FIG. 4-32. Transient response of an example of proportional-type control loop for various

step reactivity inputs.

of a reactor having an I* of 4 X 10~4 sec, controlled by a fast hydraulic

servo loop having constants f = 0.5, 01 = 26 radians/sec, T = 0.069 sec,

KoK. = 120F0, and K0K,KB = 1.62. Figure 4-32 shows the response of

this system to step reactivity changes.


AUTOMATIC REACTOR CONTROL 83

It can be seen in this approximate solution that the peak heights of

the transient response are a direct function of the input Sk step amplitude

and that the time for the transient to return to its steady point is approx-

imately constant. It must be remembered, however, that the transfer

function of the reactor was derived for small oscillations. Consequently,

the response, as determined in this manner, will be correct only for rela-

tively small values of Sk steps.

Elementary Control System for Reactor with Negative Temperature Coeffi-

cient. Approximate solutions for the stability and transient response of

a proportional-type reactor control loop in which the reactor has a nega-

tive temperature coefficient are ac-

complished in a similar manner to

that just presented. In the stabil-

ity considerations the block diagram

of Fig. 4-33 may be used and the

open-loop gain obtained and ex-

amined. It will usually be found

that greater gain can be permitted

DEMAND

<8>

FIG. 4-33. Elementary block diagram for

transient analysis of reactor control loop

considering temperature coefficient.

for the control loop with the nega-


Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

tive temperature coefficient. The

transient analysis becomes much

more complex, and another method of handling this will be presented in

Sec. 4-10.

4-10. Determination of Control-loop Performance by Simulation

Technique. In order to eliminate many of the approximations we have

just made in the transient analysis of an actual system, we can resort to

a change in technique and attempt to find more exact solutions to the

transient problem using analogue-computing methods. First, however,

it becomes necessary to define in more detail the problem of what is

required from an over-all reactor control system. It has been seen, from

the elementary transient analysis just performed, that many parameters

are involved. Each one of these parameters can conceivably be opti-

mized one against another, provided that a set of standards are laid down.

Although some reactors possess no significant temperature coefficient, the

majority of the reactors in use at present have substantially large nega-

tive temperature coefficients. The magnitude and the effective time con-

stant of this temperature coefficient must be taken into account, and

many items added to our elementary control-loop block diagram.

In general, a system of this sort may be examined by four criteria:

1. Absolute stability—this criterion means no exponentially increasing

oscillations can exist.

2. Relative stability—what is the number of oscillations before a tran-

sient has essentially died out?


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

3. Maximum power-level excursion—the peak power reached by the

reactor for a fixed-shape input disturbance.

4. The restoring time of the system—the time for the transient to die

out essentially.

It is apparent, then, that various systems and components may be

compared by means of these criteria and decisions made on an over-

all basis as to what is the best control configuration for a given

problem.

Specific Example. In order to grasp the magnitude of this problem,

let us follow through this examination procedure for a specific example.

Assume that a control system has a block diagram such as is given in

Fig. 4-34. In this block diagram we have a reactor that, for the sake of

COMPARATOR

CONTROL ROD SERVO

SYSTEM

FIG. 4-34. Servo block diagram for simulator setup.

example, has a mean neutron lifetime of 10~4 sec and, consequently, the

transfer function given in Figs. 3-15 and 3-16. The reactor also has a

negative temperature coefficient, and this negative temperature coeffi-

cient will be treated as a separate servo system around the reactor as

indicated in Sec. 4-3. Consequently, it will have a given time constant,

or bandwidth, that depends upon the type of reactor construction.

The heat-transfer coefficient from the fuel of the reactor to the mod-

erator will principally determine the time constant of the temperature


Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

coefficient effect. Actually, in a practical plant, as mentioned in Sec. 4-3,

there will be two time constants involved. The temperature coefficient

is one of the most important variables in the plant, and it is desired to

study its effect upon the control system from ranges of KTC = 0 to

approximately 0.045fc/(n/n ~ no).

Control-rod Servo System. The control-rod servo system transfer func-

tion to be examined in this example is

Sk/(n — no)

84

(4-60)
AUTOMATIC REACTOR CONTROL 85

, jr . . ,- .. -, Sk/(n — n0)

where K = gain in reactivity per unit error,

HQ

fn = <0n/2ir = undamped natural resonant frequency of the control-

rod servo system, cycles/sec

f = damping ratio of the servo system

/o = l/2irro = an integrating break frequency, cycles/sec

The choice of this form of the transfer function of the control-rod system

is made for several reasons. First, it is an excellent approximation to

some of the practical servo systems now in use in reactor control systems.

Second, a quadratic form of this sort should closely approximate the

behavior of almost any type of physical equipment which might be chosen

to drive a control rod. That is, an electric motor or hydraulic system

should be essentially of this form. The (TOS + I)/TOS portion of this

transfer function is used principally to eliminate the steady-state error

between n and n0. It will be recognized from an examination of the

block diagram of Fig. 4-34 that an error can exist between n and n0 and

yet the reactivity requirements that k = 1 for steady state of the reactor

be completely satisfied by the temperature coefficient effect system alone.

The integration thus provided by this transfer function eliminates a

steady-state error, but for many control systems this portion of the trans-

fer function might not be used.

It can now be seen that to study this given reactor control system

completely requires an understanding of the effects and optimization of


Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the following seven variables:

1. The magnitude of the temperature coefficient

2. The time constant of the temperature coefficient

3. K, the system gain

4. /n, the natural frequency of the control system

5. f, the damping ratio

6. /0, the integrating break frequency

7. The shape of the transient disturbance

It is possible to examine the stability and the transient response by

the methods outlined in Secs. 4-7, 4-8, and 4-9. To determine the stabil-

ity and to solve for the transient responses for a range of values of all

the various parameters within reason for a system of this sort by means

of these methods is a long and tedious procedure. With this large num-

ber of variables, an empirical approach may be tried and methods of

analogue computation and simulation are quite attractive. Chapter 11

outlines the means of designing simulation apparatus. We shall assume

that a specific simulator has been designed for the example of Fig. 4-34

and that all of the seven above-mentioned parameters may be varied.

Let us now carry through this specific problem. The method to be

used will be to set up the simulated plant, inject given types of disturb-
86 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

ances, and study the effects of changing the parameters one at a time.

It would be anticipated that the form of the transient output disturbance

would be essentially that of Fig. 4-32. The particular local conditions

we can select for this study are that the reactor is operating at full power,

the power demand n0 fixed at full power, and no manipulation is to be

permitted to the coolant flow through the reactor. A positive reactivity

disturbance will then be inserted into the reactor, and the form of the

power output transient will be examined.

Input Disturbance. The input disturbance to be used will be, as before,

one in which the reactivity into the reactor is changed in a prescribed

manner. Since one of the evaluation criteria of the performance of the

servo system is based upon its response to this transient disturbance, it

is clear that the transient disturbance must be specified accurately. The

form of the disturbance to be used is a first-order time delay of the form

Sk = 0.003(1 — e~l/r). This type of disturbance is chosen because, by

selection of the proper value of the parameter r, an input is obtained that

is reasonably typical of the form of transients to be expected in actual

physical systems. The value of T = 0 denoted a 0.0035& step function

which is obviously physical fiction. Nevertheless, responses to this

type of input are of interest because the analytical calculations for this

type of response are comparatively simple. In addition, a step input

gives the worst transient response in most physical systems, and thus an

upper limit is established on the system performance. The use of a

constant amplitude of reactivity change of 0.0035A; is quite severe and


Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

probably represents a worse case condition in that reactivity changes of

0.003 in Sk are very unlikely in most reactors operating at a constant

power level.

Temperature Coefficient. It will be recognized from Fig. 4-34 that the

effect of the negative temperature coefficient is an inherently stabilizing

one in the over-all reactor control system. In fact, for many types of

reactors there is no need to have an external servo system for transient

protection when the negative temperature coefficient is of sufficient mag-

nitude. However, with analogue computational techniques it is just as

simple to examine the effect of the negative temperature coefficient acting

alone or acting in conjunction with an external servo system, and this is

the procedure that will be used. The temperature coefficient magnitude

will be defined by the use of KTc — ATC, as indicated in Sec. 4-3.

Procedure. We can first optimize the constants of the external control-

rod servo system. It will be found in this type of problem that the damp-

ing factor, the gain, and the integrating break frequency can all be set

by cut-and-try methods on the analogue simulator to effective optima.

These optimum constants can then be used while the natural frequency

response of the servo loop is examined. For these optimum conditions


AUTOMATIC REACTOR CONTROL

87

the control-rod servo system may be specified by one parameter alone,

that is, the natural frequency of the control servo, /n.

Damping. We can now proceed with the optimization process. Fig-

ure 4-35 shows the effect of varying -f from 0.5 to 1.5, when a 0.0035A;

step transient is inserted into a system having a temperature coefficient

gain of KTC = 0.039, a natural frequency of 0.3 cycle/sec, and an inte-

grating break frequency of 0.032 cycle/sec. It can be seen that the over-

damped servo provides the best of the three damping coefficients con-

sidered. This figure also shows that a higher gain is permissible for the

same stability in the overdamped condition. The time to reach a steady

state following the transient differs little between the cases shown in

170

100

34567

TIME, SECONDS

FIG. 4-35. Control-system transient response to 0.0035(t step function as damping coeffi-

cient is varied.

Fig. 4-35 in that the damping ratio affects principally the extent of the

oscillation following the transient. Although these results are indicated

for only one value of /n and temperature coefficient, the performance of

the system as a function of damping constant would be found to be similar

for other natural frequencies and temperature coefficients.

Gain. As Fig. 4-35 indicates that f = 1.5 gives the best system
Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

response, only data taken from f = 1.5 need be considered from here on.

The next parameter to optimize is the loop gain K. The effects of vary-

ing K are indicated in Fig. 4-36. This figure illustrates that the loop

gain is not a critical adjustment within limits. A factor of 10 in gain

reduction from a stable maximum does not cause extreme differences in

transient shape. The primary effect of higher gain is to reduce the ampli-

tude of the tail of the transient without seriously affecting its slope. If

the gain is increased higher and higher, however, instability ultimately


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

100J—

10

11

4567

TIME. SECONDS

FIG. 4-36. Control-system transient response to 0.0035k step function as loop gain is

varied.

4567

TIME, SECONDS

10 1)

FIG. 4-37. Control-system transient response to 0.0035fc step function as ratio of inte-

grating break frequency to natural frequency is varied.

Sk

will result. As indicated from Fig. 4-36, gains of K = 0.152 7 ^7—

\n — nQ)/no

might be permitted.

88

Integrating Break Frequency. We have just seen that the gain and the

damping may be set empirically to optimum values in that neither is

particularly critical. The integrating break frequency /0 can next be


Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated on 2011-10-24 11:44 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

control servo alone. The measure of the servo performance of the tem-

perature coefficient alone can be obtained by extrapolating the curves

containing temperature coefficients to the left to a zero value of /n, which

effectively corresponds to the elimination of the control servo. If the

peak power values thus obtained are projected horizontally to the right,

the intercepts with the zero temperature coefficient curve give the equiv-

alent control-system performance natural frequency. For example, the

temperature coefficient gain of KTC = 0.009 has the equivalent servo

natural frequency of 0.08 cycle/sec. The temperature coefficient effect

of KTC = 0.039 has the equivalent servo frequency of 0.4 cycle/sec.

300

0.1 1.0

CONTROL ROD SERVO UNDAMPED RESONANT FREQUENCY, fn

FIG. 4-41. Reactor peak power level reached as a function of natural frequency and

temperature coefficient in optimized control system with 0.003 step input.

4-11. Peak Limiting by Negative Temperature Coefficient. Before

attempting to evaluate the usefulness of these curves for design purposes,

an examination of the information just supplied by analogue-computer

studies indicates that the negative temperature coefficient alone is capable

of limiting the peak power output value which a reactor can attain under

the condition of step input transients. This answer can also be obtained

analytically. Let us consider a bare reactor having no control system,

and let us mathematically ignore its temperature coefficient for the

moment, except to state that it is a strong one. For the case of a step
Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

input we can again write the familiar kinetic equations

dn

Tt

Sk —

n+

..=i

(4-61)

(4-62)

92

~dt
AUTOMATIC REACTOR CONTROL 93

Initially in the steady-state condition

= 0 Sk = 0 \id = U<> (4-63)

If we now inject a step Sk into the reactor, we can isolate the steady-

state and variable portion of n as before into

n = n0 + Sn (4-64)

Let us assume that following an input step of reactivity, the delayed-

emitter concentration remains the same for a short interval of time.

Then

d(Sn) _

- v TT-^ ) ("o + Sn) + £ n0 (4-65)

ttL \ V J V

Sk 8 - Sk

T* o 7*

The solution for Sn then becomes

Sn = ^^r n0(l - e'W-'W"'!') (4-66)

fj — O/C

and the maximum initial value that Sn can obtain for positive steps in

Sk without the contributions of the delayed emitters becomes

Sk

n0 (4-67)

mn ,

In a reactor with a negative temperature coefficient the reactivity

change created by the rise in power level is in the opposite direction to


Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the step in reactivity. Therefore, it tends to cut back the peak Sn as

given in Eq. (4-67). The time constants are the important factors here.

If in a reactor having a strong temperature coefficient the time constant

of this temperature coefficient is in the order of a few tenths of a second,

then the peak in Sn can never rise above the value given in Eq. (4-67).

As an example, let us insert a step Sk of 0.003 into a reactor having an

I* of 10~4 sec. The time constant for the initial rise l*/(fi — Sk) becomes

22 milliseconds (msec). As the effects of the delayed neutrons do not

become apparent in most practical reactors for approximately 0.3 sec

(see Sec. 3-2), reactors having fast temperature coefficient time constants

are effectively protected by themselves against large Sk step transients.

Figure 4-42 plots Eq. (4-67) and indicates the peak powers that can be

attained by reactors having strong, fast negative temperature coefficients

under the influence of step input reactivity transients.


94 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

An examination of the peak levels involved in Figs. 4-38 and 4-39

indicates that the temperature coefficient of Fig. 4-38 is effectively limit-

ing the peak power to the value indicated for a 0.0035A; step in Fig. 4-42.

The curves of Fig. 4-39, however, indicate that the temperature coeffi-

cient corresponding to a KTc of 0.009 is not quite strong enough to con-

tain the initial dn to the values of Fig. 4-42.

4-12. Evaluation of Transient Response. Although we have estab-

lished four criteria in Sec. 4-10 for comparing reactor control systems,

we have not as yet determined how

these criteria are involved in the

over-all picture of reactor and plant

operation. The question of philos-

ophy of control, coupled with many

other problems such as mechanical

22

I"

£

1.4

1.2

1.0.

BREflKDOWN

LEVEL

0.004

0.002
Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

8k-STEP

FIG. 4-42. Maximum possible transient

peak power level for a reactor with a

strong negative temperature coefficient as

a function of input reactivity step.

NORMAL RATING PLUS

SAFETY FACTOR

PERMISSIBLE OPERATING TIME

FIG. 4-43. Permissible power output from

any type of generating apparatus as a

function of permissible operating time.

damage to the reactor, metallurgical difficulties, and plant limitations,

must be considered.

The first problem concerns itself with the question: What are the lim-

itations of the reactor that the control system must protect? Basically,

a reactor is no different from any other power-producing device insofar

as its ability to handle an overload is concerned. Figure 4-43 illustrates

the form of the permissible amount of power that may be taken from any

piece of generating apparatus as a function of time. For example, if the

apparatus under consideration is an electrical generator, the generator is

designed to operate continuously at a fixed rated power output. Actu-

ally, in a practical design a small safety factor is usually permitted.

Therefore, the generator may operate continuously at some rating slightly

Q
AUTOMATIC REACTOR CONTROL 95

above its normal full power rating. If one now considers operation for

short times, that is, times of the order of the time constant of the thermal

heating of the generator, a slightly greater output from the machine may

be obtained. If the generator will not reach its normal internal tempera-

ture ratings for 5 min at full power, then conceivably one might take more

than full power from the generator for a time shorter than 5 min. This

relationship is presumably hyperbolic up to the point where some other

factor, such as insulation breakdown, occurs. That is, if full power

might be drawn from the machine for 5 min, one could conceivably take

twice full power from the machine for 2)^ min or four times full power for

1y± min. At some power level breakdown would occur regardless of

how short the interval of time was made.

In the case of a reactor the ultimate limitation may be the destruction

of the reactor structure or melting down of the fuel. In addition, power

reactors may be of such a size that they effectively possess comparatively

little heat capacity. This means that the time scale on Fig. 4-43 might

be shifted so that the continuous rating plus the safety factor must be

carried down into times comparable to 0.1 sec. If this is the case, for

times shorter than 0.1 sec it is indeed possible to run power reactors at

higher power level than their rated level up to the point where something

distorts or melts. Actually, it is the integrated area under a curve

similar to that of Fig. 4-43 which is of importance. The number of watt

seconds of energy available is what will ultimately cause damage. For

examining control-system transients most breakdown phenomena can


Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

be translated into power level for use with our four basic criteria. For

example, the initial indication of a failure might be the melting of the

moderator. This melting can be translated into a power level.

Let us now go back to our example as illustrated by the transient-

response curves in Figs. 4-38, 4-39, and 4-40. Two sets of information

are at once apparent. The first is that under some conditions no exter-

nal control system is needed to restrain severe transients. The tempera-

ture coefficient by itself is quite capable of handling a wide range of

input disturbances. The output transients and power level encountered

with modest input transients and no control system would not appear to

be too severe for many types of reactor plants. The second obvious

information that may be derived from these curves is that if one does use

an external control system, the higher the natural frequency the better;

or the faster the control system acts, the smaller will be the over-all

transient disturbance.

The question then arises: Why not build extremely fast control sys-

tems which have natural frequencies in the order of 100 cycles/sec?

Here over-all safety philosophy dictates against such systems. First,

the components which are used in these fast systems are not at the
96 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

moment as reliable as those used in slower systems. Second, a failure

in a fast-acting system can in itself inject a fast transient disturbance.

The faster the capabilities of the system, the more severe the transient.

Therefore, the argument may be used that if nothing about the reactor

control system is capable of changing at a fast rate, no fast accidents

can occur. From a safety point of view one should then design any

needed control system to be as slow as possible and still satisfy the over-

all performance requirements.

It will be realized that reactors must have an external control system

of one sort or another for purposes other than transient-disturbance

control. That is, means must be provided to extract control rods for

startup processes, for poison compensation, and for temperature control.

Control-rod mechanisms will be needed to shut down the reactor for

normal operation or emergencies. Consequently, some combination of

basic control-loop elements that are capable of being connected together

either manually or automatically in a control circuit will exist around a

reactor.

4-13. Procedure for the Selection of Control-system Constants.

Granting that a reactor control system is needed, the following procedure

may be used as a guide to determining the control-system constants:

1. Determine the temperature coefficient magnitude and time constant

•or the ranges that the temperature coefficient might have during the

reactor lifetime.

2. Determine the integrated or peak energy permitted from the reactor


Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

as a function of time.

3. Decide the form of the control-system mechanisms, and derive the

transfer functions.

4. Determine the adequacy of existing components of the type chosen

as a function of the natural frequency of the control loop.

5. Consider the performance of the reactor control system by simula-

tion techniques or others, and evaluate by means of the four criteria given

in Sec. 4-10 for the type of desired response.

6. Choose the slowest, safest system that fits all the above conditions.

REFERENCES

1. Brown, G. S., and D. P. Campbell: "Principles of Servomechanisms," John

Wiley & Sons, Inc., New York, 1951.

2. Chestnut, H., and R. W. Mayer: "Servomechanisms and Regulating System

Design," John Wiley & Sons, Inc., New York, 1951.

3. James, H. M., N. B. Nichols, and R. S. Phillips: "Theory of Servomechanisms,"

McGraw-Hill Book Company, Inc., New York, 1947.

4. Stephenson, R.: "Introduction to Nuclear Engineering," McGraw-Hill Book

Company, Inc., New York, 1954.


AUTOMATIC REACTOR CONTROL 97

5. Harrer, J. M., and J. A. Deshong, Jr.: Discontinuous Servo for Control of Power

Reactors, Nucleonics, vol. 12, no. 1, p. 44, 1954.

6. Grace, J. N.: Synthesis of Control Systems for Nuclear Power Plants, "Conven-

tion Record of the IRE 1954 National Convention," pt. 9, "Medical and Nuclear

Electronics," Institute of Radio Engineers, New York, 1954.

7. Kochenburger, R. J.: A Frequency Response Method for Analyzing and Synthesiz-

ing Contactor Servomechanisms, Trans. AIEE, vol. 69, pt. 1, pp. 270-284, 1950.

8. Churchill, R. V.: "Modern Operational Mathematics in Engineering," McGraw-

Hill Book Company, Inc., New York, 1944.

9. MacColl, L. A.: "Fundamental Theory of Servomechanisms," D. Van Nostrand

Company, Inc., New York, 1945.

10. Bode, H. W.: "Network Analysis and Feedback Amplifier Design," D. Van

Nostrand Company, Inc., New York, 1945.

11. Lauer, H., R. Lesnick, and L. E. Matson: "Servomechanism Fundamentals,"

McGraw-Hill Book Company, Inc., New York, 1947.

12. Moore, R. V.: The Control of a Thermal Neutron Reactor, Proc. Inst. Elec. Engrs.

(London), vol. 100, pt. 1, p. 90, 1953.

13. Bowen, J. H.: Automatic Control Characteristics of Thermal Neutron Reactors,

Proc. Inst. Elec. Engrs. (London), vol. 100, pt. 1, p. 102, 1953.

14. Nyquist, H.: Regeneration Theory, Bell System Tech. J., vol. 11, p. 126, January,

1932.

15. Weinberg, A. M., and W. K. Ergen: Some Aspects of Non-Linear Kinetics, Proc.

Kjeller Conference on Heavy Water Reactors, JENER, 1953.

16. Lipkin, H. J.: A Study of the Non-Linear Kinetics of the Chatillon Reactor, J.
Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Nuclear Energy, vol. 1, pp. 203-213, 1955.


CHAPTER 5

REACTOR CONTROL MECHANISMS

In Chap. 4 reactor control loops were examined and it was shown that

the output of the control system was usually a control-rod motion which

changed reactivity. Before going on to plant control, it is of interest to

examine some of the actual types of mechanisms that may be used to

actuate these control rods. Of course, moving control rods is not the

only way of changing reactivity in a control sense. For example, neutron-

absorbing chemicals might be inserted into a reactor and change its

reactivity with no physical rod motion involved. In this chapter, how-

ever, we are going to consider only the motions of neutron-absorbing

types of control rods. That is, we are not going to consider forms of

reflector control or other schemes that change the leakage of the reactor,

but the assumption will be made in all cases that moving a control rod into

the reactor structure reduces the reactivity and tends to turn the reactor

off.

Physically, there have been as many different types of control rods as

there have been reactors built. The only requirement for the absorptive

type of rod is that the rod be black or at least "dark gray" to thermal

neutrons. Many shapes of control rods have also been used. Some

have been round, some square, some hollow, and some solid, all depending

upon the design of the individual reactors. For thermal reactors, boron

and cadmium are the most commonly used absorbing materials, and

these materials may be alloyed into some other metal or may be clad by

another material. In this chapter we shall consider some of the mechan-


Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ical requirements of the mechanisms that move these rods in and out of

the reactor. Brief descriptions will be given of the features of the simpler

typical systems.

5-1. General Requirements of Control-rod Mechanisms. The con-

trol rods themselves may be classified in three groups according to func-

tion: safety rods, shim rods, or regulator rods. Each function may call

for a specific speecTof motion. As we have seen in Chap. 4, the purpose

of the regulator rod is_jtomaintair£ a constant desired power level in the

face oiTransient reactivity changes and to assist in making a change in


REACTOR CONTROL MECHANISMS 99

this power level. The control aspects and speed requirements of the

regulator rocTwere shown to be variable, but in general the speed was to

be as slow as good transient performance would permit. Movement of

the regulator rod was usually automatic, but manual operation could be

permitted.

As has been previously indicated, shim rods are used either for auto-

matic follow-up of the regulator rod or for the slow movement of larger

amounts of reactivity than can be handled by the regulator r63. Shim

rods afe~particnlarly concerned with the startup process, and as will be

shown later, their required motion is generally quite"sldw.

Safety rods, on the other hand, are for scramming or rapid shutdown

action, and their motion inward is quite fast. Details of the speed

requirements of safety rods will also be given later.

It is obvious then that different mechanical motion requirements may

exist for each of these three functions. In some reactors these functions

will be separated, and in others they may be combined. That is, during

startup a shim rod may be operated at its normal slow speed, but as soon

as it is extracted from the reactor, it conceivably may have its function

changed so that it becomes a safety rod. The usual situation in most

reactors is that regardless of the normal function of the control rod, it will

also have provision for some form of fast insertion into the reactor.

For many purposes it is also desirable to move more than one rod at

a time; consequently, mechanisms must be designed which permit either

mechanical or electrical ganging. The electrical ganging systems are


Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

usually more flexible in that it is easier to change the number of rods in

any given bank. It can be seen that these functional requirements

dictate the possible forms of mechanical configuration that a control-rod

mechanism may take.

Environmental Conditions. The type of reactor also affects the design

of the control-rod mechanism in that the reactor purpose and operation

create different types of environmental conditions. Most reactors can

be divided into two environmental classifications: nonpressurized and

pressurized. The nonpressurized reactors at present are a group of

experimental or research reactors which sit at room temperature and

room pressure ambient. Thus they can havfe their mechanisms out in

the open where they are easily accessible and where lubricants can be

applied to given portions of the mechanisms.

The AEG has announced that the first commercial central station

power reactor will be a pressurized water reactor. Many of the other

conceivable power reactors are also of the pressurized type. The coolant

or the combination moderator coolant may be some substance such as

water or gas, and this substance is kept at a high pressure in order to

increase the thermal efficiency of the plant. Implying that a reactor is


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

100

pressurized, however, means that either part of the mechanism must be

inside a pressure shell or some complex means used to control the reactor

from the outside of the pressure shell.

For the pressurized type of reactor the mechanism may be buried

completely inside a large pressure vessel type of tank or, as is more com-

mon, some form of thimble may be used in which the mechanism can

operate. Figure 5-la indicates an internal thimble configuration, and

Fig. 5-16 an external thimble setup. In Fig. 5-la the entire mechanism

-CONTROL ROD

DRIVE

MECHANISM

EACTOR

ESSURE

ESSEL"^,

CONTROL ROD

/////////////

\l

r/////// '//////

\\

INTERNAL ^

THIMBLE X

11
Generated on 2011-10-24 11:45 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

1 REACTOR

'////////////,

/////

/////////////

EXTERNAL

THIMBLE

REACTOR

- CONTROL

ROD DRIVE

MECHANISM

,REACTOR

PRESSURE

VESSEL

CONTROL

ROD

(a) (b)

FIG. 5-1. Pressurized reactor thimbles, (a) Internal thimble, (b) External thimble.

and control rod are actually outside the reactor pressure vessel, but the

pressure vessel wall is extended down to where it is a part of the reactor

structure. In the external type of thimble the control rod alone extends

into the reactor but the thimble contains a part of the drive mechanism.

Some form of coupling then must be arranged through the thimble wall

to the basic control-rod drive. It can also be seen that the portion of the

mechanism inside the reactor pressure vessel enclosure must operate

without any maintenance and in high-power reactors without the benefit

of lubricants because of the deleterious effects of gamma radiation.

On the other hand, the internal thimble configuration requires more

fissionable material to compensate for additional neutron-absorptive loss

in the thimble shell. It is also more difficult mechanically to fabricate


REACTOR CONTROL MECHANISMS 101

and assemble the pressure vessel and reactor structure for the internal

thimble system when there are many rods used in the reactor.

Other Mechanical Requirements. In addition to the environmental

requirements and to the speed of motion requirements imposed upon a

control-rod mechanism, there are other factors which also must be con-

sidered in the design of these mechanisms. One basic problem which

always exists is that of providing some form of energy storage either in the

mechanism itself or in the control system feeding the mechanism. The

purpose of this energy storage is simply that in the event of a power

failure it is most desirable still to be able to have a means of inserting the

rods completely into the reactor. Many forms of energy storage, such

as gravity, springs, or inertia, may be used, and these will be discussed

later.

Also involved in the design of the mechanism is the problem of the

expected life of the reactor and the mechanism. Because of the difficulty

of servicing a reactor and its associated parts, the mechanisms usually

must be designed for perfect operation for a long time. This problem

is complicated by the fact that the scramming operation often is a violent

one involving large mechanical forces. Severe damage may be done to

the mechanism or the reactor structure if a reactor is scrammed too often.

Means for softening these forces are usually provided in most mechanism

designs.

Another requirement which is usually built into the mechanism is the

requirement of rod position indication. For neutron physics reasons,


Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

such as reactivity measurements, or for control information, such as to

provide position feedback, it is desirable to know the position of a control

rod quite accurately. In nonpressurized reactors this poses no problem,

but in the case of a pressurized reactor system, accurate determination

of the position of a control rod inside a pressure vessel may be a difficult

design limitation.

Operational requirements may also interpose special design require-

ments. In reactors used in military plants it is conceivable that the

control mechanism might have to operate under external shock and vibra-

tion conditions. Conflicting limitations may arise because of this require-

ment. For example, take the case of scramming the control rod. From

a safety point of view it might be desirable to have a mechanism of the

so-called hair-trigger type, with which just a very small motion will

cause the rod to scram, but this type of mechanism could also cause a

scram in the event of shock. If a different mechanism were designed

such that it did not unlatch easily under shock for its normal operation, it

might take too long in releasing the rod for a scramming operation.

5-2. Motors and Mechanisms for Control Rods: Nonpressurized

Systems. The motors and mechanisms that may be used for control-rod
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

drives can be arbitrarily divided into two groups: those mechanisms used

for nonpressurized reactors and those used for pressurized reactors.

Actually, all the pressurized systems may be used for nonpressurized

reactors. However, some of these drives are particularly well adapted

to pressurized conditions and consequently will be treated separately here.

Electric motors are generally used in the basic power arrangement for

the control-rod drive. The simplest cases are those whereby only a con-

stant speed is required of the motor. The motor is then geared down

through the rod-driving mechanism to set the rod speed at the desired

value. Three-phase induction motors are common for this application.

AC LINE

-L 90° PHASE

SHIFT

FIG. 5-2. Control loop for two-phase a-c servomotor drive system.

Single-phase induction motors are probably better in that the loss of a

phase in a three-phase motor causes the motor to lose its ability to change

the direction of travel of the control rod. A-c induction motors are

preferred over d-c motors in that no accident can cause them to operate

faster than synchronous speed. Also, their maintenance is smaller.

The a-c motor, however, must be reversible.

Two-phase induction motors are easily reversed and have the advan-

tage that certain designs may be used directly as servo control motors.

Figure 5-2 indicates a control loop such as might have been used in Chap.

4 for a regulator-rod drive, using a two-phase a-c servomotor. In this


Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

diagram it will be noted that the d-c output of the neutron detector must

be converted into alternating current to provide a signal to drive one of

the phases of the servomotor. Switching arrangements may be supplied

as desired to operate manually or reverse the direction of the motor drive.

102
REACTOR CONTROL MECHANISMS

103

For nonpressurized reactors gearing arrangements are the simplest

forms of mechanisms used following the drive motor. Rack-and-pinion

drives and screw-type drives are quite common. Figure 5-3 indicates the

simple rack and pinion, whereby a drive motor of any type drives a pin-

ion through a gearbox and a clutch. The clutch provides a convenient

means of slip at the ends of the rod travel and also provides one means of

= CLUTCH

PINION

-RACK

-CONTROL ROD

REACTOR

FIG. 5-3. Elementary rack-and-pinion drive for control rod.

direct gravity scram by releasing the control rod from the gearbox. This

type of mechanism is particularly adapted to low-cost research reactors.

Two screw-and-nut arrangements are indicated in Fig. 5-4. Fig-

ure 5-4a shows the conventional movable nut arrangement, whereas

Fig. 5-46 indicates the movable screw arrangement. Figure 5-46 can

be used with a conventional motor, or if it is desired to save height in

the mechanism structure, hollow-shafted drive motors may be used.

Unless the shape of the control rod is round, some form of keyway is

usually provided to prevent the control rod from turning.

For discontinuous motions, magnetically operated screw jacks have

also been suggested. These drives are similar to the conventional auto-
Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
REACTOR CONTROL MECHANISMS

105

cable arrangement, but any of the other means may be used. The neces-

sary bearings and supports are not indicated in Fig. 5-5.

Synchronous Reluctance Motor Drives. For cases where extremely slow

rotary motion is required inside a pressure shell, special types of motors

have been developed.1 One such type is the so-called canned synchro-

nous reluctance motor. This device is inherently capable of operating

ELECTROMAGNET

NON MAGNETIC

PRESSURE SHELL

FIG. 5-5. Magnetic feedthrough for pressurized reactor drive system.

at very slow speeds, and the amount of gearing that must be used can

be kept to a minimum. For nonlubricated systems this is a distinct

advantage.

The operating principle of the reluctance motor is indicated in Fig. 5-6.

Here three-phase windings are shown, and we can first consider the case

where the windings are supplied from a d-c bus. The armature can be
Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

considered to be a solid piece of magnetic material, indicated as a two-

pole bar in Fig. 5-6. If the windings are fed as shown with winding A

and winding B placed across the d-c line and winding C being left open,

a fixed field configuration will be set up inside the motor and the bar will

align itself in a given position in this

field. If, now, by means of mechan-

ical switching, windings A and C are

fed from the line and winding B left

open, a different field configuration

will be set up and the bar armature

will swing around to a displacement of

120° from its former position. In the

event that some stepdown gearing is

involved in the subsequent drive mech-

anism, step rotary motion of this sort

may be tolerated.

By feeding full current through one phase winding and half current

through the other phases, 60° steps may be obtained. Figure 5-7 shows

a commutator switching arrangement of this sort. Motions of 60° in

FIG. 5-6. Three-phase synchronous re-

luctance motor connection.

FIG. 5-7. Stepped drive system for synchronous reluctance motor.

106

magnitude are feasible for many applications. It will be noted that the
Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

windings are energized in the following sequence:


Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the motor air gap there is a 0.02-in. Inconel can which protects the

stator windings and punchings from the water and also acts as the pres-

sure wall. The can is, of course, backed up to take the actual pressure

by the stator core. The leads of the stator windings are brought out

from the top of the frame through high-pressure Kovar glass terminal

insulators. The outside frame, which is made of stainless steel, acts as a

pressure backup again for the stator and prevents release of any contam-

inated water to the outside atmosphere in case a leak develops in the can.

The rotor and the shaft are made of a magnetic stainless steel. The

bearings, as indicated, are of conventional ball-bearing design, except

that they are made of special materials to operate at rated temperatures.

To illustrate the performance of the motor of Fig. 5-8, Table 5-1 presents

a summary of its design characteristics as given by the manufacturer.

TABLE 5-1

Design characteristic

State

Rating

Speed, rpm

Normal

High-speed operation

Maximum

Holding (continuous)

0-16

53
Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

1,800

Torque, Ib-ft

Running (intermittent)

Normal

0 15

Cooling-water flow gpm

Minimum

Normal

0.109

120

Cooling-water inlet temperature, °F

Maximum

Normal

130

Winding temperature, °C

Maximum (intermittent)

150

250

2,500

Pressure, psi

Linear Reluctance Motor. The problem of converting rotary motion

to linear motion inside a pressure shell is an extremely difficult one because

of the environmental conditions. An unusual electric motor has been

developed which creates linear motion directly without the need of going

through rotating motion first.2-3's Linear motors have been used for

some time for other applications such as catapult drives4 and X-ray

tables. They are very simply adapted to pressurized reactor work. To

obtain a crude feeling for the operation of the linear motor consider first

the standard reluctance motor just described. If this cylindrical motor

were figuratively to be sliced open at one axial place, flattened out, and

then rolled up again at right angles to the previous cylinder, a structure

similar to the cross section of Fig. 5-9 would be obtained. Here the field

coils are nothing but circular doughnuts around a long tube. The arma-

ture consists of a series of ringlike poles on a long bar. The winding slots

108
Generated on 2011-10-24 11:46 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

POSITION

INDICATOR

DEMAND

NEUTRON

SIGNAL

CONTROL

ROD

FIG. 5-10. Control system for linear motor drive.

'////////////- //,

HYDR

P\P€.

4ULIC

LINES

/A y////////////

HYDRAULIC

MOTOR

* GEARING

, REACTOR

/ PRESSURE

/ VESSEL

REA

^-CONTROL
Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ROD

:TOR

//////////////////////////////////,

110

FIG. 5-11. Pressurized system using internal hydraulic motor rod drive.
REACTOR CONTROL MECHANISMS

111

penetrated many times. Hydraulic and pneumatic drive systems are

available which are complete analogues to electrical driving systems.

In the higher horsepower ratings some advantage in size may be obtained

by using hydraulic actuators. Whether one uses electrical, pneumatic,

or hydraulic systems will depend upon the load, the accessibility of the

DEMAND

^CONTROL

ROD

REACTOR

Y/////////////////////////////////'/

FIG. 5-12. Internal hydraulic piston driving system.

components, the speed of scramming required, and to some extent the

past experience of the designer.

Figure 5-11 shows one arrangement for using a hydraulic motor com-

pletely inside a pressure shell. An arrangement of this sort is partic-

ularly advantageous if the hydraulic fluid can be the same fluid as the

coolant or moderator so that any leakage from the hydraulic motor does

not contaminate the system. The hydraulic motor may be of any con-

ventional type such as a vane or reaction type motor.


Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
REACTOR CONTROL MECHANISMS 113

leakage is permitted, radioactive coolant may escape and soon make the

premises uninhabitable. Consequently, the specifications for a high-

pressure seal could conceivably be as severe as "should not leak one drop

in a month."

Figure 5-13 indicates one of the forms that a rotating shaft seal might

take. A conventional motor is used to drive a shaft into the pressure

vessel. Leakage along the shaft is blocked by a series of packing glands.

The packing glands serve also to reduce the pressure from the inside to

the outside in gradual steps until ultimately the last gland is at atmos-

pheric pressure. Some leakage will still result through these glands, and

tap-offs are provided at various points between glands so that this leakage

may be bled off into a storage tank. Figure 5-13 indicates the final tap-

off at atmospheric pressure. From the storage tank another line and

pump are provided to reinsert any coolant leakage back into the primary

system. Similar seals for direct transmission of linear motion are also

available.

5-4. Scramming Mechanisms. Thus far some of the types of mecha-

nisms that can be used to provide the normal slow drives for control rods

have been indicated. As previously mentioned, most of these drives will

have some form of scramming mechanism built in, either directly at the

rod or indirectly in the control-drive feed. There are two distinct types

of scramming systems that must be considered, and which of these is

used depends upon whether the rod is inserted into the reactor vertically

or horizontally. When the rod is inserted vertically, gravity is almost


Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

always used to assist in the scramming process. When the rod is inserted

horizontally, some form of power drive is involved.

Direct Magnet Release. The simplest form of scram employing gravity

only is where the rod is connected directly to a mechanism by means of

an electromagnet. Breaking the current through the magnet causes the

rod to release by gravity. Figure 5-5 shows such a system for the pres-

surized type drive, and simpler systems can be devised for nonpressurized

reactors.

Latch Relays. The disadvantage of the direct-magnet release lies in

the fact that some designs require a large amount of magnet current

which is difficult to supply and also difficult to break quickly because of

the L/R time constant. A small magnet may be used to release a

mechanical latch and accomplish the same effect, usually with more

speed and with large decreases in driving power. One such scheme is

indicated in Fig. 5-14a. Here a large arm is held against a detent in the

control-rod shaft by means of the force created in a small solenoid coil.

When the current is disconnected from this coil, a spring forces the large

arm outward. Gravity again produces the actual scram.

It will be shown later that gravity scrams are sometimes inadequate,


Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
M5

REACTOR CONTROL MECHANISMS

tage is taken of the fact that the reluctance motor type of motor drive

described in Sec. 5-3 is a synchronous device and consequently will oper-

ate effectively at any line frequency supplied to it. Under normal con-

ditions the supply frequency might be a few cycles per second. However,

for scramming purposes, proper switching may be arranged so that for

inward motion only the 60-cycle line will be connected to the reluctance

motor. In this way the control-drive mechanism will operate at highly

accelerated rates.

Figure 5-156 indicates a system using two motors. One motor is the

normal driving motor connected to some form of mechanical speed

fa)

DIFFERENTIAL

GEARING

(b)

FIG. 5-15. Powered scram drives, (a) Two-frequency reluctance motor operation, (fa)

Two-motor system.

reducer driving differential gearing. Feeding the differential gearing

also is a high-speed induction motor which is capable of rotating in one

direction only. Again proper switching throws one motor or the other

onto the line.

5-5. Energy Storage Devices. All reactor control systems must pro-

vide for some form of energy storage in order to move the rods to their

shutdown position in the reactor in the event of power failure. The


Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

simplest and most direct energy storage means is gravity, when the

mechanism design permits its use. For reactors having vertical control

rods, gravity dropping of the rods in the event of power failure is very

practical. For reactors with horizontal rods, gravity may also be used,
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

but the system is more complicated. Any type of energy storage may be

used, and either potential or kinetic energy can be employed.

The type of storage used when gravity storage is not directly applicable

is usually hydraulic or electrical, and each specific system suggested has

a direct analogy in the other medium. For electrical systems, storage

batteries used with d-c motors and appropriate switching provide a good

combination. The hydraulic analogy to this system is the gravity

accumulator supplying hydraulic fluid to a hydraulic motor through

WEIGHT

PISTON

ACCUMULATOR

OPERATING

""" VALVE

HYDRAULIC

PUMP

HYDRAULIC

CONTROL

VALVES

DUMP

VALVE

(b)

FIG. 5-16. Hydraulic tank energy storage systems, (a) Electrical output, (b) Hydraulic

output.

appropriate valving. Hydraulic tanks or accumulators are used in some


Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

present-day reactors and may be connected to supply either electrical or

hydraulic power. Figure 5-16 indicates schematically the two types of

analogous designs. In this figure a weighted piston is used to supply

hydraulic fluid at the proper pressure to the generator or pump. These

elements then can ultimately drive control-rod motors. Another form

of accumulator is the spring-loaded accumulator in which the weight is

replaced by a spring or series of springs. The outputs are essentially the

same.

116

Figure 5-17 shows another way in which energy may be stored in

springs. In this figure the rod is connected through a clutch or differen-


REACTOR CONTROL MECHANISMS

117

tial arrangement to either of two drives. One of these drives is the

normal motor driving system, and the other is the scramming drive,

which is powered by a previously wound spring motor. The spring drive

may be connected automatically to the rod in the event of power failure.

This system is most suitable for use with small light rods.

0,.^..

SCRAM

TWO WAY

NORMAL

SPRING

SLOW

SPEED

MOTOR

DRIVE

SYSTEM

FIG. 5-17. Spring-wound-motor energy-storage system.

FLY WHEEL

ELECTRIC

MOTOR

ELECTRIC

GENERATOR

CONTROL
Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

SWITCHING

ROD

DRIVE

ELECTRIC

MOTOR

(a)

FLY WHEEL

ELECTRIC

MOTOR

HYDRAULIC

PUMP

(b)

CONTROL

VALVES

ROD

DRIVE

HYDRAULIC

MOTOR

FIG. 5-18.

output.

Flyw

heel

in'er

tia storage systems, (a) Electrical output, (b) Hydraulic

The above schemes are potential energy schemes. The Brookhaven

reactor employs the inertia of a flywheel in a kinetic energy system.

Figure 5-18 indicates two flywheel inertia schemes. Figure 5-18a is an

electrical system in which an electric motor is used to bring a flywheel up

to speed. On the same shaft is an electric generator which is normally

used to supply the power for a rod drive motor or scramming system.
118 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

If the charging motor power fails, the inertia of the flywheel will keep the

generator operating for several minutes after the power failure.

Figure 5-186 shows the hydraulic analogy of the same system. Both

of these inertia systems are such that there is a finite time after power

failure in which reactor shutdown may be accomplished. For example,

either the rods are inserted within 5 min after power failure or the fly-

wheel system will coast to a complete stop. The accumulator schemes

have a similar restriction if automatic operating valves actuated by power

failure are used. But if manual valving is employed, the gravity storage

is available at any time.

Energy storage may be provided from one source to supply several

rods, or each rod may have its own storage system. The gravity or

spring-loaded accumulator is more easily adapted for ganged storage,

whereas the flywheel system can be placed very simply on each rod drive.

