,

INTRODUCTION
TO
OPERATIONS
RESEARCH
JOSEPH G. ECKER
MICHAEL KUPFERSCHMID
Rensselaer Polytechnic Institute

JOHN WILEY & SONS

NEW YORK CHICHESTER
BRISBANE TORONTO
SINGAPORE

;1

t..

Copyright

1988, by John Wiley & Sons, Inc.

All rights reserved.

Published

simultaneously

in Canada.

Reproduction or translation of any part of

this work beyond that permitted by Sections
107 and 108 of the 1976 United States Copyright
Act without the permission of the copyright
owner is unlawful. Requests for permission
or further information should be addressed to
the Permissions Department,
John Wiley & Sons.
Library of Congress Cataloging in Publication Data
Ecker, Joseph G.
Introduction to operations research.
Includes bibliographies and indexes.
1. Operations research.
I. Kupferschmid,
II. Title.
T57.6.E331988
Printed

658.4'03'4

in the Republic

of Singapore

10 9 8 7 6 5 4 3

87-23147

Michael.

ABOUT THE
AUTHORS

,
IJ

(
,
I

is Professor and Chairman of the Department of Mathematical Sciences at Rensselaer Polytechnic Institute in Troy, New York. All of
his degrees were received from the University of Michigan-a B.A. in mathematics in 1964, an M.S. in mathematics in 1966, and a Ph.D. in mathematics in
1968. Dr. Ecker has been a faculty member at RPI since 1968 and teaches a
variety of courses in operations research and mathematics. He is also an affiliated
faculty member in RPI's Department of Decision Sciences and Engineering
Systems. In 1975 he was awarded a Fulbright-Hays Research Award and a
NATO Postdoctoral Fellowship in Science, and spent the year as a visiting
professor in the Center for Operations Research and Econometrics at the Catholic University of Louvain in Belgium. In 1983 he was a visiting professor in the
mathematics department of the Ecole Polytechnique Federale de Lausanne in
Switzerland. His research interests are in operations research and include the
theory and applications of linear and nonlinear programming, multiple objective
programming, geometric programming, and algorithm development. He is the
author or coauthor of more than 45 research publications and is a consultant
for several major corporations. Dr. Ecker has been an associate editor of Operations Research and is currently an editor of SIAM Review.

JOSEPH

G. ECKER

KUPFERSCHMID is a computing
nology Services Department at Rensselaer
faculty member in RPI's Department of
Systems. He received a B.S. in electrical

MICHAEL

consultant in the Information TechPolytechnic Institute and an adjunct

Decision Sciences and Engineering
engineering from RPI in 1968 and
Vll

viii

ABOUT THE AUTHORS

worked for the next three years at Sikorsky Aircraft designing and flight testing
helicopter autopilots. After returning to RPI for a Master of Engineering degree,
which was awarded in 1972, he studied theatre engineering at the Yale School
of Drama and worked for six more years in industry as a control systems engineer.
In 1978 he resigned his position as design supervisor for the controls division of
the J. R. Clancy Company and returned to RPI for graduate work in operations
research and statistics, leading to a M.S. in 1980 and a Ph.D. in 1981. Dr.
Kupferschrnid has taught courses in operations research and in computing at
RfI and is an author of seven research papers. His research interests are in the
experimental evaluation of algorithm performance, the development of nonlinear optimization methods, and the applications of mathematical programming.
Dr. Kupferschmid is a registered Professional Engineer.

PREFACE

This textbook is intended for use in a two-semester sequence of courses introducing the mathematical methods of operations research. Part I can also be used
alone for a one-semester course on linear programming. We have chosen to
provide deep and thorough coverage of the most important methods in operations research rather than a superficial treatment of a larger number of topics.
The level of exposition is appropriate for juniors and seniors who are majoring
in engineering, computer science, mathematics, and quantitative methods in
management.
The basic techniques of operations research are simple and straightforward,
and only a small amount of advanced mathematics is needed for a technically
accurate introduction to the subject. This textbook assumes a knowledge of high
school elementary algebra and a familiarity with simple matrix notation such as
would be introduced in the first class of an undergraduate linear algebra course.
In addition, Chapters 9 and 10 assume a knowledge of elementary differential
calculus, and Part III assumes a knowledge of elementary probability and sta.tistics. A concise appendix on matrix notation makes the book accessible to
students who have not previously had any linear algebra.
Both the style of exposition and the mathematical notation have been chosen
to reflect the simplicity of the subject, and the readability of the text is considered
more important than rigor. Examples are used extensively to introduce and
motivate the topics. In this way, the presentation reflects the inductive process
of scientific discovery rather than imitating the retrospective deduction that is
typical in research papers. This approach gives the student the opportunity to
LX

