OPERATIONS

RESEARCH
 About the Author
P. Sankara Iyer is presently Retired Reader in Mathematics at Alagappa University, Karaikudi in Tamil
Nadu. After holding the position of a Lecturer in Mathematics at Alagappa College from 1966 to 1981,
he became Professor in 1981. He then joined Alagappa University in 1985 and retired as Reader in
Mathematics in the year 2001. He is M.Sc. (Mathematics), M.Phil. (Mathematics) and Ph.D. (Computer
Applications in Operations Research). He has done a Certificate course in Computer Programming and
a Certificate course in Evaluation Methodology and Examination by A.I.U. His teaching experience is
vast—B.Sc. (19 yrs), M.Sc. (19 yrs), M.C.A. (15 yrs) and M.Phil. (18 yrs). He has taught all subjects at
B.Sc. level; Operations Research, Differential Equations, Differential Geometry, Classical Mechanics,
Space Mechanics, Quantum Mechanics, Algebra, Statistics and Numerical Analysis at M.Sc. level;
Resource Management Techniques (OR) and Computer based Numerical Methods at M.C.A. level; and
Algebra at M.Phil. level. His study material for M.Sc. Mathematics (Operations Research and Differential
Equations), M.Sc. Industrial Mathematics (Operations Research)—2 books, M.C.A. (Resource
Management Techniques) and M.B.A. (Operations Research) of Alagappa University book are very
popular among the students. He is a member of (i) Operations Research Society of India, (ii) Ramanujan
Mathematical Society and (iii) Forum for Advancement of Science and Human Development.
Prof. Sankara Iyer’s primary area of research is Operations Research: Fractional Programming
(Computer Applications). He has published seven research papers and guided more than 50 students for
M.Phil. Research Project. He has attended twelve conferences and seminars in Algebra, Operations
Research, Analysis Statistics and Stochastic Processes. He has been the Cultural Coordinator of Alagappa
University for 15 years, Director of Social Service in Alagappa College for 10 Years, Director of Fine
Arts in Alagappa College for 8 Years and Secretary of Staff Club for 6 Years. He is musically inclined
and has composed and tuned nearly 200 songs and given public performances. He has written and
directed social dramas (besides acting as the main protagonist), promoting national integration, communal
harmony and eradication of dowry. Tamil Isai Sangam, Karaikudi conferred the title ‘MUTHAMIZH
SELVAR’ (year 2001), in appreciation of his service in the field of literature, music, and drama. He
gives a musical discourse every year on the life of Saint Thyagaraja during Thyagaraja Aradhana Festival
at Tiruvaiyaru in Thanjavur District, Tamil Nadu.
 OPERATIONS
RESEARCH

P. Sankara Iyer
Reader in Mathematics (Retd)
Alagappa University
Karaikudi, Tamil Nadu

Tata McGraw-Hill Publishing Company Limited

NEW DELHI
McGraw-Hill Offices
New Delhi New York St Louis San Francisco Auckland Bogotá Caracas
Kuala Lumpur Lisbon London Madrid Mexico City Milan Montreal
San Juan Santiago Singapore Sydney Tokyo Toronto
 Tata McGraw-Hill
Published by Tata McGraw-Hill Publishing Company Limited,
7 West Patel Nagar, New Delhi 110 008.
Copyright © 2008, by Tata McGraw-Hill Publishing Company Limited.
No part of this publication may be reproduced or distributed in any form or by any means, electronic, mechanical,
photocopying, recording, or otherwise or stored in a database or retrieval system without the prior written permission of
the publishers. The program listings (if any) may be entered, stored and executed in a computer system, but they may not
be reproduced for publication.
This edition can be exported from India only by the publishers,
Tata McGraw-Hill Publishing Company Limited.
ISBN (13): 978-0-07-066902-4
ISBN (10): 0-07-066902-3
Managing Director: Ajay Shukla
General Manager: Publishing—SEM & Tech Ed: Vibha Mahajan
Sponsoring Editor: Shalini Jha
Editorial Executive: Sandhya Chandrasekhar
Senior Copy Editor: Dipika Dey
Production Executive: Suneeta S Bohra
General Manager: Marketing—Higher Education & School: Michael J Cruz
Product Manager: SEM & Tech Ed: Biju Ganesan
Controller—Production: Rajender P Ghansela
Asst. General Manager—Production: B L Dogra
Senior Production Manager: P L Pandita

Information contained in this work has been obtained by Tata McGraw-Hill, from sources believed to be reliable.
However, neither Tata McGraw-Hill nor its authors guarantee the accuracy or completeness of any information
published herein, and neither Tata McGraw-Hill nor its authors shall be responsible for any errors, omissions, or
damages arising out of use of this information. This work is published with the understanding that Tata McGraw-Hill
and its authors are supplying information but are not attempting to render engineering or other professional services.
If such services are required, the assistance of an appropriate professional should be sought.

Typeset at Text-o-Graphics, B-1/56, Arawali Appartment, Sector 34, Noida 201 301, and printed at
Lalit Offset, 219, F.I.E., Patpar Ganj Industrial Area, Delhi - 110 092

Cover Design: Rashtriya Printers

RZLCRRAXRQBLY
 Foreword
I had the golden opportunity of being a student of Prof. P. Sankara Iyer, the author of this book. He is a
model teacher and we (all his students) consider him as our great Guru and Guide. I feel proud to write
this foreword for a book written by my beloved teacher.
Mathematics is the source of all knowledge. It is a vast ocean and the author of this book has dived
deep into a particular branch of this subject, namely Operations Research.
The contents of this book have been arranged on the principle of gradation. The chapters are in a
logical sequence: Introduction, Linear Programming, Integer Programming, Transportation, Assignment,
Decision Theory, Games Theory, Dynamic Programming, Sequencing, Queueing, Inventory, Replacement,
PERT and CPM, and Simulation.
There are two distinctive features that make this book stand apart. The first one is that the examples
and exercises provided in each chapter are indeed a boon to the students and they can assess their
understanding of each chapter clearly. Another unique feature is that the entire book has been based on
a self-taught new method, culminating from Prof. Sankara Iyer’s rich teaching experience of 35 years.
Also, while structuring this book, he has kept both the instructor and the student in his purview, and
hence it will prove to be a valuable resource for them.
The book covers the syllabi of almost all Indian universities. Therefore not only students of
Mathematics but also students of M.C.A., M.B.A., C.A. and I.C.W.A. will benefit from this book.
I earnestly wish that the student community and faculty members derive maximum benefit from this
magnificent creation.

Dr V. VEMBAIYAN
Principal,
Alagappa Govt. Arts College,
Karaikudi, Tamil Nadu.
 Preface
In recent years, Operations Research—the mathematical analysis of a process, used in making decisions—
has become a very popular branch of mathematics studied by undergraduate and postgraduate students.
This interdisciplinary branch of Applied Mathematics has wide applications in various types of real life
situations. A large number of research papers have been published on this subject and a great deal of
research is being carried out.
Purpose
The primary objective of writing this book is to provide the students with a basic knowledge of the
processes involved in Operations Research and the enormous practical applications of the techniques to
daily life problems. Presently, there are limited good books in the market, with even fewer offering well
illustrated problems and also, the style of presentation in foreign books is not very helpful to our Indian
students—this generates a need for a single book that covers all the topics of this subject in order to
cater to the requirement of students in various universities. This book fulfills the purpose.
Users
This book is primarily meant for the course on Operations Research to be studied by the engineering
students of all branches. It offers coverage of different topics pertaining to the syllabus prescribed for
the students of Mathematics and Statistics (UG and PG), M.C.A., M.B.A., M.Com., B.E. and B.Tech. of
Indian Universities and also, C.A. and I.C.W.A. This book is a fine reference for competitive (UG and
PG levels) and I.A.S. examinations.
Highlights
What makes this book unique is its lucid presentation of the subject aided by plenty of solved and
unsolved problems. Brief yet adequate theory clearly explained in simple language supported by worked
out examples (with step by step solution) and exercise problems facilitate quick learning by the students.
The chapters have been structured in a planned manner so that the students can study the subject
independently.
It is a comprehensive text which helps students in understanding problem solving methods based on
model formulation, solution procedure and analysis. The salient features of the book are
(i) Provides sound knowledge of the OR techniques in managerial decision making.
(ii) Incorporates important chapters like Sequencing Problems, Replacement Models, Project
Scheduling, Artificial Variables and Bounded Variable Techniques.
 viii Preface

(iii) Numerous solved examples give an overall view of the applications of OR techniques. Solved
examples from various university question papers are an effective tool for examinations.
(iv) To clarify concepts, figures and tables are used exhaustively.
(v) To carry out self evaluation, students can solve multitude of exercises that are provided in each
chapter along with their answers.

The Sigma Format

This book is based on the Sigma Format that offers a clear understanding of fundamental theory supported
with exhaustive examples, solved using a step-by-step approach. To get a better grasp on the subject,
lots of exercises along with answers are provided in the chapters. This well structured format has proven
to be a handy resource material for students as well as teachers.

Chapters Organisation
In the first chapter, the importance of the study of OR and its various applications have been discussed.
Chapter 2 describes the method of mathematical formulation of a real life problem (in the form of a
Linear Programming Problem) and the graphical method of solving an L.P.P having two decision variables.
Simplex Method, an important method of solving a general L.P.P has been discussed in detail in chapter
3. Various modifications of simplex method such as Big M method, Two Phase Method, Dual Simplex
Method, Principle of Duality, and Revised Simplex Method have been covered in the subsequent chapters
4, 5, 6 and 7. When the variables are restricted to assume values within certain given limits, we have to
apply Bounded Variable Technique described in chapter 8. Another restriction that the variables can
take only integer values, leads to Integer Programming which is given in chapter 9. Also, a study of the
effect on the optimal solution due to changes in the values of the parameters, which is called Sensitivity
Analysis, has been discussed in chapter 10.
Then follow chapters 11, 12, 13, 14, 15 and 16 dealing with applications of OR techniques namely
Transportation and Assignment Problems, Decision Theory, Game Theory, Dynamic Programming and
Sequencing of Jobs. Another set of applications have been provided in detail in chapters 17 (Queueing
theory), 18 (Inventory models), 19 (Replacement theory), and 20 (P.E.R.T - C.P.M). Simulation Techniques
of solving different types of problems have been taken up for discussion in the last chapter (chapter 21).

Acknowledgements
I would like to express my gratitude to all the authors who were a source of inspiration while writing this
book. I wish to express my indebtedness to my teachers and well wishers. I wish to acknowledge my
sincere thanks to late Dr M K Venkataraman for his inspiration and encouragement in this project. My
thanks are also due to my wife and children, and M.Phil. Scholar, R G N Nagarajan of Alagappa University
for their continuous help and cooperation in fulfilling the project. I express my deepest sense of gratitude
to the following reviewers, whose suggestions have helped me bring the text to its present form.

J S Oberoi S K Tiwari
Baba Banda Singh Bahadur Institute of National Institute of Technology, Jalandhar
Engineering, C M Agarwal
Fatehgarh sahib, Punjab Maulana Azad National Institute of Technology,
Bhopal
 Preface ix

Rajiv Lochan S Srinivasan

The Institute of Chartered Financial Analysts of Crescent Engineering College, Chennai
India, Jaipur S Nadarajan
S R Kapoor Dept. of Mechanical Engineering
Dept. of Electrical Engineering GCT, Coimbatore
Kota Engineering College, Kota A Rajesh
Sinha Roy Dept. of Mechanical Engineering
Dept. of Mathematics Kumaraguru College of Technology
Future Institute of Technology, Kolkata Coimbatore
Shila Das A Ramakrishna Rao
Dept. of Statistics Adi Shankara Instiute of Engineering and
Dibrugarh University, Assam Technology,
Chakradhar Dash Kalady
N M Institute of Technology K R Divakara Roy
Khandagiri, Bhubaneswar Dept. of Mechanical Engineering,
Reena Pant College of Engineering
Mumbai University, Mumbai Andhra University, Visakhapatnam

Arun Kumar, T. Chandrashekar

M H Saboo Siddik College of Engineering, Dept. of Mechanical Engineering,
Mumbai M S Ramaiah Institute of Technology, Bengaluru

I am truly grateful to the publishers at McGraw-Hill Education India for coming forward to publish
this book and for their efficient management of the project. I am sure that this book will receive good
response from the teachers and the taught. Suggestions from the readers for the improvement of the
book are most welcome and will be duly acknowledged and incorporated.

P. Sankara Iyer
Email id: psankaraiyer@yahoo.com
 Contents
Preface vii

1 OPERATIONS RESEARCH 1
1.1 What is Operations Research? 1
1.2 Characteristics of OR 1
1.3 Phases of OR 2
1.4 Scope of OR 2
1.5 Drawbacks and Difficulties of OR 2

2 LINEAR PROGRAMMING 3
2.1 Introduction 3
2.2 Formulation 3
2.3 Graphical Method of Solution 13
Exercises 20
Answers 22

3 SIMPLEX METHOD 23
3.1 Introduction 23
3.2 Simplex Method 24
Short Answer Questions 35
Exercises 36
Answers 39

4 ARTIFICIAL VARIABLES 41
4.1 Artificial Variables 41
4.2 Big M Method (Charne’s Penalty Method) 41
4.3 Two-Phase Method 46
Short Answer Questions 49
Exercises 50
Answers 53
 xii Contents

5 DUAL SIMPLEX METHOD 54

5.1 Dual Simplex Method 54
Short Answer Questions 57
Exercises 57
Answers 58

6 DUALITY 59
6.1 Rules for Writing the Dual of a Primal 59
6.2 Duality Theorems 62
6.3 Principle of Duality 63
6.4 Economic Interpretation of the Dual 66
6.5 Complementary Slackness Theorem 67
Short Answer Questions 67
Exercises 68
Answers 69

7 REVISED SIMPLEX METHOD 71

7.1 Step-by-Step Procedure 71
7.2 Advantages of Revised Simplex Method Over the Ordinary Simplex Method 79
Exercises 80
Answers 81

8 BOUNDED VARIABLE TECHNIQUE 82

8.1 Step-by-Step Procedure of Bounded Variable Technique 82
Exercises 88
Answers 88

9 INTEGER PROGRAMMING 89
9.1 Gomary’s Cutting Plane Method 90
9.2 Gomary’s Algorithm (Step-by-Step Procedure) 90
9.3 Mixed Integer Programming 93
9.4 Branch and Bound Method 97
Exercises 95, 101
Answers 96, 102

10 SENSITIVITY ANALYSIS 103

10.1 Discrete Changes in cj 103
10.2 Discrete Changes in bi 105
10.3 Changes in the Matrix Elements aij 106
10.4 Addition of a Variable 107
10.5 Addition of a Constraint 108
Exercises 109
Answers 111
 Contents xiii

11 TRANSPORTATION PROBLEM 112

11.1 Mathematical Formulation 112
11.2 Initial Basic Feasible Solution 113
11.3 Testing for Optimality 120
11.4 Optimization—MODI Method (Modified Distribution Method) 121
11.5 Degeneracy 123
11.6 Unbalanced Transportation Problem 125
11.7 Maximization Type 128
Exercises 110, 129
Answers 120, 132

12 ASSIGNMENT PROBLEMS 134

12.1 What is an Assignment Problem? 134
12.2 Mathematical Formulation 134
12.3 Method of Solving an AP 135
12.4 Step-by-Step Procedure of Hungarian Method (Minimization Problem) 135
12.5 Maximization Assignment Problem 140
12.6 Unbalanced Assignment Problem 141
12.7 Travelling Salesman Problem 142
Exercises 144
Answers 147

13 DECISION THEORY 149

13.1 What is Decision Theory? 149
13.2 Steps in Decision Making 149
13.3 Maximin and Maximax Criteria 149
13.4 Hurwicz Criterion and Laplace Criterion 150
13.5 Minimax Regret Criterion (Savage Criterion) 151
13.6 Decision-Making Under Risk 152
13.7 Expected Opportunity Loss Criterion (EOL) 154
Exercises 156
Answers 158

14 THEORY OF GAMES 159

14.1 Introduction 159
14.2 Two-Person Zero Sum Games 159
14.3 Games with Saddle Point (Pure Strategy) 160
14.4 Games without Saddle Points (Mixed Strategy) 161
14.5 Formula for Finding the Value of the Game in Case of 2 × 2 Games
without Saddle Point 161
14.6 Dominance Property 163
14.7 Graphical Method 166
14.8 Algebraic Method 170
 xiv Contents

14.9 Linear Programming Method 171

Exercises 173
Answers 175

15 DYNAMIC PROGRAMMING 178

15.1 Introduction 178
15.2 Dynamic Programming Method 178
15.3 Applications to Mathematical Problems 185
15.4 Solution of Linear Programming Problems 187
Exercises 189
Answers 193

16 SEQUENCING PROBLEMS 194

16.1 Introduction 194
16.2 Processing n Jobs through Two Machines 194
16.3 Processing n Jobs on Three Machines 197
16.4 Processing n Jobs through m Machines 199
Exercises 201
Answers 203

17 QUEUEING THEORY 205

17.1 Introduction 205
17.2 Characteristics of a Queueing System 205
17.3 Symbols and Assumptions 206
17.4 Transient and Steady States 206
17.5 Model I (M/M/1/• Queueing Model) 206
17.6 Model 2—M/M/1/N (Queue with Finite Capacity N) 212
17.7 Model 3—(Multichannel System) M/M/C/• 215
17.8 Model 4—(Multichannel System
with Limited Capacity) M/M/C/N 218
Exercises 220
Answers 223

18 INVENTORY MODELS 224

18.1 Introduction 224
18.2 Costs Involved in Inventory Control 224
18.3 Characteristics of Inventory System 225
18.4 Deterministic Inventory Models with no Shortage 225
18.5 Probabilistic Model 237
Exercises 239
Answers 242
 Contents xv

19 REPLACEMENT MODELS 243

19.1 Introduction 243
19.2 Replacement of Items which
Deteriorate in Efficiency with Time 243
19.3 Replacement of Items that Fail Completely 249
Exercises 251
Answers 254

20 PROJECT SCHEDULING (PERT AND CPM) 225

20.1 Introduction 255
20.2 Network Diagrams 255
20.3 Critical Path Method 257
20.4 PERT Calculations 262
Exercises 271
Answers 276
Short Answer Questions 279

21 SIMULATION 280
21.1 Introduction 280
21.2 Event Type Simulation 280
21.3 Monte–Carlo Technique 281
Exercises 284
Answers 284

Index 286

Operations Research
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
1.1 WHAT IS OPERATIONS
RESEARCH?
The term Operations Research (OR) relates to
military operations during the second World War.
Scientists used various techniques to deal with
strategic and tactical problems during the war.
After the war military OR group scientists tried
to apply OR techniques to civilian problems
relating to business, industry, and research
development.
During the 1950s educational institutions
introduced OR in their curricula. Today service
organizations such as airlines, railways, hospitals,
libraries and banks employ OR to improve their
efficiency. OR is a science which provides
mathematical techniques and algorithms for
solving decision problems. It is also an art in the
sense that it deals with collecting data for model
construction, validation of the model and
implementation of the solution obtained.
In other words, OR is a scientific approach to
problem solving for executive management. The
problems involve integrated systems of men,
machines and materials.
1
1.2 CHARACTERISTICS OF OR
An important characteristic of OR is that it deals
with a problem as a whole. Before arriving at a
decision it takes into consideration all possible
interactions between various parts or departments
of an organization. For example, if a problem on
inventory management is taken, the production
department will be interested in continuous
production, so as to reduce set-up cost and other
maintenance costs. But this will give rise to
problems in finance, marketing and other
departments. Hence it is necessary to analyse the
factors affecting all parts of the organization and
prepare a mathematical model. A solution to this
model will optimize the profits of the organization.
Another characteristic is that OR involves a
team of experts from different scientific and
engineering disciplines. After a careful analysis
of the problem, a suitable mathematical model is
prepared and this model is solved using scientific
techniques. Thus a decision or solution is obtained
which gives an optimal solution to the problem.
The importance of OR has been felt very much
in industries. Inventory management, replacement
 2 Operations Research
of equipments, deterministic and probabilistic
decision-making in resource allocation, production
planning, sequencing of jobs, assignments of jobs
to machines and so many other problems are solved
using OR techniques.
1.3 PHASES OF OR
The first step is formulation of the problem. We
have to identify the objective, the decision
variables and the constraints involving the
variables. After that a mathematical model has to
be constructed, with an objective function to be
optimized and constraints in the form of equalities
or inequalities. The solution of the model can be
obtained by analytic, iterative or Monte-Carlo
method depending on the structure of the model.
These mathematical models are absolutely free
from the influence of qualitative or emotional
human factors which affect decision-making.
1.4 SCOPE OF OR
Some of the industrial and business problems
which can be analyzed by OR technique include
1. Finance and Accounting Investment
and portfolio management, auditing, cash
flow analysis, capital budgeting etc.
2. Marketing Product mix, marketing,
planning, sales, allocation and assignment,
advertising and media planning etc.
3. Purchasing and procurement
Optimal buying, re-ordering, transportation,
replacement etc.
4. Production Warehouse, distribution,
transportation, production planning,
blending, inventory control, maintenance,
project scheduling, allocation of resources,
etc.
5. Management Manpower planning,
project management, etc.
1.5 DRAWBACKS AND
DIFFICULTIES OF OR
There are certain aspects of OR which are difficult
and which need careful approach. First, formulation
of the problem is very important. The problem must
be properly selected and accurately defined, which
involves some difficulties. Data collection may also
consume a large portion of time and money. Also
OR analysis is mainly based on past observations.
There is no guarantee that the same set of conditions
would continue in future also. Again decisions must
be made at the right time and any delay may cause
loss or damage. Also, the cost involved in the project
is an important factor. It is the responsibility of OR
experts to translate the concepts and ideas into simple
procedures easily comprehensible to managers and
workers alike. They should also ensure that the new
proposals are properly implemented. Only then OR
study will be useful to humanity.

Linear Programming
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
2.1 INTRODUCTION
Linear programming is one of the various
techniques of optimization (Maximizing the gain
and minimizing the loss). It deals with problems
of allocating limited resources to obtain a desired
optimal solution, subject to a set of given condi-
tions (constraints).
A linear programming problem (LPP) consists
of an objective function which is to be optimized
and a set of constraints to be satisfied by the
variables. Usually the problem is of the form
Maximize (or Minimize)
Z = C1x1 + C2x2 + C3x3 + … + Cnxn
subject to the constraints
a11x1 + a12x2 + … + a1nxn ( ≥, £, or =) b1
a21x1 + a22x2 + … + a2nxn ( ≥, £, or =) b2
: : :
: : : Example 2.1 (Production Allocation Problem)
am1x1 + am2x2 + … + amnxn ( ≥, £, or =) bm
x1, x2, x3, …, xn > 0.
The objective function is Z and the variables
x1, x2, …, xn are called decision variables. The
constraints are linear inequalities or equalities and
2
they correspond to the conditions on the use of
the available resources. The objective function
may correspond to total cost, profit, total quantity
to be produced and other similar objectives.
2.2 FORMULATION
Given a situation, describing the availability of
resources and conditions and the object to be
optimized, we have to convert it into a
mathematical model. This process is called
formulation. It consists of the following steps:
Step 1 Defining the decision variables and writ-
ing the objective function.
Step 2 Expressing the conditions of availability
of the resources in the form of linear inequalities
or equalities using the decision variables.
A company manufacturers three items A, B and C.
These items are processed on three machines
M1, M2 and M3. The time required for each product
in each machine is given below. Also the total time
of availability of each machine is given.
 4 Operations Research
Table 2.1
Machine Time per unit (hours) Availability of the
A B C machine hr/day
M1 1 1.5 2 18
M2 2 1 1 20
M3 1 2 2 16
The company gets a profit of Rs 20, Rs 30 and
Rs 40 per unit of A, B and C respectively.
Determine the number of units of each product to
be manufactured per day in order that the total
profit is maximum.
Formulation The objective is to determine the
number of units to be manufactured. Hence the
decision variables are x1, x2, x3 corresponding to
the number of units of A, B and C respectively.
The profits are Rs 20, Rs 30 and Rs 40 per unit of
the products. Thus the objective function becomes
Z = 20x1 + 30x2 + 40x3
which is to be maximized.
Considering the machine M1 whose availability
is 18 hrs we have the condition that the total time
taken by M1 should not exceed 18 hrs. Therefore
we get a constraint
(1)x1 + (1.5)x2 + (2)x3 £ 18
i.e., x1 + 3/2 x2 + 2x3 £ 18
Similarly for M2 and M3 we get
2x1 + x2 + x3 £ 20
x1 + 2x2 + 2x3 £ 16
Also the number of units cannot assume
negative value and this leads to the constraints
x1 ≥ 0, x2 ≥ 0 and x3 ≥ 0.
Now the process of formulation is complete.
The linear programming problem has the form.
Maximize Z = 20x1 + 30x2 + 40x3 subject to the
constraints
x1 + 3/2x2 + 2x3 £ 18
2x1 + x2 + x3 £ 20
x1 + 2x2 + 2x3 £ 16
x1, x2, x3 ≥ 0.
Example 2.2 (Advertising media selection
problem) A company wants to work out a plan
for advertising its product in three different media,
TV, Radio and Newspaper. The details are given
by the following table:
Table 2.2
TV Radio Newspaper
Rs Rs Rs
Cost of advertising 4000 3000 1000
Number of customers 20000 30000 20000
Impact rate .8 .6 .5
The company cannot spend more than Rs 80000
for advertising. Advertisement should be given
atleast once in TV. It is required to determine the
number of times the advertisement should be given
in each medium so that the rate exposure is
maximum.
[Rate exposure = (impact rate) (number of
customers)]
Formulation It is required to determine the
number of times the advertisement is to be given
in each medium. Therefore we take the decision
variables x1, x2 and x3 to represent the frequency
in TV, Radio and Newspaper respectively.
The rate exposure of one advertisement in TV
is given by (0.8)(20000). In Radio it is given by
(0.6)(30000) and in Newspaper it is (0.5)(20000).
Therefore the total rate exposure is
(0.8)(20000)x1 + (0.6)(30000)x2 + (0.5) (20000)x3
and this is to be maximized. Hence the objective
function is
Z = 16000x1 + 18000x2 + 10000x3
The total cost of advertising is
4000x1 + 3000x2 + 1000x3 and this should not
exceed Rs 80000. This leads to the constraint
4000x1 + 3000x2 + 1000x3 £ 80000
Also at least one advertisement has to be given
in TV and this gives another constraint x1 ≥ 1.
Thus the LPP is of the form
Maximize Z = 16000x1 + 18000x2 + 10000x3,
subject to the constraints
4000x1 + 3000x2 + 1000x3 £ 80000
x1 ≥ 1
x2, x3 ≥ 0.
 Linear Programming 5

Example 2.3 A soft drink plant has two bottling Example 2.4 (Diet Problem) In our daily diet,
machines A and B manufacturing 1 litre and 2 litre proteins, fats and carbohydrates are required with
bottles. The following table gives the a minimum of 5, 2 and 3 units respectively. We
manufacturing data: consider the foodstuff consisting of bread, butter
and milk. The yields of these nutritional
Machine 1 litre bottle 2 litre bottle
requirements and the cost per unit of the food items
A 100/min 60/min are given below:
B 50/min 80/min
Table 2.3
The machines can be run 8 hours per day, 5 days
Food Yield per unit Cost per unit
per week. Weekly production of the drink cannot
Protein Fat Carbohydrate of food
exceed 3,00,000 litre and the demand for 1 litre
and 2 litre bottles are 25,000 and 8000 per week Bread 4 1 2 12
respectively. The manufacturer gets a profit of Butter 3 2 1 60
50 paise on 1 litre bottle and 75 paise on 2 litre Milk 3 2 1 7
bottle. It is required to find the number of bottles
Minimum 5 2 3
to be manufactured in each type so as to maximize requirement
the total profit.
Formulation Let us assume that x1 units of 1 litre It is required to determine the combination of
bottle and x2 units of 2 litre bottle be produced the food items such that minimum requirement
weekly. The total profit is Rs (0.5)x1 + (0.75)x2 is satisfied and the total cost of the food is
and this has to be maximized. Thus we get the minimum.
objective function Z = (0.5)x1 + (0.75)x2. Formulation Let us assume that the numbers of
The machines run for 5 days per week at the units of bread, butter and milk to be consumed are
rate of 8 hrs per day. Hence the total time they run x1, x2 and x3 respectively. The total cost of the food
per week is 40 hrs or 2400 min. On machine A, is 12x1 + 60x2 + 7x3 and this has to be minimized.
100 units of 1 litre and 60 units of 2 litre bottles Hence the objective function is
are produced per minute. Therefore we get the Z = 12x1 + 60x2 + 7x3.
constraint x1/100 + x2/60 £ 2400.
Similarly, on machine B, we get the constraint Considering the minimum requirement of
protein per day we get
x1/50 + x2/80 £ 2400.
4x1 + 3x2 + 3x3 ≥ 5.
Total weekly production cannot exceed
3,00,000 Similarly, we get the constraints for the
x1 + x2 £ 300000. minimum requirement of fat and carbohydrates
as
Considering the demand for the items in the
market, we get the constraints, x1 + 2x2 + 2x3 ≥ 2 and
x1 ≥ 25000 and x2 ≥ 8000. 2x1 + x2 + x3 ≥ 3.
Thus the formulated LPP is Thus the formulation of LPP is
Maximize Z = (0.5)x1 + (0.75)x2 subject to the Minimize Z = 12x1 + 60x2 + 7x3 subject to the
constraints constraints
x1/100 + x2/60 £ 2400 (or) 3x1 + 5x2 £ 720000 4x1 + 3x2 + 3x3 ≥ 5
x1/50 + x2/80 £ 2400 (or) 8x1 + 5x2 £ 960000
x1 + 2x2 + 2x3 ≥ 2
x1 + x2 £ 300000
x1 ≥ 25000 2x1 + x2 + x3 ≥ 3
x2 ≥ 8000. x1, x2, x3 ≥ 0.
 6 Operations Research

Example 2.5 (Blending Problem) A firm Example 2.6 Three grades of coal A, B and C
manufactures an alloy having the following contain phosphorus and ash as impurities. For a
specifications: particular industrial process a maximum of
(i) Specific gravity £ 0.95 100 tons of coal is required which should
(ii) Chromium ≥ 10% contain not more than 3% of ash and not more
(iii) Melting point ≥ 420°C. than 0.03% of phosphorus. The percentage of
Details of raw materials A, B and C and their impurities and the profits from each grade are
properties are given below: given below
Table 2.4 Table 2.5
Property Raw Material Coal Phosphorus (%) Ash (%) Profit (Rs/ton)
A B C
A 0.02 3 12
Specific gravity 0.92 0.96 1.05 B 0.04 2 15
Chromium 8% 12% 15% C 0.03 5 14
Melting point 380°C 400°C 480°C
Find the proportions in which the three grades
Cost of the raw materials are Rs 100, Rs 300 are to be used in order to maximize the profit.
and Rs 50 per ton for A, B and C respectively.
It is required to determine the proportion in Formulation Let x1, x2, x3 be the quantities in
which the raw materials are to be used to tons of grades A, B and C respectively. Then the
manufacture the alloy such that the total cost is objective function (profit) to be maximized is
minimum. Z = 12x1 + 15x2 + 14x3
Phosphorus content should not exceed 0.03%.
Formulation It is assumed that x1, x2 and x3 tons
Therefore we get a constraint
of raw materials A, B and C be used for making
the alloy. The total cost of the raw materials is 0.02x1 + 0.04x2 + 0.03x3 £ 0.03 (x1 + x2 + x3)
100x1 + 300x2 + 50x3 and this has to be minimized. Equivalently, 2x1 + 4x2 + 3x3 £ 3 (x1 + x2 + x3)
Total specific gravity is 0.92x1 + 0.96x2 + 1.05x3 or – x1 + x2 £ 0
and this should be less than 0.95. Therefore we Considering the constraint on the ash contents,
get a constraint 0.92x1 + 0.96x2 + 1.05x3 < 0.95. we get
It is given that the chromium content should be 3x1 + 2x2 + 5x3 £ 3 (x1 + x2 + x3)
greater than 10%. Hence we must have 8x1 +
12x2 + 15x3 > 10. or – x2 + 2x3 £ 0
Also the melting point of the alloy has to be Also the total quantity required is not more than
greater than 420°C. This constraint can be 100. So we have x1 + x2 + x3 £ 100.
written as Thus the formulated LPP is
380x1 + 400x2 + 480x3 ≥ 420. Minimize Z = 12x1 + 15x2 + 14x3 subject to the
Thus the LPP can be written as constraints
Minimize Z = 100x1 + 300x2 + 50x3 – x1 + x2 £ 0
subject to the constraints
– x2 + 2x3 £ 0
0.92x1 + 0.96x2 + 1.05x3 £ 0.95
x1 + x2 + x3 £ 100
8x1 + 12x2 + 15x3 ≥ 10.
380x1 + 400x2 + 480x3 ≥ 420 x1, x2, x3 ≥ 0
x1, x2, x3 ≥ 0 Example 2.7 A farmer cultivates tomato,
carrot and potato in his farm of area 100 acres.
 Linear Programming 7

Average yield per acre is 2000 kg of tomato, Labour cost = Rs 60 (5x1 + 6x2 + 5x3)
2000 kg of carrot and 1000 kg of potato. The cost The total profit P = total sales – total cost.
of the fertilizer is Rs 3 per kg and the requirement P = Rs (10000x1 + 16000x2 + 10000x3) –
of fertilizer is 100 kg each for tomato and carrot (300x1 + 300x2 + 180x3) – (300x1 +
and 60 kg for potato per acre. Labour required is 360x2 + 300x3)
5 man-days for tomatoes and potatoes and 6 man-
days for carrot per acre. A total of 400 man-days i.e. P = Rs (9400x1 + 15340x2 + 9520x3)
are available at Rs 60 per man-day. which is to be maximized.
Formulate this problem as a LPP so as to Total area £ 100 acres
maximize the farmer’s profit if he can sell the items \ x1 + x2 + x3 £ 100
at Rs 5 per kg for tomato, Rs 8 per kg for carrot Total labour is restricted to 400 man-days
and Rs 10 per kg for potato. i.e., 5x1 + 6x2 + 5x3 £ 400
Formulation Let us assume that the farmer Thus the LPP is
cultivates tomato, carrot and potato in x1, x2 and Maximize P = 9400x1 + 15340x2 + 9520x3
x3 acres respectively. Then the total sale of item is subject to the constraints
Rs (2000x1 (5) + 2000x2 (8) + 1000x3 (10)). x1 + x2 + x3 £ 100
Total expenditure for fertilizer is 5x1 + 6x2 + 5x3 £ 400
Rs 3 (100x1 + 100x2 + 60x3) x1, x2, x3 ≥ 0

EXERCISES
■ ■ ■ ■ ■ ■ ■

(Formulate into LPP) are Rs 20 for drilling, Rs 30 for shaping

1. Three articles A, B and C have weight, and Rs 30 for polishing. The capacities of
volume and cost as given below: The total each machine (parts per hour) are given
weight cannot exceed 2000 units and total below:
volume cannot exceed 2500 units. Find the Capacity per hour
number of articles to be selected from each Machine A B C
type such that the total cost is maximum.
Drilling 25 40 25
Weight Volume Cost (Rs) Shaping 25 20 20
A 4 9 5 Polishing 40 30 40
B 8 7 6
C 2 4 3 It is required to determine how many parts
of each type should be produced per hour in
2. An electronic company produces three types order to maximize the profit.
of parts for washing machine. It purchases 3. A furniture manufacturing company plans to
unfinished articles from a foundry and does make chairs and tables. The requirements are
the jobs of drilling, shaping and polishing. given below:
The selling prices of the parts A, B and C are Item Requirement Profit (Rs)
Rs 8, Rs 10 and Rs 14 respectively. Their Timber (in board ft) Man-hours
purchase costs are Rs 5, Rs 6 and Rs 10
respectively. The company has only one Chair 5 10 45
machine for each job. Machine costs per hour Table 20 15 80
 8 Operations Research

Totally 400 board feet of timber and 450 man- 7. An oil refinery has to decide on the optimal
hours are available. Determine the number of mix of two possible blending processes. The
chairs and tables to be manufactured in order input and output per production run are given
to maximize the profit. below.
4. A firm manufactures four different machine Process Input Output
parts A, B, C and D using copper and zinc.
Crude A Crude B Gasoline X Gasoline Y
The requirement of copper and zinc for each
part and their availability and the profit 1 5 3 5 8
earned from each part are given below: 2 4 5 4 4

Item Requirement (kg) Profit (Rs) The maximum quantity of crude A and B
Copper Zinc available are 200 units and 150 units
respectively. At least 100 units of X and
A 5 2 12 80 units of Y are required in the market. The
B 4 3 8 profit per production run from process 1 and
C 2 8 14
process 2 are Rs 300 and Rs 400 respectively.
D 1 1 10
Determine the optimal number of production
Availability 100 75 runs of process 1 and process 2.
How many of each part should be 8. M.K.V. Foods company is developing a diet
manufactured in order to maximize the supplement called Hi-Pro. The specifications
profit? and the calorie, protein and vitamin content
5. A company produces three types of products of three basic foods are given below:
A, B and C. The production department pro- Nutritional Units of nutritional elements per 100 gm
duces components sufficient to make 50 units elements Basic foods Hi-Pro
of A, 25 units of B and 30 units of C per day. F1 F2 F3 specifications
In the assembly department, only 100 man Calories 350 250 200 300
hours are available to assemble the products Proteins 250 300 150 200
daily. The assembly times for A, B and C are Vitamin A 100 150 75 100
0.8 hr, 1.7 hrs and 2.5 hrs respectively and Vitamin C 75 125 150 100
the profits on them are Rs 12, Rs 20 and Cost per
Rs 45 respectively. The company has an 100 g (Rs) 1.50 2.00 1.20
order commitment of 20 units of A and a
total of 15 units of B and C. How many of Determine the quantities of F1, F2, F3 to be
each product are to be produced so that the used to meet the requirements with minimum
profit is maximum? cost.
6. A company sells two products A and B 9. A sports goods company manufactures two
making a profit of Rs 40 and Rs 30 per unit types of soccer balls X and Y. There are two
respectively. The total capacity of the types of employees—semi-skilled and
skilled. The time required for each type of
production process is 30000 man-hours. It
ball and the time available per week are given
takes three hours to produce one unit of A
below:
and one hour to produce one unit of B. It has
been found that a maximum of 8000 units of Type of employee Time required Time available
A and 12000 units of B can be sold in the (hrs) (hrs per week)
market. Find the number of units of A and Ball X Ball Y
that of B to be produced so as to get maximum Semi-skilled 2 3 80
profit. Skilled 4 6 150

The wages for semi-skilled and skilled
employees are Rs 6 and Rs 10 per hour
respectively. At least 15 balls of type X and
at least 10 balls of type Y must be
manufactured in order to meet with the
weekly demand. Determine the number of
balls of each type to be manufactured so that
the cost of production is minimum.
10. A media specialist has to decide in the
allocation of advertising in three media 1, 2
and 3. For advertising once, the costs are Rs
1000, Rs 750 and Rs 500 respectively. The
total annual allotment of funds is Rs 20000.
Media 1 is a monthly magazine and it is
desired to advertise not more than once in
one issue. At least six messages should
appear in media 2 while the number of
advertisements in media 3 must be between
4 and 8. The expected effective audiences
for one advertisement in each media are
80,000, 60,000 and 45,000 respectively.
Determine the optimal allocation that would
maximize the total effective audience.
11. A company produces refrigerators in unit I
and heaters in unit II. The two products are
produced and sold on a weekly basis. The
weekly production should not exceed 25 in
unit I and 36 in unit II. Only 60 workers can
be employed at the maximum. A refrigerator
requires 2 man-weeks of labour and a heater
requires 1 man-week of labour. The profit is
Rs 1200 per refrigerator and Rs 800 per
heater. Determine the number of refrigerators
and heaters to be produced per week so that
the total profit is maximum.
12. Old hens can be bought for Rs 25 each but
young ones cost Rs 50 each. The old hens
lay 3 eggs per week and the young ones 5
eggs per week. The cost of an egg is Rs 2.
The cost of feeding a hen is Rs 5 per week.
If there are only Rs 600 for purchasing the
hens and if it is not possible to house more
than 20 hens at a time, find how many of
each kind should be purchased in order to
get maximum profit per week.
Linear Programming 9
13. A mining company is taking a certain kind
of ore from two mines A and B. The ore is
divided into three groups g1, g2 and g3 based
on the quality. Every week the company has
to supply 240 tons of g1, 160 tons of g2 and
440 tons of g3. The cost per day of running
mine A is Rs 5000 and of running mine B is
Rs 3000. Each day A will produce 60 tons of
g1, 20 tons of g2 and 40 tons of g3. The
corresponding figures for B are 20, 20 and
80. How many days per week should A and
B work in order to minimize the total cost of
production.
14. An automobile manufacturing company has
three plants A, B and C for production. Plant
A produces 40 scooters and 35 motorcycles
per week. Plant B produces 65 scooters only
and plant C produces 53 motorcycles only
per week. The cost of operating plants A, B
and C are respectively Rs 20,000, Rs 18,000
and Rs 1,60,000 per week. The company has
to fulfill a production scheme of 1500
scooters and 1100 motorcycles in 20 weeks.
Determine the production plan of minimum
cost.
15. A soap manufacturing company has two
factories A and B each producing toilet soap,
washing soap and washing powder. The
number of units produced per day are as
follows:
Item Plant A Plant B
Toilet soap 1500 1200
Washing soap 3000 2000
Washing powder pkts 2000 4000
There is a monthly demand of 25,000 toilet
soaps, 40,000 washing soap and 45,000
packets of washing powder. The operating
costs are Rs 8000 and Rs 6000 per day for
plants A and B respectively. Determine how
many days each plant should work in one
month so as to minimize the production cost.
16. A company produces both exterior and
interior house paints. Two raw materials A
and B are used for manufacturing the paints.
 10 Operations Research

The daily requirements of A and B in tons Rs 650,000 for the year 2008. There are 5
for the two types of paints are given below investment projects under consideration. The
Requirement per Availability estimated returns and expected expenditure
ton of paint (tons) of each project in two years are as follows:
Exterior Interior Project Estimated returns Cash expenditure
(in thousands) (in thousands)
A 1 2 6
Year 2007 Year 2008
B 2 1 8
A 240 120 320
The factory has capacity to produce a B 390 550 594
maximum of 3 tons of exterior and two tons C 80 118 202
of interior paints per day. The price per ton of D 150 250 340
exterior and interior paints are Rs 15,000 and E 182 324 474
Rs 10,000 respectively. Determine how much
Assuming that the returns are directly
quantity of these paints should be produced
proportional to the investment, determine the
per day in order to maximize the sales.
proportions of investment in each project.
17. A company produces washing machines and
20. An advertisement company wishes to plan
TV sets. The weekly production cannot its advertising strategy in three different
exceed 30 washing machines and 25 TV sets. media, TV, Radio and Magazine. Following
Two workers are required for a washing data has been obtained from market survey.
machine and one worker for a TV per week.
TV Radio Maga- Maga-
Totally there are 80 workers in the company.
zine 1 zine 2
The profits are Rs 500 and Rs 300 on washing
machine and TV respectively. How many of Cost of an 30,000 20,000 15,000 10,000
each item should the company produce in advertising unit
(in Rs)
order to maximize the profit. No. of customers 2,00,000 6,00,000 1,50,000 1,00,000
18. A company has two plants each of which reached
produces two products A and B. Each plant per unit
can work up to 16 hours per day. In plant I it No. of female 1,50,000 4,00,000 70,000 50,000
takes 3 hours to prepare and pack 1000 units customers reached
per unit
of A and 1 hour to prepare and pack one unit
of B. In plant II it takes 2 hours to prepare The company wants to spend not more than
and pack 1000 units of A and 1.5 hour to Rs 4,50,000 on advertising. Further require-
prepare and pack one unit of B. In plant I it ments are:
costs Rs 15,000 to prepare and pack 1000 (i) At least 10,00,000 exposures take place
units of A and Rs 28,000 to prepare and pack among female customers
one unit of B whereas these costs are (ii) Advertising in magazines is limited to
Rs 18,000 and Rs 26,000 respectively in Rs 1,50,000
plant II. The company must produce at least (iii) At least 3 advertising units be bought on
10000 units of A and 8 units of B. Determine magazine 1 and 2 units on magazine 2.
the production schedule so as to minimize The number of units in TV and Radio should
the total production cost. be each between 5 and 10. Determine the
19. An engineering company plans to diversify number of units to be bought in TV, Radio,
its operations during the year 2007–2008. An Magazine 1 and Magazine 2, such that
amount of Rs 5,15,000 has been allotted for the maximum number of customers are
this purpose during 2007 and an amount of reached.
 Linear Programming 11

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. Maximize Z = 5x1 + 6x2 + 3x3 subject to the 8. Minimize C = (1.50)x1 + (2.00)x2 + (1.20)x3
constraints subject to the constraints
4x1 + 8x2 + 2x3 £ 2000 350x1 + 250x2 + 200x3 ≥ 300
9x1 + 7x2 + 4x3 £ 2500 250x1 + 300x2 + 150x3 ≥ 200
x1, x2, x3 ≥ 0 100x1 + 150x2 + 75x3 ≥ 100
2. Maximize P = (0.25)x1 + (1.00)x2 + (0.95)x3 75x1 + 125x2 + 150x3 ≥ 100
subject to the constraints x1, x2, x3 ≥ 0
8x1 + 5x2 + 8x3 £ 200 9. Minimize C = 52x1 + 78x2 subject to the
4x1 + 5x2 + 5x3 £ 100 constraints
3x1 + 4x2 + 3x3 £ 120 2x1 + 3x2 £ 80
x1, x2, x3 ≥ 0 4x1 + 6x2 £ 150
3. Maximize P = 45x1 + 80x2 subject to the x1 ≥ 15
constraints x2 ≥ 10
5x1 + 20x2 £ 400 x1, x2 ≥ 0
10x1 + 15x2 £ 450 10. Maximize Z = 80000x1 + 60000x2 + 45000x3
x1, x2 ≥ 0 subject to the constraints
4. Maximize Z = 12x1 + 8x2 + 14x3 + 10x4 sub-
100x1 + 750x2 + 500x3 £ 20000
ject to the constraints
x1 £ 12
5x1 + 4x2 + 2x3 + x4 £ 100
x2 ≥ 6
2x1 + 3x2 + 8x3 + x4 £ 75
4 £ x3 £ 8
x1, x2, x3, x4 ≥ 0
5. Maximize P = 12x1 + 20x2 + 45x3 subject to x1, x2, x3 ≥ 0
the constraints 11. Maximize Z = 1200x1 + 800x2 subject to the
0.8x1 + 1.7x2 + 2.5x3 £ 100 constraints
x1 £ 50 2x1 + x2 £ 60
x2 £ 25 x1 £ 25
x3 £ 30 x2 £ 36
x4 ≥ 20 x1, x2 ≥ 0
x2 + x3 ≥ 15 12. Maximize P = x 1 + 5x 2 subject to the
x1, x2, x3 ≥ 0 constraints
6. Maximize Z = 40x1 + 30x2 subject to the x1 + x2 £ 20
constraints 25x1 + 50x2 £ 600
3x1 + x2 £ 3000 x1, x2 ≥ 0
x1 £ 8000 13. Minimize C = 5000x1 + 3000x2 subject to
x2 £ 12,000 the constraints
x1, x2 ≥ 0 3x1 + x2 ≥ 12
7. Maximize Z = 300x1 + 400x2 subject to the x1 + x2 ≥ 8
constraints x1 + 2x2 ≥ 11
5x1 + 4x2 £ 200 x1, x2 ≥ 0
3x1 + 5x2 £ 150 14. Minimize C = 20000x1 + 18000x2 + 16000x3
5x1 + 4x2 ≥ 100 subject to the constraints
8x1 + 4x2 ≥ 80 40x1 + 65x2 ≥ 1500
x1, x2 ≥ 0 35x1 + 53x3 ≥ 1100
 12 Operations Research

x1 £ 20 Constraints on minimum daily production

x2 £ 20 are
x3 £ 20 x1 + x3 ≥ 10
x1, x2, x3 ≥ 0 x2 + x4 ≥ 8
15. Minimize C = 8000x1 + 6000x2 subject to Total cost is Z = 15000 x1 + 28000 x2 +
the constraints 18000x3 + 26000 x4
15x1 + 12x2 ≥ 250 Thus the required formulation is
3x1 + 2x2 ≥ 40 Minimize Z = 15000 x1 + 28000 x2 + 18000x3
2x1 + 4x2 ≥ 45 + 26000x4 subject to the constraints
x1, x2 ≥ 0 3x1 + x2 £ 16
16. Let x1 tons of exterior paint and x2 tons of 2x3 + 1.5x4 £ 16
interior paint be produced. The total sales is x1 + x3 ≥ 10
Z = 15000x 1 + 10000x 2 which is to be x2 + x4 ≥ 8
maximized. Considering the availability of x1, x2, x3, x4 ≥ 0
raw materials we get 19. Let x 1 , x 2 , x 3 , x 4 and x 5 represent the
x1 + 2x2 £ 6 for A and proportions of investment in the projects A,
2x1 + x2 £ 8 for B B, C, D and E respectively.
Also, on the capacity of factory we have,
Net returns to be maximized is given by
x1 £ 3 and x2 £ 2
240x1 + 390x2 + 80x3 + 150x4 + 182x5
Thus the LPP is
The budget constraints are
Maximize Z = 15000x1 + 10000x2 subject to
120x1 + 550x2 + 118x3 + 250x4 + 324x5
the constraints
£ 515 for 2007
x1 + 2x2 £ 6
2x1 + x2 £ 8 and
x1 £ 3 320x1 + 594x2 + 202x3 + 340x4 + 474x5
x2 £ 2 £ 650 for 2008
x1, x2 ≥ 0 Also, each proportion has to be less than 1.
17. Let the company produce x 1 washing Therefore the formulated LPP is
machines and x2 TV sets Maximize Z = 240x1 + 390x2 + 80x3 + 150x4
x1 £ 30 and x2 £ 25. + 182x5 subject to the constraints
The constraint on the workers is 120x1 + 550x2 + 118x3 + 250x4 + 324x5 £ 515
2x1 + x2 £ 80 320x1 + 594x2 + 202x3 + 340x4 + 474x5 £ 650
The total profit is Z = 500x1 + 300x2 x1 £ 1, x2 £ 1, x3 £ 1, x4 £ 1, x5 £ 1
The required LPP. is x1 ≥ 0, x2 ≥ 0, x3 ≥ 0, x4 ≥ 0, x5 ≥ 0
Maximize Z = 500 x1 + 300 x2 subject to the 20. Let x1, x2, x3 and x4 represent the number of
constraints advertising units in TV , Radio, Magazine 1
2x1 + x2 £ 80 and Magazine 2 respectively. The objective
x1 £ 30 is to maximize the total number of potential
x2 £ 25 customers given by
x1 ≥ 0, x2 ≥ 0 105 (2x1 + 6x2 + 1.5x3 + x4)
18. Let x1 units of A (unit of 1000) and x2 units The constraint of the advertising budget is
of B be produced in plant I. Let x3 units (unit 30000x1 + 20000x2 + 15000x3 +
of 1000) of A and x4 units of B be produced 10000x4 £ 450000
in plant 2. The time constraints are or 30x1 + 20x2 + 15x3 + 10x4 £ 450
3x1 + x2 £ 16 The constraint on the number of female
2x3 + 1.5x4 £ 16 customers reached is given by
 Linear Programming 13

150000x1 + 400000x2 + 70000x3 + 50000x4

≥ 1000000
or 15x1 + 40x2 + 7x3 + 5x4 ≥ 100
For advertising in the magazines, we have
15000x1 + 10000x2 £ 150000
or 3x1 + 2x2 £ 30
Also, x3 ≥ 3, x4 ≥ 2 and
5 £ x1 £ 10, 5 £ x2 £ 10
Thus the required LPP is
Maximize Z = 105 (2x1 + 6x2 + 1.5x3 + x4)
subject to the constraints
30x1 + 20x2 + 15x3 + 10x4 £ 450
15x1 + 40x2 + 7x3 + 5x4 ≥ 100
3x1 + 2x2 £ 30
x3 ≥ 3, x4 ≥ 2, x1 ≥ 5, x1 £ 10
x2 ≥ 5, x2 £ 10
x1, x2, x3, x4 ≥ 0.
2.3 GRAPHICAL METHOD OF
SOLUTION
Linear programming problems involving two vari-
ables can be easily solved by a graphical procedure.
We know that any first degree equation in x and
y represents a straight line in the xy plane. If we
consider the equation 2x + 3y = 6, any point
(x1, y1) satisfying this equation lies on this line.
On the other hand, if 2x1 + 3y1 < 6 then the point
(x1, y1) lies in the region which contains the origin.
Also, 2x1 + 3y1 > 6 implies that the point
(x1, y1) lies on that side of the line not containing
Y where x and y are the points (x1, x2) and (y1, y2)
2x
+3
y =6
(0, 2)
2x + 3 y > 6
2x + 3y < 6
R
P Q X
X Y
Fig. 2.2
the origin. Thus all the points satisfying the
inequality 2x + 3y < 6 lie on that side of the line
2x + 3y = 6, which contains the origin and 2x +
3y > 6 represents the set of all points lying on the
other side which does not contain the origin.
Based on this fact we can find the region, re-
presented by a constraint inequality. For each cons-
traint we obtain a region. The intersection of all these
regions represents the set of points which satisfy all
the constraints and it is called feasible region. This
region contains the set of all feasible solutions of
the LPP and it is in the form of a polygon.
2.3.1 Convex Set
A subset X Õ R2 is said to be convex if for any two
points P and Q in X the line segment PQ is
completely contained in X.
Any point R on the line joining P and Q is given
by
lx + (1 – l)y; 0 £ l £ 1(convex combination)
respectively.
R has co-ordinates
[lx1 + (1 – l)y1 lx2 + (1 – l)y2]
X is convex if this point
lx + (1 – l)y Œ X 0 £ l £ 1
where x and y are any two points of X.

(3, 0)
O X

Fig. 2.1 Regions represented by Fig. 2.3 Convex Sets

constraints
 14 Operations Research
Fig. 2.4 Non-convex Sets
If a point in X cannot be expressed as a convex
combination of any two points of X then it is called
an extreme point of X.
Theorem 1: The intersection of two convex sets
is also a convex set.
Proof: Let X1 and X2 be two convex sets and let
X = X1 « X2.
Take x1, x2 e X. Therefore x1, x2 e X1 as well as
x1, x2 e X2.
Now, x1, x2 e X1 and X1 is a convex set. Hence
lx1 + (1 – l)x2 e X1
where 0 £ l £ 1.
Again x1, x2 e X2 and X2 is convex. Hence lx1 +
(1 – l)x2 e X2 where 0 £ l £ 1.
Thus lx1 + (1 – l)x2 e X1 « X2.
i.e. lx1 + (1 – l)x2 e X (0 £ l £ 1)
Therefore we find that
x1, x2 e X Æ lx1 + (1 – l)x2 e X (0 £ l £ 1)
which implies that X is convex.
Theorem 2: Let S be a convex set in the plane
R2 bounded by lines. Then a linear function
Z = c1x1 + c2x2
where x1, x2 e S and c1, c2 are scalars, attains its
extreme value (maximum or minimum) at the
vertices of S.
Proof: Take any two points A(x1, x2) and B(x1¢,
x2¢) of S. Then any point on the line segment AB is
given by a convex combination of the coordinates
of A and B. Let P (x, h) be any point on the line
segment AB.
\ x = lx1 + (1 – l)x1¢
h = lx2 + (1 – l)x2¢
The value of Z at A is given by (0 £ l £ l)
z1 = c1x1 + c2x2
Similarly, the value of Z at B is given by
z2 = c1x1¢ + c2x2¢
Now the value of Z at P (x, h) is given by
z = c1x + c2h
= c1(lx1 + (1 – l)x1¢) + c2 (lx2 + (1 – l)x2¢)
= l(c1x1 + c2x2) + (1 – l) ( c1x1¢ + c2x2¢)
= lz1 + (1 – l) z2
= l(z1 – z2) + z2 (0 £ l £ 1)
If z1 > z2 then z2 £ z £ z1 and if z2 > z1 then
z1 £ z £ z2 . In any case z lies between the
values of z at A and B where A and B are any two
points of S.
In case P is an interior point of S, it lies on a
line segment AB with A and B on two edges of S.
Hence the extreme value of Z cannot occur at P
B
P
S
A
Fig. 2.5
In case P lies on an edge of S, it lies on a
segment AB with A and B as two vertices and
therefore z lies between the values at the vertices.
Hence the extreme value of z cannot be obtained
at a point on an edge of S.
Thus the extreme value of Z is attained only at
one vertex of S.
2.3.2 Graphical Method
Let us consider an LPP of the form
Maximize Z = c1x1 + c2x2 subject to the con-
straints
aix1 + bix2 £ ki (i = 1, 2, 3, …, n)
x1, x2 > 0
First we plot the lines aix1 + bix2 £ ki (i = 1, 2,
3, …, n)
in the X1X2 plane and mark the regions represented
by the given inequalities. We have to consider only
the I quadrant since x1, x2 ≥ 0.
 Linear Programming 15

X2 Solution We plot the lines x1 + x2 = 12, x1 +

5x2 = 45 and 3x1 + x2 = 30 and mark the regions
represented by (1), (2) and (3) using different
shadings. The region common to all (intersection
of the shaded areas) represents the set of all points
D C forming the feasible solutions to the given LPP.
X2
B (3)
(2)
45
O A X1
(1) 40
35
Fig. 2.6 Model with 3 constraints 30
Each constraint represents a region and the 25
region common to all these (intersection of the
sets) gives the feasible region of the given LPP. 20
Find out the coordinates of the vertices of the 15
feasible region by solving the equations of the C
10
lines. Substitute the coordinates of the vertices, D
B
one by one, in the objective function. Choose the 5 (2)
vertex which gives the maximum value of Z. This
O A
gives the solution of the LPP. 5 10 15 20 25 30 35 40 45 X1
Note: (3) (1)
A
1. If the constraint is of ≥ type we have to shade
the region not containing the origin. Fig. 2.7
2. We have to find the value of Z only at the
vertices since the maximum or minimum of The region corresponding to (1) is shaded with
Z is attained only at a vertex (Theorem 2.2). slant lines, that corresponding to (2) is shaded
3. If a feasible region does not exist then the with vertical lines, and that corresponding to (3)
given LPP has no solution and if the feasible is shaded with horizontal lines. We find that the
region is not a closed polygon then the feasible region is the polygon OABCD. O is the
problem has unbounded solution. origin (0,0), A is the point (10,0) and D is the point
4. If two vertices give the same maximum (0,9). B and C are to be determined. B is the point
(or minimum) value then every point on the of intersection of (1) and (3). Solving the equations
edge joining the vertices gives the same value x1 + x2 = 12 and 3x1 + x2 = 30, we get x1 = 9 and
and the given problem has infinite number x2 = 3. Thus B is the point (9,3). Now we find C
of optimal solutions. by solving (1) and (2).
Solving the equations
Example 2.8 (Solve by graphical method) x1 + x2 = 12 and x1 + 5x2 = 45, we get x 1 =
Maximize Z = 6x 1 + 9x 2 subject to the 15/4 and x2 = 33/4.
constraints Thus C is the point (15/4, 33/4).
x1 + x2 £ 12 Now, the objective function is Z = 6x1 + 9x2
x1 + 5x2 £ 45 The value of Z at O is Z0 = 0.
3x1 + x2 £ 30 The value of Z at A is
x1, x2 ≥ 0 ZA = 6(10) + 9(0) = 60
 16 Operations Research

Similarly, Note: If one of the constraint equation is

ZB = 6(9) + 9(3) = 81 parallel to the objective function then we get
ZC = 6(15/4) + 9(33/4) = 387/4 infinite number of solutions.
ZD = 6(0) + 9(9) = 81. Example 2.10 Solve graphically
The maximum value of Z occurs at C(15/4, Maximize Z = 3x 1 + 4x 2 subject to the
33/4) and the value is Z* = 387/4. constraints
Hence the optimal solution of the given LPP is – 3x1 + 2x2 £ 6
x1 = 15/4, x2 = 33/4, Z* = 387/4. 3x1 + x2 ≥ 6
Example 2.9 Solve graphically x1 + x2 £ 8
Maximize Z = 40x1 + 100x2 subject to the x1, x2 ≥ 0
constraints Solution The regions represented by – 3x1 + 2x2
2x1 + x2 £ 500 £ 6 (1) is shaded using vertical lines. Since x1 ≥ 0
2x1 + 5x2 £ 1000 x1, x2 ≥ 0 the negative side of the X1 axis is not considered.
Solution The lines and regions represented by Horizontal lines represent the region
constraints (1) and (2) are shown in the figure. corresponding to 3x1 + x2 ≥ 6 (2) which is ≥ type.
The feasible region is given by the polygon OABC. The region given by x1 + x2 £ 8 (3) is represented
O is the point (0,0), A is (250,0) C is (0,200). B by slant lines. The feasible region is ABCD.
can be found by solving the equations 2x1 + x2 =
X2
500 and 2x1 + 5x2 = 1000. We obtain B as the point
(375/2,125). 8
7
X2
6 C
700 5
600 4 D
500 3
400 2
300 (1) (1)
1
B
200
C B –2 –1 O A X1
1 2 3 4 5 6 7 8
100 (2) (3)
O A
100 200 300 400 500 600 700 800 X1 Fig. 2.9

A is (2,0) B is (8,0) C is (2,6) which is obtained

Fig. 2.8
by solving (1) and (3) with equality sign. D is
(2/3,4) which is obtained by solving
Now Z = 40x1 + 100x2 – 3x1 + 2x2 = 6 and 3x1 + x2 = 6.
Z0 = 0, ZA = 40(250) + 100(0) = 10,000 Now Z = 3x1 + 4x2
ZB = 40(375/2) + 100(125) = 20,000 ZA = 3(2) + 4(0) = 6
ZC = 40(0) + 100(200) = 20,000 ZB = 3(8) + 4(0) = 24
We find that both B and C give the maximum ZC = 3(2) + 4(6) = 30
value 20,000. Therefore there is no unique optimal ZD = 3(2/3) + 4(4) = 18.
solution and any point on the line BC can be taken The maximum value of Z occurs at C (2, 6).
as an optimal solution with Z* = 20,000. Hence the solution is x1 = 2, x2 = 6, Z* = 30.
 Linear Programming 17

Example 2.11 Solve graphically Minimum value is 5.