Ganged storage is somewhat less desirable than individual storage in that

a failure in the system could affect not one rod but all of them. For a

given amount of energy storage, it can usually be shown that the flywheel

system occupies less space and has less weight than the accumulator.

The gravity system may be somewhat more reliable basically in that the

flywheel must operate continuously and ultimately may have bearing

trouble. The flywheel drive also will take more power from the supply

lines, but this is generally of small consequence. In any event, the choice

of one energy storage form over another will depend upon space and

weight requirements and upon whether ganged or individual energy sup-


Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ply is desired.

The above considerations are essentially for horizontal rod drives.

For a vertical rod drive system, simple or accelerated gravity drop upon

power failure is preferred.

5-6. Buffers. The use of aided power drives or accelerated gravity

scramming drives means that control rods are inserted into a reactor with

considerable force. After the rod has reached the end of its useful travel,

it must be stopped gently, as expediently as possible, so that neither the

reactor nor its structure will be damaged. Spjrings and dashpot arrange-

ments and conventional automobile-type^ shock absorbers have been used

for this purpose. FigureT5-19 indicates an elementary form of spring~ahd

dashpot arrangement. The principal concern in the use of these devices

is that they be mounted on such a place that bowing or bending of the

drive shaft mechanism is minimized.

5-7. Rod Position Indication. In most reactors it is desirable to know

where each rod is at all times in order to anticipate the state of criticality

of the reactor or to determine some corollary parameter such as the

amount of depletion or the amount of poisoning. Position indication is

no problem at all for nonpressurized systems in that conventional a-c


REACTOR CONTROL MECHANISMS

119

synchros or a-c and d-c potentiometers may be used. Proper gearing or

multiple units can provide any reasonable degree of accuracy of position

indication.

For pressurized systems this problem may become formidable. The

accurate transmission of position through a pressure shell wall is usually

accomplished by magnetic devices involving rather large air gaps. These

devices take the form of canned synchro generators or canned differential

ROD RELEASE

MAGNET

SPRING AND

DASHPOT

CONTROL ROD

FIG. 5-19. Spring and dashpot buffer arrangement.

transformers, or they may even be built in as an integral part of the drive

motor. Figure 5-8 indicates a drive motor having the rotor of a position

indicator built in as an integral part of the drive motor shaft. A multi-

pole pickup coil is used with this motor on the outside of the structure.

Figure 5-20 indicates a differential transformer type of direct linear

motion position indicator used with a linear motor. Here a change in

inductance caused by the position of a magnetic slug is a direct measure

of the position of the control rod. In many cases the indication through
Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
Generated on 2011-10-24 11:47 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
REACTOR CONTROL MECHANISMS

121

let us assume that we have an automatic control system of any of the

types indicated in Chap. 4 and that the reactor has no negative tempera-

ture coefficient. The curve of Fig. 5-21a indicates the shape of the tran-

sient response we would expect from our control system, with P being the

peak power excursion reached by the transient for an input reactivity step.

The control system removes the power transient by moving a rod in

such a manner as to return the net reactivity to zero. As it would be

impossible to move the rod in a stepwise motion, it is more reasonable

to assume that the reactivity change created by the rod would be of the

shape indicated in Fig. 5-216. If reactivity and rod position are linearly

POWER

LEVEL

REACTIVITY

CHANGE

OR

ROD

MOTION

(a)

TIME

- 1 CYCLE

FIG. 5-21. Control-rod motion during transient elimination.

related, the rod would start to move slowly, gather speed, and then slow

down as the desired end point was approached. For simplicity let us
Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

assume that this required motion is sinusoidal about the halfway point

and that the sine wave has a peak half amplitude of A. The time taken

by the transient to die out effectively may be considered one cycle of the

rod motion sine waves.7 This method of analysis is not strictly correct.

The method used in Chap. 4 of denning a natural frequency fn for a given

control loop is a better one. However, in practice the frequency of a

sine wave derived in this manner and /n are of the same order of magni-

tude and either will do for rough horsepower calculations.

Now assume that the driving motor horsepower requirements must be

such that this sinusoidal motion of a given frequency / and a given ampli-

tude A can be followed by the regulator rod. If

(5-1)

co = 2rr/
122 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

rod displacement

s = A sin uit (5-2)

rod velocity

v = Aw cos t (5-3)

and rod acceleration

a = —A co2 sin ait (5-4)

For sinusoidal motion, horsepower

hp = Fv (5-5)

where the force on the rod F = ma, and if the rod mass = m, in poundals,

then

mA*w* sin tat cos tat

hp =

550

when the appropriate constants are in feet and seconds. Then as

sin cot cos wt = Yt, sin 2tat, the peak horsepower required from the driv-

ing system to contain the transient is

_ , , ,_ _.

Peak hp = - (5-7)

To obtain an idea of the magnitude of the parameters involved, let us

try a numerical example. Assume the transient of Fig. 4-32, and arbi-

trarily state that a complete sine wave cycle would take 0.15 sec or

/ = 6.67 cycles/sec. Let us further assume that to correct for a 0.0035&

step we might have to move a 200-lb control rod 2 ft.


Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Then the maximum horsepower would be

_ 200 X &r3 X (6.67)3

Peakhp= --- 32 X 1,100 (5-8)

= 418 hp

which is a staggering amount for most servo motors. Apparently either

the constants of our example are impractical or we are asking for too much

by way of transient response. It can be seen from Eq. (5-7) that the

horsepower requirements depend most strongly on ta and A. If, in our

example, ta were decreased by a factor of 10 and A by a factor of 2, the

peak horsepower requirement would come down to approximately 0.1 hp,

a far more manageable number. It will be recalled from Sec. 4-12 that

the slowest possible control system should be chosen for safety reasons.

Apparently the slowest possible system should also be chosen in order to

obtain a drive of practical physical size.

The above argument has been presented on the basis of the reactor

having no temperature coefficient. It will be recalled that the effect of


REACTOR CONTROL MECHANISMS 123

the temperature coefficient was principally upon the front end of the tran-

sient, and the time for the disturbance to be returned to zero depended

upon the external control system. Therefore, by virtue of the way in

which the above horsepower requirements were obtained, the negative

temperature coefficient of the reactor does not affect the answer.

REFERENCES

1. Esselman, W. H., and W. H. Hamilton: Position Control in Sealed Systems,

"Proceedings of the 1953 Conference on Nuclear Engineering," University of

California Press, Berkeley, Calif., 1953.

2. Shaw, V. G., G. M. Anderson, R. G. Durnal, and F. J. Walcott: A Canned Linear

Reluctance Motor, WAPD-T-5, Westinghouse Atomic Power Division, Pittsburgh,

Pa.

3. Robinson, R. C., and W. E. McCown: Direct Linear Motion in Sealed Systems,

WAPD-T-39, Westinghouse Atomic Power Division, Pittsburgh, Pa.

4. Jones, M. F.: Launching Aircraft Electrically, Aviation, vol. 45, pp. 62-65, October,

1946.

5. Catalogue, Autoclave Engineers, Erie, Pa.

6. Catalogue, American Instrument Company, Silver Springs, Md.

7. Harrer, J. M.: Controlling a Power-producing Nuclear Reactor, Nucleonics, vol.

6, no. 3, p. 58, 1950.

8. Anderson, G. M.: Analysis of the Linear Reluctance Motor, WAPD-T-6, Westing-

house Atomic Power Division, Pittsburgh, Pa.

9. Harrer, J. M.: Control Rod Mechanisms, Nucleonics, vol. 13, no. 6, p. 48, June,

1955.
Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CHAPTER 6

NUCLEAR POWER PLANT CONTROL

Until now we have been discussing reactors from which no power is

being taken. The reactors of Chap. 4 might be test reactors whereby

the neutrons from the reactor are being used for experimental purposes

or to create radioisotopes. Throughout the discussion on reactor control

it has been presumed that although the reactor is giving off heat, no use

is being made of this heat. This situation corresponds in our d-c gen-

erator analogy to having a generator operating at no load. We expect,

of course, that if we load a d-c generator, we shall change its internal

characteristics. Consequently, it is natural to expect that as we load

up a reactor and attempt to obtain power from it, we shall also affect its

internal characteristics. In this chapter we shall examine the operation

and control of power reactors from which useful heat power is being

extracted.

6-1. Introduction. We might now look broadly at the mechanism of

the reaction of a load upon a reactor. For the sake of illustration an

elementary block diagram of a possible nuclear power plant is shown in

Fig. 6-1. Here the reactor is considered merely as a source of heat energy.

This heat is extracted by passing a coolant through the reactor. The

heat energy is then transferred via a heat exchanger to a turbine system,

and the output shaft of the turbine can be made to drive many types of

working apparatus. The primary coolant considerations are basic to

this discussion, and this coolant may be gas, water, or a liquid metal.

Two external paths exist for affecting the reactor. One is a local path
Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

through its own control-rod system, and we have examined in Chap. 4

the principles of automatic control by means of control rods. The other

path is the direct connection to the plant via the coolant system. Any

variation in the coolant system parameters such as temperature, pressure,

or flow will affect the reactivity of the reactor. Variations in the coolant

parameters do not directly influence reactivity unless the coolant is also

the moderator. It is the effect upon the moderator and fuel that usually

causes a reactivity change, and any external influence that changes mod-

erator characteristics can be expected to change reactivity.

124
NUCLEAR POWER PLANT CONTROL

125

The largest plant effect is usually one caused by a change in tempera-

ture. We have already seen that the temperature coefficient is an impor-

tant safety and control element. The temperature of the moderator and

the fuel can be affected by the internal heat of the reactor, by external

variations in the load, or by a combination of these two effects.

Another effect may be the pressure of the primary system. In power

plants where the coolant or moderator is either a gas or water, the primary

circulating system is usually pressurized to increase the plant efficiency.

The reactivity of a reactor can be a distinct function of primary system

pressure, and consequently, external changes in the pressurizing system,

as a function of the plant load or the pressurizer controls, can affect

reactivity.

FIG. 6-1. Elementary block diagram of nuclear power plant showing essential elements.

In order to isolate these two variables we may define two coefficients.

These are the temperature coefficient as the |d5fc/dT|con8tp and the pres-

sure coefficient as the \dSk/dp\.uuttT.

6-2. Description of Basic Elements of a Nuclear Power Plant. In

order to get a better feel for the large number of items that may affect

these two coefficients, we might look at a more complex block diagram

than the elementary one given in Fig. 6-1. Figure 6-2 is a block diagram

of a nuclear power plant indicating some of the more important auxiliary

systems that may be present. In an actual power plant hundreds of

auxiliary systems may be used, and each of these can conceivably affect
Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

reactor operation. We shall now briefly describe some of the more

important systems of the plant.

Nuclear Instrumentation System. The nuclear instrumentation system

is the multiplicity of apparatus that is used to monitor nuclear radiations

over the complete range of reactor operation. It consists usually of sev-

eral radiation-detection instruments and associated electronic circuitry

to indicate the reactor power level and the reactor period at any instant.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

This system can affect the reactor principally by giving erroneous read-

ings. A false value of n from the nuclear instrumentation creates a

wrong power demand error signal and thus causes control-rod motion.

The nuclear detectors are usually calibrated against thermal detectors,

and the calibration circuits are a permanent part of the system.

Temperature and Flow System. At power operating levels the power

output and all the characteristics of the plant are determined_by tem-

perature and now measurements. Measurements of these variables are

complicated by the presence oTTadioactive coolant with no leakage what-

ever of the coolant usually being permitted. Because this system fur-

I CONTROLS I

FIG. 6-2. Block diagram of a nuclear power plant indicating some of the important

auxiliary systems.

nishes the basic indication of reactor power output, it will obviously be

involved in the response of the secondary system output.

Reactor Shutoff and Scramming System. One control requirement

peculiar to the nuclear power plant is the instrumentation setup for

alarm, cutback, and scramming circuits. Alarm circuits are common in

the power mcttisliy, butrthe" pecuKaTliotion of "scram" is one limited

to the reactor field. In certain types of reactors the safety requirements

are such that it is necessary to employ these types of scramming devices.

However, from a practical point of view it is obvious that the plant must

be kept in an operable state as continuously as possible. The plant must

not be shut offevery time some minor control in a secondary loop maP
Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

functions. The" control designer's "problem is to limit "to~an absolute

126
NUCLEAR POWER PLANT CONTROL 127

minimum the number of controls which can cause a scram. For most

accidents and component failures it is usually possible to cut back the

reactor power level at a slower rate than scram rate or merely to ring an

alarm and have an operator manually make the necessary adjustment or

shutdown.

Pump or Blower Controls. The power output of the reactor depends

directly upon the rate of coolant flow. Consequently, this flow is also

involved in reactions upon the reactor. The coolant flow may be fixed,

continuously variable, or stepped in accordance with a demanded sched-

ule. As the power input to a pump varies with the cube of its output, it

can be seen that to operate the plant at low power outputs without cut-^f\'" .

ting back coolant flow usually results in a loss of efficiency. Pumping \_/

power can become greater than 5 percent of the plant power output if

precautions to change the flow are not taken. The pump controls there-

fore may become quite complex in that multiple units may be switched

on and off the line and variable flow schedules can be used in addition.

Pressurizer System. If the primary loop requires a pressurizer, an

interesting problem in reactor dynamics can occur. The pressurizer in

its elementary form might be a tank partially filled with coolant which,

in turn, is usually covered by a blanket of gas under a specific pressure.

This tank is connected to a high point in the primary coolant system via

a pipe. It can be seen that the coolant in the tank and in the pipe can

represent an inertia. The gas pressure above the coolant in the pres-

surizer acts as a spring."^ Consequently, u insufficient damping exists Tn


Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the connecting pipe, the system conditions are proper for a pressure oscil-

lation to occur between the pressurizer and thg_primary loop' This pres-

sure oscillation obviously creates a reactivity oscillation in the reactor

through the pressure coefficient. If this problem is not considered care-

fully, the time delays may be such that the initial pressure oscillation

can be reinforced by the reactor power oscillation. The pressurizer sys-

tem then becomes an auxiliary requiring most careful design attention

in that it can violently affect the basic performance of the reactor.

Secondary System. The secondary system of the nuclear power plant

resembles a conventional steam plant. All the necessary auxiliaries that

are used in modern steam practice will be used for the nuclear power

plant. The control problem for the nuclear plant becomes very similar

to the control problem of the steam plant. A great deal must be known

about the characteristics of the output load, however, before the reactions

back upon the reactor can be explored. In the case of an aircraft power

plant this information might be specified in terms of aircraft speed as a

function of turbine output. For a shipboard plant the information might

be supplied in the form of propeller revolutions per minute versus turbine

output.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

6-3. Steady-state Programming. The pattern that the temperatures,

pressures, and flows throughout the plant assume as a function of power

output is called the program. Figure 6-3 indicates again a basic plant,

PUMP

FIG. 6-3. Elementary block diagram of nuclear power plant indicating the symbols used

for temperature, pressure, and flow.

and the symbols that will be used are

Tc = inlet coolant temperature to the reactor

Th = outlet coolant temperature from the reactor

Tav = average coolant temperature = (Tc + 7\)/2

F = primary coolant flow

F, = steam flow

p = absolute steam pressure

T, = steam temperature

A distinct relationship exists between the temperatures and the pressures

of the primary and secondary sides of the plant. The difference between

the average temperature of the primary coolant in the boiler and the~

steam temperature is directly proportional to the power being transferred

from the primary coolant to the steam side.

P=

- T.)

(6-1)

where P is the total power and K is a proportionality factor which depends

on the heat-transfer characteristics of the boiler. The value of K varies


Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

slightly with coolant flow rate, thermal power, and boiler level but may

be assumed constant for a first approximation. The values of T^ and

T, at full power can be determined from saturation data from steam

tables when the steam pressure for 100 percent power is specified. As

zero output power is approached, the difference between T^ and T, con-

verges to zero and the value at which the temperatures converge may be

128
NUCLEAR POWER PLANT CONTROL

129

arbitrarily located on a temperature scale, assuming the presence of a

control system to enforce the selection. The point to be made is that

once the constants of the primary side of the plant are known, the second-

ary constants are automatically specified and vice versa. For example,

Eq. (6-1) states that if one desires a plant in which the average tempera-

ture of the primary coolant is constant with power level, then the tem-

perature program of the steam is automatically specified as dropping off

as the power increases. Conversely, if one desires a given program of

steam temperature as a function of power level, the average temperature

program of the primary coolant is

immutably fixed as long as we as-

sume a constant K.

Constant-T^ Program.1 One type

of desirable program for a nuclear

power plant is the program whereby

the average temperature of the pri-

mary coolant is constant regardless

of the power output. This type of

program is shown in Fig. 6-4 for_ |5

constant primary coolant flow. For

reactors having a negative temper-

ature coefficient this is the natural

program of the reactor and the one


Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

that requires the least amount of

external control.

POWER OUTPUT

FIG. 6-4. Variations in temperatures and

pressures as a function of power output

for constant-average-temperature pro-

gram with fixed coolant flow.

Let us examine the simple plant shown in Fig. 6-3, making the follow-

ing assumptions. The steady-state program desired is the one of Fig. 6-4

with constant-coolant flow. The reactor has a negative temperature

coefficient, and the plant is operating so that a steady output is being

supplied to the load. Now suppose that more output is required by

the load. This greater loading causes more heat to be extracted from

the heat exchanger, and for a shortjjeriod of time, the heat capacity of the

heat exchanger and coolant can usually supply the additional load.

However, this extra energy extracted from the system requires that The

temperature of the coolant into the reactor must drop. If the reactor

has a negatrve_temperature coefficient, the dropping of the inlet tempera-

ture causes a drop in the average temperature. anoTconsequently the reac-

tor will possess more reacfivTtyl If the reactor was initially in a critical

state, ft now temporarily "becomes supercritical. The output temperature

of the coolant rises, and more energy is then available from the reactor.

Finally, in the steady state the reactor returns to its critical condition

with the average coolant temperature the same as it was initially. It

will be noted that without any control mechanism whatever, the reactor
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

system has stabilized itself about a given average temperature and auto-

matically supplies a reasonable demand placed upon it. In other words,

if the plant programming calls for the T.v to be constant, no control-rod

motions or no external control means

are necessary to handle changes in

power demand.

Let us now assume that for some

reason other than the reactor control

system, it is desirable to establish a

different pattern of coolant tempera-

tures. A reasonable pattern might

be that shown in Fig. 6-5. This fig-

ure shows a simple system which

might be called for by some structural

condition desiring to keep the outlet^

coolant temperature fixed at a given

.figure 6-tJ shows

POWER OUTPUT

FIG. 6-5. Variations in temperatures and

pressures as a function of power output

for constant-outlet-temperature program

with fixed coolant flow.

maximum level.

a block diagram of a control system


Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

indicating how this condition might

be achieved. Let us assume the

same conditions in that our reactor possesses a negative temperature

coefficient and is operating critical at a given output level. Again, as

more load is required, the inlet temperature to the reactor would drop.

More reactivity would be inserted into trie reactor because of the negative

FIG. 6-6. Block diagram of control system for obtaining constant-outlet-temperature

program.

130

temperature coefficient. The outlet coolant temperature would tend to

rise,~but^iow we must insert a control on the outlet temperature that

measures th~e" temperature ^and then varies the cooTant pump speed in

such a~manner~as to hold this outlet temperature constant.


NUCLEAR POWER PLANT CONTROL

131

In this way it is very feasible to hold the steady-state outlet tempera-

ture constant as a function of load as indicated in Fig. 6-5. However,

it will be noted that the average coolant temperature must drop with

load, meaning that the reactor^because of its negative temperature coeffi-

cient^ would now attempt to operate in a supercritical condition. Con-

sequently, another control loop, usually_the^ automatic external control^

rod loop such as described in ChapT 4, must be added to the system to

extract the reactivity inserted by the changeTn average coolant tempera-

ture. It can be seen then that any program other than the constant-

average-temperature program will usually require an external' reactivity

control to be used when the steadv-statejeyel changes.

The advantages of the constant-average-temperature program are

important. If the effects of aging and poisoning could be ignored, no

external reactor control would be necessary during normal operation.

The reactor would simply follow throttle changes in a stable manner by

means of the negative temperature coefficient. Another significant

advantage of the constant-T.v program concerns the pressurizer. _If

r.v is constant, the volume of coolant in the primary loop'is~essehtially

constaTir^an^thej'equired size for the pressurizer is at a minimum.

However, there are also some ^disadvantages to this program. The

principal one is the large change in steam pressure over the power range,

as seen in Fig. 6-4. A wide pressure

range means larger and heavier steam,


Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

piping, automatic throttling devices^

special turbines, and boiler-feed pump

problems^ Furthermore, since the

effects of aging and poisoning cannot

be ignored, there always must be a

feedback control loop to hold T.v con-

stant even in a simple plant. This

control loop, however, can be very

slow and might be manual.

Constant-pressure Program. An-

other extreme in plant programming

is to hold the steam pressure con-

stant as a function of power level and

permit the primary system tempera-

tures to rise. This is the program of

Fig. 6-7. The constant-T.v program is the one preferred by the reactor.

The constant-pressure program is the one preferred by the secondary

steam system. The advantages and disadvantages of the two systems

are reversed. That is, the constant-pressure program permits optimum

design of the steam plant in that it eliminates unconventional problems

POWER OUTPUT

FIG. 6-7. Variations in temperatures and

pressures as a function of power output

for constant-pressure program with fixed

coolant flow.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

132

in automatic throttling devices, special turbines, and boiler-feedwater

pumps. On the other hand, there is a large reactivity change required

because of the T.v change, and in some cases serious primary coolant

pressurizer problems can arise because of the large pressurizer volume

needed.

Another disadvantage is that in going from one steady-state power

level to another, the plant must fight the tendency of the negative tem-

perature coefficient to hold T.v constant. Also a much larger amount of

520

500

4RO

^^

la)

ATOV=

4 psi

±±

< 540

U.

CO
Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

t! 520

ct

' 500

fr

H»

4801-

ATOV=46.3«F

Ap=0

I- 520

soo

20 40 60 80 100

LOAD, PER CENT POWER

FIG. 6-8. Numerical example of eon-

stant-r.v program, constant-p program,

and compromise program for hypo-

thetical plant operating around 500°F.

8k CHANGE

TOTAL ATOV

FIG. 6-9. Variations in system parameters

as a function of change in Tar.

control-rod motion would usually be needed in a constant-pressure plant.

This control-rod motion again might be made very slow.

Compromise Programs. We find ourselves, then, in a position where

there are two logical extremes to the programming, and on the surface

these two programs are incompatible. One program is favored by the

primary loop; the other by the secondary loop. In most practical plants

a compromise appears inevitable. There are two obvious methods of

making this compromise. The first of these is to set up a steady-state

program which is part way between these two extremes. That is, a

program might be chosen in which a moderate variation in !T.v and steam

O.
133

NUCLEAR POWER PLANT CONTROL

pressure would both be taken and no attempt made to hold either con-

stant. This sort of approach is shown in Fig. 6-8, using typical numbers

around 500°F for a hypothetical plant. Figure 6-8a indicates the

constant-average-temperature program, Fig. 6-86 a constant-steam-

pressure program, and Fig. 6-8c an attempted compromise. The prac-

tical numbers obtained indicate that a design can be made which,

although it is not optimum, could be satisfactory to advocates of both

primary and secondary loop optimization. A variation on these schemes

is, of course, possible by varying boiler level and consequently K. This

system is not considered in this discussion.

FIG. 6-10. Block diagram of plant and universal control system capable of setting up any

type of control program.

The method that is used at a given power level and at a given flow is to

plot a series of curves of the various parameters involved against a trial

ATav. A curve of this sort is shown in Fig. 6-9, and at any AT.v the

designer of the plant may barter Sk changes against temperature and

pressure variations.

Time Constant Compromise.2 The second type of compromise that

might be made is on a time-constant basis. As yet we have said nothing

about the transient-response consequences of steady-state programs, and

this will be discussed in some detail later. From an elementary point of

view, however, we can quickly see that the proper speed of response to

transient changes can ease programming compromises.


Generated on 2011-10-24 11:48 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Let us assume that we have an elementary plant and control system,

such as is given in Fig. 6-10, which is capable of setting up and demanding

any conceivable useful type of control program. To tie back to our pre-
134 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

vious case, let us assume that this system has been set up to demand the

constant-steam-pressure program of Fig. 6-7. We now make the assump-

tion that the demand signal no out of our universal control device into

the comparator is velocity limited. That is, let us assume by way of

illustration that irrespective of any instantaneous demand changes in F,,

p, T,, Tc, Th, or T.v, the TIO power demand signal cannot change from,

say, 3 to 100 percent full power in less than 1 min.

Now let us follow through some of the typical demands that might be

made on this system. Suppose we are operating at a low power level and

the throttle is opened slowly. Opening the throttle slpwly in times of

minutes causes the demand signal no to rise slowly and the nuclear loop

to reposition the control rods so as ultimately to match the constant-

pressure reference demand. Suppose now that the throttle is opened

suddenly, say instantaneously. No new signal comes into the compara-

tor instantaneously from the demand loop because of the velocity limit

we have imposed on n0. However, this quick transient-throttle change

is reflected through the boiler into the primary loop as a change in aver-

age temperature of the coolant in the reactor, and the negative tempera-

ture coefficient of the reactor supplies a change in reactivity to match

the load on a transient basis. In other words, for transient throttle

operation the plant behaves as though it were operating without controls,

stabilized only by the negative temperature coefficient, and for transients

of short duration it behaves completely as though it were operating as a

constant-average-temperature system. Ultimately a demand signal


Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

appears into the comparator and the entire system slowly changes over

to a constant-pressure program.

The system responds to a change in reactivity in the reactor in a similar

manner. If this reactivity change is a slow one, such as caused by aging

and poisoning, the n loop signal, which is likely to be also velocity limited

by virtue of the fact that the control rods are not permitted to change

reactivity very fast, will handle this change slowly and safely. If a fast

change in reactivity occurs, the negative temperature coefficient again

takes over and the system behaves as though no external controls were

present.

Apparently, then, a plant possessing a negative temperature coefficient

behaves as a constant-average-temperature system for transient operation

provided the control system which ultimately sets steady-state conditions

isslowT EffecTfvely, then, one can have the advantages of both constant-

average-temperature and constant-steam-pressure systems with proper

design.

Flow Considerations. As has been mentioned, the rate at which energy

is carried by the primary coolant from the reactor to the boiler must be

equaled by the power generated by the reactor. This is expressed by the


NUCLEAR POWER PLANT CONTROL

135

simple relationship

P = F(Th -

(6-2)

CONSTANT

FLOW

VARIABLE

FLOW

POWER OUTPUT

(O)

There may be reasons why Th must never exceed a certain value. Prob-

lems of stress, corrosion, and wear may all be involved. Also, as pre-

viously mentioned, it is inefficient to operate the primary loop at low power

output with the flow needed for full

power output. These facts mean

that some form of flow control will be

used in the plant, and this control

concerns the programming system.

Figure 6-11 indicates two types of

programming changes, as a result of

flow considerations, for a constant-

average-temperature system. In

Fig. 6-1 la, the flow is constant from

zero power to a given power level,


Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

and then the flow is increased linearly

with power output to full power.

In this way, Th and Tc are held con-

stant, but the power output continues

to rise because of the flow change.

Continuously variable speed control

for pumps and blowers is more com-

plicated than stepped control, and

Fig. 6-116 indicates a two-speed type

of control. Here again the flow is

constant until a given power level is

reached, and at this point the pump

or blower speed is increased by a

fixed amount, thus dropping the

temperature difference required for

a given output. Many variations

and combinations of these two types

of controls are, of course, possible.

6-4. Elementary Thermodynamics of the Basic Loop. We have

broadly examined the stability and transient response of a nuclear power

plant and have come to the conclusion that the basic plant, consisting of

a reactor, coolant, heat exchanger, and steam system, in itself possesses

regulating ability and consequently is a basic control loop. In order to

study this basic loop as a servomechanism, we must examine more care-

fully the elementary reactor thermodynamic structure. We shall not

be able to handle anything but the simplest approximations to an actual

POWER OUTPUT

FIG. 6-11. (o) Constant-Tjv program in

which flow is variable above a given

power output, (b) Constant-Tnr program

in which flow is changed stepwise at a

given output level.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

plant, but the approach to the determination of the stability for a prac-

tical plant will be indicated. The method we shall use will be to develop

the differential equations for the thermodynamics of the reactor, the

transport delays in the loop, the transfer of energy in the boiler, and the

use of this energy in the steam system. Once the basic form of the equa-

tions is determined, we can again change over to a transfer function

approach and examine the response of an elementary nuclear power plant

as a basic control loop. Figure 6-12 indicates the loop and the symbols

we shall be examining.

In order to simplify the situation we shall immediately make the

assumption that the coolant and the moderator of our reactor are one

BASIC

CONTROL

L00P

PUMP

BOILER FEED

WATER PUMP

FIG. 6-12. Elementary power plant as a basic control loop.

and the same. In this way we shall eliminate double heat transfer

mathematics. Many practical plants may be constructed using this

system with either gas or water as the moderator. For our thermo-

dynamics considerations we shall use both words interchangeably. It

is also implied throughout this discussion that unless otherwise specified,

the reactor has a negative temperature coefficient.


Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

In the core of a reactor the fission of uranium 235 nuclei liberates

energy. In addition, energy is liberated by the decay of fission products.

This total energy manifests itself in a rise in the temperature of the fuel.

If no energy is withdrawn from the fuel, the temperature will continue to

rise indefinitely as long as fission continues. In power reactors a tempera-

ture difference between the fuel and the surrounding coolant causes heat

to flow from the fuel to the coolant.

136

Reactor Thermal System. Figure 6-13 shows a typical section of a

simplified reactor core configuration. The fuel may be a mixture of

uranium and other metals in an alloy, or there may be cladding involved.

We make the simplifying assumption, however, that an over-all fuel

element possesses constant density and constant specific heat. In Fig.


Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
138 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

(6-4)

where the time constant n is defined as dfcftf/2K.

We have now replaced the partial derivative by the ordinary derivative,

as ff is a function of time only. Actually this process is crude, as the

spatial variations and local effects in a reactor are going to be smeared

when we consider that we shall ultimately feed back into our reactor a

reactivity depending on T^..

We can now set up the heat-transfer balance in a similar manner for

the heat flowing out of the fuel and being carried away by the coolant.

Again, we shall assume an integrated situation. This heat is the last

term of Eq. (6-3), and it is carried off into the flow of coolant as

2K(ff - T.v) = dm«cM<tm.d + ™™™ (Th - T.) (6-5)

at TO

where dm<1d = density of the moderator coolant

cmod = specific heat of the moderator coolant

tmod = thickness of moderator-coolant passage as indicated in

Fig. 6-13

TO = time for a unit volume of coolant to pass through the reactor

core

One of the temperatures involved may be eliminated if we return to

the previous symmetrical definition of

rri \ rp

r«=-^y- (6-6)

Introducing this factor, we can transform Eq. (6-5) into the form of
Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Eq. (6-4) giving

+ -2 + T = ft + Tc (6-7)

TO / at TO

where the time constant r2 = dmodCmodtmod/2K.

Transport Delays. With Eqs. (6-4) and (6-7) we have now related

the reactor heat output to Tc, Th, and T^. We now must transport this

heat to the boiler via a pipe, as indicated in Fig. 6-12. We shall assume

that all our heat transfer occurs at a point in the reactor and ignore the

time delays in transporting the average heat from one position in the

reactor to another.

We can now examine the temperatures involved in the transportation

of the heat from the reactor to the boiler. The temperatures can be

given as

Tn(0 = r»(< - T,) (6-8)


NUCLEAR POWER PLANT CONTROL '139

or the inlet coolant temperature to the boiler Tu is of the same form as

the reactor outlet temperature, only it attains a given value after a fixed

transport delay r3. We are assuming no heat loss in the pipe transporta-

tion system. This situation is a good approximation if our connecting

pipe is well insulated thermally.

Pure transport delays may be difficult to handle analytically or by

analogue methods, and for some purposes it may be desirable to approx-

imate the time delay by a differential equation form. We shall use both

forms later depending on the problem. To develop the approximate

equations we may rearrange terms and expand in a Taylor series

T»(0 = TK(t + r3) (6-9)

T ff\ -t dT"® J. T'2 d*Tbi^ j fK*n\

= Tbi(t) + r, —^- + ^T —%- + . . . (6-10)

For slow changes in temperature we can ignore all terms beyond the

second. Consequently, we may use the following first-order differential

equation to indicate the time delay between reactor outlet and boiler

inlet temperatures.

J/TT

T" + T» II? " T» " ^ - T°

In a similar manner we might describe the transport delay from the

boiler back to the inlet of the reactor either by

Tt(£) = T^t ~ T4) (6-12)

dT

or by Tc + r4 ^f = Tbo (6-13)
Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

where T^ is the outlet temperature of the boiler and r4 the transport delay

between the boiler outlet and the reactor inlet.

We shall define boiler average temperature similarly to the way in

which we defined reactor average temperature so that

fb = Tb° + Tbi (6-14)

tt

Therefore our approximate form for the reactor inlet temperature as a

function of boiler temperatures becomes

dT

Tc + Tt^T = 2Tb~ Tbi

Boiler Equations. We can now use the same type of logic and approx-

imations in order to write the heat balance equations for the boiler.
140 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

First, on the primary coolant side of the boiler we have

Rc^K - Tba) = KbAb(Tb - T.) (6-16)

or 2Rc(TK - fb) = KbAb(Tb - T.) (6-17)

where Rc = mass rate of coolant flow

KI = effective boiler heat-transfer coefficient

Ab = effective area of the boiler tubes

On the steam side of the boiler we may express the fact that the rate

of heat storage in the boiler metal, steam, and water is equal to the differ-

ence between the rate at which heat is transferred across the boiler tubes

and the rate at which it is delivered to the turbine. Or

flT

(Mmcm + Msca) ^ = KbAb(Tb - T.) - KaAp (6-18)

where Mm = mass of boiler tube metal

M, = mass of steam and water in boiler

cm = specific heat of boiler tube metal

c, = average specific heat of steam and water in boiler

Ka = a throttle constant

A = relative throttle opening (full open corresponds to A = 1)

p = steam pressure = p(T,), a function of steam temperature

Several simplifying assumptions are again made in obtaining Eqs.

(6-17) and (6-18). First, the heat transfer between the primary coolant

side and the steam side of the boiler is represented by means of a single

area and a single heat-transfer coefficient. Second, the thermal capacity

of the boiler metal is lumped with that of the steam and water. Actually
Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

there is a time delay associated with a transfer of heat between metal and

steam. Neglecting this time delay will change the shape of the initial

transient of output power encountered with thermal changes, but this

approximation has little or no effect on the basic kinetics of the over-all

loop. The power output has been taken to be proportional only to the

flow of steam through the throttle, assuming that the enthalpy of the

saturated steam is constant. The flow of steam is then taken as the prod-

uct of throttle opening and steam pressure, using a suitable proportional-

ity constant. The further assumption has been made that the time spent

by the coolant in passing through the boiler is negligible in comparison

with the time spent in the pipes and that the heat exchange in the boiler

is integrated and treated as though it occurred at one point. These

assumptions limit the accuracy of the descriptive equations. However,

the errors involved in these assumptions are usually less than the amount

of uncertainty in the engineering values of the coefficients used.

6-5. Transfer Function Representation of Basic Plant Components.

Having the basic plant performance in mind, we can now proceed to


NUCLEAR POWER PLANT CONTROL

141

develop the form of the transfer functions for each of the components in

the basic loop. We may first start with the reactor. It will be recalled

from Sec. 4-3 that the elementary reactor with a negative temperature

coefficient could be treated as a simple parallel system containing a reactor

transfer function and a temperature coefficient feedback path. The feed-

back path was fed a variation in neutron power, and from its output came

a reactivity change caused by a change in average coolant temperature.

In the case of a reactor connected to a power plant, the situation

becomes more complex. Figure 6-14 indicates that in addition to the

simple T.v feedback loop, there exists another loop containing both Tc

8k IN

(OUT OF PLANT)

Th

To

(IN FROM PLANT)

FIG. 6-14. Temperature coefficient feedback loop.

and TK. The situation is even more complicated because of the fact that

Tc and T'.v are not independent.

Reactor-thermal-system Transfer Function. We can now develop a

more sophisticated, but still approximate, transfer function for the reactor

thermal system. The equations of heat flow from the reactor to the

coolant were developed in Sec. 6-4 and are repeated below

r, + 2is
Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

TO

Tc

(6-19)

(6-20)

(6-21)

We may eliminate the average fuel temperature Tf from these equations,

change to Laplace-transform notation, and derive the following relation-

ship between inlet and outlet temperature:


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

!_» fa + i\ _ Tl~\ s - i j Tc(s) + yQ(s)

2 \T2 / 1 J

where

Factoring

and

7=

(6-22)

(6-23)

(6-24)

we arrive at the form of the transfer function equation

«)

(6-26)

The block diagram representation of this transfer equation is shown in

Fig. 6-15. Here the over-all reactor thermal transfer function artificially

Q(s)

T,(s)

Th(s)

FIG. 6-15. Block diagram servo representation of a simple reactor thermal system.

appears to be in two parts, an inlet part and an outlet part, with the

reactor heat feeding in between them. The generation of Tav is, of course,

a separate external problem and its transfer function will later be obtained

by combining Tc(s) and Th(s).

The method for refining the over-all thermal transfer function of a

reactor is now also apparent. Let us assume that a reactor is divided


Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

into a discrete number of series parts or sections, each section being

farther along in the coolant flow path. Now a transfer equation of iden-

tical form to Eq. (6-26) can be derived for each individual section of the

reactor. The assumption is again made that the average heat transfer

from fuel to coolant occurs at a point in each individual section. The

heat is then transported from section to section, and the final outlet tem-

perature is a summation from the heats of all the individual sections.

142
143

NUCLEAR POWER PLANT CONTROL

Figure 6-16 indicates a block diagram of a two-section reactor. Here

Qiji = 6272, and (?i2(s) represents the transport delay between the first

section and the second. As many sections as desired may be added,

depending upon the degree of refinement needed in the design of a control

system for a particular plant. The tediousness of the calculations rapidly

exceeds the additional utility of the more accurate results when too many

sections are used.

n(s)

Q,(s)

02(s)

GM(s)

G,2 (s)

Gc2(s)

Th(s)

FIG. 6-16. Block diagram servo representation of a two-section reactor thermal system.

Transport Systems Transfer Functions. There are two principal delays

in the primary loop of Fig. 6-12, and both of these have transfer functions

of essentially the same form. First, it is necessary to transfer the heat

from the reactor to the boiler, and second, after the heat exchange process

has occurred, the coolant must be transferred back from the boiler to the

reactor.

The equation denoting the transport delay from reactor to boiler is

Eq. (6-8)
Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

= Th(t - T.)

and from boiler to reactor Eq. (6-12)

rff\ rp (t _ \

c\l) — J- bo\v — T4y

The transforms of these equations become

rw(s) = Th(s)e~"'

IT (o\ — T, (<>)p-'ri

J. cv«Jy J. oo \« /C

(6-27)

(6-28)

The transfer functions then become the output divided by the input or

ifeS = .-"' (6-29)

(6-30)

Tc(s)

This is an interesting form of transfer function in that it will be recog-

nized that there is no attenuation involved, only phase shift. The

amount of phase lag in our elementary system is directly proportional


144 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

to T3 + T4. Physically we have insulated our pipes perfectly and have

lost no heat in getting from the reactor outlet to the boiler inlet and fro.n

the boiler outlet to the reactor inlet. The phase lag increases linearly

with the length of the pipes.

The approximate differential equations expressing the equivalent delays

are Eqs. (6-11) and (6-13), and the approximate transfer functions are at

once recognizable as

(6-31)

Th(s) I + r3s

and -ff. 7 r -~ ^ :

This form is the so-called resistance-capacitance time delay and has a

maximum of 90° phase lag as a function of frequency rather than the

unlimited phase lag of the exact delay function.

Boiler Transfer Functions. We can develop the boiler transfep function

in a manner completely analogous to the way we determined the transfer

functions for the reactor thermal system.

The heat-flow equations for the boiler were Eqs. (6-16), (6-18), and

(6-14). They are repeated below.

R.(TH - Tbo) = KbA,,(fb - T,) (6-33)

dT

(Mmcm + M,c.) 2±i = KbAb(Tb - T,) - KaAp (6-34)

Tb = Tbo + Tti (6-35)

Steam pressure p is a function of steam temperature T, through the

saturation relationship. Eliminating T, we may derive the following


Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Laplace-transform equation in which the variable Tbl(s) and T&<,(s) now

represent small deviations from their steady-state values at the power

level at which the constants listed below are evaluated:

„ TM , ,

= K-B i — i - (O-dO;

1 + rtos

+i+

where KB = - ( bb - - - (6-37)

K.AB \_

KbAb V

00. s^s I "L m^m ,„ no\

TU = F —— =i (o-3o)

•vr ji I i i "^^ a-*i -*^ i ^-

Kb b [ ~K^ + (2Re/KbAt) -
145

Tbo =

B=

NUCLEAR POWER PLANT CONTROL

Msc, + Mmcm

KaAB

dTssat

(2Rc/KbAb)

at corresponding value of Ts

(6-39)

(6-40)

The transfer function of the boiler is then a distinct function of the

power level at which the output system operates. Equation (6-37)

indicates the variation in the gain term as a function of level. The

variable corresponding to power level is A, the throttle opening. It is of

interest to note that when A = 0, the throttle closed completely, KB = 1.

In other words, at zero power level there is no attenuation in the boiler.

For A > 0 some attenuation will be provided by the boiler.

1.0

O.fl

. 0.6

J0.4

0.2
Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

0.1

K0AB

100

FIG. 6-17. Boiler gain as a function of throttle opening.

Figure 6-17 shows boiler attenuation as a function of KaAB/KbAb,

which is related, as indicated, to the throttle opening or load. The curve

is presented for a given boiler with heat transfer characteristics such that

2Rc/KbAb = 4. It can easily be seen from this illustration that the boiler

does not provide much fixed attenuation even for the heaviest loads.

Equation (6-37) indicates that the maximum attenuation at full load

cannot be more than

(2Rc/KbAb) - 1

(2Rc/KbAb)

(6-41)

6-6. General Recapitulation of the Dynamic Performance of the

Basic Plant. The general dynamic requirements of the plant are con-

cerned with the stability of the complete system and the rate and mag-

nitude of changes in output power expected. The requirement of stabil-

ity suggests application of the familiar stability criteria of servo theory.

The required speed of response depends on the particular application.


146 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

For example, at a central station power installation it may be satisfactory

if power level can be changed only slowly, whereas a plant used to propel

a vehicle might be required to change power rapidly.

The ability of the system to handle changes in load is affected by the

bandwidth of the control system and by the thermal capacity of the pri-

mary coolant. The coolant acts as a reservoir of energy which can absorb

differences between reactor power and boiler power occasioned by rela-

tively fast changes in output demand. Thus reactor power may be per-

mitted to lag behind the power delivered to the load, thereby relieving

somewhat the speed of reactor control required. However, unbalances

in power result in changing coolant temperatures, and a transient is not

over until the temperatures have been reset to the values specified by the

steady-state program. Hence the required speed of the control system

depends on the magnitude and duration of coolant temperature excur-

sions which can be tolerated.

The familiar frequency-response method of stability analysis is applica-

ble for studying the natural stability of the plant and for synthesis of a

control system. The procedure is to compute the response of plant

variables to sinusoidal deviations of reactor power, about an average

value. The plant variables of interest are those which are fed back

through the inherent feedback paths of the plant and those which are

being considered for feedback through automatic control circuits. The

analysis is usually accomplished for a fixed setting of the steam throttle,

corresponding to a constant average output power. Because of non-


Generated on 2011-10-24 11:49 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

linearities which inevitably arise, the frequency-response type of analysis

is valid only for amplitudes of power excursions that are small compared

with the average value. The analysis may be repeated for a sequence of

values of throttle setting in order to cover adequately a given power

range of interest.

The source of power, the reactor, responds to changes in reactivity.

The reactivity Sk depends on the ratio of neutron-production rate to the

rate of loss, which in turn depends on the geometry of the reactor and on

variables which affect the absorption and leakage of neutrons. These

variables are effectively the temperature and pressure of the coolant, as

described in Sec. 6-1, and the concentration of neutron-absorbing poisons.

Since reactivity depends on plant variables, which in turn depend on

reactor power, the plant contains inherent feedback, as shown in Fig. 6-18.

The dependence of reactivity on the plant variables is expressed in the

form of coefficients of reactivity, assuming linearity. The dependence

of the plant variables on reactor power is determined from the frequency-

response calculations.