PREFACE

rediscover the important results personally instead of merely reading about them
in theorems. Proofs are given for some results, but only after the result has been
illustrated by example, and only when the proof provides a constructive method
for solving problems. Thus the book is not a treatise on mathematical theory.
The simplicity of the methods used in the book means that they can be
deeply understood even by beginning students of the subject, and the treatment
is mathematically precise even though the results are often stated informally.
Thus the discussion is not a cookbook tabulation of trite formulae, and the
I
stydent should reasonably be expected to understand the mathematical basis for
the techniques in addition to being able to apply them.
Answers to selected exercises are given at the end of the book, and a
separate answer book is available that contains complete solutions to all of the
exercises.
The development of this text benefited greatly from the comments and
suggestions made by our colleagues during its use in preliminary form over the
past four years. In particular, we express our gratitude to Professors Carlton E.
Lemke (Rensselaer Polytechnic Institute), Richard T. Wong (Purdue University), and Thomas M. Liebling (Ecole Polytechnique Federate de Lausanne).
We also thank the many RPI students who used the preliminary versions, proofread the text, and tested out the exercises. Special thanks are due to Richard
Sych, Lori Grieb, Laura Ripans, Carla Bryan, and Robert Bosch.

Joseph G. Ecker
Michael Kup[erschmid

Ii

CONTENTS

XU

CONTENTS

Selected References
Exercises

3
3.1

~.2

29
30

THE SIMPLEX ALGORITHM

PROGRAMMING
Standard Form and Pivoting

FOR LINEAR

3.4

3.5

,
3.6
3.7

3.8

37
40
41

Canonical Form

42
42
44

45
46

Optimal, Unbounded, and Infeasible Forms

47

Optimal Form
Unbounded Form
Two Infeasible Forms

47
48
49

Solving Linear Programs in Canonical Form

51

Pivoting to Optimal Form

What Can Go Wrong: Degeneracy and Cycling
Ways to Prevent Cycling
Convergence in the Nondegenerate
Case
Convergence in the Degenerate Case

51
52
54
55

Obtaining Canonical Form from Standard Form

58

Getting an Identity
The Subproblem Technique
Pivoting to Form a Subproblem
Summary of the Subproblem Technique

58
62
62

The Simplex Algorithm

63

The Simplex Algorithm

63

55

59

Reformulating Any Linear Program into Standard Form

64

Maximization Problems
Inequality Constraints
Free Variables

64
64
65

The Method of Artificial Variables

69

Getting b ~ 0
The Artificial Problem
Feasibility of the Original Problem
Canonical Form for the Original Problem

3.9

37

Standard Form
The Simplex Tableau
Pivoting on a Simplex Tableau
Canonical Form
Finding a Better Basic Feasible Solution
The Simplex Rule for Pivoting
The Geometry of a Pivot

3.3

37

69

70
71
71

Pivot Matrices and the Revised Simplex Method

74

Pivot Matrices
The Revised Simplex Method
Tableaus with the Same Basic Sequence

74

3.10 Computer Solution of Linear Programs

Computer Cycling
Controlling Roundoff Errors
Tolerances and Errors
Other Practical Considerations

Selected References
Exercises

76

78

79
79
80
81
82
83
83

CONTENTS

4
4.1

4.2
I

4
4.3

GEOMETRY OF THE SIMPLEX ALGORITHM

Geometry of Pivoting

91
91

Graphical Representation of the Feasible Set

Extreme Points and Basic Feasible Solutions
The General Case
Alternative Representation of the Feasible Set
Graphical Interpretation of Canonical Form Tableaus

