BOOK REVIEWS
ference methods, parabolic equations, and nonlinear equations. Some
general computer programs are reproduced in the Appendix.
This book should prove to be of considerable value to practitioners
doing computions concerned with the solution of difficult partial
differential equations over irregular domains. The mathematical back-
ground required of the prospective reader is nominal (elementary
differential equations), so the book should be accessible to just about
anyone reading this review. All in all, it is a refreshing addition to the
computing/applied mathematics literature and should stimulate much
associated mathematical research and development over the coming
years.
Fault Diagnosis of Digital Systems-H. Y. Chang, E. G. Manning,
and G. Metze (New York: Wiley, 1970, 159 pp.). Reviewed by Stephen
Y. H. Su, Department of Computing and Information Sciences, Case
Western Reserve University, Cleveland, Ohio 44106.
Recently, the reliability and availability of computing systems
have received a great deal of attention. The aim is to design a digital
system which stays operational despite failure of components. Two
major existing techniques for increasing the availability are redun-
dancy and fault diagnosis. Fault diagnosis is a new area in computer
science. It is only a little over ten years old. This is the first book in the
fault diagnosis area. The authors' aim is to provide the readers with
an introduction to fault diagnosis so that they will be able to read
and understand the literature.
In Chapter 1, the authors point out that the main topic of this
book is the development of techniques for the generation, selection,
and evaluation of automated diagnosis procedures for digital systems.
The background material of fault diagnosis is presented in Chapter 2.
Definitions of faults, tests, fault detection and location, and diagnostic
resolution are given. The combinational and sequential testing pro-
cedures are described with their comparisons. Chapter 3 presents the
generation, selection, and verification of tests for detecting single
stuck-type logical faults. Path sensitizing, D algorithm, the equivalent
normal form, and Poage's method of generating tests for fault detection
of combinational networks are given. For testing sequential circuits,
Poage's method and Seshu's heuristics are briefly described. Methods
for minimizing the number of tests are given. Manual and physical
fault simulations and their comparisons are described.
Chapter 4 presents Seshu's sequential analyzer, which is a set of
computer programs that can generate tests and fault-simulation data
for switching networks. Interpretations of the simulation data to
provide repair information are presented in Chapter 5. The research
problems include improved simulation techniques, the development of
design principles to yield more diagnosable digital systems, the de-
velopment of language for describing digital systems, and the hard-
ware compiler for implementing digital networks. The book closes
with the following speculations on the future: 1) greater automation
of fault diagnosis procedure, 2) an increasing interest in fault-tolerant
design, and 3) further research in the area of the application of graph
theory, algebra, and combinatorial analysis to the diagnosis problem.
The book provides concise introductory material to the work in
the fault diagnosis area up to about 1969. The presentation is clear.
For instance, many students have found the section on D algorithm
very helpful for their understanding of the method (perhaps because
the presentation is not as formal as the original paper). It provides
the readers with a good foundation in fault diagnosis. The chapter on
recent development is interesting. The authors point out: "At present
designs are produced by hand, which seems unsatisfactory. An auto-
mated design facility, or 'Hardware Compiler' would produce designs
more cheaply and quickly, and could be made to respect design
constraints much more scrupulously than could a human designer."
Logic designers may not agree with this point of view, but in spite
301
of the difficulties, this reviewer feels that the automated logic design
of diagnosable digital systems should be an important goal to achieve
in the near future.
Testing of multiple faults, intermittent failures, and nonlogical errors
in logic networks is not covered since no research work had been
published at the time the book was written. This book can be used
as the textbook for the first course in fault diagnosis. Homework
problems should be supplemented. For research work, the student
should consult original papers after reading this book.
Electrical Transients in Power Systems-Allan Greenwood (New York:
Interscience, 1971, 544 pp.). Reviewed by Amos Selzer, Cutler-Hammer
Inc., Milwaukee, Wis.
This book is aimed at both students and practicing engineers. It
consists of 17 chapters, four appendixes, and an index. Its purpose is
"to teach students and engineers the fundamentals of this vital subject
and to equip them to recognize and solve transient problems...."