In the following discussion we shall assume that the inherent feedback

due to changes in poison concentration has a negligible effect on stability

over the frequency range of interest when compared with temperature


NUCLEAR POWER PLANT CONTROL

147

and pressure feedback. This assumption is justified either if the changes

in poison concentration are small in magnitude or if the poisoning changes

occur slowly compared with the temperature and pressure changes.

REACTIVITY

KTC.KP AND KX

DENOTE

COEFFICIENTS

OF REACTIVITY

CONCENTRATION

FIG. 6-18. Inherent feedback loops of a nuclear power plant.

TEMP. COEF.

REACTIVITY

FEEDBACK

I LOOP

/__SISJE^_J I

v/

\ TEMPERATURE

FEEDBACK LOOP

FIG. 6-19. Basic plant control loop showing temperature coefficient reactivity feedback

loop and temperature feedback loop.

As we have already determined the frequency response of the reactor

in Chap. 3, we must now find the response of average coolant tempera-

ture and pressure to oscillations in reactor power. This step is accom-


Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

plished by straightforward computation in an elementary plant by using

the set of differential equations or transfer functions just derived in

Secs. 6-4 and 6-5.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

6-7. Temperature Feedback Loop Analysis. Figure 6-19 now shows a

block diagram of the basic plant control loop that will be analyzed and

whose component transfer functions have been developed in Sec. 6-5.

Our basic control loop consists essentially of two loops: first, the tem-

perature coefficient reactivity feedback loop involving the reactor and,

second, a plant temperature feedback loop. The temperature feedback

loop is the new portion of our study and is most important to stability

considerations. First a quick look at this loop alone shows that it is

regenerative; that is, an increase in temperature in an element of coolant

at Th is fed around through the boiler and back into the reactor as an

FIG. 6-20. Servo block diagram of temperature feedback loop.

increase in temperature at Tc. The only fixed attenuation in the system

is at the boiler, and at low power levels, as just shown, there is only a very

small amount of fixed attenuation present. The heat-transfer lags in the

reactor thermal system and in the boiler, however, contain some attenua-

tion at high frequencies.

The temperature feedback loop gain figures very prominently in all

attempts to relate neutron-level changes to any temperature change in

the plant. Figure 6-20 shows the temperature feedback loop with all

of the transfer functions just derived. Suppose it is desired to relate the

small signal sinusoidal response of the neutron level to a temperature T

in the loop. Then, if the temperature feedback loop gain is denoted by

148

f An aside is necessary here concerning notation. The transfer functions that we


Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

have derived and have been using throughout this text have been in Laplace-transform
NUCLEAR POWER PLANT CONTROL U9

T(s) _ open-loop gain from n to T

n(s) 1 - KLGL(s)

(6-42)

The minus sign indicates that the feedback is regenerative. For example,

variations in T\ as a function of variations in n become from Fig. 6-20

^>

n(a) l-KLGL(s)

The term 1/[1 — KLGL(s)] appears very prominently in all stability

calculations and will be used directly to determine the over-all tempera-

ture coefficient feedback. In addition, it will be needed later when con-

trols are tied on the basic loop.

Consequently, it is now pertinent, in order to get a feeling for the

properties of 1/[1 — KiX?z,(s)], to look at an example. Let us therefore

compute 1/[1 — KLGL(s')] for a fictitious reactor plant, using simple num-

bers. Referring to Fig. 6-20, the following constants will be chosen for

the inlet portion of the reactor thermal transfer function. Let

£ - 4 g - 3

For the outlet portion of the reactor thermal transfer function let

[i + («/«»)][! + (»/«*•)] = [i+j(//8)][i+y(//2)i (6"45)

where

For the boiler constants we also arbitrarily choose

KB \ j"â„¢8 = 0.794 \

1 + TboS I

Where ™ = 2.X 0.15 Tb° = 20^01


Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

and KB is chosen as

0.794 = -2db

form as functions of a complex variable s. Strictly speaking, in using these transfer

functions, we are discussing transfer functions for sinusoidal signals of frequency /.

We therefore should be substituting for s, jZ-n-f or ju. Actually, the symbol s has no

meaning where we are performing nonlinear operations and is used in these following

sections only to simplify notation. For all the operations involving frequency ju

is implied.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

150

12

a4

uf

§

UJ-4

!r-8

-12

-16

Hi-//

^U \ , . \....

COMBINED -

AMPLITUDE

- FUNCTION -

0.001

1OO

0.01 0.1 1 10

FREQUENCY, CYCLES PER SECOND


Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

FIG. 6-21. Relative amplitude response of transfer function of temperature feedback loop

Ki.GL(s) showing the individual section responses.

100

50

UJ

t -50

^-100

a.

-150

-200

SHIFT WITHOUT

TRANSPORT DELAYS

0.001

100

0.01 0.1 1 10

FREQUENCY, CYCLES PER SECOND

CO

FIG. 6-22. Phase shift of transfer function of temperature feedback loop Kr.Gds) showing

individual phase shifts but omitting linear phase shift of transport delay.
NUCLEAR POWER PLANT CONTROL 151

The coolant transport time delay constants are chosen equal to be

TS = 5 sec TI — 5 sec

Coolant flow is assumed to be constant at one fixed value.

Using these constants, Fig. 6-21 shows the form of the relative ampli-

tude response of KLGL(s). The attenuation constant KB is not indicated

in this figure, as it serves only to move the combined amplitude function

down on the plot. It will be noted from this figure that as the frequency

increases, the relative amplitude response does not go to zero aa would be

expected for most physical systems. In this case the amplitude levels

off to a finite value as a function of frequency.

0.001

0.01

0.1 1

FREQUENCY, CYCLES PER SECOND

FIG. 6-23. Transfer function of 1/[1 — KiGz,(s)] without mixing.

100

Figure 6-22 indicates the phase shift in the system without the trans-

port delays of the pipe. It can be seen that the phase of this remainder

of the system settles out at 180° phase shift at high frequencies. The

transport delay phase shift that occurs is linear with frequency and will

add directly onto the phase shift of Fig. 6-22.

Figure 6-23 shows a plot of the amplitude and phase of 1/[1 — KLGL(S)]

for our example. The cyclic nature of the loop transfer function is due

to the linear time delays causing the phase angle of KI,GL(S) to increase
Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

steadily with frequency. The peaks and valleys in the plot of the ampli-

tude of 1/[1 — KL,GL(S)] occur at frequency intervals which are roughly

multiples of l/2rt, where n is the total time delay taken by the coolant in

going around the primary loop. Physically speaking, the peaks and

valleys occur at frequencies at which the temperature signal carried

around the loop either reinforces or opposes the temperature signals intro-
Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
NUCLEAR POWER PLANT CONTROL

153

TEMPERATURE T0 •

VOLUME

of Eqs. (6-13) and (6-15) might therefore really be a better representation

of the two pipe delays than the pure time delay formula we have been

using. Regardless of the pipe delays, however, we can certainly assume

some mixing at the inlet to the reactor and the boiler and most likely

also some mixing at the outlets of these

components.

Mixing Transfer Functions. To de-

velop a transfer function for these

mixing effects let us consider the situ-

ation of Fig. 6-25. We shall continue

to use the pure time delay representa-

tion for our pipes and assume that

the mixing takes place in the inlet of

the reactor or boiler. In Fig. 6-25 the

large volume indicated may be the

coolant in a bell or header of the reac-

tor or boiler. We shall also assume

that substantially complete mixing oc-

curs within this volume.

Let Fi = Fz = F = rate of flow of coolant in cubic feet per second and

consider an interval of time At which is very small and can be made to


Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

approach zero. The input quantity of coolant then becomes F At.

At the end of At sec, the following volume and temperature conditions

exist: There are F At cu ft of coolant at temperature T7,; there are

V — F At cu ft of coolant at temperature TV These two volumes are

mixed completely to give V cu ft of coolant at temperature T0 + AT0, or

FIG. 6-25. Mixing volume representa-

tion.

or

F(?'0 + A7Y) = F At

AT,

At

"V*

i + (V - F At)T0

F_

T0

Letting A To and At approach zero,

dT0 , F „. F

In Laplace notation

where the initial conditions are zero, and

TVs) _ F/V

(6-47)

(6-48)

(6-49)

(6-50)

(6-51)

Tt(8) s + (F/V) 1 + (V/F)s

As V has the dimensions of cubic feet and F is in cubic feet per second,
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

FIG. 6-26. (a) Symbolic representation of mixing in inlet and outlet of reactor and boiler,

(b) Block diagram of temperature feedback loop indicating placement of mixing transfer

functions.

'bo

BOILER

MIXING

FIG. 6-27. Transfer function representation of a temperature feedback loop with two

mixing terms.

V/F has the dimensions of time and consequently may be considered as a

conventional time constant, or

r.(«) _ i

IXs) 1 + rs

(6-52)

\S4

Temperature Feedback Loop Containing Mixing. From a practical

point of view we can now insert this type of mixing transfer function into

the temperature feedback loop. Figure 6-26 shows one way this might
Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
NUCLEAR POWER PLANT CONTROL

be accomplished. Of course, what we are doing is inserting some high-

frequency attenuation in the open-loop gain in order to give our plant

more practical characteristics.

-4

-^^n.

x\^

1 1 1 1 1 III

. -i — r T-[ i in

i .,.,, nn

- """^

\\

\\

g -12

~100 at

cc

_J
Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

-150 §

LJ

1-20

i \-r- PH

«E SHIFT

WITHOUT

PORT DELAYS

-200 1.

bj

AMPLITUDE—

\ \ TRANS

\\

-24

\\

-28

i1ii11

\\

-300

..iiiiii

! 1 1 LM

-35O

0.001

100

0.01 0.1 1 10

FREQUENCY, CYCLES PER SECOND

FIG. 6-28. Transfer function example of temperature feedback open-loop gain KtG/,(s)

with mixing.

5n

— R^

1 — 1 1 1 1 1 11

1 — r-mnr

0 10

i5

^V

Ke= 0.794 /

\/

A/W-

3 °
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

temperature may be smeared promptly and not be permitted to cause

rapid changes in reactivity through the temperature coefficient.

We can now return to our previous example, using the loop block

diagram of Fig. 6-27, with the constants assigned from Eqs. (6-44), (6-45),

and (6-46), and denote simple values to r6 = T6 = (1/ir) sec (break fre-

quency 0.5 cycle/sec). We can again carry through the analysis for

1/[1 — KLGL(S)], first obtaining the amplitude and phase response of

KL.GL(S) shown in Fig. 6-28. Now the open-loop gain decreases satis-

factorily toward zero with frequency. Figure 6-29 shows the amplitude

and phase response of 1/[1 — KI.GL(S)] for our simple example, and now

the periodic oscillation has been damped out to where the function is

plausible.

10

^' '""

^\

iii11111

1 ,11

Iiiiii!t

iii.

\/

A AAAn,

POWER LEVEL P
Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

I ~5

2-io

*J

VZ

\JV

FLOW

I 15

i 10

15

-5

-^^

^\

V_X v

/\A

POWER LEVEL P/2

FLOW F/2

-10

0.(

)01 0.01 0.1 1 10 1C

1'!11111

tiiti11

FREQUENCY, CYCLES PER SECOND

FIG. 6-30. Amplitude of 1/[1 — Ki,Gz,(s)] as the flow and power level are changed.

6-9. Flow Changes. In the previous sections we have been consider-

ing the case where the flow of coolant has been fixed at a given value and

the power output of the plant has also been fixed. Let us assume that

it is desired to change the power output of the plant by effectively chang-

ing the coolant flow. For illustrative purposes let us cut the flow in half,

thus changing the power output by a factor of 2. If one were to examine

the magnitude and phase of 1/[1 — KLGL(S}] of our simple example, it

can be seen that if the frequency scale values of both the magnitude and

phase were divided by 2, the resulting curves would very closely approx-

imate the transfer functions of the new conditions.

156

Figure 6-30 illustrates the change in amplitude of this function for a

given power and a given flow as against one-half of that power and one-

half of the initial flow. It can be seen that the curves are similar, as it is
NUCLEAR POWER PLANT CONTROL 157

apparent that all of the linear time delays in the primary coolant loop are

inversely proportional to the coolant flow rate. The gain constant KL of

the loop is roughly equal to that of the boiler gain KB, which from Fig.

6-18 does not change very much for a power output change of 2 to 1.

Therefore, changing the flow through the primary loop by a limited

amount changes for the most part the frequency scale only of the transfer

function. The general form of 1/[1 — KLGL(s)] remains the same.

It should be emphasized again that this analysis is proper only for fixed

fuel reactors. Other types of reactors, such as homogeneous reactors or

boiling reactors, can be handled with the same general type of approach,

but here reactivity would also be a function of flow.

6-10. Analysis for Multiple-secHon Reactor. Figure 6-16 indicated

the form of the reactor thermal transfer function when the reactor was

treated in sections. The analysis we have been carrying through up to

the present has been for a simple one-section reactor. Before we proceed

with the study of the reactivity feedback loop, it is of interest to point out

the method of analysis that would be used for a more complicated reactor

system. In order to get a more accurate representation of the reactor

thermal system, the reactor is split up into sections. Figure 6-31 shows

a block diagram of a temperature feedback loop in which the reactor is

broken up into four series sections. The other elements in the loop are

the same as indicated in Fig. 6-27. The neutron flux from the reactor

may be assumed to produce equal heat in each of the four sections of the

reactor, that is, Qi = Q2 = Q3 = Qi. Or if it is desired to taper the heat


Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

input into the reactor representation in an effort to take care of unsym-

metrical flux distribution, these quantities may have different values.

We can now use the principle of linear superposition in the following

manner. Let us assume that Qi is finite and Q2, Qa, and Qt = 0. The

open-loop transfer function between n(s) and any temperature in the

loop T(s) is simply the product of all the transfer functions between

n(s) and T(s) in the direction of signal flow. As indicated previously, the

feedback loop transfer function KLGL(s) is the product of all the series

transfer functions in the primary coolant loop. Then

n(s)

a loop, Q,.Q,,Q,=0

Q!,Qj.Qi=0

Similarly if Q2 is finite and Qi, Q3, Qi = 0,

T(s)

n(s)

[T(s)/n(s)l

.pen loop. Qi,(l,,Qf= 0

(6-53)

(6-54)

Corresponding expressions are obtained for finite Q3 and Qt. Therefore,

assuming we are dealing with linear transfer functions, we may apply the
6
O -I

158
Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
NUCLEAR POWER PLANT CONTROL

theory of superposition and find that

\S9

n(s)

T(s)

n(s)

Q-.Qt.Q4-*0

T(s)

n(s)

T(s)

T(s) 1 [

n(s) Ql,Ql,Qf,0 1 - KLGL(s) L

T(s)

-)_ ^' _i_

peoloop. O1.el,«4=0 ^(S) open loop. «!,Q,,Qi—0

T(s)

n(s)

T(s)

n(s) open loop. Q,,Qt,Q(=0

T(s)

open loop<

'. Ol.O,.Ot-oJ

(6-55)

It is evident that an analysis of this sort can become quite tedious in

order to obtain great accuracy. It is apparent also that if this type of


Generated on 2011-10-24 11:50 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

analysis is carried into too many sections, distributed reactor representa-

tion should be developed. Some methods for obtaining a distributed

reactor representation have been worked out in the classified literature

but are beyond the scope of this book.

6-11. Application and Limitations of Temperature Feedback Loop

Transfer Functions. As the relationship of all the temperatures in the

temperature feedback loop to the neutron-level variations can now be

derived, we can return our attention to the remainder of the basic con-

trol loop (see Fig. 6-19). This is the portion of the loop whereby the

average temperature via the negative temperature coefficient changes

the reactivity of the reactor. In order to obtain reactivity we note that

Sk = (rC)(Tar - To)

(6-56)

and To

where (TC) is the temperature coefficient of reactivity

is a reference temperature.

It is apparent that we must derive an expression for Tn(s)/n(s), where

s again is to represent j'co and the entire operation is for small signal varia-

tions about given values of 77.v and n.

Using the simple example of the block diagram of Fig. 6-27

TVs)

n(s)

Gh(s)

and

and

Therefore 2

or

L - KLGL(s)

T.(s) Th(s)KLGL(s)

n(s) ' G.(s)

2r.v(«) = Th(s) + Tc(s)

n(s)

- KLGL(s)

n(s)

'

'

Gc(s)

(6-57)

(6-58)

(6-59)

(6-60)

(6-61)

(6-62)

ITVs)

2 n(s)
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Tav may be developed in a similar manner for more complex divided

reactor thermal systems by the superposition technique developed in

Sec. 6-10.

10

— J^J , ,.

^\

X. / 1 /I AA A (.

II!..,

iii.

I.,!.

1IiI1

•o 0

l\T!

\^ j M

.AMPLITUDE

UJ -10

50
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

|-20

<

/\

/\

/An

/ at

-30

O.C

^\^_

101 0.01 0.1 1 10 1C

'^

X-- PHASE

v, ,

-100

-150

)0

FREQUENCY, CYCLES PER SECOND

FIG. 6-32. Transfer function example Th(s)/n(s), KB = 0.794.

1O

iiii11M

I1I1111

1 J i 1 1 11 1

1 1 t .1 1 i i 1

"S^^\

^o

-50

Ct

'. -10

a?

g -2°

vS /

/In
NUCLEAR POWER PLANT CONTROL

161

frequency in the same manner as Gh(s). The function Tc(s)/n(s) can

then be obtained from Eq. (6-58) and is shown in Fig. 6-33. Here the

amplitude response as a function of frequency falls off much more rapidly

at high frequency because of the additional boiler and mixing high-

frequency dropoffs. In other words, as the coolant progresses through

the loop, high-frequency variations in coolant temperature become

more and more attenuated. Consequently, the high-frequency response

of r.v(s)/n(s), shown in Fig. 6-34, closely resembles the response of

Th(s)/n(s) at high frequency.

We might now examine the over-all situation, again restating the pur-

poses and assumptions we have made in dealing with the temperature

10

|^' ' '""

^xT

1iil

n,

JU
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

/I/W-

ii.111

i i in

JD

"° 10

\/

\r

^XsX' AMPLIT

JDE

CT^S

\y

uj -20

i-

_j

s 30

<

-40

^\

^x

I\

~~i

tr

I «

fln

1 mn.

yll K

5,^ PHASE

"X^\

50

-50
162 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

point in the reactor is carried on to another point by the primary coolant.

There the signal may again reinforce or oppose the temperature signal of

the initial point, depending upon the phase shift acquired in passing from

one point to another. The linear time delays in the primary coolant

loop may cause the above-mentioned reinforcement and opposition

action to occur repeatedly at regular intervals of frequency with respect

to each section input of the multiple reactor.

With proper interpretation the transfer functions just derived in Figs.

6-32, 6-33, and 6-34 can provide valuable insight into the behavior of the

nuclear power plant. In the first place, they may be helpful in obtaining

a "feel" for the transient response of the nuclear power plant to external

system disturbances without any control system. Second, Nyquist plots

of this data provide a measure of the degree to which the bare reactor is

stabilized by the negative temperature coefficient of reactivity. Third,

the transfer functions provide basic data upon which to design a satis-

factory control scheme. Finally, they may help to point out adjustments

and modifications which may improve the dynamic performance of a

contemplated power plant. In order to interpret these transfer functions

correctly, constant reference must be made to the physical system and

to the conditions and assumptions upon which these transfer functions

are based.

Primarily, it must be remembered that the transfer functions or plots

of frequency which have been presented in this analysis have been based

on a simple example. Two basic conditions were involved: First, the


Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

throttle opening was kept constant, and second, the primary coolant flow

rate was constant. The results obtained are confined solely to relating

the response of the plant to reactivity or reactor power disturbances.

The response to changes in throttle setting or coolant flow rate have not

been specifically considered here.

These additional responses may be developed in a similar manner to

those just presented. It must be emphasized again that the frequency

response characteristics have been calculated for very small sinusoidal

variations in the power plant variables about a specified steady-state

condition for the entire plant. Predictions of the transient behavior of

the plant based on the frequency response characteristics must therefore

generally be confined to small transients about the steady-state condi-

tions specified.

6-12. Temperature Coefficient Reactivity Feedback Loop. In most

present-day reactors the temperature coefficient of reactivity at constant

pressure is negative and the net effect is to make the reactor self-regu-

lating. This negative temperature coefficient is, in servo terminology,

degenerative and has a stabilizing influence on an unstable physical sys-

tem. The actual form of the negative feedback is not derivable from the
NUCLEAR POWER PLANT CONTROL

143

kinetic equations of a bare reactor but depends solely on the characteristics

of the physical plant associated with the reactor. That is, an increase in

reactor power results in a more rapid absorption of heat by the heat

transfer medium or coolant. This increased rate of energy transfer

eventually results in a higher average temperature of the coolant modera-

tor and effects a change in reactivity. It is apparent that the form of

feedback effect depends upon the characteristics of the associated plant

and load rather than on the nuclear characteristics of the reactor.

This negative feedback, which is an inherent characteristic of the

plant, is a control system which is already built into the plant and which

cannot be removed or disconnected for either good or bad. It has one

COOLANT

FLOW RATE

CONTROL

ROD

8k

?\ 8k

"

^'^

PRIM
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

'

BOILER

LOOP

ARY "" "*"

<1

ruRBINE

TEMP. ( KTGT(S) ,

COEF. \ J

T,

TEMPERATURE

COEFFICIENT

Tov

FIG. 6-35. Block diagram of nuclear power plant indicating plant temperature coefficient

feedback loop.

advantage over an external control system in that it is ever reliable and

never out of order. Any external control system that is desired must be

designed to operate in parallel with the inherent control action in such a

fashion that the over-all performance is satisfactory. Consequently,

before designing an external reactor control system, one must first exam-

ine the relative stability of the reactor plant system before any controls

are added.

We are now in a position to make such an analysis. A new block

diagram of the plant showing how the negative temperature coefficient

ties back around the reactor through the plant is given in Fig. 6-35.

From this figure we see that in order to determine the stability of the

complete reactor plant system about a given nuclear power level no with

constant coolant flow rate, throttle opening, and control-rod positions,

we must examine the temperature coefficient reactivity feedback loop

transfer function

n(s) 3P.v(«)

= (TO

(6-63)

Sk(s) n(s)
164 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

where n(s)/8k(s) is the bare reactor transfer function as has been devel-

oped in Sec. 3-4. The poisoning feedback combination could also be

added if desired. It will be recalled from Sec. 3-4 that the reactor itself

is a nonlinear element, its gain being proportional to neutron-power level.

For sinusoidal signals of small amplitude at a given average power level

n0, the bare reactor was represented by the transfer function of Eq. (3-41)

(6-64)

«(.) <*sn < v ft

Plots of the magnitude and phase of the bare reactor transfer function

were presented as Figs. 3-15 and 3-16. The open-loop transfer function

of Eq. (6-63) is thus the product of the temperature coefficient (TC), the

transfer function for T^(s)/n(s), and the bare reactor transfer function.

As we have already developed T^(s)/n(s) in graphical form for our

elementary numerical example, we may continue the example again by

using the curves of Figs. 3-15 and 3-16, I* = 10~4 sec for the bare reactor

transfer function and thus compute XrGr(s) for the example. Figures

6-36 and 6-37 indicate the magnitude and phase of KTGT(s)/(TC) thus

obtained. From these curves it is easily seen that the system of our

example possesses great stability. Negative temperature coefficients for

practical reactors might range between zero and 10~3Sfc/°F. The ampli-

tude curve of Fig. 6-36 is at +15 db when the phase shift reaches —180°

at 11 cycles/sec. If the temperature coefficient of the reactor under

consideration were 10~35fc/°F, then 60 db must be subtracted from the

curve of Fig. 6-36 in order to obtain the total gain of the over-all tem-
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

perature coefficient feedback loop. Then, at —180° the gain would be

— 45 db, a tremendous margin of stability. If the temperature coeffi-

cient were smaller, this margin would be still larger and the stability of

the plant would be even greater.

The equation for the reactor feedback loop gain, Eq. (6-63), can be

rewritten as

K^T(s) = (TC) m , _ g* ... 1 + ^^ (6-65)

Our analysis has been at a constant power level whereby KB = 0.8 and

is at a constant flow. Let us examine now what happens if these param-

eters are changed. Assume that the power level is reduced. The reactor

gain n(s)/Sk(s) is nonlinear, as its gain is directly proportional to n0, the

power level at which the reactor is operating. In other words, if the

power level is dropped, the gain portion of the transfer function depending

upon the reactor will be reduced in a linear manner. The gain of KLG^s)
NUCLEAR POWER PUNT CONTROL

165

100

0.001 0.01 0.1 1

FREQUENCY, CYCLES PER SECOND

FIG. 6-36. Open-loop amplitude response of transfer function of temperature coefficient

reactivity feedback loop KrGr(s)/(TC).

50

-50

_,- -100

-150

-200

-250

0.001

0.01

10

100

0.1 1

FREQUENCY, CYCLES PER SECOND

FIG. 6-37. Open-loop phase shift of transfer function of temperature coefficient reactivity

feedback loop KTGM/(TC).

is also affected by power level in that it depends mostly upon boiler

attenuation KB. KB varies inversely with power level in accordance


Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

with Eq. (6-37), so that as the power level is dropped, the gain of KLGL(s)

goes up from this cause. This effect is very noticeable because the pri-

mary coolant loop provides a regenerative feedback path as previously

described. The term 1/[1 — KLGL(s)} is very dependent upon KB. For

example, from Fig. 6-17, the boiler attenuation is about 20 percent at


166 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

KaAB/Kt,Ab = 0.5. That is, the boiler gain is 0.8 at zero frequency at

this power level. Therefore the value of 1/[1 — KI£TL(S)] = 5 at zero

frequency. At KaAB/KbAb — 0.03 the boiler attenuation is only 2 per-

cent and 1/[1 — KLGL(s)\ = 50. In other words, a portion of the over-

all loop transfer function gain is reduced with power level whereas another

portion is increased. Whether a net positive or negative gain change

results depends upon the individual system constants.

The above reasoning was based upon constant coolant flow. If the

coolant flow is changed at the same time that the power level is changed,

there are two more effects which can result. The gain constant 7, which

relates the primary coolant temperature to reactor power, is inversely

proportional to coolant flow, for it will be recalled that y was denned in

Eq. (6-23) as y = tfro/tmoddmodcmod- As TO is the time for a unit volume of

coolant to pass through the reactor core, it is evident that TO increases

directly as the flow rate is decreased. Consequently, the loop gain

becomes larger inversely with flow. KB also changes slightly with flow,

but this is usually a secondary effect.

Therefore, in examining the stability of a given reactor plant system

it may be more appropriate to conduct the analysis at the lowest usable

power level rather than at full power output if the flow is reduced at low

power level. If the system can then be shown to be stable at its minimum

usable output, it usually can be assumed, for simple systems, that this is

a worse case and the system is stable a priori at full output. For com-

plex systems in which intricate flow patterns are called for in the pro-
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

gramming, it is necessary to examine how the heat-transfer coefficients

change as a function of flow before this assumption can be taken for

granted.

6-13. Transient Analysis. The procedure for analyzing the stability

of a reactor plant system without controls has just been indicated. We

would also, of course, like to examine analytically the transient response

of this system. The stability of a simple plant has been shown to be

very high for reasonable values of temperature coefficient. For small

temperature coefficients the absolute stability of a plant will be even

higher, but now the transient response should be poorer. What we have

is a degenerative feedback system in which the magnitude of the tempera-

ture coefficient controls the feedback factor. With a very small tempera-

ture coefficient the system is extremely stable as far as sustained oscilla-

tion is concerned, but a transient injected into the system would create

large overshoots in all the parameters. As the temperature coefficient

becomes larger, the system will still possess good stability, but now tran-

sients cause only moderate overshoots. As the temperature coefficient

becomes extremely large, sustained oscillations will result, and although

the transient response may be improved, it is immaterial. An optimum


NUCLEAR POWER PLANT CONTROL 167

temperature coefficient or range of temperature coefficients must then

exist. A minimum value should be present to prevent dangerously high

transients, and the maximum value must not approach the sustained

oscillation conditions.

A word denning stability is in order at this point. In conventional

terms, if a system is stable, no susta.ineH nsp;ilafir.r.a ^QT1 rpR1I]j- An

oscillation is caused when the gain of the system is greater than unity at

ISO^phase shut. Now, in the case of a bare reactor or for a reactor with

a temperature coefficient feedback loop with a zero or positive tempera-

ture coefficient, it is recognized that at zero frequency an infinite gain

results. This condition is also termed unstable, even though at zero

frequency only —90° of phase shift occurs. Actually the addition of

only the most minute amount of negative temperature coefficient makes

the loop of Eq. (6-65) stable. Its performance, with respect to transient

response, would still be very poor in that large excursions, but not infinite

excursions, of all parameters would occur. Such systems can be labeled

as having insufficient stability. A more proper description might be

merely to say that systems having large parameter excursions, because

of transient disturbances, simply have poor performance. Rigorous

mathematical representation of these explanations is available6 if

desired by the reader.

The transient analysis of even a simple system such as has been used

in the stability analysis of this chapter presents considerable analytical

difficulty. From 10 to 20 simultaneous differential equations would be


Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

involved, depending upon how closely one wished to approximate the

reactor and on how many mixing terms were used. The obvious solution

is to use the technique of Sec. 4-10 and set up an analogue computer or

simulator to examine the transient response. The method of designing

such a simulator is given in Chap. 11. Most of the components involved

in the example we have used have quite simple analogues with the excep-

tion of the linear time delays. These delays must be represented quite

accurately, as a wrong representation at low power level can easily throw

the temperature feedback loop KLGL(S) into continuous oscillation.

Magnetic tapes, phonograph records, or elaborate electronic networks4

may be used to approximate the linear delays.

Two types of transients are of interest. One is a transient change in

throttle opening, and the other is a transient change_m ^reactivity similar

to "the transient used in checking reactor response. What is necessary

is tcTexamine Lhe response of the Teactor power output, the thermal plant

output, and the significant temperatures in the plant as a function of

these two types of input disturbance. As the computer set up for such a

program involves obtaining numerial design data for a complete specific

plant system with all its auxiliary components, we shall merely outline
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

here the relative types of responses to be expected from a simple plant

without using an actual numerical example.

Consider the plant of Fig. 6-38, which will be of the same type that we

have been using previously, having a common coolant moderator and an

elementary temperature feedback control loop. Let us assume that this

plant is operating at a fixed power output level and that we suddenly

open up the throttle. This effectively inserts a step change in power

demand. Figure 6-39 indicates qualitatively the results of this operation.

The first effect noted is an instantaneous increase in thermal output

accompanied by a less rapid decrease in steam temperature. The power

delivered to the turbine is available very fast because of the release of

energy stored in the heat capacity of the boiler and coolant. The

THROTTLE

COOLANT

CIRCULATOR

BOILER FEED

WATER PUMP

FIG. 6-38. Block diagram of elementary plant showing mixing.

decrease in steam temperature causes a decrease in primary coolant

temperature, since more heat is extracted from the coolant. The change

in temperature of primary coolant produces a significant reactivity change

at once, and the reactor power starts to rise almost immediately. The

thermal behavior of the fuel is of importance from a materials point of

view, and its general temperature pattern is similar to the pattern of the
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

average coolant temperature. It will be noted that after the transient

has passed, the average coolant temperature comes back to its initial

value, but now the power delivered to the turbine is the greater required

amount. A slight overshoot is seen to be present in the reactor power

output, and this overshoot serves the purpose of recharging the stored

energy in the boiler system.

168

For the illustration used, the plant has a moderate-sized temperature

coefficient, and the entire plant appears to be highly overdamped.

Becaxise of this fact, it would appear possible to take a plant of this type
NUCLEAR POWER PLANT CONTROL

169

and increase its speed of response to a transient by means of an external

control system.

The other type of disturbance we wish to examine is the disturbance

caused by a change in reactivity for any reason whatever. A simple

case to consider is that of a step withdrawal of a control rod. It has been

shown in Sec. 3-2 that when a control rod is suddenly withdrawn, the

ill

TIME

FIG. 6-39. Transient response of plant

parameters to a positive step in throttle

opening.

TIME

FIG. 6-40. Transient response of plant

parameters to a positive step in reactivity.

reactor starts up on a positive period, the leading edge of which starts to

follow the relationship

dn Sk

~Al = ]*n (6-66)

Thus, in a very short amount of time a bare reactor would quickly reach

a very high power level. However, what actually happens in the case

of a reactor plant with negative temperature coefficient is that before the


Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

power level gets very high, the heat flow through the reactor fuel to the

coolant raises the coolant temperature enough to stop the rapid rise.

Figure 6-40 shows the effect of a step change in reactivity upon the sys-

tem. The rising coolant temperature extracts reactivity and brings the

reactor power level down. A small secondary peak may exist because
CONTROL Of NUCLEAR REACTORS AND POWER PLANTS

of the emission of delayed neutrons produced at the initial high peak of

power. The reactor fuel acts as a filter between the reactor power and

the reactivity change. A transient heat spike can be seen in the fuel

temperature, but it is completely filtered out by the time the transient

arrives at reactor coolant temperature. The steam temperature and,

therefore, the thermal power follow the transient much more slowly.

After the transient has passed, the reactor coolant temperature must

be above its original value in order to offset the increase in reactivity

caused by the rod motion. This condition causes all the temperatures

of the plant to rise, and as a consequence, the thermal output power and

the reactor power will rise somewhat above their original values. A com-

pletely similar, but opposite, set of curves results when negative throttle

changes or negative reactivity steps are inserted.

6-14. External Reactor Control System. Having developed the tem-

perature coefficient reactivity feedback loop, we are now in a position to

n0 SIGNAL DERIVED FROM MEASUREMENTS

OF THE STEAM PLANT VARIABLES

FIG. 6-41. Block diagram indicating the feedback loops which affect reactor stability.

examine a combined reactor and plant to see how the reactor responds

to external controls in a manner similar to the analysis of Sec. 4-7. It

will be recalled that in this previous section the temperature coefficient

was handled as a simple local feedback loop around the reactor and the

feedback gain was combined with the reactor gain by means of conven-

tional servo analysis for two elements in parallel. Then when a com-
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

bined reactor temperature coefficient transfer function was developed, an

external control loop could be tied around the combined system. Figure

6-41 indicates how a plant can be set up to cause an external reactor con-

trol loop to control a combined reactor plant system. It is necessary,

however, to determine the combined transfer function of the reactor sys-

170
171

NUCLEAR POWER PLANT CONTROL

tem indicated by the dotted box of Fig. 6-41. The reactor and its feed-

back loops again may be linearized for small amplitude sinusoidal varia-

tions at an average power level no. These transfer functions are indicated

in the diagram of Fig. 6-42a. The system may then be reduced to the

equivalent form of Fig. 6-426, in which the reactor and its temperature

coefficient feedback system have been reduced to a single block. The

function n(s)/5fcrod(s) is the representation of the new combined reactor

REACTOR

^rodl+A

^6K^

n(s)

6k (s)

„ "(=).

1 +.

1
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

1_

bK P

TEMP.

COEF

EXTERN

LAIN REACTOR

EMP. FEEDBACK

, ToV(s)

TC n(s)

AL CONTROL

FEEDBACK

SYSTEM

KBGB(s)

(a)

FIG. 6-42. Equivalent block diagrams of the reactor with temperature coefficient and

control-rod feedback loops for small sinusoidal signals.

and temperature coefficient transfer function. KgGs(s) again represents

the external reactor control system. Then

n(s)/<5fc(s)

(TC)[T\v(S)/n(s)] 1 +

(6-67)

(6-68)

We have then reduced the problem of reactor control in a power plant to

the more simple problem of reactor control as analyzed in Secs. 4-7 and

4-8. The bare reactor had a given transfer function, and it was shown

that this reactor could be controlled either with a proportional control

system or with an on-off step type control system. When a simple reac-

tor system containing a negative temperature coefficient, but no coolant

flow, was examined, it was shown that all that was needed was to modify

the reactor transfer function by combining it with a parallel reactor tem-

perature coefficient feedback system. Poison feedback was handled in a

similar manner. Then a new combined reactor function with all these

feedbacks could be controlled by an external loop. In the case of a full


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

plant we have now shown that the bare reactor transfer function need

only be combined in parallel again with a plant temperature coefficient

feedback system and an external control, either proportional or on-off

type, may be analyzed in a completely analogous manner to the simpler

systems previously presented.

SYSTEM

COMBINATION

CONTROL

6k (s)

n(s)

la)

5k(s)

KRGR(s)

n(s)

ARE REACTOR

KTC GTC

LOCAL

TEMP. COEF.

FEEDBACK

(b)

6k (S)

n(s)

BARE REACTOR
Generated on 2011-10-24 11:51 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

PLANT TEMP.

FEEDBACK

(c)

REACTOR

CONTROL SYSTEM

FIG. 6-43. Combinations of reactor and various feedback loop transfer functions, (a)

Elementary reactor, (b) Local temperature coefficient feedback, (c) Plant temperature

coefficient feedback.

Figure 6-43 indicates the sequence of combinations just described.

Figure 6-43a shows a bare reactor and then an external reactor control

system transfer function of any type KB(?B(S). Figure 6-436 shows the

combination of this bare reactor and a local temperature coefficient effect.

This is the condition of most reactors operating without external plants.

Figure 6-43c shows an example of reactor control when a full plant is

present. The poison feedback loop again is not shown. It can easily

be seen that the analysis for any external reactor control system can

always be made to have the same form by proper combination.

172
173

NUCLEAR POWER PLANT CONTROL

For illustrative purposes we can plot Eq. (6-68) for the example we have

been using for a temperature coefficient of (TC) = 10~35fc/°F. Figure

6-44 indicates this plot of the combined reactor and plant temperature coef-

ficient transfer function. It is of interest to note that this transfer func-

tion is of the same general form as the combined reactor, local temperature

coefficient feedback system of Figs. 4-4 and 4-6. In particular, the trans-

fer function amplitude approaches a finite value at zero frequency. The

phase shift approaches zero at zero frequency and —90° at infinite fre-

quency. Thus we see again that the example of the reactor plant nega-

tive temperature coefficient system we have been using is a most stable

one, neglecting any possible long-term poison oscillations.

59

7- 50

—

? J 45

I" 40

LU

S 35

5 30

i 25

25

C1-

In -50
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

I -75

-100

^APPROACHES

0.001

APPROACHES 90°-:

0.01

10

100

0.1 1

FREQUENCY, CYCLES PER SECOND

FIG. 6-44. Magnitude and phase of over-all reactor transfer function with temperature

coefficient, (TC) = 10-35fe/°F.

6-15. Automatic Plant Control.6 Reactor control by itself in a plant is

of little importance. The essential problem is to obtain useful power

from the plant. Reactor control then, at most, becomes a minor auxiliary

loop in a power plant. Plant control must be all-prevailing and overrule

local situations. We have just examined the natural stability of a plant

and found that the reactor could be treated as a component in a plant

system. We must now examine the basic plant when an attempt is

made to superimpose an external control system upon it. Although most

plants will be quite stable by themselves, it is very possible to upset this

natural stability by an improperly designed external control system.

The function of an automatic plant control system is twofold. First,

it must set up the ultimate steady-state temperature, pressure, and flow

programs desired; second, it should improve, or at least not detract from,

the transient penormance of the basic plant.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

To accomplish these functions requires the study of a new control loop.

We have determined that the reactor plant and reactor control could be

synthesized into a single loop. We now must superimpose a demand

loop upon the reactor and the temperature coefficient feedback loop.

Figure 6-45 shows an arrangement of a plant illustrating the position of

the new demand loop.

Two different situations can arise. These will depend on whether the

plant by itself has a large amount of natural stability or has poor transient

performance or so-called insufficient stability. Since the type of control

system required depends upon whether the plant is inherently stable or

has poor performance, the two situations will be discussed separately.

FIG. 6-45. Block diagram of power demand loop.

If an analysis shows that the plant by itself has poor performance,

then complete automatic control is necessary. The automatic system

must control the reactor power level according to the output demanded,

in addition to maintaining the required temperatures. These require-

ments suggest a control system of the type shown in Fig. 6-46. Reactor

power is maintained equal to a generated power demand signal by con-

trolling the reactivity, usually by manipulation of the control rods in the

reactor core. The indication of reactor power is obtained from nuclear

instruments. The power demand signal is generated from measurements

of plant variables and must be a measure of the load on the plant and the

deviation of a variable from the value specified by the steady-state pro-

gram. Usually the steam flow is a good indication of the load require-
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ments. In fact^ if the enthalpy risethrough the steam generator is a

constant independent of load, the power delivered is directly proportional

to the steam flow. Throttle position also may be used. TEe particular

174
NUCLEAR POWER PLANT CONTROL

175

variable used to indicate deviation from the set program might be any

coolant temperature in the loop or the steam pressure. Because of the

relationships that exist at steady-state conditions, only one variable is

needed to control the program.

This system of control reactor power through a demand loop and a

neutron-measuring reactor control loop is a necessity for a reactor con-

taining either a positive temperature coefficient or a very small negative

temperature coefficient. After a control program for such a system has

been decided upon, the stability of the complete system may be studied

on a frequency response basis. Again we must stipulate complete lin-

earity. Consequently, the results are strictly valid only for small devia-

tions. Satisfaction of familiar stability criteria and attainment of desired

4»

MEASURED ERRO

REACTOR SIGN/!

POWER

SIGNAL

L REACTIVITY g|<

MECHANISMS ^

REACTOR

KEACTOR POWER ..

PLANT

j
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

n0

INHERENT

POWER

DEMAND

FEEDBACK

POWER DEMAND SIGNAL

GENERATING CIRCUITS

MEASURED

CONTROL

VARIABLES

FIG. 6-46. Block diagram of control system for a plant containing inadequate natural

stability.

margins of stability are readily accomplished using standard servo tech-

niques. Whether the addition of derivative control (anticipation), inte-

gral control (reset), or other compensating networks is necessary can be

determined from Nyquist and Bode diagrams in the usual manner.

If the analysis of the basic plant, such as has been carried out in the

previous sections of this chapter, indicates that the plant is inherently

sufficiently stable, the automatic control problem is simplified. The

plant effectively controls itself supplying whatever load is imposed. Any

difference between reactor power and steam load results in a rate of

change of average coolant temperature. The reactivity change caused

by the temperature change then alters reactor power, thus resetting the

temperature and the heat balance. The speed of transient response

obtained may in itself be adequate for the system application, so that in

many cases automatic control of the power level for improved transient

performance may not be required. The use of nuclear measurements

and the generation of a neutron power demand signal become unneces-

sary. Thus the role of the automatic system is reduced to that of main-

taining the steady-state temperature program.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The requirements for a plant that is inherently sufficiently stable sug-

gest a control system of the type shown in Fig. 6-47. Here the tempera-

ture or pressure error is controlled directly by insertion and withdrawal

of reactivity by control-rod motion. The inherent stability of the plant

compensates for any change in reactivity by shifting the temperature

level until the net reactivity is again reduced to zero. Thus by moving

the control rods a given distance, the temperatures shift to a new con-

stant value, and direct control of the program is accomplished with no

neutron information being required. However, as in the previous exam-

ple, the over-all system stability must again be studied and optimized by

a frequency response method.

PROGRAM

REFERENCE

FIG. 6-47. Block diagram of control system for a plant containing adequate natural

stability.

In summary, if the plant has sufficient natural stability, direct control

of plant parameters may be attempted. If the plant has poor inherent

transient performance, neutron-level control should also be used.

Control Systems for Plants with Insufficient Inherent Stability. It will

be recalled that many forms of control programs may be demanded

from a plant. Obviously, each program can call for a completely differ-

ent type of control, but there are recognizable patterns of control systems

which can fit frequently occurring situations. The most obvious type

of control program is the constant-T^ program. As has been pointed out


Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

in Sec. 6-3, this is the natural program of the reactor. In addition, there

is the constant-steam-pressure program, which is the preferred program

of the secondary system. Many in-between compromises are possible.

We shall now examine a few types of control system setups for the con-

stant-T.v program, the constant-steam-pressure program, and varying

degrees of compromise when there is a neutron control loop involved.