91
93
96
97
98

Convex Sets

99

Definition of a Convex Set

Convexity of the Feasible Set

99
100

Multiple Optimal Solutions

101

Finding All Optimal Solutions

Convexity of the Set of Optimal Solutions
Optimal Rays

101
102
103
105
105

Selected References
Exercises

5
5.1
5.2

5.3
5.4

5.5
5.6
5.7

6
6.1
6.2

6.3

I
\

Xlll

DUALITY IN LINEAR
The Standard Dual Pair

PROGRAMMING

109
109

Duality Relations

110

Getting the Dual Solution from a Primal Solution

111

Relationships between Optimal Tableaus

Constructing an Optimal Dual Vector

111
113

Economic Interpretation

115

of Dual Variables

The Complementary Slackness Conditions

118

Finding the Dual of Any Linear Program

119

Dual of a Linear Program in Standard Form

A More Complicated Example
Dual of the Transportation Problem

120
121
122

The Dual Simplex Method

124

Another Example of Dual Simplex Pivoting

Using the Dual to Computational Advantage

127
US

A Difficult Primal Problem Having an Easy Dual

128

Theorems of the Alternative

130

Selected References
Exercises

132
133

SENSITIVITY ANALYSIS IN LINEAR

PROGRAMMING
The Brewery Problem
Changes in Production Requirements

139

Changes in Nonbasic Variables

Increasing Basic Variables
Decreasing Basic Variables
When a Nonbasic Variable Exceeds Its Minimum Ratio

141
142
144
145

Changes in Available Resources

145

Changes in One Resource

Simultaneous Changes in Several Resources
Computing Shadow Prices

146
148
150

139
141

XLV

6.4
6.5

CONTENTS

Changes in Selling Prices

151

Simultaneous Changes in Prices

153

Adding New Products or Constraints

154

New Products
Technology Changes
New Constraints

154
155
156
157
157

Selected References
Exercises
I

J,
7.1
7.2

7.3
7.4
7.5

7.6
7.7

LINEAR PROGRAMMING MODELS FOR

NETWORK FLOW PROBLEMS
The Transportation Problem

161

The Transportation Tableau

166

Using the Dual to Improve the Current Solution

167

Obtaining an Optimal Transportation Tableau

A Final Comparison with the Simplex Algorithm

174
175

The Transportation Algorithm

Finding an Initial Basic Feasible Solution
Variations of the Transportation Problem

176
178
181

Unequal Supply and Demand

The Transshipment Problem
Multiple Optimal Solutions

181
183
186

The Assignment Problem

General Network Models

187
190

The General Network Flow Model

Spanning Trees and Basic Feasible Solutions
Solving a General Network Flow Problem
Summary of the General Network Flow Algorithm
Finding an Initial Feasible Spanning Tree

191
193
195
198
199

Selected References
Exercises

201
201

PART II
INTEGER, NONLINEAR AND DYNAMIC
PROGRAMMING
8
8.1
8.2
8.3

8.4

161

209

INTEGER PROGRAMMING
The Integer Programming Problem
Implicit Enumeration
Solution by Branch and Bound

211
211
214
217

The Branch-and-Bound Algorithm

A Branch-and-Bound Example in Detail
The Order of Selecting Unfathomed Nodes
Practical Considerations in Using Branch and Bound

217
218
225
227

0-1 Integer Programs

227

A 0-1 Example
Looking Ahead
How Far to Look Ahead
Getting Nonnegative, Increasing Cost Coefficients
The Branch-and-Bound Algorithm for 0-1 Programs

228
233
237
238
239

CONTENTS

8.5

8.6
I

Ii

Integer Programming Formulation Examples

242

The Oakwood Furniture Problem Revisited

The Knapsack Problem
Capital Budgeting
Facility Location
The Traveling Salesman Problem
A Scheduling Problem

242
243
243
244
246
247

Integer Programming Formulation Techniques

Writing General Integer Programs as 0-1 Programs

250

Selected References
Exercises

250
251
251
253
253

NONLINEAR PROGRAMMING
A Nonlinear Programming Problem

259
259

Contour Plots
Assumption of Continuity
Algorithms for Nonlinear Programming

261
263
263

Unconstrained Minimization

264

Maxima and Minima in One Dimension

Maxima and Minima in n Dimensions

264
267

Equality Constraints

272

Parametric Representation
of Equality Constraints
Several Equality Constraints
Several Parameters
The Method of Lagrange
Classifying Lagrange Points
An Important Use of the Lagrange Multiplier Theorem

273
278
280
280
283
286

Inequality Constraints

287

Active and Inactive Constraints

The Orthogonality Condition
The Karush-Kuhn- Tucker Conditions

288
290
291

Convexity

295

Convex Functions
Convexity and Minima
Checking Whether a Function is Convex

295
296
297

Karush-Kuhn- Tucker Theory of Nonlinear Programming

299

The KKT Method

Strategy in Solving the KKT Conditions

300
302

Numerical Methods of Nonlinear Programming

304

Line Searching
The Method of Steepest Descent
The Generalized Reduced-Gradient
The Ellipsoid Algorithm
Conclusion