Minimize Z = 5x1 + 2x2 subject to the constraints Hence the solution is x1 = 1, x2 = 0, Z* = 5.
3x1 + x2 ≥ 3
Example 2.12 Solve graphically
3x1 – 2x2 £ 6
Minimize Z = 5x1 + 4x2 subject to the constraints
x1 + x2 £ 4
x1, x2 ≥ 0. x1 – 2x2 £ 1
Solution The following graph represents the x1 + 2x2 ≥ 3
regions corresponding to the constraint inequalities x1, x2 ≥ 0.
and the feasible region of the given LPP. X2
The feasible region is given by the polygon
ABCDE. Solving the equations 3x1 – 2x2 = 6, x1 + (0,4)
x2 = 4 we get x1 = 14/5, x2 = 6/5, C is (14/5, 6/5).
(0,3)
X2
D C
(0,4)
(0,2)
(1)
(0,3)
(2) ( )
0 , 3–
2 A
E (0,1)

(0,2) B ( ,–)
2 1
2

(0,1)
C ( –,– )
14 6
5 5 O (1,0) (2,0) (3,0) (4,0) X1
( )
0 , 1–
2
(2)

O A B X1
(1,0) (2,0)
(1)
Fig. 2.10
Thus we have the vertices A (1, 0), B (2, 0),
C (14/5, 6/5), D (0, 4) and E (0, 3)
Now Z = 5x1 + 2x2
ZA = 5(1) + 2(0) = 5;
ZB = 5(2) + 2(0) = 10;
ZC = 5(14/5) + 2(6/5) = 82/5;
ZD = 5(0) + 2(4) = 8;
ZE = 5(0) + 2(3) = 6.
(3,0) (4,0) Fig. 2.11
(3)
Solution Draw the lines x1 – 2x2 = 1 and x1 +
2x2 = 3 and mark the regions x1 – 2x2 £ 1 and x1 +
2x2 ≥ 3. We find that the common region (feasible
region) is not closed above. The feasible region is
unbounded. We have only two vertices A and B.
A is the point (0, 3/2) and by solving the equations
we get B as the point (2, ½).
Now Z = 5x1 + 4x2,
ZA = 5(0) + 4(3/2) = 6,
ZB = 5(2) + 4(1/2) = 12.
The minimum value is 6. Therefore the solution
of the given LPP is
x1 = 0 x2 = 3/2, Z* = 6.
Note: If the question is given as “Maximize
Z = 5x 1 + 4x 2 this problem has unbounded
solutions.
Example 2.13 Solve graphically
Maximize Z = 2x 1 + 3x 2 subject to the
constraints
 18 Operations Research

2x1 + x2 £ 2
3x1 + 4x2 ≥ 12 X2
x1, x2 ≥ 0.
40
Solution The following graph gives the regions
represented by the constraints.
X2 30
(0,26)
(3)
(0,4)
20
(6,11)
(0,3)
10 C B (8,10)
(0,5) D
(0,2)
B (13,0)
O 10 20 30 40 X1
(–5,0)
(0,1) (1) (2)

O (1,0) (2,0) (3,0) (4,0) X1 Fig. 2.13

(1) (2) O is (0,0); A is (13, 0); D is (0, 5).

Fig. 2.12
From the graph we find that there is no common
region between the two. That is to say that there is
no point (x 1 , x 2 ) which satisfies both the
constraints. Hence there is no feasible solution.
Thus the given LPP has no solution.
Example 2.14 Solve graphically
Maximize Z = 10x 1 + 15x 2 subject to the
constraints
2x1 + x2 £ 26
2x1 + 4x2 £ 56 or (x1 + 2x2 £ 28)
– x1 + x2 £ 5
x1, x2 ≥ 0.
Solution The graph showing the feasible region
is given below.
The feasible region (solution space) is enclosed
by the polygon OABCD.
B is obtained by solving 2x1 + x2 = 26 and
x1 + 2x2 = 28.
B is (8, 10).
C is given by x1 + 2x2 = 28 and – x1 + x2 = 5.
C is (6, 11)
Now Z = 10x1 + 15x2
ZO = 10(0) + 15(0) = 0
ZA = 10(13) + 15(0) = 130
ZB = 10(8) + 15(10) = 230 Æ Maximum
ZC = 10(6) + 15(11) = 225
ZD = 10(0) + 15(5) = 75.
Hence the solution is x1 = 8 x2 = 10 Z* = 230.
Example 2.15 A manufacturer of furniture
makes chairs and tables. Processing of these
products is done on two machines A and B. A chair
requires 2 hours on machine A and 6 hours on B.
A table requires 5 hours on A and 6 hours on B.
There are 16 hours of time available on machine
A and 30 hours on machine B. At the most 4 chairs
are to be manufactured. The manufacturer gets a
profit of Rs 50 on a chair and Rs 100 on a table.
Determine the number of chairs and tables to be
manufactured so as to maximize the profit.
Solution First we formulate this into an LPP. Let
x1 chairs and x2 tables be produced. The total profit
is P = 50x1 + 100x2 which is the objective function
to be maximized. Considering the time taken by
x1 chairs and x2 tables on machine A we get the
constraint 2x1 + 5x2 < 16.
 Linear Programming 19

The next constraint on the time taken on less than 50 kg. At least 20 kg of X and not more
machine B is given by than 40 kg of Y can be used. The cost of X is Rs 10
6x1 + 6x2 £ 30 or (x1 + x2 £ 5) per kg and that of Y is Rs 25 per kg. Determine
Also, x1 £ 4. how much of X and Y are to be used to minimize
Finally the non-negativity constraints are the total cost.
x1 ≥ 0 and x2 ≥ 0. Solution
Hence we get the LPP
Maximize P = 50x1 + 100x2 subject to the Formulation
constraints. Let us assume that x1 kg of X and x2 kg of Y
2x1 + 5x2 £ 16 are used to form the product. At least 20 kg of X
x1 + x2 £ 5 and not more than 40 kg of Y are to be used.
x1 £ 4 Therefore we have x1 ≥ 20 and x2 £ 40.
x1, x2 ≥ 0. Also the weight of the product must be not less
Let us represent the problem graphically. than 50 kg. This implies that
x1 + x2 ≥ 50.
The total cost of the product is 10x1 + 25x2
which is to be minimum.
Thus the LPP is
Minimize C = 10x 1 + 25x 2 subject to the
constraints
x1 ≥ 20, x2 £ 40, x1 + x2 ≥ 50, x1, x2 ≥ 0

Fig. 2.14

The solution space is represented by the shaded

region OABCD. The point B is obtained by solving
the equations (2) and (3). B is (4, 1). Similarly
solving (1) and (2) we get the point C (3, 2).
Now, P = 50x1 + 100x2
PO = 0 PA = 50(4) + 0 = 200
PB = 50(4) + 100(1) = 300 Fig. 2.15 Graphical representation
PC = 50(3) + 100(2) = 350 Æ Maximum
PD = 0 + 100(3) = 300. A is the point (20, 40), B is (20, 30) and C is
Hence the solution is x1 = 3, x2 = 2, P* = 350. (50, 0).
3 chairs and 2 tables are to be manufactured in C = 10x1 + 25x2
order to get the maximum profit of Rs 350. CA = 10(20) + 25(40) = 1200
CB = 10(20) + 25(30) = 950
Example 2.16 A company combines factors X
and Y to form a product which must weigh not CC = 10(50) + 25(0) = 500 Æ Minimum.
 20 Operations Research

Thus the optimal solution is x1 = 50 x2 = 0 order to minimize the total cost and the minimum
C* = 500. cost is Rs 500.
Therefore 50 kg of X alone has to be used in

EXERCISES
■ ■ ■ ■ ■ ■ ■

Solve the following LP problems graphically: Canned pine- Canned Available

1. Maximize Z = 3x 1 + 4x 2 subject to the apple juice resources
constraints
Labour 2 3 12
2x1 + x2 £ 40
(man hours)
2x1 + 5x2 £ 180 Equipment 3 2 12
x1,x2 ≥ 0 (machine hours)
2. Maximize Z = 3x 1 + 2x 2 subject to the Raw material (unit) 2 2 9
constraints
– 2x1 + x2 £ 1 7. A company manufactures two types of
x1 £ 2 printed circuits A and B. The following table
x1 + x2 £ 3 gives the requirements of transistors, resistors
x1, x2 ≥ 0 and capacitors for each type and other
3. Maximize Z = 3x 1 + 4x 2 subject to the particulars. How many circuits of each type
constraints should the company produce from the
x1 + x2 £ 450 available stock in order to earn maximum
2x1 + x2 £ 600 profit.
x1, x2 ≥ 0 Circuit
4. Maximize Z = 15x1 + 10x2 subject to the A B Stock available
constraints
2x1 + 3x2 £ 180 Transistor 15 10 180
x1 £ 60 Resistor 10 20 200
Capacitor 15 20 210
x2 £ 40
Profit (Rs) 5 8
x1, x2 ≥ 0
5. Maximize Z = 2x 1 + x 2 subject to the 8. Maximize Z = x 1 – 2x 2 subject to the
constraints constraints
x1 + 2x2 £ 10 – x1 + x2 £ 1
x1 + x2 £ 6 3x1 + 2x2 ≥ 12
x1 – x2 £ 2 0 £ x1 £ 5
x1 – 2x2 £ 1 2 £ x2 £ 4
x1, x2 ≥ 0. 9. Maximize Z = 7x 1 + 3x 2 subject to the
6. A pineapple firm produces canned pineapple constraints
and canned juice and the requirement and x1 + 2x2 ≥ 3
availability of the resources are given below: x1 + x2 £ 4
The profit per unit of canned pineapple and 0 £ x1 £ 5/2
one unit of canned juice are Rs 2 and Rs 5 0 £ x2 £ 5/2
respectively. Determine the production 10. Minimize Z = 60x1 + 40x2 subject to the
schedule which maximizes the profit. constraints
 Linear Programming 21

3x1 + x2 ≥ 40 should be purchased in order to minimize the

2x1 + 5x2 ≥ 44 total cost?
3x1 + 3x2 ≥ 40 18. Maximize Z = 6x 1 + 4x 2 subject to the
x1, x2 ≥ 0 constraints
11. Minimize Z = 7x 1 + 8x 2 subject to the x1 – x2 £ 2
constraints 3x1 + 4x2 ≥ 12
3x1 + x2 ≥ 9 3x1 + 2x2 £ 18
x1 + 3x2 ≥ 9 x1, x2 ≥ 0
x1, x2 ≥ 0 19. Minimize Z = 2x 1 – 10x 2 subject to the
12. Find the maximum and minimum values of constraints
Z = 5x1 + 3x2 subject to the constraints x1 – x2 ≥ 0
x1 + x2 £ 6 x1 – 5x2 ≥ – 5 (– x1 + 5x2 £ 5)
2x1 + 3x2 ≥ 6 0 £ x1 £ 2
0 £ x1 £ 3 x2 ≥ 0
x2 ≥ 0 20. Maximize Z = 40x1 + 80x2 subject to the
13. Find the maximum and minimum values of constraints
Z = 20x1 + 10x2 subject to the constraints 2x1 + 3x2 £ 48
x1 + 2x2 £ 40 0 £ x1 £ 15
3x1 + x2 ≥ 30 0 £ x2 £ 10
4x1 + 3x2 ≥ 60 21. Maximize Z = 40x1 + 100 x2 subject to the
x1, x2 ≥ 0 constraints
14. Minimize Z = 80x1 + 120x2 subject to the 12x1 + 6x2 £ 3000
constraints 4x1 + 10x2 £ 2000
x1 + x2 £ 9 2x1 + 3x2 £ 900
2x1 + 5x2 £ 36 x1, x2 ≥ 0
0 £ x1 £ 2 22. Maximize Z = 3x 1 + 2x 2 subject to the
0 £ x2 £ 3 constraints
15. Maximize Z = 3x 1 + 2x 2 subject to the – 2x1 + 3x2 £ 9
constraints 3x1 – 2x2 ≥ –20
x1 – x2 £ 1 x1, x2 ≥ 0
x1 + x2 ≥ 3 23. Maximize Z = 10x1 + 15x2 subject to the
x1, x2 ≥ 0 constraints
16. Maximize Z = 6x 1 + 4x 2 subject to the 2x1 + x2 £ 26
constraints x1 + 2x2 £ 28
x1 + 2x2 £ 2 – x1 + x2 £ 5
2x1 + 5x2 ≥ 10 x1, x2 ≥ 0
x1, x2 ≥ 0 24. Maximize Z = 3x 1 + 2x 2 subject to the
17. A person requires 10,12 and 12 units of constraints
chemicals A, B and C respectively for his 5x1 + x2 ≥ 10
garden. A liquid product contains 5, 2 and 1 x1 + x2 ≥ 6
units of A, B and C respectively per jar. A x1 + 4x2 ≥ 12
dry product contains 1, 2 and 4 units of A, B x1, x2 ≥ 0
and C per packet. If the liquid product is 25. Maximize Z = – x1 + 2x2 subject to the
Rs 3 per jar and the cost of the dry product is constraints
Rs 2 per packet find out how many of each – x1 + 3x2 £ 10
 22 Operations Research

x1 + x2 £ 6 x1 + 3x2 ≥ 400
x1 – x2 £ 2 x1 + 2x2 £ 350
x1, x2 ≥ 0 x1 + x2 ≥ 200
26. Maximize Z = 20x1 + 10 x2 subject to the x1, x2 ≥ 0
constraints 29. Maximize Z = 50x1 + 30x2 subject to the
x1 + 2x2 £ 40 constraints
3x1 + x2 ≥ 30 2x1 + x2 ≥ 18
4x1 + 3x2 ≥ 60 x1 + x2 ≥ 12
x1, x2 ≥ 0 3x1 + 2x2 £ 34
27. Maximize Z = 2x 1 + 3x 2 subject to the x1, x2 ≥ 0
constraints 30. Maximize Z = x1 + 3x2 subject to the con-
x1 + x2 £ 30 straints
12 ≥ x2 ≥ 3
x1 + x2 £ 7
x1 – x2 ≥ 0
x2 £ 5
x1 ≥ 0
3x1 + 10x2 ≥ 30
28. Minimize Z = 200x1 + 400x2 subject to the
constraints x1, x2 ≥ 0

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. x1 = 5/2, x2 = 35, Z* = 147.5 18. There are infinite number of solutions.

2. x1 = 2, x2 = 1, Z* = 8
3. x1 = 0, x2 = 450, Z* = 1800
4. x1 = 60, x2 = 20, Z* = 1100
5. x1 = 4, x2 = 2, Z* = 10
6. x1 = 0, x2 = 4, Z* = 20
7. x1 = 2, x2 = 9, Z* = 82
8. x1 = 5, x2 = 2, Z* = 1
9. x1 = 5/2, x2 = 3/2, Z* = 22
10. x1 = 12, x2 = 4, Z* = 880
11. x1 = 9/4, x2 = 9/4, Z* = 135/4
12. x1 = 0, x2 = 2, Min Z = 6
x1 = 3, x2 = 3, Max Z = 24
13. x1 = 6, x2 = 12, Min Z = 240.
x1 = 40, x2 = 0, Max Z = 800.
14. x1 = 2, x2 = 3, Z* = 520.
15. The solution is unbounded.
16. The problem has no solution.
17. Liquid 1, dry 5, Min cost = 13
Z* = 36
Note that the objective function 6x1 + 4x2 is
parallel to 3x1 + 2x2 = 18.
19. There are infinite number of solutions.
Z* = – 10
(2x1 – 10x2 is parallel to x1 – 5x2 = – 5)
20. x1 = 9, x2 = 10, Z* = 1160
21. There are infinite number of solutions
Z* = 20000.
22. There is no solution.
23. x1 = 8, x2 = 10, Z* = 230
24. x1 = 1, x2 = 5, Z* = 13
25. x1 = 2, x2 = 0, Z* = –2
26. x1 = 6, x2 = 12, Z* = 240
27. x1 = 18, x2 = 12, Z* = 72
28. x1 = 100, x2 = 100, Z* = 60000
29. x1 = 10, x2 = 2, Z* = 560
30. x1 = 2, x2 = 5, Z* = 17
 3
Simplex Method

CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■

3.1 INTRODUCTION Similar examples can be given in the case of

three variables also.
If a set of simultaneous equations is given, in But if the number of equations is less than the
general we can solve them and obtain the values number of variables then the variables in excess
of the variables when the number of equations is are called non-basic variables and in the process
equal to the number of variables. The solution is of solving the equations they are treated as
unique if the matrix of the coefficients is non- constants. The variables taken for solution are
singular. For example consider two equations in called basic variables. For example, consider the
two variables: equations
4x1 + x2 = 9 2x1 + x2 + x3 = 14
x1 + 2x2 = 4 x1 + 2x2 + x3 = 16
The solution is x1 = 2, x2 = 1, we find that Since we have two equations we can solve for
4 1 two variables, keeping the third one as constant.
=7¹0 The basic solution with x1 and x2 as basic variables
1 2
is given by solving
Now if we consider the equations
2x1 + x2 = 14 – x3
x1 + x2 = 3
x1 + 2x2 = 16 – x3
2x1 + 2x2 = 6
we get x1 = 4 – x3/3; x2 = 6 – x3/3
we find that they are not independent. There exists
Taking x1 and x3 as basic variables, we get
an infinite number of solutions in this case:
(x1 = 1, x2 = 2), (x1 = 2, x2 = 1), (x1 = 0, x2 = 3) and 2x1 + x3 = 14 – x2
so on. x1 + x3 = 16 – 2x2
But the equations The corresponding basic solution is
x1 + x2 = 3 x1 = – 2 + x2; x3 = 18 – 3x2
2x1 + 2x2 = 5 Also taking x2 and x3 as basic we obtain the
are inconsistent and they have no solution. solution
 24 Operations Research

x2 = 2 + x1, x3 = 12 – 3x1 We assign zero value to the non-basic variables

Thus if we have n variables and m equations and express the three basic solutions as
(n > m) then we have m basic variables and n – m (i) x1 = 4, x2 = 6, x3 = 0
non-basic variables. In the example given above (ii) x1 = –2, x2 = 0, x3 = 18
there are three variables and two equations. The (iii) x1 = 0, x2 = 2, x3 = 12
two basic variables are selected in 3C2 = 3 ways In a linear programming problem if a set of
and accordingly we get three basic solutions. For values of the variables satisfy all the constraints
m equations in n variables (n > m) we get, nCm then it is called a feasible solution. If a basic
basic solutions. solution is also a feasible solution then it is called
We can represent the equations, given in the a basic feasible solution. A basic feasible solution
example, in a matrix form. which optimizes the objective function is called
The equations are the optimal solution. In an LPP the variables
2x1 + x2 + x3 = 14 should satisfy the non-negativity constraints also
x1 + 2x2 + x3 = 16 [i.e. xi > 0 "i]. In the above example the second
In matrix form we write, solution is basic but not feasible, since x1= –2 (<0).
æ x1 ö
æ 2 1 1ö ç ÷ æ 14 ö 3.2 SIMPLEX METHOD
çè 1 2 1÷ø ç x2 ÷ = çè 16 ÷ø
çè x ÷ø
3 3.2.1 Standard form of a given
The first solution with x 1 and x 2 as basic LPP
variables can be written in the form
Consider the LPP
x1 + 0x2 + x3/3 = 4
Maximize Z = C1x1 + C2x2 + ... + Cnxn subject
0x1 + x2 + x3/3 = 6
to the constraints
In the matrix form it can be written as a11x1 + a12x2 + … + a1nxn £ b1
æ x1 ö a21x1 + a22x2 + … + a2nxn £ b2
æ 1 0 1 / 3ö ç ÷ æ 4 ö
çè 0 1 1 / 3÷ø ç x2 ÷ = çè 6 ÷ø
: : : :
çè x ÷ø : : : :
3 am1x1 + am2x2 + … + amnxn < bm
Here the identity matrix x1 x2 … xn ≥ 0
æ 1 0ö In order to solve this LPP we have to express
çè 0 1 ÷ø the problem in the standard form. The standard
form of LPP has the following characteristic
corresponds to the basic variables (x1, x2) and it is
properties:
called basis matrix.
(i) The objective function should be of
The solution having x1 and x3 as basic variables maximization type. If it is given, as
can be written as Minimize Z = C1x1 + C2x2 + … + Cnxn
æ x1 ö we change it to maximization type by taking
æ 1 -1 0 ö ç ÷ æ -2 ö
çè 0 3 1÷ø ç x2 ÷ = çè 18÷ø
– Z = – C1x1 – C2x2 – … – Cnxn
çè x ÷ø and write
3 Minimize W = – C1x1 – C2x2 – …– Cnxn
and the solution with x2 and x3 as basic variables where W = – Z
is given by Max (W) = – Min (– W) and
æ x1 ö Min (W) = – Max (– W)
æ -1 1 0 ö ç ÷ æ 2 ö We find that Min(Z) = – Max (– Z)
çè 3 0 1÷ø ç x2 ÷ = çè 12 ÷ø = – Max (W)
çè x ÷ø
3 and thus Min(Z) can be derived from Max W.

(ii) All the constants bj (j = 1, 2, 3, 4,…, m) must
be positive. If there is a negative constant
then multiply the constraint inequality
on both sides by –1. For example the
constraint 3x1 + 2x2 + 5x3 £ –2 can be written
as – 3x1 – 2x2 – 5x3 ≥ 2
[when multiplied by –1, the £ sign changes
to ≥ and vice versa].
(iii) All the variables must be non-negative. If one
variable, say xr, is negative then put xr¢ = – xr
and replace xr with – xr¢ in the problem. Now
xr¢ ≥ 0. Again if a variable xr is unrestricted
in sign (it can assume either +ve or –ve value)
then put xr = ur – vr, where both ur and vr are
non-negative. ur > vr Æ xr is +ve and ur < vr
Æ xr is –ve.
(iv) Finally all the constraints must be equalities.
If a constraint inequality is of upper bound
type (£) then we add a new variable sj called
slack variable and convert it into equality.
The constraint
a11x1 + a12x2 + … + a1nxn £ b1can be changed
into
a11x1 + a12x2 + … + a1nxn + s1 = b1
Similarly, in case of lower bound type
constraints (≥) we subtract a new variable sj
called surplus variable and obtain the
equality. Accordingly, the constraint
a11x1 + a12x2 + … + a1nxn ≥ b1can be changed
into
a11x1 + a12x2 + … + a1nxn – s1 = b1
It should be noted that all the variables should
satisfy the non-negativity constraint.
Example 3.1 Rewrite the following LPP in the
standard form:
Maximize Z = 2x1 + 3x2 + 4x3 subject to the
constraints
x1 + 3x2 + 5x3 £ 10
3x1 – x2 – 2x3 £ –4
x1 £ 0 x2, x3 ≥ 0
Solution Multiplying the second constraint by
–1, we get
–3x1 +x2 + 2x3 > 4
Given x1 £ 0. Therefore put x1 = –x1¢
Thus the problem becomes
Simplex Method 25
Maximize Z = –2x1¢ + 3x2 + 4x3 subject to the
constraints
– x1¢ + 3x2 + 5x3 £ 10
3x1¢ + x2 + 2x3 ≥ 4
x1, x2, x3 ≥ 0
Finally, we change the constraints to equalities.
By introducing a slack variable s1 and a surplus
variable s2 we get the standard form as
Maximize Z = –2x1¢ + 3x2 + 4x3 subject to the
constraints
– x1¢ + 3x2 + 5x3 + s1 = 10
3x1¢ + x2 + 2x3 – s2 = 4
x1, x2, x3, s1, s2 ≥ 0
Example 3.2 Change the following LPP into
standard form:
Minimize Z = 2x1 + 2x2 + 4x3 + x4 subject to the
constraints
3x1 + 2x2 + x3 + 9x4 £ 16
7x1 + 2x2 + 4x3 + x4 £ 24
x1. x2, x3 ≥ 0 x4 unrestricted.
Solution The objective function has to be written as
Maximize W = – Z = – 2x1 – 2x2 – 4x3 – x4
Since x4 is unrestricted, write
x4 = u4 – v4
The problem becomes
Maximize W = –2x1 – 2x2 – 4x3 – (u4 – v4)
subject to the constraints
3x1 + 2x2 + x3 + 9 (u4 – v4) £ 16
7x1 + 2x2 + 4x3 + u4 – v4 £ 24
x1, x2, x3, u4, v4 ≥ 0
Introducing slack variables s1 and s2 we get the
standard form as
Maximize W = –2x1 – 2x2 – 4x3 – u4 + v4 subject
to the constraints
3x1 + 2x2 + x3 + 9u4 – 9v4 + s1 = 16
7x1 + 2x2 + 4x3 + u4 – v4 + s2 = 24
x1, x2, x3, u4, v4, s1, s2 ≥ 0
Note: If a constraint is given in the form of
equality then we can express it as two inequalities,
one (£ ) type and the other (≥) type.
Example 3.3 Rewrite in the standard form:
Maximize Z = 5x1 + 2x2 + x3 subject to the
constraints
2x1 + x2 – x3 = 6
x1 + x2 + x3 £ 8
x1, x2, x3 ≥ 0
 26 Operations Research
Solution The first constraint is given as
2x1 + x2 – x3 = 6
This can be written as two inequalities
2x1 + x2 – x3 £ 6
2x1 + x2 – x3 ≥ 6
Introducing slack and surplus variables. We get,
2x1 + x2 – x3 + s1 = 6
2x1 + x2 – x3 – s2 = 6
Adding a slack variable in the second constraint
we get
x1 + x2 + x3 + s3 = 8
Thus the standard form of the given LPP is
Maximize Z = 5x1 + 2x2 + x3 subject to the
constraints
2x1 – x2 – x3 + s1 = 6
2x1 + x2 – x3 – s2 = 6
x1 + x2 + x3 + s3 = 8
x1, x2, x3, s1, s2, s3 ≥ 0
Note: An LPP is said to be in canonical form if
(i) the objective function is of maximization type
(ii) all the constraints are (£) type
(iii) all the variables are non-negative.
3.2.2 Simplex Algorithm
Consider the LPP (in the standard form)
Maximize Z = c1x1 + c2x2 + c3x3 subject to the
constraints
a11x1 + a12x2 + a13x3 + s1 = b1
a21x1 + a22x2 + a23x3 + s2 = b2
x1, x2, x3, s1, s2 ≥ 0.
Since we have two equations we can have two
basic variables. Taking s1 and s2 as the basic
variables we have a solution
s1 = b1 – a11x1 – a12x2 – a13x3
s2 = b2 – a21x1 – a22x2 – a23x3.
Assigning zero values to the non-basic variables
x1, x2 and x3 we obtain a basic feasible solution.
s1 = b1 and s2 = b2
The objective function can be written as
Z = c1x1 + c2x2 + c3x3 + c4s1 + c5s2
Therefore for s1 = b1 and s2 = b2 we get the
value of Z as Z0 = c4b1 + c5b2. We want to find out
other basic feasible solutions which would increase
the value of Z. We shall take a non-basic variable
into the basis and examine whether the new basic
feasible solution gives an increased value for Z.
Substituting for s1 and s2 in Z, we get
Z = c1x1 + c2x2 + c3x3 + c4(b1 – a11x1 – a12x2 –
a13x3) + c5(b2 – a21x1 – a22x2 – a23x3)
= c4b1 + c5b2 – {c4a11 + c5a21 – c1}x1 – {c4a12
+ c5a22 – c2}x2 – {c4a13 + c5a23 – c3}x3.
s1 and s2 are the current basic variables and c4
and c5 are their cost co-efficients in the objective
function. Taking
æ s1 ö
X B = ç ÷ and CB = (c4 , c5 )
è s2 ø
We have CBXB = c4s1 + c5s2
= c4b1 + c5b2 = Zo
æ a11 ö æ a12 ö æ a13 ö
A1 = ç ÷ A2 = ç ÷ A3 = ç
è a21 ø è a22 ø è a23 ÷ø
æ a11 a12 a13 ö
of the matrix A = ç
è a21 a22 a23 ÷ø
\ c4a11 + c5a21 = CBA1
c4a12 + c5a22 = CBA2
c4a13 + c5a23 = CBA3
We denote these values by Z 1, Z 2 and Z 3
respectively.
Thus
Z = Zo – (Z1 – c1)x1 – (Z2 – c2)x2 – (Z3 –c3)x3
3
= Zo - å (Z j - cj ) x j
j =1
For some j, if Zj – cj is negative then the value
of Z will increase if the corresponding non-basic
variable xj becomes a basic variable. The most
negative value of Zj – cj will indicate the variable
which will cause maximum increase in Z, when
included in the basis.
Suppose Z1 – c1 is the most negative then x1
should be taken into the basis. The present value
of x1 (non-basic) is zero. Let us assign a value q to
x1. Then we get the solution
x1 = q, x2 = 0, x3 = 0, s1 = b1 – a11q and
s2 = b2 – a21q
Since x1 has been taken into the basis and the
basis can consist of only two variables, one of the
current basic variables should leave the basis and
become non-basic. We have to determine which
of the variables s1 and s2 should become non-basic
and assume zero value
 Simplex Method 27

s1 = 0 Æ b1 – a11q = 0 In the above example we assumed that both a11

Æ q = b1/a11
s2 = 0 Æ b2 – a21q = 0
Æ q = b2 / a21
If a11 is negative then q becomes negative and
same is the case if a21 is negative. Therefore let us
assume that both a11 and a21 are positive.
Let b1/a11 < b2/a21. We take q = b1/a11 and s1
becomes zero. On the other hand, if we take q =
b2/a21 then
s1 = b1 – a11q
= b1 – a11 b2/a21
= (a21b1 – a11b2) / a21 < 0
Æ ¨ [since b1/a11 < b2/a21]
Hence we take the minimum ratio as q. Now s1
becomes non-basic with zero value
x1 = q, x2 = 0, x3 = 0, s1 = 0, s2 = b2 – a21q
is the new solution which gives
Z = Z0 – (Z1 – c1)q > Z0 since Z1 – c1 < 0
Note: If Zj – cj ≥ 0 for all j then it is not possible
to increase Z and the current solution itself is
optimal.
While taking the ratio we consider only the aij
which is positive. If all the aij are negative, Z can
be increased to any value and in this case the
solution is unbounded.
and a21 are positive and obtained the value of q as
the minimum of b1/a11 and b2/a21. If a11 is negative,
and a21 is positive we have to take q = b2/a21 as it
is the only possibility.
3.2.3 Simplex Tableau
The simplex algorithm can be carried out
conveniently by representing the given objective
function, constraint equalities, basic variables,
basic solution, etc. in a tabular form known as
simplex tableau or simplex table.
Consider an LPP in the standard form:
Maximize Z = c1x1 + c2x2 + … + cnxn subject to
a11x1 + a12x2 + … + a1nxn + s1 = b1
a21x1 + a22x2 + … + a2nxn + s2 = b2
. . . .
. . . .
. . . .
am1x1 + am2x2 + … + amnxn + sm = bm
x1, x2, x3 … xn, s1, s2, … sm ≥ 0
We find that (s1, s2, …, sm) can be taken as the
basic vector B and their coefficients in the
objective function are zeroes, i.e. CB = (0, 0, 0 …0).
The simplex table representing the initial
solution is given below.

Table 3.1
x1 x2 x3 .... xs .... xn s1 s2 .... .... sm Const
CB B c1 c2 c3 .... cs .... cn 0 0 .... .... 0 0
0 s1 a11 a12 a13 .... a1s .... a1n 1 0 .... .... 0 b1
0 s2 a21 a22 a23 .... a2s .... a2n 0 1 .... .... 0 b2
0 s3 a31 a32 a33 a3s a3n 0 0 1 0 b3
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
0 sr ar1 ar2 ar3 . ars . arn . . . . . br
. . . . . . . . . . . . . . .
0 sm am1 am2 am3 . ams . amn 0 0 . . 1
Zj–cj –c1 –c2 –c3 .... –cs .... –cn 0 0 .... .... 0 0

The starting solution is Zj = CB Aj = 0 " j since CB = (0,0,0,…..,0) in

s1 = b1, this table.
s2 = b2, …sm = bm, Therefore Zj – cj = – cj. Zj – cj is called relative
Z0 = 0
cost factor and is denoted by c j . Thus c j = –cj in
The column corresponding to the basic
variables form an identity matrix Im. the starting table.
 28 Operations Research

3.2.4 Testing for Optimality The starting simplex table is

In the starting table if all the relative cost factors Table 3.2
are non-negative ( c j ≥ 0 " j) then it is not possible Z x1 x2 s1 s2 s3 Const
to increase Z further and the current solution itself
CB B 3 2 0 0 0 0
is the optimal solution.
0 s1 –2 1 1 0 0 1
Suppose there exist some c j < 0. Then select
0 s2 1 0 0 1 0 2
the most negative of them, say cs . The
0 s3 1 1 0 0 1 3
corresponding column As is called the pivot column
and the corresponding variable (non-basic) enters Zj – cj –3 –2 0 0 0 0
the basis and becomes a basic variable for the
improved solution. Thus xs is the entering variable. Most negative Zj – cj is –3 and hence A1 is the
Now find the ratios bi/ais for all ais > 0 in the As pivot column. x1 enters the basis. In the pivot
column and choose the minimum ratio, say br/ars. column the negative element –2 should not be
Then the corresponding rth row is called the pivot taken for finding the ratio. The other two ratios
row and the element ars is called the pivot element are 2/1 and 3/1 and the minimum ratio is 2/1 which
which indicates that the variable sr leaves the basis corresponds to the 2nd row and hence the 2nd row
and becomes non-basic in the improved solution. is the pivot row. Therefore s2 is the leaving variable
Since xs becomes a basic variable, the pivot and the pivot element is 1. Since x1 enters the basis
column should have all the elements equal to zero the first column should be transformed into
except the pivot element which should become
unity. In order to obtain this, we make elementary æ 0ö
row transformations. This process is called ç 1÷
ç ÷
iteration. In the modified table xB contains xs as a è 0ø
basic variable in place of sr and CB contains cs in using row transformations.
place of 0. Thus an improved solution is obtained. We make the transformations R1¢ = R1 – (– 2)/
Once again we find Zj – cj and examine whether (1) R2 i.e. R1 = R1 + 2R2 and this changes –2 to 0.
the current solution is optimal. If there is possibility Also we take R3¢ = R3 – (1)/ (1) R2 i.e. R3¢ = R3 – R2
of improving further proceed to find out the which reduces 1 to 0. The pivot element is 1 and
entering variable and leaving variable and modify needs no change. The new table is
the table. This process is repeated until we obtain
the optimal table in which c j ≥ 0 " j and the Table 3.3
corresponding solution is the optimal solution. Z x1 x2 s1 s2 s3 Const

Example 3.4 Solve by simplex method CB B 3 2 0 0 0 0

Maximize Z = 3x1+2x2 subject to the constraints 0 s1 0 1 1 2 0 5
– 2x1 + x2 £ 1 3 x1 1 0 0 1 0 2
x1 £ 2
0 s3 0 1 0 –1 1 1
x1+x2 £ 3
x1, x2 ≥ 0 Zj–cj 0 –2 0 3 0 6
Solution The standard form of the LPP is
Maximize Z = 3x1 + 2x2 subject to We have obtained an improved solution s1 = 5,
– 2x1 + x2 + s1 = 1
x1 + s2 = 2 x1 = 2, s3 = 1, Z = 6. There is one negative c j = –2.
x1 + x2 + s3 = 3 Hence this solution is not optimal and it can be
x1, x2, x3, s1, s2, s3 ≥ 0 improved further.
 Simplex Method 29

The pivot column is the x2 column. Hence x2 x1 + x2 + x3 + s1 = 10

enters the basis. Taking the ratios 5/1, 2/0 and 1/1 2x1 – x3 + s2 = 3
we find that the minimum is 1 corresponding to s3 2x1 – 2x2 + 3x3 + s3 = 0
row and therefore s3 leaves the basis. The pivot
x1, x2, x3, s1, s2, s3 ≥ 0
column should be changed to
The starting simplex table is
æ 0ö
ç 0÷ Table 3.5
ç ÷
è 1ø Z 1 –1 3 0 0 0
by row transformations. Since there is already 0
CB B x1 x2 x3 s1 s2 s3 Const
as the 2nd element, the x1 row needs no change.
Take R1¢ = R1 – R3. This reduces 1 to 0. Also 0 s1 1 1 1 1 0 0 10
the pivot element itself is 1. 0 s2 2 0 –1 0 1 0 3
The new table is 0 s3 2 –2 3 0 0 1 0
Table 3.4 Zj – cj –1 1 –3 0 0 0 0

Z x1 x2 s1 s2 s3 Const
The most negative c j is –3 and therefore x3
CB B 3 2 0 0 0 0
enters the basis. Minimum ratio is obtained as Min
0 s1 0 0 1 3 –1 4 (10/1, 0/3) = 0 corresponding to s3. Hence s3 leaves
3 x1 1 0 0 1 0 2 the basis. The necessary row transformations are
2 x2 0 1 0 –1 1 1 given by
Zj – cj 0 0 0 1 2 8 R1¢ = R1 – (1/3) R3; R2¢ = R2 + (1/3)R3;
R3¢ = (1/3)R3
Here all the Zj – cj are non-negative and hence
The new table is (I Iteration)
the current solution is optimal. The optimal
solution is Table 3.6
s1 = 4, x1 = 2, x2 = 1
Since the coefficient of the slack variable s1 is Z 1 –1 3 0 0 0 0
zero in the objective function it need not be taken CB B x1 x2 x3 s1 s2 s3 Const
into account. We give the solution in terms of the
0 s1 1/3 5/3 0 1 0 – 1/3 10
decision variables x1 and x2 only. Thus the optimal
solution is 0 s2 8/3 – 2/3 0 0 1 1/3 2
x1 = 2, x2 = 1, Z* = 8 3 x3 2/3 – 2/3 1 0 0 1/3 0
Example 3.5 Maximize Z = x1 – x2 + 3x3 subject Zj – cj 1 –1 0 0 0 1 0
to the constraints
x1 + x2 + x3 £ 10
2x1 – x3 £ 3 Pivot element is 5/3. x2 enters the basis and s1
2x1 – 2x2 + 3x3 £ 0 leaves.
x1, x2, x3 ≥ 0 Performing R1¢ = R1 (3/5) we get (II Iteration)
Solution The given LPP can be expressed in the R2¢ = R2 – (– 2/3)/(5/3)R1
standard form as follows: (i.e. R2¢ = R2 + (2/5)R1)
Maximize Z = x1 – x2 + 3x3 + 0s1 + 0s2 + 0s3 R3¢ = R3 – (– 2/3)/(5/3)R1
subject to (i.e. R3¢ = R3 + (2/5)R1)
 30 Operations Research

Table 3.7 Table 3.9

Z 1 –1 3 0 0 0 0 Z 2 2 4 0 0 0

CB B x1 x2 x3 s1 s2 s3 Const CB X B x1 x2 x3 s1 s2 s3 Const Ratio

–1 x2 1/5 1 0 3/5 0 – 1/5 6 0 s1 5/3 8/3 0 1 –1/3 0 200 120
0 s2 14/5 0 0 2/5 1 1/5 6 4 x3 1/3 1/3 1 0 1/3 0 100 300
3 x3 4/5 0 1 2/5 0 1/5 4 0 s3 2/3 8/3 0 0 –1/3 1 140 210

Zj–cj 6/5 0 0 3/5 0 4/5 6 Zj–cj –2/3 –2/3 0 0 4/3 0 400

All the relative cost factors are non-negative. x1 and x2 both have the same relative cost factor
Hence this solution is optimal. and hence the selection of entering variable is at
The solution is our choice. Let us choose x1 to enter the basis. The
x1 = 0, x2 = 6, x3 = 4, Z* = 6 minimum ratio is 120 and hence s1 leaves the basis.
Note: One of the decision variables x1 = 0. This Perform the operations
is called a degenerate solution. R1¢ = (3/5)R1; R2¢ = R2 – (1/5)R1,
Example 3.6 Maximize Z = 2x1 + 2x2 + 4x3 R3¢ = R3 – (2/5)R1
subject to the constraints We get the II iteration
2x1 + 3x2 + x3 £ 300
x1 + x2 + 3x3 £ 300 Table 3.10
x1 + 3x2 + x3 £ 240 Z 2 2 4 0 0 0
x1, x2, x3 ≥ 0
Solution The standard form of the LPP is CB B x1 x2 x3 s1 s2 s3 Const
Maximize Z = 2x 1 +2x 2 +4x 3 +0s 1 +0s 2 +0s 3 2 x1 1 8/5 0 3/5 – 1/5 0 120
subject to 4 x3 0 – 1/5 1 – 1/5 2/5 0 60
2x1 + 3x2 + x3 + s1 = 300 0 s3 0 8/5 0 – 2/5 – 1/5 1 60
x1 + x2 + 3x3 + s2 = 300
Zj – cj 0 2/5 0 2/5 6/5 0 480
x1 + 3x2 + x3 + s3 = 240
x1, x2, x3, s1, s2, s3 ≥ 0 Optimality is reached. The optimal solution is
The starting table is x1 = 120, x2 = 0, x3 = 60, Z* = 480
Table 3.8 (This is a degenerate solution since x2 = 0)
Example 3.7 Maximize Z = 2x1+4x2+x3 subject
Z 2 2 4 0 0 0 0
to the constraints
CB B x1 x2 x3 s1 s2 s3 Const x1 + 3x2 £ 4
0 s1 2 3 1 1 0 0 300 2x1 + x2 £ 3
0 s2 1 1 3 0 1 0 300 x2 + 4x3 £ 3
0 s3 1 3 1 0 0 1 240 x1, x2, x3 ≥ 0
Solution Standard form
Zj–cj –2 –2 –4 0 0 0 0
Maximize Z = 2x1 + 4x2 + x3 + 0s1 + 0s2 + 0s3
subject to the constraints
x3 enters the basis since – 4 is the most negative x1 + 3x2 + s1 = 4
relative cost factor. The minimum among the 2x1 + x2 + s2 = 3
ratios {300/1, 300/3, 240/1} is 100 which x2 + 4x3 + s3 = 3
corresponds to s2 and hence s2 leaves the basis.
x1, x2, x3, s1, s2, s3 ≥ 0
Take R 1¢ = R1 – (1/3) R2; R 2¢ = (1/3)R2; R3¢ =
R3 – (1/3)R2. We get the table (I iteration) Initial table
 Simplex Method 31

Table 3.11 Note: If we divide the third constraint of the

problem by 4, we get
Z 2 4 1 0 0 0 0
(1/4)x2 + x3 + (1/4)s3 = 3/4
CB B x1 x2 x3 s1 s2 s3 Const Ratio We get x3 itself as one of the starting basic
0 s1 1 3 0 1 0 0 4 4/3 variables instead of s3. As a result we can solve
0 s2 2 1 0 0 1 0 3 3/1 the problem in two iterations.
0 s3 0 1 4 0 0 1 3 3/1 3.2.5 Step by Step Procedure of
Zj – cj – 2 – 4 –1 0 0 0 0 Simplex Algorithm
x2 enters and s1 leaves Step 1 Rewrite the given LPP in the standard
R1¢ = R1/3; R2¢ = R2 – (1/3)R1; form.
R3¢ = R3 – (1/3)R1 Step 2 Form the simplex table giving the coeffi-
cients of all the variables along the corresponding
Table 3.12 columns (as given below)
I Iteration
Table 3.15
Z 2 4 1 0 0 0 0
Z c 1 c2 … c n 0 0 … 0 0
CB B x1 x2 x3 s1 s2 s3 Const Ratio
Coefficient Basic
4 x2 1/3 1 0 1/3 0 0 4/3 – of basic variables
0 s2 5/3 0 0 – 1/3 1 0 5/3 – variables
0 s3 – 1/3 0 4 – 1/3 0 1 5/3 5/12 CB B x 1 x2 xn s1 s2 sm Const
Zj – cj – 2/3 0 –1 4/3 0 0 16/3 0 s1 a11 a12 … a1n 1 0 … 0 b1
0 s2 a21 a22 … a2n 0 1 … 0 b2
R3¢ = R3/4 . . . . . . . . . . .
. . . . . . . . . . .
Table 3.13
0 sm am1 am2 … amn 0 0 … 1 bm
II Iteration
4 x2 1/3 1 0 1/3 0 0 4/3 4
Step 3 Write down the relative cost factor Zj – cj
for each column. If all the relative cost factors are
0 s2 5/3 0 0 – 1/3 1 0 5/3 1
non-negative, stop. The current solution is opti-
1 x3 –1/12 0 1 –1/12 0 1/4 5/12 – mal. Otherwise choose the most negative Zj – cj.
Zj– cj –5/12 0 0 4/3 0 1/4 69/12 The corresponding non-basic variable (say xs) en-
ters the basis. This column is the pivot column.
R1¢ = R1 – (1/5)R2, R2¢ = R2(3/5)
R3¢ = R3 + (1/20)R2 Step 4 Find the ratios bi/ais for positive ais in the
pivot column. Choose the minimum ratio, say br/ars.
Table 3.14 Then rth row is the pivot row and the corresponding
III Iteration basic variable leaves the basis. ars is the pivot element.
4 x2 0 1 0 2/5 –1/5 0 1 Step 5 In the pivot column, all the elements ex-
2 x1 1 0 0 – 1/5 3/5 0 1 cept the pivot element must be reduced to zero by
1 x3 0 0 1 –1/10 1/20 1/4 1/2 suitable row transformations. For the ith row, make
Zj– cj 0 0 0 11/10 9/20 1/4 13/2 the transformation
Ri¢ = Ri – (ais/ars)Rr i = 1 2 3 ….. m (i π r)
All Cj ≥ 0. Optimality is reached. The solution is For the pivot row Rr¢ = Rr/ars
x1 = 1, x2 = 1, x3 = 1/2, Z* = 13/2 Carry out the operations for all the rows,
This is a non-degenerate solution. including the constants.
 32 Operations Research

Step 6 In the new table include xr as a basic vari- x2 enters the basis and s1 leaves.
able and cr as its coefficient, in place of the leav- R1¢ = R1; R2¢ = R2 + R1; R3¢ = R3 – R1
ing variable. Find the relative cost factors Zj – cj.
This completes one iteration. Table 3.17

Step 7 Examine all the Zj – Cj. If all of them are I Iteration

non-negative, stop. The current solution is opti- Z 1 2 1 0 0 0 0
mal. If there exists negative Zj – Cj, repeat the pro- CB B x1 x2 x3 s1 s2 s3 Const Ratio
cedure from step 3.
2 x2 2 1 –1 1 0 0 2
Step 8 Once the optimality is reached write
0 s2 4 0 4 1 1 0 8 8/4 = 2
the solution, by giving the values of the basic
variables, from the constants column and the 0 s3 2 0 2 –1 0 1 4 4/2 = 2
optimum Z or Z* is given by the constant of the Zj–cj 3 0 –3 2 0 0 4
Zj – cj row. We can check the value of Z* by R1¢ = R1 + R3/2, R2¢ = R2 – 2R3, R3¢ = R3/2
substituting the values of the optimal basic
variables in the objective function. Table 3.18
Example 3.8 Maximize Z = x1 + 2x2 + x3 subject II Iteration
to the constraints 2 x2 3 1 0 1/2 0 1/2 4
2x1 + x2 – x3 £ 2
0 S2 0 0 0 3 1 –2 0
– 2x1 + x2 – 5x3 ≥ – 6
4x1 + x2 + x3 £ 6 1 X3 1 0 1 –1/2 0 1/2 2
xi ≥ 0 i = 1, 2, 3 Zj–cj 6 0 0 1/2 0 3/2 10
Solution Standard form of the LPP All Zj – c j ≥ 0. Optimality is reached. The
Maximize Z = x1 + 2x2 + x3 + 0s1 + 0s2 + 0s3 optimal solution is
subject to the constraints x1 = 0; x2 = 4; x3 = 2; Z* = 10
2x1 + x2 – x3 + s1 = 2
2x1 – x2 + 5x3 + s2 = 6 Example 3.9 Maximize Z = 12x1+15x2+14x3
subject to the constraints
4x1 + x2 + x3 + s3 = 6
x1, x2, x3, s1, s2, s3 ≥ 0 –x1 + x2 £ 0
It should be noted that the constraint –2x1 + x2 –x2 + 2x3 £ 0
– 5x3 ≥ –6 has to be changed as x1 + x2 + x3 £ 100
2x1 – x2 + 5x3 £ 6 and then s2 is added, since the x1 unrestricted.
constant should be non-negative. x2, x3 ≥ 0
Starting table Solution Given that x1 is unrestricted. Hence we
take x1 = u1 – v1 where u1 ≥ 0 and v1 ≥ 0. The
Table 3.16 standard form of the problem is
Z 1 2 1 0 0 0 0 Maximize Z = 12(u1 – v1) + 15x2 + 14x3 subject
to the constraints
CB B x1 x2 x3 s1 s2 s3 Const Ratio
– (u1 – v1) + x2 + s1 = 0
0 s1 2 1 –1 1 0 0 2 2/1 – x2 + 2x3 + s2 = 0
0 s2 2 –1 5 0 1 0 6 – (u1 – v1) + x2 + x3 + s3 = 100
0 s3 4 1 1 0 0 1 6 6/1 u1, v1, x2, x3, s1, s2, s3 ≥ 0
Zj–cj –1 –2 –1 0 0 0 0 The starting table is
 Simplex Method 33

Table 3.19 u1 = 40, v1 = 0, x2 = 40, x3 = 20, Z* = 1360

i.e. x1 = 40, x2 = 40, x3 = 20, Z* = 1360.
Z 12 –12 15 14 0 0 0 0
CB B u1 v1 x2 x3 s1 s2 s3 Const Ratio Example 3.10 Minimize Z = x1–3x2+2x3 subject
to the constraints
0 s1 –1 1 1 0 1 0 0 0 0
3x1 – x2 + 2x3 £ 7
0 s2 0 0 –1 2 0 1 0 0 –
–2x1 + 4x2 £ 12
0 s3 1 –1 1 1 0 0 1 100 100
–4x1 + 3x2 + 8x3 £ 10
Zj–cj –12 12 –15 –14 0 0 0 0 x1, x2, x3 ≥ 0
R1¢ = R1, R2¢ = R2 + R1, R3¢ = R 3 – R 1 Solution We have to change the problem into
maximization type. Take W = –Z.
Table 3.20 Maximize W = –x1 + 3x2 – 2x3
Therefore, the standard form is
I Iteration
Maximize W = –x1 + 3x2 – 2x3 subject to the
15 x2 – 1 1 1 0 1 0 0 0 – constraints
0 s2 – 1 1 0 2 1 1 0 0 – 3x1 – x2 + 2x3 + s1 = 7
0 s3 2 –2 0 1 –1 0 1 100 50 –2x1 + 4x2 + s2 = 12
Zj–cj –27 27 0 –14 15 0 0 0 – 4x1 + 3x2 + 8x3 + s3 = 10
x1, x2, x3, s1, s2, s3 ≥ 0
R1¢ = R1+(1/2)R3, R2¢ = R2 + (1/2)R3, The initial table is
R3¢ = (1/2)R3
Table 3.23
Table 3.21
Z –1 3 –2 0 0 0 0
II Iteration CB B x1 x2 x3 s1 s2 s3 Const Ratio
15 x2 0 0 1 1/2 1/2 0 1/2 50 100 0 s1 3 –1 2 1 0 0 7 –
0 s2 0 0 0 5/2 1/2 1 1/2 50 100/3 0 s2 –2 4 0 0 1 0 12 3
12 u1 1 –1 0 1/2 –1/2 0 1/2 50 100 0 s3 –4 3 8 0 0 1 10 10/3
Zj–cj 0 0 0 –1/2 3/2 0 27/2 1350 Zj–cj 1 –3 2 0 0 0 0 –
R1¢ = R1 – (1/5) R2, R2¢ = R2(2/5), R1¢ = R1 + (1/4)R2, R2¢ = R2/4,
R3¢ = R3 – (1/5)R2 R3¢ = R3 – (3/4)R2

Table 3.22 Table 3.24

III Iteration I Iteration

15 x2 0 0 1 0 2/5 –1/5 2/5 40 – 0 s1 5/2 0 2 1 1/4 0 10 4
14 x3 0 0 0 1 1/5 2/5 1/5 20 – 3 x2 –1/2 1 0 0 1/4 0 3 –
12 u1 1 –1 0 0 –3/5 –1/5 2/5 40 – 0 s3 –5/2 0 8 0 –3/4 1 1 –
Zj–cj 0 0 0 0 8/5 1/5 68/5 1360 – Zj–cj –1/2 0 2 0 3/4 0 9 –
All Zj – cj ≥ 0. Optimal solution reached. The R1¢ = R1(2/5), R2¢ = R2 + R1/5,
optimal solution is R3¢ = R3 + R1
 34 Operations Research

Table 3.25 8x1 + 2x2 + 3x3 + s1 = 250

4x1 + 3x2 + s2 = 150
II Iteration 2x1 + x3 + s3 = 50
Z –1 3 –2 0 0 0 0 x1, x2, x3, s1, s2, s3 ≥ 0
CB B x1 x2 x3 s1 s2 s3 Const Ratio
The starting simplex table is
–1 x1 1 0 4/5 2/5 1/10 0 4 – Table 3.27
3 x2 0 1 2/5 1/5 3/10 0 5 – Z 20 6 8 0 0 0 0
0 s3 0 0 10 1 –1/2 1 11 – CB B x1 x2 x3 s1 s2 s3 Const Ratio
Zj–cj 0 0 12/5 1/5 4/5 0 11 – 0 s1 8 2 3 1 0 0 250 250/8
All Zj – cj ≥ 0. The optimal solution is reached. 0 s2 4 3 0 0 1 0 150 150/4
The optimal solution is 0 s3 2 0 1 0 0 1 50 25
x1 = 4, x2 = 5, x3 = 0, W* = 11 Zj–cj –20 –6 –8 0 0 0 0 0
The solution of the given LPP is
x1 = 4, x2 = 5, x3 = 0, Z* = –11 R1¢ = R1 – 4R3, R2¢ = R2 – 2R3 ,
R3¢ = R3/2
Example 3.11 An auto parts company produces
three types of parts A, B, C. The capacity of the Table 3.28
machines and the number of machine hours I Iteration
required for one unit of each part is given below
0 s1 0 2 –1 1 0 –4 50 50/2
Table 3.26 0 s2 0 3 –2 0 1 –2 50 50/3
Machine type Available Productivity in 20 x1 1 0 1/2 0 0 1/2 25 –
machine machine hours Zj–cj 0 –6 2 0 0 10 500 –
hours per unit
per week R1¢ = R1 – (2/3)R2, R2¢ = R2/3, R3¢ = R3
A B C
Table 3.29
Hobbling 250 8 2 3
II Iteration
Shaping 150 4 3 0
Grinding 50 2 – 1 0 s1 0 0 1/3 1 –2/3 –8/3 50/3 50
6 x2 0 1 –2/3 0 1/3 –2/3 50/3 –
The profit is Rs 20, Rs 6 and Rs 8 for one unit
of A, B and C respectively. Find out how much of 20 x1 1 0 1/2 0 0 1/2 25 50
each part the company should produce per week Zj–cj 0 0 –2 0 2 6 600 –
in order to maximize the profit.
R1¢ = R1 – (2/3)R3, R2¢ = R2 + (4/3)R3,
Solution Formulation
R3¢ = 2R3
Maximize Z = 20x1 + 6x2 + 8x3 subject to the
constraints Table 3.30
8x1 + 2x2 + 3x3 £ 250 III Iteration
4x1 + 3x2 £ 150
0 s1 –2/3 0 0 1 –2/3 –3 0 –
2x1 + x3 £ 50 6 x2 4/3 1 0 0 1/3 0 50 –
x1, x2, x3 ≥ 0 8 x3 2 0 1 0 0 1 50 –
The standard form is Zj–cj 4 0 0 0 2 8 700 –
Maximize Z = 20x1+6x2+8x3 subject to the
constraints The optimal solution is reached.
 Simplex Method 35

The optimal solution is x1 = 0, x2 = 50, x3 = 50, The starting table is

Z* = 700
Table 3.31
The company has to produce 50 units of part B
and 50 units of part C and need not produce part Z 2 3 4 0 0 0 0
A. The maximum profit is Rs 700. CB B x1 x2 x3 s1 s2 s3 Const Ratio
0 s1 –1 –5 –9 1 0 0 2 –
Example 3.12 Maximize Z = 2x1 + 3x2 + 4x3 0 s2 3 –1 1 0 1 0 10 10
subject to the constraints 0 s3 2 3 –7 0 0 1 0 –
–x1 – 5x2 – 9x3 £ 2 Zj–cj –2 –3 –4 0 0 0 0 –
3x1 – x2 + x3 £ 10 R1¢ = R1 + 9R2, R2¢ = R2, R3¢ = R3 + 7R2
2x1 + 3x2 – 7x3 £ 0
Table 3.32
x1, x2, x3 ≥ 0
I Iteration
Solution The standard form of the LPP is
0 s1 26 –14 0 1 9 0 92 –
Maximize Z = 2x1 + 3x2 + 4x3 subject to the
4 x3 3 –1 1 0 1 0 10 –
constraints
0 s3 23 –4 0 0 7 1 70 –
–x1 – 5x2 – 9x3 + s1 = 2
Zj–cj 10 –7 0 0 4 0 40 –
3x1 – x2 + x3 + s2 = 10
All the elements of the pivot column are
2x1 + 3x2 – 7x3 + s3 = 0
negative. Therefore the problem has unbounded
x1, x2, x3, s1, s2, s3 ≥ 0 solution.