176

Figure 6-48 indicates two types of control systems. Figure 6-48a rep-

resents a scheme whereby the power demand signal no is generated accord-

ing to the equation


177

NUCLEAR POWER PLANT CONTROL

=K

(6-69)

The no signal generated in this manner is then compared with reactor

power, assumed proportional to the neutron flux level, and the resultant

error signal actuates the control rods. The integral of !F.v error is neces-

sary to provide a steady-state no signal, since no Tav error exists in the

steady state. This system, while simple, is at a disadvantage because

a poor transient response results from throttle changes. As the throttle

COMPARATOR

?&[>— 0*--

t Tn

-T~L Tovref

' 1 — '

l>

II

ROD

Lhj '

rVc

LOAD

REACTOR |*

BOILER

<
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

(o)

COMPARATOR

DRIVE

f~

IS

T-

BOILER

LOAD

n0=K1Fs+K2[(ToVre(-Tov) + i (TOVref-Tov)dt]

(t»

FIG. 6-48. Two types of control systems for constant-average-temperature program,

(a) Indirect control, (b) Direct control.

opening is increased, steam flow increases immediately, thus withdrawing

more power from the primary coolant and lowering the average coolant

temperature. The reactor power is increased to supply the demand,

preventing further depletion of the reservoir of energy stored in the cool-

ant. The sooner reactor power equals the steam demand, the shorter

the duration of the transient. The control system of Fig. 6-48a requires

that an error in Tav exist for a length of time following a throttle change,

in order to furnish the necessary change in demand signal n0) since the

accumulated integral of the T^ error is the only contributor to no in

the steady state. In other words, this system behaves pretty much in the
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

same manner as would the natural plant stability by itself. The Tm,

signal ensures that the long-term depletion and poisoning changes are

wiped out.

Figure 6-486 shows a control scheme that does not have the limitations

just described. In this system we again assume that the enthalpy of the

steam generator is essentially constant throughout the power range.

Since power at any level is then the product of steam flow and enthalpy,

the thermal power being delivered is directly proportional to steam flow

F,. The generated demand signal then becomes

(Tmr.rtf -

dt

(6-70)

Here the demand signal is nearly directly proportional to steam flow and

a change in steam demand immediately changes HQ. The reactor power

COMPARATOR

COMPARATOR

n0=K,Fs+K[(pref-p)+l^pref-p)df]

(b)

FIG. 6-49. Two types of control systems for constant-steam-pressure program, (a)

Indirect control, (b) Direct control.

178

is adjusted much more rapidly than in the system of Fig. 6-48a, where

a temperature error was required for a length of time. The integral of

IT.v error is included to supplement the steam-flow signal in regions where


Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the steam-flow indication may be in error. The value of Ki should be


NUCLEAR POWER PLANT CONTROL

179

very close to unity; that is, Ki = 1 percent reactor power demand per

1 percent steam flow. The scheme of Fig. 6-486 may be called the direct

method of controlling no; that of Fig. 6-48a the indirect method.

Two comparable schemes can be set up for controlling a constant

steam pressure program, and these are indicated in Figs. 6-49a and 6-496.

Figure 6-49a employs the pressure error to generate the power demand

signal which is then compared with reactor power in a similar manner as

was the Tav error. This system has the same inherent limitations as

Fig. 6-48a in that an error is required during a transient to supply the

steady-state n0 signal. The circuit of Fig. 6-496 again controls no

directly, using the steam-flow measurement to represent the actual power

COMPARATOR COMPUTER

DRIVE

ROD

In > 1

Pref IF,

^L

^* 1 — 1

LOAD

<

1
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

(3

REACTOR

=J--J

BOILER

n0=K,Fs+K2[(pref-p) + iy"(pref_p)dt] WHERE Pref=f(Fs)

FIG. 6-50. Universal control system based on a variable pressure reference.

demand, plus a function of the pressure error to reset the power demand

signal.

A system capable of handling any program ranging from constant Tav

to constant p is indicated in Fig. 6-50. The scheme essentially is the

same as that of Fig. 6-496, except now the pressure reference is not con-

stant but may be programmed in any predetermined manner, linear or

nonlinear, by the steam flow. In other words, if a constant- T.v program

is desired from this system, the pressure reference can be tapered to drop

off in the proper manner as a function of power demand. By means of a

universal control scheme of this type, any program can be set up by meas-

uring and controlling from secondary system variables. No measure-

ments of primary loop temperatures are thus required.

Control System for Plant Having Sufficient Inherent Stability. Once it

has been determined that the natural stability of a plant by itself will be

satisfactory for the intended service, the superimposed control system

becomes considerably simpler. No reactor control loop as such is needed,

and direct control of the parameters is possible.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The first system that comes to mind is the very elementary one of

manual control. Figure 6-51 shows two possible manual control schemes,

one for holding the average temperature constant and the other for hold-

ing steam pressure constant. The human link in the system is specified

to be deliberately slow, and his action consists merely of setting a meter

reading. The plant takes care of itself through its own natural stability,

and only occasional shimming by the operator is necessary.

ROD

DRIVE

Q- O

ROD

MANUAL

CONTROL

REACTOR

'INDICATOR

BOILER

(a)

LOAD

ROD

DRIVE

-o
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

(/*]

-Pref

MANUAL PRESSURE

CONTROL ERROR

INDICATOR

FIG. 6-51. Two manual control systems, (a) Constant-average-temperature program,

(b) Constant-steam-pressure program.

If, on the other hand, automatic control is desired, systems similar to

those of Figs. 6-52 and 6-53 may be used. Here a temperature or pres-

sure error is made to actuate control rods directly and thus take care of

removing the specified system parameter error. Any neutron indication

that is present in the system is used only for such information as giving

alarms and is not tied back into the control. Both systems of Figs. 6-52

and 6-53 are slow shimming operations and may be used either with

proportional control or with on-off control.

180
NUCLEAR POWER PLANT CONTROL

181

TEMPERATURE

ERROR

Th REFERENCE

FIG. 6-52. Automatic control system for a plant with sufficient inherent natural stability

using outlet temperature as the primary control variable.

TAPERED PRESSURE

REFERENCE

SOME FUNCTION

OF F,

Fs MEASURED

STEAM FLOW

(b)

FIG. 6-53. Automatic control system for a plant with sufficient inherent natural stability,

(a) Primary control variable, steam pressure, (b) Primary control variable, steam flow.

6-16. Stability Analysis for Demand Loop. The analysis of any of the

plant control systems just presented should again be in two parts, that is,

a stability analysis based on frequency response and a transient analysis.

As the transient analysis would in practice be accomplished by simulation

technique on analogue computing machines for a very specific reactor


Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

plant, we shall confine our attention to a general approach to the stability

problem.

As before, two variations of the problem exist in that there are two

different types of demand loop, depending upon whether the plant by

itself has adequate natural stability or not. In an actual plant design,

at this point the problem becomes one of synthesizing the demand loop.

The process is a tedious one of actually optimizing component constants

for maximum stability and for meeting operating specifications.

Frequency Analysis for Plant with Reactor Control Loop. Let us first

consider the case of a plant having inadequate natural stability where a

FIG. 6-54. Power demand control loop for a plant with insufficient natural stability.

reactor control loop is specified. The power demand loop for this con-

dition is shown in general form in Fig. 6-54. The stability analysis for

this loop is outlined in the following manner. Let us break the loop at

some convenient place such as between the power demand signal generat-

ing circuit and the comparator at the point marked X on Fig. 6-54. We

can now examine the power demand open-loop gain and phase shift by

injecting a small sinusoidal variation in no into the comparator and

examining the output of the power demand signal generating circuit.

The over-all loop gain is then examined for phase margin and gain mar-

gin, and if necessary, appropriate compensating circuits can be added to

the loop. Symbolically the open-loop gain consists of two parts:

Open-loop gain =

n(s)
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

n(s)

(6-71)

182

where n(s)/nc(s) is crudely the reactor system portion of the over-all

transfer function and n0(s)/n(s) is the plant portion of the power demand
NUCLEAR POWER PLANT CONTROL 183

loop transfer function. The term n(s)/no(s) consists of an external reac-

tor control loop around the combined reactor plant temperature coeffi-

cient feedback loop transfer function n(s)/5/crod(s). If we denote this

neutron loop gain by KnGn(s), then

n0(s) l+KnGn(s)

KnGn(s), of course, contains all of the nonlinearities previously examined

in that the reactor gain is a function of level, the comparator gain an

inverse function of level, and the control amplifier, if it is of the contactor

type, may have its gain of the form (V/b)Ab/a(\x\/b), where the output

is a discontinuous function of its input. Nevertheless, how these non-

linearities may be handled has been pointed out in Chap. 4, and KnGn(s)

may be found in a logical manner.

The power demand signal n0 will usually be similar in form to the

following terms, as indicated in Sec. 6-14:

n0 = /i(F.) + fi(T) or /,(p) (6-73)

That is, the power demand signal first consists of some term generally

directly dependent on thermal power output or its direct relations such

as steam flow or throttle opening. Second, there will be an error-reducing

term which will be used to set up the plant program around some variable

such as a plant temperature or steam pressure. The transfer function

can be found as

For a more specific example let us assume that the plant operating

conditions call for T\ to remain constant as a function of power level.

The following equation might then apply for a given control system:
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

n0 = KJ, + K2(Th.ref - Th) (6-75)

then "^ \=K^-K^ (6-76)

n(s) n(s) n(s)

As all the other terms of this equation are essentially fixed by the prin-

cipal components of the plant, the synthesis of the demand loop would

usually consist of optimizing the constants Ki and K^.

Two details must be mentioned at this point. The first is that, al-

though the procedure for obtaining T\(s)/n(s), Tc(s)/n(s), and T^(s)/n(s)

has been previously outlined, no mention has been made of F,(s)/n(s).

The method used to determine Fe(s)/n(s) is taken directly from the dif-
1 84 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

ferential equation of (6-18),

- T.) - KaAp

(Mmcm + M,c.) ^

and the additional relationship

F, = (BKaA)T,

(6-77)

(6-78)

where the symbols are as defined in Sec. 6-4 and the constants are greatly

dependent on the specific secondary plant used.

Another detail which must be included in the demand loop over-all

gain is the fact that the information on plant conditions is not available

instantly. For example, a thermometer of some sort would be used to

TC

COMPENSATION

IF NEEDED

DEMAND \

R P LOOP .

CONTROL

AMPLIFIER

LINEAR OR

NON LINEAR

REACTIVITY

ACTUATOR
Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

MECHANISMS

REACTOR

COMBINED WITH

PLANT TC

FEEDBACK

DEMAND SIC

GENERAâ„¢

CIRCUITS A

COM PA RAT

NAL uc.

G T0 OR ME

- X s,6

ASUREMENT

NAL DELAYS

MEASURED

CONTROL

VARIABLES

PLANT

Tref

OR Pref

FIG. 6-55. Demand control loop for a plant with sufficient natural stability.

measure primary loop temperature. In conventional practice such a

thermometer might have a time constant ranging from 0.2 to 30 sec.

This time constant can have a very serious effect on the stability of the

demand loop. In addition, in some types of hydraulic or pneumatic

measuring devices, a transport time delay may also be involved before

the temperature or pressure signal arrives at the power demand signal

generating circuit. Therefore, in general, additional delays of the form

1/(1 + TS) and e~â„¢ in combination may be expected in a practical plant

demand loop.

Frequency Analysis for Plant with Sufficient Natural Stability. For

this second type of control system we may examine the block diagram of

Fig. 6-55. Here the separate reactor control loop is eliminated and a

comparison is made directly between some demanded temperature or

pressure reference (linear or nonlinear with power output) and the actual
NUCLEAR POWER PLANT CONTROL 185

temperature or pressure existing in the plant. The demand open-loop

gain now becomes

Open-loop gain = (6-79)

or Open-loop gain = - (6-80)

(ps) n(s)

depending on whether we are using a temperature or a pressure as a

reference. The term n(s)/[T(s) or p(s)] then becomes the combined

reactor plant temperature coefficient feedback transfer function in series

with the control amplification and actuation mechanism. The demand

signal may take forms such as

T = Tref - To (6-81)

where TO may be any function Tc, Th, or T.v, or

leading to transfer functions for the plant portion of the loop of the form

n(s) n(s) n(s) n(s)

or for the case where pressure is the reference

PM KlF_M + K^ (6.83)

n(s) n(s) n(s)

It must be pointed out again that all analyses such as have been indi-

cated above should be examined for stability at various power levels, as

most of the parameters being used are a function of level or flow or both.

Analyses conducted in the above manner will usually result in a limita-

tion being placed on the reactor system for stability. This limitation

can, in most cases, be boiled down to the simple problem of permissible

rate of reactivity change. The condition that is found in a practical case


Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

is that above a given control-rod speed of motion the system is unstable.

It can be shown that below this speed the system will be stable. It is

important to note at this point that practical reactor operating conditions,

such as transient poison burnout, also impose requirements on control-

rod reactivity change rates from a reactor safety point of view. Some

of these requirements may call for the control rods moving at a faster

rate than a given amount. These requirements will be examined in

detail in Chap. 9. Thus, offhand, the over-all problem of plant control

system design for power range operation is going to depend upon finding
186 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

conditions of overlap between control system stability and reactor opera-

tional limitations.

REFERENCES

1. Schultz, M. A.: Trans. AIEE, vol. 71, paper 53-31, 1952.

2. Schultz, M. A.: Temperature Programs and Control Systems for a Nuclear Power

Plant, "Proceedings of the 1953 Conference on Nuclear Engineering," University

of California Press, Berkeley, Calif., 1953.

3. Brown, G. S., and D. P. Campbell: "Principles of Servomechanisms," John Wiley

& Sons, Inc., New York, 1951.

4. Single, C., and G. Stubbs: "Transport Delay Simulation Circuits," WAPD-T-38,

Westinghouse Electric Corporation, Pittsburgh, Pa.

5. Grace, J. N.: Synthesis of Control Systems for Nuclear Power Plants, "Conven-

tion Record of the IRE 1954 National Convention," pt. 9, "Medical and Nuclear

Electronics," Institute of Radio Engineers, New York, 1954.

6. Truxal, J.[G: "Automatic Feedback Control System Synthesis," McGraw-Hill

Book Company, Inc., New York, 1955.

7. Stubbs, G. S.: Constant Reactor Outlet Temperature Control System, IRE

Trans, on Nuclear Science, vol. NS-1, no. 1, September, 1954.


Generated on 2011-10-24 11:52 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CHAPTER 7

REACTOR CONTROL RADIATION DETECTORS

We have seen in the preceding chapter that certain types of control

systems call for the use of a radiation-measuring loop. The radiation-

measuring devices involved are unique and therefore will be described

separately in this chapter as control components. Actually many types

of radiation detectors exist, but very few are suitable for control purposes.

We shall attempt to define the requirements and operational ranges for

the instruments. We shall also look at the limitations of radiation-

detection instruments as power indicators and control elements. Finally

we shall examine the problems involved in calibration of these detectors.

7-1. Measurement Problem. In considering what to measure in a

reactor in order to obtain a signal for control purposes, it is evident that

one desires to obtain a measurement of reactor power output. In a

complete plant operating at a high power level it seems possible that a

thermal measurement of the temperature rise across the reactor times

the flow of coolant might be used. This method has two disadvantages:

First, the information is slow; second, over much of the reactor operat-

ing range at low power levels there is no appreciable temperature rise

involved. Also, for reactors without a plant attached, this method may

be unfeasible.

In searching for a suitable parameter to use for power-measuring pur-

poses, it is evident that the total power emanating from the reactor con-

sists of the sum of all the energies involved in the nuclear disintegrations

in the reactor. As over 90 percent of the energy involved comes from


Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the direct fission process and less than 10 percent by indirect beta and

gamma radiations, it seems reasonable to use the incidental neutrons

from the fission process as a measurement parameter. In addition, the

beta and gamma power does not directly follow the fission power time-

wise but may come from long-lived decay activities. Consequently,

attempting to monitor on gamma radiation alone, for example, would

lead to an erroneous answer.

The power level of a nuclear reactor is then generally assumed to be

proportional to the number of neutrons in its core. Operating conditions

187
188 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

arise where the relationship between neutrons and power is much more

complex, but for most purposes the simple assumption of direct propor-

tionality is essentially correct.

The problem exists of measuring a given fixed fraction of reactor neu-

trons with conventional radiation detectors and to relate the output of

these detectors to reactor power. It can be assumed to start that the

neutron flux distribution in a reactor is not a constant but varies with

time and position in the core. In order to read power output directly,

it is then necessary to measure continuously the average core flux. Two

means of making an average flux measurement are at once apparent.

First, a large number of detectors might be placed as uniformly as pos-

sible throughout the core and an average taken of their output readings.

Second, a detector might be placed so far from the core that the core can

be considered a point source of neutrons and local neutron density varia-

tions in the core will be insignificant. Both methods require compromises

to be at all applicable, with the second approach being the one most gen-

erally used.

7-2. Ranges of Measurements. The fact that the detecting instru-

ments are usually placed at some distance from the reactor imposes

severe limitations on instrument sensitivity and range. First a nuclear

reactor is really never shut off. In a conventional power plant whose

normal full power output might be 1,000 kw, if the output were turned

down to 1 kw, one would consider that the plant was essentially shut

off and could be ignored. In the case of a nuclear plant operating at


Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the same level, because of the unique self-multiplying characteristics of

the medium at any power level 1 kw still represents a great many neu-

trons. If these neutrons are not monitored continuously and the pos-

sibility of creating a multiplication factor greater than unity exists, then

the plant is capable of ultimately destroying itself, starting from any

conceivable power level. Neutrons must then be monitored at all times

and at all levels. In the case of a power reactor this requirement may

call for as much as 10-decade instrumentation for complete monitoring

of the entire operating range from source strength to full power output.

No one present-day instrument is capable of covering such a range.

In conventional instrumentation, instruments usually can operate use-

fully over 3 or at most 4 decades. Consequently, to cover the wider

requirements of the nuclear power plant, over-all coverage may be

obtained by using multiple instrumentation with overlapping ranges,

or a very sensitive instrument may be used and its position changed with

respect to the reactor as a function of power level. Figure 7-1 indicates

the multiple instrument overlapping-range system. Figure 7-2 illus-

trates the movable instrument system. The diagram of Fig. 7-2 shows

an interesting servo-type instrumentation system in which the detector


REACTOR CONTROL RADIATION DETECTORS

189

position is continually moved so that the detector always operates in a

constant neutron flux. In this way the position of the instrument with

respect to the reactor becomes a measurement of the neutron level.

The advantage of the movable instrument system is that only one type

of detector is needed to cover the entire operating range. The principal

disadvantage is that the neutron level as a function of distance from the

10"

UJ

UJ

io-

io-

10"

io-

UJKO

O.UJO;

Q UJ

Oo

0.0:

z uj

uj o

t>
Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

28

ro

Ul

S*uj

g8

FULL

POWER

. CRITICAL

SOURCE

LEVEL

FIG. 7-1. Typical instrument-range coverage for a multiple fixed instrument system.

reactor is usually very nonlinear and may contain sharp slope changes at

the boundaries of different materials. Although these effects may be

calibrated out of the instrument, they are apt to change with time.

Consequently, the overlapping-range type of instrumentation is the more

prevalent system at present. However, moving an instrument a fixed

amount to reduce its sensitivity or to protect it from high neutron or

gamma radiation levels is done quite frequently.

Figure 7-1 also indicates the conventionally used terms to denote the

operating ranges of the reactor. Three ranges are usually considered,


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

although separate instruments may not necessarily be used to cover each

range as such. These ranges are the source, or startup, range, the period

range, and the power operation range.

The source range is the range of neutron level between the points

where the reactor is shut down as completely as possible and where the

reactor is critical. At the complete shutdown level the detector sees only

its fraction of those neutrons emanating from the source as multiplied by

the subcritical multiplication factor of the reactor. The point at which

a reactor goes critical depends upon how fast reactivity is removed from

the reactor. This level is usually between 20 and 1,000 times the source

i THIS DISTANCE IS MEASURE

REACTOR

("

/ Up ^

MOVABLE FIXED

NEUTRON DETECTOR , NUT SCREW

1|

I ' isl i 1

iiihh iiiiiiiii|i * i|i i ii'hii — 1 1 ih i rnjj i iiiii i ; i|i|iMi|i i|i|i|i|

= . SERVO

g MOTOR *

NEUTRON

CIRCUIT

~\
Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

REFERENCE

LEVEL

FIG. 7-2. Block diagram of movable instrument system in which the detector is auto-

matically positioned to sit in a constant neutron flux.

level. Once criticality is reached, the reactor is brought up to a level

where thermal power can be measured. This rise is usually restricted to

a given rate or period. This interval in level is called the period range.

The period used for specific reactors depends upon their operational

requirements and may vary from 3 sec to a few minutes. In this range

circuits providing both level and period information are usually con-

nected to the detector.

The power operation range is where the reactor is producing an amount

of power close to its rating. This range usually extends downward no

more than 2 decades from full power rating.

190

7-3. Description of Instruments. The many types of instruments that

might be considered for reactor control can be divided into two types,

classified by the operation of the external circuit to which they are

connected. These are pulse-counting types of circuits and current-


Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

A picture of a conventional BF3 proportional counter is given in Fig.

7-3. Several other configurations are presented in the literature.1 The

instrument has a plateau in counting

rate as a function of applied voltage,

and the plateau of a typical detec-

tor is presented in Fig. 7-4. These

instruments are usually connected to

an electronic circuit which selects

pulses above a given size to be counted.

In this way spurious pulses and noise

may be partially eliminated. The

performance of a detector and circuit

connected in this manner is deter-

mined by a so-called integral bias

curve. Such a curve is presented in

Fig. 7-5 for a typical BF3 counter.

Sensitivity of tubes of the type of

Fig. 7-3 ranges from 0.1 to 2 counts

per unit of neutron flux.

Boron-lined Counter.1 The boron-lined counter usually consists of

cylindrical electrodes coated with boron, which are enclosed in an inert

600

Q 400

o
Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ro

2200

2600

2400

VOLTAGE

FIG. 7-4. Counting rate versus voltage

for a typical BF3 counter.

3000

1000

100

10

0 200 400 600

PULSE HEIGHT SELECTOR SETTING, ARBITRARY UNITS

FIG. 7-5. Integral bias curve for typical BF3 counter showing counting rate as a function of

pulse height selector setting.

192

gas atmosphere. Its external appearance and performance are similar

to the BFa counter. Its advantages over the BF3 counter are that it
REACTOR CONTROL RADIATION DETECTORS

193

does not contain a fine wire electrode and that it usually operates at a

somewhat lower voltage between electrodes. Its principal disadvantage

is that for equal volumes it is usually less sensitive than the BF3 counter

and has a shorter voltage plateau. This shorter plateau usually means

that the tube will perform more poorly in the presence of large gamma

fields.

Fission Counter. The fission counter detects neutrons by utilizing the

highly ionized fission fragments that result when a neutron causes a

nucleus to undergo fission. In general, the instrument consists of an

ELECTRODE NO. 1

ELECTRODE

NO. 2

SPIRALS INSULATED FROM EACH OTHER AND

COATED ON BOTH SIDES WITH FISSIONABLE

MATERIAL

FIG. 7-6. Cross-sectional view of spiral type fission counter.

ionization chamber type detector which contains fissionable material in

the form of a thin foil or coating on the counter electrodes. By a proper

choice of this coating material and by the use of neutron shields placed

around the outside of the counter it is possible to design a tube that is

sensitive to neutrons of a particular energy range. At present it is the

measurement of thermal energy range neutrons that is of importance in

most reactor control problems.


Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Two types of fission counters are available, a spiral type1 and a cylin-

drical type.6 Figure 7-6 indicates a drawing cross section of the spiral

type of fission counter. Two separate foils, usually coated with uranium,

are wound concentrically and electrically insulated from each other with

a very small spacing of about 0.02 in. The spirals are mounted into a
Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
REACTOR CONTROL RADIATION DETECTORS

195

100

10

-NEUTRONS:

-ALPHAS

ure 7-9 shows an integral bias curve for a fission counter which uses a

uranium 235 coating. As this material gives off a spontaneous alpha

particle of fixed energy, the tube circuit must be biased to select pulses

above a given amount to eliminate counting these alpha particles. A

ratio range 10:1 to 100:1 in neutron 500

to alpha-particle output is usually set

up by the pulse height selector circuit.

lonization Chamber. In the period

range and power range, neutron-

sensitive ionization chambers are nor-

mally used for neutron flux measure-

ments in reactors. These are the

instruments which are usually con-

nected to the automatic control sys-

tem as indicators of power level.

The detector consists of chambers in

which large surfaces may be coated


Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

with boron 10 and the ionization

from the recoil alpha particles again

collected as current.1

The range of intensity over which

neutron flux may be detected is

limited by the gamma-ray background

in which the chamber must oper-

ate. The ratio of neutron flux to gamma flux at the normal operating

position of an ionization chamber in a reactor core or shield structure is

usually such that the signal caused by the neutron flux is about 1,000

times that caused by the gamma rays. Therefore, when the reactor is in

the power range, gamma-ray background is no problem. However, the

gamma-ray intensity does not vary directly with reactor power as does

the neutron intensity. Thus, a lower limit is set for neutron detection

at the level where the ionization caused by neutrons equals that caused

by gamma rays.

lonization chambers at present are of the parallel-plate type or cylin-

drical type. That is, the internal electrodes may consist of a series of

parallel disks or concentric cylinders or cups coated with boron. The

literature6'26'27 contains in detail the proper design procedures to obtain

maximum sensitivity from an instrument. The tubes usually may be

treated as constant-current sources having characteristics as shown in

Fig. 4-16.

100 150 200

PULSE HEIGHT SELECTOR

SETTING, ARBITRARY UNITS

FIG. 7-9. Integral bias curve for ura-

nium 235 coated cylindrical fission

counter, showing counting rate as a

function of pulse height selector setting.

Compensated lonization Chamber. The neutron-detection ranges of an

ionization chamber can be increased by gamma-ray compensation, that

is, by balancing out the component of the signal caused by the gamma
Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
REACTOR CONTROL RADIATION DETECTORS 197

Neutron Thermopiles. The instruments just mentioned are the ones

principally used in reactor control. However, many other devices have

found considerable use. One of these detectors is the neutron thermo-

pile.9'10'27 This instrument consists of several thermocouple junctions

connected in series, alternate junctions of which are coated with boron.

The instrument is a small one and in the past has taken two forms, one a

cylinder roughly the size of a fountain pen, the other a disk about the

size of a half dollar. These instruments are rugged and reliable and do

not require the high-voltage insulation of the previous instruments.

One disadvantage of the instrument is that its design usually must make

a compromise between sensitivity and response time. And as the sen-

sitivity of the instrument is quite small because the volume of boron

present is small, most of the instruments attempt to favor increased sen-

sitivity. Consequently, time constants in the order of one to several

seconds result and the instruments have not been used in the past for

direct reactor control. Another disadvantage of these instruments is

that they operate in quite high neutron fluxes and ultimately suffer from

boron depletion.

Scintillation Counter. Although the scintillation counter has received

wide publicity as a prospecting instrument and laboratory tool, it has not

as yet seen much service as a reactor control device. The principal

reason is that most scintillation crystals or liquids which are sensitive to

neutrons are also very sensitive to gamma rays. Also, the multiplier

portion of the detector is very susceptible to heat in that the noise level
Generated on 2011-10-24 11:53 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

of the tube increases greatly with temperature. As the vicinity of most

power reactors is thermally warm, the scintillation counter is at a

disadvantage for power reactor usage. However, scintillation counters

are used as gamma detectors in low-power reactor safety setups. Sev-

eral excellent descriptions of scintillation counters are found in the

literature.11~14

7-4. Effect of Gamma Radiation on Instrument Responses.f A prob-

lem always encountered in determining the usefulness of a neutron-

detecting instrument in a reactor system is the ability of the instrument to

operate under gamma irradiation. The existence of large gamma inten-

sities around a power reactor must be taken for granted, and provisions

made to live with them. The gamma radiation might originate in the

decay of fission products or from reactor materials which capture the

incident neutrons. The instruments are most likely to be affected by

capture gammas in the shield or the instrument itself. For most purposes

the gamma equilibrium radiation at any instant in the reactor operating

cycle can be considered as a first approximation to fall off exponentially

f Sections 7-4 to 7-7 were originally presented by J. C. Connor and the author in

Reactor Power Calibration, Nucleonics, vol. 12, no. 2, February, 1954.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

1000

2000

1000

20 40 60

DISTANCE FROM REACTOR, INCHES

FIG. 7-12. Gamma intensity versus distance, during reactor operation and shutdown.

with distance from the reactor. Gamma-level problems then are sub-

stantially reduced as the instru-

ment is placed farther away from

the reactor. Figure 7-12 shows a

typical gamma flux falloff with dis-

tance for a water-iron shielding mix-

ture under two operating conditions.

Figure 7-13 shows how the gamma

radiation at any given point in a

shield might drop off as a function

of time after shutdown of a reactor.

Here the initial sharp dropoff is a

function of the shutoff of the fission-

product gamma radiation. The

slow tail of the curve depends

upon the half lives of the mate-

100

10
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

0 20 40

TIME AFTER SHUTDOWN, HOURS

FIG. 7-13. Gamma intensity versus time

after reactor shutdown at a fixed point in a

reactor shield.

198

rials at the particular point under

consideration. The curve shown

is for a structural material contain-

ing iron and manganese.

Gamma radiation affects the response of the neutron-detecting instru-

ments or their circuits directly in an adverse manner. The effect of


REACTOR CONTROL RADIATION DETECTORS

199

gamma radiation is different on each of the instruments described in

Sec. 7-3. A brief discussion follows on how BF3 proportional counters,

fission counters, and ionization chambers are disturbed.

Present-day BF3 proportional counters suffer direct characteristic

changes from high gamma irradiations. Figure 7-14 indicates a typical

integral bias curve for a BF3 counter and amplifier operating with and

without a superimposed gamma field. With no gamma radiation present

10,0 00 r |-I

1000 r

NEUTRON RESPONSE WITH

BACKGROUND GAMMA ONLY

; 100 r

200 400 600

PULSE HEIGHT SELECTOR SETTING

(ARBITRARY UNITS)

FIG. 7-14. Integral bias curve for BF3 counter with and without gamma radiation back-

ground.

other than the normal laboratory background, a conventional plateau

type curve is obtained, giving counting rate as a function of pulse height

discriminator settings. As the gamma background is increased, the tail

of the curve at high discriminator settings begins to fall off and ultimately
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

reaches a point where no usable plateau range exists. This point may

occur between 50- and 1,000-r/hr gamma background, depending upon

the design of the counter. No permanent damage appears to occur to

the instrument at these gamma levels. However, it is known that at

106 r/hr some form of semipermanent damage does occur. After even

short exposures to these very high levels the counters do not operate

properly, and several weeks may be required for the instrument to return
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

to its original operating characteristics. BF3 counters are ordinarily

not used at such high gamma levels, but the damage may occur even if

the counter is disconnected and sitting in this ambient. The mechanism

of this instrument damage is not fully understood but is related to decom-

position of the BF3 gas and the amount of gas amplification which is

obtained in the tube.

Fission counters have been irradiated in gamma fluxes approximating

5 X 106 r/hr. No permanent dam-

age occurs under these irradiations.

However, the electronic circuits

which connect to the fission counter

suffer a form of partial blocking

because of the immense gamma

pulse pile-up. In order to minimize

the blocking effect it is usually

necessary to use extremely short

time constants in the electronic cir-

cuits. These differentiating net-

works are generally placed as close

to the detecting instruments as

possible.

Figure 7-15 indicates the per-

formance of a typical fission counter

under high gamma radiation con-


Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ditions. From the curve it can be

seen that operation at higher pulse

height discriminator settings reduces

the gamma contribution to the

counting rate much more rapidly

than the contribution of the neu-

trons. Any reduction in neutron

sensitivity caused by operating at

higher pulse height settings must

NEUTRONS+a ;

.+NOISE

7+NOISE,

GAMMA LEVEL

=1.7x105 R/HR -

100 150 200

PULSE HEIGHT SELECTOR SETTING

(ARBITRARY UNITS)

FIG. 7-15. Integral bias curve for fission

counter operating under gamma radiation

background.

be charged directly to the gamma background radiation.

lonization chambers, as described in Sec. 7-3, collect all ion pairs

created within them. A current proportional to neutron flux is usually

obtained by a secondary process such as the ionization from the alpha

particle emitted by the reaction of a thermal neutron with boron 10.

This secondary process must compete directly with the ionization caused

by the gamma flux inside the volume of the chamber. An ionization

chamber may be designed which will have a 3- to 4-decade neutron range

in the presence of a gamma field of 200 r/hr. To extend this range

farther it is necessary to compensate the instrument against the gamma

200
REACTOR CONTROL RADIATION DETECTORS

201

radiation. This compensation is usually accomplished by providing two

operating chamber volumes—one volume which is sensitive to neutrons

plus gamma rays and the other sensitive to gamma rays alone. A sub-

traction process then enables the instrument neutron range to be extended

for roughly 2^ more decades. A typical response of an ionization

chamber operating on a reactor shutdown problem is shown in Fig. 7-16',

and it will be noted that the amount of compensation shifts as a function

o 10"

CD

10"

1CT'

•UNCOMPENSATED

-COMPENSATED

AFTER LONG-

VX^-TIME REACTOR

\V IRRADIATION

ORIGINAL —

COMPENSATION

0 5 10 15 20 25
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

TIME AFTER REACTOR SHUTDOWN (MINUTES)

FIG. 7-16. Ionization chamber current versus time after reactor shutdown for various

degrees of compensation.

of time. This shift is usually caused by changes in the gamma gradient

adjacent to the instrument as the reactor builds up the longer half-life

emitters to a fixed activity.

7-5. Effects of Reactor Operation and Rod Shadowing on Neutron

Measurements. Having once established the calibration and placement

of the nuclear detector, it is necessary to see how this calibration changes

with time and operating conditions. For proper calibration of a thermal-

neutron detector, the thermal-neutron flux at the detector location must

be a function of power only. As far as the response of a particular

instrument is concerned, the reactor may be regarded as a neutron source

surrounded by an attenuating medium. Consider for a moment that


202 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the reactor is a point source at a fixed position emitting a constant number

of neutrons each second. The neutron attenuation between the source

and a particular point in the shielding structure depends upon distance,

the scattering and absorption cross sections of the shielding materials,

and the physical arrangement of the shielding structure. Since a reactor

is far from a monoenergetic neutron source, it is clear that even if the

rate of emission of neutrons remains constant, changes in their energy

distribution will result in flux variations at the detector, because the

neutron cross sections of the shielding material are a function of energy.

In addition to energy variation problems, a reactor ordinarily cannot

be considered as a point source of neutrons unless the distribution of

neutrons as a function of their position in the core remains constant for

all operating conditions. As has been mentioned, a power-producing

reactor must contain enough excess fissionable material to maintain crit-

icality as fuel depletion occurs. Fuel depletion reduces reactivity, and

if the reactor has control rods of any type, the depletion may be compen-

sated for by rod motion. A rod movement can cause a change in the

neutron flux distribution within the core. It may be necessary or desir-

able to move control rods for any one or a combination of the following

reasons:

1. To calibrate control-rod effectiveness or reactivity

2. To compensate for fission-product poisoning

3. To compensate for fuel depletion

4. To compensate for reactor temperature changes


Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

5. To change the flux distribution within the core in such a manner that

fuel depletion occurs uniformly or in a prescribed manner

6. To change the power level

The relative effect of these rod movements on neutron instrument

calibration depends principally upon the design of the reactor in question.

Figures 7-17a and 7-17& are two views of a thermal reactor containing

four absorber type control rods which can move parallel to the Z axis.

Loss of reactivity for reasons such as fuel depletion or fission-product

poisoning must be compensated for by withdrawing these control rods.

The curves to the right of Fig. 7-17a show what might happen to the neu-

tron flux distribution along the Z axis in the core if all four control rods

are moved uniformly from position A to position B. It is apparent that

the reading of a neutron detector located beneath the reactor decreases,

since the peak in the flux distribution shifts up and away from the instru-

ment. The reading of a detector located on the X axis would increase,

since the flux peak moves toward the axis of the instrument.

The curves plotted to the right of Fig. 7-176 show what might happen

to the neutron flux distribution along the Y axis for a nonuniform rod

withdrawal. Curve C shows the symmetrical flux distribution present


REACTOR CONTROL RADIATION DETECTORS

203

if all rods are at a uniform height. Curve D shows that the peak in the

flux distribution shifts toward a detector located on the Y axis if reactiv-

ity changes are compensated by withdrawing rods 1 and 2 while rods 3

and 4 remain at the same height. Other rod withdrawal programs might

be hypothesized, but it is clear that no detector location or rod program

exists for a reactor of the design shown which is not subject to some

neutron-detector sensitivity shift because of changes in flux distribution.

Instrument calibration changes caused by rod motion are called rod

shadow effects. Planning an experiment to determine the magnitude

of a particular effect may be complicated by the simultaneous presence

DETECTOR

BL

t—X

NEUTRON

FLUX

®DETECTOR

fa)

SIDE VIEW OF

FOUR ROD REACTOR

®DETECTOR

(b)

TOP VIEW OF
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

FOUR ROD REACTOR

FIG. 7-17. Reactor flux distributions as a function of control-rod positions.

of many other variables. Some idea of the order of magnitude of the

calibration changes because of rod movements alone may be obtained

from the following example. Suppose we have a reactor, as shown in

Fig. 7-18, containing four control rods equally spaced radially from the

center of the core. If the reactor contains a separate neutron source, a

particular instrument may have a given source-level counting rate with

the reactor subcritical and all control rods inserted. Figure 7-18 is a

plot of the counting rate data that might be obtained by withdrawing and

reinserting one rod at a time as a function of angular position of the rod

measured with respect to a line between the detector and the center of

the core. Note that the maximum effect occurs for those rods closest

to the detector. The dashed curve through the four rod position points

indicates the counting rates which might be obtained if the reactor had

more than four control rods.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Rod shadow effects of this nature are greater the closer the instrument

is to the core. At points in the shield far from the reactor, the neutron

flux is more closely proportional to the total number of neutrons in the

core. Unfortunately, choice of distance between core and instrument is

usually dictated by instrument sensitivity. The highest power level

detectors, however, may be placed farthest away from the core, thus

minimizing rod shadow effects for power operation.

/ii^V

ieo- — 1 — k~-j —

i— 0DETECTOR

EACTOR

l^

-90°

TOP VIEW

OF REACTOR

ROD z^x'

s'

'XNx

xv.v

^
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ROD 1 ^"'

^J3

^^ ROD 4

no=

COUNTING RATE WITH

n = COUNTING RATE WITH INDIVIDUAL ROD OUT

ei

5 MEASURED WITH RES

ALL RODS IN

PECT TO DETECTOR M

IS

-180°

-90°

+ 90°

+180°

FIG. 7-18. Magnitude of instrument reading change as control rods are pulled one at a

time.

7-6. Temperature Effects on Neutron-measuring Instruments. Most

pile operating conditions, in addition to causing control-rod movement,

create more complex disturbances on the neutron-detector calibrations.

For example, consider the case of a change in reactor temperature. The

neutron flux at a detector location in the shielding structure depends to

a large extent on the thermalizing of fast neutrons from the reactor in

the immediate vicinity of the detector. If the temperature of a reactor

changes sufficiently to cause the density of the moderator to decrease, it

is apparent that the number of fast neutrons reaching the detector will

increase. Changes in the energy distribution of neutrons in the reactor

may accompany temperature variations. The leakage flux of fast neu-

trons and hence the neutron-detector calibrations depend upon the mag-

nitude of these effects.

204
REACTOR CONTROL RADIATION DETECTORS

205

In addition, temperature may affect the instrument reading directly,

even assuming a constant flux. Figure 7-19 shows the operation of a

commercial BF3 counter at various temperatures. It can be seen that

unless great care is taken in the instrument design and usage, counting

rate is tremendously dependent on temperature.

100

2200

2800

2400 2600

VOLTAGE

FIG. 7-19. Characteristic curves of commercial BF3 counter as a function of temperature.

7-7. Instrument Calibration and Intel-calibration. Despite the many

difficulties pointed out above with neutron-radiation detectors, they are

still most useful, particularly once they have been directly calibrated in

terms of power level. Three basic types of power calibration are usually

available. These are foil measurements at low power levels, a direct

energy calibration generally made at low power levels, and a conventional

calorimetric calibration usually made at high levels. With multiple-

instrument arrangements such as are indicated in Fig. 7-1, intercalibra-

tion between any of these methods and the neutron detectors can be

accomplished.

Low-level Calibrations. Before bringing a newly designed reactor safely

to its rated power, it is desirable to check such items as the adequacy of


Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the biological shield, control effectiveness, and criticality configurations.


206 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Experiments of this type are most easily conducted at low power when

radiation levels around the core are low enough to permit physical access

to all parts of the plant. Data obtained at low levels can then be extrap-

olated to higher power operations. Thus instrument readings and cal-

ibrations at low power are important to assure that such reactor design

items as shield adequacy will be proper at higher power.

The first calibration method, experimentally troublesome but capable

of yielding reasonable accuracy in a thermal reactor, is to determine

the average thermal-neutron flux by irradiating foils at various locations

in the core for a known length of time and measuring the induced activ-

ity.2 The power P developed within the core is given by the product of

the number of fissions occurring per second, and Q, the energy release

per fission event.

P = nvZfQ (7-1)

where n is the total number of neutrons in the core, v is their velocity in

centimeters per second, and 2/ is the macroscopic cross section for fis-

sion. 2/ may be calculated for the particular reactor in question, and

Q (3.2 X 10~1 1 watt-sec per fission) is a known constant. Unfortunately,

the usual indium-cadmium-foil counting techniques determine only that

portion of the neutron flux below an energy of 0.6 electron volt. It is

possible, however, to measure average thermal flux to within 10 percent

by this method. A large number of measurements are required to deter-

mine average core flux if the core structure is at all complicated. Access

to the core for removal and insertion of foils must also be possible.
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

A second method of low-level calibration which does not require a

highly accessible core is to bring the reactor to exact criticality, insert a

neutron source of known strength, and plot the resultant rise in counting

rate of the neutron instruments as a function of time. By solving the

pile kinetic equations subject to the proper boundary conditions, it can

be shown that the rate of change of neutrons with time after an initial

transient, and consequently the rate of power increase, is linear and pro-

portional to the source strength. The following type of derivation may

be used. We may start with the familiar kinetic equations

dn Sk — 0 v „ „ m 0.

n + ) \iLi + b (7-2)

dt I*

t=i

dc< _ & „ > c (7 ^

— .--n-XiG; I,-.,\

where the symbols have their usual meanings and S is the neutron source

strength.

These equations are to be solved subject to the following conditions.

The source S is to be inserted into a critical reactor (Sk = 0) at time


REACTOR CONTROL RADIATION DETECTORS 207

zero. The boundary conditions are that when t = 0, n = n0, (7, = Cl0,

dn/dt = S, dd/dt = 0. The terms n0 and CM are the values of the

neutron density and delayed-emitter concentration present prior to the

insertion of the source and are constants.

We can first try a simplified solution using lumped delayed emitters,

having a single decay constant X « 0.1 sec~1. For this situation we find

that

n=

"' "

_ _ . __

[X + (ft/I*)]* X + (ft/I*)

when t "2> 10 sec. In this type of solution for large values of t and any

reasonable value of I*, the negative exponent of the transient term in the

solution becomes quite large, and after an initial waiting time, dn/dt

approaches the value

dn _ X „ ,_ ,

~dt X + (ft/I*) ° U"&)

In other words, after an initial transient, the reactor neutron level rises

linearly with time.

If the kinetic equations are solved using all six delayed-emitter groups,

the expression for n would contain six transient terms, each of which

decays after several half lives to a negligible value. Then, after a given

time determined by the half life of the longest-lived delayed emitter,

dn/dt is given by
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

i i 0i i 0i . i i 06 \ dn „ . .

1 • + + + ' + = s ("<)J

K%-S (7-7)

For a graphite-moderated reactor of I* = 10~3 sec, K « 100, and for

this type of reactor

lâ„¢ (7-8)

The expression for dn/dt may be integrated, giving

n = no + 10r*St (7-9)

where no is the neutron density at a time t = 0 chosen after the initial

transient is past and the reactor is rising linearly with time. For our

example, if D is the time required for the counting rate of a neutron-

measuring instrument to double

n0 = 10-2SD (7-10)
208 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

To maintain a consistent set of units, n must be the total number of

thermal neutrons in the core if S is the total source strength in the core

expressed in neutrons per second.

We must now relate power output to neutron level by means of Eq.

(7-1), P = nvI,fQ. In order to put the equation for P in a more useful

form, it is necessary to develop an equivalent expression for i>2/.

The mean lifetime I* of a thermal neutron in a given core may be

defined as

I* = - (7-11)

where ft is the mean free path and v is the velocity of a thermal neutron

in the core. The probability that a neutron will be absorbed is simply the

ratio of absorption to production. If 20 is the macroscopic cross section

for absorption in nonfissionable material or the probability of such absorp-

tion, then

= absorption

production

If t is the number of neutrons produced per fission,

envSf

Solving for «2/,

i>2/ = — or v"Lf = -p (7-14)

The equation for power level now becomes

P = ^ (7-15)

We can continue our example for the case of the graphite-moderated


Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

reactor and find that

The accuracy of power calibration measurements made in this manner

depends upon the validity of the assumptions made in the derivation, the

power level at which the measurements are made, and the strength and

calibration accuracy of the neutron source.