The subject matter is well organized and clearly written. Each chapter
is supplemented with an excellent list of references which allows the
interested reader to study the subject more specifically.
A point of criticism is that there is no effort to categorize the
transient sources. A list of natural sources and man-made sources
could show the sources we have to live with and those that might be
eliminated.
The first nine chapters in this book illustrate the fundamentals of
electrical transient phenomena and are discussed individually. Chapter
1 introduces the circuit parameters-the mathematical statement of a
transient problem and its physical interpretation. Chapter 2 deals
with the Laplace transform method, which the author uses through-
out the book whenever a mathematical analysis is needed. From a
practical point of view, this method is well justified.
Chapter 3 introduces transient phenomena in ac circuits. The
problems described are those of closing a switch and opening a
circuit breaker to clear a fault. The terms "recovery voltage" and
"R.R.R.V." are defined and discussed.
Chapter 4 is devoted to the study of the series and parallel RLC
circuits. The highlights of this chapter are the dimensionless solutions
of most common transforms described by the use of generalized
graphs. Solutions to practical problems can easily be extracted from
these curves. There are some typographical errors in the equations
in this chapter.
Chapter 5 treats abnormal switching transients such as current
chopping, capacitance switching, ferroresonance, and transformer
magnetizing inrush current. Chapter 6 discusses the transient problem
in three-phase circuits, and Chapter 7 deals with transients in dc
circuits. As in ac circuits, most transients here occur by switching
operations. The problem of the "commutation transient" is well
covered.
Chapter 8 is devoted to the study of electromagnetic and electro-
static fields and their influence on transient behavior. This chapter is
unique because the emphasis is on the fields and not on circuit
aspects. This approach is more fundamental and gives a better under-
standing of the physical phenomena of electrical transients.
Chapter 9 covers the topic of traveling waves on transmission lines.
It treats the problem of switching on and off a transmission line and
discusses the transient problem that is associated with such an action.
The next eight chapters are more specifically oriented. Chapter 10
is devoted to lightning. In Chapter 11, the behavior of windings under
transient conditions is discussed. The protection of system and equip-
ment against transient overvoltages is discussed in Chapter 12. The
problem of transients in the integrated power system is covered in
Chapter 13. In Chapter 14, computer-aided techniques to solve elec-
trical transient problems are examined. The reviewer is in complete
302
agreement with the author's warning that ". the computer is an
important adjunct to analysis and experiments, not a substitute for
it."
The transformation from theory to practice is done in Chapter 15,
where practical circuit-parameter values are used to solve transient
problems. Chapters 16 and 17 deal with equipment and techniques that
are used to measure transients.
This book is well suited as a textbook and is an asset to the practicing
engineer.
Introduction to Mathematical Techniques in Pattern Recognition-
Harry C. Andrews (New York: Interscience, 1972, 255 pp.). Reviewed
by Earl E. Swartzlander, Jr., Hughes Aircraft Company, Culver City,
Calif. 90230.
This book introduces the basic concepts of pattern recognition and
provides a consistent notation which is especially useful to students
and others who wish an introductory guide to the pattern recognition
literature.
The first chapter is an introduction to the notation and the concepts
of pattern space, feature space, and classification space.
Feature selection is surveyed in Chapter 2. First, various linear
transformations (i.e., simple diagonal transformations, Karhunen-
Loeve rotational expansions, and generalized linear transformations)
are described; then transformations based on minimization of the
divergence of a Gaussian pattern distribution are presented; and
finally nonlinear transformations are introduced. This chapter also
describes a method for adaptive feature selection. The author is
careful to emphasize the problem sensitivity of feature selection (i.e.,
most feature selection techniques are well suited to certain problems
with which they produce considerably better results than they do in
general).
Chapter 3 introduces distribution-free classification where dis-
criminants are developed on the basis of known prototypes. Linear
and 'D discriminants are described, as is a one-at-a-time training
algorithm (and its convergence theorem). A brief survey of other
classification techniques (i.e., piecewise linear discriminants, nearest
neighbor rules, and the method of potential functions) conludes the
chapter.