S H O RT A N S W E R Q U E S T I O N S
■ ■ ■ ■ ■ ■ ■

1. Define feasible solution 15. Fill in the blanks:

2. Define optimal solution (i) The number of basic solutions to a
3. Define slack variable system of 3 simultaneous equations in 4
4. Define surplus variable unknowns is _________.
5. Define unrestricted variable (Ans: 4C3 = 4)
6. Define degenerate solution (ii) The set of all feasible solutions to an LPP
7. When do you say that an LPP has an is a ________ set. (Ans: convex)
unbounded solution? (iii) If none of the decision variables is zero
8. When do you say that an LPP has infinite then the solution is called a _______
number of solutions? solution. (Ans: non-degenerate)
9. Give the matrix form of an LPP. (iv) If some decision variable is zero the
10. State the characteristics of the standard form solution is called a __________.
of an LPP? (Ans: degenerate)
11. How is the pivot column selected? (v) The entering variable is the non-basic
12. How is the pivot row selected? variable corresponding to ________.
13. How do you find out whether a solution is (Ans : most negative Zj – cj)
optimal or not. (vi) The entering variable column is called
14. What is the test of optimality? a________. (Ans: pivot column)
 36 Operations Research

(vii) The leaving variable row is called a
__________. (Ans: pivot row)
(viii) A constraint of £ type is changed to
equality by adding a _________.
(Ans: slack variable)
(ix) A constraint of ≥ type is reduced to
equality by introducing a _________.
(Ans: surplus variable)
(x) A slack variable is introduced in £
inequality constraint to obtain a
__________. (Ans: basic solution)
16. State whether True or False
(viii) Every feasible solution is optimal
(i) The main components of an LPP include
the decision variables, constraints and
the objective function. (Ans: True)
(ii) A feasible solution satisfies all the
constraints. (Ans: True)
(iii) Graphical method is useful when the
problem involves two variables.
(Ans: True)
(iv) Optimal solution occurs at one of the
vertices of the feasible region.
(Ans: True)
(v) An LPP may have more than one optimal
solution. (Ans: True)
(vi) Every basic solution is a feasible
solution. (Ans: False)
(vii) An optimal solution is a feasible
solution. (Ans: True)
(Ans: False)
(ix) In an LPP of standard form all the
variable are non-negative (Ans: True)
(x) In the simplex iteration the pivot element
cannot be zero or negative. (Ans: True)

EXERCISES
■ ■ ■ ■ ■ ■ ■

Rewrite the following LP problems in the 3x1 + 4x2 + x3 £ 8

standard form: x1, x2 ≥ 0, x3 £ 0
1. Maximize Z = 3x 1 + 2x 2 subject to the 5. Maximize Z = x1 + 2x2 + 5x3 + x4 subject to
constraints the constraints
–2x1 + x2 £ 1 2x1 + x2 + 3x3 – x4 £ 10
x1 £ 2 x1 + 3x2 + x3 + 2x4 £ 16
x1 + x2 £ 3 x1 unrestricted, x2, x3, x4 ≥ 0
x1, x2 ≥ 0 Solve the following using simplex method:
2. Maximize Z = 5x 1 + 7x 2 subject to the 6. Maximize Z = 4x1 + 10x2 subject to the
constraints constraints
x1 + x2 £ 8 2x1 + x2 £ 50
3x1 + 4x2 £ 20 2x1 + 5x2 £ 100
6x1 + 7x2 £ 24 2x1 + 3x2 £ 90
x1, x2 ≥ 0 x1, x2 ≥ 0
3. Minimize Z = 2x1 + 3x2 + x3 subject to the 7. Maximize Z = 3x1 + 2x2 + 5x3 subject to the
constraints constraints
x1–2x2 + 3x3 ≥ –4 x1 + 4x2 £ 420
2x1 + 5x2–7x3 £ 8 3x1 + 2x3 £ 460
x1, x2, x3 ≥ 0 x1 + 2x2 + x3 £ 430
4. Maximize Z = 4x1 + 2x2 + 6x3 subject to the x1, x2, x3 ≥ 0
constraints 8. Maximize Z = 6x 1 + 9x 2 subject to the
2x1 + 3x2 + 2x3 ≥ 6 constraints
 Simplex Method 37

x1 + x2 £ 12 15. Maximize Z = 5x1 + 3x2 + 7x3 subject to the

x1 + 5x2 £ 44 constraints
3x1 + x2 £ 30 x1 + x2 + 2x3 £ 22
x1, x2 ≥ 0 3x1 + 2x2 + x3 £ 26
9. Minimize Z = 8x 1 – 2x 2 subject to the x1 + x2 + x3 £ 18
constraints x1, x2, x3 ≥ 0
–4x1 + 2x2 £ 1 16. Minimize Z = 3x1 – 2x2 – x3 subject to the
5x1 – 4x2 £ 3 constraints
x1, x2 ≥ 0 –x1 + 2x2 + 3x3 £ 7
10. Maximize Z = 300x1 + 200x2 subject to the 4x1 – 2x3 £ 12
constraints 3x1 + 8x2 + 4x3 £ 10
5x1 + 2x2 £ 180 x1, x2, x3 ≥ 0
x1 + x2 £ 45 17. Maximize Z = 9x1 + 2x2 + 5x3 subject to the
x1, x2 ≥ 0 constraints
11. Maximize Z = 100x1 + 200x2 + 50x3 subject 2x1 + 3x2 – 5x3 £ 12
to the constraints 2x1 – x2 + 3x3 £ 3
x1 + x2 + 2x3 £ 200 3x1 + x2 – 2x3 £ 2
10x1 + 8x2 + 5x3 £ 2000 x1, x2, x3 ≥ 0
2x1 + x2 £ 100 18. Maximize Z = 2x + 3y + 4z subject to the
x1, x2, x3 ≥ 0 constraints
12. Maximize Z = 3x1 + 2x2 + 5x3 subject to the 3x + 2z £ 41
constraints 2x + y + z £ 35
x1 + 2x2 + x3 £ 43 2y + 3z £ 30
3x1 + 2x3 £ 46 x, y, z ≥ 0
x1 + 4x2 £ 42 19. Maximize Z = x1 + 4x2 + 5x3 subject to the
x1, x2, x3 ≥ 0 constraints
13. A firm manufactures two products A and B 3x1 + 6x2 + 3x3 £ 22
using wood, plastic and steel. The availability x1 + 2x2 + 3x3 £ 14
and requirements are given below. 3x1 + 2x2 £ 14
x1, x2, x3 ≥ 0
Availability (kg) Requirement 20. A farmer has a 1000 acres of land on which
for one unit he grows corn, wheat or soyabeans.
A B
The particulars are given below.
Wood 240 1 3
Plastic 370 3 4 Crop Cost of preparation Number of man- Profit per
Steel 180 2 1 per acre (Rs) days of work per acre (Rs)
acre
The company gets profits of Rs 4 and Rs 6 Corn 100 7 30
on one unit of A and B respectively. Wheat 120 10 40
Determine the number of units of A and B to Soyabeans 70 8 20
be produced in order to get maximum profit. If the farmer has Rs 1,00,000 for preparation
14. Maximize Z = 3x1 + 2x2 + 5x3 subject to the and can count on 8000 man-days of work
constraints how many acres should be allotted to each
x1 + x2 + x3 £ 9 crop in order to maximize profit?
2x1 + 3x2 + 5x3 £ 30 21. Three grades of coal A, B and C contain
2x1 – x2 – x3 £ 8 phosphorus and ash as impurities. For a
x1, x2, x3 ≥ 0 particular industrial process a maximum of
 38 Operations Research

100 tons of coal is required which should 27. Maximize Z = 2x 1 + 3x 2 subject to the
contain not more than 3 per cent of ash and constraints
not more than 0.03 per cent of phosphorus. –x1 + 2x2 £ 4
The percentage of impurities and the profits x1 + x2 £ 6
from each grade are given below. x1 + 3x2 £ 9
x1, x2 ≥ 0
Coal Phosphorus % Ash % Profit Rs/ton 28. Maximize Z = 4x1 + 10x2 subject to the
A .02 3 12 constraints
2x1 + x2 £ 10
B .04 2 15
2x1 + 5x2 £ 20
C .03 5 14 2x1 + 3x2 £ 18
Find the proportions in which the three x1, x2 ≥ 0
grades are to be used in order to maximize 29. Maximize Z = 2x 1 + x 2 subject to the
the profit. constraints
x1 – 2x2 £ 1
22. Maximize Z = 2x 1 + 3x 2 subject to the
x1 + x2 £ 6
constraints
x1 + 2x2 £ 10
–x1 + 2x2 £ 4
x1 – x2 £ 2
x1 + x2 £ 6 x1, x2 ≥ 0
x1 + 3x2 £ 9 30. Maximize Z = 4x 1 + 6x 2 subject to the
x1, x2 unrestricted constraints
23. Minimize Z = 3x + 2y subject to the x1 + 3x2 £ 240
constraints 3x1 + 4x2 £ 270
–2x + 3y £ 9 2x1 + x2 £ 180
–x + 5y £ 20 x1, x2 ≥ 0
x, y ≥ 0 31. Maximize Z = 30x1 + 40x2 + 20x3 subject to
24. Maximize Z = 6x1 + 8x2 + 4x3 + 3x4 subject the constraints
to the constraints 10x1 + 12x2 + 7x3 £ 10000
2x1 + 3x2 + x3 £ 900 7x1 + 10x2 + 8x3 £ 8000
x2 + 2x3 £ 600 x1 + x2 + x3 £ 1000
3x1 + 2x2 + 2x3 £ 1200 x1, x2, x3 ≥ 0
3x3 + x4 = 100 32. Maximize Z = 2x1 + 4x2 + 3x3 subject to the
x1, x2, x3, x4 ≥ 0 constraints
[Hint: x4 can be taken as an initial basic 3x1 + 4x2 + 2x3 £ 60
variable] 2x1 + x2 + 2x3 £ 40
25. Maximize Z = 5x 1 + 2x 2 subject to the x1 + 3x2 + 2x3 £ 80
constraints x1, x2, x3 ≥ 0
33. Maximize Z = 10x1 + x2 + 2x3 subject to the
2x1 + x2 £ 8
constraints
x2 £ 4
x1 + x2 – 2x3 £ 10
x1 £ 3 4x1 + x2 + x3 £ 20
x1 ≥ 0, x2 unrestricted. x1, x2, x3 ≥ 0
26. Maximize Z = 4x 1 + 3x 2 subject to the 34. Maximize Z = x1 + x2 + 3x3 subject to the
constraints constraints
2x1 + x2 £ 1000 3x1 + 2x2 + x3 £ 3
x1 + x2 £ 800 2x1 + x2 + 2x3 £ 2
x1, x2 ≥ 0 x1, x2, x3 ≥ 0

35. Maximize Z = 3x1 + 5x2 + 4x3 subject to the
constraints
2x1 + 3x2 £ 8
2x2 + 5x3 £ 10
3x1 + 2x2 + 4x3 £ 15
x1, x2, x3 ≥ 0
36. Maximize Z = 4x1 + x2 + 3x3 + 5x4 subject to
the constraints
4x1 – 6x2 – 5x3 – 4x4 ≥ – 20
–3x1 – 3x2 + 4x3 + x4 £ 10
–8x1 – 3x2 + 3x3 + 2x4 £ 20
x1, x2 , x3, x4 ≥ 0
37. Maximize Z = 2x1 + x2 – 3x3 + 5x4 subject to
the constraints
x1 + 7x2 + 3x3 + 7x4 £ 46
3x1 – x2 + x3 + 2x4 £ 8
2x1 + 3x2 – x3 + x4 £ 10
x1, x2 , x3, x4 ≥ 0
ANSWERS
■ ■ ■ ■ ■ ■ ■
1. Maximize Z = 3x1 + 2x2 + 0s1 + 0s2 + 0s3
subject to the constraints
–2x1 + x2 + s1 = 1
x1 + s2 = 2
x1 + x2 + s3 = 3
x1, x2, s1,s2, s3 ≥ 0.
2. Maximize Z = 5x1 + 7x2 + 0s1 + 0s2 + 0s3
subject to the constraints
x1 + x2 + s1 = 8
3x1 + 4x2 + s2 = 20
6x1 + 7x2 + s3 = 24
x1, x2, s1, s2, s3 ≥ 0
3. Maximize (–Z) = –2x1 – 3x2 – x3 subject to
the constraints
–x1 + 2x2 – 3x3 + s1 = 4
2x1 + 5x2 – 7x3 + s2 = 8
x1, x2, x3, s1, s2 ≥ 0
4. Maximize Z = 4x1 + 2x2 – 6x3¢ + 0s1 + 0s2 +
0s3 subject to the constraints
2x1 + 3x2 – 2x3¢ – s1 = 6
3x1 + 4x2 – x3¢ + s2 = 8
x1, x2, x3¢, s1, s2 ≥ 0 (x3¢ = –x3)
Simplex Method 39
38. Maximize Z = 3x1 + 4x2 + x3 + 5x4 subject to
the constraints
8x1 + 3x2 + 2x3 + 2x4 £ 10
2x1 + 5x2 + x3 + 4x4 £ 5
x1 + 2x2 + 5x3 + x4 £ 6
x1, x2 , x3, x4 ≥ 0
39. Maximize Z = 15x1 + 6x2 + 9x3 + 2x4 subject
to the constraints
2x1 + x2 + 5x3 + 6x4 £ 20
3x1 + x2 + 3x3 + 25x4 £ 24
7x1 + x4 £ 70
x1, x2 , x3, x4 ≥ 0
40. Maximize Z = 2x1 – 4x2 + 5x3 – 6x4 subject
to the constraints
x1 + 4x2 – 2x3 + 8x4 £ 2
–x1 + 2x2 + 3x3 + 4x4 £ 1
x1, x2 , x3, x4 ≥ 0
5. Maximize Z = u1–v1 + 2x2 + 5x3 + x4 + 0s1 +
0s2 subject to the constraints
2u1 – 2v1 + x2 + 3x3 – x4 + s1 = 10
u1 – v1 + 3x2 + x3 + 2x4 + s2 = 16
u1, v1, x2, x3, x4, s1, s2 ≥ 0 (x1 = u1 – v1)
6. x1 = 0, x2 = 20, Z* = 200
7. x1 = 0, x2 = 100, x3 = 230, Z* = 1350
8. x1 = 4, x2 = 8, Z* = 96
9. x1 = 0, x2 = 1/2, Z* = –1
10. x1 = 30, x2 = 15, Z* = 12000
11. x1 = 0, x2 = 100, x3 = 50, Z* = 22500
12. x1 = 0, x2 = 10, x3 = 23, Z* = 135
13. x1 = 30, x2 = 70, Z* = 540
14. x1 = 5, x2 = 0, x3 = 4, Z* = 35
15. x1 = 6, x2 = 0, x3 = 8, Z* = 86
16. x1 = 0, x2 = 5/4, x3 = 0, Z* = –5/2
17. x1 = 0, x2 = 12, x3 = 5, Z* = 49
18. x = 11, y = 9, z = 4, Z* = 65
19. x1 = 0, x2 = 2, x3 = 10/3, Z* = 74/3
20. Maximize Z = 30x1 + 40x2 + 20x3 subject to
the constraints
10x1 + 12x2 + 7x3 £ 10000
 40 Operations Research

7x1 + 10x2 + 8x3 £ 8000 26. x1 = 200, x2 = 600, Z* = 2600

x1 + x2 + x3 £ 1000 27. Unbounded solution
x1, x2, x3 ≥ 0 28. x1 = 15/4, x2 = 5/2, Z* = 40
Also x1 = 250, x2 = 625, x3 = 0, Z* = 32500 29. x1 = 4, x2 = 2, Z* = 10
21. Maximize Z = 12x1 + 15x2 + 14x3 subject to 30. x1 = 30, x2 = 70, Z* = 540
the constraints 31. x1 = 250, x2 = 625, x3 = 0, Z* = 32500
–x1 + x2 £ 0 32. x1 = 0, x2 = 20/3, x3 = 50/3, Z* = 250/3
–x2 + 2x3 £ 0 33. x1 = 10, x2 = 0, x3 = 0, Z* = 50
x1 + x2 + x3 £ 100 34. x1 = 0, x2 = 0, x3 = 1, Z* = 3
35. x1 = 89/41, x2 = 50/41, x3 = 64/41,
x1, x2, x3 ≥ 0
Z* = 765/41
Also x1 = 40, x2 = 40, x3 = 20, Z* = 1360
36. Unbounded solution
22. x1 = 9/2, x2 = 3/2, Z* = 27/2 37. x1 = 0, x2 = 12/7, x3 = 0, x4 = 34/7, Z* = 26
23. Unbounded solution 38. x1 = 15/14, x2 = 0, x3 = 0, x4 = 5/7,
24. x1 = 1800/7, x2 = 900/7, x3 = 0, x4 = 100, Z* = 95/14
Z* = 20100/7 39. x1 = 4, x2 = 12, x3 = 0, x4 = 0, Z* = 132
25. x1 = 3, x2 = 2, Z* = 19 40. x1 = 8, x2 = 0, x3 = 3, x4 = 0, Z* = 31

Artificial Variables
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
BIG M METHOD AND
TWO-PHASE METHOD
4.1 ARTIFICIAL VARIABLES
In the previous chapter we considered LP problem
involving upper bound constraints (£ type) only.
These constraints were changed to equalities by
introducing slack variables. These slack variables
become the initial basic variables and provide a
feasible solution to start with. Now we consider
problems having lower bound constraints (≥ type)
and equality constraints. The inequality of ≥ type
can be changed to equality by subtracting a surplus
variable. But its coefficient is –1 and hence it has
a negative value, and therefore it gives rise to an
infeasible solution. Thus a surplus variable cannot
become a starting basic variable. In order to obtain
an initial basic feasible solution we introduce
another variable called artificial variable.
For example, consider the constraint
2x1 + 3x2 + 4x3 ≥ 10
Introducing a surplus variable s we get
2x1 + 3x2 + 4x3 – s = 10
\ s = 2x1 + 3x2 + 4x3 – 10
4
Assigning zero values to the non-basic variables
we get s = – 10 (negative value)
Therefore we introduce an artificial variable R
and write
2x1 + 3x2 + 4x3 – s + R = 10
\ R = 10 – 2x1 – 3x2 – 4x3 + s
Now R becomes a starting basic variable with
a value 10.
4.2 BIG M METHOD (CHARNE’S
PENALTY METHOD)
The purpose of introducing the artificial variable
is just to obtain an initial basic feasible solution.
But already we have equality with a surplus
variable. Adding one more variable to only one
side of this equality causes violation of the equality
constraint. In order to overcome this difficulty we
need to get rid of these variables and not to allow
them to appear as optimal basic variables. If at all
an artificial variable becomes a basic variable in
the final table, its value must be zero. This will
ensure a solution to the problem. To achieve this,
these artificial variables are assigned a large
penalty cost coefficient (–M) in the objective
function to be maximized. However, if there is an
 42 Operations Research

artificial variable in the optimal basis with positive R1¢ = R1 + R2,

value then it indicates that the given problem has R2¢ = 2R2,
no solution. R3¢ = R3
Example 4.1 Maximize Z = 3x1 – x2 subject to
Table 4.3
the constraints
2x1 + x2 ≥ 2 II Iteration
x1 + 3x2 £ 3
Z x1 x2 s1 s2 s3 R Cons Ratio
x2 £ 4
x1, x2 ≥ 0 CB B 3 –1 0 0 0 –M 0
Solution The standard form of the problem is 3 x1 1 3 0 1 0 0 3 –
Maximize Z = 3x1 – x2 + 0s1 + 0s2 + 0s3 – MR 0 s1 0 5 1 2 0 –1 4 –
subject to the constraints
0 s3 0 1 0 0 1 0 4 –
2x1 + x2 – s1 + R = 2
x1 + 3x2 + s2 = 3 Zj – cj 0 10 0 3 0 M 9 –
x2 + s3 = 4
x1, x2, s1, s2, s3, R ≥ 0 All c j ≥ 0 and therefore optimality is reached.
Here s1 is a surplus variable, s2 and s3 are slack Optimal solution is x1= 3, x2= 0, Z*= 9
variables and R is an artificial variable with a Example 4.2 Minimize Z = 5x1 + 4x2 + 3x3
penalty cost – M in the objective function. M is a
subject to the constraints
very large number tending to infinity. The starting
simplex table is x1 + x2 + x3 = 100
x1 £ 20
Table 4.1 x2 ≥ 30
Z x1 x2 s1 s2 s3 R Cons Ratio x3 £ 40
CB B 3 –1 0 0 0 –M 0 x1, x2, x3 ≥ 0
–M R 2 1 –1 0 0 1 2 1 Solution Here the first constraint is an equality
0 s2 1 3 0 1 0 0 3 3 and hence we introduce an artificial variable only.
In the 2nd and 4th constraint slack variables are
0 s3 0 1 0 0 1 0 4 –
introduced. In the third constraint we have to
Zj–cj –2M–3 –M + 1 M 0 0 0 –2M
introduce a surplus variable and an artificial
R1¢ = (1/2) R1, variable.
R2 = R2 – (1/2) R1, The standard form of the LPP is
R3¢ = R3 Maximize W = (–Z) = –5x1 – 4x2 – 3x3 + 0s1 +
0s 2 + 0s 3 – MR 1 – MR 2 subject to the
Table 4.2
constraints
I Iteration x1 + x2 +x3 + R1 = 100
3 x1 1 1/2 –1/2 0 0 1/2 1 – x1 + s1 = 20
0 s2 0 5/2 1/2 1 0 –1/2 2 4 x2 – s2 + R2 = 30
0 s3 0 1 0 0 1 0 4 – x3 + s3 = 40
Zj – cj 0 5/2 –3/2 0 0 3/2+M 3 x1, x2, x3, s1, s2, s3, R1, R2 ≥ 0
 Artificial Variables 43

The initial table is

Table 4.4
Z x1 x2 x3 s1 s2 s3 R1 R2 Const Ratio
CB B –5 –4 –3 0 0 0 –M –M 0
–M R1 1 1 1 0 0 0 1 0 100 100
0 s1 1 0 0 1 0 0 0 0 20 –
–M R2 0 1 0 0 –1 0 0 1 30 30
0 s3 0 0 1 0 0 1 0 0 40 –
Zj – cj –M + 5 –2M + 4 –M + 3 0 M 0 0 0 –130M
R1¢ = R1 – R3 No other change
Table 4.5
I Iteration
–M R1 1 0 1 0 1 0 1 –1 70 70
0 s1 1 0 0 1 0 0 0 0 20 –
–4 x2 0 1 0 0 –1 0 0 1 30 –
0 s3 0 0 1 0 0 1 0 0 40 –
Zj – cj –M + 5 0 –M + 3 0 –M + 4 0 0 2M – 4 –120 – 70M –
s2 is chosen to enter the basis in order to get rid of R1 : R3¢ = R3 + R1
Table 4.6
II Iteration
0 s2 1 0 1 0 1 0 1 –1 70 70
0 s1 1 0 0 1 0 0 0 0 20 –
–4 x2 1 1 1 0 0 0 1 0 100 100
0 s3 0 0 1 0 0 1 0 0 40 40
Zj – cj 1 0 –1 0 0 0 –4 + M M –400 –

R1¢ = R1 – R4 R3¢= R3 – R4
Table 4.7
III Iteration
0 s2 1 0 0 0 1 –1 1 –1 30 –
0 s1 1 0 0 1 0 0 0 0 20 –
–4 x2 1 1 0 0 0 –1 1 0 60 –
–3 x3 0 0 1 0 0 1 0 0 40 –
Zj – cj 1 0 0 0 0 1 –4 + M M –360 –
The optimal solution is
x1= 0, x2 = 60, x3 = 40, Z* = –W* = 360
 44 Operations Research

Example 4.3 Maximize Z = x1 + 2x2 + 3x3 – x4 2x1 + x2 + 5x3 – s1 + R2 = 20

subject to constraints x1 + 2x2 + x3 + x4 – s2 + R3 = 10
x1 + 2x2 + 3x3 = 15 x1, x2, x3, x4, s1, s2, R1, R2, R3 ≥ 0
2x1 + x2 + 5x3 ≥ 20
Note: In the simplex table the column corre-
x1+ 2x2 +x3 +x4 ≥ 10
sponding to x4 is
x1, x2, x3, x4 ≥ 0
Solution Introducing surplus variables and æ 0ö
artificial variables, we get the standard form of ç 0÷
the LPP as ç ÷
è 1ø
Maximize Z = x1 +2x2 + 3x3 –x4 + 0s1 + 0s2 –
and therefore we can take x4 itself as a starting
MR1– MR2 – MR3 subject to constraints
basic variable. R3 need not be introduced.
x1 + 2x2 + 3x3 + R1 = 15

Table 4.8
Z x1 x2 x3 x4 s1 s2 R1 R2 R3 Const Ratio
CB B 1 2 3 –1 0 0 –M –M –M 0
–M R1 1 2 3 0 0 0 1 0 0 15 15/3
–M R2 2 1 5 0 –1 0 0 1 0 20 20/5
–M R3 1 2 1 1 0 –1 0 0 1 10 10/1
Zj – cj –4M – 1 –5M – 2 –9M – 3 –M + 1 M M 0 0 0 –45M

R1¢ = R1 – ( 3/5) R2, R2¢ = (1/5) R2, R3¢ = R3 – (1/5)R2

Table 4.9
I Iteration
–M R1 –1/5 7/5 0 0 3/5 0 1 –3/5 0 3 15/7
3 X3 2/5 1/5 1 0 –1/5 0 0 1/5 0 4 20
–M R3 3/5 9/5 0 1 1/5 –1 0 –1/5 1 6 30/9
Zj – cj (–2M+1)/5 (–16M – 7)/5 0 –M + 1 (–4M – 3)/5 M 0 (9M + 3)/5 0 –9M + 12

R1¢ = (5/7) R1, R2¢ = R2 – (1/7)R1, R3¢ = R3 – (9/7)R1

Table 4.10
II Iteration
2 x2 –1/7 1 0 0 3/7 0 5/7 –3/7 0 15/7
3 x3 3/7 0 1 0 –2/7 0 –1/7 2/7 0 25/7
–M R3 6/7 0 0 1 –4/7 –1 –9/7 4/7 1 15/7
Zj – cj –6M/7 0 0 –M+1 4M/7 M (16M+7)/7 3M/7 0 –15M/7+15

x4 enters, R3 leaves (no change)

 Artificial Variables 45

Table 4.11
III Iteration
Z x1 x2 x3 x4 s1 s2 R1 R2 R3 Const Ratio
CB B 1 2 3 –1 0 0 –M –M –M 0
2 x2 –1/7 1 0 0 3/7 0 5/7 –3/7 0 15/7 –
3 x3 3/7 0 1 0 –2/7 0 –1/7 2/7 0 25/7 25/3
–1 x4 6/7 0 0 1 –4/7 –1 –9/7 4/7 1 15/7 15/6
Zj – Cj 0 0 0 4/7 1 (16M + 7)/7 –4/7 + M –1 + M 90/7

R1¢ = R1 + (1/6)R3, R2¢ = R2 – (1/2)R3, R3¢ = (7/6)R3

Table 4.12
IV Iteration
2 x2 0 1 0 1/6 1/3 –1/6 1/2 –1/3 1/6 5/2
3 x3 0 0 1 –1/2 0 1/2 1/2 0 –1/2 5/2
1 x1 1 0 0 7/6 –2/3 –7/6 –3/2 2/3 7/6 5/2
Zj – cj 0 0 0 1 0 0 1+M M M 15

The optimal solution is Table 4.13

x1 = 5/2, x2 = 5/2, x3 = 5/2, x4 = 0, Z* = 15 Z x1 x2 s1 s2 s3 R Const Ratio
Note: When an artificial variable leaves the
basis we may drop that artificial variable and omit CB B 1 4 0 0 0 –M 0 –
all the entries corresponding to its column from 0 s1 3 1 1 0 0 0 3 3/1
the simplex table. 0 s2 2 3 0 1 0 0 6 6/3
Example 4.4 Maximize Z = x1 + 4x2 subject to –M R 4 5 0 0 –1 1 20 20/5
the constraints Zj – cj –4M – 1 –5M – 4 0 0 M 0 –20M –
3x1 + x2 £ 3
2x1 + 3x2 £ 6 R1¢ = R1 – (1/3)R2, R2¢ = (1/3)R2,
4x1 + 5x2 ≥ 20 R3¢ = R3-(5/3)R2
x1, x2 ≥ 0 Table 4.14
Solution Standard form of the problem is
I Iteration
Maximize Z = x1 + 4x2 + 0s1 + 0s2 + 0s3 – MR
0 s1 7/3 0 1 –1/3 0 0 1 3/7
subject to the constraints
3x1 + x2 + s1 = 3 4 x2 2/3 1 0 1/3 0 0 2 3

2x1 + 3x2 + s2 = 6 –M R 2/3 0 0 –5/3 –1 1 10 15

4x1 + 5x2 – s3 + R = 20 Zj–cj –(2M+5)/3 0 0 (5M+4)/3 M 0 –10M+8 –
x1, x2, s1, s2, s3, R ≥ 0 R1¢ = (3/7)R1, R2¢ = R2 – (2/7)R1,
The starting table is R3¢ = R3 – (2/7)R1
 46 Operations Research

Table 4.15 Example 4.5 Solve using, two-phase simplex

method
II Iteration Maximize Z = 5x 1 + 8x 2 subject to the
Z x1 x2 s1 s2 s3 R Const Ratio
constraints
CB B 1 4 0 0 0 –M 0 3x1 + 2x2 ≥ 3
1 x1 1 0 3/7 –1/7 0 0 3/7 – x1 + 4x2 ≥ 4
4 x2 0 1 –2/7 3/7 0 0 12/7 – x1 + x2 £ 5
–M R 0 0 –2/7 –11/7 –1 1 68/7 –
x1, x2 ≥ 0
Solution Assigning zero costs to the slack and
Zj–cj 0 0 (2M–5)/7 (11M+11)/7 M 0 (51–68M)/7 –
surplus variables and –1 to the artificial variables
Optimality is reached. But the artificial variable in the objective function we get the standard form
R is present in the optimal basis with a value 68/7. Maximize Z = 5x1 + 8x2 + 0s1 + 0s2 +0s3 – R1 – R2
Hence the problem has no feasible solution and subject to
hence no solution. 3x1 + 2x2 – s1 + R1 = 3
x1 + 4x2 – s2 + R2 = 4
4.3 TWO-PHASE METHOD x1 + x2 + s3 = 5
The two-phase method is another method to solve x1, x2, s1, s2, s3, R1, R2 ≥ 0
LP problems involving artificial variables. This Phase I Consider the auxiliary LPP with the
method consists of two phases. In the first phase objective function W = –R1 – R2 subject to the
we consider an auxiliary LPP with the objective given constraints. The starting table is
function.
W = 0x1 + 0x2 + … + 0xn – R1 – R2… Table 4.16
where R1, R2,… are all artificial variables. We
W x1 x2 s1 s2 s3 R1 R2 Const Ratio
maximize W using simplex method under the given
constraints. Two cases arise. CB X B 0 0 0 0 0 –1 –1 0
Case (I) Max W < 0 and at least one artificial –1 R1 3 2 –1 0 0 1 0 3 3/2
variable appears in the optimal basis with a posi- –1 R2 1 4 0 –1 0 0 1 4 4/4
tive value. In this case the given LPP does not 0 S3 1 1 0 0 1 0 0 5 5/1
possess any feasible solution. cj –4 –6 1 1 0 0 0 –7
Case (II) Max W = 0 and no artificial variable
appears in the optimal basis or Max W = 0 and R1¢ = R1 – (1/2)R2
some artificial variable appears at zero level in R2¢ = (1/4)R2
the optimal basis. In this case the problem has a R3¢ = R3 – (1/4)R2
solution and we may proceed to phase II.
Table 4.17
In phase II we use the optimal solution of phase
I as the starting solution. Assign the actual costs I Iteration
to the objective variables. If there is an artificial
–1 R1 5/2 0 –1 1/2 0 1 –1/2 1 2/5
variable at zero level, assign zero cost to that
artificial variable. Apply simplex method and 0 x2 1/4 1 0 –1/4 0 0 1/4 1 4
obtain the optimal solution. We may omit all the 0 s3 3/4 0 0 1/4 1 0 –1/4 4 16/3
artificial variables from the table which are non- c j –5/2 0 1 –1/2 0 0 3/2 –1
basic at the end of phase I.
 Artificial Variables 47

R1¢ = (2/5) R1 Table 4.21

R2¢ = R2 – (1/10)R1 II Iteration
R3¢ = R3 –(3/10)R1
0 s2 3 0 0 1 4 16
8 x2 1 1 0 0 1 5
Table 4.18 0 s1 –1 0 1 0 2 7
II Iteration Zj – cj 3 0 0 0 8 40 Optimal

W x1 x2 s1 s2 s3 R1 R2 Const Ratio The solution is x1 = 0, x2 = 5, Z* = 40.

CB XB 0 0 0 0 0 –1 –1 0 Example 4.6 Solve using the two-phase method
0 X1 1 0 –2/5 1/5 0 2/5 1/5 2/5 Minimize Z = 5x1 – 6x2 – 7x3 subject to the
0 X2 0 1 1/10 –3/10 0 –1/10 3/10 9/10
constraints
x1 + 5x2 – 3x3 ≥ 15
0 s3 0 0 3/10 1/10 1 –3/10 –1/10 37/10
5x1-6x2+10x3 £ 20
cj 0 0 0 0 0 1 1 0 x1 + x2 + x3 = 5
Solution The standard form of the problem is
Optimality is reached. There is no artificial
Maximize Y = (–Z) = –5x1 + 6x2 + 7x3 + 0s1 +
variable in the optimal basis. We proceed to phase
0s2 – R1 –R2 subject to the constraints
II, omitting R1 and R2 from the table.
x1 + 5x1 – 3x3 – s1 + R1 = 15
Phase II Maximize Z = 5x 1 + 8x 2 + 0s 1 + 5x1 – 6x2 + 10x3 + s2 = 20
0s2 + 0s3 x1 + x2 + x3 + R2 = 5
The starting table is x1, x2, x3, s1, s2, R1, R2 ≥ 0

Table 4.19 Phase I Consider the auxiliary LPP with the

objective function W = –R1 – R2 subject to the given
Z x1 x2 s1 s2 s3 Const Ratio constraints. The starting table is
CB B 5 8 0 0 0 0
Table 4.22
5 x1 1 0 –2/5 1/5 0 2/5 2
W x1 x2 x3 s1 s2 R1 R2 Const Ratio
8 x2 0 1 1/10 –3/10 0 9/10 –
CB B 0 0 0 0 0 –1 –1 0
0 s3 0 0 3/10 1/10 1 37/10 37
–1 R1 1 5 –3 –1 0 1 0 15 15/5
Zj – cj 0 0 –6/5 –7/5 0 46/5
0 S2 5 –6 10 0 1 0 0 20 –
R1¢ = 5R1, R2¢ = R2 + (3/2)R1, –1 R2 1 1 1 0 0 0 1 5 5/1
R3¢ = R3 – (1/2)R1 cj –2 –6 2 1 0 0 0 –20

Table 4.20 R1¢ = (1/5)R1, R2¢ = R2 + (6/5)R1,

R3¢ = R3 – (1/5)R1
I Iteration
Table 4.23
0 S2 5 0 –2 1 0 2 –
I Iteration
8 x2 3/2 1 –1/2 0 0 3/2 –
0 s3 –1/2 0 1/2 0 1 7/2 7 0 x2 1/5 1 –3/5 –1/5 0 1/5 0 3 –
Zj – cj 7 0 –4 0 0 12 0 S2 31/5 0 32/5 –6/5 1 6/5 0 38 190/32
–1 R2 4/5 0 8/5 1/5 0 –1/5 1 2 10/8
R1¢ = R1 + 4R3, R2¢ = R2 + R3, R3¢ = 2R3 c j –4/5 0 –8/5 –1/5 0 1/5 0 –2
 48 Operations Research

R1¢ = R1+(3/8)R3, R2¢ = R2 – 4R3, Solution Introducing surplus variables with zero
R3¢ = (5/8)R3 costs and artificial variables with costs –1 in the
objective function we obtain the standard form of
Table 4.24 the problem as
II Iteration Maximize Z = – 4x1 – 3x2 – 9x3 + 0s1 + 0s2 – R1 –
R2 subject to the constraints
W x1 x2 x3 s1 s2 R1 R2 Const Ratio 2x1 + 4x2 + 6x3 – s1 + R1 = 15
CB B 0 0 0 0 0 –1 –1 0 6x1 + x2 + 6x3 – s2 + R2 = 12
0 x2 1/2 1 0 –1/8 0 1/8 3/8 15/4 x1, x2, x3, s1, s2, R1, R2 ≥ 0
0 s2 3 0 0 –2 1 2 –4 30
Phase I The objective function is W = –R1 – R2
0 x3 1/2 0 1 1/8 0 –1/8 5/8 5/4
with the given constraints. The starting table is
cj 0 0 0 0 0 1 1 0
Table 4.26
Optimality is reached. There is no artificial
variable in the optimal basis. Hence we proceed W x1 x2 x3 s1 s2 R1 R2 Const Ratio
to phase II omitting R1and R2 from the table.
CB B 0 0 0 0 0 –1 –1 0
Phase II Maximize Y = –5x1 + 6x 2 + 7x 3 + –1 R1 2 4 6 –1 0 1 0 15 15/6
0s1 + 0s2 –1 R2 6 1 6 0 –1 0 1 12 12/6
The starting table is
cj –8 –5 –12 1 1 0 0 –27
Table 4.25
R1¢ = R1 – R2, R2¢ = (1/6)R2
Y x1 x2 x3 s1 s2 Const Ratio

CB B –5 6 7 0 0 0 –
Table 4.27
6 x2 1/2 1 0 –1/8 0 15/4 – I Iteration
0 s2 3 0 0 –2 1 30 – –1 R1 –4 3 0 –1 1 1 –1 3 3/3
7 x3 1/2 0 1 1/8 0 5/4 – 0 x3 1 1/6 1 0 –1/6 0 1/6 2 12
cj 23/2 0 0 1/8 0 125/4 – cj 4 –3 0 1 –1 0 1 –3
Optimality is reached.
R1¢ = (1/3)R1, R2¢ = R2 – (1/18)R1
Solution
x1 = 0, x2 = 15/4, x3 = 5/4, Table 4.28
Z* = –Y* = –125/4
II Iteration
Example 4.7 (Solve using the two-phase
method) 0 x2 –4/3 1 0 –1/3 1/3 1/3 –1/3 1
Maximize Z = – 4x1 – 3x2 – 9x3 subject to the 0 x3 11/9 0 1 1/18 –2/9 –1/18 2/9 11/6
constraints cj 0 0 0 0 0 1 1 0
2x1 + 4x2 + 6x3 ≥ 15
The optimality is reached. No artificial variable
6x1 + x2 + 6x3 ≥ 12
appears in the optimal basis. Hence we proceed to
x1, x2, x3 ≥ 0 phase II
 Artificial Variables 49

Phase II Z = –4x1 – 3x2 – 9x3 + 0s1 + 0s2 Solution The standard form of the LPP is
The starting table is Maximize Z = 5x1 + 3x2 + 0s1 + 0s2 – R1 subject
to the constraints
Table 4.29
2x1 + x2 + s1 = 1
Z x1 x2 x3 s1 s2 Const Ratio
x1 + 4x2 – s2 + R1 = 6
CB B – 4 –3 –9 0 0 0 x1, x2, s1, s2, R1 ≥ 0
–3 x2 –4/3 1 0 –1/3 1/3 1 – The objective function of the auxiliary LPP is
–9 x3 11/9 0 1 1/18 –2/9 11/6 9/6
W = –R1
cj –3 0 0 1/2 1 –39/2
The starting table is
R1¢ = R1 + (12/11)R2, R2¢ = (9/11)R2
Table 4.31
Table 4.30
W x1 x2 s1 s2 R1 Const Ratio
I Iteration
CB B 0 0 0 0 –1 0
Z x1 x2 x3 s1 s2 Const Ratio
0 s1 2 1 1 0 0 1 1/1
CB B –4 –3 –9 0 0 0 –1 R1 1 4 0 –1 1 6 6/4
–3 x2 0 1 12/11 –3/1 1/11 3 cj –1 –4 0 1 0 –6
–4 x1 1 0 9/11 1/22 –2/11 3/2
R1¢ = R1, R2¢ = R2 – 4R1
cj 0 0 27/11 7/11 5/11 –15

The optimality is reached. Solution of the given Table 4.32

LPP is I Iteration
x1 = 3/2, x2 = 3, x3 = 0, Z* = –15
0 x2 2 1 1 0 0 1
Example 4.8 (Solve using the two-phase
method) –1 R1 –7 0 –4 –1 1 2
Maximize Z = 5x 1 + 3x 2 subject to the cj 7 0 4 1 0 –2
constraints
2x1 + x2 £ 1 All Zj – Cj ≥ 0. The optimality is reached. But
x1 + 4x2 ≥ 6 the optimal basis has an artificial variable R1 with
x1, x2 ≥ 0 a value 2. Hence the given LPP has no solution.

S H O RT A N S W E R Q U E S T I O N S
■ ■ ■ ■ ■ ■ ■

1. Define surplus variable. Fill in the blanks

2. Explain: artificial variable, big-M method,
two-phase method.
3. Explain how the two-phase method helps to
find out the feasibility of an LPP.
4. Explain the advantages of the two-phase
method over the big-M method.
5. An artificial variable is introduced in an
equality constraint to obtain a ___________.
(Ans: basic solution)
6. Artificial variable can be dropped from the
simplex iteration table, when it becomes
_________. (Ans: non-basic)
 50 Operations Research

7. An LPP has no solution if an artificial 8. The number of basic solutions to an LPP is

variable appears in the optimal basis with __________. (Ans: finite)
____________. (Ans: positive value)

EXERCISES
■ ■ ■ ■ ■ ■ ■

Solve using big-M method x1 + 1/2x2 + 1/2x3 £ 1

1. Minimize Z = 5x1 – 6x2 – 7x3 (3/2)x1 + 2x2 + x3 ≥ 8
subject to the constraints x1, x2, x3 ≥ 0
x1 + 5x2 – 3x3 ≥ 15 8. Maximize Z = 5x1 – 2x2 + 3x3 subject to the
5x1 – 6x2 + 10x3 £ 20 constraints
x1 + x2 + x3 = 5 2x1 + 2x2 – x3 ≥ 2
x1, x2, x3 ≥ 0 3x1 – 4x2 £ 3
2. Maximize Z = 3x 1 + 2x 2 subject to the x2 + 3x3 £ 5
constraints x1, x2, x3 ≥ 0
2x1 + x2 £ 2 Solve using the two-phase method.
3x1 + 4x2 ≥ 12 9. Minimize Z = 12x1 + 20x2 subject to the
x1, x2 ≥ 0 constraints
3. Minimize Z = 4x 1 + 3x 2 subject to the 6x1 + 8x2 ≥ 100
constraints 7x1 + 12x2 ≥ 120
2x1 + x2 ≥ 10 x1, x2 ≥ 0
–3x1 + 2x2 £ 6 10. Minimize Z = 2x1 + 4x2 + x3 subject to the
x1 + x2 ≥ 6 constraints
x1, x2 ≥ 0 x1 + 2x2 – x3 £ 5
4. Maximize Z = 2x1 + x2 + x3 subject to the 2x1 – x2 + 2x3 = 2
constraints –x1 + 2x2 + 2x3 ≥ 1
4x1 + 6x2 + 3x3 £ 8 x1, x2, x3 ≥ 0
3x1 – 6x2 – 4x3 £ 1 11. Minimize Z = x1 + x2 + 2x3 subject to the
2x1 + 3x2 – 5x3 ≥ 4 constraints
x1, x2, x3 ≥ 0 x1 + x2 + x 3 £ 9
5. Minimize Z = 2x 1 + 3x 2 subject to the 2x1 – 3x2 + 3x3 = 1
constraints –3x1 + 6x2 – 4x3 = 3
x1 + x2 ≥ 5 x1, x2, x3 ≥ 0
x1 + 2x2 ≥ 6
Solve the following LPP
x1, x2 ≥ 0
12. Minimize Z = 2x1 + x2 – x3 – x4
6. Maximize Z = 3x 1 + 2x 2 subject to the
subject to the constraints
constraints
x1 – x2 + 2x3 – x4 = 2
2x1 + x2 £ 1
2x1 + x2 – 3x3 + x4 = 6
3x1 + 4x2 ≥ 4
x1 + x2 + x3 + x4 = 7
x1, x2 ≥ 0
x1, x2, x3, x4 ≥ 0
7. Minimize Z = x1 + 2x2 + x3 subject to the
13. Minimize Z = 3x 1 + 2x 2 subject to the
constraints
constraints
 Artificial Variables 51

7x1 + 2x2 ≥ 14 3x1 + x2 ≥ 27

–2x1 – x2 £ – 6 –x1 – x2 £ 21
x1 + x2 ≥ 5 x1 + 2x2 ≥ 30
x1, x2 ≥ 0 x1, x2 ≥ 0
14. Minimize Z = 4x 1 + x 2 subject to the 22. Maximize Z = 2x 1 + x 2 subject to the
constraints constraints
3x1 + x2 = 3 3x1 + x2 = 3
4x1 + 3x2 ≥ 6 4x1 + 3x2 ≥ 6
x1 + 2x2 £ 3 x1 + 2x2 ≥ 3
i1, x2 ≥ 0 x1, x2 ≥ 0
15. Maximize Z = 4x1 + 5x2 + 2x3 subject to the 23. Minimize Z = 4x 1 + 2x 2 subject to the
constraints constraints
2x1 + x2 + x3 £ 10 3x1 + x2 ≥ 27
x1 + 3x2 + x3 = 12 –x1 – x2 £ 21
x1 + x2 + x3 = 6 x1 + 2x2 ≥ 30
x1, x2, x3 ≥ 0 x1, x2 ≥ 0
16. Maximize Z = 3x1 + x2 + 2x3 subject to the 24. Food X contains 6 units of vitamin A per gram
constraints and 7 units of vitamin B per gram and costs
12 paise per gram. Food Y contains
4x1 + x2 + 2x3 = 9
8 units of vitamin A per gram and 12 units of
8x1 + x2 – 4x3 £ 10
vitamin B per gram and costs 20 paise per
x1, x2, x3 ≥ 0
gram. The daily minimum requirements of
17. Minimize Z = 2x1 + 9x2 + x3 subject to the vitamin A and vitamin B are 100 units and
constraints 120 units respectively. Find the minimum
x1 + 4x2 + 2x3 ≥ 5 cost of product mix by simplex method.
3x1 + x2 + 2x3 ≥ 4 25. A company has two factories and each
x1, x2, x3 ≥ 0 produces three products A, B, C from
18. Maximize Z = 5x1 – 2x2 + 3x3 subject to the common raw material. Production per hour,
constraints costs per hour, and orders received for the
2x1 + 2x2 – x3 ≥ 2 products are given below.
3x1-4x2 £ 3
x2 + 3x3 £ 5 Product Operating Cost
x1, x2, x3 ≥ 0 Per hour (Rs)
X Y Z
19. Maximize Z = 6x 1 + 4x 2 subject to the
constraints Plant I 2 4 3 9
Plant II 4 3 2 10
2x1 + 3x2 £ 30
Order 50 24 60
3x1 + 2x2 £ 24 received
x1 + x2 ≥ 3
x1, x2 ≥ 0 Find the number of production hours required
20. Maximize Z = 2x1 + x2 + 3x3 subject to the to fulfill the orders at minimum cost.
constraints 26. Maximize Z = 8x2 subject to the constraints
x1 + x2 + 2x3 £ 5 x1 – x2 ≥ 0
2x1 + 3x2 + 4x3 = 12 2x1 + 3x2 £ –6
x1, x2, x3 ≥ 0 x1, x2 unrestricted
21. Maximize Z = 4x 1 + 2x 2 subject to the 27. Minimize Z = 3x 1 + 2x 2 subject to the
constraints constraints
 52 Operations Research

3x1 + x2 ≥ 12 4x1 + x2 + x3 £ 6
x1 + x2 ≥ 8 x1, x2 , x3 ≥ 0
x1 + 2x2 ≥ 11 35. Maximize Z = 2x1 + 3x2 + 5x3 subject to the
x1, x2 ≥ 0 constraints
28. Maximize Z = 2x1 + 3x2 – 5x3 subject to the 3x1 + 10x2 + 5x3 £ 15
constraints 33x1 – 10x2 + 9x3 £ 33
x1 + x2 + x3 = 7 x1 + 2x2 + x3 ≥ 4
2x1 – 5x2 + x3 ≥ 10 x1, x2 , x3 ≥ 0
x1, x2, x3 ≥ 0 Solve using the two-phase method
29. Minimize Z = 3x 1 + 2x 2 subject to the 36. Maximize Z = 3x1 + 2x2 + 2x3 subject to the
constraints constraints
x1 + x2 ≥ 2 5x1 + 7x2 + 4x3 £ 7
x1 + 3x2 £ 3 –4x1 + 7x2 + 5x3 ≥ –2
3x1 + 4x2 – 6x3 ≥ 29/7
x1 – x2 = 1
x1, x2 , x3 ≥ 0
x1, x2 ≥ 0
37. Maximize Z = 5x 1 + 3x 2 subject to the
30. Maximize Z = 2x 1 + x 2 subject to the
constraints
constraints
x1 + x2 £ 6
3x1 + x2 = 3
2x1 + 3x2 ≥ 3
4x1 + 3x2 ≥ 6
x1 ≥ 3
x1 + 2x2 £ 3 x2 ≥ 3
x1, x2 ≥ 0 x1, x2 ≥ 0
31. Maximize Z = 5x 1 + 8x 2 subject to the 38. Minimize Z = 20x1 + 10x2 subject to the
constraints constraints
3x1 + 2x2 ≥ 3 x1 + 2x2 £ 40
x1 + 4x2 ≥ 4 3x1 + x2 ≥ 30
x1 + x2 £ 5 4x1 + 3x2 ≥ 60
x1, x2 ≥ 0 x1, x2 ≥ 0
32. Minimize Z = –2x 1 – x 2 subject to the 39. Maximize Z = 2x 1 + 3x 2 subject to the
constraints constraints
x1 + x2 ≥ 2 x1 + x2 £ 30
x1 + x2 £ 4 x1 – x2 ≥ 0
x1, x2 ≥ 0 x1 £ 20, x2 ≥ 3, x2 £ 12
33. Minimize Z = 12x1 + 20x2 subject to the x1, x2 ≥ 0
constraints 40. Maximize Z = –3x1 + 6x2 subject to the
6x1 + 8x2 ≥ 100 constraints
7x1 + 12x2 ≥ 120 x1 + 2x2 + 1 ≥ 0
x1, x2 ≥ 0 –4x1 + x2 + 23 ≥ 0
34. Maximize Z = x1 + 2x2 + x3 subject to the 2x1 + x2 – 4 ≥ 0
constraints x1 – 4x2 + 13 ≥ 0
2x1 + x2 – x3 £ 2 x1 – x2 + 1 ≥ 0
–2x1 + x2 – 5x3 ≥ –6 x1, x2 ≥ 0
 Artificial Variables 53

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. x1 = 0, x2 = 15/4, x3 = 5/4, Z* = –125/4 x2 = 42/5, Z* = 48. Since there are two optimal
2. No solution solutions it means that there are infinite number
3. x1 = 4, x2 = 2, Z* = 22 of solutions.
4. x1 = 9/7, x2 = 10/21, x3 = 0, Z* = 64/21 20. x1 = 3, x2 = 2, x3 = 0, Z* = 8
5. x1 = 4, x2 = 1, Z* = 11 21. Unbounded solution
6. x1 = 0, x2 = 1, Z* = 2 22. x1 = 3/5, x2 = 3, Z* = 21/5
7. No solution 23. x1 = 24/5, x2 = 63/5, Z* = 222/5
85 24. x1 = 15, x2 = 5/4, Z* = 205
8. x1 = 23/3, x2 = 5, x3 = 0, Z* = 25. x1 = 35/2, x2 = 15/4, Z* = 195
3
9. x1 = 15, x2 = 5/4, Z* = 205 26. x1 = –6/5, x2 = –6/5, Z* = –48/5
10. x1 = 0, x2 = 0, x3 = 1, Z* = 1 27. x1 = 2, x2 = 6, Z* = 18
11. x1 = 0, x2 = 13/6, x3 = 5/2, Z* = 43/6 28. x1 = 45/7, x2 = 4/7, x3 = 0, Z* = 102/7
12. x1 = 3, x2 = 0, x3 = 1, x4 = 3, Z* = 2 29. x1 = 3/2, x2 = 1/2, Z* = 11/2
13. x1 = 1, x2 = 4, Z* = 11 30. x1 = 3/5, x2 = 6/5, Z* = 12/5
14. x1 = 3/5, x2 = 6/5, Z* = 18/5 31. x1 = 0, x2 = 5, Z* = 40
15. x1 = 3, x2 = 3, Z* = 27 32. x1 = 4, x2 = 0, Z* = –8
16. x1 = 0, x2 = 0, x3 = 3/2, Z* = 3 33. x1 = 15, x2 = 5/4, Z* = 205
17. x1 = 0, x2 = 0, x3 = 5/2, Z* = 5/2 34. x1 = 0, x2 = 4, x3 = 2, Z* = 10
18. x1 = 23/3, x2 = 5, x3 = 0, Z* = 85/3 35. No solution
19. x1 = 8, x2 = 0, Z* = 48 36. x1 = 1, x2 = 2/7, x3 = 0, Z* = 25/7
In exercise No. 19, we find Z j – C j = 0 37. x1 = 3, x2 = 3, Z* = 24
corresponding to the non-basic variable x2 in the 38. x1 = 6, x2 = 12, Z* = 240
optimal table. Hence we can bring x2 into the basis 39. x1 = 18, x2 = 12, Z* = 72
and obtain another optimal solution x1 = 12/5, 40. x1 = 3, x2 = 4, Z* = 15

Dual Simplex Method
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
5.1 DUAL SIMPLEX METHOD
The simplex method starts with an initial basic
feasible solution and moves towards optimality.
If the constants bi are all non-negative (bi ≥ 0) the
solution is feasible. Simplex method is terminated
when an optimal solution is reached (Zj – cj ≥ 0
for all j).
In certain cases some constants bi may be
negative (infeasibility), but at the same time
optimality conditions are satisfied (Zj – cj ≥ 0 for
all j) in the starting table. In other words we start
with an infeasible but optimal solution. Here
regular simplex method cannot be used. Instead
we apply dual simplex method and work towards
feasibility. Finally we arrive at an optimal and
feasible solution [Zj – cj ≥ 0 and bi ≥ 0 " i, j].
5.1.1 Dual Simplex Algorithm
Step 1 Rewrite the LPP in the standard form. Con-
vert ≥ type constraints to £ type by multiplying both
sides by –1. Introduce slack variables, form the sim-
plex table and obtain the initial basic solution.
Step 2 If all the constants are non-negative and
Zj – cj ≥ 0 " j the solution is optimal. Stop.
5
Step 3 If all the constants are non-negative and
some Zj – cj < 0 then apply simplex method for
optimization.
Step 4 If bi < 0 for some i and Zj – cj < 0 for
some j then there exists no solution. Stop.
Step 5 If bi < 0 for some i and Zj – cj ≥ 0 " j then
choose the most negative bi, say br. The corre-
sponding row is the pivot row and the correspond-
ing basic variable leaves the basis.
Step 6 Examine the elements of the pivot row.
If all the arj ≥ 0 there exists no solution. Stop.
Step 7 If some a rj < 0 then find the ratios
| (Zj – cj)/arj | where arj < 0 and select the minimum
among them, say | Zs – cs /ars |, then the sth column
is the pivot column and the corresponding non-
basic variable enters the basis. ars is called the pivot
element.
Step 8 Perform row operations, making the ele-
ment ars as 1 and the other elements of the pivot
column, zero. We get the improved solution.
Repeat steps 5 to 8 till optimality is reached
[i.e. bi > 0 " i and Zj – cj ≥ 0 " j].
 Dual Simplex Method 55

Example 5.1 Apply the dual simplex method to R1 – (1/3) R2, R2/(–3), R3 – (2/3) R2
solve:
Minimize Z = 2x1 + x2 subject to the constraints Table 5.2
3x1 + x2 ≥ 3 I Iteration
4x1 + 3x2 ≥ 6
x1 + 2x2 ≥ 3 Z x1 x2 s1 s2 s3
x1, x2 ≥ 0 XB
CB B –2 –1 0 0 0
Solution Converting the objective function to 0 s1 –5/3 0 1 –1/3 0 –1
maximization type and the constraints to £ type –1 x2 4/3 1 0 –1/3 0 2
we get
0 s3 5/3 0 0 –2/3 1 1
Maximize W = – Z = – 2x1 – x2 subject to the
constraints Zj–cj 2/3 0 0 1/3 0 –2
–3x1 – x2 £ –3
s1 leaves the basis.
– 4x1 – 3x2 £ –6
Min {|(2/3) / (–5/3)|, | (1/3)/(–1/3) |}
–x1 – 2x2 £ –3
= Min {2/5,1} = 2/5
x1, x2 ≥ 0
Hence x1 enters the basis. The pivot element
The standard form is
is –5/3.
Maximize W = –2x1 – x2 + 0s1 + 0s2 + 0s3
The necessary row transformations are
subject to the constraints
R1(–3/5), R2 + (4/5)R1, R3 + R1
–3x1 – x2 + s1 = –3
The resulting table is
–4x1 – 3x2 + s2 = – 6
–x1 – 2x2 + s3 = –3 Table 5.3
x1, x2, s1, s2, s3 ≥ 0
The initial simplex table is Z x1 x2 s1 s2 s3
XB
CB B –2 –1 0 0 0
Table 5.1
–2 x1 1 0 –3/5 1/5 0 3/5
Z x1 x2 s1 s2 s3 –1 x2 0 1 4/5 –3/5 0 6/5
XB
CB B –2 –1 0 0 0 0 s3 0 0 1 –1 1 0
0 s1 –3 –1 1 0 0 –3 Zj–cj 0 0 2/5 1/5 0 –12/5
0 s2 –4 –3 0 1 0 –6
0 s3 –1 –2 0 0 1 –3 bi ≥ 0 " i, Zj – cj ≥ 0 " j. Hence optimality is
Zj – cj 2 1 0 0 0 0 reached. The solution is x 1 = 3/5, x 2 = 6/5,
Z* = 12/5.
Note: We use the notation XB to represent the Example 5.2 Solve using the dual simplex
values of the current basic variables. method
All Zj – cj ≥ 0 and all the bi < 0 the solution is Maximize Z = –3x 1 – 2x 2 subject to the
not optimal. The most negative bi is –6. Hence the constraints
corresponding basic variable s2 leaves the basis. x1 + x2 ≥ 1
ìï Z j - c j üï x1 + x2 £ 7
Min í : a2 j < 0 ý x1 + 2x2 ≥ 10
ïî a2 j ïþ
x2 £ 3
= min {|2/(–4)| , |1/(–3)|}
= min {1/2, 1/3} = 1/3 x1, x2 ≥ 0
Hence the corresponding variable x2 enters the Solution Rewrite the problem as
basis. –3 is the pivot element. Make the row Maximize Z = –3x1 – 2x2 subject to the cons-
transformations traints
 56 Operations Research

–x1 – x2 £ –1 Table 5.6

x1 + x2 £ 7
Z x1 x2 s1 s2 s3 s4
–x1–2x2 £ –10 XB
x2 £ 3 CB B –3 –2 0 0 0 0
x1, x2 ≥ 0 0 s1 0 0 1 0 –1 –1 6
The standard form is 0 s2 0 0 0 1 1 1 0
Maximize Z = –3x 1 – 2x 2 subject to the –2 x2 0 1 0 0 0 1 3
constraints –3 x1 1 0 0 0 –1 –2 4
–x1 – x2 + s1 = – 1
Zj–cj 0 0 0 0 3 4 –18
x1 + x2 + s2 = 7
–x1 – 2x2 + s3 = –10 Optimality reached. The optimal solution is
x2 + s4 = 3 x1 = 4, x2 = 3, Z* = – 18.
x1, x2, s1, s2, s3, s4 ≥ 0
The starting table is Example 5.3 Using the dual simplex method
solve:
Table 5.4 Minimize Z = x1 + 2x2 + 3x3 subject to the
constraints
Z x1 x2 s1 s2 s3 s4
XB x1 – x2 + x3 ≥ 4
CB B –3 –2 0 0 0 0 x1 + x2 + 2x3 £ 8
0 s1 –1 –1 1 0 0 0 –1 x2 – x3 ≥ 2, x1, x2, x3 ≥ 0
0 s2 1 1 0 1 0 0 7
Solution The standard form of the problem is
0 s3 –1 –2 0 0 1 0 –10
Maximize W = –Z = –x1 – 2x2 – 3x3 subject to
0 s4 0 1 0 0 0 1 3
the constraints
Zj–cj 3 2 0 0 0 0 0 –x1 + x2 – x3 + s1 = –4
s3 leaves the basis x1 + x2 + 2x3 + s2 = 8
Min {|3/(–1)|, |2/(–2)|} = Min {3, 1} = 1. –x2 + x3 + s3 = –2
The corresponding variable x2 is the entering x1, x2, x3, s1, s2, s3 ≥ 0
variable and the pivot element is –2. The necessary The initial table is
row transformations are R1 – (1/2)R3, R2 + (1/2)R3,
Table 5.7
R3/(–2), R4 + (1/2)R3.
We get the revised table W x1 x2 x3 s1 s2 s3
XB
CB B –1 –2 –3 0 0 0
Table 5.5
0 s1 –1 1 –1 1 0 0 –4
Z x1 x2 s1 s2 s3 s4 0 s2 1 1 2 0 1 0 8
XB
CB B –3 –2 0 0 0 0 0 s3 0 –1 1 0 0 1 –2
0 s1 –1/2 0 1 0 –1/2 0 4
cj 1 2 3 0 0 0 0
0 s2 1/2 0 0 1 1/2 0 2
–2 x2 1/2 1 0 0 –1/2 0 5 R1(–1), R2 + R1
0 s4 –1/2 0 0 0 1/2 1 –2
–1 x1 1 –1 1 –1 0 0 4
Zj – cj 2 0 0 0 1 0 –10 0 s2 0 2 1 1 1 0 4
s4 leaves and x1 enters the basis. The pivot 0 s3 0 –1 1 0 0 1 –2
element is –1/2. Performing the transformations cj 0 3 2 1 0 0 –4
R1 – R4, R2 + R4, R3 + R4, R4(–2), we get the new
table R1 – R3, R2 + 2R3, R3(–1)
 Dual Simplex Method 57

W x1 x2 x3 s1 s2 s3 XB –2x1 – x2 + s1 = –2
–1 x1 1 0 0 –1 0 –1 6 x1 + x2 + s2 = –1
0 s2 0 0 3 1 1 2 0 x1, x2, s1, s2 ≥ 0
–2 x2 0 1 –1 0 0 –1 2 The starting simplex table is

cj 0 0 5 1 0 3 –10 Table 5.8

W x1 x2 s1 s2
Optimality is reached. The optimal solution is XB
x1 = 6, x2 = 2, x3 = 0, W* = – 10, Z* = 10 CB B –1 –1 0 0
0 s1 –2 –1 1 0 –2
Example 5.4 Solve using the dual simplex 0 s2 1 1 0 1 –1
method
cj 1 1 0 0 0
Minimize Z = x 1 + x 2 subject to the con-
straints x1 enters and s1 leaves R1/(–2), R2 + (1/2)R1
2x1 + x2 ≥ 2
–x1 – x2 ≥ 1 –1 x1 1 1/2 –1/2 0 1
x1, x2 ≥ 0 0 s2 0 1/2 1/2 1 –2
Solution Rewriting the problem, we get cj 0 1/2 1/2 0 –1
Maximize W = –Z = –x1 – x2 subject to the cons-
traints We see that the pivot row corresponds to the
–2x1 – x2 £ –2 basic variable s2. Hence s2 leaves the basis. In order
x1 + x2 £ –1 to determine the entering variable we need to take
x1, x2 ≥ 0 the ratio of the c j ’s with the negative elements of
The standard form is the pivot row. Since there is no negative element
Maximize W = –x1 – x2 + 0s1 + 0s2 in the pivot row we cannot find the ratios. Hence
subject to the constraints there is no feasible solution to the given problem.