High-level Calibration. The difficulties experienced in obtaining a

power calibration at low levels largely disappear at high powers where

the reactor is developing sufficient heat to allow calorimetric measure-

ments to be made. The procedure is first to obtain intercomparison


REACTOR CONTROL RADIATION DETECTORS

209

data between whatever overlapping low- and high-level neutron instru-

mentation is present at some convenient set of power points. The high-

power instruments can be calibrated at a given reactor power output by

measuring coolant flow and temperature rise across the core, power being

proportional to the product of these two variables. Having obtained a

calibration of the high-level instruments, the low-level detectors may be

intercalibrated. The accuracy of power calibrations made in this man-

ner may be as good as that of the temperature- and flow-measuring

instruments.

It is of interest to examine how the thermal calibration standard itself

may vary as a function of time and operation and to see also the variations

THERMAL

POWER

CORRECTION

COMPARATOR

NEUT

POWE

~1

!0

RON I

* "" I AMPLIFIER

1 GAIN

| CHANGER
Generated on 2011-10-24 11:54 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

n<

BTU

CALCULATOR

,i

AMPLIFIER

_., 1

NEUTRON

DETECTOR

II

""1

REACTOR

ii1

HEAT

IXCHANGER

STEAM

PLANT

^ PHI

"\

IP]

FIG. 7-20. Block diagram of automatic calibration system for nuclear instrumentation.

that might occur in the process of intercalibration. Let us assume that

we have a nuclear power plant containing a neutron-measuring system

which is used to read the power output of the reactor. It is desired to

set up some form of continuous power-level calibration. Figure 7-20

indicates a block diagram of a nuclear power plant which attempts to

provide such a calibration. Thermal power in the reactor loop is

obtained directly by means of a so-called Btu calculator. This calcula-

tor computes the product of the coolant flow and the temperature differ-

ence across the reactor. Neutron power is measured by a neutron detec-

tor. The signal is amplified and then compared with the output signal

from the Btu calculator in a correction comparator. The output of this

comparator then adjusts the gain of the amplifier in such a manner that

the neutron power reads the same as the calculated thermal power.
210 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Let us examine now what the Btu calculator actually reads. It cer-

tainly reads all of the thermal input to the coolant loop. This includes

pump power, incidental beta-gamma power from the reactor, and fission

power from the reactor. The neutron detector reads essentially fission

power as modified by all of the local operating conditions previously men-

tioned. Let us assume that the amplifier gain changing mechanism sets

the neutron power measurement to be equal to the calculated over-all

thermal power measurement at full power output of the plant. In some

plants under these conditions, pump heat input might be estimated at 10

percent of the calculated thermal output, and beta-gamma power might

be 6 percent of the total output. The neutron instruments would then

be automatically set up to read high by a factor of 16 percent compared

with fission power.

Let us now suppose that a power demand change is made on the system

calling for 10 percent power output instead of 100 percent. Immediately

after this change is met by the plant, the beta-gamma power given off by

the reactor has not changed appreciably from its high power level and

might now represent a full 60 percent of the actual thermal power output.

The pumping losses at low power outputs can probably be considered

negligible if the coolant flow is reduced in the power-changing operation.

The neutron instruments will now be reading fission power of 4 percent

full power plus a 16 percent correction factor and would indicate that

the power output is 4.5 percent instead of 10 percent. Ultimately the

Btu calculator time constants, which should be quite long in order not to
Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

disturb the control system, will catch up with this situation and all meters

will read 10 percent. This correction might conceivably take as long as

J^ hr. Of course, at this new level of operation the beta-gamma power

would eventually decay out and ultimately become 6 percent of the new

power output level, and further continuous corrections would have to be

applied to the neutron-measuring instruments.

7-8. Instrument Circuits. The amplification, counting, and integrating

circuits which follow the detecting elements just described are a complex

subject, and considerable literature is available on their design and

usage.16'17 These circuits fall into two categories: pulse counting and

current integrating types. The essential thing, from a control point of

view, is that the output of these circuits is always a d-c signal which is

in some manner proportional to a neutron signal into the detector.

The circuits which are peculiar to reactor control are the period cir-

cuits which give outputs proportional to rate of change of neutron level.

Period outputs are obtained by first starting with a signal proportional

to level. The logarithm of this signal is then obtained in a logarithmic

amplifier, and finally the derivative of the logarithmic signal is taken to

provide an inverse period signal. The heart of this operation is the log-
REACTOR CONTROL RADIATION DETECTORS 211

arithmic amplifier which also is used to compress the wide range required.

Considerable work has been accomplished in the field of logarithmic

amplifiers.1s'19'23'24-2s

The principal design problem is the one of obtaining utmost reliability

in the circuits. Multiple circuit configurations are generally used,20 and

efforts are made to use as few vacuum tubes as possible.

REFERENCES

1. Rossi, B., and H. H. Staub: "lonization Chambers and Counters," McGraw-Hill

Book Company, Inc., New York, 1949.

2. Curran, S. C., and J. D. Craggs: "Counting Tubes: Theory and Application,"

Butterworth & Co. (Publishers) Ltd., London, 1949.

3. Korff, S. A.: "Electron and Nuclear Counters: Theory and Use," D. Van Nos-

trand Company, Inc., New York, 1946.

4. Wilkinson, D. H.: "lonization Chambers and Counters," Cambridge University

Press, New York, 1950.

5. Baer, W., and R. T. Bayard: A High Sensitivity Fission Counter, Rev. Sci. Instr.,

vol. 24, no. 2, pp. 138-140, February, 1953.

6. Goodman, C. (ed.): "The Science and Engineering of Nuclear Power," vol. 1,

Addison-Wesley Publishing Company, Cambridge, Mass., 1946.

7. Murray, R. L.: "Introduction to Nuclear Engineering," Prentice-Hall, Inc., New

York, 1954.

8. McCreary, H. S., Jr., and R. T. Bayard: A Neutron-sensitive lonization Chamber

with Electrically Adjusted Gamma Compensation, Rev. Sci. Instr., vol. 25, no. 2,

pp. 161-164, February, 1954.


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

9. Barbaras, G., et al.: The Design and Construction of Boron Coated Thermopiles

for Use in Neutron Fields, AECD-2485, 1949.

10. Barbaras, G., et al.: The Design and Construction of Boron Coated Thermopiles

for Use in Neutron .Fields, AECD-2975, 1950.

11. Marshal, F., J. W. Coltman, and A. I. Bennett: Photomultiplier Radiation Detec-

tor, Rev. Sci. Instr., vol. 19, pp. 744-770, November, 1948.

12. Gilette, R. H.: Calcium and Cadmium Tungstate as Scintillation Counter Crystals

for Gamma-ray Detection, Rev. Sci. Instr., vol. 21, pp. 294-301, April, 1950

13. Morton, G. A.: Photomultipliers for Scintillation Counting, RCA Rev. vol.. 10,

no. 4, p. 526, 1949.

14. Hofstadter, R., et al.: Detection of Slow Neutrons, Phys. Rev., vol. 82, p. 749,

June 1, 1951.

15. Schultz, M. A., and J. C. Connor: Reactor Power Calibration, Nucleonics, vol.

12 no. 2, February, 1954.

16. Elmore, W. C., and M. Sands: "Electronics: Experimental Techniques," McGraw-

Hill Book Company, Inc., New York, 1949.

17. Lapp, R. E., and H. L. Andrews: "Nuclear Radiation Physics," Prentice-Hall,

Inc., New York, 1954.

18. Meagher, R. E., and E. P. Bently: Vacuum Tube Circuit to Measure the Loga-

rithm of a Direct Current, Rev. Sci. Instr., vol. 10, p. 336, 1939.

19. Wade, E. J.: Instruments Used with Experimental Reactors, "Convention Record

of the IRE 1954 National Convention," pt. 9, " Medical and Nuclear Electronics,"

Institute of Radio Engineers, New York.


212 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

20. Stephenson, R.: "Introduction to Nuclear Engineering," McGraw-Hill Book

Company, Inc., New York, 1954.

21. Dunlap, G. W.: Detection and Measurement of Nuclear Radiation, Elec. Eng.,

August, 1948.

22. Cochran, D., and C. A. Hansen, Jr.: Instrumentation for a Nuclear Reactor,

Nucleonics, August, 1949.

23. Connor, J. C., R. G. Durnal, and V. G. Shaw: Photomultiplier Log Level-Period

Meter for Reactor Control, Nucleonics, vol. 11, no. 11, November, 1953.

24. James, W. G.: Logarithms in Instrumentation, ORNL 413, Oak Ridge National

Laboratories, Oak Ridge, Tenn.

25. Cooke-Yarborough, E. H., and E. W. Pulsford: An Accurate Logarithmic Count-

ing Rate Meter Covering a Wide Range, Proc. Inst. Elec. Engrs. (London), vol.

98, pt. 2, p. 191, 1951.

26. Jaques, T. A. J., H. A. Ballinger, and F. Wade: Neutron Detectors for Reactor

Instrumentation, Proc. Inst. Elec. Engrs. (London), vol. 100, pt. 1, p. 110, 1953.

27. Carmichael, H.: Design of the Chalk River Ion Chambers, Natl. Research Council

Can. Kept. 276.

28. Jordan, W. H., H. B. Frey, and G. Kelley: An Instrument for Measuring the

Logarithms of Neutron Level and the Period of a Pile, Oak Ridge Natl. Lab. Rept.

110, November, 1948.


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CHAPTER 8

OPERATIONAL CONTROL PROBLEMS: STARTUP

In considering the startup of a nuclear reactor the primary fact exists

that operation is required over many decades. As we have shown, the

range from zero to full power is a range covering nuclear fissions resulting

in a few neutrons per second to many billion neutrons per second. Unless

adequate control is maintained, particularly in the startup range, an

accident is most likely to occur. The primary reason that startup range

is the most dangerous one is because here the measuring instruments are

poorest. This is not because the instruments are poor in a physical

sense, but at low neutron levels the statistics of the instrument circuits

usually demand instruments having slow responses. In the event that

the neutron level in a reactor starts to rise rapidly in the startup range,

the information may not be received by a control or an observer until it

is too late to do anything about it.

The basic startup problem is first getting the initial neutron level up

high enough that the detecting instruments can measure their fraction

of the neutrons with good statistics. Many factors enter into setting

this initial startup level, and these will be examined in this chapter.

We are not concerned here with the startup of a nuclear power plant,

but rather with starting a reactor by itself. A power-plant startup is a

complete series of details depending upon the design of the plant and

generally can be considered as a form of preheating the plant. For

example, the temperatures of large metallic masses such as are involved

in shielding structures, pressure vessels, or boilers cannot be changed too


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

rapidly for fear that dangerous stress forces might develop. Conse-

quently, before a reactor is brought up to power, the plant is usually

made ready for it by proper preheating and pressurizing where

necessary.

There are essentially two types of reactor startups that will be de-

scribed: first, the initial cold startup of the reactor and, second, subse-

quent hot startups in a ready plant. We shall first discuss sources of

neutrons that set instrument levels, and then we shall examine these

213
214 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

types of startups. A large portion of the chapter will be spent on the

so-called startup accident. Finally we shall outline some of the systems

under which a reactor may be brought to power level.

8-1. Neutron Sources. As the power level of a reactor must be under

observation at all times, there must be available neutrons at all times for

the instruments to measure. Actually cosmic rays or spontaneous fis-

sion will create a few neutrons in a reactor, but their number is usually

too few to be satisfactorily detected. Consequently, a physical source

of neutrons is deliberately inserted into a reactor, and it is the multiplica-

tion of the neutrons from this source which is usually observed by the

startup range-detecting instruments. In a practical operating reactor

there are usually two other types of created sources of neutrons. These

are, first, photoneutrons created by gamma radiation and, second, the

delayed neutrons emanating from fission-product decays. We shall

examine these three sources in order.

Radioactive Sources. There are several nuclear reactions whereby a

free neutron is given off when a given substance is bombarded by another

particle. (a,n) reactions are very prominent, and the Be(a,n)C reaction

is the one generally used. The alpha particle can be obtained from the

natural radioactive decay of Ra or Po. The most common type of neu-

tron source is the Po210-Be source. This source is used in preference to

the Ra-Be source in large sizes because it is relatively inexpensive.

The number of neutrons emanating from a Po2IO-Be source depends

upon the number of curies of activity in the initial Po and upon the
Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

efficiency of converting the alpha particles into neutrons. Geometric

considerations are most important. In a typical source the efficiency

might be such that one curie of Po-Be can be made to give off in the

neighborhood of 2 X 106 neutrons per second.

Actually it is not this source strength which is important, nor even the

neutron flux created by these emanations, but rather the number of

counts per second or current produced in a neutron-detecting instrument

in a given fixed position with respect to the reactor. The detecting

instrument is usually calibrated in terms of neutron flux, and as has been

shown in Chap. 7, fairly elaborate means must be used for calibration.

It is quite obvious, however, that the larger the source rating in curies,

the more counts per second will be indicated by the detecting instru-

ments. The only limit to the size of a radioactive source is the practical

one of how large a source can be safely handled in the fabricating and

installation technique of placing the source into the reactor. On occa-

sion a small factor may be gained by splitting the source into a number

of discrete pieces, each piece being more easily handled than the over-all

source itself. Physically the source can be made quite small. Com-

mercial sources of neutron radiation are available as small capsules


OPERATIONAL CONTROL PROBLEMS: STARTUP

215

roughly J4 in- m diameter by a few inches long, or they are also avail-

able in thin sandwich type plates.

Because of the fact that Po210 is radioactive with a half life of only

138 days, the neutron emission from the Po-Be source drops off rapidly

with time. It is actually necessary to consider reactor assembly time

and rated operational life in the design of the low-level instrumentation

because of this source strength dropoff. Figure 8-1 indicates the rela-

tive source strength of a Po210-Be source as a function of time. If the

source must be installed in an early stage of the reactor construction, far

more sensitive instruments may have to be used. For example, if the

time from the initial insertion of the source to the end of the life of the

1.0

0.5

0.2

0.1

0.05

0.02

INITIAL SOURCE STRENGTH

0 200 400 600

TIME, DAYS

FIG. 8-1. Radioactive decay of polonium 210~beryllium source as a function of time.

reactor is 2 years, then the source will have fallen off by a factor of approx-

imately 30. Obtaining an instrument sensitivity increase of a factor of


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

30 may well be an impossible task. Consequently, it is of interest to

examine induced source levels in a reactor that has operated for some time.

PhotoneutronSources.2'4'B After a reactor has been runat power level, it

is desirable that some form of induced radioactivity be built up to take

the place of the decaying fixed radioactive source. As will be recalled

from Chap. 7, an inevitable large gamma flux exists around power reac-

tors. Use may be made of this gamma radiation in an effort to create

new neutrons by a (y,ri) reaction. In order to produce a neutron by

this method the gamma energy must be at least equal to the binding

energy of the neutron to the nucleus in question. This energy is about

8 Mev for most nuclei except for some of the light ones. Deuterium and

beryllium have threshold energies of 2.21 and 1.62 Mev respectively,

and for this reason they are feasible for use in induced neutron sources.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The type of neutrons emitted from these reactions is usually monoener-

getic and does not have the fission neutron energy spectrum.

The process involved in this artificial source is that following a shut-

down of the reactor after any appreciable power operation a large gamma

flux exists in and around the core. Now if the reactor contains either

beryllium or water in its structure in any form, new neutrons will appear

in this structure. Heavy water exists in normal water in a concentration

AFTER 10O HOURS

OPERATION AT FIXED

POWER LEVEL

AFTER 10 HOURS

OPERATION AT SAME

FIXED POWER LEVEL

AFTER ONE HOUR

OPERATION AT SAME

FIXED POWER LEVEL

AFTER 0.1 HOUR

OPERATION AT SAME

FIXED POWER LEVEL

20 40

TIME AFTER SHUTDOWN, HOURS

FIG. 8-2. Relative core flux after shutdown caused by photoneutron production in D2

for various power-level operation times.


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

by weight of 0.016 percent. In a water-moderated plant, many thou-

sands of pounds of water would be available and consequently a great

many deuterium nuclei would be available for the (7,71) reaction. The

cross section for the (7,71) reaction with deuterium is sufficiently high

that a large photoneutron source exists for a considerable time after

shutdown. It will be recalled from Fig. 7-12 that the gamma radiation

drops off rapidly after shutdown. It is consequently to be expected that

the photoneutrons would also drop off. Figure 8-2 indicates the relative

behavior of the photoneutrons as a function of time after the reactor

has been shut down. This series of curves shows that the number of

216
217

OPERATIONAL CONTROL PROBLEMS: STARTUP

10"

photoneutrons depends upon the past history of operation of the reactor;

and the longer the reactor has been operated at power levels prior to

shutdown, the higher will be the photoneutron flux level.

At present beryllium and water have existed in most of the reactors

designed, and consequently it has not been necessary deliberately to

design materials into a reactor for specifically increasing the number of

neutrons at shutdown. The curves of Fig. 8-2 also indicate that when-

ever possible after a reactor has been shut down, it should be started up

immediately. Obviously after a sufficiently long time has elapsed, the

photoneutron source will completely die out.

It should be noted in passing that gamma radiation existing in the

shield or external structure of a reactor container is not useful from the

standpoint of photoneutron production. Only neutrons that can be

multiplied by the reactor fuel assist the instrument-level measurements.

Fission-product Delayed Neutrons. After any reactor operation,

delayed neutrons are emitted from the fission products. These delayed

neutrons are given off at various times, but after a sufficient time interval

the number of neutrons present in

the core from fission products has

been shown to decay on a fixed 80-

sec period. After a power level


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

operation, then, in going to a shut-

down condition the initial reduction

in power level is a function of how

much negative reactivity is injected

into the reactor in a given amount

of time. After a few minutes, how-

ever, the 80-sec period decay pre-

dominates. Figure 8-3 indicates the

relative amount of fission-product

delayed neutrons decaying on this

80-sec period, and for most reactors

at the end of 20 min there is no

substantial measurable contribution

from the fission-product delayed neu-

trons. Another bench mark may

then be set for desirable reactor oper-

ation. That is, in the event a reac-

tor must be turned off from its power-level operation, all efforts should

be made to start it up again in less than 20 min in order to use the fission-

product delayed-neutron source.

10-

icr

«r

24

4 8 12 16 20

TIME AFTER SHUTDOWN. MINUTES

FIG. 8-3. Relative core flux after shut-

down caused by fission-product delayed

emission.

8-2. Initial Reactor Startup. After a reactor has been assembled, the

dramatic moment finally comes when it is necessary to start it up.


218 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Many questions are in the minds of the designers. The first question is:

Will the reactor go critical at the point that the calculations have indi-

cated? The reactor at this stage is presumably cold and at room tem-

perature. Any plant that might be used with the reactor is not in opera-

tion. The instruments are set at their most sensitive positions and

probably would be connected to scalers rather than counting-rate meters

in that higher sensitivities are more easily obtained. The use of the

scalers means that the operation will take a fairly long time. Low count-

ing rates will initially be used, and large time constants will be necessary

to obtain proper readings.

Bringing the reactor to initial criticality can be done in a number of

ways depending upon its design.1 Some reactors, such as the water

boiler, add liquid uranium fuel mixture to a container until a sufficient

amount is present that the reactor becomes critical. Other reactors such

as the NRX reactor add moderator in the form of heavy water to such a

level that the system becomes critical. Still other heterogeneous reac-

tors have fixed fuel and moderator and remove control rods so that the

multiplication is increased to where the reactor system becomes critical.

In any case, the effect is always the same. A multiplication factor is

brought from some fractional value to unity. As the initial operation

is apt to be a very slow and careful one, the performance of the reactor is

such that it appears to be operating on the subcritical multiplication

formula. That is, the instruments will be reading a signal proportional

to 1/(1 — fc), or the counting rate from an instrument channel will be
Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

c.r. - A r^1 (8-1)

where c.r. is the counting rate and A an instrument constant. As k

approaches unity very slowly, the counting rate approaches infinity.

It is customary to plot 1/c.r. as a function of the process that is changing

the multiplying factor. In this way, when 1/c.r. approaches zero, the

reactor approaches criticality. In a small water-boiler-type reactor the

plot might be 1/c.r. as a function of grams of uranium added. For

the large heterogeneous reactor the abscissa of the plot might be rod travel

in inches. Figure 8-4 shows such a plot for bringing a water boiler to

criticality. As the uranium is being added, the curve constantly can be

extrapolated to zero to predict where the reactor will go critical. The

shape of the curve as the reactor approaches criticality will depend on the

position and type of detecting instruments. If an instrument is located

very close to the reactor, it will effectively see the source for a long period

of time and the shape of the curve will be concave downward.

As will be recalled from Sec. 2-7, as criticality is approached in this

manner, the reactor will take longer and longer to settle out at a fixed
OPERATIONAL CONTROL PROBLEMS: STARTUP

219

neutron level or counting rate. At criticality the reactor level will con-

tinue to rise indefinitely because of the additive neutrons from the source.

Actually, criticality will be reached at such a high level that the source

probably will be contributing only in the neighborhood of a fraction of a

percent of the over-all counts read by the instruments.

If the source is removable from the reactor, then another technique

can be used. The reactor is brought close to criticality by the above-

described methods and then the source is quickly removed from the reac-

tor. If the reactor were exactly at critical, the counting rate would stay

-INSTRUMENT TOO CLOSE

TO SOURCE

INSTRUMENT

FARTHER AWAY

FROM SOURCE

THAN NECESSARY

0 0.2 0.4 0.6 0.8 1.0

KILOGRAMS OF URANIUM

FIG. 8-4. Plot of 1/c.r. versus amount of

uranium solution for a small water boiler

reactor, showing the effect of instrument

placement.

-CRITICAL

-SUBCRITICAL
Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

TIME

FIG. 8-5. Counting rate versus time after

source is dropped for a reactor at or near

critical.

constant. If the reactor were subcritical a slight amount, the counting

rate would drop off slowly, and if the reactor were supercritical, the level

would rise. Curves of the form of Fig. 8-5 might be obtained. In a

supercritical case, before any rapid rises are permitted to occur, some

reactivity would be removed from the reactor. Experimental reactors

would usually have provisions for removing the source. Large power

reactors probably would have fixed sources.

Once initial criticality has been obtained, it is desirable to examine the

reactor's characteristics. Measurements would then be made of such

parameters as the effectiveness of the control rods and their shutdown


220 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

capabilities, relative rod effectiveness, rod calibrations, instrument rod

shadowing, and other theoretical calculations.

8-3. Subsequent Reactor Startups. After the initial startup or several

initial startups of the type just described, confidence is gained in the

repeatability of reactor startup phenomena and the reactor can be taken

gradually to higher and higher power levels. After rated power-level

operation has been reached and the reactor then shut down, there are

many other contributors to the reactor shutdown neutron-level instru-

ment indications. The three source contributions will be as described

in Sec. 8-1. The second major effect will be the change in subcritical

multiplication caused by changes in reactivity, which in turn are caused

by temperature and poisoning.

The magnitude of these effects may best be seen by a simple example.

Let us assume that we have a reactor whose design is such that the nor-

mal cold shutdown multiplication factor fc = 0.9. The subcritical mul-

tiplication factor of this shutdown reactor would then be 10. Let us

now assume that the reactor has a negative temperature coefficient of

10~45fc/°F and that the normal hot operating temperature of the reactor

is 1000°F. Then an additional negative reactivity of 0.1 would be placed

into the reactor in changing its temperature from cold to hot. The

over-all hot k would now be 0.8, and the subcritical multiplication factor

would be 5. In other words, if the shutdown reactor were preheated

up to its normal operating temperature, the counting rates of the instru-

ments would drop off by a factor of 2.


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Similarly, as shown in Sec. 2-14, after a reactor has been operated at

a high power level for a while, the so-called peak xenon effect will build

up in accordance with a curve similar to that of Fig. 2-9. A peak of

negative reactivity would be inserted into the core at approximately

11 hr after shutdown. The amount of negative reactivity this poisoning

would insert would depend upon the reactor design, its sepecific power,

the length of operation at any given power level, and the time after shut-

down that the reactor is started up. It is well within conceivable prob-

ability to have the poisoning effect also change the counting rate up to

another factor of 3 or 4.

After long-term operation at power level some of the fuel has been

burned up. This uranium depletion also changes the subcritical multi-

plication factor but usually no more than a few percent.

We therefore find that the startup neutron level of a reactor depends

very much upon the time after shutdown the reactor is again started up.

The fixed source, photoneutron sources, delayed-neutron sources, tem-

perature, poisoning, and depletion all may change the readings of the

neutron-detecting instruments. The degree of variability may be plus

or minus from the initial startup cold level and may be measured in thou-
OPERATIONAL CONTROL PROBLEMS: STARTUP

221

sands of times the initial level. This degree of variability is a severe

headache to the instrumentation and control designer, and efforts to cut

this range by reactor design and operational procedures are usually

greatly appreciated.

Getting the Reactor Critical. A starting level having been established

at a given time, and with some knowledge and confidence of the reactor

and the plant, the first problem is: How fast should the reactor be brougth

up to critical? We shall look at

this problem from the reverse point

of view of seeing how fast the reac-

tor becomes critical under given

operating conditions, and we shall

later examine how proper operating

specifications and conditions should

be laid down.

Figure 3-9 is repeated as Fig. 8-6

in order that we may describe it in

terms of a startup problem. This

figure shows that at infinitely slow

rates of change in reactivity the

reactor never becomes critical and

follows a hyperbolic curve ap-

proaching criticality. At finite


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

rates of reactivity change the reac-

tor becomes critical at lower and

lower levels as the rate is increased.

At large values of subcriticality all

the curves merge and the rates of

change of reactivity do not matter.

-4 -2

PER CENT REACTIVITY

FIG. 8-6. Power level of a reactor as func-

tion of percent reactivity remaining in the

reactor for given linear rates of change of

reactivity.

Figure 8-7 indicates the period that

might be obtained by a given reac-

tor as a function of the reactivity

remaining in the reactor, as this

reactivity is inserted linearly up to the critical point. Here different

rates of reactivity are shown, indicating clearly the relationship of Eq.

(2-11) that the period is

P = d(Sk)/dt (8"2)

That is, the period changes linearly with the reactivity as long as the

reactivity is being inserted at a given linear rate. This equation holds

only for large values of subcriticality, as when the reactor approaches

critical at a finite rate the curves start to bend over, as is seen in Fig. 8-7.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Equation 8-2 may be combined with the subcritical multiplication

factor equation giving the level

L=

-_

and the relationship between period and level derived as

L=

[d(6k)/dt]P

(8-3)

(8-4)

In other words, any subcritical period may be obtained at any subcritical

level depending upon how fast reactivity is inserted into the reactor.

-12

-14

-4 -6 -8 -10

PER CENT, REACTIVITY

FIG. 8-7. Period versus reactivity of a reactor for given rates of change of reactivity.

Equation (8-4) obviously also does not hold when the level gets too close

to critical, but for operation at 60-sec periods or longer the formula is

usually quite valid. It can be seen that instrument sensitivity in counts

per second can therefore be swapped off directly against time. For

example, a given reactor with normal rod extraction rates might go

critical in 10 to 20 min. One can gain approximately a factor of 10 in

the instrument counting rate at any reasonable period by taking 2 hr to


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

get this reactor to its critical level.

222

Obviously, then, the reactor can also go critical at any desired period.

Figure 8-8 indicates the type of curve that can be obtained showing
223

OPERATIONAL CONTROL PROBLEMS: STARTUP

reactor period at criticality as a function of reactivity insertion at various

rates. This curve is based on approximate calculations for a given reac-

tor. This curve changes slightly as a function of I*, but the general shape

of the curve would be as shown in Fig. 8-8.

Rod Effectiveness. Until now we have been talking about linear rates

of change of reactivity. In a practical reactor, linear rates exist only

under short-term conditions. In problems involving large motions of

control rods, the reactivity as a func-

tion of time is mostly nonlinear.

These nonlinearities are important

from three points of view: (1) that

of startup, (2) that of automatic con-

trol, and (3) for scramming con-

siderations. In this section we shall

discuss only changes in rod effec-

tiveness from the startup point of

view.

The effectiveness of a control rod

depends upon its position in the core

and the value of the neutron flux at

that point of insertion of the rod. An

approximate formula can be used

that the effectiveness of a control rod


Generated on 2011-10-24 11:55 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

varies as the square of the neutron

flux in which it is placed. Thus a

rod or group of rods is most effective

at the center of a reactor and least

effective in the outer edges or in the

reflector. The control rods in a ther-

mal reactor reduce the reactivity in two ways. They absorb thermal neu-

trons at a rate which is proportional to the thermal-neutron flux at the

point in question before the insertion of the rod, and they also increase

the thermal leakage from the reactor. Physically the presence of the rod

decreases the number of fissions in its vicinity. Thus for a given total

power output the flux must be higher in some other portion of the reactor.

In a large reactor containing many rods the total effectiveness of a bank

of rods is only slightly less than the sum of the individual values for each

rod used alone. This is the condition whereby the rods are not too close

together, since the shadowing of one rod by another will depend upon the

distance between rods. For a small reactor the effectiveness of the rods

is not even approximately additive, as the insertion of a single rod greatly

distorts the flux distribution throughout the entire reactor.

0.16

0.14

0.12

0.10

0.08

j 0.06

0.04

0.02

0 4 8 12 16 20

CRITICAL PERIOD, SECONDS

FIG. 8-8. Period on which o reactor

would go through criticality by inserting

reactivity at given linear rates.

The effectiveness of a control rod then is -usually fairly small when


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

it is completely inserted into a reactor. It becomes larger as the rod is

extracted to approximately halfway out, and then near where it is com-

pletely withdrawn from the reactor its effectiveness again is quite reduced.

For approximate calculations a rod worth value of the form J sin2 x dx

can be used where x is the distance the rod is extracted out of the reactor.

This form is illustrated in Fig. 2-10.

From a startup point of view it is the rate of reactivity change of a

control rod when the reactor is close to criticality that is important.

From Fig. 8-6 it has been shown that at greatly subcritical operation

almost any reasonable reactivity rate produces the same results but close

to criticality the level and period are sharply dependent upon reactivity

TIME

FIG. 8-9. Power level versus time for a reactor being started up on either linear rate of

change of reactivity or linear rod motion.

rate. Therefore the effectiveness of the control rods must usually be

considered as that value of reactivity rate which the rods possess as the

reactor approaches criticality. If the rods are at their maximum effec-

tiveness, criticality will be reached at lower levels and shorter periods.

Another effect of the rod effectiveness is to lengthen the startup time.

With all of the variability in startup level that exists and additional var-

iability in rod worths and effectiveness as a function of position and poi-

soning, etc., reactor startup systems usually must be designed, for safety

reasons, on the basis of maximum rod effectiveness. Consequently, the

rods are at a lower effectiveness most of the time. Hence, it takes longer
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

for the reactor to attain criticality. Figure 8-9 illustrates this effect by

showing a reactor startup to criticality using linear reactivity rates and

then linear rod motion rates on a rod setup whereby the rods have an

224
OPERATIONAL CONTROL PROBLEMS: STARTUP

223

effectiveness of the form / sin2 x dx with the maximum effectiveness

equal to the previous linear reactivity rate. The length of time to reach

critical for this particular startup from —13 percent in reactivity is nearly

doubled.

It should be mentioned in passing that intermittent rod motion, such

as pulling a rod for a given length of time and then stopping and then

pulling again, etc., is sometimes used. For analysis purposes the rates

of change of reactivity and consequent levels and periods for this type of

motion can be obtained to a good degree of approximation by merely

taking the straight-line average of the intermittent reactivity change.

8-4. Operational Startup Requirements. The usual approach to a

given power level is on a period basis; that is, the reactor will be brought

140

60

40

20

TIME

FIG. 8-10. Period versus time for a reactor starting from shutdown and going to power

level.

up to its power operating point from criticality on perhaps a 30-sec period

or possibly a 20-sec or shorter period. As these values of period can be

reached before the reactor goes critical at given rates of reactivity with-

drawal, control-rod motion may be slowed down in actual practice or


Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

stopped before criticality is reached. Figure 8-10 indicates a possible

period versus time sequence in a reactor startup. From this figure,

once a 20-sec period is reached, either an operator or an automatic circuit

manipulates the control rods to maintain this period until the desired

power level is attained. In this type of operation criticality is of no

importance whatever. The reactor is brought through the critical

point without the operator's knowing or caring where it is.

There are several requirements upon which the startup time or the

required reactivity rates of change of a reactor control system may be

based. These considerations usually are of three types: (1) operational

considerations, (2) safety startup considerations, or (3) poisoning over-

ride considerations. We shall discuss these considerations in order.


226 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Operational Considerations. If one were permitted by operational

requirements to start up a reactor in a day's time, the operational and

safety requirements would be unimportant and only poisoning override

would have to be considered in startup problems. In discussing a power-

producing nuclear plant for electrical generation or a military propulsion

plant, it is quite apparent that this plant must be in a power operative

condition for as large a percentage of the time as possible. Scramming

such a reactor is most undesirable, and shutting the reactor down to fixed

source level is equally undesirable. If the reactor plant is such that it

must supply power continuously to some of its own auxiliaries, then a

fixed minimum operating level exists. This level might conceivably be

a few percent of the plant full-power output rating. It then would take

the reactor only a minute or so to get up to a useful output level from this

stand-by type of state. This step presupposes that the steam plant and

auxiliaries are always ready to run. Starting from source level is rarely

done for this type of plant.

On the other hand, research or experimental reactors need such opera-

tion cycles that are turned on every morning and turned off every night.

Startup times should be minimized so that the day is not wasted in getting

to an operational level. Other types of output programming are pos-

sible, and the operational startup requirements for each reactor must be

considered on an individual basis.

In order to get a feel for the numbers involved let us again examine an

elementary example. Let us assume that we have a reactor which for


Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

some reason must be brought up to power in 20 min. If one considers

a 10-decade startup, roughly 2 decades will be necessary to get to critical

and 8 decades will be used in bringing the reactor up to power. If the

power level is attained on a 30-sec period, 9.2 min will be consumed by

this operation. Therefore 10.8 min are available to bring the reactor to

critical. If the startup is made from a shutdown value of reactivity of

—13 percent, such as has been used in some of the previous examples, the

average reactivity rate required for this startup would be 2 X 10~45fc sec~1.

This elementary type of reasoning may be what determines reactivity

insertion rates for a reactor plant provided the safety and poisoning con-

siderations can be met.

8-5. Safety Startup Considerations. The fundamental premise upon

which safety startup considerations are based is that the speed at which

reactivity may be inserted into a reactor can be that speed at which the

protection system ensures a safe shutdown in the event of accident. The

faster reactivity is inserted into a reactor, the shorter will be the period

at a given power level and the greater the possibility that the reactor

power level might overshoot any safety stops. Normally it is assumed

that nothing would go wrong in starting up a reactor, but many possible


OPERATIONAL CONTROL PROBLEMS: STARTUP 227

accidents may be hypothesized. One of these accidents appears to be

much more severe than the others. This is the so-called startup accident,

and we shall examine it in detail as an extreme case.

The startup accident is that condition in which reactivity is continually

inserted into a reactor at a given rate and nothing is done to stop it. In

the case of a reactor being started up by withdrawing its control rods,

the situation might easily be hypothesized that a relay controlling a motor

driving the rods becomes stuck. The motor keeps turning, and the rods

are extracted from the reactor until they are all the way out. In this way

the reactor is brought from a subcritical condition through critical and

through prompt critical. It is presumed that safety stops or emergency

means of stopping rod motion are not available until the reactor power

level reaches a given overpower amount.

The problem then becomes: How bad is this accident, and what sort

of periods are involved? If the level of reactor power rises too rapidly,

an overshoot of the top protection level signal may result and the reactor

be damaged. The parameter that is important is the period attained by

the reactor in that this is a measure of the time available to do something

about the accident.

Two methods of analysis of reactor power level and period are available.

The first method is to attempt to solve the reactor kinetic equations for

various rates of reactivity change starting from various subcritical levels.

The method of solution would be similar to that outlined in Sec. 3-3.

This method is tedious and restrictive in that the entire range of probable
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

periods and levels involved cannot be seen without many numerical cal-

culations. Another method that might be used is to solve for the bound-

ary cases of maximum and minimum periods that may be involved in a

startup accident and to attempt to design a startup system to handle the

worst case.

We shall indicate here how to develop such period bounds, f Intui-

tively there will be two such bounds, the first an upper bound on how short

the period can get as the reactor becomes greatly supercritical and the

second a lower bound at greatly subcritical conditions. The upper bound

should be dependent in some way upon neutron lifetime I*. On the

other hand, for greatly subcritical conditions there must be a certain

minimum period involved which will depend upon the rate of insertion of

reactivity.

The upper bound will be handled in the following manner: We may

start by first rewriting the familiar pile kinetic equation

t The development of the period upper bound follows that of H. W. Newson in the

classified literature. The lower bound analysis presented was first accomplished by

F. Engel, Jr.
228 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

dn Sk - /3 . V % r< i a to a\

n + ) \id + S (8-5)

dt I

If we note that the last two terms on the right-hand side of the equation

are always positive, then

dn ^ Sk - /3 ,0 „.

_>__n (8_6)

As the period P is defined as

P = -T^TT, (8-7)

dn/dt

it follows that

7*

P<

for Sk > /?. For Sk < ft the inequality of Eq. (8-8) loses its significance

with regard to fixing the magnitude of the period. However, this inequal-

ity establishes a real upper bound for the period when the reactor is above

prompt critical.

We have then established a maximum limit on the period a reactor

can possess by saying that within a given upper criticality range the

period cannot be larger than a specified amount.

A lower bound to the period in the subcritical range may also be estab-

lished. It can be seen from the subcritical multiplication formula


Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

g£

P = ,,,.. ,,. that this lower bound will depend on the rate of change of

reactivity. Defining this rate of reactivity change by the symbol y, in

Sk sec~',

P = =^ (8-9)

as the subcritical relationship.

For reactors subcritical by a large amount it has been shown that this

formula is exact, but as the reactor approaches criticality, the formula

does not hold. The formula is inaccurate because it assumes that all

neutrons are effectively prompt neutrons. As criticality is approached

in a startup, the delayed neutrons become more and more important, in

that the rates of change of level become comparable with delayed-neutron-

emission times. It is felt intuitively from a control point of view that the

effect of the delayed neutrons is to slow down any level changes, thus

increasing the reactor period. Therefore Eq. (8-9) represents an intui-

tive lower bound. A rigorous analysis will now be presented which sub-

stantiates this feeling.


OPERATIONAL CONTROL PROBLEMS: STARTUP 229

For simplicity, consider the pile kinetic equations for the lumped-

delayed-emitter situation; that is

^=«*_IL2B + XC + S (8-10)

f = ^-AC (8-11)

For the subcritical condition the reactor must ultimately reach some

equilibrium. This equilibrium may be denned as when the concentration

of the delayed-neutron emitters C(t) is constant. Here neutrons are lost

at the same rate they are being produced.

Thus at equilibrium

= Jn + XC + (S = 0 (8-12)

and — = ^ n - \C = 0 (8-13)

CLv t

Substituting Eq. (8-13) into Eq. (8-12)

dn Sk dC . „ Sk . „ _

di=pn-lTt+s = i*-n + s = 0

n=

which must hold in the equilibrium state and is independent of whether

we use one lumped delayed emitter or six individual ones. If we now

make a small change in Sk about this equilibrium state,

(*-^)

\5fco + A5fc Sk0/

An = n - n(0) = SI* - ) (8-16)

The time required for Sk to change by an amount ASk is

, = ^ (8-18)
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

where 7 is the previously defined rate of change of reactivity. Then if

we are considering a startup whereby we are inserting reactivity, the

equilibrium value of the neutron flux level must be greater than the level

before equilibrium is established for any subcritical Sk. The greatest

possible value for the time rate of change of the neutron flux level then
230 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

would be

A. Sl*Mk y (8_19)

5fc0(5fco +

Sl*y

Sko(Sko -

If one obtains the time derivative of n in the following manner

the period lower bound becomes

(8-20)

n _ SI* (Skr _ _ Sk

dn/dt > Sk Sl*y y (*~~

This expression could have been obtained by a direct differentiation of

Eq. (8-15), but the above process appears to have greater validity.

An upper and a lower bound for the period of a reactor involved in a

startup type accident have thus been established. Figure 8-11 illustrates

these bounds for several conditions. It will be noted that the upper

bound depends only on I* and the amount of reactivity present in the

core at a given time. The lower bound, however, is a function of the

rate of insertion of reactivity. Figure 8-11 also indicates a possible

startup accident when reactivity is being inserted at the linear rate of

1.2 X 10~45fc sec~1 into a reactor having an I* of 10~4 sec. At large sub-

criticalities the reactor follows the lower period bound. Above prompt

critical the period closely follows the upper bound.

It will be noted, of course, that the upper bound is really in the wrong

direction as far as safety design considerations are concerned. That is,


Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

it would be desirable to know that a reactor could never get on a period

shorter than a given amount. However, above prompt critical, the

period follows the upper bound as developed so closely that this bound

may also be used as a practical value of the minimum period. Using this

limit type of approach, the severity and time involved in a startup acci-

dent may be seen without the necessity of performing a complete calcula-

tion of the kinetic equations for ramp input functions.

It is to be stressed that no temperature coefficient is considered in this

approach as presented. The lower bound picture would probably not be

affected by temperature coefficient, but the upper bound would be dras-

tically changed.

The period attained in the intermediate region between the bounds as

developed depends upon how far subcritical the reactor is at the start

and the reactivity change rate. Consequently, the period attained at


OPERATIONAL CONTROL PROBLEMS: STARTUP

231

any level above the startup level is a function of from how far below that

level the initial reactivity insertion started. Let us assume, for example,

that we have established by the safety system a protection level above

100

50

20

10

in

0.5

0.2

0.1

0.05

0.02

0.01

\II

V«—7=10~58k/SEC

PERIOD—A

LOWER \
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

BOUND \

-b. 6k V-

7=10-46k/SEC

i =1.2x10-4 6k/SEC-

l*=10-4 SEC

=10-38k/SEC

-4

-2

PERIOD

BOUND

r^ I*

r'

SEC

SEC

X10

,-s

0246

REACTIVITY, 8k

FIG. 8-11. Period as a function of reactivity indicating upper and lower bounds for a

startup accident.

which a signal is given for a scram or some emergency means to reverse

rod motion. A good example might be that when the reactor power level

exceeded 200 percent of full power, a signal would be provided to the

scramming system to insert all the control rods as quickly as possible.

The problem then exists of how much time is available from when the
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

dangerous level is noted by an instrument and the rods actually start

moving in. This available time depends upon the period that the reactor

is on at this level. In turn, the period that may be attained depends

upon how far below this protection level the startup accident began.

Figure 8-12 indicates the period attained by a reactor at a given protec-

tion level as a function of the decades of rise in reactor power level before

the protection level is reached. The curves again are for the example

that the reactor was originally —13 percent subcritical. It can be seen

from Fig. 8-12 that if only a few decades of power-level change are

-, 0 PERIOD ATTAINED, SECONDS

Db0PPr

.o ot .— w cn-'ro oiOC

INSE

-APPRO

\
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

EAcnvr

RTION

< 4xicr4

rv

WE---

Sk/SEC

REACTIVITY

--INSERTION RATE -

APPROX. 1CT4 Sk/SEC

>x

^^^ -

18 !

46 8 10 12

DECADES OF RISE

BEFORE REACHING A PROTECTION POWER LEVEL

FIG. 8-12. Period attained by a reactor during a startup accident at a given fixed pro-

tection level versus the number of decades below the protection level reactivity insertion

started. Startup from — 13 percent in reactivity. No temperature coefficient.

involved between the startup point and the protection level, the periods

attained are quite modest and a comparatively long amount of time is

available to initiate some protection device. On the other hand, if an

accident occurs when the reactor is started possibly 10 decades below the

protection level, very fast periods can occur, particularly at high reactiv-

ity insertion rates.

232

Once a period has been established in this manner or from boundary

conditions, then the amount of time available to do something about

protection is directly established. Figure 8-13 indicates the relative

power level that a reactor would attain above the protection trip level as

a function of the time delay in doing something about the rising power

level. In other words, while waiting for the protection signal to act
OPERATIONAL CONTROL PROBLEMS: STARTUP

233

important, as the power level could conceivably become very high during

this delayed interval. Conversely, Fig. 8-13 indicates that if a 200 per-

cent absolute power level is needed to protect a reactor, then if a protec-

tion signal is provided at 100 percent full power, the protection system

must reverse the reactivity and cause the power level to start coming

down with a delay of less than 0.7 sec, provided that the reactor is on a

1-sec period. We shall discuss these overshoots in greater detail in

Chap. 10.