While the methods of Chapter 3 do not require knowledge of pat-
tern statistics, the methods of the fourth chapter require statistical
data. The Bayes classifier (i.e., maximum likelihood ratio) is derived
and various parametric and nonparametric classifiers are developed.
Some of the nonparametric classifiers described include probabilistic-
descent method, potential functions with varied kernels, and an adaptive
histogram method which is suitable for approximating nonstationary
densities.
Nonsupervised learning (or clustering) is introduced in the penulti-
mate chapter. Various heuristic methods, a vector field approach
(similar to potential function methods), mode estimating procedures,
and minimum spanning tree concepts which are useful for clustering
are described. The author then explains the notion of intrinsic dimen-
sionality. This chapter is concluded with a brief description of statistical
methods for nonsupervised learning.
The final chapter is concerned with sequential learning schemes.
Wald's sequential probability ratio test and Fu's modified sequential
probability ratio test are described and illustrated with a Gaussian
example. Other concepts treated in this chapter include dynamic pro-
gramming, use of the Lehmann alternative, and supervised and non-
supervised sequential Bayesian learning methods.
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS, MAY 1973
Appendixes on the use of multidimensional rotations for feature
selection and a rate distortion criterion for feature selection conclude
the book. Of specific interest is a feature selection experiment per-
formed on handwritten numerals which dramatically illustrates the
dimensionality reduction-correct classification tradeoff which is
implicit in feature selection.
This book is well suited as a text for senior or graduate level courses
in pattern recognition; nearly 20 percent of the text is devoted to
problems and references. There is sufficient material for a one-year
sequence of courses. It is a pleasure to report that this book attains
the author's goal of providing a pleasantly readable introduction to
the pattern recognition literature.
Asynchronous Sequential Switching Circuits-S. H. Unger (New York:
Wiley, 1969, 290 pp.). Reviewed by Stephen Y. H. Su, Department of
Computing and Information Sciences, Case Western Reserve University,
Cleveland, Ohio 44106.
Switching systems include digital computers, telephone switching
systems, and many types of control systems. A switching system is
often made up of many switching circuits. There are two types of
switching circuits: combinational and sequential. A combinational
circuit has no memory. In a sequential circuit, the present outputs
depend upon the present inputs as well as the previous history of the
circuits, i.e., it has memory. There are two types of sequential circuits:
synchronous and asynchronous. In a synchronous circuit, clock pulses
are used to synchronize the circuit operations. No matter how soon
the current operation is finished, no action will take place before the
next clock pulse comes along. In an asynchronous circuit, as soon as
the present circuit operation is finished, a signal indicating the
completion is sent out which starts the next operation immediately.
Asynchronous circuits are of interest for at least two reasons: l)
theoretically, they are faster than the synchronous circuits; and 2) fre-
quently, it is important to have a circuit which can operate asyn-
chronously, such as the interface circuit controlling the interactions
between high-speed digital computers or other switching systems and
low-speed peripheral devices such as card readers, printers, etc.
The approach of this book is different from most other books since
instead of starting with some definitions and theorems, the author
starts with a practical interesting problem of designing a switching
circuit for the control of highway traffic. He goes through the whole
design process step by step, until the logic circuit is obtained. Then he
points out the undesirable operations, such as hazards and critical
races, which can occur in asynchronous circuits and presents methods of
eliminating them. The rest of the chapters of the book treat various
steps of synthesizing asynchronous circuits in detail.
This book is different from most of the other books in the switching
theory area in that instead of covering many different topics, it covers
the topic of asynchronous sequential switching circuits thoroughly and
informally. The thorough treatment provides the reader with a clear
deep understanding of the subject. The informal treatment along
with many examples, makes the book easier to read without requiring
special mathematical background. With some general knowledge of
combinational circuits and the Kanaugh map, a reader who is not
in the switching theory area can understand most of the material. The
book contains updated material and the presentation is clear. It
contains solutions to about one-third of the problems and a large
number of references. The book is suitable for self-study. It is a good
reference book for people in the control area who are interested in the
application of switching theory to control circuits. It can also be used
as a text for a second course in switching theory.