S H O RT A N S W E R Q U E S T I O N S
■ ■ ■ ■ ■ ■ ■

1. What is the difference between the regular 2. State the optimality condition in the dual
simplex method and the dual simplex method. simplex method.

EXERCISES
■ ■ ■ ■ ■ ■ ■

Applying the dual simplex method solve: x1 + x2 ≥ 1

1. Minimize Z = 2x1 + 2x2 + 4x3 2x1 + 3x2 ≥ 2
subject to the constraints x1, x2 ≥ 0
2x1 + 3x2 + 5x3 ≥ 2 3. Minimize Z = 20x1 + 16x2 subject to the
constraints
3x1 + x2 + 7x3 £ 3 x1 + x2 ≥ 12
x1 + 4x2 + 6x3 £ 5 2x1 + x2 ≥ 17
x1, x2, x3 ≥ 0 x1 ≥ 5/2
2. Maximize Z = –3x 1 – x 2 subject to the x2 ≥ 6
constraints x1, x2 ≥ 0
 58 Operations Research

4. Minimize Z = 10x1 + 6x2 + 2x3 subject to the 3x1 + x2 + 7x3 £ 3

constraints x1 + 4x2 + 6x3 £ 5
–x1 + x2 + x3 ≥ 1 x1, x2 , x3 ≥ 0
3x1 + x2 – x3 ≥ 2 11. Minimize Z = 2x 1 + x 2 subject to the
x1, x2 ≥ 0 constraints
5. Maximize Z = – 4x1 – 6x2 – 18x3 subject to x1 + x2 ≥ 2
the constraints x1 + x2 £ 4
x1 + 3x3 ≥ 3 x1, x2 ≥ 0
x2 + 2x3 ≥ 5 12. Minimize Z = 12x1 + 20x2 subject to the
x1, x2, x3 ≥ 0 constraints
6. Maximize Z = –2x 1 – x 3 subject to the 6x1 + 8x2 ≥ 100
constraints 7x1 + 12x2 ≥ 120
x1 + x2 – x3 ≥ 5 x1, x2 ≥ 0
x1 – 2x2 + 4x3 ≥ 8 13. Maximize Z = –2x1 – 3x2 subject to the
x1, x2, x3 ≥ 0 constraints
7. Minimize Z = x 1 + 2x 2 subject to the x1 + x2 ≥ 5
constraints x1 + 2x2 ≥ 6
2x1 + x2 ≥ 4 x1, x2 ≥ 0
x1 + 2x2 £ 7 14. Minimize Z = 4x 1 + 2x 2 subject to the
x1, x2 ≥ 0 constraints
8. Minimize Z = 3x1 + 5x2 + 4x3 subject to the 3x1 + x2 ≥ 27
constraints x1 + x2 ≥ 21
–2x1 – x2 + 5x3 ≥ 2 x1 + 2x2 ≥ 30
3x1 + 2x2 + 4x3 ≥ 16 x1, x2 ≥ 0
x1, x2, x3 ≥ 0 15. Maximize Z = –5x1 – 4x2 – 3x3 subject to
9. Maximize Z = –3x 1 – x 2 subject to the the constraints
constraints x1 + x2 + x3 ≥ 100
x1 + x2 ≥ 1 x1 £ 20, x2 ≥ 30, x3 £ 40
2x1 + 3x2 ≥ 2 x1, x2 , x3 ≥ 0
x1, x2 ≥ 0 16. Minimize Z = 2x1 + 9x2 + x3 subject to the
10. Minimize Z = 2x1 + 2x2 + 4x3 subject to the constraints
constraints x1 + 4x2 + 2x3 ≥ 5
2x1 + 3x2 + 5x3 ≥ 2 3x1 + x2 + 2x3 ≥ 4
x1, x2 , x3 ≥ 0

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. x1 = 0, x2 = 2/3, x3 = 0 Z* = 4/3 9. x1 = 0, x2 = 1, Z* = –1
2. x1 = 0, x2 = 1, Z* = – 1 10. x1 = 0, x2 = 2/3, x3 = 0, Z* = 4/3
3. x1 = 5, x2 = 7, Z* = 212 11. x1 = 0, x2 = 2, Z* = 2
4. x1 = 1/4, x2 = 5/4, x3 = 0, Z* = 10 12. x1 = 15, x2 = 5/4, Z* = 205
5. x1 = 0, x2 = 3, x3 = 1, Z* = – 36 13. x1 = 4, x2 = 1, Z* = –11
6. x1 = 0, x2 = 14, x3 = 9, Z* = – 9 14. x1 = 3, x2 = 18, Z* = 48
7. x1 = 0, x2 = 2, Z* = 4 15. x1 = 0, x2 = 60, x3 = 40, Z* = –360
8. x1 = 0, x2 = 0, x3 = 4, Z* = 16 16. x1 = 0, x2 = 0, x3 = 5/2, Z* = 5/2
 6
Duality

CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■

Given any LPP we can associate another LPP with 6.1 RULES FOR WRITING THE
it. The given problem is called primal and the DUAL OF A PRIMAL
newly constructed problem is called dual. If the
primal has n variables and m constraints then the Primal Dual
dual has m variables and n constraints. The
primal and dual are so related that we can get (i) Maximization Z Æ Minimization W
the solution of the primal from the solution of (ii) Minimization Z Æ Maximization W
(iii) Coefficients of the ü
the dual. Also if the number of constraints in
objective function ý Æ Constants
the primal is greater than the number of variables þ
(iv) Constants Æ Coefficients of the
(m > n) then in the dual the number of constraints objective function
n will be less than the number of variables m. It (v) n variables Æ n constraints
is easier to solve the dual with less number of x1, x2, …, xn
constraints (manipulation work is reduced). Hence (vi) m constraints Æ m variables y 1 ,
we prefer to solve the dual and from the solution y2,…, ym
of the dual we can obtain the solution of the given ü
(vii) Matrix of constraint ýÆ AT
primal. coefficients A þ
In an LPP if the objective is maximization, (viii) ith variable ≥ 0 Æ ith constraint is of
the constraints are £ type and the variables ≥ 0 standard type
it is called the standard type of problem. If the ith variable £ 0 Æ ith constraint not of
objective is minimization then the problem is standard type
of standard type if the constraints are ≥ type ith variable Æ ith constraint is an
and the variables ≥ 0. We give below the rules unrestricted equality
for constructing the dual problem, given a (ix) jth constraint: Æ jth variable ≥ 0
primal. standard type
 60 Operations Research

jth constraint: not ü th x1 + 3x2 + 3x3 £ 16

of standard type ýþ Æ j variable £ 0 x1 £ 0, x2, x3 ≥ 0
jth constraint: Æ jth variable Solution The objective function of the dual is
equality unrestricted. Minimize W = 6y1 + 4y2 + 12y3 + 16y4
These rules can be easily understood from the The first variable x 1 £ 0. Hence the first
following examples. constraint of the dual is not of the standard type
Example 6.1 Write the dual of the LPP \ 2y1 + 3y2 + 5y3 + y4 £ 2
Maximize Z = x1 + 2x2 + 3x3 subject to the x2, x3 ≥ 0 Æ. The other constraints are of
constraints standard type
3x1 + x2 + x3 £ 12 \ y1 + 2y2 – 4y3 + 3y4 ≥ 4
x1 + 2x2 + 4x3 £ 20 y1 – y2 + 2y3 + 3y4 ≥ 5
2x1 + 5x2 – x3 £ 18 Again the first constraint is £ type (standard)
x1, x2, x3 ≥ 0 Hence, y1 ≥ 0
Solution The constants become objective But the second constraint is ≥ type (not
coefficients maximize Æ minimize. xi Æ yj. standard). Therefore the 2nd variable y2 £ 0. Since
Hence the objective of the dual is the other constraints are of standard type (£ ) y3,
Minimize W = 12y1 + 20y2 + 18y3 y4 ≥ 0.
Hence the dual problem is
æ 3 1 1ö æ 3 1 2ö Minimize W = 6y1 + 4y2 + 12y3 + 16y4 subject
A = ç 1 2 4 ÷ ® AT = ç 1 2 5÷ to the constraints
ç ÷ ç ÷ 2y1 + 3y2 + 5y3 + y4 £ 2
è 2 5 -1ø è 1 4 -1ø y1 + 2y2 – 4y3 + 3y4 ≥ 4
Objective coefficients 1, 2, 3 become constants. y1 – y2 + 2y3 + 3y4 ≥ 5
All the variables in the primal are ≥ 0 and hence y1 ≥ 0, y2 £ 0, y3, y4 ≥ 0
all the constraints are of standard type ≥ (dual is Note: In order to solve the dual we have to
minimization). Therefore the constraints are reduce it to the standard form.
3y1 + y2 + 2y3 ≥ 1 Example 6.3 Write the dual of the LPP
y1 + 2y2 + 5y3 ≥ 2 Minimize Z = 2x1 + 5x2 + x3 subject to the
y1 + 4y2 – y3 ≥ 3 constraints
Again all the constraints of the primal are of 9x1 – 3x2 + 5x3 ≥ 7
standard type £ (maximization problem) and hence 6x1 – 4x2 – 3x3 £ 2
all the variables of the dual are ≥ 0. y1, y2, y3 ≥ 0. x1, x2 ≥ 0
Thus the dual of the given primal is \ x3 unrestricted
Minimize W = 12y1 + 20y2 + 18y3 subject to Solution Minimize Æ Maximize
the constraints Constants Æ Objective coefficients
3y1 + y2 + 2y3 ≥ 1 The objective function of the dual is
y1 + 2y2 + 5y3 ≥ 2 Maximize W = 7y1 + 2y2
y1 + 4y2 – y3 ≥ 3 The first constraint is of ≥ type (standard)
y1, y2, y3 ≥ 0 \ y1 ≥ 0
The second constraint is £ type (not standard)
Example 6.2 Write the dual of \ y2 £ 0
Maximize Z = 2x1 + 4x2 + 5x3 subject to the x1, x2 ≥ 0 Æ the first two constraints of the dual
constraints are of the standard type (£ for maximization).
2x1 + x2 + x3 £ 6 Therefore the first two constraints are
3x1 + 2x2 – x3 ≥ 4 9y1 + 6y2 £ 2
5x1 – 4x2 + 2x3 £ 12 – 3y1 – 4y2 £ 5
 Duality 61

x3 is unrestricted Æ the third constraint of the y1 + 5y2 + 7y3 £ 5

dual is equality x3 unrestricted Æ third constraint is an equality
i.e. 5y1 – 3y2 = 1 y1 – 2y2 + 4y3 = –3 or – y1 + 2y2 – 4y3 = 3
Thus the dual problem is The first constraint of the primal is an equality
Maximize W = 7y 1 + 2y 2 subject to the \ y1 is unrestricted.
constraints The second constraint is of £ type (not standard)
9y1 + 6y2 £ 2 \ y2 £ 0
– 3y1 – 4y2 £ 5 The third constraint is ≥ (standard)
5y1 – 3y2 = 1 \ y3 ≥ 0
y1 ≥ 0, y2 £ 0 Hence the dual is
Example 6.4 Write the dual of Maximize W = 22y1 + 65y2 + 120y3 subject to
Maximize Z = 5x 1 + 6x 2 subject to the the constraints
constraints y1 + 3y2 + y3 £ 4
x1 + 2x2 = 5 y1 + 5y2 + 7y3 £ 5
– x1 + 5x2 ≥ 3 y1 – 2y2 + 4y3 = – 3 (–y1 + 2y2 – 4y3 = 3)
x1 unrestricted, x2 ≥ 0. y1 is unrestricted, y2 £ 0, y3 ≥ 0.
Solution The objective function of the dual, is Example 6.6 Using the previous example verify
Minimize W = 5y1 + 3y2 that the dual of the dual is the primal.
The first constraint is equality. Hence the Solution Given primal is
corresponding variable of the dual is unrestricted, Minimize Z = 4x1 + 5x2 – 3x3 subject to the
i.e. y1 is unrestricted. constraints
The second constraint is of ≥ type (not of x1 + x2 + x3 = 22
standard type) and hence y2 £ 0. 3x1 + 5x2 – 2x3 £ 65
x1 is unrestricted and hence the first constraint x1 + 7x2 + 4x3 ≥ 120
of the dual is an equality. y1 – y2 = 5. x1, x2 ≥ 0, x3 unrestricted
x2 ≥ 0 Æ the second constraint of the dual is of The dual problem is
standard type ≥ (minimization) Maximize W = 22y1 + 65y2 + 120y3 subject to
2y1 + 5y2 ≥ 6 the constraints
Thus the dual problem is y1 + 3y2 + y3 £ 4
Minimize W = 5y 1 + 3y 2 subject to the y1 + 5y2 + 7y3 £ 5
constraints y1 – 2y2 + 4y3 = – 3
y1 – y2 = 5 \ y1 is unrestricted, y2 £ 0, y3 ≥ 0
2y1 + 5y2 ≥ 6 Now we write the dual of the dual, considering
\ y1 is unrestricted. y2 £ 0. it as the primal.
Example 6.5 Write the dual of the following LPP For the dual the objective function is
Minimize Z = 4x1 + 5x2 – 3x3 subject to the Minimize Z = 4z1 + 5z2 – 3z3
constraints y1 unrestricted Æ first constraint is an equality
x1 + x2 + x3 = 22 z1 + z2 + z3 = 22
3x1 + 5x2 – 2x3 £ 65 y2 £ 0 Æ second constraint is not of the standard
x1 + 7x2 + 4x3 ≥ 120 type. Hence it should be £
x1, x2 ≥ 0 3z1 + 5z2 – 2z3 £ 65
\ x3 unrestricted. y3 ≥ 0 Æ the third constraint is of standard
Solution The objective function of the dual is type ≥
Maximize W = 22y1 + 65y2 + 120y3 z1 + 7z2 + 4z3 ≥ 120
x1, x2 ≥ 0 Æ, the first two constraints of the The first two constraints of the primal are £
dual, are of standard type £ (maximization) (standard)
i.e. y1 + 3y2 + y3 £ 4 \ z1 ≥ 0, z2 ≥ 0
 62 Operations Research

The third constraint is an equality. Hence in the a11x1 + a12x2 + … + a1nxn £ b1

dual z3 is unrestricted. a21x1 + a22x2 + … + a2nxn £ b2
Thus the dual is M M M
Maximize Z = 4z1 + 5z2 – 3z3 subject to the am1x1 + am2 x2 + … + amnxn £ bm
constraints x1, x2, … xn ≥ 0
z1 + z2 + z3 = 22 which is the given primal.
3z1 + 5z2 – 2x3 £ 65 Hence the dual of the dual is the primal.
z1 + 7z2 + 4z3 ≥ 120 Theorem 2: The value of the objective function Z
z1, z2 ≥ 0 z3 unrestricted. for any feasible solution of the primal is less than
Changing the symbols of the variables or equal to the value of the objective function W
z1 Æ x1, z2 Æ x2, z3 Æ x3, Z Æ Z for any feasible solution of the dual.
We find that this is the same as the given primal. Proof: Let the primal problem be
Hence the result. Maximize Z = c1x1 + c2x2 + … + cnxn subject
to the constraints
6.2 DUALITY THEOREMS a11x1 + a12x2 + … + a1nxn £ b1
Theorem 1: The dual of the dual is the primal. a21x1 + a22x2 + … + a2nxn £ b2
Proof: Let the given primal problem be written in M M M
the canonical form (standard type), as am1x1 + am2x2 + …. + amnxn £ bm
Maximize Z = c1x1 + c2x2 + … + cnxn subject x1, x2, …, xn ≥ 0 (1)
to the constraints The dual problem is
a11x1 + a12x2 + … + a1nxn £ b1 Minimize W = b1y1 + b2y2 + … + bmym subject
a21x1 + a22x2 + … + a2nxn £ b2 to the constraints
a11y1 + a21y2 + ... + am1ym ≥ c1
M M M
am1x1 + am2x2 + … + amnxn £ bm a12y1 + a22y2 + ... + am2ym ≥ c2
x, x2, x3,…, xn ≥ 0 (1) M M M
The dual of this problem is a1ny1 + a2ny1 + … + amnym ≥ cn
Minimize W = b1y1 + b2y2 + … + bmym subject y1, y2, …, ym ≥ 0 (2)
to the constraints Multiplying the first constraint of (1) by y1, the
a11y1 + a21y2 + … + am1ym ≥ c1 second constraint by y2, etc. and adding all of them
a12y1 + a22y2 + … + am2ym ≥ c2 we get
M M M (a11x1y1 + a12x2y1 + … + a1nxny1)+
a1ny1 + a2ny2 + … + amnym ≥ cn (a21x1y2 + a22x2y2 + … + a2nxny2)+
y1, y2,…ym ≥ 0 (2) M M M
Now consider (2) as the primal and write its (am1x1ym + am2x2ym + … + amnxnym)
dual as £ b1y1 + b2y2 + … + bmym (3)
Maximize U = c1t1 + c2t2 + … + cntn subject to Similarly, multiplying the first constraint of (2)
the constraints by x1, the second constraint by x2, etc. and adding
a11t1 + a12t2 + … + a1ntn £ b1 them all we get
a21t1 + a22t2 + … + a2ntn £ b2 (a11x1y1 + a21x1y2 + … + am1x1ym)+
M M M (a12x2y1 + a22x2y2 + … + am2x2 ym)+
am1t1 + am2t2 + … + amntn £ bm (a1nxny1 + a2nxny2 + … + amnxnym)
t1, t2,…tn ≥ 0 (3) ≥ c1x1 + c2x2 + … + cnxn (4)
Writing Z for U and x1, x2…xn instead of t1, The sums on the LHS of (3) and (4) are equal
t2…tn \ c1x1 + c2x2 + …cnxn £ a11x1y1 + … +
We get amnxnym
Maximize Z = c1x1 + c2x2 + … + cnxn subject £ b1y1 + b2y2 + … + bmym
to the constraints \ Z£W
 Duality 63

i.e. any feasible solution to the primal is less in the optimal solution of the primal. If x i
than or equal to any feasible solution to the dual. corresponds to si, then xi = a If xi corresponds to
Note: If W Æ • (unbounded solution to the dual) Ri then the value of xi in the optimal solution is
then the primal cannot have a feasible solution. Also xi = b. This fact can be stated as follows. “In the
if Z Æ • (unbounded solution to the primal) solution of the dual find the relative cost factors
then the dual cannot have a feasible solution. of the starting basic variables in the optimal table.
Primal unbounded Æ Dual infeasible In case of a slack variable it represents the optimal
Dual unbounded Æ Primal infeasible. value of the corresponding primal variable. In case
of an artificial variable delete M and change the
Theorem 3 (Fundamental theorem): If both
sign. It gives the optimal value of the corres-
primal and dual have optimal solutions then their
ponding primal variable. This is called principle
optimum values must be equal.
of duality. It helps us to give the solution of the
i.e. Z* = W*.
primal from the solution of the dual without
Proof: Let Z* and W* be the optimum values of Z
solving the primal directly. Also Z* = W*.
and W respectively. The primal is a maximization
problem and the dual is a minimization problem. Example 6.7 Solve the following problem by
Hence for any feasible solutions Z and W of the solving its dual
primal and dual we have Z £ Z* and W* £ W. Maximize Z = x1 + 2x2 subject to the constraints
But we have proved that Z £ W. – x1 + 3x2 £ 10
Since Z* and W* are also feasible solutions we x1 + x2 £ 6
find that Z* £ W* x1 – x2 £ 2
Thus Z £ Z* £ W* £ W. x1 + 3x2 £ 6
Suppose Z* < W*. Then there is a value of Z, 2x1 + x2 £ 4
say Z1 such that Z1 £ W* and Z* < Z1. But Z1 is a x1, x2 ≥ 0
feasible solution of the primal and max Z = Z*. Solution The dual of the given problem is
Hence Z1 £ Z*. But this is a contradiction. Minimize W = 10y1 + 6y2 + 2y3 + 6y4 + 4y5
Hence Z* = W* subject to the constraints
–y1 + y2 + y3 + y4 + 2y5 ≥ 1
6.3 PRINCIPLE OF DUALITY 3y1 + y2 – y3 + 3y4 + y5 ≥ 2
y1, y2, y3, y4, y5 ≥ 0
Each variable xi of the primal corresponds to a The standard form of the dual is
constraint ci of the dual. This constraint contains Maximize W1 = –W = –10y1 – 6y2 – 2y3 – 6y4 –
a slack or artificial variable (si or Ri) of the dual 4y5 subject to the constraints
which becomes a starting basic variable. In the – y1 + y2 + y3 + y4 + 2y5 – s1 + R1 = 1
optimal table of the dual note down the relative 3y1 + y2 – y3 + 3y4 + y5 – s2 + R2 = 2
cost factor of si (say a) or Ri (say M – b). This y1, y2, y3, y4, y5, s1, s2, R1, R2 ≥ 0
corresponds to the value of the primal variable xi The simplex table (Big M method) is
Table 6.1
W1 y1 y2 y3 y4 y5 s1 s2 R1 R2
XB
CB B – 10 –6 –2 –6 –4 0 0 –M –M
–M R1 –1 1 1 1 2 –1 0 1 0 1
–M R2 3 1 –1 3 1 0 –1 0 1 2
cj 10 – 2M 6 – 2M 2 6 – 4M 4 – 3M M M 0 0 – 3M

R1 – (1/3)R2 R2/3
 64 Operations Research

y1 y2 y3 y4 y5 s1 s2 R1 R2
XB
CB B –10 –6 –2 –6 –4 0 0 –M –M
–M R1 –2 2/3 4/3 0 5/3 –1 1/3 1 –1/3 1/3
–6 y4 1 1/3 –1/3 1 1/3 0 –1/3 0 1/3 2/3
c j 4 + 2M 4 – (2M/3) 4 – (4M/3) 0 2 – (5M/3) M 2 – (M/3) 0 4M/3 – 2 – 4 – M/3

R1(3/5), R2 – (1/5)R1
–4 y5 –6/5 2/5 4/5 0 1 –3/5 1/5 3/5 –1/5 1/5
–6 y4 7/5 1/5 –3/5 1 0 1/5 –2/5 –1/5 2/5 3/5
cj 32/5 16/5 12/5 0 0 6/5 8/5 M – 6/5 M – 8/5 – 22/5

The optimality is reached. Relative cost factors Maximize W1 = – y1 + 4y2¢ – 3y3 + 0s1 + 0s2
of the starting basic variables R1 and R2 in the – MR1 – MR2 subject to the constraints
optimal table are M – 6/5 and M – 8/5. Deleting M y1 – y2¢ + y3 – s1 + R1 = 3
and changing the sign we get 6/5, 8/5. Variables – y1 – y2¢ – 3y3 – s2 + R2 = 4
corresponding to the constraints are x1 and x2. y1, y2¢, y3 ≥ 0
Therefore, x1 = 6/5 and x2 = 8/5 are the values of The simplex table is
x1 and x2 giving optimal solution of the primal.
Thus x1 = 6/5, x2 = 8/5, Z* = W* = – W1* = 22/5 is Table 6.2
the solution of the given LPP. W1 y1 y 2¢ y3 s1 s2 R1 R2
XB
Example 6.8 Using the duality theory solve: CB B –1 4 –3 0 0 –M –M
Maximize Z = 3x 1 + 4x 2 subject to the –M R1 1 –1 1 –1 0 1 0 3
constraints –M R2 –1 –1 –3 0 –1 0 1 4
x1 – x2 £ 1
x1 + x2 ≥ 4 cj 1 2M – 4 2M + 3 M M 0 0 –7M
x1 – 3x2 £ 3
The current solution is optimal. But R1 and R2
x1, x2 ≥ 0
are present in the optimal basis at + ve level. Hence
Solution The given problem (primal) is
W1 Æ – •, which shows that the primal problem
Maximize Z = 3x 1 + 4x 2 subject to the
has no feasible solution.
constraints
x1 – x2 £ 1 Example 6.9 Using duality theory solve
x1 + x2 ≥ 4 Maximize Z = 4x 1 + 3x 2 subject to the
x1 – 3x2 £ 3 constraints
x1, x2 ≥ 0 2x1 + x2 £ 10
The dual is x1 + x2 £ 8
Minimize W = y1 + 4y2 + 3y3 subject to the x1, x2 ≥ 0
constraints Solution The dual problem is
y1 + y2 + y3 ≥ 3 Minimize W = 10y 1 + 8y 2 subject to the
– y1 + y2 – 3y3 ≥ 4 constraints
y1 ≥ 0, y2 £ 0, y3 ≥ 0 2y1 + y2 ≥ 4
Take y2¢ = – y2 y1 + y2 ≥ 3
The standard form of the dual is y1, y2 ≥ 0
 Duality 65

The standard form is Solution The dual of the given problem is

Maximize W1 = –10y1 – 8y2 + 0s1 + 0s2 – MR1 Maximize W = 5y1 + 12y2 + 4y3 subject to the
– MR2 subject to the constraints constraints
2y1 + y2 – s1 + R1 = 4 y1 + 2y2 + y3 £ 5
y1 + y2 – s2 + R2 = 3 2y1 – y2 + 3y3 £ 2
y1, y2, s1, s2, R1, R2 ≥ 0 y1, y2, y3 ≥ 0
The simplex table is The standard form of the dual is
Maximize W = 5y1 + 12y2 + 4y3 + 0s1 + 0s2
Table 6.3 subject to the constraints
W1 y1 y2 s1 s2 R1 R2 XB
y1 + 2y2 + y3 + s1 = 5
2y1 – y2 + 3y3 + s2 = 2
CB B –10 –8 0 0 –M –M 0
y1, y2, y3, s1, s2 ≥ 0
–M R1 2 1 –1 0 1 0 4 The simplex table is
–M R2 1 1 0 –1 0 1 3
Table 6.4
cj 10 – 3M 8 – 2M M M 0 0 –7M
W1 y1 y2 y3 s1 s2 XB
R1/2, R2 – (1/2) R1
CB B 5 12 4 0 0 0
–10 y1 1 1/2 –1/2 0 1/2 0 2
0 s1 1 2 1 1 0 5
–M R2 0 1/2 1/2 –1 –1/2 1 1
0 s2 2 –1 3 0 1 2
cj 0 3 – M/2 5 – M/2 M (3M/2) –5 0 –M – 20
cj –5 – 12 –4 0 0 0
R1 – R2 (R2)2
R1/2 R2 + (1/2)R1
–10 y1 1 0 –1 1 1 –1 1
12 y2 1/2 1 1/2 1/2 0 5/2
–8 y2 0 1 1 –2 –1 2 2
0 s2 5/2 0 7/2 1/2 1 9/2
cj 0 0 2 6 M–2 M–6 –26
cj 1 0 2 6 0 30
The optimality is reached. R1 and R2 are the
The optimality is reached.
starting basic variables. The relative cost factors The relative cost factors of s1 and s2 in the
of R1 and R2 in the optimal table are M – 2, M – 6. optimal table are 6 and 0 respectively. The primal
Deleting M and the negative sign we get 2 and 6. variables corresponding to s1 and s2 are x1 and x2
R1 appears in the first constraint and R2 in the respectively. Hence the optimal solution of the
second constraint which correspond to x1 and x2 primal is
of the primal. x1 = 6, x2 = 0, Z* = 5(6) + 2(0) = 30
Hence the solution is x1 = 2, x2 = 6, Z* = 4(2) + Note: The solution of the dual is
3(6) = 26. y1 = 0, y2 = 5/2 y3 = 0, W* = 30.
Note: The solution of the dual is
Example 6.11 Obtain the solution by solving its
y1 = 1, y2 = 2, W* = –W1* = 26, Z* = W*
dual
Example 6.10 Solve using the principle of Maximize Z = 3x 1 + 2x 2 subject to the
duality constraints
Minimize Z = 5x1 + 2x2 subject to the constraints x1 + x2 ≥ 1
x1 + 2x2 ≥ 5 x1 + x2 £ 7
2x1 – x2 ≥ 12 x1 + 2x2 £ 10
x1 + 3x2 ≥ 4 x2 £ 3
x1, x2 ≥ 0 x1, x2 ≥ 0
 66 Operations Research

Solution The dual problem is Maximize W1 = y1¢ – 7y2 – 10y3 – 3y4 + 0s1
Minimize W = y1 + 7y2 + 10y3 + 3y4 subject to + 0s2 – MR1 – MR2 (y1¢ = – y1) subject to the
the constraints constraints
y1 + y2 + y3 ≥ 3 – y1¢ + y2 + y3 – s1 + R1 = 3
y1 + y2 + 2y3 + y4 ≥ 2 – y1¢ + y2 + 2y3 + y4 – s2 + R2 = 2
y1 £ 0, y2, y3, y4 ≥ 0 y1¢, y2, y3, y4, s1, s2, R1, R2 ≥ 0
The standard form is The simplex table is

Table 6.5
W1 y1¢ y2 y3 y4 s1 s2 R1 R2 XB
CB B 1 –7 –10 –3 0 0 –M –M
–M R1 –1 1 1 0 –1 0 1 0 3
–M R2 –1 1 2 1 0 –1 0 1 2
cj 2M – 1 7 – 2M 10 – 3M 3–M M M 0 0 – 5M

R1 – 1/2 R2, R2/2

–M R1 –1/2 1/2 0 –1/2 –1 1/2 1 –1/2 2
–10 y3 –1/2 1/2 1 1/2 0 –1/2 0 1/2 1
cj 4 + (M/2) 2 – (M/2) 0 (M/2) – 2 M 5 – M/2 0 (3M/2) – 5 –10 – 2M

R1(2), R2 + R1
0 S2 –1 1 0 –1 –2 1 2 –1 4
–10 y3 –1 1 1 0 –1 0 1 0 3
cj 4 –3 0 3 10 0 M – 10 M –30

R1 – R2
0 s2 0 0 –1 –1 –1 1 1 –1 1
–7 y2 –1 1 1 0 –1 0 1 0 3
cj 6 0 3 3 7 0 M–7 M –21

The optimality is reached. The starting basic 6.4 ECONOMIC INTER-

variables R1 and R2 have relative cost factors
M – 7 and M. The corresponding variables in the
primal are x1 and x2. Deleting M and changing the
sign we get the solution,
x1 = 7, x2 = 0, Z* = 21
Note: For the dual we get y1¢ = 0, y2 = 3, y3 = 0,
y4 = 0, W* = –W1* = 21
PRETATION OF THE DUAL
The interpretation of the dual variables from the
cost or economic point of view helps us in making
decisions on business processes. The objective
function represents the profit to be maximized
while the constraints give the limits of the
 Duality 67

resources. The dual objective is to minimize the
cost of the resources by considering many
alternatives. The worth of the alternative uses is
called accounting or shadow prices of the
resources. The shadow prices are determined along
with minimization objective. The management of
Proof: In the optimal table if a primal variable
xj has Zj – cj > 0 then xj is non-basic and hence
xj = 0.
Also in the optimal dual table, if a dual variable
yi has positive value then the ith primal constraint
n

the business will be interested in finding the å aij x j £ bi must become an equation since the
j =1
shadow prices so as to minimize the expenditure
associated with existing capacities of the various corresponding slack variable is zero.
If we assume that vj is the surplus variable for
resources.
the jth dual constraint and si is the slack variable
for the ith primal constraint then
6.5 COMPLEMENTARY m
SLACKNESS THEOREM vj = Zj – cj = å aij yi - c j (1)
i =1
A pair of primal and dual feasible solutions n
xi (i = 1, 2, 3….n) and yj (j = 1, 2, 3…m) are optimal and si = bi – å aij x j (2)
i =1
to their respective problems if and only if the
In other words, if
following conditions are satisfied:
vj > 0 then xj = 0 and if yi > 0 then si = 0.
æ n ö Therefore for the optimal solutions of the primal
yi ç bi - å aij x j ÷ = yi si = 0 i = 1, 2, 3, ..., m and dual
è j =1 ø
vjxj = 0 and yisi = 0 for all i and j.
æm ö This result can be restated as
x j ç å aij yi - c j ÷ = x j v j = 0 j = 1, 2, 3, ..., n é ù
è i =1 ø ém ù n
x j êå aij yi - c j ú = 0 and yi êbi - å aij x j ú = 0
where si and vj are slack and surplus variables ë i =1 û ëê j =1 ûú
associated with the primal and dual constraints. Hence the result.

S H O RT A N S W E R Q U E S T I O N S
■ ■ ■ ■ ■ ■ ■

1. What is duality? Explain. (iii) If the primal is of maximization type and

2. State the fundamental theorem of duality. the variable xk £ 0 then the corresponding
3. State the complementary slackness theorem. k th constraint of the dual is
4. Explain how you will write the dual of a _____________. (Ans: £ type)
given primal. (iv) If the primal is of maximization type and
5. What is the principle of duality? the kth constraint is of £ type then the
6. Fill in the blanks: dual variable yk is _______.
(i) If the primal has n variables and m (Ans: ≥ 0 type)
constraints then the dual has _________, (v) If the primal has unbounded solution
__________. then the dual has _________.
(Ans: m variables, and n constraints) (Ans: infeasible solution)
(ii) If the kth constraint of the primal is an (vi) If the dual has unbounded solution then
equality then the dual variable yk is the primal has _________.
__________. (Ans: unrestricted) (Ans: no feasible solution)
 68 Operations Research

(vii) If a primal variable has positive optimum constraint is __________ at the

value then the corresponding dual optimum. (Ans: an equality)

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Write the dual of the LPP x1 – x2 £ 2

Maximize Z = 2x1 + 5x2 + 6x3 x1 – 2x2 £ 1
subject to the constraints x1, x2 ≥ 0
x1 + 6x2 – x3 £ 3 (ii) Minimize Z = 2x1 + 2x2 subject to the
2x1 – 3x2 + 7x3 £ 6 constraints
x1, x2, x3 ≥ 0 2x1 + 4x2 ≥ 1
2. Write the dual of x1 + 2x2 ≥ 1
Maximize Z = 3x1 + 17x2 + 9x3 subject to the 2x1 + x2 ≥ 1
constraints x1, x2 ≥ 0
x1 – x2 + x3 ≥ 3 (iii) Minimize Z = x1 – x2 + x3 subject to the
– 3x1 + 2x2 £ 1 constraints
x1, x2, x3 ≥ 0 x1 – x3 ≥ 4
3. Write the dual of x1 – x2 + 2x3 ≥ 3
Minimize Z = 10x1 – 6x2 – 8x3 subject to the x1, x2, x3 ≥ 0
constraints (iv) Minimize Z = – x1 + 7x2 + 10x3 + 3x4
x1 – 3x2 + x3 = 5 subject to the constraints
– 2x1 + x2 + 3x3 = 8 –x1 + x2 + x3 ≥ 3
x1, x2, x3 ≥ 0 – x1 + x2 + 2x3 + x4 ≥ 2
4. Write the dual of x1, x2, x3, x4 ≥ 0
Minimize Z = 5x1 + 3x2 + 2x3 subject to the (v) Minimize Z = 4x1 + 2x2 + 3x3 subject to
constraints the constraints
2x1 + x2 + x3 = 6 2x1 + 4x3 ≥ 5
x1 – 3x2 + 2x3 ≥ 8 2x1 + 3x2 + x3 ≥ 4
3x1 + x2 + 5x3 £ 12 x1, x2, x3 ≥ 0
x1≥ 0, x2 £ 0, x3 unrestricted (vi) Maximize Z = 5x1 + 12x2 + 4x3 subject
5. Write the dual of to the constraints
Maximize Z = 5x 1 + x 2 subject to the x1 + 2x2 + x3 £ 5
constraints 2x1 – x2 + 3x3 = 2
3x1 + x2 ≥ 6 x1, x2,x3 ≥ 0
x1 – 2x2 ≥ 2 (vii) Maximize Z = 2x1 + 3x2 subject to the
2x1 + 5x2 ≥ 10 constraints
x1 unrestricted, x2 ≥ 0 – x1 + 2x2 £ 4
6. Obtain the solution of the following problems x1 + x2 £ 6
by solving their duals. x1 + 3x2 £ 9
(i) Maximize Z = 2x1 + x2 subject to the x1, x2 unrestricted
constraints (viii) Maximize z = 2x1 + x2 subject to the
x1 + 2x2 £ 10 constraints
x1 + x2 £ 6 3x1 + x2 = 3
 Duality 69

4x1 + 3x2 ≥ 6 2x1 + 2x2 ≥ 3

x1 + 2x2 ≥ 3 x1, x2 ≥ 0
x1, x2 ≥ 0 (xiv) Minimize Z = 11x1 + 10x2 + 8x3 subject
(ix) Minimize Z = 3x1 + 2x2 subject to the to the constraints
constraints x1 ≥ 40
x1 + x2 ≥ 1 x1 + 2x2 ≥ 96
x1 + x2 £ 7 x1 + x2 + 2x3 ≥ 100
x1 + 2x2 ≥ 10 x1, x2 , x3 ≥ 0
x2 £ 3 (xv) Maximize Z = x1 + 6x2 subject to the
x1, x2 ≥ 0 constraints
(x) Minimize Z = x1 + 2x2 + 3x3 subject to x1 + x2 ≥ 2
the constraints x1 + 3x2 £ 3
x1 – x2 + x3 ≥ 4 x1, x2 ≥ 0
x1 + x2 + 2x3 £ 8 (xvi) Maximize Z = 6x1 – 2x2 + 3x3subject to
x2 – x3 ≥ 2 the constraints
x1, x2, x3 ≥ 0 2x1 – x2 + 2x3 £ 2
(xi) Minimize Z = 40x1 + 200x2 subject to x1 + 4x3 £ 4
the constraints x1, x2 , x3 ≥ 0
x1 + 10x2 ≥ 40 (xvii) Maximize Z = 4x1 + 2x2 + 6x3. subject
3x1 + 10x2 ≥ 60 to the constraints
4x1 + 5x2 ≥ 40 4x1 – x2 £ 9
x1, x2 ≥ 0 – x1 + x2 + 2x3 £ 8
(xii) Maximize Z = 7x1 + 5x2 subject to the – 3x1 + x2 + 4x3 £ 12
constraints 1 £ x1 £ 3, 0 £ x2 £ 5, 0 £ x3 £ 2
3x1 + x2 £ 48 (xviii) Minimize Z = 2x1 + 6x2 subject to the
2x1 + x2 £ 40 constraints
x1, x2 ≥ 0 x1 + 2x2 + x3 ≥ 0
(xiii) Minimize Z = 10x1 + 8x2 subject to the x2 – x3 ≥ 2
constraints x1 + 6x2 + 3x3 ≥ –5
4x1 + 2x2 ≥ 5 x1, x2 ≥ 0 x3 unrestricted

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. Minimize W = 3y 1 + 6y 2 subject to the y1 + 2y3 ≥ 9

constraints y1 £ 0, y2 ≥ 0
y1 + 2y2 ≥ 2 3. Maximize W = 5y 1 + 8y 2 subject to the
6y1 – 3y2 ≥ 5 constraints
– y1 + 7y2 ≥ 6 y1 – 2y2 £ 10
y1, y2 ≥ 0 3y1 – y2 ≥ 6
2. Minimize W = 3y 1 + y 2 subject to the –y1 – 3y2 ≥ 8
constraints y1, y2 unrestricted
y1 – 3y2 ≥ 3 4. Maximize W = 6y1 + 8y2 + 12y3 subject to
– y1 ≥ 17 the constraints
 70 Operations Research

2y1 + y2 + 3y3 £ 5 (vi) x1 = 9/5, x2 = 8/5, x3 = 0, Z* = 141/5

y1 – 3y2 + y3 ≥ 3 (vii) x1 = 9/2, x2 = 3/2, Z* = 27/2
y1 + 2y2 + 5y3 = 2 (viii) x1 = 3/5, x2 = 6/5, Z* = 12/5
y1 unrestricted y2 ≥ 0, y3 £ 0 (ix) x1 = 4, x2 = 3, Z* = 18
5. Minimize W = 6y1 + 2y2 + 10y3 subject to
(x) x1 = 6, x2 = 2, x3 = 0, Z* = 10
the constraints
3y1 + y2 + 2y3 = 5 (xi) x1 = 10, x2 = 3, Z* = 1000
y1 – 2y2 + 5y3 ≥ 1 (xii) x1 = 0, x2 = 40, Z* = 200
y1, y2, y3 £ 0 (xiii) x1 = 1, x2 = 1/2, Z* = 14
6. (i) x1 = 4, x2 = 2, Z* = 10 (xiv) x1 = 40, x2 = 28, x3 = 16, Z* = 848
(ii) x1 = 1/3, x2 = 1/3, Z* = 4/3 (xv) x1 = 3/2, x2 = 1/2, Z* = 9/2
(iii) No solution (xvi) x1 = 4, x2 = 6, x3 = 0, Z* = 12
(iv) x1 = 0, x2 = 3, x3 = 0, x4 = 0, Z* = 21 (xvii) x1 = 3, x2 = 5, x3 = 2, Z* = 34
(v) x1 = 0, x2 = 11/12, x3 = 5/4 Z* = 67/12 (xviii) x1 = 0, x2 = 2/3, x3 = – 4/3, Z* = 4
 7
Revised Simplex Method

CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■

If we solve an LPP using computers, it requires Z – c1x1 – c2x2 – º – cn xn – 0xn + 1

large space in the memory of the computer to store – 0xn + 2 – º – 0xn + m = 0
the entire table. This is not possible in the case of a11x1 + a12 x2 + º + a1n xn + xn + 1 = b1
large-scale problems. Revised simplex method is a21x1 + a22 x2 + º + a2xn + xn + 2 = b2
a modification of the regular simplex method and
it computes and stores only the relevant M M M M
information required for the current iteration and am1x1 + am2x2 + º + amnxn + xn + m = bm
hence it is more economical on the computer. In x1, x2, º, xn xn + 1 , xn + 2 º xn + m = 0
the revised simplex method we need only the Take Z as x0 and cj as – a0j
following during each iteration: j = 1, 2, 3, º, (n + m).
(i) the net evaluations Zj – cj to determine the The constraints can be expressed in matrix
entering variable, form as
(ii) the pivot column, and æ 1 a01 a02 K a0 n a0, n +1 K a0, n + m ö æ x0 ö
(iii) the current basic variables and their values ç0 a a12 K a1n 1 K 0 ÷ ç x1 ÷
in order to determine the leaving variable. ç 11
÷ç ÷
In other words, we need to compute only ç 0 a21 a22 K 0 K 0 ÷ ç x2 ÷
XB, CB, B–1 and Z. We have to compute B–1 at çM M M M ÷çM ÷
each iteration from the B –1 of the previous ç ÷ç ÷
çè 0 a 1 ÷ø çè xn + m ÷ø
iteration. m1 am 2 K amn 0 K
æ0 ö
7.1 STEP-BY-STEP PROCEDURE çb ÷
ç 1÷ æ 1 a0 ö æ x0 ö æ 0 ö
= ç b2 ÷ or çè 0 A ÷ø çè x ÷ø = çè b ÷ø
Step 1 Express the given LPP in the standard çM ÷
form, taking the objective function also as one of ç ÷
the constraints, as follows:
çè b ÷ø
m
 72 Operations Research

Step 2 Obtain an initial basic feasible solution. Taking e(1) as B0(1) we can write B1 as
Initial basis matrix B = Im. Corresponding to each æ 1 -CB ö
B1 = ç
è 0 B ÷ø
column a j of the matrix A, a new (m + 1)
component vector is defined as
[a0j, a1j, a2j º amj] = [– cj, a1j, a2j º amj] B0(1) B1(1) B2(1) K Bm(1)
\ aj(1) = [– cj aj] æ 1 -CB1 -CB 2 L - CBm ö
Similarly, b(1) = [0, b1, b2 º bm]T ç0 B B12 L - B1m ÷
= ç ÷
11

= [0 b]T ç: : : : ÷
ç0 B Bmn ÷ø
The column corresponding to Z (or x 0) is è m1 Bm 2
the (m + 1) component unit vector denoted by e(1). Since B is invertible, we get
The basis matrix B1 of order (m + 1) can be
æ 1 CB B -1 ö
written as
B1 = ç
–1 ÷
B1 = [ e1(1) B1(1) B2(1) º Bm(1)] è 0 B -1 ø
The initial simplex table is

Basic B0(1)( = Z) B1(1)( = s1) B2(1)( = s2) º Bm(1)( = sm) Solution YK(1)
Variable º values
B XB(1)

Z 1 0 0 º 0 0 Zk – ck
s1 0 1 0 º 0 b1 y1k
s2 0 0 1 º 0 b2 y2k

M M M M º M M M
sm 0 0 0 º 1 bm ymk

Step 3 Select the entering variable (Pivot
column)
For each non-basic variable calculate
Zj – cj = CBB1–1 aj(1) – cj
If all the Zj – cj = 0 current solution is optimal.
If some Zj – cj < 0 then select
Zk – ck = Max {| Zj – cj | : Zj – cj < 0 }
Step 5 Update the current solution
Update the initial table by introducing the
new basic variable x k (= a k(1)) into the basis
and removing the variable xr (= Br(1)) from the
basis.
Repeat step 3 to step 5 till an optimal solution
is reached or there is an indication of unbounded
solution (yik £ 0 for all i)

Step 4 Select the leaving variable (Pivot row) Example 7.1 Solve the following LPP using the
Calculate yk(1) = B1– 1ak(1) revised simplex method.
Maximize Z = 2x1 + x2
where ak(1) = [– ck ak] subject to the constraints
Find the ratio 3x1 + 4x2 £ 6
xBr /yrk = Min {xBi/yik : yik > 0 } 6x1 + x2 £ 3
Thus the vector Br(1) leaves the basis. x1, x2 ≥ 0
 Revised Simplex Method 73

Solution Standard form of the LPP is Non-basic variables

Z – 2x1 – x2 = 0
a1(1) a2(1)
3x1 + 4x2 + s1 = 6
x1 x2
6x1 + x2 + s2 = 3
-2 -1
x1, x2, s1, s2 ≥ 0
3 4
The matrix form is
e(1) a1(1) a2(1) a3(1) a4(1) 6 1
I Iteration
æ1 -2 -1 0 0ö æZö
ç0 çx ÷ Zj – cj = (First row of B –1 1) (non-basic columns)
3 4 1 0÷
ç ÷ ç 1 ÷ æ 0ö
è0 6 1 0 1ø ç x2 ÷ = ç 6 ÷ æ -2 -1ö
çs ÷ ç ÷ = (1 0 0 ) ç 3 4 ÷ = ( -2, -1)
ç 1 ÷ è 3ø ç ÷
çè s ÷ø è 6 1ø
2
\ x1 enters the basis
æ 1 -C ö æ Z ö æ 0 ö \ k=1
çè 0 A ÷ø çè X ÷ø = çè b ÷ø xik /yik = Min {xBi/yik, yik > 0}
= Min {xB1/y11, xB2/y21}
B0(1) B1(1) B2(1)
Now yk = B1– 1ak(1)
(1)

æ1 0 0ö
æ -2 ö
Basic matrix B1 = ç 0 0÷ = I
ç
1
÷ = ak(1)= ç 3÷ for k = 1
è0 0 1ø ç ÷
è 6ø
æ 1 CB B -1 ö xB(1) = B – 1b
B1–1 = ç ÷, æ 0ö
è 0 B -1 ø
= b = ç 6÷
ç ÷
æ1 0ö è 3ø
B= ç
è0 1 ÷ø xBr /yrk = Min {6/3 3/6}
= 3/6 = xB2/y21
CB = [ 0 0 ] \ B2(1) leaves the basis
Initial table is We get B1(1) B2(1) XB(1) y1(1)(Zk – ck)
æ 0 0 0 -2 ö
B0(1) B1(1) B2(1) XB(1) Yk(1) ç1 0 6 3÷
Basic
variables Z s1 s2 ç ÷
B è0 1 3 6ø
Z 1 0 0 0 Zk – ck Make row operations
R¢1 = R1 + (2/6) R3,
s1 0 1 0 6
R¢2 = R2 – (3/6) R3,
s2 0 0 1 3
R¢3 = R3/6
 74 Operations Research

The new table is

Basic B0(1) B1(1) B2(1) XB(1)(b) y2(1) (Zk – ck) XB(1)/y2(1)
variables Z s1 x2
B
Z 1 0 1/3 1 – 2/3 –2
s1 0 1 – 1/2 9/2 7/2 9/7
x1 0 0 1/6 1/2 1/6 6/2

Non-basic variables \ B1(1) leaves the basis

a4(1) a2(1) We get
s2 x2 B1(1) B2(1) XB(1) Y2(1)
0 -1 æ0 1 3 1 -2 3ö
0 4 ç 1 -1 2 9 2 7 2 ÷
ç ÷
1 1 è 0 1 6 1 2 1 6ø
æ -1ö æ 0ö Make the row operations
= ç 4÷
a2(1) a4 = ç 0 ÷
(1)
R11 = R1 + (4/21) R2,
ç ÷ ç ÷
è 1ø è 1ø R21 = (2/7)R2,
New basic matrix R31 = R3 – (1/21)R2
æ 1 0 1 3ö
The new table is
B1 = ç 0 1 -1 2 ÷
–1
ç ÷
è 0 0 1 6ø Basic B0(1) B1(1) B2(1) XB(1)(b)
variables B Z x2 x1
II Iteration
Z 1 4/21 5/21 13/7
Zj – cj = (first row of B11) (non-basic columns)
x2 0 2/7 –1/7 9/7
æ -1 0 ö
= (1 0 1 / 3) ç 4 0 ÷ x1 0 –2/21 4/21 2/7
ç ÷
è 1 1ø Non-basic variables
= ( -2 / 3, 1/ 3) (k = 2)
a4(1) a3(1)
Zk – ck = –2/3
s2 s1
\ a2(1) enters the basis, i.e. x2 enters the basis.
Now y2(1) = B1– 1a2(1) 0 0
0 1
æ 1 0 1 3ö æ -1ö æ -2 3ö
ç ÷ 1 0
= çç 0 1 -1 2 ÷÷ ç 4÷ = ç 7 2 ÷
ç ÷ è 16ø æ 0ö æ 0ö
è 0 0 1 6 ø è 1ø
a3(1) = ç 1÷ a4(1) = ç 0÷
XBr /yrk = Min {(9/2)/(7/2), (1/2)/(1/6)} ç ÷ ç ÷
= 9/7 (r = 1) è 0ø è 1ø
 Revised Simplex Method 75

The new basic matrix where e(1) = B0(1) a3(1) = B1(1)

(1) (1)
æ 1 4 21 5 21ö a4 = B2 a5(1) = B3(1)
B1– 1 = ç0 2 7 -1 7 ÷ The basis matrix
ç ÷ B0(1) B1(1) B2(1) B3(1)
è 0 -2 21 4 21ø
æ1 0 0 0ö
III Iteration ç0 1 0 0÷
Zj – cj = (First row of B1–1) B= ç ÷
ç0 0 1 0÷
[column aj(1) (non-basic)] çè 0 0 0 1÷ø
æ 0 0ö B1(1) B2(1) B3(1)
= (1 4 21 5 21) ç 1 0 ÷ æ1 0 0ö
ç ÷ æ 1 CB B -1 ö
è 0 1ø –1
B1 = ç ç 0÷
-1 ÷ B = ç0 1 ÷
= ( 4 21 5 21)
è0 B ø è0 0 1ø
Zj – cj ≥ 0 for all j CB = ( 0 0 0 )
Hence the current solution is optimal.
The initial simplex table is
The optimal solution is
x1 = 2/7, x2 =9/7, Z* = 13/7 Basic B0(1) B1(1) B2(1) B3(1) XB(1)
variables Z s1 s2 s3
Example 7.2 Using the revised simplex B
method Z 1 0 0 0 0
Maximize Z = 3x1 + 5x2 s1 0 1 0 0 4
subject to the constraints s2 0 0 1 0 6
x1 £ 4 s3 0 0 0 1 18
x2 £ 6
3x1 + 2x2 £ 18 I Iteration
x1, x2 ≥ 0 Zj – cj = (First row of B 1– 1) (non-basic column)
Solution Standard form of the LPP is æ -3 -5ö
Z – 3x1 – 5x2 = 0 ç 1 0÷
= (1 0 0 0 ) ç ÷ = ( -3, -5 )
x1 + s1 = 4 ç 0 1÷
x2 + s2 = 6 çè 3 2 ÷ø
3x1 + 2x2 + s3 = 18 a2(1) (x2) enters the basis k = 2
x1, x2, s1, s2, s3 = 0 yk(1) = B1–1 ak(1)
The matrix form is æ -5ö
e(1) a1(1) a2(1) a3(1) a4(1) a5(1) ç 0÷
= ak(1) = a2(1) = ç ÷
æ 1 -3 -5 0 0 0 ö æ Z ö ç 1÷
æ 0ö çè 2÷ø
ç0 1 0 1 0 0 ÷ çç x1 ÷÷ ç 4÷
ç ÷
ç0 0 1 0 1 0 ÷ ç x2 ÷ = ç ÷ æ 0ö
çè 0 ç ÷ ç 6÷
3 2 0 0 1÷ø ç s1 ÷ çè 18÷ø
ç 4÷
XB(1) = B1–1 b = b = ç ÷
ç s2 ÷ ç 6÷
ç ÷ çè 18÷ø
è s3 ø
 76 Operations Research

XBr /yrk = Min {XBi /yi2, yi2 > 0} B1(1) B2(1) B3(1) XB(1) Y2(1)
= Min {XB1/y12, XB2/y22, XB3/y32} æ0 0 0 0 -5ö
= Min {4/0, 6/1, 18/2 } ç1 0 0 4 0÷
ç0 1 0 6 1÷
= 6 = xB2/y22 è0 0 1 18 2ø
(1)
Thus B2 (s2) leaves the basis.
Make row transformations
We get
R11 = R1 + 5R3, R21 = R2, R31 = R3, R41 = R4 – 2R3

Basic B0(1) B1(1) B2(1) B3(1) XB(1) Y1(1) XB(1)/y1(1)

variables Z s1 x2 s3
B
Z 1 0 5 0 30 –3 –
s1 0 1 0 0 4 0 –
x2 0 0 1 0 6 1 6/1 = 6
s3 0 0 0 1 6 3 6/3 = 2

Non-basic variables = Min {4/1, 6/0, 6/3}

a1(1) ( x1 ) a4(1) ( s2 ) = 6/3 = 2 (r = 3)
-3 0 \ B3(1) leaves the basis (s3 leaves the basis)
We get
1 0
B1(1) B2(1) B3(1) XB(1) Y1(1)
0 1
æ0 5 0 30 -3ö
3 0 ç1 0 0 4 1÷
ç ÷
New basic matrix is ç0 1 0 6 0÷
æ1 0 5 0ö çè 0 -2 1 6 3÷ø
ç0 1 0 0÷
B1–1 = ç ÷ Make row transformations
ç0 0 1 0÷ R11 = R1 + R4,
çè 0 0 -2 1÷ø R21 = R2 – (1/3)R4,
II Iteration R31 = R3,
Zj – cj = (First row of B1– 1) (non-basic columns) R41 = (1/3)R4

æ -3 0 ö The improved solution is

ç 1 0÷ Basic B0(1) B1(1) B2(1) B3(1) XB(1)
= (1 0 5 0 ) ç ÷
ç 0 1÷ variables B Z s1 x2 x1 (b)
çè 3 0 ÷ø Z 1 0 3 1 36
= (-3 5) ( K = 1) s1 0 1 2/3 – 1/3 2
\ The vector a1(1) (x1) enters the basis x2 0 0 1 0 6
XBr/yrk = Min { xBi /yi1, yi1 > 0}
x1 0 0 – 2/3 1/3 2
= Min {xB1/y11, xB2/y21, xB3/y31}
 Revised Simplex Method 77

Non-basic variables We have

a4(1) a5(1) C = ( 4 -1 -2 0 0 - M -M )
s2 s3 æ 2 -3 2 1 0 0 0 ö
0 0 A = ç -5 2 3 0 -1 1 0 ÷
ç ÷
0 0 è -3 0 2 0 0 0 1ø
1 0 æ 12 ö
æ 1 -C ö
0 1 b = ç 4÷ B1 = ç
The new basic matrix is ç ÷ è0 A÷ø
è 1ø
æ1 0 3 1ö e(1) a1(1) a2(1) a3(1) a4(1) a5(1) a6(1) a7(1)
ç0 1 2 3 -1 3÷
ç ÷ Z x 1 x 2 x 3 s1 s2 R1 R2
B1–1 = ç 0 0 1 0÷ -4
æ1 1 2 0 0 M Mö
çè 0 0 -2 3 1 3÷ø ç0 2 -3 2 1 0 0 0÷
ç ÷
III Iteration B1–1 = ç0 -5 2 3 0 -1 1 0÷
Zj – cj = (First row of B1–1) (non-basic columns) çè 0 -3 0 2 0 0 0 1÷ø
æ 0 0ö æ 0ö
ç 0 0÷ ç 12 ÷
= (1 0 3 1) ç ÷ = ( 3, 1) b1 = ç ÷
ç 1 0÷ ç 4÷
çè 0 1÷ø çè 1÷ø
Hence the current solution is optimal. The
optimal solution is æ 1 CB B -1 ö
B1–1 = ç ÷
x1 = 2, x2 = 6, Z* = 36 è 0 B -1 ø
Example 7.3 Using the revised simplex method e(1) a4(1) a6(1) a7(1)
Maximize Z = 4x1 – x2 – 2x3 æ1 0 -M -M ö
subject to the constraints ç0 1 0 0 ÷
2x1 – 3x2 + 2x3 £ 12 = ç ÷
ç0 0 1 0 ÷
– 5x1 + 2x2 + 3x3 ≥ 4 çè 0 0 0 1 ÷ø
– 3x1 + 2x3 = 1
x1, x2, x3 ≥ 0 Zj – cj = (1 0 - M -M )
(1)
Solution By introducing slack, surplus and a1 a2(1)
a3 a5(1) (1)

artificial variables, we obtain the standard form as

æ -4 1 2 0 ö
Maximize ç 2 -3 2 0 ÷
Z = 4x1 – x2 – 2x3 + 0s1 + 0s2 – MR1 – MR2 ç ÷
subject to the constraints ç -5 2 3 -1÷
çè -3 0 2 0 ÷ø
2x1 – 3x2 + 2x3 + s1 + 0s2 + 0R1 + 0R2 = 12
– 5x1 + 2x2 + 3x3 + 0s1 – s2 + R1 + 0R2 = 4 = ( – 4 + 8M, 1 – 2M, 2 – 5M, M)
– 3x1 + 0x2 + 2x3 + 0s1 + 0s2 + 0R1 + R2 = 1 2 – 5M is most negative:
x1, x2, x3, s1, s2, R1, R2 ≥ 0 Hence x3 enters the basis
 78 Operations Research

y3(1) = B1–1a3(1) Zj – cj = 1 0 - M
( ( 3M - 2 ) /2 )
æ1 0 -M -M ö æ 2ö æ -4 1 0 M ö
ç0 1 0 0 ÷ ç 2÷ ç 2 -3 0 0 ÷
= ç ÷ ç ÷ ç ÷
ç0 0 1 0 ÷ ç 3÷ ç -5 2 -1 0 ÷
çè 0 0 0 1 ÷ø
çè 2 ÷ø çè -3 0 0 1÷ø
æ -5M + 2 ö
ç 2 ÷ = ëé M 2 - 1, -2 M + 1, M , ( 5M - 2 ) /2 ûù
ç ÷ x2 enters the basis
= ç 3 ÷
çè 2 ÷ø y2(1) = B1–1 a2(1)
æ1 0 -M (3M - 2)/2 ö æ 1ö
XB(1) = B1–1b1 ç0 1 0 -1 ÷ ç -3÷
æ1 0 -M -M ö æ 0ö ç ÷ ç ÷
= ç0 0 1 -3 2 ÷ ç 2÷
ç0 0 ÷ ç 12 ÷ çè 0 1 2 ÷ø çè 0÷ø
1 0
= ç ÷ ç ÷ 0 0
ç0 0 1 0 ÷ ç 4÷
çè 0 0 0 1 ÷ø
çè 1÷ø æ -2M + 1ö
ç -3 ÷
æ -5M ö = ç ÷
ç 12 ÷ ç 2 ÷
= ç ÷ çè 0 ÷ø
ç 4÷
çè XB(1) = B1– 1b1
1÷ø
XBr /yrk = XBr /Yr3 æ1 0 -M (3M - 2)/2 ö æ 0ö
ç0 1 0 -1 ÷ ç 12 ÷
= Min (12 2 4 3 1 2 ) = 1/2 ç ÷ ç ÷
= ç0 0 1 -3 2 ÷ ç 4÷
\ r=3 \ R2 leaves the basis çè 0
We get 0 0 1 2 ÷ø çè 1÷ø

B1(1) B2(1) B3(1) XB(1) Y3(1) æ (-5M - 2)/2 ö

æ 0 -M -M -5M -5M + 2 ö ç 11 ÷
ç ç ÷
s1 1 0 0 12 2÷ = ç 52 ÷
ç ÷
R1 ç 0 1 0 4 3÷ çè 1 2 ÷ø
ç ÷
R2 è 0 0 1 1 2ø XBr /Yr2 = ( – , 5/4, •) Minimum is 5/4
Introduce x3 and drop R2. R0 + (5M/2 – 1) R3. \ R1 leaves the basis.
R1 – R3, R2 – (3/2)R3, R3/2 We get
We get B1(1) B2(1) B3(1) XB(1) Y2(1)
æ 1 0 - M (3M - 2)/2 ö æ 0 -M ( 3M - 2 ) /2 ( -5M - 2 ) /2 -2M + 1ö
ç0 1 0 -1 ÷ s1 ç 1 0 -1 11 -3 ÷
ç ÷ ç ÷
B1–1 = ç 0 0 1 -32÷ R1 ç 0 1 -3 2 52 2 ÷
çè 0 0 1 2 ÷ø ç ÷
0 x3 è 0 0 12 12 0 ø
 Revised Simplex Method 79

Introduce x2 and drop R1. We get

R0 + (M – 1/2) R2, R1 + (3/2)R2, R2/2 B1(1) B2(1) B3(1) XB(1) Y1(1)
æ 0 - 1 2 - 1 4 - 9 4 - 3 4ö
æ1 0 - 1 2 - 1 4ö ç ÷
ç0 s1 ç 1 3 2 - 13 4 59 4 17 4 ÷
1 3 2 -13 4 ÷
B1–1 = ç ÷ x2 ç 0 1 2 -3 4 5 4 -1 4 ÷
ç0 0 1 2 - 3 4÷ çç ÷
çè 0 0 0 1 2 ÷ø
x3 è 0 0 12 1 2 - 3 2 ÷ø
Introduce x1 and drop s1. R0 + (3/17)R1
Zj – cj = (1 CB B -1 ) (a1(1) a5(1) a6(1) a7(1) ) R1 ( 4 17 ) , R2 + (1 17 ) R1 , R3 + ( 6 17 ) R1
æ -4 0 M M ö We get
ç 2 0 0 0÷ æ 1 3 17 - 4 17 - 14 17 ö
= (1 0 - 1 2 - 1 4 ) ç ÷ ç 0 4 17
ç -5 -1 1 0 ÷ 6 17 -13 17 ÷
çè -3 0 0 1÷ø B1– 1 = ç ÷
ç 0 1 17 10 17 - 16 17 ÷
çè 0 6 17 9 17 -11 17 ÷ø
= ( - 3 4 , 1 2, M - 1 2, M - 1 4 )
x1 enters the basis Zj – cj = (1 3 17 - 4 17 - 14 17 )
y1(1) = B1–1 a1(1) æ0 0 M M ö
ç 1 0 0 0÷
æ1 0 - 1 2 - 1 4 ö æ -4 ö ç ÷
ç 0 -1 1 0 ÷
ç0 1 3 2 -13 4 ÷ ç 2 ÷ çè 0 0 0 1÷ø
= ç ÷ ç ÷
ç0 0 1 2 - 3 4 ÷ ç -5÷
çè 0 ç ÷ = ( 3 17 4 17 M - 4 17 M - 14 17 )
0 0 1 2 ÷ø è -3ø
All are non-negative.
\ Current solution is optimal.
æ - 3 4ö
ç 17 4 ÷ æ 1 3 17 - 4 17 - 14 17 ö æ 0 ö
ç ÷ ç 0 4 17 6 17 -13 17 ÷ ç 12 ÷
= ç -1 4 ÷ B1– 1 b1 = ç ÷ ç ÷
çè - 3 2 ÷ø ç 0 1 17 10 17 - 16 17 ÷ ç 4 ÷
çè 0 6 17 ç ÷
9 17 -11 17 ÷ø è 1ø
XB(1) = B1–1b1 æ - 6 17 ö
ç 59 17 ÷
æ1 0 - 1 2 - 1 4ö æ 0ö
= ç ÷
ç0 1 3 2 -13 4 ÷ ç 12 ÷ ç 36 17 ÷
= ç ÷ ç ÷ çè 97 17 ÷ø
ç0 0 1 2 - 3 4÷ ç 4÷
çè 0 ç ÷
0 0 1 2 ÷ø è 1ø \ Solution is
x1 = 59/17, x2 = 36/17, x3 = 97/17, Z* = – 6/17.
æ - 9 4ö
ç 59 4 ÷ 7.2 ADVANTAGES OF REVISED
=ç ÷ SIMPLEX METHOD OVER
ç 5 4÷
çè 1 2 ÷ø THE ORDINARY SIMPLEX
METHOD
XBr /Yr1 = (–, (59/4)/(17/4), –) 1. In the case of LP problems having larger
s1 leaves the basis number of variables than the number of
 80 Operations Research

constraints the amount of computations is 3. Since calculations are carried out on a

very much reduced.
2. When working with computers, in the revised
simplex method, only the basic variables,
inverse of the basis matrix and the constants
are to be stored in the memory.
column only if it is ready to enter the basis
computing errors are minimum.
4. The data can be stored accurately and
compactly since the revised simplex method
works only with the original data.