306
307
311
315
322

Mixed-Integer Programs
Enforcing Logical Conditions

9
9.1

9.2

9.3

9.4

9.S

9.6

9.7

9.8

xv

Method

Nonlinear Programs with Special Structure

322

Quadratic Programming
Geometric Programming
Separable Programming
Other Theoretical Aspects of Nonlinear

322
326
329
332

Programming

XVl

CONTENTS

10
10.1

110.2
10.3
10.4
10.5
10.6

Selected References
Exercises

333
334

DYNAMIC PROGRAMMING
An Introductory Example

347
347

The Stagecoach Problem

The Backward Recursive Relations
The Forward Recursive Relations
Basic Features of a Dynamic Programming

347
349
351
352

Formulation

A Loading Problem
The Boxes Problem
An Equipment Replacement Problem
Problems with Several State Variables

353
356
359
363

The Curse of Dimensionality

364

Continuous State Dynamic Programming Problems

364

A Simple Example
A More Complicated

364
366

Example

Selected References
Exercises

PART III
PROBABILISTIC
11

11.1
11.2

369
369

MODELS

375

QUEUEING MODELS
A Simple Example and Basic Terminology
A Simple Model Suggested by Observation

377
377
379

Consequences of Exponentially Distributed Interarrival

Special Case 1: Arrivals But No Departures
Special Case 2: Departures But No Arrivals

11.3
11.4
11.5
11.6
11.7
11.8
11.9
11.10

12.1

382
387
390

A More General Arrivals and Departures Model

Steady-State Queueing Systems
A Single-Server System with Constant Rates
Operating Characteristics from Basic Principles
Multiple-Server Queues
Finite Queues
Limited-Source Queues
Optimization in Queues

391
394
397
399
402
405
407
409

Optimizing over the Number of Servers, s

Optimizing over the Mean Service Rate, J.L
Optimizing over the Mean Arrival Rate, A

411
412
412

11.11 Queueing Models Involving Nonexponential

12

Times

Distributions

413

The Kendall Notation

414

Selected References
Exercises

415
415

INVENTORY MODELS
Economic Order-Quantity Models

421
421

Economic
Economic
Economic
Economic

Order-Quantity
Order-Quantity
Order-Quantity
Order-Quantity

Models
Models
Models
Models

with
with
with
with

No Shortages Allowed
Shortages Allowed
Price Breaks
Several Inventories

422
424
426

427

,
r

CONTENTS

12.2

Dynamic Inventory Models with Periodic Review

A Single-Period Model
A Two-Period Model

429
432
437
441
441
449

Selected References
Exercises

452
452

SIMULATION
Next-Event Simulation

Selected References
Exercises

457
458
459
459
463
463
464
464
464
465
466
467
468
469
471
473
475
475
477
480
480
484
487
488

APPENDIX

491

Concave Cost Functions

Convex Cost Functions

12.3

IJ

13
13.1

Stochastic Inventory Models

Observing a Queueing System

Constructing an Event List

13.2

Statistical Analysis of Simulation Results

Estimating Land L,
Estimating Wand
Estimating Server Utilization
Estimating Probability Distributions
Checking Random Interevent Times

w,

13.3

Generating Pseudorandom Numbers

Random and Pseudorandom Number Sequences
Random-Number Generators
The Multiplicative Congruential Algorithm
Other Uniform Random-Number
Generators
Nonuniform Distributions

..

13.4

Monte Carlo Simulations

Evaluation of an Integral
The Metropolis Algorithm

13.5

Practical Considerations in Simulation Modeling

Computer Programs for Simulation
Logical and Statistical Validity

A.1
,

'1

NOTATIONAL
ALGEBRA
Matrices

CONVENTIONS

FOR MATRIX

Equality of Two Matrices

Multiplying Two Matrices

t
A.2
A.3
A.4
A.5
A.6

XVll

Systems of Linear Equations

The Sum of Two Matrices
Multiplying a Matrix by a Real Number
The Transpose of a Matrix
Some Simple Matrix Identities
Selected References
Answers to Selected Exercises
Index

491
492
492
493
494
494
494
495
495
497
505