1000

500

0.4 0.8 1.2

DELAY TIME, SECONDS

FIG. 8-13. Relative power level attained by a reactor on a given period as a function

of the time delay in doing something about it.

Level Considerations with Temperature Coefficient. The considerations

until now have been based upon the reactor having no temperature coeffi-

cient. We have seen in the case of automatic control that the addition

of a temperature coefficient has a tremendous effect. In startup control

also the no-temperature-coefficient case is a very pessimistic one, and

with any sized negative temperature coefficient the situation improves

considerably. It can be shown that for a given negative temperature

coefficient there exists a corresponding reactivity insertion rate under

which the reactor can be protected from a startup accident by virtue of


Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the temperature coefficient alone. Subsequently, for the same given


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

temperature coefficient there exists a higher reactivity insertion rate

during startup, whereby the reactor system can be protected either by

stopping the control rods or reversing their motion during a startup acci-

dent. Obviously a still higher reactivity insertion rate can be used in

starting a reactor, provided both a negative temperature coefficient and

a scram are used as the ultimate protection system.

NO PROTECTION

SMALL NEGATIVE

TEMPERATURE

COEFFICIENT

LARGER NEGATIVE

TEMPERATURE

COEFFICIENT

PROTECTION SIGNAL LEVEL

POWER

RANGE

TEMPERATURE COEFFICIENT

PLUS ROD REVERSAL

SCRAM

TIME

FIG. 8-14. Reactor power level as a function of time for several types of protection

systems.

Let us crudely examine the process that exists during the startup acci-

dent for several types of protection systems. Figure 8-14 illustrates some
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

of these situations. Let us assume that the reactor comes into the power

range on a given period. Curves similar to those in Fig. 8-12 can be used

to determine this period. With no protection system, obviously the

reactor power level would continue to rise on a fast exponential until the

reactor destroyed itself. If even the smallest negative temperature coeffi-

cient is available, then a form of protection exists in that ultimately the

temperature of the reactor will rise to such a point that sufficient negative

reactivity will be inserted into the reactor to overcome the positive reac-

234
OPERATIONAL CONTROL PROBLEMS: STARTUP 235

tivity put in by the startup accident. When this occurs, the power level

must reverse itself and the reactor will tend to reach an equilibrium posi-

tion between the startup accident reactivity and the negative reactivity

being inserted by the temperature coefficient. Eventually, however, if

rod pulling continues, the average temperature of the reactor will con-

tinue to rise to match this rod pulling and the power level will gradually

follow, assuming some simple coolant flow pattern. This second rise

following the first sharp peak is usually a long time effect taking in the

order of minutes, and presumably several different types of protection

systems may be employed to handle this second rise.

As the temperature coefficient is increased, the peak level reached in the

startup accident becomes smaller and smaller until with a very large

temperature coefficient the actual useful power output of the reactor

itself may be limited. As a matter of fact, in some reactors this is the

method of startup and operation. That is, the rods are pulled all the

way out deliberately and the power level rises to such a point whereby

the negative reactivity set by the average temperature comes into balance

with the positive reactivity set up by rod removal at some maximum

power level. However, for most reactors having large excess reactivities

there are temperature limitations that make this type of operation

dangerous.

If the temperature coefficient is insufficient to restrain the power level

to a safe amount under any attainable period, then it is necessary to apply

external means for protection. One simple means is to stop the rods
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

externally from moving by any convenient mechanical device. A second

scheme is to reverse the rod motion at the same speed at which the rods

are being extracted. The ultimate in protection is, of course, attained

by a scram in which all available reactivity is injected into the reactor as

violently as possible. It will be noted in Fig. 8-14 that for internal pro-

tection the power level remains high at some point where a balance in

reactivity can be obtained whereas for external protection schemes involv-

ing rod movement the reactor is ultimately shut down.

From these types of considerations design criteria curves against the

startup accident may be derived. Figure 8-15 indicates such curves.

It is first necessary to assume that a situation exists below which the

reactor is always protected or safe and above which some damage may be

caused. This situation may be defined in terms of reactor fuel tempera-

ture, moderator temperature, or power level. These parameters, of

course, are related. For convenience, the safe condition is usually defined

in terms of a power level. To go back to our previous example, the design

of a reactor may be such that below 200 percent of its rated power level

the reactor is always safe. If the power level is permitted to rise above

200 percent full power, something will occur that will damage the system.
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
OPERATIONAL CONTROL PROBLEMS: STARTUP 237

is a well-known phenomenon, as explained in Sec. 2-14, that after reactor

operation at power level, a given amount of so-called equilibrium xenon

poisoning builds up in thermal reactors. A thermal reactor also loses

reactivity for approximately 11 hr after shutdown, following which time

the reactivity is then returned to its normal value in approximately

60 hr after shutdown. Figure 2-9 illustrates this peak poisoning effect.

The peak poisoning after shutdown (as well as the equilibrium poisoning)

depends upon the previous long-time operating flux. If we define poison-

ing as the ratio of the thermal neutrons absorbed by the poison to the

TABLE 8-1

Long-time operating flux

Equilibrium poison

Maximum poison after shutdown

1012

1013

10"

2 X 1014

0.0054

0.028

0.045

0.046

0.0055

0.03

0.20
Generated on 2011-10-24 11:56 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

0.38

thermal neutrons absorbed in the fissionable material, Table 8-1, showing

the relative amounts of poison, can be computed from Glasstone and

Edlund.7

Once the poisoning has reached its peak, it starts to decay back to its

normal condition. During the time of this decay, reactivity is being

added to the reactor in a manner similar to actually pulling control rods.

The maximum slope of this reactivity insertion may occur at approx-

imately 20 hr after a power-level shutdown. If at this time the reactor

is at critical or close to critical, this can be a dangerous condition. The

addition of this amount of reactivity must be compensated for by the

insertion of control rods or other poisons, or else the power level will rise

rapidly under conditions similar to a startup accident. In other words,

the control rods must be capable of being inserted at a given minimum

rate in order to counteract the poisoning effect. We have previously

been discussing how fast the control rods might be extracted from a

reactor, but this present consideration concerns how fast they must be

inserted during startup.

A special extreme case exists, which may or may not be taken into

account, depending upon the automatic control philosophy selected for

power operation. The curve of Fig. 2-9 is for the case of poison build-up

with the reactor at zero power. The second case is the condition whereby

the reactor is turned on at the time peak xenon is reached and then

immediately run at full power. Here the xenon poisoning is burned out

very rapidly. Strictly speaking, this is not a startup problem and will

be mentioned again under operational power problems. Once full power


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

is reached, these fast rates of rod motion will have to be considered by

any automatic or manual control system.

To determine the speed of control rods for any of these poison condi-

tions, minimum rod effectiveness must be considered. We can conceiv-

ably get into a situation whereby it is necessary to move the rods to com-

pensate for a given amount of reactivity per second insertion by the

poison. At maximum rod effectiveness it might be very easy to over-

come this poison, but at minimum rod effectiveness it might well be

impossible. If the reactor structure is at all complicated, the rods can

conceivably be at any position at critical. Therefore provisions must be

made in the control system for maintaining at least sufficient rod effec-

tiveness which can be inserted at a given rate to take care of this poison-

ing decay during startup. Scramming to override this burnout is, of

course, undesirable.

EXCESS REACT VITY

10 20 30

TIME AFTER SHUTDOWN, HOURS

40

FIG. 8-16. Poisoned startup example in which the reactor is not capable of overriding

peak xenon.

Although the problem of safety does not exist, if one attempts to over-

ride the initial portion of the xenon build-up curve of Fig. 2-9 between

1 and 10 hr, there may be an operational reason for inserting reactivity

at a rate faster than the poisoning is inserting negative reactivity. The


Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

front edge of this curve has a slope approximately 10 times that of the

rear end, and if this poison effect must be overridden, a rate 10 times faster

than that previously required would have to be used.

238

Many reactors do not provide for complete xenon override. That is,

they do not have sufficient excess reactivity to overcome the poison even

when their rods are all the way out. For these reactors, in the event

they are shut down after an extended operational period, a startup must

be made within a given time; otherwise a waiting period is involved. Fig-

ure 8-16 indicates an example of this type of operation. Assume that a

reactor has built in a maximum excess reactivity of 10 percent and the


OPERATIONAL CONTROL PROBLEMS: STARTUP 239

xenon poisoning is capable of building up to 20 percent after shutdown

following a given power-level operation. Figure 8-16 indicates that

unless this reactor is started up within 1^ hr after shutdown, a wait will

be necessary until 31 hr after shutdown before the reactor can again be

brought to critical. Once any substantial power level is reached, the

poison will rapidly be burned out.

As also mentioned in Sec. 2-14, the problem of xenon override is partic-

ularly severe in reactors operating on an 8- to 12-hr day and then being

shut down. It is quite evident that this sort of operation requires that

the reactor always start up near peak xenon poisoning.

8-7. Startup Control Systems. The specifications for control-rod

speeds during startup having been examined, we can now briefly look at

startup control systems. The first and most obvious control system

would be a manual one. The problem is simply to increase the multi-

plication factor of a reactor to such a state that the reactor power level

rises on a given period until a desired power output is reached. Figure

8-10 indicates a period versus time curve for such an operation. A man-

ual control on reactor rod motion would simply be first to extract the

rods at a given rate until the reactor reached the specified period, in the

case of Fig. 8-10, in 20 sec, and then rod motion would be stopped or

changed manually to maintain this period. The operator would watch

two meters, a period meter and a power-level indicator. The reactor

would be maintained at the fixed period until the specified power level was

reached. Then the reactor would be returned to exact criticality, and


Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

this power level maintained.

Because of the hazards involved in manual reactor startup, some form

of override or interlock protection is usually provided. The most com-

mon system is of the permissive type. In this sort of startup operation,

control rods may be moved in or out provided the reactor period is not

shorter than a fixed amount. Once shorter periods are achieved, the

rods are automatically stopped by the interlocks or they may be moved

in, either manually or automatically. Figure 8-17 illustrates a permissive

type of operation about a 20-sec period. Here the reactor is started up

and is gradually brought from long periods toward a 20-sec period. For

periods longer than 25 sec the control rods may be moved either in or out

by an operator. As the period shortens to 20 sec, interlocks prevent the

rods from being withdrawn any farther and rod motion is stopped. In

the event that a 15-sec period is reached, interlocks take over and permit

control-rod motion inward only. This form of semiautomatic control

provides a simple, safe means of getting the reactor to power level.

Automatic Period Control. An automatic control loop can, of course,

be used to replace the manual or permissive operation just described.

Figure 8-18 shows such an automatic control system for period control.
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

This loop is very similar to the rod control loops of Chap. 4. The prin-

cipal difference is that instead of the level demand being placed into the

comparator, a period demand signal is used. For the example of Fig.

8-10, a period demand of 20 sec would be set into the comparator as a

fixed voltage. The reactor, through the neutron detector and period cir-

50

40

vi

£20

Q.

10

_PERM ISSIV E_ ROD_ MOTION,JN_0_R _OUT _ TO POWER

ROD MOTION STOPPED

LEVEL

PERMISSIVE ROD MOTION, IN ONLY

TIME

FIG. 8-17. Permissive type startup control. Period controls direction of rod motion.

CONTROL

RODS

REACTOR

NEUTRON
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

DETECTOR

ACTUATOR

PERIOD

CIRCUITS

PERIOD

SIGNAL

ERROR

SIGNAL

AMPLIFIER

COMPARATOR

PERIOD

DEMAND

FIG. 8-18. Automatic control system for period control.

240

cuits, would create a matching voltage, and any error signal would be

amplified to initiate control-rod motion to reduce this error signal to zero.

Similar types of analysis to those presented in Chap. 4 may be used to

determine the constants and compensations needed for this type of con-

trol loop. The error-signal amplifier may be of the on-off type or pro-

portional type.
OPERATIONAL CONTROL PROBLEMS: STARTUP

241

Combination Period and Level Startup Control.3^ It is desirable, if the

complexity of the control system permits, to provide a startup control

whereby both period and level are used to control the motion of the rods.

For the purpose of automatic control it is essential only that the reactor

be at such a level that good period information is obtainable from the

instruments. It is unimportant whether an initial startup or startup

after scram is involved. Consider the block diagram of Fig. 8-19a. This

control loop is similar to the previously described loop of Fig. 8-18, with

the exception that a level and a period detector and a level and a period

demand signal have been inserted to feed information into the compara-

tor. Let us now follow a startup program by observing the operation of

PERIOD

DEMAND

0-10 SEC

0-10 V

DESIRED POWER LEVEL

' CONSTANT

10 SECOND

PERIOD

TIME

(b)

(a)

FIG. 8-19. Combination period and level control, "magic-number" system, (a) Block
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

diagram of control system, (b) Approach to power-level operating curve.

this loop. If the reactor is initially at a very low level, we can assume

that no significant output is available from the level detector. We

might also arbitrarily assign numbers on the basis of the voltages that

one might expect in the comparator. That is, for example, a variation

of level from 0 to 100 percent full power might be interpreted in terms of

0 to 100 volts into the comparator. A variation in period or period

demand from infinity to 10 sec might correspond to a variation in voltage

from 0 to 10 volts. The level detector may be expected to give out infor-

mation only during the top 2 decades of power operation. If we now

start at some arbitrary low level, for example 8 decades below full power,

and we demand into the comparator a 10-volt signal corresponding, for

example, to either 10 percent of full power or a 10-sec period, then the

rods will be extracted from the reactor by the actuator mechanism which

has been given an error signal. This signal corresponds to the fact that

f This section was originally presented by the author in AECD-31&3, 1950.


242 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

insufficient voltage is coming out of the period detector to match the

10-volt input to the comparator. It will be recalled from Figs. 8-9 and

8-10 that even though the rods are pulled at some maximum permitted

speed, for a long time the reactor will not be able to attain a 10-sec period.

Consequently, a large error signal is maintained into the amplifier and

actuator mechanism, and the rods are removed from the reactor at max-

imum permitted speed. However, as soon as the period detector gives

out a 10-volt signal, the error is reduced to zero and the rod motion is

stopped or possibly reversed. This condition presupposes that the servo

loop has a fast time constant compared with a 10-sec period. The period

of the reactor has now leveled off at 10 sec and is maintained by the servo

loop. The reactor goes through criticality on a 10-sec period and on up

in power for approximately 4 more decades on this same 10-sec period

with the servo system acting completely on period information alone.

As soon as the reactor approaches the power operating range, some

information begins to come out of the level detector. This information,

plus what is coming out of the period detector, must now match up

against the 10 volts. This situation automatically calls for the output

of the period detector to be reduced, and consequently, the pile must now

operate on a longer period. As the level increases still farther, the servo

system gets less and less information from the period channel and more

and more information from the level detector. This action is as shown

in Fig. 8-196. Here the reactor comes up through the startup range and

period range on a 10-sec period, and this period slowly levels off into
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

infinity as the reactor approaches the desired power.

The system just described is a so-called magic number system wherein,

by the proper choice of voltages corresponding to period and level, only

one input is needed to the comparator. Other automatic systems are

possible, and a very flexible one is a system whereby a discrete period

and level can be demanded independently. A third system of merit is

one whereby period information is the controlling factor until the power

level is reached and then level information only takes over and period

information drops out. The system that the designer selects is one which

fits the time constants of the measuring instruments and amplifiers as

well as the desired operating program. For example, if the level informa-

tion is available from the instruments in a matter of milliseconds and the

period information is not available in less than 10 sec, it is futile to build

a fast servo system for level control and expect to be able to use period

information with this system. In general, with the instruments avail-

able at present, there is a higher resistance and capacitance time constant

associated with period information than with level information. If the

time constants of both pieces of information are comparable, then the

same speed servo system obviously may be used. If the two time con-
OPERATIONAL CONTROL PROBLEMS: STARTUP

243

stants differ widely, then modifications have to be made to the servo

system to set up a different response time in the period range from that

in the power range.

Another type of interesting comparator circuit for use when the time

constants are not widely different is shown in Fig. 8-20. Here power and

period information are received from the respective measuring circuits.

The circuits give voltage outputs proportional to neutron density and

reactor period. Period demand is varied by changing the amount of

reference voltage used to balance the period information. Power demand

POWER

INFORMATION

(IOOV = IOO%

POWER)

PERIOD

INFORMATION

(+IOV=5 SECOND

PERIOD)

POSITIVE

REFERENCE

VOLTAGE

NEGATIVE
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

REFERENCE

VOLTAGE

FIG. 8-20. Simplified schematic diagram of a comparator using both period and level

information.

is varied by changing the gain of the information received from the power-

measuring circuit. The diodes separating the demand circuits are used

in such a manner that only that voltage which is most positive will assume

control of the rod-actuating mechanism. The negative voltage reference

shown is to set up a zero position for accompanying amplifiers. A system

of this type effectively switches over automatically from period to power

demand at about 90 percent of the power demand level. Yet a rein is

held by the period control which prevents the reactor from ever changing

its level too fast.

REFERENCES

1. Murray, R. L.: "Introduction to Nuclear Engineering," Prentice-Hall, Inc., New

York, 1954.

2. Bernstein, S., et al.: Yield of Photoneutrons from U236 Fission Products in Heavy

Water, Phys. Rev., vol. 71, p. 573, 1947.


244 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

3. Schultz, M. A.: Automatic Control of Power Reactors, AECD-3163, 1950.

4. Bernstein, S., et al.: Yield of Photoneutrons from UMi Fission Products in Be,

AECD-1S33, Feb. 20, 1948.

5. Ergen, W. K.: Hard Gamma Emitters among Fission Fragments, ANP-59, May

3, 1951.

6. Moore, R. V.: The Control of a Thermal Neutron Reactor, Proc. Inst. Elec. Engrs.

(London), vol. 100, pt. 1, p. 90, 1953.

7. Glasstone, S., and M. C. Edlund: "The Elements of Nuclear Reactor Theory,"

D. Van Nostrand Company, Inc., New York, 1952.


Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CHAPTER 9

OPERATIONAL CONTROL PROBLEMS:

POWER OPERATION

In this chapter we shall consider the operation of a power reactor at

its rated power level. It will be recalled that the power range was denned

as the top 2 decades of output below the full power rating of the reactor.

It is the problems that occur within this range which will be discussed.

The operation of a stable reactor plant in the power range is a dull

process. Once the plant achieves this operating range, the temperature

coefficient plus the control system maintains any desired power level.

Occasional control-rod motions of a slow sort are required, and it is the

causes of these motions plus some of the physical means for creating the

reactivity changes that will occupy our attention.

9-1. Requirements for Reactivity Changes at Power Level. There are

six items that might cause a reactor system to require a change in reactiv-

ity at power level. These items are programming, temperature changes,

pressure changes, depletion, poisons, and flux distribution control. These

may appear singly or in combination and may be compensated for man-

ually or automatically. Each item is important, both from an opera-

tional point of view and from a design point of view, in that each individ-

ual condition sets a requirement for a maximum change rate of reactivity

that must be designed into the control system.

Programming. Problems in programming arise directly when a reactor

power-level change of any sort is required. The operational program

may require any of the types of temperature and pressure variations called
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

for in Chap. 6. The question to be answered is: What are the limitations

that should be imposed upon a reactor plant in changing from one power

level to another? In a large plant one usually looks outside the reactor

for the limitations. It has been shown that the reactor can change its

level quite fast. The boiler may have a temperature change rate limita-

tion, or the turbine may not be able to handle power as fast as the pri-

mary system can deliver it. In any event, it is reasonable to assume that

some type of limitation will be placed on the system.

245
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

In casually considering what sort of a limitation should be imposed

upon reactor power changes, we might first look at a period limitation

such as we had in the startup range. For illustrative purposes let us say

that the reactor plant should not be capable of having its power level

changed from 1 percent of rated output to full output in a time faster than

1 min with a minimum period restriction being imposed. Figure 9-1

illustrates the ultimate condition of making this entire level change at the

fastest possible period. From the figure this period can be seen to be

approximately 13 sec. This type of period limitation means that in the

first 13 sec the power level would increase from 1 to 2.7 percent of rated

100

50

20

10

50

60

20 30 40

TIME, SECONDS

FIG. 9-1. Permitted power-level change

versus time for a fixed minimum period

operation.

POWER LEVEL

FIG. 9-2. Permitted period as a function of

power level.
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

power and in the next 13 sec from 2.7 to 7.4 percent, and so on. In

other words, the absolute power level rises very slowly at first and then

much faster on a percentage basis as the level is increased. In the last

13 sec of the operation the power level would go from 37 to 100 percent.

This sort of period limitation is rather useless, then, as it appears to be

in the wrong direction. It is far more desirable to slow up the rate of

change of power level as full power is approached to prevent overshoot if

nothing more.

246

The reactor is a somewhat unique device in that it effectively operates

on an exponential basis rather than on a linear basis. If it turns out that

the reactor is the limiting component of the system in the speed of chang-

ing power level, then a combination of linear level and period type of

limitation may be usefully considered. It will be recalled from the

startup accident that the danger which exists in going from one power

level to another is a function of the distance between these levels. There-


OPERATIONAL CONTROL PROBLEMS: POWER OPERATION

247

fore if only a comparatively small level change is called for, this change can

be handled safely at a faster period than if a larger change is made.

Figure 9-2 illustrates a type of combination level-period restriction

that permits a given safe period operating area for each power level. At

any given operating power level, period signals longer than a prescribed

amount can be tolerated. As the power level approaches full power, a

minimum period situation is set up. However, if the reactor is at a low

power level, only slow rates of change are safely permitted. This system

has the same disadvantage as the one above of permitting an operation

into full power at an ever-increasing percentage change as a function of

time.

Most physical apparatus is apt to have a percentage power type of

limitation. A boiler might have a limitation that it can handle power

changes of, say, 1 percent of full power per second or a similar type of

rating. We can see, therefore, that operational power-level changes as a

POWER LEVEL OR TIME ROD POSITION

FIG. 9-3. Rod position used as a thermometer for constant-Ti, constant-coolant flow

program.

function of time should not have limitations similar to startup range

limitations. Instead, percentage change limitations appear more useful.

The maximum reactivity change rate required from the control rods

to change level depends directly upon the type of program. For the

constant-average-temperature program, with negative temperature coeffi-


Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

cient in the reactor, the level change can be made without any rod motion

at all. For other programs some maximum rate of reactivity change or

rod motion will be needed.

Temperature Changes. Changes in reactor temperature can cause

changes in reactivity and vice versa through the temperature coefficient

effect. We can see that the control-rod position can actually be used as a

thermometer for short-term variations once the temperature coefficient

is established. We might use an example with the constant-T» constant

flow program such as is given in Fig. 9-3. As power-level changes are

made, Tav is changed, thus causing a reactivity change and ultimately a

shifting of control-rod position to compensate for this temperature shift.

If the reactor in question has a temperature coefficient of 10~65fc/°F and

a control-rod worth over a small range can be considered linear and worth
248 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

10~65fc per inch of motion, then this control rod as a thermometer has the

calibration of 1°F per inch of motion.

Pressure Changes. If the reactor plant system is of the pressurized

type, using gas or water as a moderator, changes in pressure can also

cause changes in reactivity. Reactivity pressure coefficients are usually

quite small, and they may be either positive or negative.

There are many causes for a pressure change in a reactor primary sys-

tem. The principal one is likely to be the pressurizer control system.

Variations in the pressurizer tank level caused by the local control are

reflected back into the reactor. If insufficient damping exists between

the pressurizer and the reactor, it is even possible to have an oscillation

set up.

In a manner similar to the temperature effect, control rods may also

be used as pressure gauges for short-term operations providing the tem-

perature is held constant. For example, if a reactor has a pressure coeffi-

cient of 10~65fc/psi and a control rod has a linear effectiveness of 10~45/c/in.,

then the calibration of this rod in terms of pressure is 100 psi/in.

Depletion. Depletion, sometimes called burnup, does not usually fig-

ure as a factor in day-to-day reactor operation. The amount of deple-

tion in terms of reactivity is a major design problem concerned with

refueling cycles and how much burnup of fuel is permitted. From an

operational point of view the effect of depletion is noted only on a long-

time basis. Therefore, rod motions caused by depletion can usually be

neglected when compared with those caused by temperature and pres-


Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

sure changes.

Poisons. By far the largest factor in causing rod motion on a steady-

state basis is the build-up of poisons. It will be recalled from Secs. 2-14

and 4-4 that these poisons originate chiefly in a thermal reactor from

xenon 135 and the concentration varies both at a steady power level and

after reactor shutdown. Figure 9-4 indicates how equilibrium xenon

builds up during operation at a fixed power level. In this figure, after

8 to 10 hr the xenon reactivity change has reached a fixed limit, but

because operation continues at a fixed power level, a small amount of

depletion reactivity continues to be lost. In a practical operation one

cannot isolate the depletion from the equilibrium poison build-up, and

the solid line curve of Fig. 9-4 is the one usually measured experimentally.

Figure 9-5 shows again the familiar xenon build-up and decay curve

after shutdown. In the curve shown here the peak xenon builds up to

20 percent in reactivity. It was mentioned under the startup accident

that this peak xenon could be burned out very rapidly at full power and

that this rate in some reactors constituted a maximum requirement on

the rate of change of reactivity. Figure 9-5 indicates this rate of burnout

for full power operation and a corresponding burnout rate for a partial
OPERATIONAL CONTROL PROBLEMS: POWER OPERATION

249

power operation. If we scale this curve, assuming a 20 percent peak

xenon, we find that the burnout rate is approximately 7 X 10~65fc/sec

for full-power burnout. It will be noted that in the process of burnout

0 2 4 6 8 10 12 14

TIME AT A FIXED POWER LEVEL, HOURS

FIG. 9-4. Equilibrium xenon and depletion build-up versus time for a continued fixed

power-level operation.

-PEAK XENON

NORMAL DECAY

10

30

40

20

TIME, HOURS

FIG. 9-5. Peak xenon build-up and burnout for different power operations.

for the illustration given, the reactivity change drops below that caused

by equilibrium xenon during burnout but, after a long-term steady

operation at the burnout rate, equilibrium xenon level is again achieved.

The equilibrium xenon depends upon the neutron flux level as indicated

in Eq. (2-26).
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

It can be seen from Fig. 9-5 that the xenon reactivity is constantly

changing and calling for rod motion. A good example of how this level

changes on a daily cyclical basis can be seen from Fig. 9-6. Here a

program is assumed such that the reactor is on for 8 hr a day and turned

off for 16 hr. It is assumed that the on time is at full power level and the

off time is at 10 percent power level, possibly needed to supply the auxil-

iary system. The build-up of poison reactivity is given in Fig. 9-6 as a

function of this operating schedule. It can be seen that at no time during

the 24 hr does the reactivity remain constant.

100

Q.

10

EXTENDED

OPERATION

AT THIS LEVEL

XENON

HOURS

16

HOURS

HOURS

16

HOURS
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

FIG. 9-6. Basic xenon build-up and decay for square wave on-off operation.

Flux Distribution Control. Fundamentally the limit on the power

output from a reactor is the same as that of any other large piece of power-

handling apparatus. That is, the limitation is apt to be a hot-spot tem-

perature. Reactor flux distribution, hence temperature distribution,

greatly depends upon the position of the control rods. It is possible to

make a reactor critical with many combinations of control-rod configura-

tions. However, there are certain configurations that lead to so-called

flatter flux distributions, and these distributions permit higher power

outputs in that the peak-to-average ratio of the flux is minimized. Fig-

ure 9-7 indicates some of the flux distributions that can be obtained with

a fictitious three-rod slab type reactor. Figure 9-7a indicates the three

rods inserted all the way into the reactor, and the flux distribution is

examined at a cross section going through the middle of the slab. For

this type of rod configuration the flux is apt to be reasonably flat. Fig-

ure 9-76 shows the shape of the flux distribution for all the rods out. The

250
OPERATIONAL CONTROL PROBLEMS: POWER OPERATION

251

shape will be remembered as being roughly of the form of a squared cosine.

Figure 9-7c illustrates a situation whereby the two outer rods are in and

the center rod is part way out. The flux distribution here indicates two

peaks in the center of the reactor. Figure 9-7d indicates a tipped flux

type of distribution whereby a nonsymmetrical condition is set up. This

is the type of flux distribution which is usually avoided in power reactor

operation, and whenever this sort of situation appears, control rods are

usually moved to reflatten the flux. The reactor, of course, can be oper-

ated with a tipped flux distribution. Usually, however, it is necessary

CONTROL

ROD

r\

REACTOR

r\

I-NEUTRON

FLUX

(a)

r\

(c) (d)

FIG. 9-7. Flux distributions for three-rod slab reactor.

to operate at a lower power output because of the limitation on the hot-

spot temperature caused at the tipped peak. Obviously no speed require-

ment for changing the rods to compensate for flux distribution is involved.
Generated on 2011-10-24 11:57 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

An alert operator will occasionally trim rod positions to keep his reactor

always in the best possible flux configuration.

9-2. Automatic Control. The required reactivity rates for power level

operation will then be the sum of the six above-mentioned items. It

may seem possible to have all these items add up at the same time and

require very large reactivity rates of change. Actually, the time scales

involved are usually quite different. Programming, temperature, and

pressure changes are likely to be on a minute-by-minute basis, whereas


252 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the xenon poisoning, which may be the largest reactivity change in a

thermal power reactor, has been shown to be on an hour-by-hour basis.

In any event, a summation of the above items calls for a maximum rate

of change of reactivity to be designed into the control system. If the

reactivity cannot be changed fast enough during the xenon burnout, the

reactor could conceivably run away.

We have seen from Chaps. 4 and 6 that reactivity rates of change

corresponding to given rod motion rates constitute a gain term in a servo

loop. If this gain is too high, the system can conceivably oscillate.

Therefore the designer is faced with the problem of first summing up all

the reactivity rates required by the power-level changes and arriving at a

maximum rod motion rate. He then designs a control system in accord-

ance with the natural stability of his reactor and the required program

and comes up with a permitted maximum rod motion rate for stability.

He hopes that these two numbers overlap in the sense that his maximum

permitted reactivity for stability will provide him with plenty of reactiv-

ity change rate for his operational and safety requirements. If he has a

wide overlap, particularly one that is wide enough to take care of all his

variables, including rod effectiveness, his problem is finished. If a suit-

able overlap does not exist, he has several choices. First, he may rede-

sign the control system or provide proper compensation for this system

to permit it to operate at a higher gain. Second, he may insist upon

operational restrictions on the way the reactor will be run. For example,

if the limiting rate of reactivity is at full power burnup of peak xenon,


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

operational plans might be stipulated that the plant is not to be started

up at peak xenon. Or full power burnup will not be permitted at peak

xenon, but rather a slower burnup rate must be used. If these opera-

tional tactics are not sufficient, a third alternative is available, in that

manual overrides in the negative reactivity direction may be permitted.

Here, in the event that a combination of circumstances calls for it, rod

motion at a faster rate than normal is permitted in the negative reactivity

direction on a manual basis.

Regulator Rod. To accomplish automatic reactivity changes one con-

trol rod may be designated as a regulator rod. This rod is then tied to

the automatic control system loop. The regulator rod is designated as

such because it has special characteristics. First, the reactivity in this

rod is likely to be limited. Usually under no set of circumstances may

this rod be permitted to contain as much as prompt critical in reactivity.

This step is taken on the assumption that if the automatic mechanism

fails in some manner, prompt critical reactivity or greater cannot be

inserted quickly into the reactor. A practical number for the reactivity

of a regulator rod such as 0.0065A; may be selected. That is, the max-

imum rod travel is held to +0.0035A;.


OPERATIONAL CONTROL PROBLEMS: POWER OPERATION

253

Rod effectiveness is very important in the case of a regulator rod,

again because of the local control-loop gain. Figure 9-8 illustrates a good

regulator rod effectiveness setup in that here the regulator rod has a

restricted travel to limit its reactivity and its effectiveness does not vary

very much over its operating range. Figure 9-9 illustrates a poor regula-

tor rod setup. Here the regulator rod is forced to travel above and below

SHIM

SAFE

RO

TY^

fy

REGUl

ROD T

.A

RA

••-REGU.I

TOR — •


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

VEL ,

Al

OR RC

REACTOR

REGULATOR ROD

EFFECTIVENESS

FIG. 9-8. Regulator-rod effectiveness for shim rods all the way in.

SHI

SAFE!

OR

RODS

'

\_

Rl

RC

:GULATO

REGULATOR ROD /

EFFECTIVENESS /

V/

'

D TRAVE

REACTOR

\
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

rods may be used to follow up the regulator rod. Figure 9-10 crudely

illustrates this process. When the regulator rod approaches the end of

its permitted travel, switches are closed which operate a control system

to cause motion of the shim follow-up rods in the same direction as the

regulator rod. In this way the regulator rod will ultimately be returned

to its center position with the bulk of the reactivity being supplied by

the shim rods. For this type of operation the regulator rod has by far

the fastest motion in reactivity change per second. For stability pur-

poses it is usually desirable that the shim-rod follow-up reactivity be at

a rate between 10 and 100 times slower than that of the regulator rod.

REGULATOR

ROD L^

\J

SHIM ROD

CONTROL

\
Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

»

SHIM

(ODS '',

FIG. 9-10. Shim-rod follow-up of regulator-rod movement.

Under special design conditions a rate of only three to four times slower

may be used if needed.


OPERATIONAL CONTROL PROBLEMS: POWER OPERATION

255

motors are fed from a common power supply which has poor regulation.

In this way, as more and more rods are thrown onto the line, the motor

speeds become slower and the reactivity rate slows down, with the max-

imum reactivity rate obtainable being the maximum rate that permits

the system to operate safely.

MAXIMUM

PERMISSIBLE RATE

14

16

0 2 4 6 8 10 12

NUMBER OF CONTROL RODS USED

FIG. 9-11. Reactivity rote as a function of the number of rods moved.

REFERENCES

1. Harrer, J. M.: Controlling a Power Producing Reactor, Nucleonics, vol. 6, no. 3,

p. 58, 1950.

2. Trimmer, J. D., and W. H. Jordon: Instrumentation and Control of Reactors,

Nucleonics, vol. 9, no. 4, p. 60, 1951.

3. Schultz, M. A.: Automatic Control of Power Reactors, AECD-3163, 1950.

4. Cole, T. E.: Design of a Control System for a Low-cost Research Reactor, Nucle-

onics, vol. 11, no. 2, p. 32, 1953.

5. Commonwealth Edison et al.: "Reports to the United States Atomic Energy

Commission on Nuclear Power Reactor Technology," Government Printing Office,

Washington, May, 1953.


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

6. Moore, R. V.: The Control of a Thermal Neutron Reactor, Proc. Inst. Elec. Engrs.

(London), vol. 100, pt. 1, p. 90, 1953.

7. Glasstone, S., and M. C. Edlund: "The Elements of Nuclear Reactor Theory,"

D. Van Nostrand Company, Inc., New York, 1952.


CHAPTER 10

OPERATIONAL CONTROL PROBLEMS: SHUTDOWN

As has been pointed out, it is really impossible to shut off a reactor

once it has operated at any power level. Shutdown does not mean

removing the power to an amount that cannot be read by thermal instru-

ments but rather means shutting down to some sort of neutron-source

level. There are two types of shutdown that are of interest. One is

the normal reactor shutdown; the second is a shutdown under duress, the

so-called scram. Normal shutdown is a prosaic operation. The opera-

tor merely inserts the control rods at their usual rate, and the neutron-

power level falls off reasonably fast at first and then ultimately settles

down to falling off at the delayed neutron period of 80 sec. The problem

which is of much more interest is that of enforced shutdown. A scram

has been defined before as a violent insertion of control rods in an effort

to cut down the neutron level as quickly as possible.

In this chapter only enforced types of shutdowns will be discussed.

Shutdown philosophy and fundamentals of scram protection will be exam-

ined. Once these ground rules are laid down, there will be a brief discus-

sion of possible accidents and finally some of the elementary circuit con-

figurations used in scramming protection systems will be presented.

10-1. Shutdown Philosophy. There are two basic approaches that are

used in the development of a shutdown philosophy. The most prevalent

at this writing has been mentioned in Chap. 1 and is a semipolitical

approach. It is based upon the need of protecting a reactor plant at all

costs. Nuclear power plants differ from conventional power plants or


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

chemical plants in that an accident in a nuclear power plant would have

severe political repercussions as well as create local physical damage.

Because of this fact, the designers of all reactors to date have been

extremely safety conscious. As a matter of fact, this safety conscious-

ness is actually required by law. The use of scramming systems stems

directly from this all-pervading need to protect the plant, coupled with

the additional factor that in early nuclear power plants many of the

possible operational hazards were obscure. The scram therefore pro-

256
OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 257

vides an excellent hedge in satisfying legal requirements and as a contin-

gency against possible unknowns.

The second philosophy, with a diametrically opposed objective, is based

on the need for providing continuous service. As with the case of a radio

broadcasting station, the nuclear power plant must be kept "on the air"

at all times. If the nuclear plant is being used to supply electric power

to a consumer, the consumer expects to find voltage in the receptacle

every time he plugs into it. The excuse that the reactor has just been

scrammed because a pump failed would hardly satisfy him.

We have, then, the objectives of two philosophies. The first philos-

ophy says that the reactor must be saved in all failure events. Any plant

failure, any significant auxiliary component failure, or any item that could

conceivably cause radioactivity to get out into the atmosphere must be

so interconnected with the reactor control system that the reactor is

always shut off as quickly as possible. The second philosophy is the one

whereby each component in the plant system must be able to take care

of itself. If the turbine fails, it must protect itself; if a pump fails, means

must be provided for throwing another pump on the line. If a reactor

fails, it either must be designed so as to be able to protect itself internally

or it must have a local protective system. It is fairly apparent that as

the nuclear industry progresses, the tendency will be away from the first

philosophy and ultimately toward the more conventional second type.

In any event, a great deal of thought must be given to protecting the

reactor regardless of whether the reactor sits alone or is a component of a


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

complex plant. Three basic situations are involved, all of which can

be related. The first type of incident that may require the shutting

down of a reactor is that sort of incident which may be caused by catas-

trophe, sabotage, or military action. As the first large-scale nuclear

power plants are to go aboard naval vessels, it is most conceivable that

during the course of the operation of the vessel, naval action might rup-

ture the plant system. In cases of this sort, where the primary loop may

be violently torn apart, there really is not very much that can be done

toward saving the reactor and a discussion of scramming systems is

futile. Minor damage may possibly be relieved by scramming.

The second situation that may require a scram consists of a class of

accidents which can be prevented by cutting down reactivity quickly.

Accidents involving improper operation or malfunctioning of the control

rods, accidents whereby malfunctioning of an auxiliary system might

reflect back into the reactor, and any conceivable accident that creates

a rapid neutron rise may be prevented by proper rod motions or scram-

ming techniques. The distinctive fact about these accidents is that a

neutron-power level is involved. If the neutron-power level can be shut

down, the accident might be avoided.


258 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

A third set of conditions that may be hypothesized are incidents

involving the reactor and its structure, but not the neutron level. A good

example of this type of accident may be as follows: Let us assume that

in a nuclear power plant the coolant through the reactor is suddenly lost

for some reason. The reactor may be shut down neutronwise, either by

the loss of the coolant or by scramming the rods, but the reactor might still

be burned up by the excessive power coming from the beta-gamma radia-

tion. This power has been indicated as Pt ~ 0.07PoA°'2 where Po is the

long-time power level in watts and PI is the beta-gamma power t sec after

shutdown.12 A large power reactor might not be able to dissipate even

a small fraction of its normal power output without designed coolant

conditions, and thus the reactor might destroy itself even though the

neutron level had been shut off most satisfactorily by a scramming system.

The large majority of conceivable accidents are of the second type,

that is, the one in which the accident might be prevented by cutting down

reactivity quickly. Philosophically, it can also be argued that it does

no harm whatever to shut down the neutron level quickly in the event of

any real accident. We shall therefore consider in some detail this situa-

tion of quickly lowering the neutron level.

We have the further choice of three subphilosophies to consider in

preventing accidents: (1) fast insertion of control rods, (2) slow movement

of reactivity, and (3) continuous monitoring of the neutron level. We

shall discuss each of these types of operation below.

Fast Rod Insertion. Operating on the basis that a fast rod insertion
Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

is needed, we can examine the requirements of safety rods. We have

previously considered the requirements for normal speed of travel of con-

trol rods and found that there was usually a direct basis for setting this

speed. This basis was that each particular operating condition called

for a given maximum and minimum rate of change of reactivity. In the

case of safety rods, however, the approach historically has been a reverse

one. Rather than asking: What are the requirements of a scramming

system? the question is: How fast can one actually insert rods of a given

type into a reactor? This rate being the best obtainable has then become

the actual requirement. The time interval involved in starting the move-

ment of the rods and the rate of motion of the control rods are severely

restricted by mechanical considerations. Control rods are apt to be

relatively heavy pieces of neutron-absorbing material. Insertion time

limits in decades are likely to fit into the sort of pattern given below. To

insert rods into a reactor in say 10~3 sec is likely to be impossible. To

assure getting rods fully inserted in 10~2 sec would probably require

explosives. To assure getting the rods into the reactor in lO~1 sec would

require a very complicated mechanical design, but to get the rods into

the reactor in 1 sec is relatively simple and probably would not require
OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 259

more than a gravity drop. It can be seen, therefore, that most practical

scramming systems based on this reverse requirement will have an inser-

tion time between 0.1 and 1 sec. Fast rod insertion requirements will

gradually drift toward a more positive approach wherein the question is

asked: What is actually needed by way of scramming speed to protect

the plant adequately?

Slow Changes of Reactivity. Every reactor control designer has a men-

tal plaque over his desk that reads: "If nothing is permitted to move

fast, there can be no fast accidents." This statement refers to control

rods, turbine controls, valve openings, or power-level changes. Any

device that is inherently capable of quickly changing a plant loop param-

eter has the possibility of malfunctioning and causing a fast accident.

In Chaps. 4 and 6 it was emphasized that automatic controls should be

made as slow as possible. In Chap. 8 it was pointed out that in the case

of a startup accident, if reactivity were changed at a slow enough rate,

the natural protection of a temperature coefficient was sufficient to safe-

guard a plant without the use of a scram. The simple corollary is that

any reactivity rate of change should never be more than the protection

system can safely shut down.

There are two reasons why the ideal safe-slow condition does not

usually exist in a practical design. First, in order to be absolutely safe,

operational maneuvers such as startup would take a very long time.

If a plant required several hours to start up, its utility in a generating

system or naval vessel might be questionable. Actually, from a safety


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

point of view a plant that takes a long time to start up is not quite so safe

as first appears. It has been shown that in startup operations which

take longer than an hour or two, in which the operator sees effectively

nothing happening, he sometimes feels compelled to do something about

it. He examines his instruments and comes to a false conclusion that

possibly the plant is far more subcritical than it actually is. If such a

conclusion is drawn and the operator makes a mistake, severe accidents

are possible. For this reason, unless the plant is fully automatic, the

control designer sadly shakes his head at the realization that his plant

may be safer if he makes its controls just a little faster.

The second reason for faster control, particularly in the negative reac-

tivity direction, is that despite the fact that everything about the plant

may be designed to move slowly, there are just some unknown accidents

which conceivably could create a fast condition. The fear of the conse-

quences of such an accident causes a scram or some other means of fast

negative reactivity insertion to be designed into the system.

Continuous Monitoring and Correction. The third method of plant

protection consists of always watching the neutron level with fast instru-

mentation and always having available fast means of taking corrective


260 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

action. If a period becomes too fast or any rapid movement is made in

the plant, control rods are always available to initiate instantaneous

corrective action. Levels are not particularly involved. It is period

that is carefully watched. Any fast rate of change involves a fast rate

of change in the opposite direction.

This system suffers from the same difficulty as the previous ones; that

is, the failure of a fast corrective circuit is apt in itself to create a fast

accident.

10-2. Fundamentals of Scram Protection. The environmental usage

of fast safety rods will now be examined. In an actual reactor plant some

rods may be specifically designated as safety rods and others as control

rods. Another system which might be used is that all rods might be

safety rods. They are then withdrawn completely at startup, and the

reactor obtains power-level variations through the negative tempera-

ture coefficient. Still another more versatile setup exists whereby any

rod may be either a safety rod or a control rod, depending upon its elec-

trical circuit connections.