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Discuss the advantages of revised simplex 3x1 + 4x2 £ 6

method over the ordinary simplex method. 6x1 + x2 £ 3
Solve the following LPP using the revised x1, x2 ≥ 0
simplex method. 8. Maximize Z = 3x1 + 5x2
2. Maximize Z = x1 + x2 subject to the constraints
subject to the constraints x1 £ 4
3x1 + 2x2 £ 6 x2 £ 6
x1 + 4x2 £ 4 3x1 + 2x2 £ 18
x1, x2 ≥ 0 x1, x2 ≥ 0
3. Maximize Z = 6x1 – 2x2 + 3x3 9. Minimize Z = 2x1 + x2
subject to the constraints subject to the constraints
2x1 – x2 + 2x3 £ 2 3x1 + x2 = 3
x1 + 4x3 £ 4 4x1 + 3x2 ≥ 6
x1, x2, x3 ≥ 0 x1 + 2x2 £ 3
4. Maximize Z = x1 + x2 x1, x2 ≥ 0
subject to the constraints 10. Maximize Z = x1 + x2 + 3x3
2x1 + 5x2 £ 6 subject to the constraints
x1 + x2 ≥ 2 3x1 + 2x2 + x3 £ 3
x1, x2 ≥ 0 2x1 + x2 + 2x3 £ 2
x1, x2, x3 ≥ 0
5. Minimize Z = x1 + 2x2
subject to the constraints 11. Maximize Z = 30x1 + 23x2 + 29x3
2x1 + 5x2 ≥ 6 subject to the constraints
x1 + x2 ≥ 2 6x1 + 5x2 + 3x3 £ 26
x1, x2 ≥ 0 4x1 + 2x2 + 5x3 £ 7
6. Maximize Z = x1 + 2x2 x1, x2, x3 ≥ 0
subject to the constraints 12. Maximize Z = 2x1 + 3x2
x1 + x2 £ 3 subject to the constraints
x1 + 2x2 £ 5 x2 – x1 ≥ 0
3x1 + x2 £ 6 x1 £ 4
x1, x2 ≥ 0 x1, x2 ≥ 0
7. Maximize Z = 2x1 + x2 13. Maximize Z = 5x1 + 3x2
subject to the constraints subject to the constraints
 Revised Simplex Method 81

4x1 + 5x2 £ 10 x1 + 4x3 £ 420

5x1 + 2x2 £ 10 x1, x2, x3 ≥ 0
3x1 + 8x2 £ 12 15. Minimize Z = 12x1 + 20x2
x1, x2 ≥ 0 subject to the constraints
14. Maximize Z = 3x1 + 2x2 + 5x3 3x1 + 4x2 ≥ 50
subject to the constraints 7x1 + 12x2 ≥ 120
x1 + 2x2 + x3 ≥ 430 x1, x2 ≥ 0
3x1 + 2x2 £ 460

ANSWERS
■ ■ ■ ■ ■ ■ ■

2. x1 = 8/5, x2 = 3/5, Z* = 11/5 9. x1 = 3/5, x2 = 6/5, Z* = 12/5

3. x1 = 4, x2 = 6, x3 = 0 Z* = 12 10. x1 = 0, x2 = 0, x3 = 1, Z* = 3
4. x1 = 3, x2 = 0, Z* = 3 11. x1 = 0, x2 = 7/2, x3 = 0, Z* = 161/2
5. x1 = 4/3, x2 = 2/3, Z* = 8/3 12. Unbounded solution
6. x1 = 1, x2 = 2, Z* = 5 13. x1 = 28/17, x2 = 15/17, Z* = 185/17
7. x1 = 2/7, x2 = 9/7, Z* = 13/7 14. x1 = 0, x2 = 100, x3 = 230, Z* = 1350
8. x1 = 2, x2 = 6, Z* = 36 15. x1 = 15, x2 = 5/4, Z* = 205
 8
Bounded Variable
Technique
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■

In general the variables of a linear programming
problem have to satisfy the non-negativity
constraint xj ≥ 0. In certain LP problems, some or
all of the variables may have lower and upper limits
to their values. Such a problem can be written as
Maximize Z = CX
subject to AX = b
lj £ xj £ uj ( j = 1, 2, 3, …, n)
where lj and uj are the lower and upper limits of
the variable xj. We may convert these constraints
to equalities by introducing slack and surplus
variables as follows:
xj + x¢j = uj
xj – x≤j = lj
This gives rise to 3n variables and m + 2n
constraints. This problem can be reduced in size
as follows.
xj – x≤j = lj can be written as
xj = lj + x≤j for lower bound constraint
This eliminates xj from all the other constraints
and the problem can be solved as usual. If we take
xj = uj – x¢j for upper bound constraint then a
difficulty arises. There is no guarantee that uj – x¢j
is non-negative and hence the problem may
become infeasible. Therefore we use a special
technique for upper bound constraints.
Instead of taking the constraint
xj + x¢j = uj
We modify the feasibility condition of the
simplex method. In the regular simplex method
all non-basic variables are assigned zero values.
But in the upper bound technique a basic variable
becomes non-basic with its upper bound value.
Also when a non-basic variable enters the basis
its value should not exceed its upper bound.
8.1 STEP-BY-STEP PROCEDURE
OF BOUNDED VARIABLE
TECHNIQUE
Step 1 Write the given LPP in the standard form
by introducing slack and surplus variables.
Step 2 If any variable xj has lower bound it
should be substituted with xj = lj + x¢j where x¢j ≥ 0.
Step 3 Obtain an initial basic solution.
Step 4 If Zj – cj = 0 " j, then the current solution
is optimal. Otherwise choose the most negative
 Bounded Variable Technique 83

Zj – cj say Zr – cr. Then the corresponding non- The initial table is

basic variable xr enters the basis.
x1 x2 x3 s1 s2 XB
Step 5 Compute the quantities CB B 1 3 –2 0 0
q1 = Min {XBi /air, air > 0} 0 s1 0 1 –2 1 0 1
i
0 s2 2 1 2 0 1 8
q2 = Min {ui – XBi/–air, air < 0}
i Zj – cj –1 –3 2 0 0 0
q = Min {q1, q2, ur} where ur is the upper uj 1 3 2 • •
bound of the entering variable xr
[If air ≥ 0 " i then q2 = •] I Iteration x2 enters the basis
q1 = min {XB1/a12, XB2 /a22} = min {1/1, 8/1}
Step 6 If q = q1, corresponding to XBk /akr then
= 1 corresponding to s1
xk leaves the basis. Regular row operations can be
performed. q2 = •
q3 = u2 = 3
If q = q2 corresponding to XBk /akr then xk leaves
the basis, xk becomes non-basic at its upper bound. q = min {q1, q2, q3} = min {1, •, 3} = 1
Substitute \ s1 leaves the basis
xk = uk – x¢k (0 £ x¢k £ uk) x1 x2 x3 s1 s2 XB
Introduce xr and drop x k and perform row
CB B 1 3 –2 0 0
operations. Substitute (XBi)1 = XBi – aikuk and
Z¢0 = Z0 – (Zk – ck) uk 0 s1 0 1 –2 1 0 1¨
0 s2 2 1 2 0 1 8
If q = ur, then xr is substituted at its upper bound.
i.e. xr = ur – xr¢ and xr remains non-basic. (0 £ x¢r £ Zj – cj –1 –3 ≠ 2 0 0 0
ur). uj 1 3 2 • •

Step 7 Go to step 4 and repeat the procedure till Introduce x2 and drop s1 . R2 – R1
optimality is reached.
x1 x2 x3 s1 s2 XB
Example 8.1 Use bounded variable technique
CB B 1 3 –2 0 0
and solve:
Maximize Z = x1 + 3x2 – 2x3 3 x1 0 1 –2 1 0 1
0 s2 2 0 4 –1 1 7
subject to the constraints
Zj – cj –1 0 –4≠ 3 0 3
x2 – 2x3 £ 1
uj 1 3 2 • •
2x1 + x2 + 2x3 £ 8
0 £ x1 £ 1, 0 £ x2 £ 3, 0 £ x3 £ 2 II Iteration x3 enters the basis
q1 = min {7/4} = 7/4 corresponding to s2
Solution The standard form of the LPP is
q2 = min {u2 – XB1/ – a13} = (3 – 1)/– (–2)
Maximize Z = x1 + 3x2 – 2x3 + 0s1 + 0s2
= 2/2 = 1 corresponding to x2
subject to the constraints
q3 = u3 = 2
0x1 + x2 – 2x3 + s1 = 1
q = min {q1, q2, q3} = min {7/4, 1, 2} = 1
2x1 + x2 + 2x3 + s2 = 8 q = q2
0 £ x1 £ 1, 0 £ x2£ 3, 0 £ x3 £ 2,0 £ s1 £ •, \ x2 leaves the basis.
0 £ s2 £ • Introduce x3 and drop x2 R1/ – 2, R2 + 2R1
 84 Operations Research

x1 x2 x3 s1 s2 XB Zj – cj ≥ 0 " j. Hence the current solution is

CB B 1 3 –2 0 0
optimal
x¢1 = 0, x¢2 = 0, x3 = 1, Z* = 8
–2 x3 0 –1/2 1 –1/2 0 –1/2
0 s2 2 2 0 1 1 9 But x1 = u1 – x¢1 = 1 – 0 = 1
Zj – cj –1 –2 0 1 0 1 x2 = u2 – x¢2 = 3 – 0 = 3
uj 1 3 2 • • \ Solution is x1 = 3, x2 = 3, x3 = 1, Z* = 8
Since x2 is non-basic it must be substituted at Example 8.2 Solve using the bounded variable
its upper bound by using technique:
x2 = u2 – x¢2 = 3 – x¢2 Maximize Z = 3x1 + 2x2
X¢B1 = XB1 – a12u2 = –1/2 – (–1/2)3 = 1 subject to the constraints
X¢B2 = XB2 – a22u2 = 9 – 2(3) = 3 x1 – 3x2 £ 3
Z¢0 = Z0 – (Z2 – C2)u2 = 1 – (–2)3 = 7
x1 – 2x2 £ 4
Thus the simplex table is updated as follows
2x1 + x2 £ 20
x1 x2¢ x3 s1 s2 XB x1 + 3x2 £ 30
CB B 1 3 –2 0 0 –x1 + x2 £ 6
–2 x3 0 1/2 1 –1/2 0 1 0 £ x1 £ 8, 0 £ x2 £ 6
0 s2 2 –2 0 1 1 3
Solution Introducing slack variables, we get the
Zj – cj –1≠ 2 0 1 0 7
standard form and the initial table as follows:
uj 1 3 2 • • Maximize
x1 enters the basis Z = 3x1 + 2x2 + 0s1 + 0s2 + 0s3 + 0s4 + 0s5
q1 = min {3/2, 1/0} = 3/2
q2 = • x1 x2 s1 s2 s3 s4 s5 XB
q3 = u1 = 1 CB B 3 2 0 0 0 0 0
ì3 ü 0 s1 1 –3 1 0 0 0 0 3
q = min í , ¥, 1ý = 1
î 2 þ 0 s2 1 –2 0 1 0 0 0 4
q = u1 0 s3 2 1 0 0 1 0 0 20
Therefore x1 is substituted at its upper bound 0 s4 1 3 0 0 0 1 0 30
but it remains non-basic
0 s5 –1 1 0 0 0 0 1 6
\ x1 = u1 – x¢1 = 1 – x¢1
X¢B1 = XB1 – a11u1 = 1 – 0(1) = 1 Zj – cj –3≠ –2 0 0 0 0 0 0
X¢B2 = XB2 – a21u1 = 3 – 2(1) = 1 uj 8 6 • • • • •
Z¢0 = Z0 – (Z1 – c1)u1 = 7 – (–1)1 = 8
I Iteration x1 enters the basis
The updated table is
q1 = min {3/1 4/1 20/2 30/1}
x1 x2 x3 s1 s2 XB = 3 corresponding to s1
CB B 1 3 –2 0 0 q2 = min {(• – 6)/ – (–1)} = •
–2 x3 0 1/2 1 –1/2 0 1 q3 = u1 = 8
0 s2 –2 –2 0 1 1 1 q = min {3, •, 8}
Zj – cj 1 2 0 1 0 8 = 3 corresponding to s1
uj 1 3 2 • • \ s1 leaves the basis
 Bounded Variable Technique 85

x1 x2 s1 s2 s3 s4 s5 XB x1 x2 s1 s2 s3 s4 s5 XB
CB B 3 2 0 0 0 0 0 CB B 3 2 0 0 0 0 0
3 x1 1 –3 1 0 0 0 0 3 0 s1 –1/2 0 1 –3/2 0 0 0 –3
0 s2 0 1 –1 1 0 0 0 1 2 x2 –1/2 1 0 –1/2 0 0 0 –2
0 s3 0 7 –2 0 1 0 0 14 0 s3 5/2 0 0 1/2 1 0 0 22
0 s4 0 6 –1 0 0 1 0 27 0 s4 5/2 0 0 3/2 0 1 0 36
0 s5 0 –2 1 0 0 0 1 9 0 s5 –1/2 0 0 1/2 0 0 1 8
Zj – cj –0 –11 3 0 0 0 0 9 Zj – cj –4 0 0 –1 0 0 0 –4
uj 8 6 • • • • • uj 8 6 0 0 0 0 0
R2 – R1, R3 – 2R1, R4 – R1, R5 + R1
x1 becomes non-basic. It must be substituted at its
upper bound. Hence we get,
II Iteration x2 enters the basis.
q1 = min {1/1 14/7 27/6} x1 = 8 – x1¢
= 1 corresponding to s2 X¢B1 = XB1 – a11u1 = –3 – (–1/2)8 = 1
q2 = min {(8 – 3)/ – (–3), •} = 5/3 X¢B2 = XB2 – a21u1 = –2 – (–1/2)8 = 2
q3 = u2 = 6 X¢B3 = XB3 – a31u1 = 22 – (5/2)8 = 2
q = min {1, 5/3, 6} X¢B4 = XB4 – a41u1 = 36 – (5/2)8 = 16
= 1 corresponding to s2
X¢B5 = XB5 – a51u1 = 8 – (–1/2)8 = 12
\ s2 leaves the basis
Z¢0 = Z0 – (Z1 – c1)u1 = – 4 – (– 4)8 = 28
x1 x2 s1 s2 s3 s4 s5 XB The updated table is
CB B 3 2 0 0 0 0 0
x1¢ x2 s1 s2 s3 s4 s5 XB
3 x1 1 0 (–2) 3 0 0 0 6
0 x2 0 1 –1 1 0 0 0 1 CB B –3 2 0 0 0 0 0
0 s3 0 0 5 –7 1 0 0 7 0 s1 1/2 0 1 –3/2 0 0 0 1
0 s4 0 0 5 –6 0 1 0 21 2 x2 1/2 1 0 –1/2 0 0 0 2
0 s5 0 0 –1 2 0 0 1 11 0 s3 –5/2 0 0 1/2 1 0 0 2
Zj – cj 0 0 –8 ≠ 11 0 0 0 20 0 s4 –5/2 0 0 3/2 0 1 0 16
uj 8 6 • • • • • 0 s5 1/2 0 0 1/2 0 0 1 12
–4 0 0 –1 ≠ 0 0 0 28
R1 + 3R2, R3 – 7R2, R4 – 6R2, R5 + 2R2

III Iteration s1 enters IV Iteration s2 enters the basis

q1 = min {7/5, 21/5} = 7/5 corresponding to s3 q1 = min {2/(1/2), 16/(3/2), 12/(1/2)}
q2 = min {(8 – 6)/2, (6 – 1)/1, •} = min{ 4, 32/3, 24} = 4 corresponding to s3
= 1 corresponding to x1 q2 = min {•, (6 – 2)/1/2}
q3 = • = 8 corresponding to x2
q = min {7/5, 1, •} = q2 = 1 corresponding to x1 q3 = •
\ x1 leaves the basis q = min {4, 8, • } = q1 = 4
R1/(–2), R2 – R1/2, R3 + (5/2)R1, R4 + (5/2)R1, \ s3 leaves the basis
R5 – R1/2 R1 + 3R3, R2 + R3, R3(2), R4 – 3R3, R5 – R3
 86 Operations Research

x1¢ x2 s1 s2 s3 s4 s5 XB Updating the table we get,

CB B –3 2 0 0 0 0 0 x1¢ x2¢ s1 s2 s3 s4 s5 XB
0 s1 –7 0 1 0 3 0 0 7 CB B –3 –2 0 0 0 0 0
2 x2 –2 1 0 0 1 0 0 4 0 s1 0 7/2 1 0 –1/2 0 0 14
0 s2 –5 0 0 1 2 0 0 4 –3 x1¢ 1 1/2 0 0 –1/2 0 0 1
0 s4 5 0 0 0 –3 1 0 10 0 s2 0 5/2 0 1 –1/2 0 0 9
0 s5 3 0 0 0 –1 0 1 10 0 s4 0 –5/2 0 0 –1/2 1 0 5
Zj – cj –1 ≠ 0 0 0 2 0 0 32 0 s5 0 –3/2 0 0 1/2 0 1 7
uj 8 6 • • • • • Zj – c j 0 1/2 0 0 3/2 0 0 33
uj 8 6 • • • • •
V Iteration x¢1 enters the basis
q1 = min {10/5, 10/3} = 2 corresponding to s4 The optimality reached since Zj – cj ≥ 0 " j
q2 = min {•, (6 – 4)/2, •} The solution is x¢1 = 1, x¢2 = 0, Z* = 33
x1 = 8 – 1, x2 = 6 – 0, Z* = 33
= 1 corresponding to x2
x1 = 7, x2 = 6, Z* = 33
q3 = 8
q1 = min {2, 1, 8} = q2 Example 8.3 Solve using the bounded variable
technique
= 1 corresponding to x2 Maximize Z = 3x1 + 5x2 + 3x3
\ x2 leaves the basis subject to the constraints
R1 – (7/2)R2, R2/(–2), R3 – (5/2)R2, x1 + 2x2 + 2x3 £ 14
R4 + (5/2)R2, R5 + (3/2)R2 2x1 + 4x3 + 3x3 £ 23
0 £ x1 £ 4, 2 £ x2 £ 5, 0 £ x3£ 3
x1¢ x2 s1 s2 s3 s4 s5 XB Solution The variable x2 has a lower bound 2.
CB B –3 2 0 0 0 0 0 Hence we introduce a variable y2 where y2 = x2 – 2
or x2 = y2 + 2 (0 £ y2 £ 3). Introducing y2 and slack
0 s1 0 –7/2 1 0 –1/2 0 0 –7 variables we get the starting table as:
–3 x1¢ 1 –1/2 0 0 –1/2 0 0 –2
0 s2 0 –5/2 0 1 –1/2 0 0 –6 x1 y2 x3 s1 s2 XB
0 s4 0 5/2 0 0 –1/2 1 0 20 CB B 3 5 3 0 0
0 s5 0 3/2 0 0 1/2 0 1 16
0 s1 1 2 2 1 0 10
Zj – c j 0 –1/2 ≠ 0 0 3/2 0 0 30 0 s2 2 4 3 0 1 15
uj 8 6 0 0 0 0 0
Zj – cj –3 –5 –3 0 0 10
Since x 2 becomes non-basic, it must be uj 4 3 3 • •
substituted at its upper bound. Hence we get
I Iteration y2 enters the basis
x2 = 6 – x¢2 q1 = min {10/2, 15/4} = 15/4
X¢B1 = XB1 – a12u2 = –7 – (–7/2)6 = 14 q2 = •
X¢B2 = XB2 – a22u2 = –2 – (–1/2)6 = 1 q3 = u2 = 3
X¢B3 = XB3 – a32u2 = –6 – (–5/2)6 = 9 q = min {q1, q2, q3]
X¢B4 = XB4 – a42u2 = 20 – (5/2)6 = 5 = min {15/4, • , 3}
q = q3 = 3
X¢B5 = XB5 – a52u2 = 16 – (3/2)6 = 7
Therefore y2 is substituted at its upper bound
Z¢0 = Z0 – (Z2 – c2)u2 = 30 – (–1/2)6 = 33 and remains non-basic. We get
 Bounded Variable Technique 87

y2 = u2 – y¢2 = 3 – y¢2 (0 £ y¢2 £ 3) We get

X¢B1 = XB1 – a12u2 = 10 – 2(3) = 4
x1 y2¢ x3 s1 s2 XB
X¢B2 = XB2 – a22u2 = 15 – 4(3) = 3
Z¢0 = Z0 – (Z2 – c2)u2 = 10 – (–5)3 = 25 CB B 3 –5 3 0 0

The updated table is as follows. 0 s1 0 0 1/2 1 –1/2 5/2

3 x1 1 –2 3/2 0 1/2 3/2¨
x1 y2¢ x3 s1 s2 XB Zj – cj 0 –1 3/2 0 3/2 59/2
≠
CB B 3 –5 3 0 0 uj 4 3 3 • •
0 s1 1 –2 2 1 0 4
0 s2 2 –4 3 0 1 3¨
IV Iteration y2¢ enters the basis
q1 = •,
Zj – cj –3 5 –3 ≠ 0 0 25
q2 = ( 4 – 3/2)/2 = 5/4
uj 4 3 3 • •
q3 = 3
There is a tie for the entering variable. Take x3 q1 = min {•, 5/4, 3} = 5/4
as the entering variable. = q2 corresponding to x1
\ x1 leaves the basis
II Iteration R2/(–2) is the only operation.
q1 = min {4/2, 3/3} = 1 corresponding to s2
We get
q2 = •
q3 = 3 x1 y¢
2 x3 s1 s2 XB
q1 = min {1, •, 3} = 1 corresponding to s2 CB B 3 –5 3 0 0
\ s2 leaves the basis. 0 s1 0 0 1/2 1 –1/2 5/2
R1 – 2/3R2, R2/3 –5 y2¢ –1/2 1 –3/4 0 –1/4 –3/4
Zj – cj –1/2 0 3/4 0 5/4 115/4
x1 y2¢ x3 s1 s2 XB uj 4 3 3 • •
CB B 3 –5 3 0 0
Since q = q2 corresponding to x1, we find that
0 s1 –1/3 2/3 0 1 –2/3 2 the leaving variable x1 should be substituted at its
3 x3 2/3 –4/3 1 0 1/3 1¨ upper bound 4.
Zj – cj –1 ≠ 1 0 0 1 28 \ x1 = 4 – x¢1
uj 4 3 3 • • X¢B1 = XB1 – a11u1 = 5/2 – 0.4 = 5/2
X¢B2 = XB2 – a21u1 = –3/4 – (–1/2)4 = 5/4
x1 enters the basis
Z¢0 = Z0 – (Z1 – c1) u1 = 115/4 – (–1/2)4 = 123/4.
III Iteration The updated table is
q1 = min {1/(2/3)} = 3/2 corresponding to x3
q2 = min {•} = • x1¢ y2¢ x3 s1 s2 XB
q3 = 4 CB B –3 –5 3 0 0
ì3 ü
q = min í , ¥, 4ý = 3/2 corresponding to x3 0 s1 0 0 1/2 1 –1/2 5/2
î2 þ –5 y2¢ 1/2 1 –3/4 0 –1/4 5/4
x3 leaves the basis
R1 + R2/2, R2 (3/2). Zj – cj 1/2 0 3/4 0 5/4 123/4
 88 Operations Research

The optimality reached. x2 = 2 + y2 = 15/4

The solution is x1 = 4, x2 = 15/4, x3 = 0, Z* = 123/4
x¢1 = 0 y¢2 = 5/4
x1 = 4 – 0 y2 = 3 – 5/4 Solve the following LPP using the bounded
x1 = 4 y2 = 7/4 variable technique.

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Maximize Z = x2 + 3x3 4x1 – x2 £ 9

subject to the constraints –x1 + x2 + 2x3 £ 8
x1 + x2 + x3 £ 10 –3x1 + x2 + 4x3 £ 12
x1 – 2x3 £ 0 1 £ x1 £ 3, 0 £ x2 £ 5, 0 £ x3 £ 2
2x2 – x3 £ 10 4. Maximize Z = 3x1 + 5x2 + 2x3
0 £ x1 £ 8, 0 £ x2 £ 4, x3 ≥ 0 subject to the constraints
2. Maximize Z = 4x1 + 4x2 + 3x3
x1 + x2 + 2x3 £ 14
subject to the constraints
2x1 + 4x2 + 3x3 £ 34
– x1 + x2 + 3x3 £ 15
–x2 + x3 £ 4 0 £ x1 £ 4, 7 £ x2 £ 10, 0 £ x3 £ 3
2x1 + x2 – x3 £ 6 5. Maximize Z = 4x1 + 5x2
x1 – x2 + 2x3 £ 10 subject to the constraints
0 £ x1 £ 8, x2 ≥ 0, 0 £ x3 £ 4 2x1 + 3x2 £ 12
3. Maximize z = 4x1 + 2x2 + 6x3 2x1 + x2 £ 6
subject to the constraints x1 ≥ 0, 1 £ x2 £ 3

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. x1 = 20/3, x2 = 0, x3 = 10/3, Z* = 10 4. x1 = 4, x2 = 35/4, x3 = 0, Z* = 223/4

2. x1 = 17/5, x2 = 16/5, x3 = 4, Z* = 192/5 5. x1 = 3/2, x2 = 3, Z* = 21
3. x1 = 3, x2 = 5, x3 = 2, Z* = 34

Integer Programming
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
In linear programming each variable is allowed to
take any non-negative real (fractional) value. But
there are certain problems in which the variables
cannot assume fractional values. For example if
x 1 represents the number of workers and x 2
represents the number of machines then their
values can be only integers. It is meaningless to
take 8.7 workers or 3.2 machines. Hence in such
situations the variables must have only integer
values. In a problem if all the decision variables
are required to assume non-negative integer values
it is called pure integer programming. If, on the
other hand, only some of the variables are
restricted to assume non-negative integer values
it is called mixed integer programming problem.
Problems involving allocation of goods, men,
machines, etc. are integer programming problems.
Suppose that the given pure integer program-
ming problem is
Maximize Z = c1x1 + c2x2 + … + cnxn
subject to the constraints
a11x1 + a12x2 + … + a1nxn = b1
a21x1 + a22x2 + … + a2nxn = b2
: : : systematic and efficient method to solve an integer
am1x1 + am2x2 + … + amnxn = bm
9
x1, x2, x3, …, xn ≥ 0 and are all integers.
We may be tempted to solve this problem by
the simplex method, ignoring the integer restriction
and then round off the non-integral values of the
variables in the optimal solution. But then the
integer valued solution may not satisfy all the
constraints, leading to an infeasible solution. For
example, consider the LPP.
Maximize Z = x1 + x2
subject to the constraints
3x1 + 4x2 £ 14
x1, x2 ≥ 0 and are integers.
The optimal solution of this problem is found
(by graphical method) to be
x1 = 14/3, x2 = 0, Z* = 14/3
If we round off the solution as x1 = 5, x2 = 0
then the constraint 3x1 + 4x2 £ 14 is not satisfied.
On the other hand if we take x1 = 4 then Z* = 4
which is not optimal.
Thus we find that it is necessary to develop a
programming problem (IPP).
 90 Operations Research

9.1 GOMARY’S CUTTING PLANE Step 3 If some variables assume fractional

METHOD values, choose the variable having the largest
fractional part, say x r . Write the Gomary’s
In this method we generate new constraints so as constraint in the form
to ensure integral solution of the given LPP. These s – S frj xj = – fr {0 < fr < 1, 0 < frj < 1}
additional constraints do not cut off that portion
of the original feasible region which contains a Step 4 Add the new constraint at the bottom of
feasible integer point. Also it cuts off the current the simplex table and find a new optimal solution
non-integer solution of the LPP. using the dual simplex method. If the new solution
First we solve the LPP, ignoring the integer gives integer values to the variables it is the
requirement, by the usual simplex method. If the required solution. Otherwise go to step 3. This
optimal solution contains integer values for the process is repeated until all the variables assume
variables then it is itself the solution of the IPP. If integer values.
some of the variables have fractional values then
choose the variable having the largest fractional Example 9.1 Solve the following IPP using
part, say xr. The corresponding rth constraint can Gomary’s cutting plane method
be written as Maximize Z = x1 + x2
br = xr + ar1x1 + ar2x2 + ... subject to the constraints
n 3x1 + 2x2 £ 5
= xr + å arj xj ( j π r) x2 £ 2
j =1
x1, x2 ≥ 0 and are integers.
If [br] denotes the integral part of br and [arj]
denotes the integral part of arj, then we can write Solution Applying simplex method we get the
br = [br] + fr, 0 < fr < 1 and initial table (standard form)
arj = [arj] + frj, 0 < frj < 1
x1 x2 s1 s2 XB
Therefore we get
CB B 1 1 0 0
[br] + fr = xr + S {[arj] + frj} xj ( j π r)
\ fr + {[br] – xr – S [arj] xj} = S frj xj ( j π r) 0 s1 3 2 1 0 5
0 s2 0 1 0 1 2
Since {[br] – xr – S [arj]xj} ≥ 0 we have
Zj – cj –1 –1 0 0 0
fr £ S frj xj
Introducing a slack variable, we get R1/3
fr + s = Sfrj xj 1 x1 1 2/3 1/3 0 5/3
0 s2 0 1 0 1 2
i.e. s – S frj xj = – fr
This is a new constraint called Gomary’s Zj – cj 0 –1/3 1/3 0 5/3
constraint, which represents a cutting plane. R1 – (2/3)R2
1 x1 1 0 1/3 –2/3 1/3
9.2 GOMARY’S ALGORITHM 1 x2 0 1 0 1 2
(STEP-BY-STEP PROCEDURE) Zj – cj 0 0 1/3 1/3 7/3

Step 1 Solve the given LPP using the simplex In the optimal solution obtained x2 assumes
method ignoring the integer constraint. integer value. But x1 has a fractional value. Hence
Step 2 If the optimal values of all the variables we generate a Gomary’s cut. The constraint
are integers it is itself the solution of the LPP. Stop. corresponding to x1 is
 Integer Programming 91

x1 + (1/3)s1 – (2/3) s2 = 1/3 x1 x2 s1 s2 s3 XB

1/3 = x1 + (1/3)s1 – (2/3)s2 CB B 1 2 0 0 0
(0 + 1/3) = (1 + 0)x1 + (0 + 1/3)s1 0 s1 1 1 1 0 0 7
+ (–1 + 1/3)s2 0 s2 2 0 0 1 0 11
i.e., 1/3 < (1/3) s1 + (1/3) s2 0 s3 0 2 0 0 1 7
Using a slack variable s, we get Zj – cj –1 –2 0 0 0 0
1/3 + s = (1/3)s1 + (1/3)s2 R1 – R3/2 R3/2
i.e., s – (1/3)s1 – (1/3)s2 = –1/3 0 s1 1 0 1 0 –1/2 7/2
Introducing this constraint, we get the table as 0 s2 2 0 0 1 0 11
2 x2 0 1 0 0 1/2 7/2
x1 x2 s1 s2 s XB
Zj – cj –1 0 0 0 1 7
CB B 1 1 0 0 0
R2 – 2R1
1 x1 1 0 1/3 –2/3 0 1/3
1 x2 0 1 0 1 0 2 1 x1 1 0 1 0 –1/2 7/2
0 s 0 0 –1/3 –1/3 1 –1/3 0 s2 0 0 –2 1 1 4
Zj – cj 0 0 1/3 1/3 0 7/3 2 x2 0 1 0 0 1/2 7/2
Zj – cj 0 0 1 0 1/2 21/2
Apply the dual simplex method.
s leaves the basis and s1 enters Optimality reached. The variables x1 and x2
R1 + R3, R3(–3). have fractional values 7/2 and 7/2 each. Both have
fractional part 1/2. We can select one variable at
The new table is our choice. Let us take the value x2 = 7/2. The
x1 x2 s1 s2 s XB corresponding constraint is
CB B 1 1 0 0 0 x2 + (1/2)s3 = 7/2
1 x1 1 0 0 –1 1 0
7/2 = x2 + (1/2)s3
1 x2 0 1 0 1 0 2 3 + (1/2) = (1 + 0)x2 + (0 + 1/2)s3
0 s1 0 0 1 1 –3 1 1/2 < 1/2s3
Zj – cj 0 0 0 0 1 2 Introducing slack variable s
1/2 + s = (1/2)s3
Optimal solution of IPP is s = (1/2)s3 – 1/2
x1 = 0, x2 = 2, Z* = 2 (–1/2)s3 + s = –1/2 is the Gomary’s constraint.
Example 9.2 Solve the IPP The new table is
Maximize Z = x1 + 2x2
x1 x2 s1 s2 s3 s XB
subject to the constraints
CB B 1 2 0 0 0 0
x1 + x2 £ 7
1 x1 1 0 1 0 –1/2 0 7/2
2x1 £ 11
0 s2 0 0 –2 1 1 0 4
2x2 £ 7 2 x2 0 1 0 0 1/2 0 7/2
x1, x2 ≥ 0 and are integers. 0 s 0 0 0 0 –1/2 1 –1/2
Solution After introducing the slack variables Zj – cj 0 0 1 0 1/2 0 21/2
we get the initial table as
 92 Operations Research

s leaves and s3 enters the basis The new table is

R1 – R4, R2 + 2(R4), R3 + R4, R4(–2)
x1 x2 s1 s2 s XB
x1 x2 s1 s2 s3 s XB CB B 1 4 0 0 0
CB B 1 2 0 0 0 0 4 x2 1/2 1 1/4 0 0 7/4
1 x1 1 0 1 0 0 –1 4 0 s2 7/2 0 – 3/4 1 0 39/4
0 s2 0 0 –2 1 0 2 3 0 s – 1/2 0 – 1/4 0 1 – 3/4
2 x2 0 1 0 0 0 1 3
Zj – cj 1 0 1 0 0 7
0 s3 0 0 0 0 1 –2 1
Zj – cj 0 0 1 0 0 1 10 (Optimal)
Applying the dual simplex method, we find that s
leaves and x1 enters the basis
The solution of the IPP is
x1 = 4, x2 = 3, Z* = 10 R1 + R3, R2 + 7R3, R3( –2) gives

Example 9.3 Solve the IPP x1 x2 s1 s2 s XB

Maximize Z = x1 + 4x2 CB B 1 4 0 0 0
subject to the constraints 4 x2 0 1 0 0 1 1
2x1 + 4x2 £ 7 0 s2 0 0 – 5/2 1 7 9/2
5x1 + 3x2 £ 15
1 x1 1 0 1/2 0 –2 3/2
x1, x2 ≥ 0 and are integers.
Zj – cj 0 0 1/2 0 2 11/2
Solution The initial simplex table is
x1 x2 s1 s2 XB Choose the x1 row for constructing the Gomary
CB B 1 4 0 0 constraint
0 s1 2 4 1 0 7 1 + (1/2) = x1 + (1/2)s1 – 2s
0 s2 5 3 0 1 15 1/2 < 1/2 s1. Introduce slack variable t,
Zj – cj –1 –4 0 0 0 1/2 + t = 1/2s1
R1/4 R2 – (3/4)R1 t – 1/2s1 = – 1/2 is the Gomary cut.
4 x2 1/2 1 1/4 0 7/4 The new table is
0 s2 7/2 0 –3/4 1 39/4
x1 x2 s1 s2 s t XB
Zj – cj 1 0 1 0 7
CB B 1 4 0 0 0 0
Both constants have the same fractional part 4 x2 0 1 0 0 1 0 1
3/4. Choose the x2 row. We have 0 s2 0 0 – 5/2 1 7 0 9/2
(1/2)x1 + x2 + (1/4)s1 = 7/4 1 x1 1 0 1/2 0 –2 0 3/2
7/4 = (1/2)x1 + x2 + (1/4)s1 0 t 0 0 – 1/2 0 0 1 – 1/2
1 + 3/4 = (0 + 1/2) x1 + x2 + (0 + 1/4) s1
Zj – cj 0 0 1/2 0 2 0 11/2
æ 1ö æ 1ö
3/4 < çè ÷ø x1 + çè ÷ø s1
2 4 By the dual simplex method, t leaves and s1
3/4 + s = (1/2)x1 + (1/4) s1 enters the basis.
i.e., s – (1/2)x1 – (1/4)s1 = – 3/4 is the Gomary R2 – 5R4, R3 + R4, R4( –2)
constraint.
 Integer Programming 93

x1 x2 s1 s2 s t XB we have frj – xj £ – 1 + fr
CB B 1 4 0 0 0 0 Multiply both sides by the negative number
(fr /fr – 1).
4 x2 0 1 0 0 1 0 1
We get (fr /fr – 1) S frj – xj ≥ fr (2)
0 s2 0 0 0 1 7 –5 7
1 x1 1 0 0 0 –2 1 1
Combining (1) and (2) we find that any feasible
solution must satisfy the inequality,
0 s1 0 0 1 0 0 –2 1
S frj + xj + (fr /fr – 1) S frj – xj ≥ fr
Zj - cj 0 0 0 0 2 1 5 (Optimal)
Introducing a slack variable s,
Solution of the given IPP is s + fr = S frj + xj + (fr /fr – 1) S frj – xj
x1 = 1, x2 = 1, Z* = 5 or s – S frj + xj – (fr /fr – 1) S frj – xj = – fr 0 < fr < 1
This is the Gomary’s constraint.
9.3 MIXED INTEGER
Adding this new constraint to the table find
PROGRAMMING a new optimal solution using dual simplex
In the previous section we discussed Gomary’s method.
method to solve IPP where all the variables Example 9.4 Solve the following mixed integer
including slack variables were restricted to take programming problem:
integer values. In the mixed integer programming, Maximize Z = – 3x1 + x2 + 3x3
only some of the decision and slack variables are
subject to the constraints
restricted to assume integer values.
Ignoring the integer restriction, we solve the – x1 + 2x2 + x3 £ 4
LPP. In the optimal table suppose xr has the largest 2x2 – (3/2)x3 £ 1
fractional part. Then the corresponding constraint x1 – 3x2 + 2x3 £ 3
can be written as
x1, x2, x3 ≥ 0 x3 is an integer.
br = xr + S arj xj + S arj xj (j π r)
(arj ≥ 0 ) (arj < 0) Solution The standard form is
[br] + fr = (1 + 0) xr + S {[arj] + frj} xj (arj ≥ 0) + Maximize Z = –3x1 + x2 + 3x3 + 0s1 + 0s2 + 0s3
S{[arj] + frj}xj (arj < 0). subject to the constraints
Rewriting this equation such that all the integer – x1 + 2x2 + x3 + s1 = 4
coefficients appear on the RHS, we get 4x2 – 3x3 + s2 = 2
S + frj xj + S – frj xj = fr + [br] – xr – arj + – arj – x1 – 3x2 + 2x3 + s3 = 3
= fr + I (0 < fr < 1) x1, x2, x3, s1, s2, s3 = 0 x3 is an integer.
where I is the integer value. LHS is positive or The simplex table is
negative according as fr + I is positive or negative.
x1 x2 x3 s1 s2 s3 XB
Case (I) Let fr + I be positive.
CB B –3 1 3 0 0 0
Then S frj+ xj + S frj – xj ≥ fr 0 s1 –1 2 1 1 0 0 4
Since S frj + xj ≥ S frj + xj + S frj–xj 0 s2 0 4 –3 0 1 0 2
we have S frj + xj ≥ fr (1) 0 s3 1 –3 2 0 0 1 3

Case (II) Let fr + I be negative. Zj – cj 3 –1 –3 0 0 0 0

Then S frj + xj + S frj – xj £ –1 + fr R1 – R3/2, R2 + (3/2)R3, R3/2
Now S frj – xj £ S frj + xj + S frj – xj (Contd)
 94 Operations Research

x1 x2 x3 s1 s2 s3 XB x 1 x 2 x3 s1 s2 s3 s XB
CB B –3 1 3 0 0 0 CB B –3 1 3 0 0 0 0
0 s1 –3/2 7/2 0 1 0 –1/2 5/2
1 x2 –1/3 1 0 1/2 0 0 –1/2 1
0 s2 3/2 –1/2 0 0 1 3/2 13/2
3 x3 1/2 –3/2 1 0 0 1/2 3/2 0 s2 1/3 0 0 –2 1 0 5 4
Zj – cj 9/2 –11/2 0 0 0 3/2 9/2 3 x3 –1/3 0 1 0 0 0 1 2

R1(2/7), R2 + R1/7, R3 + 3R1/7 0 s3 2/3 0 0 3/2 0 1 –7/2 2

Zj - cj 5/3 0 0 1/2 0 0 5/2 7 (Optimal)
1 x2 –3/7 1 0 2/7 0 –1/7 5/7
0 s2 9/7 0 0 1/7 1 10/7 48/7 The solution is x1 = 0, x2 = 1, x3 = 2, Z* = 7
3 x3 –1/7 0 1 3/7 0 2/7 18/7
Example 9.5 Solve the following mixed IPP
Zj – cj 15/7 0 0 11/7 0 5/7 59/7
Maximize Z = x1 + x2
x 3 must be an integer. The corresponding subject to the constraints
constraint is 2x1 + 5x2 £ 16
18/7 = – (1/7)x1 + x3 + (3/7)s1 + (2/7)s3 6x1 + 5x2 £ 30
2 + (4/7) = – (1/7)x1 + x3 + (3/7)s1 + (2/7)s3 x1, x2 = 0 x1 is an integer
fr = 4/7. The coefficient of x1 is –1/7. Solution The starting table is
Applying the rule we get
4/7 £ (3/7)s1 + (2/7)s3 + [(4/7) / {(4/7) – 1}] x1 x2 s1 s2 XB
(–1/7)x1 CB B 1 1 0 0
£ (3/7)s1 + (2/7)s3 + (4/21)x1 0 s1 2 5 1 0 16
Introducing slack variable s 0 s2 6 5 0 1 30
4/7 + s = (4/21)x1 + (3/7)s1 + (2/7)s3 Zj –cj –1 –1 0 0 0
The Gomary constraint is R1/5, R2 – R1
– (4/21)x1 – (3/7)s1 – (2/7)s3 + s = – 4/7 1 x2 2/5 1 1/5 0 16/5
Adding this new constraint to the table 0 s2 4 0 –1 1 14
we get Zj –cj –3/5 0 1/5 0 16/5

x1 x2 x3 s1 s2 s3 s XB R1 –R2/10, R2/4

CB B –3 1 3 0 0 0 0 1 x2 0 1 3/10 –1/10 9/5

1 x1 1 0 –1/4 1/4 7/2
1 x2 –3/7 1 0 2/7 0 –1/7 0 5/7
Zj – cj 0 0 1/20 3/20 53/10 (Optimal)
0 s2 9/7 0 0 1/7 1 10/7 0 48/7
3 x3 –1/7 0 1 3/7 0 2/7 0 18/7 Since x1 must have an integer value we take
0 s – 4/21 0 0 –3/7 0 –2/7 1 – 4/7 the corresponding constraint.
Zj –cj 15/7 0 0 11/7 0 5/7 0 59/7
x1 – (1/4)s1 + (1/4)s2 = 7/2
7/2 = x1 – (1/4) s1 + (1/4)s2
R1 – R4/2, R2 + 5R4, R3 + R4, R4( –7/2) 3 + (1/2) = x1 – (1/4)s1 + (1/4)s2
Apply dual simplex method. s leaves and s3 1/2 £ [(1/2)/((1/2) – 1)] ( –1/4)s1 + (1/4)s2
enters the basis. We get £ (1/4)s1 + (1/4)s2
 Integer Programming 95

Introduce s. Perform the operations

1/2 + s = (1/4) s1 + (1/4)s2 R1 + (6/5) R3, R2 – R3, R3( –4).
s – (1/4)s1 – (1/4)s2 = –1/2 is the Gomary We get
constraint.
x1 x2 s1 s2 s XB
Adding this constraint to the table we get
CB B 1 1 0 0 0
x1 x2 s1 s2 s XB 1 x2 0 1 0 –2/5 6/5 6/5
CB B 1 1 0 0 0 1 x1 1 0 0 1/2 –1 4
1 x2 0 1 3/10 –1/10 0 9/5 0 s 0 0 1 1 –4 2
1 x1 1 0 –1/4 1/4 0 7/2 Zj – cj 0 0 0 1/10 1/5 26/5
0 s 0 0 –1/4 –1/4 1 –1/2
Zj – cj 0 0 1/20 3/20 0 53/10 The optimal solution is
x1 = 4, x2 = 6/5, Z* = 26/5
Apply the dual simplex method
s leaves, s1 enters.

EXERCISES
■ ■ ■ ■ ■ ■ ■

Solve the following Integer Programming 5. Maximize Z = 3x1 + 4x2

Problems
1. Maximize Z = 2x1 + 2x2
subject to the constraints
5x1 + 3x2 £ 8
2x1 + 4x2 £ 8
x1, x2 ≥ 0 and are integers.
2. Minimize Z = –2x1 – 3x2
subject to the constraints
2x1 + 2x2 £ 7
x1 £ 2
x2 £ 2
x1, x2 ≥ 0 and are integers.
3. Maximize Z = x1 + x2
subject to the constraints
2x1 + 5x2 £ 16
6x1 + 5x2 £ 30
x1, x2 ≥ 0 and are integers.
4. Maximize Z = x1 + 2x2
subject to the constraints
x1 + x2 £ 7
2x1 £ 11
2x2 £ 7
x1, x2 ≥ 0 and are integers.
subject to the constraints
3x1 + 2x2 £ 8
x1 + 4x2 £ 10
x1, x2 ≥ 0 and are integers.
6. Maximize Z = 7x1 + 9x2
subject to the constraints
–x1 + 3x2 £ 6
7x1 + x2 £ 35
x1, x2 ≥ 0 and are integers.
7. Maximize Z = 9x1 + 10x2
subject to the constraints
4x1 + 3x2 ≥ 40
x1 £ 9
x2 £ 8
x1, x2 ≥ 0 and are integers.
8. Maximize Z = 2x1 + 20x2 – 10x3
subject to the constraints
2x1 + 20x2 + 4x3 £ 15
6x1 + 20x2 + 4x3 = 20
x1, x2, x3 ≥ 0 and are integers.
9. Minimize Z = x1 – 3x2
subject to the constraints
 96 Operations Research

x1 + x2 £ 5 x2 £ 5
–2x1 + 4x2 £ 11 x1, x2 ≥ 0 and are integers.
x1, x2 ≥ 0 x2 is an integer. 16. Minimize Z = 3x1 + 2.5x2
10. Minimize Z =10x1 + 9x2 subject to the constraints
subject to the constraints x1 + 2x2 ≥ 20
x1 £ 8 3x1 + 2x2 ≥ 50
x2 £ 10 x1, x2 ≥ 0 and are integers.
5x1 + 3x2 ≥ 45 17. Maximize Z = 2x1 + 3x2
x1, x2 ≥ 0 x1 is an integer. subject to the constraints
11. Maximize Z = 4x1 + 6x2 + 2x3 x1 + 3x2 £ 9
subject to the constraints 3x1 + x2 £ 7
4x1 – 4x2 £ 5 x1 – x2 £ 1
–x1 + 6x2 £ 5 x1, x2 ≥ 0 and are integers.
–x1 + x2 + x3 £ 5 18. Maximize Z = 7x1 + 6x2
x1, x2, x3 ≥ 0 x1, x3 are integers. subject to the constraints
12. Maximize Z = 20x1 + 17x2 2x1 + 3x2 £ 12
subject to the constraints 6x1 + 5x2 £ 30
4x1 + 3x2 £ 7 x1, x2 ≥ 0 and are integers.
x1 + x2 £ 4 19. Maximize Z = 5x1 + 4x2
x1, x2 ≥ 0 and are integers.
subject to the constraints
13. Maximize Z = 4x1 + 3x2
subject to the constraints x1 + x2 ≥ 2
5x1 + 3x2 £ 15
x1 + 2x2 £ 4
3x1 + 5x2 £ 15
2x1 + x2 £ 6
x1, x2 ≥ 0 and are integers.
x1, x2 ≥ 0 and are integers.
14. Maximize Z = 3x1 + 2x2 20. Maximize Z = 3x1 + 2x2 + 5x3
subject to the constraints subject to the constraints
2x1 + x2 £ 60 5x1 + 3x2 + 7x3 £ 28
x1 £ 25 4x1 + 5x2 + 5x3 £ 30
x2 £ 35 x1, x2 , x3 ≥ 0 and are integers.
x1, x2 ≥ 0 and are integers. 21. Maximize Z = 3x1 + 6x2 + 8x3
15. Maximize Z = 3x1 + 5x2 subject to the constraints
subject to the constraints 5x1 + 4x2 + 8x3 £ 24
2x1 + 4x2 £ 25 2x1 + 4x2 – x3 £ 7
x1 £ 8 x1, x2 , x3 ≥ 0 and x3 is an integer.

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. x1 = 1, x2 = 1, Z* = 4 [x1 = 0, x2 = 2, Z* = 4] 6. x1 = 4, x2 = 3, Z* = 55
2. x1 = 1, x2 = 2, Z* = – 8 7. x1 = 9, x2 = 2, Z* = 101
3. x1 = 3, x2 = 2, Z* = 5, [x1 = 4, x2 = 1, Z* = 5] 8. x1 = 2, x2 = 0, x3 = 2, Z* = –16
4. x1 = 4, x2 = 3, Z* = 10 9. x1 = 1/2, x2 = 3, Z* = –17/2
5. x1 = 0, x2 = 3, Z* = 12 10. x1 = 8, x2 = 5/3, Z* = 95
 Integer Programming 97

11. x1 = 2, x2 = 1, x3 = 6, Z* = 26 xk £ [xk] and xk ≥ [xk] + 1

12. x1 = 1, x2 = 1, Z* = 37
13. x1 = 3, x2 = 0, Z* = 12
14. x1 = 13, x2 = 34, Z* = 107
15. x1 = 8, x2 = 2, Z* = 34
16. x1 = 14, x2 = 4, Z* = 52
17. x1 = 0, x2 = 3, Z* = 9
18. x1 = 5, x2 = 0, Z* = 35
19. x1 = 3, x2 = 0, Z* = 15
20. x1 = 0, x2 = 0, x3 = 4, Z* = 20
21. x1 = 0, x2 = 2, x3 = 2, Z* = 28
9.4 BRANCH AND BOUND
METHOD
Branch and bound method can be used to solve
all integer as well as mixed integer programming
problems. We divide the entire feasible solution
space into smaller parts and then find an optimal
solution for each sub-problem. We start by
imposing feasible upper and lower bounds for the
decision variables in each sub-problem and thus
the number of iterations are reduced by discarding
those sub-problems having non-integer and
infeasible solutions. The integer valued solution
which gives the best optimum value for the
objective function is selected as the optimal
solution of the given IPP.
A disadvantage of this method is that it requires
the optimal solution of each sub-problem.
Whenever the optimum value of the objective
function of a sub-problem is less than (for
maximization problem) that of the best available
solution (already obtained) then this sub-problem
is called fathomed and is dropped from
consideration.
9.4.1 Step-by-step Procedure
Step 1 Solve the given LPP ignoring the integer
restrictions. If the optimal solution satisfies the
integer restrictions then the optimal integer solu-
tion is obtained. Stop. Otherwise go to step 2.
Step 2 Let the optimum Z be Z1. Let xk be the
basic variable having largest fractional value. We
add two mutually exclusive constraints.
This gives rise to two sub-problems
n
(1) Maximize Z = å cj xj
j =1
subject to å aij xj = bi
xk £ [xk]
xj ≥ 0
(2) Maximize Z = å cj x j
subject to å aij x j = bi
xk ≥ [xk] + 1
xj ≥ 0
Step 3 Obtain the optimal solutions of the sub-
problems (1) and (2). Optimum values of Z
obtained be Z2 and Z3. The best integer solution
value becomes the lower bound say ZL.
Step 4 If any sub-problem has an infeasible
solution delete it. If a sub-problem has feasible
but non-integer solution return to step 2. If a sub-
problem yields an integer solution, equal to Z1,
then optimal solution is reached. If this value is
not equal to the upper bound but greater than ZL
then take this value as the new upper bound. Return
to step 2. Finally if it is less than ZL, then terminate
this branch.
Step 5 This procedure of branching
and bounding continues until no further sub-
problem remains to be examined. Now the integer
solution corresponding to the current lower bound
is the optimal integer solution of the given
problem.
Example 9.6 Use the branch and bound method:
Maximize Z = 2x1 + 2x2
subject to the constraints
5x1 + 3x2 £ 8
x1 + 2x2 £ 4
x1, x2 ≥ 0 and are integers
Solution The standard form of the problem is
Maximize Z = 2x1 + 2x2 + 0s1 + 0s2
subject to the constraints
5x1 + 3x2 + s1 = 8
 98 Operations Research

x1 + 2x2 + s2 = 4 X2
x1, x2, s1, s2 ≥ 0 4
Z x1 x2 s1 s2 XB
CB B 2 2 0 0 (0, –83 ) 3
0 s1 5 3 1 0 8
0 s2 1 2 0 1 4
2
cj –2 –2 0 0 0
R1/5, R2 – R1/5 (1,1) (3)
(0,1) 1
2 x1 1 3/5 1/5 0 8/5
0 s2 0 7/5 –1/5 1 12/5
cj 0 – 4/5 2/5 0 16/5 O 1 2 3 4 X1
( –8 , 0
) (1) (2)
R1 – (3/7)R2, R2(5/7) 3
2 x1 1 0 2/7 –3/7 4/7
Fig. 9.1
2 x2 0 1 –1/7 5/7 12/7
cj 0 0 2/7 4/7 32/7(Optimal) The solution is x1 = 1, x2 = 1, Z* = 4 which is an
integer solution. The lower bound of Z is 4. No
The non-integer optimal solution is sub-problem is to be taken. It is fathomed.
x1 = 4/7, x2 = 12/7, Z* = 32/7 Solution of the sub-problem (2)
x2 has the maximum fractional part 5/7
X2
x2 = 12/7, 1< x2 < 2
Take two constraints 4
x2 £ 1 and x2 ≥ 2
Sub-problem (1) (0, 83– ) 3
Maximize Z = 2x1 + 2x2
subject to the constraints ( –25 , 2) (3)
5x1 + 3x2 £ 8 (0,2) 2
x1 + 2x2 £ 4
x2 £ 1 (1,1)
1
x1, x2 ≥ 0
Sub-problem (2)
Maximize Z = 2x1 + 2x2 O 1 2 3 4 X1
subject to the constraints (1) (2)
5x1 + 3x2 £ 8
x1 + 2x2 £ 4 Fig. 9.2
x2 ≥ 2 The optimal integer solution is
x1, x2 ≥ 0
x1 = 0, x2 = 2, Z* = 4. It is fathomed. Hence the
Solution of sub-problem (1) (by graphical integer solution is given by
method) x1 = 1, x2 = 1, Z* = 4. (or) x1 = 0, x2 = 2, Z* = 4
 Integer Programming 99

Example 9.7 (Use branch and bound method) 3x1 + 2x2 ≥ 50

Minimize Z = 3x1 + 2.5 x2 x2 ≥ 3
subject to the constraints x1, x2 are non-negative integers.
x1 + 2x2 ≥ 20
3x1 + 2x2 ≥ 50 Solution of sub-problem (1) (Graphical method)
x1, x2 are non-negative integers. X2
Solution We obtain a non-integer optimal 40
solution by graphical method.

30
(0,25)
20

10
(16,2) (3)

O 10 20 30 40 X1
(2) (1)

Fig. 9.4

The solution is x1 = 16, x2 = 2, Z2 = 53

Fig. 9.3
Solution of sub-problem (2) (Graphical method)

The solution is X2
x1 = 15, x2 = 2.5, Z1 = 51.25 40
Z1 = 51.25 is the lower bound. Z value in the
subsequent sub-problems cannot be less than
51.25. 30
x2(= 2.5) is the only non-integer value. Hence
(0,25)
we get two sub-problems (1) and (2).
20
Sub-problem (1)
Minimize Z = 3x1 + 2.5x2
subject to the constraints
x1 + 2x2 ≥ 20
10 (44
— , 2)
3
3x1 + 2x2 ≥ 50 (3)
x2 £ 2 O 10 20 30 40 X1
x1, x2 are non-negative integers. (2) (1)
Sub-problem (2)
Minimize Z = 3x1 + 2.5x2 Fig. 9.5
subject to the constraints
x1 + 2x2 ≥ 20 The solution is x1 = 44/3, x2 = 3, Z3 = 51.5
 100 Operations Research

Since the solution of sub-problem (1) is all Solution of Sub-problem (4)

integer no more sub-division is necessary. X2
(4)
Z3 = 51.5 is the new lower bound. 40
Also Z3 < Z2.
Hence we have to subdivide problem (2) into
problem (3) and (4). 30

Sub-problem (3) (0,25)

Minimize Z = 3x1 + 2.5x2 20
subject to the constraints
x1 + 2x2 ≥ 20
3x1 + 2x2 ≥ 50 10
x2 ≥ 3 (15,3) (3)
x1 £ 14
O 10 20 30 40 X1
x1, x2 ≥ 0 and are integers. (2) (1)
Sub-problem (4)
Fig. 9.7
Minimize Z = 3x1 + 2.5x2
subject to the constraints The solution is x1 = 15, x2 = 3, Z5 = 52.5
x1 + 2x2 ≥ 20 Both problems (3) and (4) have all integer
3x1 + 2x2 ≥ 50 solutions and therefore branch and bound
x2 ≥ 3 algorithm is terminated.
The optimal all integer solution is given by sub-
x1 ≥ 15 problem (3), which is x1 = 14, x2 = 4, Z4 = 52
x1, x2 ≥ 0 and are integers.
Solution of problem (3)

X2
(4)
40

30
(0,25)
20

Fig. 9.8
10
Example 9.8 Branch and Bound Method
(14,4) (3)
Maximize Z = 7x1 + 9x2
subject to the constraints
O 10 20 30 40 X1
(2) (1)
–x1 + 3x2 £ 6
7x1 + x2 £ 35
Fig. 9.6
x1 £ 7
x2 £ 7
Solution is x1 = 14, x2 = 4, Z4 = 52 x1, x2 are non-negative integers.
 Integer Programming 101

Solution x1 = 9/2, x2 = 7/2, Z1 = 63 Sub-problem (3)

Maximize Z = 7x1 + 9x2
Sub-problem (1)
Maximize Z = 7x1 + 9x2 subject to the constraints
subject to the constraints –x1 + 3x2 £ 6
–x1 + 3x2 £ 6 7x1 + x2 £ 35
7x1 + x2 £ 35 x1 £ 4
x2 £ 7 x2 £ 3
x1 £ 4 x1, x2 are non-negative integers.
x1, x2 are non-negative integers. The solution is x1 = 4, x2 = 3, Z4 = 55
The solution is x1 = 4, x2 = 10/3, Z2 = 58
Sub-problem (4)
Sub-problem (2) Maximize Z = 7x1 + 9x2
Maximize Z = 7x1 + 9x2 subject to the constraints
subject to the constraints –x1 + 3x2 £ 6
–x1 + 3x2 £ 6 7x1 + x2 £ 35
7x1 + x2 £ 35 x1 £ 4
x2 £ 7 x2 ≥ 4
x1 ≥ 5
x2 £ 7
x1 £ 7
This has no solution (infeasible)
x1, x2 are non-negative integers.
\ The optimal integer solution is
The solution is
x1 = 4, x2 = 3, Z4 = 55
x1 = 5, x2 = 0, Z3 = 35 (fathomed).

EXERCISES
■ ■ ■ ■ ■ ■ ■

Using the branch and bound technique solve the x1 – x2 £ 1

following IPP:
1. Maximize Z = 3x1 + 4x2
subject to the constraints
3x1 + 2x2 £ 8
x1 + 4x2 £ 10
x1, x2 ≥ 0 and are integers.
2. Maximize Z = 110x1 + 100x2
subject to the constraints
6x1 + 5x2 £ 29
2x1 + 7x2 £ 24
x1, x2 ≥ 0 and are integers.
3. Maximize Z = 2x1 + 3x2
subject to the constraints
x1 + 3x2 £ 9
3x1 + x2 £ 7
x1, x2 ≥ 0 and are integers.
4. Minimize Z = 9x1 + 10x2
subject to the constraints
x1 £ 9
x2 £ 8
4x1 + 3x2 ≥ 40
x1, x2 ≥ 0 and are integers.
5. Minimize Z = x1 + 4x2
subject to the constraints
2x1 + 4x2 £ 7
5x1 + 3x2 £ 15
x1, x2 ≥ 0 and are integers.
6. Maximize Z = x1 + x2
subject to the constraints
4x1– x2 £ 10
 102 Operations Research

2x1 + 5x2 £ 10 2x1 + x2 £ 6

2x1 – 3x2 £ 6
x1, x2 ≥ 0 and are integers.
7. Maximize Z = x1 + 2x2
subject to the constraints
2x1 £ 7
x1 + x2 ≥ 7
2x1 ≥ 11
10. Maximize Z = 20x1 + 17x2
x1, x2 ≥ 0 and are integers.
8. Maximize Z = 4x1 + 3x2
subject to the constraints
x1 + 2x2 £ 4
x1, x2 ≥ 0 and are integers.
9. Maximize Z = 3x1 + 2x2
subject to the constraints
2x1 + x2 £ 60
x1 £ 25
x2 £ 35
x1, x2 ≥ 0 and are integers.
subject to the constraints
4x1 + 3x2 £ 7
x1 + x2 £ 4
x1, x2 ≥ 0 and are integers.

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. x1 = 0, x2 = 3, Z* = 12 6. x1 = 2, x2 = 1, Z* = 3
2. x1 = 4, x2 = 1, Z* = 540 7. x1 = 4, x2 = 3, Z* = 10
3. x1 = 0, x2 = 3, Z* = 9 8. x1 = 3, x2 = 0, Z* = 12
4. x1 = 9, x2 = 2, Z* = 101 9. x1 = 13, x2 = 34, Z* = 107
5. x1 = 1, x2 = 1, Z* = 5 10. x1 = 1, x2 = 1, Z* = 37
 10
Sensitivity Analysis

CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■

In a linear programming problem we have the values 10.1 DISCRETE CHANGES IN cj

of cost coefficient cj, availability of resources bi and
constraint coefficient aij which contribute very much Let the cost coefficient ck be changed to ck + Dck.
in arriving at an optimal solution. If there is any Two cases arise:
change or error in any of these parameters it gives
Case (I) ck does not belong to CB.
rise to a change in the optimal solution. Also it may
sometimes be required to analyse the effect of In this case the relative cost factor ck becomes
changes in these parameters on the optimal solution. zk – (ck + Dck). Hence the current solution continues
Such changes may be discrete or continuous. A to be optimal if
study of the effect on the optimal solution due to Zk – (ck + Dck ) ≥ 0 or Zk – Ck ≥ Dck
discrete changes in the parameters is called Case (II) ck ΠCB
sensitivity analysis or post-optimality analysis. In m
the case of continuous changes it is called
parametric programming. Here we shall discuss the
Now c j ¢ = å CBi aij - c j¢
i =1
sensitivity analysis only. In most of the cases it is
not necessary to solve the problem again. Some cj ¢ = å CBi aij + (ck + Dck )akj - c j
simple calculations applied to the current optimal i¹k
solution will give the desired result. m
There are five types of changes which are = å CBi aij + Dck akj - c j
possible in an LPP: i =1
(i) change in cj = Zj – cj + Dckakj
(ii) change in bi Hence the current solution continues to be
(iii) change in aij optimal if Zj – cj + Dckakj ≥ 0
(iv) addition of a new variable i.e. Dck akj ≥ – (Zj – cj)
(v) addition of a new constraint i.e. Dck ≥ – (Zj – cj)/akj
 104 Operations Research

i.e. Max {–(Zj – cj)/akj} £ D ck £ Min {–(Zj – cj)/akj } x1 + x2 + x3 £ 3

akj > 0 akj < 0 x1 + 4x2 + 7x3 £ 9
Example 10.1 Consider the LPP x1, x2, x3 ≥ 0
Minimize Z = 3x1 + 5x2 subject to Find the effect on the optimality of the current
3x1 + 2x2 £ 18 solution when the objective function is changed
x1 £ 14 to 3x1 + 8x2 + 5x3
x2 £ 6 Solution The optimal solution table obtained by
x1, x2 ≥ 0 the simplex method is
Discuss the change in cj on the optimal solution Table 10.2
of the problem.
Solution The optimal solution of the problem x1 x2 x3 s1 s2
(using the simplex method) is XB
CB B 4 6 2 0 0
x1 = 2, x2 = 6, Z* = 36 4 x1 1 0 –1 4/3 –1/3 1
The optimal table is
6 x2 0 1 2 –1/3 1/3 2
Table 10.1 cj 0 0 6 10/3 2/3 16
x1 x2 s1 s2 s3
XB Replacing the objective function with the new
CB B 3 5 0 0 0 one, we have
3 x1 1 0 1/3 0 –2/3 2
Table 10.3
0 s2 0 0 –2/3 1 4/3 0
5 x2 0 1 0 0 1 6 x1 x2 x3 s1 s2
XB
cj 0 0 1 0 3 36 CB B 3 8 5 0 0
3 x1 1 0 –1 4/3 –1/3 1
For the basic variable x1(k = 1) we get the
8 x 0 1 2 –1/3 1/3 2
condition
Max { –(Zj – cj)/a1j} £ D c1 £ Min { –(Zj – cj)/a1j } cj 0 0 8 4/3 5/3 19
a1j > 0 a1j < 0
–1/(1/3) £ Dc1 £ – 3/(–2/3) a13 = 1/3 c j = Zj – cj ≥ 0 " j. Therefore the optimal
–3 £ Dc1 £ 9/2 a15 = –2/3 solution remains the same
But a13 = 1/3 and a15 = –2/3 correspond to the x1 = 1, x2 = 2, Z* = 19
non-basic variables s1 and s3 and hence the current Note: We find that changes in the objective
optimal solution remains the same so long as
coefficients cj do not affect the feasibility of the LPP.
c1 £ 3 + 9/2 and c1 ≥ 3 – 3 since c1 = 3
i.e. 15/2 ≥ c1 ≥ 0 Example 10.3 Consider the LPP
For the other basic variable x2, k = 3 Maximize Z = 2x1 + x2 + 4x3 – x4 subject to the
we get constraints
–3/1 £ Dc2 £ – 1/–0 x1 + 2x2 + x3 – 3x4 £ 8
–3 £ Dc2 £ • and c2 = 5
–x2 + x3 + 2x4 £ 0
\ c2 £ 5 + • and c2 ≥ 5 – 3
i.e. 2 £ c2 £ • 2x1 + 7x2 – 5x3 – 10x4 £ 21
Thus the optimal solution is unaltered if 0 £ c1 x1, x2, x3, x4 ≥ 0
£ 15/2 and 2 £ c2 < • Discuss the effect of changing (c1 c2 c3 c4) to
Example 10.2 Consider the LPP (1 2 3 4 ) on the optimal solution.
Maximize Z = 4x1 + 6x2 + 2x3 subject to the Solution The optimal table obtained by the
constraints simplex method is
 Sensitivity Analysis 105

Table 10.4 This solution is optimal. The new optimal

solution is
x1 x2 x3 x4 s1 s2 s3
XB x1 = 0, x2 = 21/2, x3 = 11/10, x4 = 47/10,
CB B 2 1 4 –1 0 0 0
Z* = 431/10
2 x1 1 0 3 1 1 2 0 8
10.2 DISCRETE CHANGES IN bi
1 x2 0 1 –1 –2 0 –1 0 0
0 s3 0 0 –4 2 –2 3 1 5 We know that the optimal solution is given by XB
= B–1 b where B–1 is the matrix of coefficients corres-
cj 0 0 1 1 2 3 0 16 ponding to the slack variables in the optimal table.
For the new objective function the table If the resource bk is changed to bk + Dbk then
becomes b¢ = (b1, b2, ….. bk + Dbk, …,bm)T
æ b1 ö
Table 10.5 çb ÷ æ 0 ö
ç 2÷ ç 0 ÷
x1 x2 x3 x4 s1 s2 s3 ç . ÷ ç ÷
CB B 1 2 3 4 0 0 0
XB \ b¢ = ç ÷ + ç D bk ÷ = b + Db
1 x1 1 0 3 1 1 2 0 8 ç bk ÷ ç 0 ÷
ç . ÷ çç ÷
÷
2 x2 0 1 –1 –2 0 –1 0 0 ç ÷ è 0 ø
è bm ø
0 s3 0 0 –4 2 –2 3 1 5
Now XB¢ = B–1 (b + Db)
cj 0 0 –2 –7 1 0 0 8 = B–1 b + B–1 Db
R1 – (1/2)R3, R2 + R3, R3/2 æ 0 ö
The solution is not optimal. In the iteration x4 ç 0 ÷
ç ÷
-1 ç . ÷
enters and s3 leaves the basis
= XB + B ç ÷
Table 10.6 ç D bk ÷
ç 0 ÷
I Iteration ç ÷
x1 x2 x3 x4 s1 s2 s3 è 0 ø
XB = XB + bk(Dbk)
CB B 1 2 3 4 0 0 0
where bk is the kth row of B–1
1 x1 1 0 5 0 2 1/2 –1/2 11/2
The new value of Z is
2 x2 0 1 –5 0 –2 2 1 5 Z¢ = CB XB¢
4 x4 0 0 –2 1 –1 3/2 1/2 5/2 = CBXB + CB B–1Dbk
cj 0 0 –16 0 –6 21/2 7/2 51/2 = Z + CB B–1 (Dbk)
Note: If we get bi negative then we can apply
R1/5, R2 + R1, R3 + (2/5)R1 the dual simplex method.

Table 10.7
Example 10.4 Consider the LPP
Maximize Z = 4x1 + 6x2 + 2x3 subject to the
II Iteration constraints
3 x3 1/5 0 1 0 2/5 1/10 –1/10 11/10 x1 + x2 + x3 £ 3
2 x2 1 1 0 0 0 5/2 1/2 21/2 x1 + 4x2 + 7x3 £ 9
4 x4 2/5 0 0 1 –1/5 17/5 3/10 47/10
x1, x2, x3 ≥ 0
Discuss the effect on the optimal solution when
c j 16/5 0 0 0 2/5 189/10 19/10 431/10 the constants are changed from (3, 9) to (9, 6).
 106 Operations Research

Solution The optimal solution obtained by the Solution By using the big M method the
simplex method is following optimal solution is obtained

Table 10.8 Table 10.10

x1 x2 x3 s1 s2 x1 x2 x3 s1 s2
XB XB
CB B 4 6 2 0 0 CB B 5 12 4 0 –M
4 x1 1 0 –1 4/3 –1/3 1 12 x2 0 1 –1/5 2/5 –1/5 8/5
6 x2 0 1 2 –1/3 1/3 2 5 x1 1 0 7/5 1/5 2/5 9/5
cj 0 0 6 10/3 2/3 16 cj 0 0 3/5 29/5 M–2/5 141/5

x1 = 1, x2 = 2, x3 = 0, Z*=16 The solution is x 1 = 9/5, x 2 = 8/5, x 3 = 0,

If new values of the constants are (9, 6) then in Z*= 141/5
the optimal table the new values of x1 and x2 are æ 7ö
given by If the requirement vector becomes ç ÷ then
è 2ø
æ x1 ö æ 4 / 3 -1 / 3ö æ 9 ö æ 10 ö æ 2 / 5 -1 / 5ö æ 7 ö æ 12 / 5ö
çè x ÷ø = çè -1 / 3 1 / 3 ÷ø çè 6 ÷ø = çè -1÷ø X B = B -1b = ç =
2 è 1 / 5 2 / 5 ÷ø çè 2 ÷ø çè 11 / 5 ÷ø
Since x2 is negative we apply the dual simplex Since x1 ≥ 0 and x2 ≥ 0 this solution is feasible.
method The new optimal solution is
Table 10.9 x1 = 11/5, x2 = 12/5, x3 = 0
x1 x2 x3 s1 s2 Z* = 5(11/5) + 12(12/5) + 4(0) = 199/5.
XB
CB B 4 6 2 0 0
10.3 CHANGES IN THE MATRIX
4 x1 1 0 –1 4/3 –1/3 10
ELEMENTS aij
6 x2 0 1 2 –1/3 1/3 –1
Let the coefficient ark be changed to ark + Dark.
cj 0 0 6 10/3 2/3 34 Two cases arise:
R1 + 4R2, R2(–3) (i) column ak is a non-basic column.
(ii) column ak is basic.
4 x1 1 4 7 0 1 6
Case (I) If ak is non-basic (the column ak is not
0 S1 0 –3 –6 1 –1 3
in the basis matrix) we find XB = B–1b and hence
cj 0 10 26 0 4 24 the only effect of the change is on the optimality
condition. The optimality condition is Zk¢ – ck ≥ 0,
The new optimal solution is x1 = 6, x2 = 0, where Zk¢ = CBB–1ak¢
x3 = 0, Z*= 24. \ Zk¢ = CBB–1(ak + Dak)
Example 10.5 Consider the LPP = CBB–1ak + CBB–1(D ark)
Maximize Z = 5x1 + 12x2 + 4x3 subject to the =>Zk + CBbk (Dark)
constraints Zk¢ – ck ≥ 0 => Zk + CBbk (Dark) – ck ≥ 0
x1 + 2x2 £ 5 i.e. Max {–(Zk – ck)/CBbk; CBbk > 0} £ Dark £
5x1 – x2 + 2x3 = 2 Min {–(Zk – ck)/CBbk; CBbk > 0}
x1, x2, x3 ≥ 0 If this condition is satisfied then the current
Analyze the effect of changing the requirement solution continues to be optimal.
æ 5ö æ 7ö Case (II) If ak is basic, we define
vector from ç ÷ to ç ÷ on the optimal solution. l = (Zj – cj) bkk – ykj CBbk
è 2ø è 2ø
 Sensitivity Analysis 107

The condition for the current optimal solution 10.4 ADDITION OF A VARIABLE
to remain optimal, is
Max {–(Zj – cj)/l) £ Dark = Min {–(Zj – cj)/l} Suppose we add a new variable with coefficients
l>0 l<0 column an+1 and cost cn+1 in the objective function.
We compute
Example 10.6 Consider the LPP yn + 1 = B–1an + 1
Maximize Z = 5x1 + 12x2 + 4x3 subject to the and Zn + 1 – cn + 1 = CB yn + 1 – cn + 1
constraints If Zn + 1 – cn + 1 ≥ 0 the current solution continues
x1 + 2x2 + x3 £ 5 to be optimal.
2x1 – x2 + 3x3 = 2 If Zn + 1 – cn + 1 < 0 apply the simplex method
x1, x2, x3 ≥ 0 and obtain a new optimal solution.
Discuss the effect of changing the column
Example 10.7 Consider the LPP
æ 1ö æ -5ö Maximize Z = 3x1 + 5x2 subject to the constraints
a3 = ç ÷ to ç ÷
è 3ø è 2ø 3x1 + 2x2 £ 18
on the optimal solution. x1 £ 4
Solution Applying big-M method we get the x2 £ 6
x1, x2 ≥ 0
optimal solution as
Discuss the change in the optimal solution when
Table 10.11 a new variable x3 is introduced with cost coefficient
2 and
x1 x2 x3 s1 R1
XB æ 1ö
CB B 5 12 4 0 –M
a3 = çç 1÷÷
12 x2 0 1 –1/5 2/5 –1/5 8/5
è 1ø
5 x1 1 0 7/5 1/5 2/5 9/5
Solution Introducing slack variables s1, s2, s3 and
cj 0 0 3/5 29/5 M–2/5 141/5 applying the simplex method we get the optimal
solution.
æ -5ö Table 10.13
When the x3 column is changed to ç ÷ the
è 2ø x1 x2 s1 s2 s3
rd
XB
new 3 column is CB B 3 5 0 0 0
3 x1 1 0 1/3 0 –2/3 2
æ a13¢ ö æ 2 / 5 -1 / 5ö æ -5ö æ -12 / 5ö
= = 0 s2 0 0 –2/3 1 4/3 0
èç a23¢ ø÷ èç 1 / 5 2 / 5ø÷ èç 2 ø÷ èç -1 / 5ø÷ 5 x2 0 1 0 0 1 6
The table becomes Cj 0 0 1 0 3 36

Table 10.12 The solution is x1 = 2, x2 = 6, Z*=36

After the addition of a new variable x3 the
x1 x2 x3 s1 R1 problem becomes
XB
CB B 5 12 4 0 –M Maximize Z = 3x1 + 5x2 + 2x3 subject to the
12 x2 0 1 –12/5 2/5 –1/5 8/5 constraints
3x1 + 2x2 + x3 £ 18
5 x1 1 0 –1/5 1/5 2/5 9/5
x1 + x3 £ 4
Cj 0 0 –169/5 29/5 M–2/5 141/5 x2 + x3 £ 6
All the elements of the pivot column are x1, x2, x3 ≥ 0
negative and hence the problem has unbounded The coefficient column of x3 in the optimal table
solution. is given by
 108 Operations Research

y3 = B–1a3
æ yj ö
æ 1 / 3 0 -2 / 3ö æ 1ö æ -1 / 3ö = ç ÷
è m +1, j
a - y jø
= ç -2 / 3 1 4 / 3 ÷ ç 1÷ = ç 5 / 3 ÷
a
ç ÷ç ÷ ç ÷
è 0 0 1 ø è 1ø è 1 ø Now Zj* – cj = Z*yj* – cj
Z3 – c3 = CBy3 – c3 æ yj ö
æ -1 / 3ö = [CB CBm +1 ] ç - cj
è am +1, j - a y j ÷ø
= (3 0 5 ) çç 5 / 3 ÷÷ - 2 > 0
= CByj + CBm+1 am+1, j – CB m+1 a yj – cj
è 1 ø
Hence the current solution itself is optimal. where CBm+1 is the cost associated with the new
variable introduced in the basis. Since the
10.5 ADDITION OF A new variable is a slack or surplus variable its
CONSTRAINT coefficient in the objective function is zero.
Therefore, CBm+1= 0
If a new constraint am + 1, 1 x1 + am + 1, 2 x2 +… +
a m + 1, n x n £ b m + 1 is added to the system of \ Zj*–cj = CB yj – cj = Zj – cj
constraints of an LPP then two cases arise. Therefore the values of c j are the same for the
Case (I) The optimal solution of the original prob- new problem also.
lem satisfies the new constraint. In this case the cur- æX ö
rent optimal solution continues to be optimal. Also Z * = [CB 0 ] ç B ÷ = CB X B = Z
è s ø
Case (II) The optimal solution XB of the original
Thus the optimum value remains unchanged.
problem does not satisfy the new constraint. In
Since the optimal solution of the original problem
this case we obtain a new optimal solution to the
modified problem. Let B be the basis matrix for does not satisfy the new constraints the slack or
the original problem and B1 be the basis matrix surplus variable possesses negative value. Hence
for the new problem with m + 1 constraints. we have to use the dual simplex method and obtain
the new optimal solution.
æ B 0ö
\ B1 = ç Note: If the new constraint is an equation then
è a 1 ÷ø an artificial variable appears in the basis at
where a = (am + 1, 1 am + 1, 2 …am + 1, n) and the 2nd a positive value and Big-M method can be
æ 0ö applied.
column ç ÷ corresponds to the slack or surplus
è 1ø Example 10.8 Consider the LPP
variable added in the new constraint. Maximize Z £ 3x1 + 5x2 subject to the constraints
æ B -1 0ö 3x1 + 2x2 £ 18
Now B1–1 = ç ÷ x1 £ 4
è -a B -1 1 ø
x2 £ 6
and yj* = B1–1aj* x1, x2 ≥ 0
æ B -1 0ö æ a j ö Discuss the effect of adding a new con-
= ç ÷ç ÷ straint 2x 1 + x 2 £ 8 to the given set of con-
-1
è -a B 1 ø è am +1, j ø straints.
æ B -1a j ö Solution The given problem has an optimal
= ç -1 ÷ simplex table as given below.
è -a B a j + am +1, j ø
 Sensitivity Analysis 109

Table 10.14 Table 10.16

x1 x2 s1 s2 s3 x1 x2 s1 s2 s3 s4
XB XB
CB B 3 5 0 0 0 CB B 3 5 0 0 0 0
3 x1 1 0 1/3 0 –2/3 2 3 x1 1 0 1/3 0 –2/3 0 2
0 s2 0 0 –2/3 1 4/3 0 0 s2 0 0 –2/3 1 4/3 0 0
5 x2 0 1 0 0 1 6 5 x2 0 1 0 0 1 0 6
cj 0 0 1 0 3 36 0 s4 0 0 –2/3 0 1/3 1 –2

The optimal solution is x1 = 2, x2 = 6, Z* = 36. cj 0 0 1 0 3 0 36

The new constraint is 2x1 + x2 £ 8. s4 leaves and s1 enters the basis. (Apply the dual
This constraint is not satisfied by the current simplex method)
solution. Hence we add a slack variable s4 to this Take R11 = R1+(1/2)R4
constraint. The new table is R21 = R2 – R4
R41 = R4 (–3/2)
Table 10.15
The new table becomes
x1 x2 s1 s2 s3 s4
XB Table 10.17
CB B 3 5 0 0 0 0
3 x1 1 0 1/3 0 –2/3 0 2 x1 x2 s1 s2 s3 s4
XB
0 s2 0 0 –2/3 1 4/3 0 0 CB B 3 5 0 0 0 0
5 x2 0 1 0 0 1 0 6 3 x1 1 0 0 0 –1/2 1/2 1
0 s4 2 1 0 0 0 1 8 0 s2 0 0 0 1 1 –1 2
cj 0 0 1 0 3 0 36 5 x2 0 1 0 0 1 0 6
0 s1 0 0 1 0 –1/2 –3/2 3
The basis matrix has been disturbed due to the Cj 0 0 0 0 7/2 3/2 33
4th row. a41 and a42 coefficients must be reduced
to zero. Take The new optimal solution is
R41 = R4 – R3 – 2R1 x1 = 1, x2 = 6, Z* = 33
The table becomes Note that the introduction of a new constraint
has reduced the optimum Z from 36 to 33.

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. In the LPP c1 can be changed without affecting the

Maximize Z = 15x1 + 45x2 subject to the optimal solution?
constraints 2. For the LPP
x1 + 16x2 £ 250 Maximize Z = – x1 + 2x2 – x3 subject to the
5x1 + 2x2 £ 162 constraints
3x1 + x2 – x3 £ 10
x2 £ 50
–x1 + 4x2 + x3 ≥ 6
x1, x2 ≥ 0 x2 + x3 £ 4
If c2 is kept fixed at 45 determine how much x1, x2, x3 ≥ 0
 110 Operations Research

Determine the ranges for discrete changes in (2,2,2) T and coefficient in the objective
b1 and b2 in order to maintain the optimality function, 5.
of the current solution. 7. Solve the LPP
3. Consider the problem, Maximize Z = 4x1 + 6x2 + 2x3 subject to the
Maximize Z = 3x 1 + 5x 2 subject to the constraints
constraints x1 + x2 + x3 £ 3
3x1 + 2x2 £ 18 x1 + 4x2 + 7x3 £ 9
x1 £ 4 x1, x2, x3 ≥ 0
x2 £ 6 (i) What is the new optimal solution when
x1, x2 ≥ 0 c3 is increased from 2 to 12?
Discuss the effect on the optimality of the (ii) Find the effect of changing the objective
solution when the objective function is function to Z = 2x1 + 8x2 + 4x3
changed to 3x1 + x2. 8. Consider the LPP
4. Consider the LPP Maximize Z = 45x1 + 100x2 + 30x3 + 50x4
Maximize Z = 3x 1 + 5x 2 subject to the subject to the constraints
constraints 7x1 + 10x2 + 4x3 + 9x4 £ 1200
x1 £ 4 3x1 + 40x2 + x3 +x4 £ 800
3x1 + 2x2 £ 18 x1, x2 , x3, x4 ≥ 0
x1, x2 ≥ 0 Find the change in the optimal solution when
Solve the LPP. If a new variable x3 is added a new variable x 5 is added with cost
æ 1ö æ 10 ö
to the problem with column ç ÷ and cost coefficient 120 and column ç ÷
è 2ø è 10 ø
coefficient 7 find the change in the optimal 9. Solve the LPP
solution. Maximize Z = 3x 1 + 5x 2 subject to the
5. In the problem constraints
Maximize Z = 5x1 + 12x2 + 4x3 subject to the x1 £ 4
constraints 3x1 + 2x2 £ 18
x1 + 2x2 + x3 £ 5 x1, x2 ≥ 0
2x1 + x2 + 3x3 = 2 Discuss the change in the optimal solution
x1, x2, x3 ≥ 0 when
Discuss the change in the optimal solution (i) a constraint x2 £ 10 is added
æ 1ö æ 2ö (ii) a constraint x2 £ 6 is added.
when a3 is changed from ç ÷ to ç ÷
è 3ø è 5ø 10. Solve
6. Given the LPP Maximize Z = 3x1 + 2x2 + 5x3 subject to the
Minimize Z = 3x 1 + 8x 2 subject to the constraints
constraints x1 + 2x2 + x3 £ 430
x1 + x2 = 200 3x1 + 2x2 £ 460
x1 £ 80 x1 + 4x2 £ 420
x2 ≥ 60 x1, x2, x3 ≥ 0
x1, x2 ≥ 0 Find the new optimal solution when x 1
discuss the effect on the optimality by adding column is changed to (1,1,6)T , c2 = 1 and
a new variable x3 with column coefficients c3 = 3.

ANSWERS
■ ■ ■ ■ ■ ■ ■
1. 15 – (195/16) £ c1 £ 15 + 195/16.
2. b2 < 10 –5/2 £ b3 £ 6
3. The original optimal solution is
x1 = 2, x2 = 6, Z* = 36
The new optimal solution is
x1 = 2, x2 = 6, Z* = 12
4. The optimal solution of the original problem
is
x1 = 0, x2 = 9, Z* = 45
For the modified problem, the solution is
x1 = 0, x2 = 5, x3 = 4, Z* = 53
5. There is no change in the current optimal
solution.
6. The original optimal solution is
x1 = 80, x2 = 120, Z* = 1200
The optimal solution of the new problem is
x1 = 20, x2 = 120, x3 = 70, Z* = 1120
7. The original optimal solution is
x1 = 1, x2 = 2, x3 = 0, Z* = 16
Sensitivity Analysis 111
If c3 becomes 12, the solution changes to
x1 = 2, x2 = 0, x3 = 1, Z* = 20
If the new objective function is 2x1 + 8x2 +
4x3 the solution becomes
x1 = 1, x2 = 2, x3 = 0, Z* = 18
8. The original optimal solution is
x1 = 0, x2 = 40/3, x3 = 800/3, x4 = 0,
Z* = 28000/3
The new optimal solution is
x1 = 0, x2 = 0, x3 = 400/3, x4 = 0, x5 = 200/3,
Z* = 12000
9. The original optimal solution is x1 = 0, x2 = 9,
Z* = 45
(i) There is no change in the optimal
solution.
(ii) The new solution is x1 = 2, x2 = 6, Z* = 36.
10. The original optimal solution is x 1 = 0,
x2 = 100, x3 = 230, Z* = 1350
The new solution is x1 = 4, x2 = 99, x3 = 228,
Z* = 795
 11
Transportation Problem

CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■

Transportation problem deals with the trans- i.e. xi1 + xi2 + xi3 + … + xin = ai (1)
portation of commodity from different sources to (i = 1, 2, 3, …, m)
different destinations. Similarly, the total demand at the destination
Given m sources and n destinations, the total Dj is bj.
quantity available at each source, the total quantity i.e. x1j + x2j + x3j + … xmj = bj (2)
required at each destination, and the costs of ( j = 1, 2, 3, …, n)
transportation of unit commodity from each source m
to each destination, it is required to determine the Also the total availability å ai and the total
quantity of commodity to be transported from each i =1
n
source to different destinations such that the total
cost of transportation is minimum. This is demand å bj must be equal
j =1
transportation problem.
m n

11.1 MATHEMATICAL i.e. å ai = å bj (3)

i =1 j =1
FORMULATION These are the constraints to be satisfied. Also
Suppose there are m sources S1, S2,…, Sm and n there is the non-negativity constraint
destinations D 1, D 2, …, D n. Let the cost of xij ≥ 0 (4)
transporting one unit of commodity from Si to Dj We form a rectangular transportation table
be cij and let xij be the quantity to be transported where the quantity xij is put in the (i, j)th cell. The
from Si to Dj. Then the problem is unit cost cij is written in a corner of the (i, j)th cell.
m n

Minimize Z =
m n
å å cij xij If å ai = å b j then the problem is called a
i =1 j =1
i =1 j =1
balanced problem. In order to obtain a feasible
The total availability at the source Si is ai solution we need a balanced problem.

Table 11.1
æm n ö
ç å ai = å b j ÷
è i =1 j =1 ø
A set of non-negative values x ij which
represents the quantity of commodity transported
from the ith source to the jth destination (i = 1, 2, 3,
…, m; j = 1, 2, 3, …, n) which satisfy the
constraints (1) (2) (3) and (4) is called a feasible
solution to the transportation problem.
While filling up each row we have (n – 1)
degrees of freedom and for each column we have
(m – 1) degrees of freedom. Hence out of the mn
cells of the table we get mn – (m – 1) (n – 1)
independent allotments. In other words the
maximum independent allotments of quantity is
mn – (m – 1) (n – 1) = m + n – 1
Thus there are (m + n – 1) values of xij > 0 the
remaining cells have xij = 0. These (m + n – 1)
cells with xij > 0 are called basic cells. In order to
solve the transportation problem we must have
(m + n – 1) basic cells. If there are less than
(m + n – 1) basic cells, it is called a degenerate
problem.
To solve a non-degenerate transportation
problem we need to start with a basic feasible
solution, then test the initial solution for optimality,
improve that solution if possible and finally we
arrive at an optimal solution.
11.2 INITIAL BASIC FEASIBLE
SOLUTION
There are different methods of obtaining an initial
solution of a given transportation problem.
Transportation Problem 113
11.2.1 North-West Corner Rule
Prepare the transportation table, consider the north-
west corner cell (1, 1) of the table. The availability
at the source S1 is a1 and the demand at the
destination D1 is b1. Choose the minimum of a1
and b1, say a1. Allot the quantity a1 to the (1, 1)
cell. This is the maximum quantity that can be
allotted to the (1, 1) cell. Having allotted a1 to the
(1, 1) cell, we find that the total availability at S1
is exhausted. Hence no more allotment can be
made in the first row. Therefore x11 = a1 and x12 =
x13 = … = x1n = 0. Now the demand at D1 becomes
b1 – a1. In the table the north-west corner cell is
(2, 1). As before choose the minimum of b1 – a1
and a2 and allot that quantity in the (2, 1) cell.
Suppose b1 – a1 is the minimum then allot b1 – a1
to the (2, 1) cell and assign zero values to
the remaining cells of the first column. Now
the availability at S2 becomes a2 – (b1 – a1). In
the new table (2, 2) cell becomes the north-west
corner cell and allot maximum possible quantity
to this cell. Proceeding like this, after a finite
number of steps we arrive at an initial basic
feasible solution. In order that it is a non-
degenerate solution we must have (m + n – 1) basic
cells.
Example 11.1 Obtain an initial basic feasible
solution, using the north-west corner rule for the
following transportation problem:
Solution In the table Sai = 30 + 40 + 50 = 120
Sbj = 35 + 28 + 32 + 25 = 120
\ Sai = Sbj
It is a balanced problem.
Min (a1, b1) = min (30, 35) = 30.
Allot 30 units to the north-west corner
cell (1, 1). Thus the availability at S1 is exhausted.
No more allotment can be made in the first
row.
 114 Operations Research

(Rs)
Now (2, 1) cell is the north-west corner cell.
S1 Æ D1 30 units : cost 30 × 6 = 180
The maximum quantity that can be allotted is the
minimum of 40 and 5. Thus we can assign 5 units S2 Æ D1 5 units : cost 5 × 5 = 25
to the (2, 1) cell and no more allotment is possible S2 Æ D2 28 units : cost 28 × 11 = 308
in the first column. S2 Æ D3 7 units : cost 7 × 9 = 63
S3 Æ D3 25 units : cost 25 × 7 = 175
S3 Æ D4 25 units : cost 25 × 13 = 325
Total cost = Rs 1076
Here x11 = 30, x21 = 5, x22 = 28, x23 = 7, x33 = 25,
and x34 = 25.
All other xij are zeroes.
Also m + n – 1 = 3 + 4 – 1 = 6. We have 6 basic
cells.
Now choose the north-west corner cell of Example 11.2 Find an initial basic feasible
the reduced table which is the (2, 2) cell. Allot solution to the following transportation problem
28 units which is the minimum of (35, 28). We get using north-west corner rule:

Solution As per north-west corner rule the

first allotment goes to (1, 1) cell. The maxi-
Next allotment is min(32, 7) to the (2, 3) cell. mum quantity that can be allotted is min
Therefore we allot 7 units to the (2, 3) cell. The (7, 5) = 5. With this allotment the table becomes
reduced table is

Now allot 25 units to the (3, 3) cell and the The next allotment is for the (2, 1) cell and
remaining 25 to the (3, 4) cell. We obtain a basic min(8, 2) = 2 and hence we can allot 2 units there.
feasible solution as Now the table reduces to
 Transportation Problem 115

Now we consider the cell (2, 2) which is the
north-west corner cell and allot 6 units there since
min(6, 9) = 6. The resulting table is
The north-west corner of the table is the (3, 2)
cell for which a maximum of 3 units can be
allotted. We get the table
O3 Æ D2 3 units : cost 3 ¥ 4 = 12
O3 Æ D3 4 units : cost 4 ¥ 7 = 28
O4 Æ D3 14 units : cost 14 ¥ 2 = 28
Total cost = Rs 102
11.2.2 Row Minima Method
In the transportation table consider the cell having
the least cost in the first row and allot the maximum
possible quantity to that cell. Then we take the
cell containing the next higher cost and allot the
maximum possible quantity. If there is a tie
between the cells choose the cell of your choice.
Repeat the process till the availability a 1 is
exhausted. Now take the second row and allot the
quantities starting from the cell having the least
cost. After exhausting a2 consider the third row
and so on. Finally we get the initial solution when
all the rows have been filled up.
Example 11.3 Obtain an initial basic feasible
solution to the following transportation problem
using the row minima method.

Solution Sai = Sbj = 35. Hence it is a balanced

problem. The least cost in the first row is 1. Allot
min (14, 15) = 14 units to the (1, 3) cell. We get

Finally we assign 4 units to the (3, 3) cell which

leads to the solution satisfying all the constraints.
The resulting initial basic feasible solution is

In the 2nd row the least cost is 2 and we can

allot 1 unit [min (16, 1)] to the (2, 3) cell. This
gives

(Rs)
O1 Æ D1 5 units : cost 5 ¥ 2 = 10
O2 Æ D1 2 units : cost 2¥3 = 6
O2 Æ D2 6 units : cost 6 ¥ 3 = 18
 116 Operations Research

The next higher cost in the 2nd row is 7 and 4
units can be allotted there. We get
exhausted the first column go to the 2nd column
and repeat the process. Finally after making the
allotments in the last column we obtain an initial
basic feasible solution.
Example 11.4 Obtain an initial basic feasible
solution to the following transportation problem
using the column minima method.

The next higher cost in the 2nd row is 8 and we

allot 6 units to the cell (2, 1)

Solution Sai = Sbj = 16. Therefore the problem

Now we can allot 5 units to the (2, 2) cell and 5
units to the (3, 2) cell which gives the initial
solution. The resulting solution is
is a balanced one. We can find a basic feasible
solution. Take the first column. The least cost is 4
in the (2, 1) cell. We make the allotment 1 in that
cell. We get the reduced table.

Consider the 2nd column. The least cost is 5 in

the (3, 2) cell. An allotment of 8 units can be made
there. The resulting table is
(Rs)
O1 Æ D3 14 units : cost 14 ¥ 1 = 14
O2 Æ D1 6 units : cost 6 ¥ 8 = 48
O2 Æ D2 5 units : cost 5 ¥ 9 = 45
O2 Æ D3 1 unit : cost 1¥2 = 2
O2 Æ D4 4 units : cost 4 ¥ 7 = 28
O3 Æ D2 5 units : cost 5 ¥ 3 = 15
Total = Rs 152 In the third column the least cost is 8 and an allot-
Also m + n – 1 = 6 and we have 6 basic cells. ment of 1 unit can be made there. The table becomes
Hence it is a non-degenerate solution.
11.2.3 Column Minima Method
This method is the same as the row minima method
except that we consider the columns one by one
instead of rows. Choose the first column and allot
the maximum possible quantity to the cell having
the least cost. Then consider the cell with the next We can allot 2 units for (1, 3) cell and 4 units
higher cost and allot the available quantity. Having for (2, 3) cell. This completes the allocation. There
 Transportation Problem 117

are m + n – 1 = 5 basic cells. The initial basic The next higher cost is 2 in the (2, 3) cell where
feasible solution obtained is we can allot 10 units. Now the table becomes

(Rs) We can assign 3 units to the (3, 1) cell having

S1 Æ D3 2 units : cost 2 ¥ 15 = 30 the next higher cost 3. Therefore we have
S2 Æ D1 1 unit : cost 1¥4 = 4
S2 Æ D3 4 units : cost 4 ¥ 16 = 64
S3 Æ D2 8 units : cost 8 ¥ 5 = 40
S3 Æ D3 1 unit : cost 1¥8 = 8
Total cost = Rs 146

11.2.4 Least Cost Method

(Matrix Minima Method) Now allot 2 units to the (2, 2) cell having the
next higher cost 4. Finally allot the remaining
Choose the cell in the table having the least cost
8 units to the (3, 2) cell. The initial solution
altogether and allot the maximum possible quantity obtained is
to that cell. Then consider the cell having the next
higher cost in the table and make the allotment there.
Use your choice when there is a tie between cells.
Make the necessary alterations in ai and bj accor-
dingly. Repeating the process for a finite number of
times we arrive at an initial basic feasible solution.
Example 11.5 Obtain an initial basic feasible (Rs)
solution to the following transportation problem S1 Æ D1 : 7 units cost 7 ¥ 1 = 7
using the least cost rule S2 Æ D2 : 2 units cost 2 ¥ 4 = 8
S2 Æ D3 : 10 units cost 10 ¥ 2 = 20
S3 Æ D1 : 3 units cost 3 ¥ 3 = 9
S3 Æ D2 : 8 units cost 8 ¥ 7 = 56
Total cost = Rs 100
11.2.5 Vogel’s Approximation
Solution Sai = Sbj = 30. Hence the problem is
balanced. The least cost in the table is 1 in the Method (VAM)
(1, 1) cell. Hence we make the first allotment of Vogel’s approximation method gives a very good
7 units there. The reduced table is initial solution. In many cases this initial solution
itself becomes an optimal solution.
Take the first row and note down the difference
between the least and the next higher cost in that
row. Then take the second row and note down the
difference between the least and the next higher
cost there. Repeat this for all the remaining rows.
 118 Operations Research

Then consider the columns one by one and note

down the difference in each column. Equal costs
should be taken as a single cost in any row or
column. We get (m + n) such values. Now choose
the largest among these numbers. If the largest
difference corresponds to the ith row then make an
allotment in the ith row in the cell having the least
cost in that row, say (i, j)th cell. Allot the minimum
of ai and bj to that cell. After making necessary The difference between the least and the
modifications in the availability and demand next higher cost is noted down for each row and
constraints, consider the row or column having the column of the reduced table. The maximum
largest difference in the reduced table. Repeat this difference is (5) in the third column. The least cost
process till all the available quantity are exhausted in the third column is 6 in the (2, 3) cell for which
and the demands are satisfied. If any two rows or 1 unit can be assigned. With this allotment the
columns have the same maximum difference then revised table is given below:
break the tie at your choice. Finally we obtain an
initial basic feasible solution.
Example 11.6 Obtain an initial basic feasible
solution to the following transportation problem
using Vogels approximation method:

(5) is the maximum difference corresponding

to the second column and the least cost is 3 in
(1, 2) cell. Hence we can allot 6 units to this cell
Solution The transportation table is Note that a1 = b2 = 6. The resulting table is

Now allot 8 units to the (3, 1) cell and the re-

In the first row the difference between the least maining 2 units to the (3, 3) cell. This yields the
and the next higher cost is 3 – 2 = (1). In the second initial solution to the given problem as given below
row the difference is 5 – 1 = (4). In the third row
the difference is 8 – 5 = (3). Similarly, the
differences in the columns are (1), (2), (5) and (6).
The maximum of all these is (6) in the 4 th
column. In the 4th column the least cost is 1 in
(2, 4) cell. Hence we allot 3 units to this cell and
S1 Æ D2 : 6 units
cancel the other cells of the 4th column. The
S2 Æ D3 : 1 unit
reduced table is
 Transportation Problem 119

S2 Æ D4 : 3 units Total cost is 18 + 6 + 3 + 40 + 30 = Rs 97.

S3 Æ D1 : 8 units Note that there are only 5 basic cells but
S3 Æ D3 : 2 units m + n – 1 = 6. Hence it is a degenerate solution.

EXERCISES
■ ■ ■ ■ ■ ■ ■

Obtain an initial basic feasible solution using Obtain an initial basic feasible solution using
north west corner rule: least cost rule.
6.
1.

2. 7.

8.
3.

Apply VAM to obtain an initial solution:

4.
9.

5. Obtain an initial basic feasible solution using

the row minima method
10.
 120 Operations Research

11. 12.

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. O1 – D1 : 5 ; O2 – D1 : 2; O2 – D2 : 6; 7. O1 – D2 : 6; O2 – D3 : 2; O2 – D4 : 6;
O3 – D2 : 3; O3 – D3 : 4; O4 – D3 : 14 O3 – D1 : 4; O3 – D3 : 6;
Total cost = Rs 102 Total cost = Rs 28
2. F1 – M1: 30; F2 – M1 : 5; F2 – M2: 28 8. A – W1 : 20; A – W3 : 10 ; B – W2 : 20;
F2 – M3 : 7; F3 – M3 : 25; F3 – M4 : 25 B – W3 : 20; B – W4 : 10; C – W2 : 20
Total cost = Rs 1076 Total cost = Rs 180
3. O1 – D1 : 6 ; O1 – D2 : 8; O2 – D2 : 2; 9. A – D1 : 5; B – D3 : 8; C – D2 : 7
O2 – D3 : 14 O3 – D3 : 1; O3 – D4 : 4 D – D1 : 2; D – D2 : 2; D – D3 : 10
Total cost = Rs 128 Total cost = Rs 80
4. O1 – D1 : 6; O2 – D1 : 1; O3 – D2 : 5; 10. S1 – D1 : 200; S1 –D2 : 50; S2 – D2 : 175;
O3 – D3 : 3; O3 – D4 : 2 S2 – D4 = 125; S2 – D3 : 275; S3 – D4 : 125
Total cost = Rs 116 Total cost = Rs 12075
5. I –B : 1; II – A : 2; II – B : 1; III – A : 2 11. A – W2 : 25; A – W3 : 9; B – W1 : 15
III – C : 2 Total cost = Rs 855 C–W1 = 4; C –W3 : 8; D – W1 : 2
6. O1 – D2 : 2; O1 – D3 : 3; O2 – D3 : 8; D – W4 : 17 Total cost = Rs 195
O3 – D2 : 7; O4 – D1 : 7; O4 – D3 : 7 12. P1 – C1 : 1; P1 – C2 : 5; P2 – C4 : 1
Total cost = Rs 83 P3 –C1: 6; P3 – C3 : 3; P3 – C4 : 1
Total cost = Rs 102

11.3 TESTING FOR OPTIMALITY
We can find an initial solution to a transportation
problem, using any one of the methods described
previously. The next step is to examine whether
the solution obtained is optimal or not.
We introduce simplex multipliers (dual
variables) ui (i = 1, 2, 3, …., m) and vj (j = 1, 2, 3,
…, n) for the m rows and n columns with the
condition that they should satisfy the equations
ui + vj = cij where cij is the cost of transportation
in the (i,j)th cell which is a basic (occupied) cell.
For (m + n – 1) basic cells we get (m + n – 1)
equations and there are (m + n) variables u1, u2,
…, um, v1, v2 ..., vn. To solve these equations we
have to reduce the number of variables to
(m + n – 1). Hence we assign zero value to one of
the simplex multipliers. In order to make the
procedure easy, we assign zero value to the simplex
multiplier which corresponds to the row or column
having the maximum number of basic cells.
Now we have (m + n – 1) variables and they
can be solved and the values of u1 and vj (for each
i and j) can be obtained.
Next we calculate the relative cost factors cij
for each non-basic cell using the formula
cij = cij – ui – vj
 Transportation Problem 121

If all the cij are non-negative then the current

solution is optimal and the total transportation cost
obtained is the least and cannot be reduced further.
On the other hand, if some cij are negative then it
indicates that the current solution is not optimal
and that it can be improved further.
The second row and fourth column have the
11.4 OPTIMIZATION—MODI
maximum difference 2. Select one of them, say
METHOD (MODIFIED 2nd row. Allot 35 units to the cell (2, 1) having the
DISTRIBUTION METHOD) least cost. Adjust the constraints and note down
the differences in the next table. We get
Select the most negative cij . The corresponding
(i, j) cell is the entering cell for the improved
solution. Let us allot a quantity q to this cell. Draw
a loop (figure having horizontal and vertical sides)
having one vertex in the entering cell and all the
remaining vertices in basic cells. Add and subtract
q from the allotments in these basic cells in order
to balance the availability and demand constraints.
Now the entering cell has allotment q and the other The first row has the maximum difference. Allot
vertices of the loop have allotments xij + q and 25 units to the (1, 4) cell. The revised table is
xij – q. Of the cells having allotments of the form
xij – q, choose the cell having the least xij. Let it be
xpq. If we take q as xpq, then the allotment in this
cell becomes zero and hence this cell becomes
non-basic (leaves the basis). With the value of
q as xpq, we get a modified table with revised
allotments. Again we test this solution for
optimality and proceed as before.
This procedure is repeated till an optimal Select the third row with the difference 2 and
solution is obtained. allot 32 units to the (3, 3) cell. The table becomes
Example 11.7 Solve the following transportation
problem

The 2nd column is automatically filled up with

Solution We obtain an initial basic feasible the remaining quantity. This is an initial basic
solution using the VAM method. The difference feasible solution with the total cost of Rs 781.
between the least and the next higher costs of each Introduce simplex multipliers u1, u2, u3 for the
row and column are noted down. rows and v1, v2, v3, v4 for the columns. Form
 122 Operations Research

the equations ui + vj = cij for the basic cells. We Taking u2 = 0 and solving these equations we
have obtain
u1 + v2 = 8 u2 + v2 = 11 u1 = – 2 u2 = 0 u3 = – 1 v1 = 5
u1 + v4 = 5 u3 + v2 = 9 v2 = 10 v3 = 8 v4 = 7
u2 + v1 = 5 u3 + v3 = 7 Using these values we calculate the relative cost
Since the 2nd column contains the maximum factors cij for the non basic cells.
number of basic cells we take v2 = 0. Substituting c11 = c11 – u1 – v1 = 3 c13 = c13 – u1 – v3 = 2
v2 = 0 in the above equations and solving them we c23 = c23 – u2 – v3 = 1 c24 = c24 – u2 – v4 = 0
get
c31 = c31 – u3 – v1 = 4 c34 = c34 – u3 – v4 = 7
u1 = 8 u2 = 11 u3 = 9 v1 = – 6
Since all the relative cost factors are non
v2 = 0 v3 = – 2 v4 = – 3
negative the current solution is optimal. The
Using these values we calculate cij = cij – ui – vj optimal solution is
for the non-basic cells. (Rs)
c11 = c11 – u1 – v1 = 4; c13 = c13 – u1 – v3 = 2 S1 Æ D2 : 10 units cost 80
c23 = c23 – u2 – v3 = 0; c24 = c24 – u2 – v4 = – 1 S1 Æ D4 : 20 units cost 100
c31 = c31 – u3 – v1 = 5; c34 = c34 – u3 – v4 = 7 S2 Æ D1 : 35 units cost 175
We find c24 is negative and hence (2, 4) is the S2 Æ D4 : 5 units cost 35
S3 Æ D2 : 18 units cost 162
entering cell. Introduce q in (2, 4) cell and form a
S3 Æ D3 : 32 units cost 224
loop as shown below.
Total cost = 776
Example 11.8 A company has three factories A,
B and C which supply units to warehouses X, Y
and Z every month. The capacities of the factories
are 60, 70 and 80 units at A, B and C respectively.
The requirements of X, Y and Z per month are
50, 80 and 80 units respectively. Transportation
The vertices (2, 2) and (1, 4) of the loop have cost per unit in rupees are given in the
5 – q, 25 – q. The least quantity is 5 and hence following table. Find out the minimum cost of
q = 5 is the maximum value of q such that all the transportation
entries are non-negative (feasible). Thus, taking
q = 5 we find that (2, 2) cell becomes non-basic
and (2, 4) cell becomes basic with an allotment of
5. The improved solution is

Solution The transportation table for the given

data is

We test whether this is an optimal solution. Write

the equations cij = ui + vj for the basic cells. We get
u1 + v2 = 8 u2 + v4 = 7 We find an initial basic feasible solution by
u1 + v4 = 5 u3 + v2 = 9 least cost rule. We get the initial solution as
u2 + v1 = 5 u3 + v3 = 7 follows:

The least cost is 5 in (1, 3) cell and (3, 3) cell.
Select (1, 3) cell and allot 60 unit. We have,
The least cost in the reduced table is 5 in (3, 3)
cell. Allot 20 units there. We get
(2, 1) cell and (3, 2) cell both have the minimum
cost 6. Select (2, 1) cell for an allotment of 50
units. The table becomes
Now we can allot 20 units to (2, 2) cell and 60
units to (3, 2) cell. The initial solution is
m + n – 1 = 5 and we have 5 basic cells.
Defining simplex multipliers ui and vj for the rows
and columns we get the equations for the basic
cells,
u1 + v3 = 5 u3 + v2 = 6
u2 + v1 = 6 u3 + v3 = 5
u2 + v2 = 8
Transportation Problem 123
Assign zero value to u2 we get the solutions
u1 = – 2 v1 = 6
u2 = 0 v2 = 8
u3 = – 2 v3 = 7
The relative cost factors for the non-basic cells
are
c11 = c11 – u1 – v1 = 4 c12 = c12 – u1 – v2 = 1
c23 = c23 – u2 – v3 = 2 c31 = c31 – u3 – v1 = 5
All cij ≥ 0. Hence this solution is itself optimal.
The optimal solution is
(Rs )
A Æ Z : 60 units Cost 300
B Æ X : 50 units Cost 300
B Æ Y : 20 units Cost 160
C Æ Y : 60 units Cost 360
C Æ Z : 20 units Cost 100
Total cost = Rs 1220
11.5 DEGENERACY
As we have discussed earlier, the number of basic
cells in a transportation table of m rows and n
columns, should be m + n – 1. If we get less than
(m + n – 1) basic cells optimality test cannot be
carried out. This is called a degenerate solution.
In order to overcome this difficulty and to solve
the degeneracy, we need (m + n – 1) basic cells.
Suppose we have only (m + n – 2) basic cells,
then we select a non-basic cell which does not form
a loop with the existing basic cells. Allot an
infinitesimal quantity Πto this selected cell and
make it a basic cell. This infinitesimal quantity
will not affect the other allotments when it is added
or subtracted. In other words it is only a dummy
allotment. But it helps us to carry out the iteration.
Now we have (m + n – 1) basic cells and we can
perform optimality test and proceed with the
iteration if necessary. Finally Πis taken as zero.
Example 11.9 Solve the following transportation
problem:
 124 Operations Research

Solution Applying VAM method we obtain an B Æ Z : 20 units Cost 180

initial basic feasible solution, given below C Æ Y : 80 units Cost 240
Total Cost = 750
Example 11.10 Solve the following TP

We have 4 basic cells but we should have

(3 + 3 – 1 = 5) 5 basic cells. Select any non-basic
cell from the set {(1, 2), (2, 2), (3, 1) and (3, 3)}
which does not form a loop with the existing basic
cells. Let us choose (2, 2) cell and make it a basic Solution An initial basic feasible solution
cell by allotting an infinitesimal quantity Πto it. obtained by the Least Cost Rule is given below
Now we get an initial basic feasible solution
having 5 basic cells.
Introduce simplex multipliers u1, u2, u3 for the
rows and v1, v2, v3 for the columns. We get 5
equations ui + vj = cij corresponding to the basic
cells.
u1 + v3 = 3 u2 + v3 = 9
u2 + v1 = 3 u3 + v2 = 3 Here m + n – 1 = 4 + 3 – 1 = 6. But we have
u2 + v2 = 8 only 5 basic cells. Hence we introduce one more
basic cell which does not form a loop with these
basic cells. Let us take the cell (2, 4) and allot an
infinitesimal quantity Πto it. Now we have

Set u2 = 0. We get the solution

u1 = – 6 v1 = 3
u2 = 0 v2 = 8
u3 = – 5 v3 = 9 Using simplex multipliers u1, u2, u3 for rows
The relative cost factors for the non-basic cells and v1, v2, v3, v4 for columns we get the equations
are C11 = 8 + 6 – 3 = 11 ui + vj = cij
c12 = 7 + 6 – 8 = 5 c31 = 11 + 5 – 3 = 13 u1 + v1 = 8 u2 + v3 = 4
u1 + v2 = 10 u2 + v4 = 7
c33 = 5 + 5 – 9 = 1 u1 + v4 = 6 u3 + v2 = 11
cij are all non negative. Hence this solution is for the basic cells.
optimal. Take Πto be zero. The optimal solution Setting u1 = 0, we get the following solution
is u1 = 0 v1 = 8
(Rs ) u2 = 1 v2 = 10
A Æ Z : 60 units Cost 180 u3 = 1 v3 = 3
B Æ X : 50 units Cost 150 v4 = 6
 Transportation Problem 125

The relative cost factors for the non-basic cells quantity available at the sources must be equal to
are given by the total quantity required at the destinations
c13 = 7 – 0 – 3 = 4 c31 = 9 – 1 – 8 = 0 m n
c21 = 12 – 1 – 8 = 3 c33 = 10 – 1 – 3 = 6 i.e. å ai = å b j
c22 = 9 – 1 – 10 = – 2 c34 = 8 – 1 – 6 = 1 i =1 j =1
Since c22 is negative (2, 2) cell enters the basis. If this condition is not satisfied, then the
Allot a quantity q to this cell. problem is called an unbalanced transportation
problem. We solve this type of problem by
changing it to a balanced problem.
When the availability is greater than the
requirement, we introduce a dummy destination
in the table. The cost of transportation from each
source to this destination is assumed to be zero.
The excess of availability is assigned as the
A loop is formed with cells (1, 2) (2, 2) (2, 4) requirement at this dummy destination. Now the
and (1, 4). We find that q = Œ. Substituting q = Œ problem is balanced and it can be solved.
we get the improved solution as given below. When the requirement is more than the
availability, we introduce a dummy source in the
table. The cost of transportation from this source
to each destination is taken as zero. The excess of
demand is assigned as the availability at this
dummy source. Thus the problem is changed
to a balanced one and can be solved in the usual
way.
The relative cost factors for the non-basic cells Example 11.11 Solve the following transporta-
are tion problem
c13 = 7 – 10 + 5 = 2 c31 = 9 – 11 + 2 = 0
c21 = 12 – 9 + 2 = 5 c33 = 10 – 11 + 5 = 4
c24 = 7 – 9 + 4 = 2 c34 = 8 – 11 + 4 = 1.
Since all the cij are non negative, the current
solution is optimal. The optimal solution is
(Rs ) Solution Here Sai = 34, Sbj = 30
W1 Æ M1 : 25 units Cost 200 Therefore the problem is unbalanced. The total
W1 Æ M2 : 2 units Cost 20 supply is 4 units in excess of total demand and
W1 Æ M4 : 23 units Cost 138 hence we introduce a dummy destination D0 with
W2 Æ M3 : 40 units Cost 160 demand 4 and change the problem into a balanced
W3 Æ M2 : 30 units Cost 330 problem. The table is
Total Cost = Rs 848

11.6 UNBALANCED
TRANSPORTATION
PROBLEM
In order that a transportation problem should yield We obtain an initial basic feasible solution by
a feasible solution it is necessary that the total VAM as given below.
 126 Operations Research

S3 Æ D1 : 4 units Cost 24
S3 Æ D3 : 6 units Cost 24
Total cost = Rs 80 (Π= 0)
Example 11.12 Solve the following trans-
portation problem
m + n – 1 = 6 + 3 – 1 = 8. But we have only A company operates three coal mines A, B, C
7 basic cells. Hence this solution is degenerate. which provide 400, 500, and 700 tonnes res-
Let us choose the cell (1, 1) which doest not form pectively per week. Orders for 500, 400, 300, 300
a loop with the basic cells. Allot an infinitesimal and 600 tonnes per week have been received from
quantity Πto this cell. Now have 8 basic cells. customers C 1 , C 2 , C 3 , C 4 , C 5 respectively.
Introduce the simplex multipliers u1, u2, u3 for Transportation costs in rupees per tonne from each
the rows and v0, v1, v2, v3, v4 and v5 for the columns. mine to each customer are given below
Form the equations ui + vj = cij for the basic cells.
We get u1 + v1 = 4 u2 + v5 = 1
u1 + v2 = 2 u3 + v0 = 0
u1 + v4 = 2 u3 + v1 = 6
u2 + v4 = 2 u3 + v3 = 4
Setting u1 = 0. We get the solution.
u1 = 0 v0 = – 2 For each tonne of coal demanded but not supplied
u2 = 0 v1 = 4 there is a loss of one rupee per tonne for the company.
u3 = 2 v2 = 2 Find the weekly shipping (transporting) schedule
v3 = 2 which minimizes the total expenses.
v4 = 2 Solution Here, the total supply is 1600 units. But
v5 = 1 the total demand is 2100 units. Hence it is an
unbalanced problem. We introduce a dummy mine
D which provides 500 tonnes per week. The loss
of one rupee per tonne is taken as the transportation
cost from this mine D to each customer. The
balanced transportation table is given below.