Let us examine a startup example involving a rod system of this third

type. Assume that a reactor has 16 rods and each rod is independently

worth 2 percent in reactivity. In a practical case the rods probably would

not be independent in that the position of any one rod would affect the

worth of the others. The amount a reactor is shut off when all the rods

are inserted is usually not known precisely because of the variations pos-

sible from temperature and poison. Assume, for our example, that the
Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

shutoff reactivity range is from 10 to 30 percent in negative reactivity.

Let us see what proper rod manipulations might be. If it were known at

all times exactly what the shutdown reactivity was, the problem would be

a simple one. Take the case of a 10 percent negative reactivity shut-

down. Here, if we desired to start up our reactor, the proper rod move-

ment sequence might be that four rods would first be withdrawn and

reconnected as a safety rod bank. The rule to adopt is that the safety

bank should always contain rods worth more in reactivity than those

which are capable of being moved in the reactor at any one time. To

provide some safety overlap we would then be permitted to move only a

maximum of three control rods when we have four of equivalent worth

in the safety bank. Practically, we might not wish to move three rods

at a time, as it is recognized, for this example, that the reactor would go

critical before three rods moved very far. However, the electrical or

mechanical interconnections should be such that with four rods out of the

reactor, only three should be permitted to move. As a matter of fact,

conservative design dictates that possibly only two rods should be per-

mitted to move, because the hypothesis may be made that the mechanism

for one of the safety rods might have failed.


OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 261

If it is known that the reactor in our example is subcritical by greater

than 18 percent, a safer situation exists in that nine rods may be pulled

out during startup to be used as the safety bank. Then, regardless of

how many other rods are moved in the reactor, there is always more

reactivity in the safety bank than can be taken out of the reactor by

improperly maneuvering control rods.

Unfortunately, unknown reactivity, loading, control-rod worth, and

poisoning usually exist. Proper instrumentation can give a feel for where

the reactor is, but as has been shown in Chap. 7, there are so many var-

iables which can affect the instruments that, at the present state of the

art, only a very rough idea is available in most reactors as to the exact

shutdown status. The only obvious safe type of startup operation then,

of the sort attempted above, is where each rod is gingerly removed at a

safe rate and committed to be a safety rod on the way up or as soon as it

is withdrawn.

In some power reactors it may be possible to "load" the reactor so

that it is always known that a given number of rods are capable of over-

riding the loading. Under these conditions, after a reactor has been shut

down, the safest operating condition is to extract immediately sufficient

rods to override the complete excess reactivity placed in the reactor by

the loading. Where this step is possible, safety rods are then cocked and

ready for any startup mishap. Startup accidents are likely to occur in

getting ready for startup in the process of cocking safety rods, partic-

ularly if the safety rods are capable of moving fast. It is rather intriguing
Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

to realize that a reactor is probably safer with some of its rods out than

with all its rods all the way in.

The question has not been discussed as to what constitutes a safe shut-

down. There is, of course, no universal agreement for all reactors.

Some number between 2 and 10 percent is generally agreed upon as being

a minimum safe value. It will be remembered that 2 percent in negative

reactivity represents a neutron multiplication of 50.

Scramming Parameters. In considering the type of scram protection

system that should be set up for a given reactor plant, the first fundamen-

tal which is encountered is that manual scram is always permitted. That

is, regardless of the condition of the plant or the status of its operating

cycle, an operator seeing anything peculiar always has the choice of turn-

ing the reactor off. Large scram buttons are placed convenient to the

operator at his control console, and in some installations scram buttons

are placed around the control room, particularly close to exit doors.

The most common type of scramming system is the so-called over-

power-level scram. Figure 10-1 indicates the operation that is involved.

It is assumed that for some reason the reactor gets on a fast period, the

curve of Fig. 10-1 being for a 0.1-sec period, and the power level of the
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

reactor quickly goes through the power range up through rated full power

to an arbitrary trip level. The trip level is indicated in this figure at

120 percent of full power, but this level may be set at any convenient

point. When the power level hits the trip level, an electrical signal is

given out which usually has to pass through such devices as a detecting

instrument, an amplifier, and a relay. The relay closes a set of contacts,

and rod motion is started. There is a time delay between when the power

level hits the trip level and when the rods actually start to move. Fig-

ure 10-1 indicates that even when there is no time delay of this sort,

there probably is some overshoot above the trip level because the power

0.1

0.3

O.4

0.2

TIME, SECONDS

FIG. 10-1. Overpower scramming operation. Reactor on 0.1-sec period, and trip level

at 120 percent full power.

level is moving fast, and a certain minimum amount of reactivity has to

be inserted by the rods before the slope of the power curve is turned

around. In other words, even with no delay time, the peak power level

reached will overshoot the trip level and a direct proportionality is usually

involved, that is,

PpeaK = KPirip (10-1)

where K is a function of the reactor period, the design of the scramming


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

mechanism, and the rod worths. It will also be noted from Fig. 10-1

that even a small delay of a mere 40 msec is sufficient to cause a tremen-

dous difference in peak power. Actually, however, it is not the peak

power that is destructive but rather the energy that is involved. The

energy in this case is the area under the curve in watt seconds. When

the energy gets above a given amount, the plant is damaged. The

energy is also usually directly related to the trip level.

262
OPERATIONAL CONTROL PROBLEMS: SHUTDOWN

263

E = Kip**, (10-2)

for a given reactor on a given period. Because energy and peak power

are thus related, through Eqs. (10-1) and (10-2), it is usually simpler to

describe accidents in terms of peak power rather than energy.

As far as limiting these peak excursions is concerned, it is most impor-

tant, as can be seen in Fig. 10-1, to cut down on the time delay involved in

the detection process and in the actuator mechanism. Having a large

amount of negative reactivity available to insert into the reactor at a fast

rate is also important. Figure 10-2 roughly indicates the effect of scram-

ming speed upon peak power attained for a level trip operation similar to

that of Fig. 10-1. Here a fixed small time delay is presupposed, and it

-TEMPERATURE COEFFICIENT ONLY

TRIP LEVEL

ROD INSERTION SPEED WHEN SCRAMMING

FIG. 10-2. Peak power attained versus rod speed when scramming. Level trip with fixed

time delay. Reactor on fixed period at trip level.

can be seen that without any scramming speed whatever, the peak power

reached by the reactor would ultimately be set by temperature coefficient

considerations alone. This figure does not indicate whether or not this

peak power would correspond to a destructive energy release. As rod

insertion speed is increased, the peak is steadily brought down to a value

very close to the trip level.

Figure 10-2 can also be interpreted as saying that in designing a scram-


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

ming mechanism against a rising level accident of this sort, there is no

point in designing the mechanism to move faster than a given amount.

Above a given rod speed the peak power reached always is roughly the

same. The major problem is usually one of taking every millisecond

possible out of the delay time and starting the rods moving as quickly as

possible. Rod speed is generally a secondary consideration. However,

studies of this sort are needed for many types of accidents if one is to

design a scramming mechanism on any basis other than as fast as possible.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Period Scram. We have seen that one of the difficulties "with level

scrams is that if they are fixed at a comparatively high level, they permit

the power level to attain momentum and overshoot the trip point. A

period scram, particularly in the source and period ranges, is more useful.

If the reactor starts to change its power level at a too-fast rate, the reac-

tor may be scrammed at a comparatively low level and the cause of the

difficulty located. Period scrams are not too useful in the power range,

as has been pointed out in Chap. 9. In this range the reactor is quite

safe using a level scram, as short periods cannot usually be attained

between levels that are close together.

The setting of the period scram trip point is not too important and

depends upon the type of reactor. Period scram settings between 1- and

10-sec periods are in common usage.

Although period scram circuits are usually arranged to trip on positive

periods, some reactor designers set their circuits to trip also on fast neg-

ative periods. The reasoning behind this setup is that if the reactor

power level is moving fast in any direction, there must be something

wrong with the reactor system and

consequently it should be shut down

and checked. The disadvantages

of period circuits have been pointed

out before.1 These disadvantages

exist because the period circuits

are inherently subject to circuit


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

noise and, because of the time

constants involved, they are apt

to give out erroneous answers in

certain types of power-level changes.

These disadvantages, coupled with

complex circuitry, sometimes out-

weigh the advantages of period

circuits, and consequently many

attempts have been made to elimi-

nate them.

Variable Level Scrams. One sys-

tem which at first appears to have

the advantages of a period scram

without being one is the so-called

variable level scram. Its operation

can be seen from Fig. 10-3 and is as

follows: Regardless of the power level of a reactor, a scram level is auto-

matically set up a given amount above this actual power level. This

amount above the actual power level can be on a linear percentage or log-

1,000,000 r

100,000 ?

10,000 r-

1000 =,

100 -

FIG. 10-3. Variable level scramming system

operation. Trip level twice actual power

level.

264
OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 265

arithmic basis. Then if the level of the reactor were to move slowly, the

variable level scram would move with it. That is, let the power level be at

100 units and the variable level scram initially set to trip at 200 units. If

the power level were to move slowly to 1,000 units, the variable level scram

would follow and automatically set itself up at 2,000 units. On the other

hand, if the power level were to change quickly, the variable level scram

would be designed so as not to be capable of changing particularly fast,

and consequently the power level would run into the tripping level for a

fast operation.

A crude elementary circuit for performing this type of manipulation is

indicated in Fig. 10-4. Let us assume that the power signal, either linear

or logarithmic, is converted into voltage and applied simultaneously to

the grid and cathode of a vacuum-tube circuit. In the plate circuit of

the vacuum tube is a scramming re-

lay, and the current relationships are

arranged so that the relay will close

when the grid and cathode voltages

are, for example, 1 volt apart. A

biasing battery is placed in the grid-

cathode circuit so that normally the

grid and cathode voltages differ by

2 volts. In the example indicated in

Fig. 10-4, the grid circuit is at an

absolute level of 8 volts and the


Generated on 2011-10-24 11:58 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

cathode circuit at 10 volts. Conse- RG 10.4 Elementary variab|e ievei

quently, there is insufficient current scramming circuit,

available to close the relay and no

scramming signal is given. A time constant network is inserted in the

cathode circuit, and in the event that power signal changes are very slow,

the capacitor involved charges up to the signal level plus the battery volt-

age. The grid and cathode voltages then effectively move together for

slow changes. In the event a fast voltage input change is involved, the

grid circuit changes directly but, because of the time delay, the cathode

circuit lags. Consequently, a trip will occur when the grid voltage and

cathode voltage are separated by the required 1 volt.

It can be seen that circuits of this type are not dependent upon the

power level but rather are dependent upon the rate of change of power

level. This is exactly the condition of a period circuit. As a matter of

fact, when most continuously variable level trip circuits are analyzed, it

turns out that they are really period circuits masquerading under another

name.

Multiple Fixed Level Scrams. Another useful type of scramming cir-

cuit of practical importance, particularly in startup problems where


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

manual operation is involved, is the multiple fixed level scramming cir-

cuit. It will be recalled that the difficulty with fixed level scramming is

that the reactor has the possibilities of attaining a short period by per-

mitting its power level to rise unchecked over several decades. The

multiple fixed level scram system is indicated in Fig. 10-5. Here fixed

scram levels are placed at convenient intervals, in this figure at every

decade. Manual operation of the reactor is presumed in startup. As

10,000,000 E

1,000,000

100,000

10,000

It

LJ

§ 1000

Q.

100

10

-FIXED SCRAM LEVEL 6-

-FIXED SCRAM LEVEL 5-

- FIXED SCRAM LEVEL 4 —

.FIXED SCRAM LEVEL 3—-

-FIXED SCRAM LEVEL Z —

-FIXED SCRAM LEVEL 1 —=

FIG. 10-5. Multiple fixed scramming level system.


Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

the reactor power level is brought up and fixed scram level 1 approached,

the operator checks to see that everything is under control and operating

properly. He then manually switches off fixed scram level 1, permitting

the reactor level to go up. As fixed scram level 2 is approached, a similar

process is gone through, and so on. In the event that the reactor gets

away, an early fixed scram level will catch it.

266

Other Types of Input Signals. Until now we have been discussing

scrams created by neutron signals. As neutron level and power are

directly related, this method provides an excellent means of keeping the


OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 267

reactor in check. However, if the philosophy is adopted that any major

difficulty in the plant shuts off the reactor automatically, other types of

input signals must be used. Some of these signals, such as a temperature

signal in the primary loop, can usually be related to power level, but there

may be a long time constant involved in the process. Some of the most

frequently considered types of external scramming signals other than

neutron-level, period, or manual signals are listed below.

1. Excessive core temperature

2. Excessive outlet coolant temperature

3. Excessive inlet coolant temperature

4. Low system pressure

5. High differential pressure across reactor

6. Pump failure

7. Electrical system power failure

8. Control-rod coolant system failure

9. Radioactivity level in secondary system too high

10. Radioactivity level in building too high

There are many other devices of this sort, all of which may be an indi-

cation that something is wrong in the plant. Generally, when a listing

of devices that can shut off a reactor becomes too long, it is found that

the cause of the shutoff is a malfunctioning of the device itself.

Rod Effectiveness. When examining reactor startup and reactor power-

level operation, we found that rod effectiveness was a very important

variable. For reactor shutdown problems as well, rod effectiveness plays


Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

an interesting role. The effectiveness of a shutoff rod enters in the follow-

ing manner. Let us assume that we have a reactor situation in which the

safety rods are fully withdrawn and their normal scram insertion method

is by gravity drop. When the rods are fully withdrawn, we can further

assume the situation given in Fig. 10-6. Here a reactor system is pos-

tulated in which the normal travel of the safety rods is 100 in. Fig-

ure 10-6a indicates rod position versus time for a simple gravity drop

containing no friction. Of course, the rods start slowly and gradually

increase speed. The rod effectiveness versus position will be assumed

to vary as the simple J sin2 x dx as previously used and shown in Fig.

10-66. Again the first several inches are useless as far as inserting much

negative reactivity into the reactor is concerned. And when one com-

bines these two effects of the initial slow rod movement rate and the low

initial rod effectiveness, one obtains the curve of Fig. 10-6c. Here it is

quite evident that the first 200 msec in time are practically worthless in

bringing negative reactivity into the reactor. This type of rod ineffec-

tiveness behaves very much like an additional delay in the scramming

circuit, and in situations such as that of Fig. 10-1, for example, a delay of

200 msec would be disastrous.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

There are two methods that may be used to combat this initial ineffec-

tiveness of the scramming rods. A simple expedient would be not to

pull the rods all the way out but to leave them partially inserted. In

the situation of Fig. 10-6, if the rods were pulled out to within 20 in. of

the top, a loss of only 10 percent of the total reactivity would be involved.

to

£ 80

~. 60

_j

LU

3 40

ac

^^"

---^

-""

^X

o 20

or

°C
Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

_— ==

>0

^—-•-

.1 0.2 0.3 0.4 0.5 0.6 0.

TIME, SECONDS

100

fek 80

|1«°

£0

S^ 40

0 _J

<r <

SS 20

__ — -

>

°<

) 20 40 60 80 100

ROD POSITION, INCHES

(b)

PERCENTAGE OF

TOTAL ROD WORTH

ro .& o> co c

0000 O C

x^"

x^

^x^

0.1

0.2

0.5

O.6

O.7

0.3 0.4

TIME, SECONDS

(o

FIG. 10-6. Rod effectiveness in a gravity drop scramming situation, (a) Rod position

versus time, (b) Rod worth versus position, (c) Rod worth versus time.
OPERATIONAL CONTROL PROBLEMS: SHUTDOWN

269

designed to attain a high initial velocity and then gravity permitted to

take over. A curve of the resultant operation is indicated in Fig. 10-7.

Here the reactivity, as a function of time, starts off in a satisfactory man-

ner. It is not important that the reactivity insertion rate slows down

from the initial rate because once the peak power has been contained,

the slope of the power versus time curve in most accidents will remain

negative.

100

5 80

ir

<n

u.

O 40

ui

5 20

or
Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

id

0.2

0.6

0.4

TIME, SECONDS

FIG. 10-7. Rod effectiveness versus time for initial motion accelerated.

10-3. Accidents. It has been pointed out previously that there are

really three categories of accidents which can exist about a reactor plant.

These categories are (1) catastrophes of sabotage or of a military nature

from which there is no protection scheme whatever available—a scram

was deemed advisable in this case merely because it could do no harm and

might possibly do some good; (2) accidents involving means of protection

other than scramming—loss of coolant was used as an example of this

type of accident, whereby some form of additional reactor cooling, either

in design or operation, had to be provided to prevent the accident; (3)

accidents involving reactivity changes directly or reflected. Here neu-

tron levels were involved and scramming would or would not be desira-

ble, depending upon how fast the level was changing.

This third category breaks up into two classes: first, those classes of

accidents which a large reactor negative temperature coefficient would

limit and, second, those classes of accidents which a large negative tem-

perature coefficient would assist. The example we have been using for

an accident that a large negative temperature coefficient would limit is

the startup accident or any similar type of accident whereby the neutron

level is capable of rising quickly. The reactor negative temperature

coefficient helps to limit this accident, because as the neutron level rises,
270 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the average temperature of the reactor system rises. Consequently,

negative reactivity is inserted into the reactor, which is thus ultimately

made subcritical and turned off.

The larger the temperature coefficient for this type of accident the

better. The larger the temperature coefficient, the smaller will be the

peak power level reached by the reactor, and conversely, for a given per-

mitted peak power level, shorter periods may be used at any level or

faster rod motions may be employed safely.

The new case that we wish to discuss is the one whereby the larger the

temperature coefficient, the worse the situation. Let us examine an

accident which we shall call the cold-coolant accident. Assume the

following situation: A reactor plant is operating at its normal temperature

when for some reason flow connections are made that would suddenly

AVERAGE COOLANT

TEMPERATURE

DROP IOO°F

TIME

FIG. 10-8. Power level versus time, cold-coolant accident.

switch new coolant into the system from an external storage tank. If

the new coolant is presumed to be at room temperature or at some tem-

perature lower than the normal reactor operating temperature, the

entrance of the new coolant into the reactor would cause the reactor

average temperature to fall. This drop in average reactor-coolant tem-

perature would now insert positive reactivity into the reactor, and the
Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

neutron level would rapidly rise. It is conceivable that this level would

rise so fast and so far that even normal rod motions, plus the later action

of the negative temperature coefficient as the coolant again is heated,

would not be able to protect the reactor.

The amount of average temperature drop occurring and the rate of

change of this temperature drop are both involved in the severity of the

accident. The total temperature drop, however, is the most important

parameter, as this represents the total amount of reactivity that must be

overcome before the accident can be restrained. Figure 10-8 illustrates

the relationship between power level and time for given average coolant

temperature drops in which cold coolant is inserted at a fixed rate. . These


Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

coefficient variation to cover the startup accident and the cold-coolant

accident for a given reactor plant. An optimum value very clearly exists.

It will be recalled that for automatic control purposes an optimum

temperature coefficient situation also existed. A too high value of neg-

ative temperature coefficient would create instability and cause the sys-

tem to oscillate. A too small value of negative temperature coefficient

would create a poor transient response and possibly xenon-poisoning

oscillations.

TEMPERATURE COEFFICIENT »-

FIG. 10-10. Optimum range of temperature coefficient to prevent startup accident and

cold-coolant accident.

10-4. Alarms and Cutbacks. Scramming is quite a severe operation.

In fact, the act of scramming, because of the violence involved, might

cause damage either to the control rods or to the surrounding structure

if done too frequently. Therefore, whenever possible, it is desirable to

limit or ward off accidents by means other than scramming. This gen-

erally is done by providing slower means of protection. Some of these

slower protections are listed below in order of increasing effectiveness.

1. Sounding or lighting up an alarm. This means permits the opera-

tor to determine what has actually happened and then to perform a

minor correction if he deems it necessary.

272

2. Stop all rod motion. It has been shown that in many cases the

negative temperature coefficient is sufficient to handle certain classes of


Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 273

accidents provided that additional complications are not introduced by

rod motion. For a reactor with a high negative temperature coefficient,

automatically stopping rod motion is usually a safe condition.

3. Permit rod motion in the negative reactivity direction only. Re-

gardless of whether the automatic control or the operator wants to per-

form a given function, once this safety condition is switched in, only neg-

ative reactivity can be obtained.

4. Automatically cause rods to move into the reactor at their normal

rate. This condition is sometimes called reverse.

5. Cause rods to move inward at a faster rate than normal. This

condition is sometimes called cutback. It is not a scram situation, as

the rods are driven inward either by the normal control motor with dif-

ferent excitation or by a separate motor with different gearing or clutch-

ing. Rod insertion rates of between two and ten times normal extraction

rates can usually restrain most accidents.

6. Scram. As previously described, this is the condition of inserting

the rods as quickly as possible.

Any given reactor system can use some or all of these devices, and they

may be used in two ways. The first method is a parallel sort of operation

whereby a given accident or a class of accidents can be assigned a given

type of correction. For example, if an electrical interlock connection is

broken, an alarm might be shown. If a pump fails, a slow inward rod

motion might be called for, and so on.

The second system is a series operation in which the weakest type of


Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

correction in the system is always used first and then the corrective action

progresses to more drastic types of protection if the severity of the acci-

dent continues to increase. For example, in the case of overpower pro-

tection, a sequence of the following type might be applied.

1. 105 percent power level—alarm

2. 110 percent power level—rod motion frozen

3. 120 percent power level—fast cutback—rods inserted above normal

rate

4. 130 percent power level—scram

These two systems are generally used in combination. Examples of

the systems presently employed are given in the literature.2 The dif-

ference in performance between the various types of protection systems

is indicated in Fig. 8-15 for the startup accident.

10-5. Last-ditch Emergency Shutoff Measures. Because of the

destructive capabilities of a reactor plant, most power reactors have last-

ditch backup safety devices which operate in the event that all of the

above-mentioned types of protection fail. These failures are presumed

to be from catastrophic causes such as explosions, earthquakes, or mili-

tary action. The assumption is made that the reactor or control-rod


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

mechanisms have been distorted or destroyed and that it is desirable to

bring the reactor to the lowest possible state of subcriticality. It is con-

ceivable, for example, that a control-rod passage might be blocked and the

reactor otherwise be in reasonably good condition. Several methods have

been mentioned as being used for last-ditch shutoff. The insertion of

chemicals with high-absorption cross sections and the use of boron shot

injected into control-rod passages are feasible. For reactors of the water

boiler type, it is also possible to dump out the fuel into an auxiliary tank

having a safe geometry. For reactors that have separate moderator and

coolant, the moderator might also be dumped. The Oak Ridge graphite

moderated reactor employs boron shot, whereas the Brookhaven reactor

contains a means of injecting liquid trichlorobenzene into tubes in the

reactor. The Canadian NRX reactor can have its moderator removed.

The speed of control required for a last-ditch emergency action usually

is not fast. Presumably the damage has already been done. All the

safety methods have been tried. These devices should not be triggered

in without some thought because of the difficulty entailed in the reverse

operation of cleaning the poison

out of a system. Times in the

order of 1 min to 1 hr could be

used. The requirement for slow

operation stipulates the use of man-

ual devices. These devices are de-

sirable because the catastrophe is


Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

likely also to have caused a power

failure. Thus manual pumps or

manually operated dump valves

may be used.

10-6. Scramming Circuits. Be-

cause of the emphasis placed upon

scramming in many reactors, some

of the elementary forms of circuits

that are used to actuate a scram

mechanism will now be examined.

Some of the requirements of these

scramming mechanisms have been

presented in Chap. 5, and it is the

feeding circuits driving these mech-

anisms that will now be examined.

Most scram actuators are schemati-

The control rod is connected to the normal

The scramming operation con-

SCRAM

MAGNET\

RELEASE

POINT

TO SCRAM MAGNET

CURRENT SUPPLY

SPRING TYPE

-MECHANISM FOR

FAST INITIAL

MOVEMENT

-CONTROL

ROD

FIG. 10-11. Elementary scramming mecha-

nism.

cally of the form of Fig. 10-11.

driving mechanism by means of a magnet.

274

sists of releasing the control rod from the direct drive and causing the rod to
OPERATIONAL CONTROL PROBLEMS SHUTDOWN

275

be accelerated as quickly as possible into the reactor. The point of release

is the scram magnet, and this magnet is normally energized. Either the

current through such a magnet may be an on-or-off proposition, or in some

circuits the condition is brought about that as the plant approaches an

unsafe condition, the current through the release magnet is reduced

toward the tripping point in some proportional manner.

Each control rod generally has its own magnet releasing system. How-

ever, these rod magnets may be connected together on one or more com-

mon power supplies. Three types of magnet feeding circuits will be

described: the relay type, the magnetic-amplifier type, and the electronic

type.

MAGNET

POWER

SUPPLY

SCRAM BUS J |_

SCRAM

SIGNAL

NO. 1

AUX. 1

RELAY

lur

SCRAM
Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

SIGNAL

NO 2

AUX. 1

RELAY

SCRAM

SIGNAL

NO. 3

AUX. ,

RELAY

y.

v,

MANUAL

SCRAMS

1'

t^O-

POWER

RELAY

RELEASE

MAGNETS

ROD

ir

RELAY

POWER

SUPPLY

..

FIG. 10-12. Poor relay-type parallel feed scramming signal circuit.

Relay-type Scramming Feed Circuits. One of the simplest types of rod

release magnet feed circuits is the relay parallel feed arrangement. This

circuit is shown schematically in Fig. 10-12. This is an on-off type feed

and may be used with any number of rods or any number of input scram-

ming signals. Because of the fact that the rod release magnets will

usually require considerable power, an auxiliary power relay may be

necessary to disconnect the magnet circuit. This type of power relay is

generally quite slow. A scramming bus is provided, to which any num-

ber of devices may be connected to close the power relay circuit, thus

causing a scram. Auxiliary relays would be used for most scramming

signals except for manual scramming. Those auxiliary relays shown in

Fig. 10-12 normally have current through them and thus keep open a

back-contact type of circuit. Manual scrams may be placed in the auxil-


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

This is a poor type of circuit for several reasons. From a safe failure

point of view, the power relay circuits, requiring current to release the

control-rod magnets, are backward. Safe failure requires that when the

power fails for any reason, the rods should scram. In Fig. 10-12 the auxil-

iary relays are connected properly in that if their power supply fails, they

close the scramming circuit. Back contacts are used throughout this

sort of circuit. This is not considered good relay practice, as back con-

tacts generally have less pressure available than front contacts. The

over-all circuit is quite slow because the auxiliary relays and the power

relay are effectively in series. The time delays in both relays add up,

SCRAM

SIGNAL

NO. 1

SCRAM

SIGNAL

NO. 2

FIG. 10-13. Relay-type series feed scramming signal circuit.

and as has been pointed out in this chapter, time delays must be kept to

a minimum. The time-delay problem is accentuated by the fact that the

power relay must be closed rather than opened to scram the circuit. For

usual relay design, opening a circuit is slightly faster than closing it.

A much better relay circuit is shown in Fig. 10-13. Here all the auxil-

iary relay contacts are normally closed and are connected in series. The

power relay is normally operated with current through its coil, and a
Generated on 2011-10-24 11:59 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

power failure of any sort releases the magnets. Front contacts are used

throughout, but two relays are still in series, which means that this is

still a slow type of circuit.

276

Magnetic-amplifier-type Scramming Circuits. In addition to the above

difficulties, relays have other troubles. Problems of contacts sticking,

welding, or needing frequent cleaning make relays not too desirable.

Vacuum-tube circuits are very fast and very versatile, but they are sub-
OPERATIONAL CONTROL PROBLEMS: SHUTDOWN

277

ject to occasional failures. Consequently, for utmost in reliability,

magnetic-amplifier types of circuits are used. Magnetic-amplifier relays

may replace electromagnetic relays on a one-to-one basis, or the unique

CONTROL

SIGNAL

LOAD

FIG. 10-14. Magnetic-amplifier relay circuit.

characteristics of magnetic-amplifier relays may be exploited to eliminate

the equivalent of many relay contacts.

Figure 10-14 shows an elementary magnetic-amplifier relay circuit.

Many more complex circuits are also available.3 The circuit of Fig. 10-14

contains an input control winding

fed by some d-c control signal such

as the current from an ionization

chamber. The output load cur-

rent can be at either a very low or

a very high value, depending upon

the control-signal current. Figure

10-15 illustrates the load current

as a function of the control-signal

current. Feedback is applied to

the magnetic amplifier in such

a manner that its core operates


Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

either unsaturated or at complete

saturation. The load current is caused to shift abruptly from one con-

dition to another as a function of signal current. A bias level is set up

whereby the relay may be placed as close to the tripping point as desired.

CONTROL CURRENT

FIG. 10-15. Magnetic-amplifier relay load

current versus signal current.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The magnetic-amplifier relay, in common with other types of relays, has

a hysteresis effect which is not shown in Fig. 10-15.

A scramming circuit employing a magnetic-amplifier power relay is

indicated in Fig. 10-16. Here the current to the rod release magnets is

either all on or nearly off, depending upon the state of the magnetic-

amplifier power relay. The control winding of the power relay is fed

from load windings of an auxiliary magnetic-amplifier relay, and the

scramming signals are fed into a scramming bus via isolating diodes.

ROD

RELEASE

MAGNETS

SCRAM

BUS

MAGNETIC

AMPLIFIER

RELAY

POWER

^1

SCRAM SCRAM SCRAM

SIGNAL SIGNAL SIGNAL

NO. 1 NO. 2 NO. 3

FIG. 10-16. Magnetic-amplifier relay scramming signal circuit.

1 r—
Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

IONIZATION

NORMAL CONDITION

ZERO CURRENT ~\

TO ALARM

INDICATION

CIRCUIT

MAGNETIC S||

AMPLIFIER T

*~ POWER D

RELAY pll

MANUAL

SCRAMS

*X0Jv-vD__^

DETECTOR

CURRENT

.5 AMPLIFIERS'

5 MAGNETIC C

MAGNETIC

"J AMPLIFIERS"

5 AUXILIARY f

_I_|-Y~«'V"Y-V_J_

CONVENTIONAL!

RELAY

ROD

RELEASE

MAGNETS

FIG. 10-17. Magnetic-amplifier scram protection channel.

These diodes are provided to prevent a scram in the event that any of the

magnetic amplifiers feeding the scramming bus inadvertently shorts out.

The speed of operation of a magnetic-amplifier relay depends upon the

frequency of its a-c power supply. Magnetic-amplifier relay closure

times are measured in cycles of the primary excitation frequency. A

closing time of 2 cycles or better can usually be obtained. This state-

ment implies that for extremely fast-acting relays, high-frequency power

supplies should be used and 400- and 800-cycle supplies are common.

278

Figure 10-17 shows a complete protection channel from an ionization

chamber employing magnetic-amplifier relays. The current from the


OPERATIONAL CONTROL PROBLEMS: SHUTDOWN

279

detector is amplified directly in a conventional magnetic-amplifier relay.

The normal condition for the auxiliary relay is to supply minimum load

current. The auxiliary relay feeds a scramming bus through the coil of

a conventional electromagnetic relay, which is employed only to indicate

which channel has caused the scram. The auxiliary relay is also isolated

from the scramming bus by means of a rectifier. The scramming bus

now feeds the power magnetic-amplifier relay which is biased so that its

load current is normally on. This load current feeds the rod release

magnet, and manual scrams are provided in the output circuit. A chan-

SCRAMMING

"SIGMA"

"<t

.Z

in

SIGNAL

NO. 1

AMPLIFIER

SCRAMMING

"SIGMA"

ELECTRONIC
Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

AMPLIFIER

"^||

5ll

ROD

RELEASE

MAGNET

SIGNAL

NO. 2

AMPLIFIER

SCRAMMING

"SIGMA"

MAGNET

AMPLIFIER

SIGNAL

NO. 3

AMPLIFIER

POWER

SUPPLY

FIG. 10-18. Electronic scramming system block diagram.

nel of this sort may be constructed from conventional components operat-

ing at 400 cycles/sec and will have an over-all time delay from detector

signal to rod motion of less than 60 msec.

Electronic Scram Circuits. When the utmost in scramming speed is

desired, electronic circuits are generally used. To take advantage of

the high speed obtainable from these circuits, special magnets with quick

release times must also be designed. In the use of electronic circuits,

problems concerned with safe failure are magnified. Direct-coupled tube

circuits are generally used and so designed that a maximum amount of

protection against tube failure is provided. The electronic type of relay

circuit is usually of a proportional type rather than on-off. However,

because of the versatility of these circuits, combinations of proportional

and nonlinear operation can be easily provided.


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Figure 10-18 shows a simplified block diagram of the basic type of cir-

cuit. Here, using the terminology originating at the Oak Ridge National

Laboratories, a so-called sigma or summation bus is set up, which is fed

from a group of summation amplifiers.4'6 This summing bus in turn

feeds a magnet holding amplifier of a vacuum-tube type. Because of

the large current requirements, more than one magnet is usually not

connected to a magnet amplifier. The current to the rod release magnet

is a rectified alternating current, and the output tubes of the magnet

amplifier are fed from a separate transformer. Manual scrams and other

low-power interlocks may be inserted by controlling the primary voltage

to this transformer. Obviously, if this primary voltage fails, the system

scrams.

SIGMA BUS

MAGNET

FIG. 10-19. Research reactor scramming system block diagram. (Courtesy of Gilbert

Goss, Radiation Counter Laboratories.)

The amplifiers supplying the current to the magnets can be controlled

from a mixing circuit. The output of the scramming signal detector is

fed into the mixer which can accept signals from several sources. The

mixer may operate in such a manner that a constant output signal is

produced until the reactor approaches some dangerous condition. Then

the output of the mixer becomes proportional to the largest signal applied

to it. The current through the release magnet is set to release the rods

at a given level. As a dangerous condition is approached, the magnet


Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

current will remain approximately constant until the tripping point is

nearly reached. Then the current may be arranged to drop nonlinearly

and sharply through the tripping point.

280

Figure 10-19 shows an elementary block diagram of a period and level

safety circuit of this type for the Oak Ridge "swimming pool" research

reactor. One period circuit and two level circuits are shown. The POP

chambers referred to in this diagram are parallel circular plate uncom-


OPERATIONAL CONTROL PROBLEMS: SHUTDOWN 281

pensated ionization chambers. Channels of the sort indicated in Fig.

10-19 have been constructed with an over-all time delay from the detec-

tor to rod motion of less than 40 msec.

REFERENCES

1. Wade, E. J.: Instruments Used with Experimental Reactors, "Convention Record

of the IRE 1954 National Convention," pt. 9, " Medical and Nuclear Electronics,"

Institute of Radio Engineers, New York, 1954.

2. Stephenson, R.: "Introduction to Nuclear Engineering," McGraw-Hill Book

Company, Inc., New York, 1954.

3. Geyger, W. A.: "Magnetic Amplifier Circuits," McGraw-Hill Book Company,

Inc., New York, 1954.

4. Cole, T. E.: Design of a Control System for a Low-cost Research Reactor, Nucle-

onics, vol. 11, no. 2, p. 32, 1953.

5. Trimmer, J. D., and W. H. Jordon: Instrumentation and Control of Reactors,

Nucleonics, vol. 9, no. 4, p. 60, 1951.

6. Cochran, D., and C. A. Hansen, Jr.: Instrumentation for a Nuclear Reactor,

Nucleonics, vol. 5, no. 21, p. 4, August, 1949.

7. Los Alamos Scientific Laboratory: An Enriched Homogeneous Reactor, Rev. Sei.

Instr., vol. 22, p. 489, 1951.

8. Dahl, A., and G. Renders: Heavy Water Reactor at Kjeller, Norway, Nucleonics,

vol. 9, no. 5, p. 5, November, 1951.

9. Colmer, F. C. W., and D. G. Littler: Gleep: Design, Construction, and Use,

Nucleonics, vol. 8, no. 1, p. 3, 1951.

10. Breazeale, W. M.: The "Swimming Pool," a Low-cost Research Reactor, Nucle-
Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

onics, vol. 10, no. 11, p. 6, November, 1952.

11. Goss, Clinton G.: Reactor Control Instruments, "Proceedings of the 1953 Con-

ference on Nuclear Engineering," University of California Press, Berkeley, Calif.,

1953.

12. Lansing, N. F., (comp.): The Role of Engineering in Nuclear Energy Develop-

ment, AEC Report T/.D-5031, p. 417.


CHAPTER 11

SIMULATORS

The transient response of nuclear reactors and their associated power

plants cannot usually be obtained by direct analysis because of the numer-

ical complexity involved. As we have seen in Chaps. 4 and 6, the solu-

tion of several simultaneous differential equations is needed, and the

transient solution of these equations is most easily accomplished by

means of analogue computing machinery. Special-purpose computing

machines for the solution of reactor plant problems are called simulators

because their performance approximates very closely that of an actual

plant. Simulators are also useful to test the controls of a given plant in

advance of its construction. That is, the simulator may be treated as

though it were the plant and actual controls connected to it, sometimes

by means of transducing elements. The complete control performance

of a plant, its instrumentation, and its external feedbacks can thus be

studied.

In this chapter we shall first briefly review elementary analogue com-

puting techniques. Then we shall use these techniques to examine the

forms of reactor and plant kinetic simulators. However, before the

reader attempts the construction of any such device, it is recommended

that a more detailed study be made of analogue computational methods.1'2

It must also be kept in mind that the simulators will be no better than

the actual equations which they simulate. If the plant is treated kinetic-

wise as a point source in a black box, the approximations in the equations

will show up as approximate answers in the output of the simulator.


Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

11-1. Elementary Analogue Computing Techniques. Kinetically many

physical systems obey the same mathematical patterns. For example,

exponential decay is exhibited by the decay of radioactive materials, the

cooling of a thin metal sheet, the deceleration of mass in a viscous fluid,

and the discharge of a capacitor through a resistor. The dimensionless

equations of these actions exhibit the same form, and any one can be used

to predict the performance of the others. Therefore any of these proc-

esses may be said to be the analogue of any one of the others.

282
SIMULATORS 283

Another definition of a simulator, then, might be that a simulator is

a convenient system used to duplicate the action of an inconvenient sys-

tem. The most convenient present-day type of simulator is the electrical

or electronic simulator. Electrical quantities are easily and accurately

measured. Electrical components are comparatively inexpensive and

can be rapidly manipulated to simulate a great variety of natural phe-

nomena. Electrical quantities can also induce mechanical motion and

hence can be transformed into solenoid displacement or motor rotation.

Conversely, mechanical motion can be transformed into an analogous

electrical quantity through use of generators or other means. Or both

mechanical and electrical systems can be used together in one simulator.

There are two types of electrical or electronic arrangements that are

generally used. These are passive electrical networks and operational

amplifier techniques.

ElN°-

« E OUT « KE IN

*«*

FIG. 11-1. Elementary multiplication process.

Linear Passive Elements. Electrical circuitry uses three basic linear

passive elements: the resistor, the capacitor, and the inductor. The cur-

rents through these elements as a function of the voltage across them are

given as / = E/R, I = C(d/dt)E, and I = 1/L ft E dt. The impedance

of any of these elements or combination of them is defined as Z = E/I.

Pure inductances are difficult to construct because of the finite amount


Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

of resistance and shunt capacitance in the windings. Large values of

inductance also are nonlinear with current. Consequently, most ana-

logue computing networks avoid the use of inductance and only resistive

or capacitive passive circuit components are generally used.

Nonlinear passive elements when needed can generally be simulated

by a diode or group of diodes. These devices are sometimes called func-

tion generators.

One other element, the coefficient potentiometer, is in common usage

to provide simple multiplication. Figure 11-1 indicates the process of

multiplication in an elementary potentiometer circuit. The potentiom-

eter tapoff, which is usually proportional to an angular displacement of


284 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the potentiometer shaft, modifies directly the input voltage so that

Emt = KEiu. Servomotors may be used to position the arm of a poten-

tiometer. Multiplication of two or more changing variables is usually

an inaccurate process, and the equations are rewritten whenever possible

to avoid such multiplications.

Operational Amplifier Techniques. The basis of the design of most

simulators depends upon the concept of the operational amplifier. In

order to understand the operation of analogue computers properly, an

understanding of operational amplifier technique is essential. Linear

passive networks may be connected together in combination, and solu-

tions to problems obtained. However, practical passive components

unfortunately load each other in the interconnection process, and conse-

quently the equations must be modified to take this loading into account.

This loading usually creates extreme complications and inaccuracies.

Therefore the operational amplifier is essentially a device designed to

prevent loading effects, but because of its unique properties, other val-

uable computing features can also be obtained.

Ideally, an operational amplifier is a direct-coupled amplifier with a

gain of 1,000 or more and a frequency response such that no instability

occurs in the circuit when negative feedback networks are connected

around the amplifier. In the design of an operational amplifier it is also

customary that the amplifier have no zero offset; that is, the output for

zero input is zero volts. The output is usually able to swing through a

range of at least 100 volts centered around zero. For the purposes of
Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

unloading interconnecting elements and accurate computation, one of

the most critical features is that the amplifier should have as high an

input impedance as possible and a low output impedance.

Figure 11-2 indicates a typical nonstabilized type of operational ampli-

fier. By "nonstabilized" is meant that the stability of the amplifier is

obtained through the stability of all its components and power supplies

alone. A stabilized operational amplifier has superimposed a-c circuitry

for correcting amplifier drift and is used when higher precision is needed.

In the amplifier of Fig. 11-2 the first stage is cathode coupled, giving the

effect of a high-impedance cathode follower input. The grid of the

second stage is used for two purposes. First, by applying a small volt-

age, usually y% volt or less, to this point, it is possible to compensate

for any zero offset caused by variation in tubes and component toler-

ances. Second, since this point is separated from the output by an even

number of inverting stages, positive feedback can be conveniently applied

here to raise the effective gain of the amplifier to a very high value.

This increased gain is used, as will be seen, to increase the accuracy of

computation.

The third stage is coupled to the second stage directly by means of a


SIMULATORS

285

voltage divider, and a similar network is inserted between the third stage

and the output stage. Capacitors are used to compensate for the input

capacitance of the tubes in order to extend the basic frequency response

INPUT

+300 V

OUTPUT

-300V

+ 300V

680K

6.3VAC

FIG. 11-2. Schematic diagram of nonstabilized operational amplifier.

If

OPERATIONAL

1'1

AMPLIFIER ^^

GAIN A^XXX^

IN I

FIG. 11-3. Generalized operational amplifier circuit.

as far as possible. To prevent oscillations when feedback is connected

around the amplifier, a network to attenuate the high-frequency response

at the rate of 6 db per octave is used. The output stage is a cathode


Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

follower to give the necessary low-impedance output. The use of nega-

tive feedback in conjunction with this type amplifier will usually further

reduce the output impedance.

Figure 11-3 shows a typical generalized operational amplifier circuit as


286 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

it is used in computational work. Associated with the amplifier is an

input impedance Z; and a feedback impedance Z/. Both of these terms

may be functions of frequency and can be treated in Laplace notation as

Zi(s) and Zf(s). Kirchhoff's law for Fig. 11-3 can be written as

&.(«) = Ii(s)Zt(a) + e(s) (11-1)

£out(s) = 7,(s)Z/(s) + e(s) (11-2)

&.,(«) = -Ae(s) (11-3)

where the voltages are as indicated on the diagram.

The input impedance of the amplifier is considered to be infinite.

Consequently, the current —//(«) must equal Ii(s). Rewriting and

combining Eqs. (11-1), (11-2), and (11-3) result in

= _/,(s) = ~-w zir:

W /0\ -S- \T?. /. oW 4 1

(11-4)

(11-5)

''A

Equation (11-5) is the transfer function of the operational amplifier and

its associated circuits, and usually the gain of the amplifier A is very high,

so that the transfer function reduces to the simple form

Z,(«)

="

(.

The operational amplifier can be used as a computing tool by the use

of discrete networks for Z/(s), and Zi(s) and/or by using multiple inputs.
Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

When multiple inputs are used,

n rr f \

outi, ) * ^ m,H ; Z;(s)

k=l

By proper use of Eqs. (11-6) and (11-7) many basic forms of computa-

tion can be accomplished. Figure 11-4 indicates some of the more com-

mon operations. Figure ll-4a indicates the process of summation. It

will be noticed again that the gain and the characteristics of the amplifier

have nothing to do with the process, and the accuracy of the computation

depends only upon the accuracy of the individual resistances and input

voltages. Because of the inverting action of the amplifier, a minus sign

usually occurs for most output processes. The sign may be reversed

quite simply by applying the output of one operational amplifier system


SIMULATORS

287

to another operational amplifier which has equal input and feedback

resistances. The output of the second amplifier will then give an inverted

answer directly, on a one-to-one basis. Figure 11-46 indicates the process

of integration. Integration is a very stable and accurate process, provid-

ing the drift of the amplifier and the leakage of the capacitor Ct are low.

R<

o—V\AAA.—i

ETC. -

-AA/WV

(a)

= -R .

R<

E,<

R2

E2o—VWW-

ETC.-1

(b)

ETC

.VWW-

oEouT=-Rf^:(C1E1+C2E2....)

ETC.

FIG. 11-4. Elementary forms of computation circuits using operational amplifier techniques,

(a) Summation, (b) Integration, (c) Differentiation, (d) Generalized multiple input
Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

solution.

Care must be taken in selecting only the highest quality capacitors for

this usage. Figure ll-4c indicates the process of differentiation. This

process is not commonly used, particularly when the equations may be

manipulated to place them in a form where integration can be used.

Figure ll-4d shows the generalized multiple input solution.

For some of the servomechanism type analyses it is desirable to sim-

ulate directly elementary transfer functions. Figure 11-5 indicates some


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

of these network simulations. In operational form Fig. 11-oa indicates

a simple lag network in which

RI

R= l +

Figure 11-56, similarly, is the simulation of a lead network whereby

Eour

EU, Ra, 1 -|- 1

Figure 1 l-5c indicates a combina-

tion network providing elemen-

tary lead and lag of the form

RlN ,.ClN

F IN 0-^AA/vJf— — —

\.

Rf

oEour

E OUT

P IN

FIG. 11-5. Operational amplifier simulation

of typical transfer functions, (a) Lag net-

work, (b) Lead network, (c) Lag lead

network.

Ein RUL 1 + RfCfS

There are many other combina-


Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

tions available to the designer.1

From these simple examples it

can be seen that the transfer

function of an operational ampli-

fier circuit can be matched to the

transfer function of most pieces

of physical equipment and a volt-

age that follows the same equa-

tions as the input to the equip-

ment may be applied to the input

impedance. The output voltage

then will follow the equations of the

equipment output. This, then,

is the basis of operational ampli-

fier technique and circuitry to sim-

ulate given pieces of apparatus.

There are a number of reactor

These simulators vary

11-2. Reactor Kinetic Simulators.!

kinetic simulators mentioned in the literature.3~6

in their complexity, depending upon the approximation to the physical

reactor that is used. For servo transient problems it is usually not neces-

sary to approximate the internal physical constants of a reactor in much

detail. Instead, the basic reactor kinetic equations that we have been

using in this text will usually prove to be sufficient. The manner in

which the power level of a reactor will increase or decrease depends upon

EM

f The derivations in the next three sections follow closely the original work of W.

Pagels in Ref. 4 and other, unpublished, references.


SIMULATORS

2*9

the past history and the effective multiplication factor of the reactor.

We may use the familiar reactor kinetic equations to describe this kinetic

performance. However, the reactor kinetic equations must be rewritten

and put into a more suitable form for computer operation as follows:

dn Sk -

v „

n + 2, XA

dd _ PJI

dt " "F

(11-8)

(11-9)

remembering that ft = ) ft, combining the two equations, and multiply-

ing by I*, we have

,e

I* -^ = Skn — I* V -TT + l*S (11-10)

dt Li dt

Equation (11-10) is now much more convenient for a computer solution.

In order, therefore, to have some idea as to the form of the networks

8k-

FIG. 11-6. Block diagram of elementary reactor kinetic simulator.

that will be used to simulate a reactor, Fig. 11-6 indicates in block form

an elementary pile simulator. The potentiometer, delay networks, cur-


Generated on 2011-10-24 12:00 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

rent source, and differentiating capacitor are the computing elements

corresponding to the mathematical operations that are indicated in the

kinetic equations. The voltage output of the operational amplifier is

double ended, giving out two voltages which are equal in magnitude and

opposite in sign. An adding bus is provided which is connected to the

input grid of the amplifier. When the gain of the amplifier is very high,

any voltage on the adding bus causes the output voltage to vary in such

a manner as to reduce this input voltage to zero. All dependent voltages


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

must then vary in accordance with the mathematical operations described

by the feedback circuits as previously indicated. The system may be

made regenerative or degenerative, depending upon the setting of the

reactivity multiplying potentiometer.

Figure 11-7 indicates the elementary circuit networks that perform the

operations mentioned above. The amplifier output voltage E represents

fn where f is a convenient proportionality constant which will be used to

assist in setting the values of the electrical components. In this instance

R9 <

I—VWV\—>^P9

I" i

XTX ^T^ ^T^ XTN '"P* ^T"*

Ic, lca Ic3 Ic4 Ic, Ic

GAIN

OPERATIONAL

AMPLIFIER

FIG. 11-7. Schematic diagram of elementary reactor kinetic simulator.

£ has the dimensions of volts per neutron. From Fig. 11-7 we can sim-

ulate the performance of Eq. (11-10) by currents flowing into a junction

point such that Kirchhoff's second law is obeyed. Then

7, = h -

(11-11)

We can now proceed to derive the simulator constants and scaling factors

for each term of Eq. (11-10).


Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

The left-hand term of Eq. (11-10) is represented by the capacitor C;

which produces a current proportional to the time derivative across it.

If the voltage E represents fn, the current

T dE /fC7\ ,.,, dn

'T = C7 -jT =* I T5" I * -JI

dt \l I dt

(11-12)

290

The symbol c will be used to denote capacitance in order not to confuse it

with the symbol C, delayed-neutron concentration. The term enclosed

in the parenthesis is a scaling factor having the dimensions amperes per

neutron.
SIMULATORS 291

The first term on the right-hand side of Eq. (11-10) is the reactivity

term and is simulated by the circuit involving R and P. The magnitude

of the current h is dependent upon the setting of the potentiometer tap

on P, the voltage E, and the ohmic values of R and P.

Let D be the proportional distance of the potentiometer tap from the

center of the potentiometer. D would then be zero at the center, + 1 at

the +E end, and —1 at the —E end. Because of the amplifier action,

the tap remains at an essentially zero potential for any D.

By direct network analysis the current flowing from the potentiometer

tap is

/s = R[l + (R/P) + (P/4B)(1 - D2)] (H-13)

When the tap is at either extreme end of the potentiometer, 7s is a max-

imum for any given E, The maximum current in Eq. (11-13) is given

when D = 1. We can then set the range of simulated reactivity varia-

tion of the simulator and the proper scaling factors by the following

method : If we set the ratio Sk/Sk^.m equal to /s/ /a.m.* and note that

(R/P)}

if again we represent the voltage E by fn then

where the term in the braces is the scaling factor.

The summation terms in Eq. (11-10) are simulated by a group of six

resistance-capacitance delay networks. The equation describing the

voltage for any of the ith-type network is

*=f+^f (11-17)
Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

where Qi is the capacitor charge in coulombs. This electrical charge is

analogous to the storing of delayed neutrons in the fission fragments

before they are emitted. The corresponding delayed-neutron equation

that this circuit is to represent is Eq. (11-9) rearranged so that

n = £xA + £^' (11-18)

Pi Pi at

Matching Eqs. (11-17) and (11-18) term for term and again introducing
292 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

the proportionality constant E = £n, we find that

0- = TW ct and R<c< = r (n-19)

PiK\ «t

for Eq. (11-17) to be the same as Eq. (11-18).

The current contributed by each simulated group of delayed neutrons

then becomes

, _ dQi (j_\ ,* dCi

dt ~ we,-/ dt

In this way the quantity in the parenthesis is a scaling factor which

can be made the same for all i species.

The sum of all the delayed currents is

/I + /2 + /, + ' ' ' =

where ftiRi =

The last term of Eq. (11-10), the source term, is simulated by introduc-

ing an independent current, h, into the adding bus. From Fig. 11-7,

if Rg is large compared with P9, the current through R<, is given as

/, = ?f (H-23)

Kg

where Dg is the fraction of P9 tapped off and V is a constant voltage

source. If Dg is used to represent S/Sm.., then

Now using Eq. (11-11) we can sum up the currents into the adding

bus and find that

/rcA dn
Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

\l*J dt

.,

R[l ^ ,

/ r \ » y dCi

i dt

Then

which now is of the same form as Eq. (11-10).


SIMULATORS 293

In order for the output current of each term to be of the correct mag-

nitude the scaling factors of each term must equal each other. That is

I* R[l +

m 26)

('

When these electrical quantities are set up, the scaling factors in Eq.

(11-25) drop out and the equation of the simulator is the same as that of

the reactor. The scaling factor, of course, can be set at any theoretical

value, but limiting values are set to a large extent by such practical con-

siderations as insulation leakage, stray capacity, and amplifier grid cur-

rent. The currents caused by these effects must be small compared with

the computing currents.

The above method outlines the design principles behind an elementary

reactor kinetic simulator. Many such devices have been built, and com-

plete circuits as well as the details of practical operation are given in the

literature.3~6

These kinetic simulators may be operated by themselves in an open-

loop condition, or they may be connected by means of control circuits

into closed-loop configurations. From the open-loop operation one can

obtain kinetic response curves such as those of Figs. 3-3 and 3-4. In the

closed-loop situation, transient responses of a reactor control system can

easily be obtained.

11-3. Subcritical Reactor Simulator. A simple variation on the


Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

reactor kinetic simulator just described is possible when problems involv-

ing negative reactivity only are being studied. For these cases a very

wide range simulator may be designed which requires no active elements.

This type of simulator is useful for shutdown problems, and the simulator

of Sec. 11-2 may be quickly modified to this form by removing the opera-

tional amplifier and modifying the multiplying potentiometer. By elim-

inating the amplifier, only passive elements remain in the circuit, noise is

minimized, and drift and extraneous source currents are eliminated.

Figure 11-8 shows the form of this simplified subcritical reactor sim-

ulator. The delay networks and prompt neutron capacitor are simulated

as before, and a high-voltage power supply is used to feed these networks

through a switch Si. The equations of this circuit with the switch Si

open are

Wk + ° ~dt +

where

and the symbols have the same meaning as given in Sec. 11-2.
294 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

The values of the elements are given by

E = fn

Qi = Yfl

Y(-Sk)

c = Yl*

c--Y&

Ct ~ Y \<

(11-29)

The symbols Y and f are the arbitrary scaling factors. It will be noted

that this particular simulation does not include a source term. If a

source is desired, it is necessary merely to feed in a constant current into

the adding bus. Substituting the relationships of Eq. (11-29) into Eq.

(11-27) results in the original reactor kinetic equation (11-10). These

equations obviously hold for negative value of Sk, since the value of Rk,

the reactivity potentiometer, must be a positive quantity. Without the

operational amplifier present, regenerative feedback is not possible.

<:

«

;R, <

.J
Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

:«2 <

;3

:RJ «

:«4 •

.J

JR5 \

\ '•

IR6

»

Si

*.

STABLE

HIGH VOLTAGE

^ DC POWER

C SUPPLY

•* x

c,

N X"

C2

.\ *.

C3

"N *

C4

.*. *

C5

•s

ce

FIG. 11-8. Schematic diagram of negative-reactivity-only simulator.

The method of operation of this simulator is quite simple. For nega-

tive step changes in Sk, Rk is first adjusted to the desired reactivity step

value. The switch Si is closed, and all the capacitors are charged up

from the high-voltage supply to some initial voltage. When the switch

is opened, the voltage decay is analogous to the decay of the neutron

population in the reactor after a negative step function of Sk is introduced.

From Figs. 3-5 and 3-6 neutron decay is rapid at first but decreases slowly

after the first few decades. By using a sensitive stable d-c amplifier in a

recorder system, this elementary simulator may be used to give accurate

answers over a total range of 5 to 6 decades. It also can provide some

flexibility in programming by varying Rk with time.

11-4. Xenon Simulator. To illustrate another form of useful simulator,

a device for the computation of the xenon poisoning of a thermal reactor

will now be discussed. It will be recalled that the effect of poisoning was

to reduce the reactivity of the reactor but changes in poisoning occurred


SIMULATORS 295

and decay of xenon were described, and these are the equations which will

be simulated.

Where the parameters vary very slowly, mechanical analogues are

sometimes more suitable than electronic ones. The electronic opera-

tional amplifier has good stable characteristics over periods of minutes,

but for periods of many hours a mechanical device such as a gearbox may

be more satisfactory. For these long-time situations a servo amplifier

driving a servomotor can be substituted for the operational amplifier, and

the feedback is provided around this amplifier by coupling devices onto

the shaft of the servomotor. For example, differentiation is performed

by coupling an electric tachometer onto the shaft of the servomotor and

the output of the tachometer is proportional to the speed of mechanical

rotation.

Again, using simple components we can proceed to design an elemen-

tary xenon simulator. The equations describing the number of xenon

atoms in a thermal reactor at any time are now repeated in generalized

form.

~ / X,7 (11-30)

^jr = 7x* + X// - \xX - *xX<i, (11-31)

The symbols have the meaning previously given in Chaps. 2 and 4.

Equation (11-30) describes the rate of production of iodine and its sub-

sequent decay into xenon. Equation (11-31) describes the production

of xenon from nuclear fission and iodine decay and its subsequent elim-

ination by natural radioactive decay and neutron absorption.


Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

The schematic diagram for an electromechanical simulator that rep-

resents these equations is shown in Fig. 11-9. The action of the circuit

is similar to one containing an operational amplifier in that the inputs to

the servo amplifiers are held effectively at zero volts with respect to

ground. Consequently, no current flows into the amplifiers. The out-

put of each of the two servo amplifiers supplies the voltage for one phase

of a two-phase low inertia servomotor. Each of these servomotors drives

a tachometer directly and one or two potentiometers through step-down

gear ratios Ni and N*.

A voltage representing thermal-neutron flux is given by Ei. The cur-

rent flowing through Ri is proportional to iodine build-up, while the cur-

rent through R2 is proportional to iodine decay. The resistance RS

passes a current equivalent to the rate of change of iodine concentration.

Similarly, the current through Rt represents the xenon build-up from

the iodine decay while Rt permits a current proportional to xenon pro-

duction from the direct fission process. The currents representing xenon
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

destruction by thermal neutrons and natural xenon decay flow through

resistors R6 and #7, respectively. The resistor Rs passes a current

equivalent to the rate of change of xenon concentration.

-««

i!

•L

. GEAR

RATIO

Ni

+ £,-«—

11

AC

LINE.

SERVO

MOTOR

TACHOMETER J, |

FIG. 11-9. Schematic diagram of xenon poisoning simulator.

Again using Kirchhoff's law for currents flowing into a junction the

simulation of Eq. (11-30) is

Rl
Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

r d6 =

dt

(11-32)

where 6 = potentiometer rotation in revolutions from the grounded end

of Pi

A^i = gear reduction ratio between tachometer TI and potentiom-

eter PI

Ki = voltage output per rpm of tachometer Ti

296

The selection of resistors is such that the effect of potentiometer load-

ing by the resistors should be made negligible. The following additional

conditions are imposed:


SIMULATORS 297

i — f <t> = voltage representing neutron flux where f is a constant

having the dimensions of volts per unit flux

i — K3Im.x = voltage representing an arbitrary maximum density

of iodine, where K3 is a constant with the dimensions of volts per

unit density of iodine and

"max ' max

By substituting the above conditions into Eq. (11-32) and rearranging,

we have

Then adjusting Ri, R*, R3, and Ni so that the quantities in the paren-

theses are equal to each other, Eq. (11-34) will be like Eq. (11-30). Sim-

ilarly, the equation for xenon in the simulator is

6 E2 cLE1_E1, « #3 , NtKzda _

~o W ' ~ "p~ p r -p- i j5 -77 — u

Pmax K\ amal /l5 Kt «maz K^ tit «'

where a = potentiometer rotation of P% and PS in revolutions from the

grounded end

A^2 = gear reduction ratio between tachometer T^, and Pz and P3

Kz = voltage output per rpm of tachometer Tz

In addition we add the following conditions: E3 = KtXm.% = voltage

representing a maximum density of xenon, where K^ is a constant with

the dimensions of volts per unit concentration of xenon. Finally

— - I- (H-36)

Substituting and rearranging, Eq. (11-35) now reads

dX / f \ / Kt \ , / X4
Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

: = 7x* + ' ' ' x

^ (H-37)

By adjusting Rt, R6, Rs, R», and Nz so that the quantities in the paren-

theses are equal, Eq. (11-37) becomes the same as Eq. (11-31).

From the above simulator equations it can be seen that the derivatives

of X and 7 are proportional to motor speed. Because of the long time

constants involved, rather large gearing ratios are required to make use

of the full speed range of the motors. Once these gearing ratios are set,
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

however, curves of X and / similar to those of Figs. 2-8, 2-9, and 8-16 can

easily be obtained.

This type of xenon simulator may be combined with the reactor kinetic

simulator just derived to give a combined thermal-reactor response under

poisoning conditions. As the purpose of these simulators is usually to

indicate transient response and not stability, the use of this combination

is of interest only for problems of many hours duration. Short-term

transients do not affect the xenon picture, and consequently the reactor

simulator by itself may be used for these transient problems.

TEMP. COEFF.

REACTIVITY

THERMAL SYSTEMS. I i ^ J

/\

\ TEMPERATURE

FEEDBACK LOOP

FIG. 11-10. Block diagram of components of nuclear-power-plant example.

11-5. Power-plant Simulators. Having seen what reactor kinetic

simulators look like and having examined some of the principles behind

their derivation, we can now attempt to set up a simulator describing the

kinetic performance of an entire power plant. For simplicity we may use

an example and attempt to simulate the fictitious power plant used in

Chap. 6. Figure 11-10 indicates the plant we shall duplicate. It will

be recalled from this figure that the plant consists of two basic feedback
Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

loops: first, a regenerative temperature feedback loop and, second, a

degenerative temperature coefficient reactivity feedback loop. The reac-

tor gives out a power level n which is changed into a thermal energy Q.

This energy heats up a coolant, and the outlet coolant temperature TH is

warmer than the inlet coolant temperature Tc, with a constant-coolant

298
SIMULATORS

299

flow being assumed. There is a time delay involved to transport the

coolant from the reactor to the boiler and from the boiler to the reactor.

Coolant mixing is assumed in both the boiler inlet and the reactor inlet.

The average coolant temperature is formed simply as (Th + Tc)/2. A

change in the average temperature causes a change in reactivity of the

reactor through the temperature coefficient. This is the system which

has been previously described in Chap. 6 and shown to be exceedingly

stable over large ranges of negative temperature coefficient.

THROTTLE

MIX

FIG. 11-11. Transfer functions of components of nuclear-power-plant example.

There are two methods of approach to simulating this system. The

first method consists of writing the differential equations for all the

processes involved. Then from these differential equations appropriate

components can be put together to solve these equations, as was done in

the case of the reactor kinetic simulator. The second method is to derive

the transfer functions for each of the components shown in the block

diagram and then to set up the corresponding operational amplifier cir-

cuits that match each of the transfer functions. For this example the

second method will be used.

Figure 11-11 indicates the transfer functions as previously described

in Chap. 6 for each of these blocks. It will be noted that a single point

reactor thermal system representation is used. Also the approximate


Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

formula for the time delays between the boiler and the reactor are stipu-
CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

lated. That is, it will be recalled that these transport time delays were

a pure time lag having a transfer function of the form e~â„¢. Upon series

expansion of this function and by using only the first two terms of the

expansion, the transfer function becomes of the form 1/(1 + TS).

A block diagram of the simulator of the plant represented in Fig. 11-11

is indicated in Fig. 11-12. In this simulation each section of the plant

is represented directly by its transfer function representation, no attempt

being made to combine transfer functions or components. In a practical

design combinations of components are obvious and are usually made in

FUNCTION ,+jQOV

6/ GENERATOR T

f 00001

^AC *

LINE

REACTOR

NEUTRON SECTION

FIG. 11-12. Schematic diagram of simulator for nuclear-power-plant example.

order to minimize the number of operational amplifiers involved. As

each amplifier has a small amount of drift, the cumulative effect of this

drift should be kept small. In Fig. 11-12 each operational amplifier is

numbered, but the resistors and capacitor values are only indicated.

Their values, of course, are dependent upon the specific design of the

plant involved.

300
Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

The simulation starts with the basic diagram of the reactor kinetic

neutron simulator. The design for this section is similar to that set

forth in Sec. 11-2, with the exception that two resistance networks are

provided to change reactivity. The first reactivity changing resistance

is labeled "Control rods," and the second one is a similar resistance net-

work marked "Temperature coefficient." The neutron output of the


SIMULATORS 301

reactor feeds the (?/,(«) reactor output thermal section which is represented

in this diagram as two simple lag terms. These are shown as the circuits

concerned with operational amplifiers 2 and 3. The transport delays

and the mixing are also indicated as elementary resistance and capac-

itance time delays. These circuits are associated with amplifiers 4, 5, 9,

and 10. The boiler is indicated as a lead network feeding a lag network

in which the gain of the lag amplifier is modified by a function generator in

its feedback loop. The purpose of this function generator is to change the

gain KB as a function of throttle position. This gain is nonlinear with

throttle setting as indicated in Eq. (6-37).

The transfer function (?c(s) is obtained in this representation by taking

the term [1 + (s/ac)][l — (s/ac')] and converting it to the equivalent

form [1 + (s/ac)] — sac'[l + (s/ac)]. The first term is simulated by a

lead network shown at amplifier 11. The output of amplifier 11 is then

differentiated in amplifier 12, and the output of amplifiers 11 and 12

added together in amplifier 13. This output is then tied back to Gh(s).

The form of computation just presented would rarely be used in practice

because the direct differentiation involved tends toward less stable opera-

tion of the system. However, this is a good illustration of some of the

operational amplifier techniques suggested in Sec. 11-1.

Finally Th and Tc are picked off and averaged in a resistor network

and a servo potentiometer arm moved to a position proportional to T.v.

This potentiometer arm is then mechanically coupled to the temperature

coefficient reactivity changing potentiometer.


Generated on 2011-10-24 12:01 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Scaling factors are obtained in a similar manner to that indicated in

Secs. 11-2, 11-3, and 11-4, and the simulated plant is then ready for opera-

tion. It is now apparent how one proceeds in nuclear power-plant sim-

ulation either from the direct differential equations or from the transfer

function representation. The transfer function representation is most

useful if measured transfer functions of actual physical components are

available. If experimentally measured transfer functions can be used,

great reliability can be placed on the accuracy of these simulators.

Transient-response curves of the form of Fig. 6-39 are very easily

obtained by use of these devices. It will be noted that no controls have

been placed upon the particular plant simulated. If it is desired to set

up any form of programming or flow pattern, the components must be

modified and suitable additions made to the circuit.

These simulators can be used extensively in synthesizing the control

system for a power plant. It is easy to optimize the system simply by

changing resistance coefficients. Another advantage accruing from the

use of these simulators is the development of engineering judgment in

connection with the power plant. While it is possible to design a power

plant solely from the mathematical expressions describing its behavior,


302 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

considerable time can be saved if the engineer can rely on intuitive judg-

ment to some extent. This judgment is gained through experience in

power-plant operation, and because of the present scarcity of nuclear

power plants, the present way to obtain experience in nuclear power-plant

control is by the use of these simulators.

REFERENCES

1. Korn, G. A., and T. M. Korn: "Electronic Analog Computers," McGraw-Hill

Book Company, Inc., New York, 1948.

2. Svoboda, A.: "Computing Mechanisms and Linkages," M.I.T. Radiation Labora-

tories Series, McGraw-Hill Book Company, Inc., New York, 1948.

3. Bell, P., and H. Straus: Electronic Pile Simulator, Rev. Sci. Instr., vol. 21, no. 8,

August, 1950.

4. Pagels, Walter: A Portable Kinetic Simulator, AIEE Trans., vol. 70, paper 51-262,

1951.

5. O'Meara, F. E.: Reactor Simulators, /. Appl. Phys., vol. 24, no. 9, September,

1953.

6. Fischbeck, Kenneth H.: Nuclear Reactor Simulators, "Convention Record of the

IRE National Convention," pt. 9, "Medical and Nuclear Electronics," Institute

of Radio Engineers, New York, 1954.


Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
PROBLEMS

Chap. 2

2-1. If a source is suddenly inserted into a subcritical reactor at t = 0, what is

the kinetic equation of neutron level as a function of time?

2-2. If a source is suddenly inserted into a supercritical reactor operating

initially at a level n = n0, what is the kinetic equation of neutron level as a func-

tion of time?

2-3. Plot Eq. (2-20) for a reactor having a mean lifetime of 10~6 sec when a step

of reactivity of 0.002 is inserted into the reactor. How far has the reactor level

risen in one second?

2-4. Why is the transient period of a reactor longer for a negative change in

reactivity than for a positive one?

2-5. What is the stable period of a reactor in which I* = 10~3 sec and Sk =

0.0012?

2-6. Given a reactor with a negative temperature coefficient, indicate the posi-

tion of the control rods as a function of power level assuming no coolant flow.

How do these control-rod positions compare with those of the case of constant

inlet temperature of Sec. 2-13?

2-7. Derive the equations for the steady-state and peak poisoning after shut-

down of samarium 149.

2-8. A given reactor has a designed excess reactivity of 5 percent and control

rods worth 7 percent. When the reactor is hot, the negative temperature coeffi-

cient affects reactivity by 3 percent. If the peak xenon poisoning after a given

shutdown is worth 9 percent, within what period of time after the shutdown

can a hot startup be accomplished? If the reactor is not started during this
Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

time interval, how long a wait is involved before it can be started cold? Hot?

2-9. Derive the peak-to-average effectiveness of a control rod whose effective-

ness varies as the sine cubed of the rod position.

Chap. 3

3-1. What is the equation for the response of a critical reactor having I* = 10~6

sec when a positive 8k step of 0.003 is inserted into it?

3-2. How high does the level in the above problem reach at the end of one

second?

3-3. Compare the response of Prob. 3-1 with the response obtained using the

approximate method of Eq. (3-6).

3-4. Plot the response of a reactor having I* = 10~3 sec to a linear reactivity

change of the form —0.1 + O.Qlt between 0 and 50 sec. Compare the levels

obtained with the subcritical multiplication factor.

303
304 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

3-5. Plot the approximate response of a critical reactor having I* = 10~6 sec to

a ramp function Sk = 0.012 sec~1.

3-6. Derive the equation of the transfer function of a reactor having I* = 10~3

sec.

Chap. 4

4-1. What are the values of the zero frequency gain and phase shift for a reactor

having an elementary negative-temperature-coefficient feedback system in which

I* = 10~4 sec, feedback time constant T = 0.1, and feedback gain factor KTC =

0.01?

4-2. What is the zero frequency gain of the xenon poisoning feedback-factor

transfer function 8X(s)/S$(s) at a flux level 4>0 = 3 X 1011?

4-3. Develop an electronic circuit for a comparator which is the analogue of the

magnetic-amplifier comparator of Sec. 4-6.

4-4. Develop an expression for the gain of an ionization chamber and compare

this gain with that of a pentode vacuum tube.

4-5. Design a phase-compensating network to give a phase shift of 0° at zero

frequency, 22° lead at 1 cycle, and 0° at an infinite frequency.

4-6. For the simulator example of Sec. 4-10, plot the peak power attained by the

reactor as a function of the control-rod servo undamped resonant frequency for

input disturbances in which r = 0.5 sec.

4-7. What is the transfer function of the samarium 149 poisoning effect?

Chap. 5

5-1. Given an elevated tank of water of 500-gal capacity, at what height must

the tank be placed to equal the stored energy in a steel flywheel of diameter 2 ft
Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

and thickness 6 in. which is rotating at 3,600 rpm?

5-2. How many automobile-type storage batteries would be required to provide

the same energy? Compare the weights of the three systems.

5-3. What is the power required to raise a 200-lb control rod at a rate of 1 ft/sec?

5-4. A reactor is 100 in. tall and has a 50-lb safety rod initially all the way out.

Scramming is by gravity plus a 50-lb spring which acts over the first foot of rod

travel. A steady frictional force of 10 Ib is also present in the scramming system.

How long does it take for the rod to be completely inserted? How fast is the rod

going when it reaches the bottom of the reactor?

5-5. A given automatic-control system removes a transient disturbance from a

reactor in 0.1 sec by moving a 25-lb regulator rod 6 in. How much horsepower is

required from the regulator-rod drive motor?

Chap. 6

6-1. Indicate the temperature and pressure programs for a pressurized water-

cooled reactor in which the reactor-inlet temperature is held constant, and there

are three step changes in pump speed as a function of power level.

6-2. How much power can be transferred from the primary loop of a water-

cooled reactor to the metal in the boiler tubes under the following conditions:
PROBLEMS 305

copper boiler tube area, 10 sq ft; boiler tubes, 1 in. inside diameter; water flow,

12 ft/sec; temperature difference between the water and metal, 50°F?

6-3. A plate-type uranium fuel element 0.125 in. thick is operating at an output

of 1,000 cal/sec per square foot of surface. What is the rate of temperature rise of

this element assuming no cooling?

6-4. Plot the response of the inlet temperature to a boiler as a function of time,

using the approximate delay formula of Eq. (6-11), if the reactor outlet tempera-

ture is suddenly raised 500°F and the delay time between reactor and boiler is

3 sec. Compare this curve with the current rise through a coil when a sudden

constant voltage is applied to it.

6-5. Indicate in block form the transfer function of a reactor thermal system for

a four-region reactor in which the region closest to the coolant inlet puts out one-

half as much power as each of the other regions.

6-6. Graphically determine 1/[1 — KL,GL(S)] for the example used in Sec. 6-8

if the mixing time constants T6 and T6 are 1 sec each.

6-7. Derive the temperature-loop feedback transfer function !Tav(s)/w(s) of

Fig. 6-19 in terms of the inlet-temperature transfer function Tc(s)/n(s).

6-8. A pressurized water-reactor plant having a negative temperature coeffi-

cient and no external control system is operating at full power output with its con-

trol rods fixed. Suddenly the pumps fail and flow stops. Sketch the behavior

of n, Tc, Th, T,, and p.

6-9. A pressurized water-reactor plant having a negative temperature coefficient

and no external control system is operating at full power conditions with fixed

coolant flow. Suddenly a control rod is dropped into the reactor. Sketch the
Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

behavior of n, Tc, Th, Ta, and p.

Chap. 7

7-1. The sensitivity of a BF3 counter using enriched B10 gas is 10 counts per

unit of flux. What is its sensitivity if it is now filled with BF3 made from naturally

occurring B?

7-2. A cylindrical enriched fission counter and a B10-lined counter have the same

surface area and volume. Which tube has the highest sensitivity? Which one

gives out the largest pulses?

7-3. A neutron has an energy of 4 ev. What is its effective velocity? With

what temperature would you associate it?

7-4. What is the maximum counting rate that may be expected from a cylin-

drical BF3 counter having a 2-in. diameter and a 12-in. length, operating in a

neutron flux of 104 per square-centimeter-second at an energy of 5 ev? Assume

100 percent enriched B10.

7-5. The thermal flux at the surface of a 2-ft-diameter reactor core is 1012

neutrons per square centimeter-second, and it drops off in accordance with

e~°-lr/r, where r is the distance into the reflector in centimeters. Where should a

boron-lined chamber having a sensitivity of 2 X 10~14 amp per unit of flux be

located to read 100 jita?

7-6. Derive a relationship between sensitivity and response time of a neutron

thermopile.
306 CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Chap. 8

8-1. Assume that one out of every 10' alpha particles from radon will produce

a disintegration in some surrounding beryllium. How many neutrons are emitted

per second from a mixture of 1 curie of radon and beryllium?

8-2. A nuclear power plant has a polonium110-beryllium source. The plant's

most sensitive detector indicates 30 counts per second when the plant is shut down.

How long will this instrument be able to function properly if the background

counting rate is 2 counts per second?

8-3. The following table indicates an instrument counting rate as a function of

control-rod position during a reactor startup. At what rod withdrawal will the

reactor go critical?

Rod position, Counting rate,

in. counts/min

0 924

4 1,990

5 2,600

6 3,540

7 7,580

7.5 16,660

7.74 48,130

8-4. A startup operation is begun from a negative reactivity of 30 percent. A

control rod is being removed from the reactor in such a manner as to produce a

linear change of reactivity of 0.0025& sec"1. On what period is the reactor after

one minute of rod pulling?


Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

8-5. Show that the shortest period a reactor can be on is greater than

27 In

where I* — mean neutron lifetime

7 = rate of increase of reactivity

HI = neutron level at which safety devices operate

n0 = startup neutron level before rods are moved

8-6. Devise an electronic circuit which will take its inputs from three period

meters and initiate a scram if any two of them indicate a shorter period than a

demand reference.

Chap. 9

9-1. The boiler in a given reactor plant limits a change in reactor power level

from 1 percent full power to full power to a time longer than 30 sec. What is the

fastest reactor period that can be used for this level change?

9-2. A thermal reactor has an excess reactivity of 5 percent. At full power out-

put at a neutron flux of 1013, it burns up reactivity at the rate of 1 percent per year.

How long can this reactor be operated continuously at full power output?

9-3. Draw a block diagram for a control system for a shim rod bank which

follows a regulator rod in a pressurized reactor. Indicate parameter values and

gains. Provide proof that the system does not oscillate.


PROBLEMS 307

Chap. 10

10-1. Given four safety rods whose reactivity values are 1, 2, 2, and 3, with

one of the rods of value 2 initially out of the reactor, devise an interlocking system

which prevents the other rods from being withdrawn unless there is more reactiv-

ity in the safety bank than in the rod being moved.

10-2. A power reactor has been operating at a steady output level of 100,000 kw

for a long time. Suddenly the coolant flow stops, and the reactor scrams. The

reactor and stagnant coolant can be considered to behave thermally like 1 ton of

copper. Plot the average temperature of the reactor and local coolant as a func-

tion of time considering the effect of the beta-gamma power.

10-3. How large a spring is needed to assist gravity to drop a 100-lb rod into a

4-ft reactor in 0.1 sec neglecting friction?

10-4. An over-power-level scram system is so designed that if a reactor on a

0.1-sec period goes through a tripping level which is set at 120 percent of full

power, the peak power reached is 180 percent. What peak power is reached if the

tripping point is set at 150 percent?

10-5. For the plant of Fig. 6-2, propose a safety system of alarms, cutbacks,

and scrams.

Chap. 11

11-1. Indicate in schematic diagram form how you would perform the follow-

ing operations with an analogue computer: 3 + 5, ax — by, dx/xt + x = 3.

11-2. Draw a schematic diagram and explain the operation of a contactor-

stabilized operational amplifier.

11-3. The operational-amplifier setup of Fig. ll-5a is to be used to simulate a


Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

mixing process whose time constant is 1 sec. What is a suitable set of constants

which might be used in this circuit?

11-4. Design a simulator to represent the decay of a polonium-beryllium source.

11-5. Design a simulator to represent samarium 149 poisoning.


Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
INDEX

Accelerated rod motion, 269

Accidents, 227-236, 257-258, 269-272

Accumulators, 116

Alarms, 272-273

Amplifier, contactor type, 71-73

error signal, 62, 66, 81, 87

logarithmic, 210-211

Analogue computing techniques, 282-288

Automatic control, 8, 61-67, 251-252, 259

Automatic period control, 239-243

Berylium, 11, 214-216

Beta and gamma power, 187, 258

Boiler gain (see Boilers, attenuation)

Boilers, attenuation, 145, 151, 157, 165

equations, 139-140

example constants, 149

level, 133

simulation, 301

temperatures, 139

transfer function, 144~145

Boron-lined counters, 192-193

Boron trifluoride counters (see Counters)

Break frequency, 44, 45, 52

Brookhaven reactor, 117, 274


Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Btu calculator, 209-210

Buffers, 118

Cadmium foils, 206

Calibration, thermal, 209-210

(See also Instruments)

Chain reaction, 11, 12

Coefficient potentiometer, 283

Cold-coolant accident, 270-272

Cold critical reactor, 22, 26

Comparator, 62-66, 134, 240, 242, 243

magnetic-amplifier type, 65-66

Compensating networks, 69, 71, 78

Computor techniques, 282-288

Control philosophy, 5-8

Control philosophy, airplane, 6-8

automatic, 8

conventional power station, 6-7

Control rods, 11, 20-21, 61, 98

actuators, 62, 66, 67

drives, horsepower requirements,

120-123

hydraulic cylinder and piston, 112

hydraulic motor, 109-111

magnetic-feed-through, 104

rack-and-pinion, 103

screw jacks, 103-104

screw-and-nut arrangement, 103

synchronous reluctance motor,

105-108

winch-and-cable,104

effectiveness, 27-28, 223-225, 253,

267-269

ganging, 99

mechanisms, requirements of, 98-99

motion, 60, 61, 202-203, 234, 235

position indication, 101, 118-120

regulator (see Regulator rods)

safety, 20, 99, 260

servo frequency response, 89-92

servo loop gain, 85, 87~88

servo system transfer function, 84—85

shadowing, 201-204

shim (see Shim rods)

simulation, 300

speed, 238, 263


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Coolant mixing, 152-156

simulation, 301

transfer function, 153-154

Counters, boron-lined, 192-193

boron trifluoride, 191-192

effect of temperature on, 205

gamma background operation,

199-200

fission (see Fission counters)

scintillation, 191, 197

Counting rate, 218-220, 222

CP-2 reactor, 46

CP-3 reactor, 70

Critical, cold, 22, 26

hot, 22, 26

prompt, 15

Critical operation, 17, 218, 221-223

Cutbacks, 272-273

Damping coefficient, 68, 85, 87

Decay constant, 14~15, 17

Delay networks, 289, 293

Delayed neutrons, 14-15, 18-19, 33, 35,

170

source, 217
Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

table of, 14

Demand loop, 134, 174-175, 181-182

Demand signal, 63-64, 66, 176-179, 183,

185

Depletion, 21, 220, 248

Detector ranges, 188-190

Detectors, power-level, 188

Deuterium, 215-216

Differential reactivity, 27

Differential transformer, 119

Discontinuous control system (see On-

off-type control system)

Emergency shutdown, 273-274

Energy release, 206

Energy storage, 101

devices, 115-118

Enthalpy, 3-5

Equilibrium xenon (see Poisoning)

Ergen, W. K., 54

Excess reactivity, 20, 26

Fission counters, cylindrical, 194-195

gamma background operation, 200

spiral, 193-194

Fission process, 11-12, 187

Fission-product poisoning, 23-26

(See also Poisoning)

Flow system (see Coolant flow)

Flux distribution, control of, 250-251

Flywheel inertia storage, 117-118

(See also Energy storage)

Foils, 205, 206

Fuel, 10-11, 136-137

Gain margin, 77, 78

Gamma radiation, 187, 196, 215-217

instrument background, 197-201

in shielding material, 198

Grace, J. N., 71

Gravity drop, 113, 114, 259, 267

Gravity storage, 115~116

Harrer, J. M., 46, 71, 120

Heat-transfer coefficient, 128, 137

Heat-transfer material, 11

Hot critical reactor, 22, 26

Hurwitz, H., Jr., 6

Hydraulic mechanism, transfer function,


INDEX

311

Kochenburger, R. J., 73

Lag network, 288

Lead network, 288

Lipkin, H. J., 54

Loading, 261

Logarithmic amplifiers, 210, 211

Lower period bound, 227-230

Magic-number control system, 241-243

Magnet release, 274, 275, 279

Magnetic amplifier, comparator, 65-66

relay, 276-279

Manual scram, 267, 275, 278, 280

Mixing (see Coolant mixing)

Moderator, 11, 21, 124, 137

Monitoring, 259-260

Motors, electric, 101-102

hydraulic, 109-111

linear reluctance, 108-109

spring-wound, 117

synchronous reluctance, 105-108

Multiple-level scrams, 265-266

Multiplication factor, 12-13, 20, 61

Natural frequency, 68-69, 85, 87, 89, 92


Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Negative temperature coefficient, control

system for reactor with, 83

definition, 21

elementary operation with, 21

feedback, 50-54

level considerations, 233-236

local transfer function, 49-54

peak limiting by, 92-94

time constant, 52-53, 85, 90, 93

Neutron detectors (see Instruments)

Neutron level, 13, 17, 53-54, 207-208

equations with delayed neutrons, 18-

20

Neutron sources (see Sources)

Neutron thermopiles, 191, 197

Neutrons, delayed (see Delayed neutrons)

fast, 12

intermediate, 12

lifetime, 13, 208

thermal, 12

Nichols charts, 52

NRX reactor, 218, 274

Nuclear instrumentation system, 125-126

Nuclear power plant, analogy to d-c

generator, 2-5

control of, 124-186

stability of (see Stability)

thermodynamic analysis, 3-5

Nyquist plots, 69

On-off-type control system, 69-78

operation, example of, 75-78

stability of, 72

Operational amplifiers, 284-288

nonstabilized, 284

Output impedance, 284-285

Overpower scram, 261

Pagels, W., 288n.

Passive elements, 283-284

Peak xenon (see Poisoning)

Period, 14, 217, 221-223, 225-232

automatic control, 239-243

circuits, 210, 264

definition, 14

demand, 241-243

scram, 264, 265


CONTROL OF NUCLEAR REACTORS AND POWER PLANTS

Pressurizer system, 127, 131

Primary coolant, 49

Programming, 128-135

compromise, 132-133

time-constant, 133-134

constant average temperature, 3,

129-131, 134, 176-179

constant outlet temperature, 130-131,

183

constant pressure, 131-132, 134, 179

limitations, 246

Prompt critical reactor, 15

Proportional counters (see Counters)

Pump control, 127, 130, 135

Pumping power, 127

Radioactive sources (see Sources)

Radium-berylium source, 214

Reactivity, 12-13, 26

change rates, 38-40, 42, 229-230, 252,

259

changes, 51, 169-170, 245

excess, 20, 26

inventory of, 26-27

shutdown, 26-27
Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Reactor, black-box concept, 29, 50

cold critical, 22, 26

as control element, 48-49

core, 11, 136-137

description, 10-11

fast, 12, 48

hot critical, 22, 26

intermediate, 12, 48

period (see Period)

power level (see Neutron level)

power-level operation of, 245~255

pressurized, 2, 99-100, 248

prompt critical, 15

ramp function response, 36-42

simulators (see Simulators)

startup (see Startup)

state, 14

step-function response, 29-36

subcritical (see Subcritical reactor)

supercritical, 17-20

thermal, 12, 48-49

Reactor control, automatic, 60-67

loop, description, 62-67

discontinuous, 62

Reactor control, loop, proportional, 62,

67-70

requirements, 60-62

stability, 67-69

transient response, 78-83, 89-92,

94-95

manual, 61

mechanisms, 98-123

Reactor stability, 50

with negative temperature coefficient,

54

with xenon feedback, 58-60

(See also Stability)

Reactor thermal system, 135-138

equations, 137-138

multiple section, 157-158

transfer function, 158-159

simulation, 301

transfer function, 141-143, 149

Reactor transfer functions, 42-46

approximate form, 46
INDEX

313

Shim rods, 20, 61, 62, 99, 253-255

follow-up, 253-254

Shutdown, 256-281

Shutdown philosophy, 256-260

Sigma bus, 279, 280

Simulators, 36, 83, 282-302

electromechanical, 295

power plant, 167, 298-302

reactor kinetic, 288-294

subcritical reactor, 293-294

techniques, 83-92

xenon, 294-298

Sources, 11, 15, 17, 206, 214-219

polonium-berylium, 214, 215

radium-berylium, 214

range, 190

removable, 219

Specific power, 137

Stability, 83

absolute, 83

insufficient, 174, 176-179, 182-184

plant, 166-167

natural, 173-174, 179-180


Generated on 2011-10-24 12:02 GMT / Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

relative, 83

sufficient, 174-175, 179-181, 184-186

Stable period, 19

Startup, 213-243, 260

accident, 227-236

control systems, 239-243

initial, 213, 217-220

operational considerations, 226

period and level, 241-243

with poisoning, 236-239

requirements, 225

safety considerations, 226-236

subsequent, 220-225

Steady period, 33

Steam flow, 3-5, 140

Steam generator, 3-5

Steam pressure, 3-5

Steam temperature, 3-4

Steam turbine, 3

Subcritical reactor, level, 15-17, 222

Subcritical reactor, period, 17, 230

Supercritical reactor, 17-20

Swimming-pool reactor, 280

Tellurium 135, 23, 24

Temperature coefficient, 21, 125

feedback transfer function, 51-52, 173

negative (see Negative temperature

coefficient)

reactivity feedback loop, 162-166

Temperature feedback loop, 148-152,

159, 162

transfer function, 151, 154-156

Thermal measurements, 187

Thermal reactor, 12, 48-49

Thermopiles, 191, 197

Thimbles, 100-101, 104

Throttle operation, 131, 134, 168

Transient period, 19

Transient response, plant system, 166-

170, 177

reactor system, 78-82, 90-92

Transport delays, 138-139

example constants, 151

simulation, 301

transfer function, 143-144, 167

You might also like