The relative cost factors for the non basic cells are
c10 = 0 – 0 + 2 = 2 c22 = 4 – 0 – 2 = 2
c13 = 3 – 0 – 2 = 1 c23 = 5 – 0 – 2 = 3
c15 = 6 – 0 – 1 = 5 c32 = 5 – 2 – 2 = 1
We find an initial basic feasible solution by
c20 = 0 – 0 + 2 = 2 c34 = 7 – 2 – 2 = 3 VAM as shown below:
c21 = 5 – 0 – 4 = 1 c35 = 3 – 2 – 1 = 0
We find that all cij ≥ 0. Hence the current
solution is optimal. The optimal solution is
(Rs)
S1 Æ D2 : 4 units Cost 8
S1 Æ D4 : 4 units Cost 8
S2 Æ D4 : 4 units Cost 8
S2 Æ D5 : 8 units Cost 8
 Transportation Problem 127

m + n – 1 = 5 + 4 – 1 = 8 and we have 8 basic

cells. Therefore it is a non-degenerate feasible
solution.
To test for optimality we introduce simplex
multipliers ui (i = 1, 2, 3, 4) and vj ( j = 1, 2, 3, 4, 5)
and form the equations, ui + vj = cij. We get the
equations
u1 + v1 = 4 u3 + v1 = 6
u1 + v3 = 1 u3 + v5 = 2 Introducing simplex multipliers ui and vj for the
u2 + v2 = 10 u4 + v1 = 1 basic cells we get the equations
u2 + v5 = 12 u4 + v4 = 1 u1 + v1 = 4 u3 + v5 = 2
Take v1 = 0 u1 + v3 = 1 u4 + v1 = 1
The values of ui and vj are u2 + v2 = 10 u4 + v4 = 1
u1 = 4 u2 = 16 u2 + v3 = 8
u3 = 6 u4 = 1 u3 + v1 = 6
v1 = 0 v2 = – 6 Take v1 = 0
v3 = –3 v4 = 0 The values of ui and vj are
v5 = – 4 u1 = 4 u2 = 11
The relative cost factors for the non-basic cells u3 = 6 u4 = 1
are v1 = 0 v2 = –1
c12 = 16 – 4 + 6 = 18 c24 = 12 – 16 – 0 = – 4 v3 = –3 v4 = 0
v5 = –4
c14 = 16 – 4 – 0 = 12 c32 = 1 – 6 + 6 = 1
The relative cost factors for the non-basic cells are
c15 = 14 – 4 + 4 = 14 c33 = 4 – 6 + 3 = 1 c12 = 16 – 4 + 1 = 13 c14 = 16 – 4 – 0 = 12
c21 = 18 – 16 – 0 = 2 c34 = 13 – 6 – 0 = 7 c15 = 14 – 4 + 4 = 14 c21 = 18 – 11 – 0 = 7
c23 = 8 – 16 + 3 = – 5 c42 = 1 – 1 + 6 = 6 c24 = 12 – 11 – 0 = 1 c32 = 1 – 6 + 1 = – 4
c43 = 1 – 1 + 3 = 3 c33 = 4 – 6 + 3 = 1 c34 = 13 – 6 – 0 = 7
(2, 3) cell has the most negative relative cost c42 = 1 – 1 + 1 = 1 c43 = 1 – 1 + 3 = 3
factor –5. Hence (2, 3) cell enters the basis. We
c45 = 1 – 1 + 4 = 4
allot a quantity q to the (2, 3) cell and form a loop
Since c32 = – 4, (3, 2) cell enters the basis. We
having all other vertices at basic cells as shown
allot a quantity q to this cell and form a loop with
below:
this cell at one vertex and basic cells at the other
vertices as given below

Of the cells with allotment of the form xij – q

(2, 5) cell has the least entry 100. Therefore q = From the entries we find that q = 100 and
100 and (2, 5) cell becomes non-basic. The revised consequently (3, 1) cell becomes non-basic. The
distribution is improved solution is
 128 Operations Research

(The entries denote profits from sale of unit

product.)

Defining simplex multipliers ui and vj and Solution First we convert the problem to
forming the equations minimization problem by subtracting each element
ui + vj = cij from the highest figure 90. We get the following
for the basic cells we obtain table:
u1 + v1 = 4 u3 + v2 = 1
u1 + v3 = 1 u3 + v5 = 2
u2 + v2 = 10 u4 + v1 = 1
u2 + v3 = 8 u4 + v4 = 1
Take u1 = 0 and solve the equations. We get
u1 = 0 u2 = 7 u3 = – 2 u4 = –3
v1 = 4 v2 = 3 v3 = 1 v4 = 4 v5 = 4 We obtain an initial basic feasible solution by
The relative cost factors for the non-basic cells are VAM as given below
c12 = 13 c14 = 12 c15 = 10 c21 = 7 (m + n – 1 = 4 + 3 – 1 = 6)
c24 = 1 c25 = 1 c31 = 4 c33 = 5
c34 = 11 c42 = 1 c43 = 3 c45 = 0
Hence the current solution is optimal.
A Æ C1 : 300 units A Æ C3 : 100 units
B Æ C2 : 300 units B Æ C3 : 200 units
C Æ C2 : 100 units C Æ C5 : 600 units
Minimum cost is 1200 + 100 + 3000 + 1600 +
100 + 1200 + (200 + 300) = Rs 7700.
Writing the equations ui + vj = cij for the basic
11.7 MAXIMIZATION TYPE cells we get
u1 + v2 = 39 u2 + v4 = 9 where u2 = 0
In general, transportation problem is motivated u2 + v1 = 10 u3 + v1 = 0
towards minimizing the transportation cost of u2 + v2 = 48 u3 + v3 = 24
goods from various sources to different The solutions are
destinations. But in certain cases the objective may u1 = – 9 v1 = 10 u2 = 0 v2 = 48
be to maximize the profit or total output. To solve u3 = – 10 v3 = 34 v4 = 9
such type of problems of maximization we change The relative cost factors for the non-basic cells
the sign of the costs to negative and then apply are
minimization algorithm. c11 = 75 + 9 – 10 = 74
If it is found to be inconvenient to deal with
c13 = 48 + 9 – 34 = 23
negative figures we can subtract all the costs in
the table from the highest cost given and then apply c14 = 57 + 9 – 9 = 57
minimization process. c23 = 64 – 0 – 34 = 30
Example 11.13 Solve the following trans- c32 = 50 + 10 – 48 = 12
portation problem to maximize the total profit. c34 = 30 + 10 – 9 = 31
 Transportation Problem 129

cij ≥ 0 for all i, j. Hence the current solution is S2 Æ M2 : 8 units
optimal and the optimal solution is S2 Æ M4 : 30 units
(Rs) S3 Æ M1 : 17 units
S1 Æ M2 : 23 units Profit 23 ¥ 51 = 1173 S3 Æ M3 : 16 units
S2 Æ M1 : 6 units Profit 6 ¥ 80 = 480
Profit 8 ¥ 42 = 336
Profit 30 ¥ 81 = 2430
Profit 17 ¥ 90 = 1530
Profit 16 ¥ 66 = 1056
Total Profit = Rs 7005

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Solve the following transportation problems: 5. Solve the TP

6. Solve the TP
2. A company has three mines A, B and C and
five factories F1, F2, F3, F4 and F5. The mines
can supply 80, 100 and 140 tonnes of ore
daily and the requirements of the factories
are 40, 50, 70 and 80 respectively. The
following table gives the unit transportation
cost of ore.
7. Solve the TP

Give a distribution plan to minimize the total

cost of transportation. 8. Solve the TP
3. Solve the TP

9. Solve the TP
4. Solve the TP
 130 Operations Research

10. Solve the TP 16. Solve

11. Solve 17. Solve

18. Solve
12. Solve

13. Solve 19. Solve the following transportation problem

to maximize the total profit.

14. Solve 20. A company has four plants manufacturing

the same product and five sales centres. The
cost of manufacturing, cost of raw materials
and capacities of the plants are given below.
Plant
1 2 3 4
Manufacturing 12 10 8 8
15. Solve cost per unit
Cost of raw 8 7 7 5
material per unit
Capacity 100 200 120 80
The sales prices, transportation costs and
demands are given in the following table
 Transportation Problem 131

Sales Transportation Unit sales Demand 25.

Centre cost per unit price
1 2 3 4
A 4 7 4 3 30 80
B 8 9 7 8 32 120
C 2 7 6 10 28 150
D 10 7 5 8 34 70
E 2 5 8 9 30 90 26.
Formulate this problem into a transportation
problem to maximize profit and find the
optimal solution.
21.

27.

22.
28.

23. 29.

24. 30.
 132 Operations Research

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. O1 – D1 : 25; O1 – D4 : 50; O1 – D5 : 25 13. S1 – D1 – 5; S2 – D2 : 8; S3 – D2 = 1

O2 – D2 : 60; O2 – D3 : 40; O2 – D4 : 25 S3 – D3 : 6; S4 – D1 : 2; S4 – D3 : 12
O3 – D1:75 Total cost = Rs 70
Total cost = Rs 615 14. S1 – D2 : 20; S1 – D3 : 3; S1 – D4 = 7
2. A – F2 : 50; A – F4 : 30; B – F4 : 50 S2 – D1 : 10; S2 – D2 : 5; S3 – D3 : 15
B – F5:50; C – F1 : 40; C – F3 : 20; Total cost = Rs 231
C – F5 : 30 15. X – B : 76; Y – B : 21; Y – C : 41
Total cost = Rs 920 Z – A : 72; Z–B:5
3. P – X : 1; Q – X : 1; Q–Y:5 Total cost = Rs 2424
Q – Z : 2; R–Z=7 16. S1 – D5 : 8; S2 – D1 : 4; S2 – D4 : 1
Total cost = Rs 142 S3 – D2 : 4; S3 – D3 : 2; S3 – D4 : 3
4. F1 – W1 : 400; F2 – W2 : 500 Total cost = Rs 92
F3 – W1 : 300; F3 – W2 : 100 17. F1 – W3 : 25; F1 – W4 : 25; F2 – W3 : 10
F3 – W3 : 200; F4 – W3 : 300 F3 – W1 : 30; F3 – W2 : 25; F3 – W4 : 15
F4 – W4 : 400; Total cost = Rs 1150
Total cost = Rs 142 18. O1 – C : 5; O2 – B : 4; O2 Æ F : 2
5. A – D1 : 4; A – D2 : 10; B – D1 : 1; O3 – A : 1; O3 – C : 1; O4 – A : 3
B – D3 : 15; C – D1 : 1; C – D4 : 4; O4 – D : 2; O4 – E : 4
Total cost = Rs 114 Total cost = Rs 112
6. P – C : 150; Q – A : 40; Q – B : 90 19. S1 – A : 20; S1 – D : 30; S2 – A : 20
Q – C : 30; R – B : 110; S – A : 130 S2 – C : 10; S3 – B : 20; S3 – C : 50;
Total cost = Rs 2980 Max. Profit = Rs 5130
7. F1 – W1 : 5; F2 – W2 : 2; F3 – W2 : 6 20.
F1 – W4 : 2; F2 – W3 : 7; F3 – W4 : 12
Total cost = Rs 743
8. P1 – D1 : 5; P2 – D2 : 3; P3 – D2 : 5
P1 – D4 : 2; P2 – D3 : 7; P3 – D4 : 13
Total cost = Rs 201
9. A1 – D2 : 3; A2 – D1 : 1; A2 – D4 : 2
A3 – D3 : 2; A3 – D4 : 4 A4 – D4 : 2
A5 – D2 : 1; A6 – D1 : 2 B – 2 : 70; B – 3 : 50; C – 1 : 100
Total cost = Rs 103 C – 2 : 40; D – 3 : 70; E – 2 : 90;
10. S1 – D1 : 2; S1 – D4 : 28; S2 – D3 : 15 21. F1 – W1 : 5; F1 – W3 : 15; F1 – W4 : 20
S3 – D2 : 25; S4 – D1 : 8; S4 – D3 : 22 F2 – W2 : 30; F3 – W1 : 15 ; F3 – W5 : 5
Total cost = Rs 193 F4 – W1 : 10
11. I – M1 : 16; I – M2 : 1; II – M3 : 13 Total cost = Rs 510
III – M4 : 12; III – M5 : 4; IV – M2 : 19 22. A – 1 : 10; A – 3 : 30; A – 4 : 40
IV – M5 : 1 B – 2 : 60; C – 1 : 30 ; C – 5 : 10 ;
Total cost = Rs 146 D – 1 : 20 ; Total cost = Rs 420
12. A – P : 7; A – Q : 4; A–R:2 23. I – 4 : 11; II – 1 : 6; II – 2 : 3
A – S : 9; B – R : 15; C–Q:8 II – 4 : 4; III – 2 : 7; III – 3 : 12
Total cost = Rs 186 Total cost = Rs 796
 Transportation Problem 133

24. A – 3 : 2; B – 2 : 1; B – 3 : 2; S2 – D4 : 40; S3 – D1 : 60; S3 – D4 : 20

C – 1 : 4; C–3:1 Total cost = Rs 8190
Total cost = Rs 29 28. S1 – D1 : 14; S1 – D3 : 4; S2 – D3 : 10
25. P – A : 10; P – B : 50; P – C : 40 S2 – D4 : 17; S3 – D2 : 12; S3 – D3 : 9
Q – C : 30; Q – E : 90; R – A : 30 Total cost = Rs 811
R – D : 90; Total cost = Rs 1400 29. W1 – M1 : 140; W1 – M3 : 60
26. W1 – S1 : 10; W1 – S4 : 40; W1 – S7 : 10 W2 – M1 : 40; W2 – M2 : 120
W2 – S7 : 5; W2 – S6 : 35; W3 – S2 : 20 W3 – M3 : 90; Total cost = Rs 5920
W3 – S3 : 45; W3 – S5 : 20; W3 – S7 : 15 30. S1 – D1 : 7; S1 – D2 : 4; S1 – D3 : 2
Total cost = Rs 580 S1 – D4 : 9; S2 – D3 : 15; S3 – D2: 8
27. S1 – D1 : 30; S1 – D2 : 70; S1 – D3 : 50 Total cost = Rs 186

Assignment Problems
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
12.1 WHAT IS AN ASSIGNMENT
PROBLEM?
Given n jobs, n workers, and the time taken by
each worker to do each job, the objective is to allot
the different jobs to the workers such that (i) no
worker is given more than one job (ii) no job is
allotted to more than one worker and (iii) the total
time taken to complete all the jobs is minimum.
This is called an assignment problem.
The problem can also have cost or profit instead
of time taken for each job. In that case we have to
maximize the total profit or minimize the total cost.
In other words we have to optimize the objective
assigning the jobs to the workers suitably. This is
the same as transportation problem except that here
the availability at each source and the demand at
each destination are taken to be one.
12.2 MATHEMATICAL
FORMULATION
We are given n jobs J1, J2, J3, …, Jn and n workers
W1, W2, W3, …, Wn. The time taken by the ith
12
worker to do the jth job (Cost of the assignment) is
given as cij, and xij denotes the assignment. The
problem can be written in the form
n n
Minimize Z = å å cij x j
i =1 j =1
subject to the constraints
n
å xij = 1 (i = 1, 2, 3, …, n)
j =1
n
å xij = 1 ( j = 1, 2, 3, …, n)
i =1
xij = 0 or 1 for all i and j.
We take xij = 1 if ith worker gets jth job.
xij = 0 otherwise
We may note that in the table of assignment
there would be one and only one assignment in
each row or column. In other words, if xij = 1 then
there is no other assignment in the ith row and no
other assignment in the j th column.
The data is given in the form of a table (matrix)
as given below:
 Assignment Problems 135

J1 J2 J3 … ... ... Jn Having finished all the rows, subtract the least
element of each column from all the elements of
W1 c11 c12 c13 . . . c1n
that column.
W2 c21 c22 c23 . . . c2n
W3 c31 c32 c33 . . . c3n Step 2 Draw horizontal and vertical lines to cover
. all the zeroes obtained in step 1. If the minimum
. number of lines required to cover all the zeroes is
. equal to n (number of jobs) then optimal solution
Wn cn1 cn2 cn3 cnn is reached. Go to step 4.
Step 3 If the minimum number of lines required
12.3 METHOD OF SOLVING AN AP to cover all the zeroes is less than n, it means that
the current solution is not optimal. Examine the
First of all the problem can be solved by finding a elements not covered by any of the lines drawn.
list of all possible solutions which means that we Select the least among these elements and subtract
have to evaluate n! assignments and choose the it from all the elements not covered by the lines.
optimal one from among them. When n is large, Add this least element to every element lying at
this method involves a large number of the intersections of the lines. This gives rise to a
calculations and hence it is not suitable for manual new table. Again draw vertical and horizontal lines
calculations. to pass through the zeroes. Check whether the
Another method of solving an AP is by the minimum number of lines required, is equal to n.
simplex method. We can treat it as a 0 – 1 integer Repeat step 3 till an optimal table is reached.
programming problem. There are n2 decision
variables and 2n equalities. This again is difficult Step 4 (Obtaining the optimal solution) In
to solve when n becomes large. the optimal table examine the first row. If there is
An assignment problem can be taken as a only one zero in the first row, square it and cancel
transportation problem wherein we have n basic all the other zeroes in that column. If there are
cells. But we must have 2n – 1 (n + n – 1) basic more than one zero in the first row, pass on to the
cells. Hence we have to introduce n – 1 dummy second row. If there is only one zero in the second
allotments in order to apply the transportation row square it and cancel all the other zeroes in
algorithm. Thus it gives rise to computational that column. In case the second row contains more
difficulties. than one zero leave it and pass on to the third row.
The Hungarian method which we are going to After exhausting all the rows, take the columns
discuss, is the best choice for solving an AP. The one by one and repeat the procedure of squaring
algorithm is very easy to apply and there is no the zeroes. If there is a tie use your choice and
difficulty in manual calculations. square the zeroes. Finally we get the optimal
solution with n zeroes (each row having exactly
12.4 STEP-BY-STEP PROCEDURE one zero and each column having exactly one
zero). A zero in the (i, j) cell indicates that ith
OF HUNGARIAN METHOD
worker is allotted with the jth job. This cell is a
(MINIMIZATION PROBLEM) basic cell. In the original assignment table add all
the times (costs) of the basic cells. This gives the
Step 1 In the assignment table subtract the least total minimum time required (minimum cost) to
element of the first row from all the elements in complete all the jobs.
the first row. Subtract the least element of the
second row from all the elements of the second Example 12.1 Solve the following assignment
row. Repeat this for all the rows one by one. problem.
 136 Operations Research

Jobs 1 J2 J3 J4
J1 J2 J3 J4 W1 0 20
W1 10 15 24 30 2
Workers

W2 16 20 28 10 W3 3 3
W3 12 18 30 16 W4 0 9 8 8
W4 9 24 32 18
Now we find that the minimum number of lines
required to cover all the zeroes is 4 (n = 4). Hence
Solution
an optimal solution is reached.
Step 1 Subtract the least element of each row
from all the elements of the respective row. We get Step 4 In the first row there are two zeroes.
Therefore take the second row and there is only
J1 J2 J3 J4 one zero in (2, 4) cell. Squaring it we have
W1 0 5 14 20
J1 J2 J3 J4
W2 6 10 18 0
W1 1 0 0 20
W3 0 6 18 4
W2 7 5 4 0
W4 0 15 23 9
W3 0 0 3 3
Subtract the least element of each column from W4 0 9 8 8
all the elements of the respective column. We have
The third row has two zeroes. Go to the fourth
J1 J2 J3 J4 row. Square the only zero in (4, 1) cell and cancel
the zero in the first column (3, 1) cell. We have
W1 0 0 0 20
W2 6 5 4 0 J1 J2 J3 J4
W3 0 1 4 4
W1 1 0 0 20
W4 0 10 9 9 W2 7 5 4 0
W3 3 3
Step 2 All the zeroes can be covered by a
W4 0 9 8 8
minimum of 3 lines as shown below.
Now examine the columns. In the first column
1 J2 J3 J4 one zero has been squared already. In the second
W1 0 20 column there are two zeroes. Go to the third
2
column. There is only one zero in the third column
W3 0 1 4 4 (1, 3) cell. Square it and cancel the zero in (1, 2)
cell. This gives
W4 0 10 9 9

Since the number of lines is 3 < 4 we go to step 3. J1 J2 J3 J4

W1 0 20
Step 3 The least element not covered by any line
W2 7 5 4 0
is 1. Subtract 1 from all the elements not on the
lines and add 1 to the elements at the points of W3 3 3
intersection of the lines. We obtain W4 0 9 8 8
 Assignment Problems 137

Now there is only one zero in the third row (or I II III IV
second column) (3, 2) cell.
A 5 0 8 10
Squaring that zero also we get the optimal
solution table. B 0 6 15 0
C 8 5 1 0
J1 J2 J3 J4 D 0 6 4 2
W1 0 20
Subtracting the least element of each column
W2 7 5 4 0
from all the elements of the same column, we get
W3 3 3
W4 0 9 8 8 I II III IV
A 5 0 7 10
The optimal assignment is
B 0 6 14 0
W1 Æ J3 : 24 C 8 5 0 0
W2 Æ J4 : 10 D 0 6 3 2
W3 Æ J2 : 18
W4 Æ J 1 : 9 Drawing the minimum number of lines to cover
all the zeroes we have
Total = 61 Optimum cost = 61
Example 12.2 The assignment costs of four II III
operators to four machines are given in the 7
following table. B 0 6 14 0
Operators C
D 0 6 3 2
I II III IV
A 10 5 13 15 We find that the minimum number of lines
Machines

B 3 9 18 3 required to cover all the zeroes is 4 which is the

C 10 7 3 2 order of the matrix. Hence optimality is reached.
D 5 11 9 7
The optimal solution is obtained as follows.
There is only one zero in the first row (1, 2) cell
Find the optimal assignment using the square it. There are two zeroes in the 2nd and 3rd
Hungarian method. row. But there is only one zero in the 4th row (4, 1)
cell. Square it and cancel the zero in the (2, 1)
Solution The assignment table is cell. We have

I II III IV I II III IV
A 10 5 13 15 A 5 0 7 10
B 3 9 18 3 B 14 0
C 10 7 3 2 C 8 5 0 0
D 5 11 9 7 D 0 6 3 2

Subtracting the least element of each row from Now consider the unassigned columns. In the 3rd
all the elements of the same row, we get column there is only one zero in the (3, 3) cell.
 138 Operations Research

Square it and cancel the zero in the (3, 4) cell. This Since the number of lines is less than 4,
gives rise to only one zero in the fourth column optimality is not reached.
(2, 4) cell. Square it. Thus the optimal assignment is Select the least element 5 not covered by the
lines and subtract from all the elements not on the
I II III IV lines and add it to the elements at the points of
A 5 0 7 10 intersection of the lines. The resulting table with
B 14 0 the minimum number of lines covering all the
C 8 5 0 zeroes is
D 0 6 3 2
M1 M2 M3 M4
The optimal solution is J1 2 6 0 0
A Æ II : 5 J2 1
B Æ IV : 3
J3 23 0 2 5
C Æ III : 3
J4
D ÆI : 5
Total = 16
Since the number of lines is 4, optimality is
Minimum total cost = Rs 16.
reached.
Example 12.3 Solve the following AP: Square the zero in (3, 2) cell and the zero in the
(4, 4) cell and cancel (1, 4) cell. Next square the
Machines
zero in (2, 1) cell and cancel (2, 3) cell. Square the
M1 M2 M3 M4 zero in (1, 3) cell. This gives the optimal solution
J1 18 26 17 11 as given below.
Jobs

J2 13 28 14 26
J3 38 19 18 15 M1 M2 M3 M4
J4 19 26 24 10 J1 2 6 0
J2 0 11 18
Solution Subtracting the least element of each J3 23 0 2 5
row and column from all the elements of the
J4 4 7 8 0
respective row and column we get the reduced
matrix as follows.
J1 Æ M3 : 17
M1 M2 M3 M4 J2 Æ M1 : 13
J1 7 11 5 0 J3 Æ M2 : 19
J2 0 11 0 13 J4 Æ M4 : 10
J3 23 0 2 0
J4 9 12 13 0 Total = 59
Example 12.4 Solve the following AP
The minimum number of lines required to cover
all the zeroes is 3 as shown below: Machines
1 2 3 4 5
M1 M2 M3 M4
A 5 5 – 2 6
Operators

J1 7 11 5 0 B 7 4 2 3 4
J2 0 11 0 3 C 9 3 5 – 3
J3 23 0 2 0 D 7 2 6 7 2
J4 9 12 13 0 E 6 5 7 9 1
 Assignment Problems 139

Solution Here we find that the operator A
cannot operate the machine 3 and the operator
C cannot operate the machine 4. Hence we assume
that the time taken by A on 3 and by C on
4 are • each.
Therefore the table is
Since the number of rows is 5 and we have only
4 lines this solution is not optimal.
Now, the least element not covered by any line,
is 1. Subtract 1 from all the elements not on the
lines and add 1 to the elements at the points of
intersection of lines. We get

1 2 3 4 5 1 2 3 4 5
A 5 5 • 2 6 A
B 7 4 2 3 4 B
C 9 3 5 • 3 C 2 0 2 • 0
D 7 2 6 7 2 D 1 0 4 4 0
E 6 5 7 9 1 E 1 4 6 7 0

Subtracting the least element from all the Again we find that 4 lines are enough to cover
elements in each row we get all the zeroes. Hence we repeat the step. The
smallest element not covered by any line is 1.
1 2 3 4 5 Subtract 1 from all the elements not on the lines
and add 1 to the elements at the intersection of the
A 3 3 • 0 4 lines. Now we have
B 5 2 0 1 2
C 6 0 2 • 0
D 5 0 4 5 0
E 5 4 6 8 0 A • 0 6
B
Subtracting the least element in each column C • 0
from all the elements of that column, we obtain D
E 0 4 5 6 0
1 2 3 4 5
Now the minimum number of lines required to
A 0 3 • 0 4 cover all the zeroes is 5 and hence optimality is
B 2 2 0 1 2 reached.
C 3 0 2 • 0 All the rows contain more than one zero. The
D 2 0 4 5 0
third column contains only one zero. Square the
E 2 4 6 8 0
zero in (2, 3) cell and cancel the zero in (2, 4) cell.
Let us try to cover all the zeroes by the minimum Now fourth column has only one zero in (1, 4)
number of vertical and horizontal lines. We find cell. Square it and cancel the zero in (1, 1) cell.
that just 4 lines are sufficient to cover all the zeroes, We have
as shown below:
1 2 3 4 5
A • 0 6
A B 1 3 0 0 3
B 2 2 0 1 2 C 1 0 1 • 0
D 0 0 3 3 0
C 3 0 2 • 0
E 0 4 5 6 0
D 2 0 4 5 0
E 2 4 6 8 0 Now there is a tie in each row and column. We
use our choice.
 140 Operations Research

Square the zero in the (3, 5) cell and cancel the 1 2 3 4

other zeroes in the 3rd row and the 5th column.
A 16 10 14 11
Now the 2nd column contains only one zero in the B 14 11 15 15
(4, 2) cell. Square it and cancel the zero in (4, 1) C 15 15 13 12
cell. The remaining only zero in (5, 1) can be D 13 12 14 15
squared now. This gives the optimal assignment
What is the optimal assignment which will yield
1 2 3 4 5 maximum profit?
A • 0 6
Solution The largest element in the table is 16.
B 1 3 0 0 3
C • 0 Subtracting all the elements from 16 we get the
D 0 0 3 3 0 table
E 0 4 5 6 0
1 2 3 4
The optimal solution is A 0 6 2 5
A Æ 4 time 2 B 2 5 1 1
B Æ 3 time 2 C 1 1 3 4
D 3 4 2 1
C Æ 5 time 3
D Æ 2 time 2 Subtract the least element of each row from all
E Æ 1 time 6 the elements of that row. We get
Total time = 15
For any other choice also we shall get the same 1 2 3 4
total time 15 even though the assignments may be A 0 6 2 5
different. B 1 4 0 0
C 0 0 2 3
12.5 MAXIMIZATION D 2 3 1 0
ASSIGNMENT PROBLEM Each column has at least one zero. Minimum
Sometimes we may have to deal with an number of lines required to cover all the zeroes is
assignment problem in which the objective is to 4 (order of the matrix). Therefore optimality is
maximize the profit, maximize the sales or reached.
maximize the total output of the machines. Since Square the only zero in (1, 1) cell of the first
max Z = –min (–Z), we multiply all the elements row and cancel the zero in (3, 1) cell. Square the
of the table by –1 and then apply the Hungarian zero in (3, 2) cell. Square the only zero of the 4th
method. Also we can apply another method. row in (4, 4) cell and cancel the zero in (2, 4) cell.
Subtract all the elements of the matrix from the Now square the zero in (2, 3) cell. We obtain the
largest element in the matrix and form a new table optimal assignment as follows.
to apply the Hungarian method.
1 2 3 4
Example 12.5 A company has 4 salesmen A, B,
A 0 6 2 5
C and D. These salesmen are to be allotted 4
B 1 4 0 0
districts 1, 2, 3 and 4. The estimated profit per day C 0 0 2 3
for each salesman in each district is given in the D 2 3 1 0
following table.
 Assignment Problems 141

(Rs) M1 M2 M3 M4
A Æ 1 Profit 16
J1 0 6 10 14
B Æ 3 Profit 15 J2 0 5 9 11
C Æ 2 Profit 15 J3 0 5 9 12
D Æ 4 Profit 15 J4
Total Profit = Rs 61.
We find that only two lines are necessary to
12.6 UNBALANCED cover all the zeroes and hence it is not optimal.
ASSIGNMENT PROBLEM Subtract the least element 5, not covered by the
lines from all the elements not on the line and add
If the number of jobs and the number of workers
5 to the point of intersection of the lines we get
(machines) are not equal it is called an unbalanced
assignment problem. As in the case of transporta-
M1 M2 M3 M4
tion problem, we introduce a dummy job or a
dummy worker in order to balance the problem. J1 5 9
We assign the cost zero for every cell in the newly J2 0 0 4 6
J3 0 0 4 7
introduced row or column. Now we can apply
J4
Hungarian method and solve this balanced assign-
ment problem.
Again we find that all the zeroes can be covered
Example 12.6 Three jobs are to be done by 4 by only 3 lines and hence optimality is not reached.
machines. Each job can be assigned to one and Subtract the least element 4 from all the
only one machine. The cost of each job on each elements not on the lines and add 4 to the points
machine is given in the following table. of intersection of the lines. We have the new table

M1 M2 M3 M4 M1 M2 M3 M4
J1 18 24 28 32 J1 0 1 1 5
J2 0 0 0 2
Job

J2 8 13 17 19
J3 0 0 0 3
J3 10 15 19 22 J4 9 4 0 0

What are the job assignments which will
minimize the total cost?
Solution It is an unbalanced problem with 3 jobs
and 4 machines. Therefore we introduce a dummy
job J4 having all the costs zero in the 4th row. We
have the balanced problem
This table requires at least 4 lines to cover all
the zeroes and hence it gives optimal solution.
Square the zero in (1, 1) cell and cancel the
other zeroes in the first column. Square the zero
in (4, 4) cell and cancel the zero in the (4, 3) cell.
We get

M1 M2 M3 M4 M1 M2 M3 M4
J1 18 24 28 32 J1 0 1 1 5
J2 8 13 17 19 J2 0 2
J3 10 15 19 22 J3 0 3
J4 9 4 0
J4 0 0 0 0
Now there is a tie. Square the zeroes in (2, 3)
Subtracting the least element from all the
and (3, 2) cells or square the zeroes in (2, 2) and
elements of each row we get
 142 Operations Research

(3, 3) cells. We get two optimal solutions given visited twice. With these restrictions we solve the
below: problem as an assignment problem. If the optimal
(i) solution does not satisfy the route (cycle) condition
we make adjustments in the assignment and make
M1 M2 M3 M4
it a cycle with minimum possible increase in the
J1 0 1 1 5 total cost.
J2 0 2
J3 3 Example 12.7 Solve the travelling salesman
J4 9 4 0 problem with the following cost matrix:
Cities
(Rs)
A B C D
J1 Æ M1 Cost 18
J2 Æ M3 Cost 17 A • 46 16 40
J3 Æ M2 Cost 15 Cities B 41 • 50 40
C 82 32 • 60
Total cost = Rs 50. D 40 40 36 •
M4 remains idle
(ii) Solution The given cost matrix is

M1 M2 M3 M4 A B C D
J1 0 1 1 5 A • 46 16 40
J2 2 B 41 • 50 40
J3 0 3 C 82 32 • 60
J4 9 4 0 D 40 40 36 •

(Rs) Subtracting the least element of each row from

J1 Æ M1 Cost 18 all the elements of that row, we get
J2 Æ M2 Cost 13
J3 Æ M3 Cost 19 A B C D
Total cost = Rs 50. A • 30 0 24
B 1 • 10 0
M4 remains idle
C 50 0 • 28
D 4 4 0 •
12.7 TRAVELLING SALESMAN
Subtracting the least element of each column
PROBLEM
from the other elements of that column, we get
A travelling salesman goes from one city to
another, visiting each city only once and returns A B C D
to the starting city after visiting all the other cities. A • 30 0 24
The cost of his travel from each city to every other B
city is given. The objective is to give a route such C
that the total cost is minimum. D 3 4 0 •
Thus the travelling salesman problem is also
an assignment problem in which (i) going from a We note that the minimum number of lines
city to itself is not allowed. Hence (i, i) cells are required to cover all the zeroes is only 3 whereas
given the cost • (ii) the travel should be a cycle the order of the matrix is 4. Hence optimality is
passing through all the cities and no city to be not reached.
 Assignment Problems 143

Subtract 3 from all the elements not on the lines Solution Assign • value to the diagonal cells
and add 3 to the points of intersection of the lines. and consider the assignment problem
This gives,
A B C D E
A B C D
A • 4 7 3 4
A • 27 0 21 B 4 • 6 3 4
B 0 • 13 0 C 7 6 • 7 5
C 49 0 • 28 D 3 3 7 • 7
D 0 1 0 • E 4 4 5 7 •

Optimality is reached. Square the zero in (1, 3) Subtracting the least element of each row from
cell and the zero in (3, 2) cell. Now square the zero all the elements of that row we obtain the table,
in (2, 4) cell and cancel the (2, 1) cell. Finally square
the zero in (4, 1) cell. The optimal solution is A B C D E
reached.
A • 1 4 0 1
B 1 • 3 0 1
A B C D C 2 1 • 2 0
A • 27 0 21 D 0 0 4 • 4
B 0 • 13 0 E 0 0 1 3 •
C 49 0 • 28
D 0 1 0 • Subtracting the least element in each column
from all the elements of that column we get
(Rs)
A Æ C 16
A B C D E
B Æ D 40
C Æ B 32 A • 1 3 0 1
D Æ A 40 B 1 • 2 0 1
Total Cost = Rs 128 C 2 1
We get A Æ C, C Æ B, B Æ D, D Æ A D 0 0 3 4
E
or AÆCÆBÆDÆA
which is a cycle. Hence the solution is A Æ C Æ
Just with 4 lines all the zeroes have been
BÆDÆA
covered (4 < 5). Hence the table is not optimal.
with minimum total cost Rs 128.
Subtract 1 from all the elements not on the lines
Example 12.8 A salesman has to visit five cities and add 1 to the elements at the points of
A, B, C, D and E. The distances (in 100 km) intersection of the lines. The resulting table is
between any two cities are given in the following
table. The salesman starts from A and has to come • 0 2 0 0
0 • 1 0 0
back to A after visiting all the other cities in a cycle.
2 1 • 3 0
Which route he has to select in order that the total 0 0 3 • 4
distance travelled by him is minimum? 0 0 0 4 •
A B C D E Since the minimum number of lines required to
A – 4 7 3 4 cover all the zeroes is 5 we have reached optimality.
Square the zero in (3, 5) cell in the 3rd row and
From

B 4 – 6 3 4
C 7 6 – 7 5 cancel the other zeroes in the 5th column. Square
D 3 3 7 – 7 the zero in (5, 3) cell in the 3rd column and cancel
E 4 4 5 7 – the other zeroes in the 5th row. We have
 144 Operations Research

A B C D E A Æ D
A • 0 2 0 0 B Æ A
B 0 • 1 0 0 C Æ E
C 2 1 • 3 0 D Æ B
D 0 0 3 • 4
E • E Æ C
We get
Now make arbitrary assignments. Square the A Æ D Æ B Æ A, C Æ E Æ C
zero in (1, 4) cell. Then we can square zero in
It is not a cycle. Hence we have to modify the
(2, 1) cell and finally square the zero in (4, 2). We
solution. The smallest non-zero element is 1 in
get an optimal solution to the assignment problem.
(2, 3) and (3, 2) cells. Let us make an assignment
to (2, 3) cell. Now we get
A B C D E
AÆD BÆC CÆE DÆB EÆA
A •
B 0 • i.e. A Æ D Æ B Æ C Æ E Æ A
C 2 1 • 3 0 which is the optimal route and the total distance is
D 0 0 3 • 4 3 + 3 + 6 + 5 + 4 = 21 (2100 km)
E •

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Solve the AP 4. I II III IV V

J1 J2 J3 J4 J5 A 8 4 2 6 1
B 0 9 5 5 4
A 20 15 25 25 29 C 3 8 9 2 6
B 13 19 30 13 19 D 4 3 1 0 3
C 20 17 14 12 15 E 9 5 8 9 5
D 14 20 20 16 24
E 14 16 19 11 22
5. J1 J2 J3 J4
2. J1 J2 J3 J4 W1 5 7 11 6
W1 20 13 7 5 W2 8 5 9 6
W2 25 18 13 10 W3 4 7 10 7
W3 31 23 18 15
W4 10 4 8 3
W4 45 40 23 21
6. Machines
3. W1 W2 W3 W4 W5 I II III IV
J1 10 15 13 15 16 1 12 30 21 15
J2 13 9 18 13 10
2 18 33 9 31
Job

J3 10 9 12 12 12
J4 15 11 9 9 12 3 44 25 24 21
J5 11 9 10 14 12 4 23 30 28 14
 Assignment Problems 145

7. The profit per day of 4 salesmen in 4 districts 11. Solve the AP

are given below. Find an optimal assignment Jobs
of the districts to the salesmen so as to I II III IV
maximize the total profit.
W1 12 9 12 9

Workers
Districts W2 15 – 13 20
1 2 3 4 W3 4 8 10 6
A 14 8 12 9
Salesman

B 12 9 13 13 12. Solve
C 13 13 11 10 M1 M2 M3 M4
D 11 10 12 13
J1 60 51 32 32
8. The efficiency of 5 machines on each of 5 J2 48 – 37 43
J3 39 26 – 33
jobs is given below. Determine an assign-
J4 40 – 51 30
ment schedule of the jobs to the machines
such that the total efficiency is maximum.
13. Solve
Job Machine
1 2 3 4 5 M1 M2 M3 M4
O1 5 5 – 2
Operators

I 62 78 50 101 82
Machine

II 70 85 60 75 55 O2 7 4 2 3
III 88 96 118 85 71 O3 9 3 5 –
IV 48 64 87 77 80 O4 7 2 6 7
V 60 70 98 66 83
14. Solve the following travelling salesman
9. There are 5 jobs and 4 machines. The problem:
expected profits on each job on each machine A B C D
is given below. Determine an optimal A – 4 7 3
assignment of the machines to the jobs so B 4 – 6 3
that the total profit is maximum. C 7 6 – 7
D 3 3 7 –
Job
1 2 3 4 5 15. Solve the following travelling salesman
I 62 78 50 101 82 problem
Machine

II 71 84 61 73 59 A B C D E F
III 87 92 111 71 81 A – 5 12 6 4 8
IV 48 64 87 77 80 B 6 – 10 5 4 3
C 8 7 – 6 3 11
Hint: Introduce a dummy machine V with D 5 4 11 – 5 8
zero profit for each job. Then convert into E 5 2 7 8 – 4
the minimization type. F 6 3 11 5 4 –

10. Solve the AP and find the minimum cost 16. Solve the following AP.
W1 W2 W3 W4 A B C
A 18 24 28 32 1 120 100 80
B 8 13 17 19 2 80 90 110
C 10 15 19 22 3 110 140 120
 146 Operations Research

17. I II III IV V 24. Job

A 10 5 13 15 16 A B C D E
B 3 9 18 13 6 M1 9 11 15 10 11

Machine
C 10 7 2 2 2 M2 12 9 – 10 9
D 7 11 9 7 12 M3 – 11 14 11 7
E 7 9 10 4 12 M4 14 8 12 7 8

18. P Q R S T 25. 1 2 3 4
A 85 75 65 125 75 A 10 24 30 15
B 90 78 66 132 78 B 16 22 28 12
C 75 66 57 114 69 C 12 20 32 10
D 80 72 60 120 72 D 9 26 34 16
E 76 64 56 112 68
26. Job
1 2 3 4
19. I II III IV V
A 4 7 5 6
A 2 9 2 7 1
B – 8 7 4
B 6 8 7 6 1
C 3 – 5 3
C 4 6 5 3 1
D 6 6 4 2
D 4 2 7 3 1
E 5 3 9 5 1
27. The expected sales by four salesmen in each
of the four zones are given below
20. A B C D E
1 2 3 4
1 160 130 175 190 200
A 42 35 28 21
2 135 120 130 160 175
B 30 25 20 15
3 140 110 155 170 185
C 30 25 20 15
4 50 50 90 80 110
D 24 20 16 12
5 55 35 70 80 105
Determine the optimal assignment so as to
21. A B C D maximize the sales.
I 200 200 400 100 28. Maximization type
II 300 100 300 300
III 400 100 100 500 A B C D E
IV 200 200 400 200 1 32 38 40 28 40
2 40 24 28 21 36
22. I II III IV 3 41 27 33 30 37
4 22 38 41 36 36
A 8 26 17 11
5 29 33 40 35 39
B 13 28 4 26
C 38 19 18 15
D 19 26 24 10 29. Maximization type
1 2 3 4 5
23. A B C D A 2 8 • 5 6
1 5 3 2 8 B 0 1 8 2 6
2 7 9 2 6 C 5 6 1 4 •
3 6 4 5 7 D 7 4 8 2 3
4 5 7 7 8 E 5 4 0 6 7
 Assignment Problems 147

30. A company has 4 plants each of which can Sales Revenue

produce any one of the 4 products A, B, C, Plant
or D. Production costs and sales revenue of 1 2 3 4
each plant are given below: A 50 68 49 62

Product
Production Cost
B 60 70 51 74
Plant
C 55 67 53 70
1 2 3 4
D 58 65 54 68
A 49 60 45 61
Product

B 55 63 45 49 Determine which product should each plant

C 52 62 49 68 produce in order to maximize the profit
D 55 64 48 66
(Profit = Revenue – Cost)

ANSWERS
■ ■ ■ ■ ■ ■ ■

13. O1 Æ M4, O2 Æ M3, O3 Æ M2,O4 Æ M1

1. A Æ J2, B Æ J5, C Æ J3, D Æ J1, Total cost = 14
E Æ J4 Total cost = 73.
14. A Æ C Æ B Æ D Æ A Total cost = 19
2. W1 Æ J1, W2ÆJ2, W3 Æ J4, W4 Æ J3
Total cost = 76. 15. A Æ B Æ F Æ E Æ C Æ D Æ A
Total cost = 30
3. J1 Æ W1, J2 Æ W5, J3 Æ W2, J4 Æ W4,
J5 Æ W3 Total cost = 48 16. 1 – C 2–B 3–A
Total time = 280 hours
4. A Æ V, B Æ I, C Æ IV, D Æ III,
E Æ II Total cost = 9. 17. A – II B–I C–V D – III
E – IV Total time = 23 hours
5. W1 Æ J1, W2 Æ J2, W3 Æ J3, W4 Æ J4
Total cost = 23. 18. A – T B–R C–S D–P
E – Q Total time = 399 hours
6. 1 Æ I, 2 Æ III, 3 Æ II, 4 Æ IV
Total cost = 60. 19. A – III B–V C–I D – IV
7. A Æ 1, B Æ 3, C Æ 2, DÆ4 E – II Total time = 13 hours
Total profit = 53. 20. 1 – E 2–C 3–B 4–A
8. I Æ 4, II Æ 2, III Æ 1, IV Æ 5, 5 – D Total time = 570 hours
V Æ 3 Total efficiency = 452 21. I – D II – B III – C IV– A
9. I Æ 4, II Æ 2, III Æ 3, IV Æ 5 Total time = 7 hours
Job 1 remains unassigned. Total profit = 376 22. A – I B – III C – II D – IV
10. A Æ W1, B Æ W3, C Æ W2 Total time = 41 hours
Total cost = 50 23. 1 – B 2–C 3–D 4 – A (or)
11. W1 Æ II, W2 Æ III, W3 Æ I 1–C 2–D 3–B 4–A
Total cost = 26 Total time = 17 hours
12. J1 Æ M4, J2 Æ M3, J3 Æ M2, J4 Æ M1 24. M1 – A M2 – B M3 – E M4 – D
Total cost = 135 (C unassigned) Total = 32
 148 Operations Research

25. A – 2 B–3 C–4 D – 1 (or) 28. 1 – B 2–A 3–E 4–C

A–3 B–2 C–4 D – 1 (or) 5 – D (or)
A–3 B–4 C–2 D–1 1–B 2–E 3–A 4–C
Total cost = 71
5 – D Total = 191
26. A – 2 B–4 C–1 D – 3 (or)
A–3 B–4 C–1 D–2 29. A – 1 B–2 C–4 D–5
Total time = 18 hours E – 3 Total = 40
27. A – 1 B–3 C–2 D – 4 (or) 30. A – 2 B–4 C–1 D–3
A–1 B–2 C–3 D–4
Total profit = 42
Total sales = 99

Decision Theory
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
13.1 WHAT IS DECISION
THEORY?
Decision theory is the process of logical and
quantitative analysis of various factors involved
in a problem of decision-making and it helps us in
making the best possible decision. There may be
a number of courses of action (strategies) before
us. The problem is to choose the best out of them
so as to maximize the gain or (minimize the loss).
There are different models of decision-making.
Mostly we make decisions by analyzing the past
situations and the availability of resources at
present.
Linear programming problems are examples of
decision under certainty. There is only one possible
outcome and the process is well defined. But when
there is uncertainty about the outcome then the
decision-maker has no knowledge about the
possibilities of the occurrence of different
outcomes and he has to decide which course of
action is to be taken. Also, the results or outcomes
of the action depend on certain factors beyond our
control. They are called events which occur
according to the laws of nature. Under these
13
circumstances we have to choose the strategy
which results in optimum pay-off (gain).
13.2 STEPS IN DECISION
MAKING
1. Make a list of all possible events which may
occur.
2. Determine all the courses of action that can
be taken in the situation.
3. Determine the pay-off for each combination
of action and event.
4. Choose the best course of action which
results in maximum pay-off.
There are several different criteria for decision-
making.
13.3 MAXIMIN AND MAXIMAX
CRITERIA
Maximin criterion is based on the assumption that
the worst possibility is going to happen. We
consider each strategy and select the minimum
possible pay off corresponding to that strategy.
From among these we select the maximum one
 150 Operations Research

(best among the worst) and choose the alternative corresponding to the maximum sum is
corresponding action. selected. The procedure is explained by the
Maximax criterion is based on the optimistic following steps.
outlook. We select the best among the best. Select
Step 1 Choose a weight a (degree of optimism)
the maximum possible pay-off for each strategy
for the maximum pay-off and a weight 1 – a
and choose the maximum from among these. The
(degree of pessimism) for the minimum pay-off
strategy giving this pay-off is selected as the course
(0 < a < 1).
of action.
Step 2 Determine h = a (maximum) + (1 – a)
Example 13.1 A business man has three
minimum for each alternative.
alternatives, X, Y and Z, each of which gives rise
to four possible events A, B, C and D. The pay-off Step 3 Select the alternative corresponding to the
(in Rs) are given below. maximum h.
The Laplace criterion assigns equal probability
Alternative Pay-off to all the events. In other words, we take the
(Action) A B C D average pay-off corresponding to each action
X 8 0 –10 6 (expected pay off). The action corresponding to
Y –4 12 18 –2 the maximum expected pay-off is selected.
Z 14 6 0 8
Example 13.2 A company wants to introduce a
Find the best possible alternative using (i) new product in place of an old one. It is to be
Maximin criterion (ii) maximax criterion decided whether the price is to be fixed as very
high, moderate or slightly increased (H, M or S).
Solution The following table gives the minimum Three possible outcomes are expected, viz.
and maximum pay-off corresponding to each increase in sales, no change in sales or decrease in
alternative sales (I, N or D). The expected sales are given in
the following table (in lakhs of rupees)
Minimum Maximum
Alternative
Pay-off Pay-off Events
X –10 8 Strategies
I N D
Y –4 18
Z 0 14 H 70 30 15
M 50 45 0
Maximum among the minimum pay-offs is 0. S 30 30 30
Hence under maximin criterion Z is to be chosen.
Maximum among the maximum pay-offs is 18. Which alternative should be chosen according
Hence under maximax criterion Y is to be chosen. to (i) Maximin criterion (ii) Hurwicz criterion with
a = 0.6 and (iii) Laplace criterion?
13.4 HURWICZ CRITERION AND Solution
LAPLACE CRITERION (i) Maximin criterion
In decision-making we cannot be completely I N D Minimum
optimistic nor pessimistic and therefore we have H 70 30 15 15
to adopt a mixture of both. In Hurwicz criterion M 50 45 0 0
we take into account both minimum and maximum
S 30 30 30 30 Æ
pay-off for each alternative and assign them some
weight or probability. We find the sum of these Maximin value is 30. Hence the company
weighted pay-offs for each alternative. The should adopt the strategy S.
 Decision Theory 151

(ii) Hurwicz criterion (a = 0.6) Pay-off Regret Pay-off Maximum

Action
A B C A B C regret
I N D Max Min H
S1 700 300 150 0 150 50 150 Æ
H 70 30 15 70 15 70(0.6) + 15(0.4) = 48Æ S2 500 450 200 200 0 0 200
M 50 45 0 50 0 50(0.6) + 0(0.4) = 30 S3 300 300 100 400 150 100 400
S 30 30 30 30 30 30(0.6) + 30(0.4) = 30
Max. pay-off 700 450 200
According to the Hurwicz criterion, strategy
H is to be adopted. Among the maximum regrets, we find that the
(iii) Laplace criterion value corresponding to S1, is the minimum. Hence
according to the Minimax regret criterion, the
I N D Expected Pay-off alternative S1 is to be chosen.
H 70 30 15 1/3 (70+30+15) = 38.3 Æ
Example 13.4 Mr Girish wants to invest
M 50 45 0 1/3 (50+45+0) = 31.6
S 30 30 30 1/3 (30+30+30) = 30
Rs 10,000 in one of the three options A, B and C.
The pay-off for his investment depends on the
Since the largest return is from H, the nature of the economy (inflation, recession or no
company should adopt strategy H. change). The possible returns under each economic
situation are given below:
13.5 MINIMAX REGRET
Nature of Economy
CRITERION (SAVAGE Strategy
Inflation: E1 Recession: E2 No change: E3
CRITERION)
A 2000 1200 1500
This method is based on the regret of the decision- B 3000 800 1000
maker after he has taken a decision, which leads C 2500 1000 1800
to loss of pay-off. He wants to minimize this loss.
For each event we define the ith regret as the What course of action has he to take according
difference between the maximum pay-off and the to
ith pay-off. This process is repeated for each event. (i) The pessimistic criterion (Maximin)
Determine the maximum regret for each (ii) The optimistic criterion (Maximax)
alternative. From among these values choose the (iii) The equally likely criterion (Laplace)
minimum one and the corresponding alternative (iv) The regret criterion.
is the best one. Solution
(i) Maximin criterion
Example 13.3 Apply the minimax regret
criterion for the following decision problem. E1 E2 E3 Min
A 2000 1200 1500 1200 Æ
Events and pay-offs
Alternatives B 3000 800 1000 800
A B C C 2500 1000 1800 1000
S1 700 300 150
S2 500 450 200 Maximum among the minimum pay-offs is
S3 300 300 100 1200. Hence the corresponding option A is
to be chosen.
Solution Consider the event A. For S1, we get (ii) Maximax criterion
the maximum pay-off 700. For S2 and S3 we get E1 E2 E3 Max
regrets 200 and 400 respectively. Similarly, we find
the regrets for the events B and C. The regret pay- A 2000 1200 1500 2000
offs for each action-event combination are given B 3000 800 1000 3000 Æ
below: C 2500 1000 1800 2500
 152 Operations Research

Maximum among the maximum pay-offs is Minimax Regret Criterion

3000. Therefore option B is to be chosen.
Pay-off Regrets Max Regret
(iii) Laplace criterion
1 2 3 4 1 2 3 4 h
E1 E2 E3 Expected Pay-off X 8 0 –10 6 6 12 28 2 28
A 2000 1200 1500 1/3(2000 + 1200 Y – 4 12 18 –2 18 0 0 10 18 Æ
+ 1500) = 1566 Z 14 6 0 8 0 6 18 0 18 Æ
B 3000 800 1000 1/3(3000 + 800 Max 14 12 18 8
+ 1000) =1600 We find that the minimax regret occurs for both
C 2500 1000 1800 1/3(2500 + 1000 Y and Z. Hence either alternative Y or Z is to be
+ 1800) =1766 Æ chosen as the course of action.
Alternative C has the maximum expected
pay-off. Hence alternative C is to be chosen. 13.6 DECISION-MAKING
(iv) Regret criterion UNDER RISK
Regret pay-off Maximum 13.6.1 Expected Monetary Value
E1 E2 E3 E1 E2 E3 regret Criterion
A 2000 1200 1500 1000 0 300 1000 In many cases the decision-maker cannot predict
B 3000 800 1000 0 400 800 800 the outcome of his action. In such situations we
C 2500 1000 1800 500 200 0 500 Æ can assign probability values to the likely
Max 3000 1200 1800 occurrence of each event. For each combination
Among the maximum regrets, we find that of action and event we assign a probability. This
500 is the minimum. Therefore alternative probability is calculated on the basis of past
C is to be chosen. experience. We calculate the conditional pay-off
for each combination of action and event. The
Example 13.5 Apply the Hurwicz criterion (a Expected Monetary Value (EMV) of an action is
= 0.7) and Minimax regret criterion to solve the the sum of all the expected conditional pay-offs
decision problem with the following pay-off table. associated with that action. The action which has
the maximum EMV has to be chosen. The step-
Events by-step procedure is as follows.
Strategy
1 2 3 4 1. Determine the conditional pay-off for each
X 8 0 –10 6 combination of act and event.
Y –4 12 18 –2 2. Calculate the expected conditional pay-offs
Z 14 6 0 8 by multiplying the conditional pay-offs by
the corresponding probabilities.
Solution Hurwicz Criterion (a = 0.7) 3. Find the EMV for each action by adding all
Pay-off these expected conditional pay-offs.
1 2 3 4 Max Min h 4. Choose the action corresponding to the
X 8 0 –10 6 8 –10 8(0.7) – 10(0.3) = 2.6 maximum EMV.
Y –4 12 18 –2 18 – 4 18(0.7) – 4(0.3) = 11.4 Æ Example 13.6 A shopkeeper buys vegetable puff
Z 14 6 0 8 14 0 14(0.7) + 0 (0.3) = 9.8 at Rs 2 and sells it at Rs 5. Unsold puffs are given
to poor people free of cost. The following is the
According to Hurwicz criterion, alternative Y
sales details during the past 100 days.
is to be chosen.
 Decision Theory 153

Daily sales of puffs 10 11 12 13 his profit is only Rs 26 (The cost of 2 unsold puffs
is Rs 4). Thus we can find the profit for each
No. of days sold 15 20 40 25 combination of action and event.
The profit (pay-off) for each combination is
Apply EMV criterion to determine how many given by the following table:
puffs the shopkeeper has to stock everyday in order Conditional pay-off
to maximize his profit.
Solution We have 4 events Conditional pay-off
E1 : Demand for 10 puffs Events (Demand)
Actions 10 11 12 13
E2 : Demand for 11 puffs
E1 E2 E3 E4
E3 : Demand for 12 puffs
E4 : Demand for 13 puffs 10 A1 30 30 30 30
(stock)
11 A2 28 33 33 33
The courses of action are
12 A3 26 31 36 36
A1 : Stock 10 puffs
13 A4 24 29 34 39
A2 : Stock 11 puffs
A3 : Stock 12 puffs From the sales details in the past 100 days we
A4 : Stock 13 puffs find that the probabilities of demand are
Stocking less than 10 or more than 13 need not 10 puffs: 15/100 = 0.15
be considered as they have zero probability of 11 puffs: 20/100 = 0.20
sales. The profit on each puff sold is Rs 3. Suppose 12 puffs: 40/100 = 0.40
that the shopkeeper stocks 10 puffs and all of them 13 puffs: 25/100 = 0.25
are sold. Then he gets a profit of Rs 30. At the Multiplying these probabilities with the pay-
same time if he stocks 12 puffs and only 10 are offs we get the expected conditional pay-offs as
sold then he loses Rs 4 on unsold puffs and hence given below:

Prob. E1 E2 E3 E4 EMV
0.15 0.20 0.40 0.25 (Total) Rs
A1 30 (.15) 30 (.20) 30 (.40) 30 (.25) 4.5 + 6 + 12 + 7.5 = 30
A2 28 (.15) 33 (.20) 33 (.40) 33 (.25) 4.2 + 6.6 + 13.2 + 8.25 = 32.25
A3 26 (.15) 31 (.20) 36 (.40) 36 (.25) 3.9 + 6.2 + 14.4 + 9 = 33.5
A4 24 (.15) 29 (.20) 34 (.40) 39 (.25) 3.6 + 5.8 + 13.6 + 9.75 = 32.75

Since the EMV for A3 is maximum, the shopkeeper should stock 12 puffs at the beginning of each day.

Example 13.7 A flower merchant purchases
roses at Rs 10 per dozen and sells them at Rs 30.
Unsold flowers are donated to a temple. The daily
demand for roses has the following probability
distribution:
Demand (in dozen) 7 8 9 10
Probability 0.1 0.2 0.4
Using EMV criterion determine how many
dozens should he purchase in order to maximize
the profit. What is the maximum expected profit?
Solution Here the events and actions are
E1: Demand for 7 dozens with prob 0.1
E2: Demand for 8 dozens with prob 0.2
E3: Demand for 9 dozens with prob 0.4
E4: Demand for 10 dozens with prob 0.3
A1: Purchase 7 dozens
0.3 A2: Purchase 8 dozens
A3: Purchase 9 dozens
A4: Purchase 10 dozens
The conditional pay-off table is as follows:
 154 Operations Research

Demand The conditional pay-off table is

7 8 9 10
E1 E2 E3 E4 Demand E1 E2 E3 E4
A1 140 140 140 140 15 16 17 18
A2 130 160 160 160 Prob. 0.1 0.2 0.4 0.3

Purchase
A3 120 150 180 180 15 A1 450 450 450 450
A4 110 140 170 200 16 A2 420 480 480 480
Expected conditional pay-offs are given by 17 A3 390 450 510 510
18 A4 360 420 480 540
Prob. E1 E2 E3 E4
0.1 0.2 0.4 0.3 Multiplying the pay-off by the corresponding
A1 14 28 56 42 probability we get the following table giving
A2 13 32 64 48 expected conditional pay-off:
A3 12 30 72 54
E1 E2 E3 E4 EMV
A4 11 28 68 60
A1 45 90 180 135 450
é Profit per dozen = Rs 20 ù A2 42 96 192 144 474
êë Loss per unsold dozen = Rs 10 úû A3 39 90 204 153 486 Æ
A4 36 84 192 162 474
EMV of A1 = Rs 140
EMV of A2 = Rs 157 Since the highest EMV is 486, the course of
EMV of A3 = Rs 168 Æ action to be chosen is A3. Thus the merchant should
EMV of A4 = Rs 167 purchase 17 baskets everyday.
We find that A3 has the maximum EMV and
hence A3 is to be chosen. In other words the 13.7 EXPECTED OPPORTUNITY
merchant should purchase 9 dozens of roses LOSS CRITERION (EOL)
everyday and the maximum expected profit is
Rs 168. This method is the same as that of EMV criterion
except that, here we choose the act which leads to
Example 13.8 A fruit merchant purchases
minimum loss. EOL is the amount of pay-off that
orange at Rs 50 per basket and sells them at
is lost by not selecting the course of action that
Rs 80. Unsold oranges are disposed off at Rs 20 per
has the greatest pay-off for the event that actually
basket the next day. Past record of sales for 120
occurs. For a given event find the difference
days is as given below:
between the maximum conditional pay-off and the
Baskets sold 15 16 17 18 pay-offs corresponding to different actions. This
is called Conditional Opportunity Loss (COL).
No. of days 12 24 48 36
Multiplying these values by the respective
Find how many baskets the merchant should probabilities we obtain the EOL. The sum of the
purchase everyday in order to maximize his profit. expected COL of an action is the EOL. The action
yielding the minimum EOL is chosen. The steps
Solution Here unsold baskets have a salvage of this procedure are
value of Rs 20 each. Therefore the loss per basket
is 50 – 20 = 30. Thus the profit is given by 1. Determine the conditional pay-off for each
30 (no. of baskets sold) – 30 (no. of baskets unsold) act event combination.
 Decision Theory 155

2. For each event find the COL values by COL table

subtracting the pay-offs of the actions from
E1 E2 E3 E4 E5
the maximum value.
3. For each action calculate the expected COL A1 0 2 4 6 8
values by multiplying with the correspond- A2 3 0 2 4 6
ing probabilities and find their sum (EOL). A3 6 3 0 2 4
A4 9 6 3 0 2
4. Select the act having the minimum EOL.
A5 12 9 6 3 0
Example 13.9 A newspaper boy purchases Prob. 0.10 0.15 0.20 0.25 0.30
magazines at Rs 3 each and sells them at Rs 5
each. He cannot return the unsold magazines. The Expected COL
probability distribution of the demand for the
magazine is given below. E1 E2 E3 E4 E5 EOL
A1 0 0.3 0.8 1.5 2.4 5
Demand 16 17 18 19 20 A2 0.3 0 0.4 1 1.8 3.5
Probability 0.10 0.15 0.20 0.25 0.30 A3 0.6 0.45 0 0.5 1.2 2.75 Æ
A4 0.9 0.9 0.6 0 0.6 3
A5 1.2 1.35 1.2 0.75 0 4.5
Determine how many copies of magazine
should he purchase (Apply EOL criterion). The course of action A3 has the minimum EOL.
Solution Therefore it is advisable that the newspaper boy
The events are should purchase 18 copies of the magazine in order
to minimize his loss.
E1: Demand for 16
E2: Demand for 17 Example 13.10 The sales of butter during 200
E3: Demand for 18 days has the following distribution:
E4: Demand for 19 Demand in kg 14 18 22 24 30 40
E5: Demand for 20 No. of days 10 20 20 80 40 30
The courses of action are
The cost of butter is Rs 80 per kg and it is sold
A1: Purchase 16 at Rs 120 per kg. Unsold stock is disposed off at
A2: Purchase 17 the rate of Rs 60 per kg the next day. Apply EMV
A3: Purchase 18 criterion and EOL criterion to determine how much
A4: Purchase 19 of butter is to be stocked at the beginning of each
A5: Purchase 20 day.
Conditional pay-off table Solution (i) The conditional pay-off table is as
follows: (Loss = 80 – 60 = Rs 20 per kg)
Demand
14 18 22 24 30 40
E1 E2 E3 E4 E5 E1 E2 E3 E4 E5 E6
16 17 18 19 20
14 A1 560 560 560 560 560 560
A1 16 32 32 32 32 32 18 A2 480 720 720 720 720 720
Purchase

A2 17 29 34 34 34 34
Stock

22 A3 400 640 880 880 880 880

A3 18 26 31 36 36 36 24 A4 360 600 840 960 960 960
A4 19 23 28 33 38 38 30 A5 240 480 720 840 1200 1200
A5 20 20 25 30 35 40 40 A6 40 280 520 640 1000 1600
Max 32 34 36 38 40
Prob. 0.05 0.10 0.10 0.40 0.20 0.15
 156 Operations Research

Expected conditional pay-off table COL table

E1 E2 E3 E4 E5 E6 EMV E1 E2 E3 E4 E5 E6

A1 28 56 56 224 112 84 560 A1 0 160 320 400 640 1040

A2 24 72 72 288 144 108 708 A2 80 0 160 240 480 880
A3 20 34 88 352 176 132 832 A3 160 80 0 80 320 720
A4 18 60 84 384 192 144 882 A4 200 120 40 0 240 640
A5 12 48 72 336 240 180 888 A5 320 240 160 120 0 400
A6 2 28 52 256 200 240 778 A6 520 440 360 320 200 0
Prob. 0.05 0.10 0.10 0.40 0.20 0.15
A5 has the maximum EMV. Hence according to
EMV criterion 30 kg of butter is to be stocked The expected COL table is given below.
everyday. E1 E2 E3 E4 E5 E6 EOL
(ii) Conditional pay-off table A1 0 16 32 160 128 156 492
A2 4 0 16 96 96 132 344
14 18 22 24 30 40 A3 8 8 0 32 64 108 220
E1 E2 E3 E4 E5 E6 A4 10 12 4 0 48 96 170
A5 16 24 16 48 0 60 164 Æ
14 A1 560 560 560 560 560 560 A6 26 44 36 128 40 0 274
18 A2 480 720 720 720 720 720
22 A3 400 640 880 880 880 880 The minimum EOL corresponds to A5. Hence
24 A4 360 600 840 960 960 960 according to EOL criterion 30 kg of butter is to be
30 A5 240 480 720 840 1200 1200 stocked.
40 A6 40 280 520 640 1000 1600 Note: We find that both EMV and EOL criteria
are consistent and they give identically the same
Max 560 720 880 960 1200 1600 results.

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Apply Maximax, Maximin and Laplace 2. Apply Hurwicz the criterion with a = 0.6
criteria for the following decision problems: for the following problems:
(a) Events (a) Strategies
E1 E2 E3 E4 A B C D
A1 400 –10 600 1800 E1 3 4 4 5
Events
Acts

A2 2000 500 40 0 E2 6 2 1 3
A3 2000 1500 –200 100 E3 1 8 8 4

(b) Events
(b) E1 E2 E3 E4 Alternatives I II III
A1 4 0 –5 3 A 3 3 7
A2 –2 6 9 1 B 4 3 2
A3 7 3 2 4 C 2 5 3
 Decision Theory 157

3. Solve the following problems using EMV (a) The pay-offs of three acts A1, A2, A3 and
criterion: the events E1, E2, E3 are given below:
(a) A business man buys an item at Rs 8
and sells it at Rs 12. Unsold items are E1 E2 E3
useless. The following table gives the A1 25 400 650
sales details for the past 100 days. A2 –10 440 740
A3 –125 400 750
Number of
25 26 27 28 29
items sold The probabilities of the events E1, E2,
No. of days 10 10 25 40 15 E3 are 0.1, 0.7 and 0.2 respectively.
Determine which act can be chosen as
Determine how many items he has to the best.
purchase in order to maximize his profit. (b) A company is proposing manufacture of
(b) The cost of a cake is Rs 2 and it is sold three products X, Y and Z. The demand
at Rs 4. Unsold cakes are given free of for each product may be good, moderate
cost to poor people at the end of the day. or poor. The probabilities of the states
The probability distribution of the of nature are given below.
demand for the cakes is given below.
Nature of demand
Demand 15 16 17 18 19
Product Good Moderate Poor
Prob. 0.1 0.2 0.3 0.25 0.15
X 0.7 0.2 0.1
Determine the number of cakes to be Y 0.5 0.3 0.2
purchased in order to get maximum Z 0.4 0.5 0.1
profit. The estimated profit or loss in thousands
(c) A farmer wants to decide which of the of rupees under the three states is given
three crops rice, wheat or corn he should below.
cultivate in his 100 acre farm. The prof-
its from the crops depend on the nature Good Moderate Poor
of rainfall: high, moderate or low. The X 30 20 10
following table gives the pay off from Y 60 30 20
the crops under different conditions Z 40 10 –15
(events).
Give your advice to the company about
Profit the choice of the product.
Rice Wheat Corn (c) A boat manufacturer finds that the
distribution of the demand for the boat
Rainfall

High 7000 2500 4000

Moderate 3500 3500 4000 is as follows:
Low 1000 4000 3000
Demand 0 1 2 3 4 5 6
The probabilities of high, moderate and Prob. 0.14 0.27 0.27 0.18 0.09 0.09 0.01
low rainfall are 0.2, 0.3 and 0.5
respectively. Determine which crop Each boat costs Rs 7000 and he sells it
should be cultivated. at Rs 10,000. Unsold boats are disposed
off at Rs 6000 each at the end of the
4. Apply EOL criterion for solving the season. How many boats should he
following decision problems: stock?
 158 Operations Research

5. Apply Minimax Regret Criterion A cake costs Rs 3 and sells for Rs 5. Apply
E1 E2 E3 EOL criterion and determine how many
A1 70 30 15 cakes should be purchased in order to
A2 50 45 10 maximize the profit (unsold cakes are given
A3 30 30 20 to poor people free of cost).
10. The cost of manufacturing an item is Rs 25
6. Apply Minimax Regret Criterion and the selling price is Rs 30. Unsold items
E1 E2 E3 E4 are sold at Rs 20 per item at the end of the
week. The following table shows the weekly
A1 40 –1 60 180
sales and the number of weeks from past
A2 200 50 4 0
experience.
A3 200 150 –20 10
Weekly sales less 4 5 6 7 greater
7. Apply EMV criterion than 3 than 7
A1 A2 A3 Prob. No. of weeks 0 10 20 40 30 0
E1 25 –10 –125 0.1
Find the optimum number of items to be
E2 400 440 400 0.7
manufactured per week.
E3 650 740 750 0.2
11. The probability distribution of monthly sales
8. A pharmacist purchases a particular vaccine of an item is as follows:
on Monday each week. Unsold vaccine Monthly sales 0 1 2 3 4 5 6
becomes useless the next week. The cost is Probabilities 0.01 0.06 0.25 0.30 0.22 0.10 0.06
Rs 2 per dose and the pharmacist sells it at
Unsold items are stored at a cost of Rs 30
Rs 4 per dose. The demand for the medicine
per unit, per month and the loss due to
over the past 50 weeks is given below:
shortage is Rs 70 per unit. Determine the
Doses per week 20 25 50 60 optimum stock per month.
No. of weeks 5 15 25 5 12. The cost of storing a TV for a week is Rs 50.
Loss due to shortage is Rs 200 per TV. The
Using EMV. criterion determine how many probability distribution of the demand is
doses he should buy every week. given below:
9. The following table gives the information
Weekly demand 0 1 2 3 4 5
on the sales of cakes in a bakery.
Probabilities 0.05 0.1 0.2 0.3 0.2 0.15
Sales per day 25 26 27 28
How many TV sets should the dealer order
No. of days 10 30 50 10 per week?

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. (a) Maximax A2, A3; Maximin A2; Laplace A3 6. Optimal strategy A1

(b) Maximax A2; Maximin A3; Laplace A3 7. Optimal strategy A2
2. (a) E3 (b) A 8. He should buy 50 doses every week.
9. Ether 26 or 27 cakes should be purchased.
3. (a) 27 items (b) 17 cakes (c) wheat
10. 6 items per week.
4. (a) A2 (b) Y (c) 3 11. 4 items to be stocked per month.
5. Strategy A1 is to be adopted. 12. 4 TV sets per week.

Theory of Games
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
14.1 INTRODUCTION
Theory of games deals with situation in which two
or more competitors are engaged in decision-
making activities involving profit. The competitors
are called players. Theory of games is one type of
decision-making technique in which one player’s
course of action (strategy) depends on all possible
strategies of the opponent. Each player will act so
as to maximize his gain or minimize his loss.
Different strategies adopted by the players lead to
different pay-offs. The strategy which optimizes
a player’s pay-off is called optimal strategy. When
each player plays his optimal strategy the resulting
pay-off is called the value of the game. In general
there are two types of strategies, viz. pure strategy
and mixed strategy. Pure strategy is a decision rule
used by a player to select the particular course of
action. Each player knows all strategies out of
which he selects the one, irrespective of the other’s
choice, which optimizes his pay-off. When both
players are left to guess the course with some
probability it is called mixed strategy. It is a
probabilistic method of deciding the optimal
strategy.
14
14.2 TWO-PERSON ZERO SUM
GAMES
A game with two players A and B in which a gain
for A is equal to the loss for B (total sum is zero) is
called a two-person zero sum game. This game is
represented by the pay-off (gain) matrix of one
player A. For B the same pay-off matrix applies
with the signs of the values changed. The strategies
of A are denoted by A1, A2, …, Am and those of B
are B1, B2, …, Bn. For each combination of Ai
(i = 1, 2, …, m) and Bj ( j = 1, 2, 3, …, n) we have
a pay-off aij. Generally we assume that aij denotes
the pay-off for A. Of course B’s pay-off is –aij.
Pay of matrix is of the following form:
B’s strategies
æ B1 B2 B3 ...... Bn ö
A1 ç a11 a12 a13 ...... a1n ÷
ç ÷
A’s strategies A2 ç a21 a22 a23 ..... a2 n ÷
.... çç .... ..... ..... ..... ..... ÷÷
A çè a
m m1 a
m2 a
m3 ..... a ÷ø
mn
Some important characteristics of two-person
zero sum game are
 160 Operations Research

(i) There are exactly two players and each player (iii) Note down the maximum pay-off in each
has a finite number of courses of action column.
(strategies). (iv) Select the minimum among these and square
(ii) The given matrix refers to the pay-off values it. It is the minimax value.
for A. (v) If the maximin and minimax values are equal
(iii) Each player knows all possible pay-offs for then mark the position of that pay-off in the
himself as well as for the other player. matrix. This element represents the value of
the game and it is called the saddle point of
14.3 GAMES WITH SADDLE the game.
POINT (PURE STRATEGY) This procedure is called solving the game with
the given pay-off matrix.
Maximin and Minimax Principle Note: In general maximin v £ v £ minimax v
Each row of the pay-off matrix represents If v = 0 = v the game is said to be fair.
the pay-off for A, corresponding to different If v = v = v the game is said to determinable
strategies of B. In the first row we find the pay-off (stable).
values of A1 corresponding to B1, B2, ..., Bn. A Example 14.1 Solve the game with the pay-off
wants to ensure the minimum profit. Hence the matrix
minimum of the I row is taken. Similarly for B1 B2 B3 B4
A2 the minimum of the II row is chosen and thus A1 1 7 3 4
we note down the minimum pay-off in each row.
Thus A is assured of these pay-offs for his actions A2 5 6 4 5
A1, A2, A3, …, Am. Now he selects the maximum A3 7 2 0 3
among these (maximin) value v and plays the
Solution
corresponding strategy. In case of B, each column
gives his loss corresponding to different strategies B1 B2 B3 B4 Row minimum
of A. He wants to minimize his loss. But A adopts A1  1 7 3 4  1
the strategy giving maximum gain. B cannot lose A2  5 6 4 5  4
more than the maximum pay-off of each column A3  7 2 0 3  0
(corresponding to B1, B2, ..., Bn). In order to
minimize his loss B has to choose the minimum Col. max 7 7 4 5
among these maximum pay-offs v . This value is Maximin value = 4 = Minimax value. v = v = v .
the minimax value and he chooses to play the The value of the game is 4. Saddle point is
corresponding strategy. (2, 3). Optimal strategy of A is A2 and that of B
If the maximin and the minimax values are is B3.
equal then the game is said to have Saddle point Example 14.2 Solve the following game:
and the corresponding pay-off is called the value
B1 B2 B3 B4
of the game v. ( v = v = v). The step-by-step
procedure is given below. A1 20 15 12 35
(i) Note down the minimum pay-off in each A2 25 14 8 10
row. A3 40 2 10 5
(ii) Select the maximum among these and square
it. It is the maximin value.
A4 −5 4 11 0
 Theory of Games 161

Solution Row minimum and column maximum Solution Ignoring p and q we find the maximin
values are given below and minimax values.
B1 B2 B3 B4 Row minimum B1 B2 B3 Row minimum
A1 20 15 12 35 12 maximin A1  2 6 5 2
A2 25 14 8 10 8  
A2  10 7 q  7 maximin
A3 40 2 10 5 0  5 p 8 5
A3  
A4 −5 4 11 0 − 5 Col. max 10 7 8
minimax
Col. max 40 15 12 35 v = v = v = 12
We find that v = v = 7. Given that the saddle
minimax point is at (2, 2). Therefore v = 7 and p £ 7 and
We find that maximin = minimax =12. q ≥ 7.
Therefore the value of the game is 12 and the
saddle point is (1, 3). The optimal strategies are 14.4 GAMES WITHOUT SADDLE
A1 of A and B3 of B.
POINTS (MIXED STRATEGY)
Example 14.3 If the following game is deter-
When there is no saddle point the players select
minable find the limits for the value of λ .
the strategies with some probabilities. A selects
B1 B2 B3 his strategies A1, A2, A3, …, Am with probabilities
A1  λ 6 4 p1, p2, …, pm respectively, where p1 + p2 + p3 +
  … + pm = 1.
A2  −1 λ −7 
B selects his strategies B1, B2, B3, …, Bn with
A3  −2 4 λ  probabilities q1, q2, q3,…, qn where q1 + q2 + …
Solution Ignoring the value of λ the maximin + qn = 1. For each strategy of A we can find the
and minimax values are expected gain.
B1 B2 B3 Row minimum
A1  λ 6 4  4 maximin
14.5 FORMULA FOR FINDING
  THE VALUE OF THE GAME
A2  −1 λ −7  −7
IN CASE OF 2 × 2 GAMES
A3  −2 4 λ  −2 WITHOUT SADDLE POINT
Col. max −1 6 4
Let the pay-off matrix for A be
minimax
maximin value = 4 B1 B2
minimax value = –1 A1  a11 a12 
If the game is determinable then v = v = v .  
A2  a21 a22 
Here v = 4 and v = –1
The strategies and probabilities are
\ l = lies between –1 and 4.
 A1 A2   B1 B2 
Example 14.4 If the saddle point is (2, 2) in the SA =  p p  and SB =  q q 
following pay-off matrix find the range of the  1 2  1 2 

values of p and q. A selects A1 with prob p1 and A2 with prob p2.

Corresponding to B’s strategy B1, the expected gain
 2 6 5 of A is E1 (p) = a11 p1 + a21 p2.
 
 10 7 q  Corresponding to B’s strategy B2 the expected
 5 p 8 gain of A is E2(p) = a12p1 + a22p2.
 
 162 Operations Research

Similarly, corresponding to A’s strategy A1 the These formulae are important and are used
expected loss of B is E1 (q) = a11q1 + a12q2. frequently in solving games without saddle point.
Corresponding to A2, expected loss of B is E2(q)
Example 14.5 Solve the following game
= a21q1 + a22q2.
If the value of the game is v, then B1 B2
E1( p) ≥ v and E2( p) ≥ v A1  8 −3 
E1( p) £ v and E2( q) £ v  
A2  −3 1 
For optimal strategy we should have E1( p) =
E2( p) = v and E1(q) = E2(q) = v. Solution First let us find whether the matrix has
\ a11p1 + a21p2 = a12p1 + a22p2. a saddle point.
p1(a11 – a12) = p2 (a22 – a21) B1 B2 Row min
a22 − a21
p1/p2 = A1  8 −3  −3
a11 − a12  
A2  −3 1  −3
a11 − a12
or p2/p1 = a Col. max 8 1
22 − a21 Since there is no saddle point we apply the
Also p1 + p2 = 1. method of mixed strategy.
a −a (a22 − a21 )
p2/p1 + 1 = 11 12 + 1 p1 =
a22 − a21 (a11 + a22 ) − (a12 + a21 )
a11 − a12 + a22 − a21
(p1+ p2)/p1 = 1+ 3
a22 − a21 = = 4/15
(8 + 1) − (−3 − 3)
(a11 + a22 ) − (a12 + a21 ) p2 = 1 – p1 = 11/15
1/p1 =
(a22 − a21 ) (a22 − a12 )
q1 =
(a22 − a21 ) (a11 + a 22 ) − (a12 + a21 )
\ p1 = (1)
(a11 + a22 ) − (a12 + a21 ) (1 + 3)
= = 4/15
(a11 − a12 ) (8 + 1) − (−3 − 3)
\ p2 = 1 – p1 = (2)
(a11 + a22 ) − (a12 + a21 ) q2 = 1 – q1 = 11/15
Again taking E1 (q) = E2 (q) we get Value of the game is
(a − a12 ) (a11a22 − a12 a21 ) 8−9
q1/q2 = 22 = = – 1/15
(a11 − a21 ) (a11 + a22 ) − (a12 + a21 ) 15
(a22 − a12 ) The optimal strategies are
and q1 = (3)
(a11 + a22 ) − (a12 + a21 )  A A2 
SA =  1 
(a11 − a21 )  4 15 11 15 
q2 = (4)
(a11 + a22 ) − (a12 + a21 )  B1 B2 
SB =  
The value of the game is given by  4 15 11 15 
v = a11p1 + a21p2
a11 (a22 − a21 ) + a21 (a11 − a12 ) Example 14.6 Solve the following game
= B1 B2
(a11 + a22 ) − (a12 + a21 )
A1  5 1
(a11a22 − a21a12 )
=
(a11 + a22 ) − (a12 + a21 )
(5) A2  3 
4
 Theory of Games 163

Solution 14.6 DOMINANCE PROPERTY

B1 B2 Row min
Dominance property is used to reduce the size of
A1  5 1  1 the pay-off matrix. In the pay-off matrix if each
  element of the ith row is less than or equal to the
A2  3 4  3
corresponding element of some other row (say jth
Col. max 5 4
row) then we say that the jth row dominates over
No saddle point.
the ith row and the ith row can be deleted from the
Now take D = (a11 + a22) – (a12 + a21)
pay-off matrix, since it implies that A’s pay-off is
= (5 + 4) – (3 + 1) = 5 less for the strategy Ai than for Aj whatever be the
p1 = (a22 – a21)/D = (4 – 3)/5 = 1/5; p2 = 4/5 strategy of B. Ai is an inferior strategy.
q1 = (a22 – a12)/D = (4 – 1)/5 = 3/5; q2 = 2/5 Similarly, if every element of the rth column is
Value = (a11a22 – a12a21) /D = (20 – 3)/D = 17/5 greater than or equal to the corresponding element
Optimal strategies are of some other column (say sth column) then B will
never play the rth strategy as it involves more loss.
 A1 A2   B1 B2  We say that sth column dominates over the rth
SA =  1 5 4 5  SB =  3 5 2 5 
    coloumn and rth column can be deleted. Br is an
inferior strategy.
Example 14.7 In a game of matching biased Also if a strategy k(Ai or Bj) is inferior to a
coins, player A gets Rs 8 if two heads turn up and convex combination of two or more of other pure
Re 1 if two tails turn up. Player B gets Rs 3 if it is strategies then the row or column corresponding
otherwise. Solve the game. to Ai or Bj can be deleted.
Solution A’s pay-offs are 8 for (HH), 1 for (TT), Note: A convex combination of Ar and As is
– 3 for (HT), and – 3 for (TH) (l1 ar1 + l2 as1, l1 ar2 + l 2 as2, ..., l1 arm + l 2 asn),
The matrix is where l 1 + l 2 = 1; 0 < l 1 < 1 and 0 < l 2 < 1.
B1 B2 Using this principle of dominance we can reduce
the pay-off matrix to a smaller one and then solve.
A1  8 −3 
  Example 14.8 Solve the following game
A2  −3 1 
B1 B2 B3
It is a game without saddle point.
D = a11 + a22 – (a12 + a21) A1  1 7 2 
 
= (8 + 1) – (– 3 – 3) = 15 A2  0 2 7 
p1 = (a22 – a21)/D = (1 + 3)/15 = 4/15; A3  5 1 6 
p2 = 11/15 The given pay-off matrix is
q1 = (a22 – a12)/D = (1 + 3)/15 = 4/15 B1 B2 B3 Row min
q2 = 11/15 A1  1 7 2  1
v = (a11a22 – a12a21)/D = (8 – 9)/15  
A2  0 2 7  0
= – 1/15 
A3  5 1 6   1
 H T  Col. max 5 7 7
SA =  
 4 15 11 15  The game has no saddle point. We find that
every element in the III column is greater than the
 H T  corresponding element in column I. Therefore,
and SB =  
 4 15 11 15  column III is inferior and it can be deleted.
 164 Operations Research

Now we have There is no saddle point.

B1 B2 Row III dominates over row I and hence row I
A1  1 7  can be deleted. Thus we get
  B1 B2 B3 B4
A2  0 2 
A2  3 4 2 4 
A3  5 1   
In this matrix every element of row II is less A3  4 2 4 0 
than the corresponding element of row I. A4  0 4 0 8 
Therefore, row II is inferior and it can be deleted. Here we find that column I is dominated by
The resulting pay-off matrix is column III and therefore we can delete column I.
B1 B 2 The resulting matrix is
A1  1 7 
  B2 B3 B4
A3  5 1 
A2  4 2 4
Considering the mixed strategy we get  
D = (1 + 1) – (5 + 7) = 2 – 12 = – 10. A3  2 4 0
p1 = (a22 – a21)/D = (1 – 5)/D A4  4 0 8 
= – 4/ – 10 = 2/5; p3 = 3/ 5
We find that column I is inferior to a convex
q1 = (a22 – a12)/D = (1 – 7)/D
combination of columns II and III.
= – 6/ – 10 = 3/5;
q2 = 2/5 4 ≥ (2 + 4)/2, 2 ≥ (4 + 0)/2,
Value of the game is (a11a22 – a12a21)/D 4 ≥ (0 + 8)/2, (l 1 = l 2 = 1/2)
= (1 – 35)/ – 10 \ column 1 can be deleted and the resulting
matrix is
= – 34/ – 10 = 17/5
The optimal strategies are B3 B4
 A1 A2 A3  A2  2 4 
SA =  2 5 0 3 5  A3  4 0 
 
 B1 B2 B3  A4  0 8 
and SB =  3 5 2 5 0  Again in this matrix, row I is dominated by a
 
convex combination of rows II and III.
Example 14.9 Solve 2 £ (4 + 0); 4 £ (0 + 8)/2
B1 B2 B3 B4 and hence row I can be deleted.
A1  3 2 4 0 Finally we get
A2  3 4 2 4
 B3 B4
A3  4 2 4 0 A3  4 0 
   
A4  0 4 0 8  A4  0 8 
Solution The given pay-off matrix is D = (4 + 8) – (0 + 0) = 12
B1 B2 B3 B4 Row min p3 = 8/12 = 2/3, p4 = 1/3
q3 = 8/12 = 2/3, q4 = 1/3
A1  3 2 4 0  0
  Value of the game is
A2  3 4 2 4  2 v = 32 / 12 = 8/3
A3  4 2 4 0  0 The optimal strategies are

 

A4  0 4 0 8  0 æ A1 A2 A3 A4 ö
SA = çè 0 0 2 3 1 3÷ø
Col. max 4 4 4 8
 Theory of Games 165

 B1 B2 B3 B4  Example 14.11 Solve

SB =  0  B
 0 2 3 1 3
æ 5 -10 9 0ö
Example 14.10 Solve ç6 7 8 1÷
B ç ÷
A ç8 7 15 1÷
æ 1 7 2ö çè 3 4 -1 4 ÷ø
A = ç 6 2 7÷
ç ÷ Solution Given matrix is
è 5 1 6ø
B1 B2 B3 B4 Row min
Solution The given pay-off matrix is A1 æ 5 -10 9 0ö -10
B1 B2 B3 Row min A2 ç 6 7 8 1÷ 1
A1 æ 1 7 2 ö 1 ç ÷
A3 ç 8 7 15 1÷ 1
A2 ç 6 2 7 ÷ 2 A4 çè 3 -1 4 ÷ø -1
ç ÷ 4
A3 è 5 1 6 ø 1 Col. max 8 7 15 4
Col. max 6 7 7 No saddle point It has no saddle point
Row II dominates over row III and therefore R3 dominates R1. R1 can be deleted.
row III can be deleted. We get B1 B2 B3 B4
We get B1 B2 B3 A2 æ 6 7 8 1ö
A1 æ 1 7 2 ö A3 ç 8 7 15 1÷
ç ÷
A2 çè 6 2 7 ÷ø A4 è 3 4 -1 4 ø
Column I dominates over column III and hence R2 dominates over R1. R1 can be deleted.
column III can be deleted. We have
We get B1 B2 B1 B2 B3 B4
A1 æ 1 7 ö A3 æ 8 7 15 1 ö
A2 çè 6 2 ÷ø A4 çè 3 4 -1 4 ÷ø
Applying mixed strategy we have C4 dominates over C2. C2 can be deleted.
D = (1 + 2) – (6 + 7) = – 10 Thus we have
p1 = (2 – 6)/ – 10 = – 4/ – 10 = 2/5; B1 B3 B4
p2 = 3/5 A3 æ 8 15 1 ö
A4 çè 3 -1 4 ÷ø
q1 = (2 – 7)/ – 10 = – 5/ – 10 = 1/2;
q2 = 1/2
C1 dominated by a convex combination C2 and C3.
Value of the game is
8 ≥ (15 + 1)/2 (l1 = l 2 = 1/2)
v = ((1) (2) – (6) (7))/ – 10 = 4
3 ≥ (–1 + 4)/2
The optimal strategies are
Hence C1 is deleted.
æ A1 A2 A3 ö Finally we get
SA = çè 2 5 3 5 0 ÷ø
B3 B4
æ B1 B2 B3 ö A3  15 1
SB = çè 1 2 1 2 0 ÷ø  
A4  −1 4 
 166 Operations Research

D = (15 + 4) – (–1 + 1) = 19 = a1j p1 + a2j (1 – p1)

p3 = (4 + 1)/19 = 5/19; p4 = 14/19 = (a1j – a2j)p1 + a2j
q3 = (4 – 1)/19 = 3/19; q4 = 16/19 Consider two vertical axes corresponding to
Value v = (60 + 1)/19 = 61/19 p1 = 0 and p1 = 1. These axes are two parallel
æ A A2
SA = 1
A3 A4 ö lines at a distance 1 unit apart. For p1 = 0 we get
çè 0 0 5 / 19 14 / 19 ÷ø the point (0, a2j) on the 0 axis and for p1 = 1 we
get the point (1, a1j) on the 1 axis. Joining these
æ B1 B2 B3 B4 ö two points we get the line corresponding to j =
and SB = çè 0 0 3 / 19 16 / 19 ÷ø 1, 2, 3, …, n. Determine the lower envelope of
all these lines. The highest point of this envelope
represents the maximum among the minimum
14.7 GRAPHICAL METHOD expected pay-off (maximin value). The lines which
The graphical method can be applied to solve the intersect at this point give the optimal strategies
game whose pay-off matrix is of the form 2 ¥ n or of B.
m ¥ 2. In this case one of the players has only two Example 14.12 Solve the following game
undominated pure strategies. Using the graphical graphically
method we can reduce the given pay-off matrix to
2 ¥ 2 type which can be solved using mixed B1 B 2 B3 B4
strategies. A1 æ 2 1 0 -2 ö
A2 çè 1 0 3 2 ÷ø
14.7.1 Procedure for 2 ¥ n Matrix
Let the 2 ¥ n pay-off matrix be Solution
B1 , B2 , B3 ..., Bn Prob. æ A1 A2 ö
SA = ç
è p1 p2 ÷ø
Let
A1 æ a11 , a12 , a13 , ..., a1n ö p1
ç ÷
A2 è a21 , a22 , a23 , ..., a2 n ø p2 We have the expected pay-off as follows:
Prob. q1 , q2 , q3 , ..., qn B’s move
A’s expected pay-off corresponding to each B1 Æ E1 (p1) = (a11 – a21)p1 + a21
strategy of B are given by = (2 – 1)p1 + 1
B’s strategy A’s expected pay-off i.e., E1 (p1) = p1 + 1 (1)
B1 a11p1 + a21p2 Similarly, B2 Æ E2 (p1) = (a12 – a22)p1 + a22
B2 a12p1 + a22p2 = (1 – 0)p1 + 0 = p1
B3 a13p1 + a23p2 E2 (p1) = p1 (2)
M M B3 Æ E3 (p1) = (a13 – a23)p1 + a23
Bn a1n p1 + a2n p2 = (0 – 3)p1 + 3
B chooses strategy Bj for which Ej ( p1) is E3 (p1) = – 3p1 + 3 (3)
minimum.
B4 Æ E4(p1) = (a14 – a24)p1 + a24
\ v = Minj {E j ( p1 )} ( j = 1, 2, …, n) = (– 2 – 2)p1 + 2
Now A chooses p1 and p2 in such a manner that E4 (p1) = – 4p1 + 2 (4)
v is as large as possible. In order to determine this Take two vertical axes p1 = 0 and p1 = 1. E1(p1)
value we shall plot each Ej (p1) as a straight line of is represented by the line joining E1(0) and E1(1).
the form E1(0) = 0 + 1 = 1, which is the point (0, 1)
Ej( p1) = a1j p1 + a2 j p2 E1(1) = 1 + 1 = 2, which is the point (1, 2).
 Theory of Games 167

Thus E1(p1) is represented by the line joining Example 14.13 Solve the following game using
the points (0, 1) on the 0 axis and (1, 2) on the the graphical method
1 axis. Now E2(p1) is the line joining (0, 0) and B1 B2 B3 B4
(1, 1)
A1  2 2 3 −2 
E3(p1) is the line joining (0, 3) and (1, 0)  
E4(p1) is the line joining (0, 2) and (1, –2) A2  4 3 2 6
Let us represent these 4 lines in the graph as given
Solution
below.
 A1 A2 
Let SA =  p 
p2 
 1
E1 (p1) = (2 – 4) p1 + 4 = – 2p1 + 4 (1)
E2 (p1) = (2 – 3)p1 + 3 = – p1 + 3 (2)
E3 (p1) = (3 – 2)p1 + 2 = p1 + 2 (3)
E4 (p1) = (– 2 – 6)p1 + 6 = – 8p1 + 6 (4)
E1 (0) = 4
E1 (1) = 2 Line joining (0, 4) and (1, 2)
E2 (0) = 3
E2 (1) = 2 Line joining (0, 3) and (1, 2)
E3(0) = 2
E3 (1) = 3 Line joining (0, 2) and (1, 3)
E4 (0) = 6
Fig. 14.1 E4 (1) = – 2 Line joining (0, 6) and (1, – 2)
From the graph we find that the highest point The graph is
of the lower envelope is H which is the intersection
of E2 (p1) and E4 (p1). Therefore B2 and B4 are the
optimal strategies of B. This leads to the 2 ¥ 2
pay-off matrix.
B2 B4
A1 æ 1 -2 ö
A2 çè 0 2 ÷ø
We have D = (1 + 2) – (0 – 2) = 5
p1 = (2 – 0)/5 = 2/5, p2 = 3/5
q2 = (2 + 2)/5 = 4/5, q4 = 1/5
Value of the game
v = (2 – 0)/5 = 2/5
The optimal strategies are
Fig. 14.2
 A1 A2 
SA =   In the graph the highest point of the lower
 2 5 3 5
envelope is H which is the intersection of (3) and
æ B1 B2 B3 B4 ö (4). Therefore B3 and B4 are the optimal strategies.
SB = ç
è 0 4 5 0 1 5 ÷ø We get the pay-off matrix.
 168 Operations Research

B3 B4 intersecting at H represent the optimal strategies

of A.
A1  3 −2 
  Example 14.14 Solve the following game
A2  2 6
B
D = (6 + 3) – (2 – 2) = 9
p1 = (6 – 2)/9 = 4/9, p2 = 5/9  −6 7 
 
q3 = (6 + 2)/9 = 8/9, q4 = 1/9  4 −5 
V = (18 + 4)/9 = 22/9 A  −1 −2 
 
 A1 A2   −2 5 
SA =  4 9 5 9   7 −6 
   
 1
B B 2 B3 B4  Solution The given matrix is
and SB =  0 0 8 / 9 1 / 9 
  B1 B2 Prob.
14.7.2 Procedure for m ¥ 2 Matrix A1  −6 7  p1
 
 4 −5 
Let the pay-off matrix be A2 p2
B1 B2 Prob. A3  −1 −2  p3
A1  a11 a12  p1  
 
A4  −2 5  p4
A2  a21 a22  p2 A5  7 −6  p5
 
.  . .  .
  Prob. q1 q2
.  . .  . E1 (q1) = a11q1 + a12q2 = – 6q1 + 7(1 – q1)
 
Am  am1 am 2  pm = – 13q1 + 7 (1)
Prob. q1 q2 E2 (q1) = a21q1 + a22q2 = 4q1 – 5(1 – q1)
B’s expected loss corresponding to each strategy = 9q1 – 5 (2)
of A are given by the following table: E3 (q1) = a31q1 + a32q2 = – q1 – 2(1 – q1)
A’s strategy B’s expected loss = q1 – 2 (3)
A1 E1 (q1) = a11q1 + a12q2 E4 (q1) = a41q1 + a42q2 = – 2q1 + 5(1 – q1)
A2 E2 (q1) = a21q1 + a22q2 = –7q1 + 5 (4)
A3 E3 (q1) = a31q1 + a32q2 E5(q1) = a51q1 + a52q2 = 7q1 – 6(1 – q1)
M M = 13q1 – 6 (5)
Am Em (q1) = am1q1 + am2q2 E1 (0) = 7,
A chooses the strategy for which E i (q 1) is E1 (1) = – 6 Line joining (0, 7) and (1, – 6)
maximum E2 (0) = – 5,
V = Maxi {Ei (q1)} i = 1, 2, 3, ..., m. E2 (1) = 4 Line joining (0, – 5) and (1, 4)
B chooses q1 and q2 such that V is as small as E3 (0) = – 2,
possible. Hence we have to find the minimax value. E3 (1) = –1 Line joining (0, – 2) and (1, – 1)
Draw two vertical lines q1 = 0 and q1 = 1 and E4 (0) = 5,
draw lines joining the pair of points. Ei(0) and Ei(1)
(i = 1, 2, 3,…, m). Find the upper envelope of all E4 (1) = –2 Line joining (0, 5) and (1, – 2)
these lines. The lowest point H of this envelope E5 (0) = – 6,
represents the minimax value. The lines E5 (1) = 7 Line joining (0, – 6) and (1, 7)
 Theory of Games 169

The graph is  B1 B2 
and SB =  11 20 9 20 
 
Example 14.15 Solve the game
B
 1 −3 
 
 3 5
 −1 6 
A  
 4 1
 2 2
 
 −5 0 
Solution Let the player B play the mixed strategy
æ B1 B2 ö
SB = ç q q2 ÷ø
è 1
B’s expected pay-offs corresponding to A’s pure
strategies are given by
A1 Æ E1 (q1) = 4q1 – 3,
E1 (0) = – 3, E1 (1) = 1
Fig. 14.3 A2 Æ E2 (q1) = – 2q1 + 5,
E2 (0) = 5, E2 (1) = 3
The minimax point H is the intersection (4) and
(5). Therefore the optimal strategies of A are A4 A3 Æ E3 (q1) = – 7q1 + 6,
and A5. E3 (0) = 6, E3 (1) = – 1
We have the pay-off matrix A4 Æ E4 (q1) = 3q1 + 1,
B1 B2 E4 (0) = 1, E4 (1) = 4
A4  −2 5 A5 Æ E5 (q1) = 2,
 
A5  7 −6  E5 (0) = 2, E5 (1) = 2
D = (– 2 – 6) – (7 + 5) = – 20 A6 Æ E6 (q1) = – 5q1,
p4 = (– 6 – 7)/ – 20 = 13/20 E6 (0) = 0, E6 (1) = – 5
p5 = 7/20 In the graph we draw the lines joining
q1 = (– 6 – 5)/ – 20 (0, – 3) and (1, 1) (1)
= 11/20 (0, 5) and (1, 3) (2)
q2= 9/20 (0, 6) and (1, – 1) (3)
v = (12 – 35)/ – 20 = 23/20 (0, 1) and (1, 4) (4)
 A1 A2 A3 A4 A5  (0, 2) and (1, 2) (5)
SA =   (0, 0) and (1, – 5) (6)
0 0 0 13 20 7 20 
 170 Operations Research

The graph is the matrix to 2 ¥ 2 type using dominance property

then the algebraic method is used.
Let the pay-off matrix be
B
æ a11 a12 .... a1n ö
ç ÷
A ç a21 a22 .... a2 n ÷
ç .. ÷
çè a ÷
m1 am 2 .... amn ø
Assume that the optimal strategies are
æ A1 A2 A3 .... Am ö
SA = ç
è p1 p2 p3 .... pm ÷ø

æ B1 B2 B3 .... Bn ö
and SB = çè q q2 q3 .... qn ÷ø
1
If the value of the game is v then
a11 p1 + a21 p2 + a31 p3 + ü
ï
K + am1 pm ³ v ï
a12 p1 + a22 p2 + a32 p3 + ïý Maximin
Fig. 14.4 K + am 2 pm ³ v ï principle
. . . . ï
The lowest point of the upper envelope is H ï
which is the intersection of (2) and (4). Hence A’s a1n p1 + a2n p2 + … + amn pm ≥ vïþ
optimal strategies are A2 and A4. We have the pay- Also
off matrix a11q1 + a12 q2 + K + a1n qn £ v ü
B1 B2 a21q1 + a22 q2 + K + a2 qn £ v ïï
Minimax
. . . ý principle
A2 æ 3 5ö . . . ï
A4 çè 4 1÷ø am1q1 + am 2 q2 + K + am qn £ v ïþ
D = (3 + 1) – (4 + 5) = – 5 In order to obtain the value of the game, we
p2 = (1 – 4) / – 5 = 3/5, p4 = 2/5 convert these inequalities into equations. Thus we
q1 = (1 – 5) / – 5 = 4/5, q2 = 1/5 get
v = (3 – 20) / – 5 = 17/5 a11p1 + a21p2 + … + am1pm = v
The optimal strategies are a12p1 + a22p2 + … + am2pm = v
æ A1 A2 A3 A4 A5 A6 ö M M M M
SA = ç
è 0 3/5 0 2/5 0 0 ÷ø a1np1 + a2np2 + … + amnpm = v
a11q1 + a12q2 + … + a1nqn = v
æ B1 B2 ö
and SB = ç
è 4 / 5 1 / 5÷ø
a21q1 + a22q2 + … + a2nqn = v
M M M M
14.8 ALGEBRAIC METHOD am1q1 + am2q2 + … + amnqn = v
In a game with m ¥ n type pay-off matrix if there Solving these equations we obtain the values
is no saddle point and if it is not possible to reduce of p1, p2, …, pm; q1, q2,…, qn in terms of v.
 Theory of Games 171

Using the equations For the given m ¥ n pay-off matrix we assume

p1 + p2 + … + pm = 1 æ A1 A2 ... Am ö
SA = ç
and q1 + q2 + … + qn = 1 è p1 p2 ... pm ÷ø
We can obtain the values of p’s and q’s and v
æ B1 B2 ... Bn ö
Example 14.16 Solve the game and SB = ç
B è q1 q2 ... qn ÷ø
æ -1 2 1ö Using maximin criterion for A we get
a11p1 + a21p2 + a31p3 + … + am1pm ≥ v
A 1 -2 2 ÷
ç
ç ÷ a12p1 + a22p2 + a32p3 + … + am2pm ≥ v
è 3 4 -3ø
M M M M
applying algebraic method.
a1n p1 + a2n p2 + …+ amn pm ≥ v
Solution where p1 + p2 + p3 + … + pm = 1, pi ≥ 0
æ A1 A2 A3 ö Dividing both sides by v, we get
SA = ç
è p1 p2 p3 ÷ø
Let
a11p1/v + a21p2/v + … + am1pm /v ≥ 1
æ B1 B2 B3 ö a12p1/v + a22p2/v + … + am2pm /v ≥ 1
SB = ç
è q1 q2 q3 ÷ø
and M M M M
We have – p1 + p2 + 3p3 = v a1n p1/v + a2n p2/v + … + amn pm/v ≥ 1
2p1 – 2p2 + 4p3 = v p1/v + p2 /v + … pm /v = 1/v
p1 + 2p2 – 3p3 = v Denote p1/v, p2/v, …, pm /v by x1 x2, …, xm.
– q1 + 2q2 + q3 = v We get the LPP,
q1 – 2q2 + 2q3 = v Minimize 1/v = Z = x1 + x2 + x3 + … + xm ü
3q1 + 4q2 – 3q3 = v (Max v ¤ Min 1/v) ï
Solving these simultaneous equations we get subject to a11x1 + a21x2 + a31x3 + … ï
p1 = 17v/30; p2 = 20v/30; p3 = 9v/30 + am1xm ≥ 1 ïý I
p1 + p2 + p3 = 1 fi v = 30/46 a12x1 + a22x2 + … + am2xm ≥ 1 ï
or v = 15/23 M M M M ï
p1 = 17/46, p2 = 20/46, p3 = 9/46 a1n x1 + a2n x2 + …+ amn xm ≥ 1 ïï
þ
Similarly, we get x1, x2, …, xm ≥ 0
q1 = 7v/15; q2 = 6v/15; q3 = 10v/15 Similarly, considering minimax criterion for B
q1 + q2 + q3 = 1 fi v = 15/23 we get
q1 = 7/23, q2 = 6/23, q3 = 10/23 a11q1/v + a12q2/v + … + a1nqn /v £ 1
Therefore a21q1/v + a22q2/v + … + a2nqn /v £ 1
æ A1 A2 A3 ö M M M M
SA = ç
è 17 46 10 23 9 46 ÷ø am1q1/v + am2q2/v +… amnqn/v £ 1
q1/v + q2/v +… qn/v = 1/v
æ B1 B2 B3 ö
Denote q1/v, q2/v,…, qn/v by y1, y2, y3,…, yn
and SB = ç
è 7 23 6 23 10 23÷ø We get the LPP
and value of the game is v = 15/23 Maximize 1/v = W = y1 + y2 + y3 + … + yn ü
(Min v ¤ Max 1/v) ï
14.9 LINEAR PROGRAMMING ï
subject to a11y1 +a12y2 + … + a1nyn £ 1
ï
METHOD a21y1 + a22y2 + … + a2nyn £ 1 ý II
ï
Problems which are solved using the algebraic M M M M ï
method can be solved using linear programming am1y1 + am2y2 + … + amnyn £ 1 ï
y1, y2, …, yn ≥ 0 ïþ
method also.
 172 Operations Research

We find that I and II are dual problems. æ B1 B2 B3 ö

SB = ç
è q1 q2 q3 ÷ø
Therefore we can solve one of these and obtain and
the solution of the other using the principle of
duality. The values of x 1, x 2, …, xm give A’s Considering the pay-off for A the LPP is
strategies and y1, y2, …, yn yield B’s strategies and Minimize Z = 1/v = x1 + x2 + x3 ü
subject to the constraints ï
the value of the game can be determined. ï
9x1 + 5x3 ≥ 1 ï
Example 14.17 Solve using the LPP method x1 + 6x2 + 2x3 ≥ 1 ý I
ï
B 4x1 + 3x2 + 8x3 ≥ 1 ï
x1, x2, x3 ≥ 0 ï
æ 9 1 4ö ïþ
ç
A ç 0 6 3÷
÷ The corresponding LPP for B’s pay-off is
è 5 2 8ø Maximize W = 1/v = y1 + y2 + y3 ü
ï
subject to the constraints ï
Solution The given pay-off matrix is 9y1 + y2 + 4y3 £ 1 ï
ý II
B1 B2 B3 6y2 + 3y3 £ 1 ï
5y1 + 2y2 + 8y3 £ 1 ï
A1 æ 9 1 4 ö y1, y2, y3 ≥ 0
ï
ïþ
ç
A2 ç 0 6 3 ÷
÷
This is the dual of problem I. Let us solve
A3 è 5 2 8 ø problem II. The standard form is
The matrix has no saddle point. Maximize W = y1 + y2 + y3 + 0s1 + 0s2 + 0s3
subject to the constraints
æ A1 A2 A3 ö 9y1 + y2 + 4y3 + s1 = 1
Take SA = ç
è p1 p2 p3 ÷ø 6y2 + 3y3 + s2 = 1
5y1 + 2y2 + 8y3 + s3 = 1
The simplex algorithm is y1, y2, y3, s1, s2, s3 ≥ 0
1 1 1 0 0 0
CB B y1 y2 y3 s1 s2 s3 XB
0 s1 9 1 4 1 0 0 1 ¨R1/9
0 s2 0 6 3 0 1 0 1
0 s3 5 2 8 0 0 1 1 R3 – 5/9R1
Cj –1≠ –1 –1 0 0 0 0
1 y1 1 1/9 4/9 1/9 0 0 1/9 R1 – 1/54R2
0 s2 0 6 3 0 1 0 1 ¨ R2 / 6
0 s3 0 13/9 52/9 –5/9 0 1 4/9 R3 – 13/54 R2
Cj 0 – 8/9≠ – 5/9 1/9 0 0 1/9
1 y1 1 0 7/18 1/9 – 1/54 0 5/54 R1 – 1/13R3
1 y2 0 1 1/2 0 1/6 0 1/6 R2 – 9/19 R3
0 s3 0 0 91/18 – 5/9 – 13/54 1 11/54 R3 × 18/91R3
Cj 0 0 – 1/9 ≠ 1/9 4/27 0 7/27
1 y1 1 0 0 2/13 0 – 1/13 1/13
1 y2 0 1 0 5/91 4/21 – 9/91 40/273
1 y3 0 0 1 – 10/91 – 1/21 18/91 11/273
Cj 0 0 0 9/91 1/7 2/91 24/91
*
Optimality reached W = 24/91
 Theory of Games 173

\ V = 91/24 2 2
y1 = 1/13 fi q1 = v/13 = 7/24 x3 = fi p3 = ¥ v = 1/12
91 91
y2 = 40/273 fi q2 = 40v/273 = 5/9 Thus we get
y3 = 11/273 fi q3 = 11v/273 = 11/72
æ A1 A2 A3 ö
By principle of duality we get (from the Cj of SA = ç
è 3 8 13 24 1 12 ÷ø
the optimal table)
9 9 æ B1 B2 B3 ö
x1 = fi p1 = ¥ v = 3/8 and SB = çè 7 24 5 9 11 72 ÷ø
91 91
1 1 v = 91/24
x2 = fi p1 = ¥ v = 13/24
7 7

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Solve the games with the following pay-off æ1 3 2 7 4ö

matrices ç3 4 1 5 6÷
(g) ç ÷
æ 1 7 3 4ö ç6 5 7 6 5÷
(a) ç 5 6 4 5 ÷ çè 2 0 6 3 1 ÷ø
ç ÷
è 7 2 0 3ø æ -3 4 2 9 ö
1 3 1 (h) ç 7 8 6 10 ÷
ç ÷
(b)  0 −4 −3 

è 6 2 4 -1ø
1 5 −1 
 æ 10 4 2 9 1ö
æ 15 2 3ö ç 7 6 5 7 8÷
(i) ç ÷
(c) ç 6 5 7 ÷ ç 3 5 4 4 9÷
ç ÷
è -7 4 0 ø çè 6 7 3 3 2 ÷ø
æ 2 6 2ö æ -1 0 0 5 3ö
(d.) ç -1 2 -7 ÷ ç 3 2 2 2 2÷
ç ÷ (j) ç ÷
è -2 4 2 ø ç -4 -3 0 2 6÷
æ 3 -1 4 6 7 ö çè 5 3 -4 2 6 ÷ø
ç -1 8 2 4 12 ÷
(e) ç ÷ æ 3 -5 0 6 ö
ç 16 8 6 14 12 ÷ ç ÷
çè 1 11 -4 2 1÷ø (k) ç -4 -2 1 2 ÷
è 5 4 2 3ø
æ 6 1 3ö æ -2 15 -2 ö
(f) ç 0 9 7 ÷ ç ÷
ç ÷ (l) ç -5 -6 -4 ÷
è 2 3 4ø è -5 20 -8ø
 174 Operations Research

æ -5 3 1 10 ö æ -2 5ö
(m) ç 5 5 4 6 ÷ ç -5 3÷
ç ÷ ç ÷
è 4 -2 0 -5ø (e) ç 0 -2÷
ç -3 0÷
æ 30 40 -80 ö ç ÷
ç ÷ çè 1 -4 ÷ø
(n) ç 0 15 -20 ÷
è 90 20 50 ø æ 2 4ö
æ 2 -2 4 1ö ç 2 3÷
ç ÷
ç 6 1 12 3÷ (f) ç 3 2 ÷
(o) ç ÷ çè -2 6 ÷ø
ç -3 2 0 6 ÷
çè 2 -3 7 1÷ø
æ 1 3 11ö
(g) ç
æ -1 -2 8ö è 8 5 2 ÷ø
(p) ç 7 5 -1÷ æ 1 3 -1 4 2 -5ö
ç ÷ (h) çè -3 5 6 1 2 0 ÷ø
è 6 0 12 ø
æ 8 10 9 14 ö æ -2 -4 3 4ö
(q) ç 10 11 8 12 ÷ (i) çè -6 -5 2 1÷ø
ç ÷
è 13 12 14 13ø æ 3 5ö
æ 2 4 3 8 4ö ç -1 6÷
ç 5 6 3 7 8÷ (j) ç ÷
(r) ç ÷ ç 4 1÷
ç 6 7 9 8 7÷ çè 2 2 ÷ø
çè 4 2 6 4 3÷ø
æ 19 6 7 5ö
æ 6 8 6ö ç 7 3 14 6 ÷
(s) ç ç ÷
è 4 12 2 ÷ø (k) ç 12 8 18 4 ÷
2. Use graphical method to solve çè 8 7 13 -1÷ø
æ 3 0 6 -1 7 ö Hint: Reduce using dominance property.
(a) ç
è -1 5 -2 2 1÷ø 3. Use algebraic method to solve
æ 1 0 4 -1ö æ 1 -1 -2 ö
(b) çè -1 1 -2 5÷ø (a) ç -1 1 1÷
ç ÷
è 2 -1 0 ø
æ 2 -4 6 -3 5ö
(c) ç æ 9 1 4ö
è -3 4 -4 1 0 ÷ø
(b) ç 0 6 3 ÷
ç ÷
æ 2 3ö è 5 2 8ø
ç 6 7÷
ç ÷ æ 4 2 4ö
(d) ç -6 10 ÷ (c) ç 2 4 0 ÷
ç -3 -2÷ ç ÷
ç ÷ è 4 0 8ø
çè 3 2÷ø
 Theory of Games 175

æ 1 -1 3ö æ -1 2 1ö
ç ÷ (b) ç 2 -2 2 ÷
(d) ç 3 5 -3÷ ç ÷
è 6 2 -2 ø è 3 3 -3ø
æ 1 -2 1ö æ 0 2 2ö
(e) ç -1 3 2 ÷ (c) ç 3 -1 3÷
ç ÷ ç ÷
è -1 -2 3ø è 4 4 -2 ø
4. Use the LPP method to solve æ 1 -1 -1ö
æ 3 -4 2 ö (d) ç -1 -1 3÷
(a) ç 1 -3 -7 ÷ ç ÷
ç ÷ è -1 2 -1ø
è -2 4 7 ø

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. (a). Saddle point (2, 3). æ A1 A2 A3 ö

Value of the game = 4 (n) SA = çè 1 5 0 4 5÷ø and
(b) Saddle points (1, 1), and (1, 3).
Value of the game = 1 æ B1 B2 B3 ö
SB = çè 0 13 15 2 15÷ø
(c) Saddle point (2, 2).
Value of the game = 5 Value of the game: 24
(d) Saddle points (1, 1), and (1, 3). Value of æ A1 A2 A3 A4 ö
the game = 2 (o) SA = çè 0 1 2 1 2 0 ÷ø and
(e) Saddle point (3, 3).
Value = 6 æ B1 B2 B3 B4 ö
(f) Saddle point (2, 3). SB = çè 1 10 9 10 0 0 ÷ø
Value = 7 Value: 3/2
(g) Saddle point (3, 2). æ A1 A2 A3 ö
(p) SA = ç
è 0 12 18 6 18÷ø
Value = 5
(h) Saddle point (2, 3).
Value = 6 æ B1 B2 B3 ö
SB = ç 0 13 18 5 18÷
(i) Saddle point (2, 3). è ø
Value = 5 Value: 10/3
(j) Saddle point (2, 3). (q) Saddle point (3, 2). Value = 12
Value of the game = 2 (r) Saddle point (3, 1). Value = 6
(k) Saddle point (3, 3). (s) Saddle point (1, 3). Value = 6
Value of the game = 2 æ A1 A2 ö
(l) Saddle point (1, 1) and (1, 3). 2. (a) SA = çè 3 7 4 7 ÷ø and
Value = – 2
(m) Saddle point (2, 3). æ B1 B2 B3 B4 ö
SB = çè 1 17 0 0 6 7 ÷ø v = 5/7
Value = 4
176 Operations Research

æ A1 A2 ö æ A1 A2 A3 A4 ö
(b) SA = çè 2 3 1 3÷ø and (j) SA = ç
2 5 0 ÷ø
and
è3 5 0
æ B1 B2 B3 B4 ö æ B1 B2 ö
SB = çè 1 3 2 3 0 0 ÷ø v = 2/3 SB = ç 4 5 1 5÷ Value = 17/5
è ø
æ A1 A2 ö æ A1 A2 A3 A4 ö
(c) SA = ç
è 4 9 5 9 ÷ø
and (k) SA = ç
0 ÷ø
and
è3 4 1 4 0
æ B1 B2 B3 B4 ö æ B1 B2 B3 B4 ö
SB = çè 4 9 0 5 9 ÷ø v = –7/9 SB = ç 0 1 4 0 3 4 ÷ø
0 è
æ A1 A2 A3 A4 A5 ö Value = 21/4
(d) SA = ç
0 ÷ø
and æ
è0 47 37 0 A A A 3 ö
3. (a) SA = ç 1 2
and
è 6 / 17 7 17 4 17 ÷ø
æ A1 A2 ö
SB = ç 5 7 2 7 ÷ v = 34/7 æ B B2 B3 ö
è ø SB = ç 1
è 2 17 9 17 6 17 ÷ø
æ A1 A2 A3 A4 A5 ö
(e) SA = ç 5 12 7 12 ÷ø and
Value = 5/17
è 0 0 0
æ A1 A2 A3 ö
æ B1 B2 ö (b) SA = ç
è 3 8 13 24 1 12 ÷ø
and
SB = çè 3 4 1 4 ÷ø v = – 1/4
æ B1 B2 B3 ö
æ A1 A2 A3 A4 ö SB = ç
è 7 24 5 9 11 72 ÷ø
v = 91/24
(f) SA = ç 1 3 0 2 3 0 ÷ø and
è
æ B1 B2 ö æ A1 A2 A3 ö
(c) SA = ç and
SB = ç
è 2 3 1 3 ÷ø
v = 8/3 è 0 2 3 1 3÷ø
æ A1 A2 ö æ B1 B2 B3 ö
SB = ç
è 0 2 3 1 3÷ø
(g) SA = ç 3 11 8 11÷ and v = 8/3
è ø
æ B1 B2 B3 ö æ A1 A2 A3 ö
(d) SA = ç and
SB = ç 0
è 2 11 9 11÷ø Value = 49/11 è 2 3 1 3 0 ÷ø
æ A1 A2 ö æ B1 B2 B3 ö
(h) SA = ç SB = ç 0 1 2 1 2 ÷ Value = 1
è 4 5 1 5÷ø
and è ø
æ B1 B2 B3 B4 B5 B6 ö æ A1 A2 A3 ö
SB = ç 0 3 5 0 (e) SA = ç 4 7 3 7 0 ÷ø and
è 2 5 0 0 ÷ø è
Value = 17/5 æ B1 B2 B3 ö
æ A1 A2 ö SB = ç 5 7 2 7 0 ÷ v = 1/7
è ø
(i) SA = ç and
è 1 3 2 3÷ø æ A1 A2 A3 ö
æ B1 B2 B3 B4 ö 4. (a) SA = ç 6 13 0 7 13÷ø and
è
SB = ç 1 5 4 5 0 0 ÷ø
è æ B1 B2 B3 ö
Value = –18/5 SB = çè 8 13 5 13 0 ÷ø v = 4/13
 Theory of Games 177

æ A1 A2 A3 ö æ B1 B2 B3 ö
(b) SA = ç 12 25 9 25 4 25÷ and SB = ç
è 5 22 8 22 9 22 ÷ø
v = 17/11
è ø
æ B1 B2 B3 ö æ A1 A2 A3 ö
SB = ç 19 50 15 50 8 25÷ v = 18/25 (d) SA = ç
è 6 13 3 13 4 13÷ø
and
è ø
æ A1 A2 A3 ö æ B1 B2 B3 ö
(c) SA = çè 6 11 3 11 2 11÷ø and SB = çè 6 13 4 13 3 13÷ø v = – 1/13

Dynamic Programming
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
15.1 INTRODUCTION
Dynamic programming is a mathematical
technique of optimization using multistage
decision process. It is a systematic procedure for
determining the combination of decisions which
maximize the overall objective. Decisions
regarding a certain problem are optimized in
stages rather than simultaneously. The original
problem is split into sub-problems such
that the outcome of each depends on the result of
the previous one. The solution is obtained in an
orderly manner by going from one stage to the
next and the complete solution is obtained at the
final stage.
Decision-making process consists of selecting
a combination of plans from a large number of
alternatives. All the combinations must be
specifically known beforehand. Then only we can
select the optimal policy. If we deal with the
problem as a whole it involves lot of computation
work and time. Certain combinations may not be
feasible. In such a situation dynamic programming
becomes very useful. We break the problem into
sub-problems (stages). Only one stage is
15
considered at a time and the infeasible
combinations are eliminated. The solution is
obtained by moving from one stage to the next
and is completed at the final stage. At each stage
there are many alternatives and the selection of
one feasible alternative is called stage decision.
The variables which specify the condition of the
process and summarize the current status (state)
are called state variables. At any stage there could
be a finite or infinite number of states.
Bellman’s principle of optimality states that “an
optimal policy has the property that whatever the
initial stages and decisions are, the remaining
decisions must constitute an optimal policy with
regards to the state resulting from the first
decision”. This implies that a wrong decision taken
at one stage does not prevent from taking optimum
decisions for the remaining stages.
15.2 DYNAMIC PROGRAMMING
METHOD
Example 15.1 (Salesmen allocation problem)
Consider the problem of allocating 9 salesmen to
three zones in such a manner that the total profits
 Dynamic Programming 179

of the company are maximized. The following Stage I (Zone I)

table gives the details about the zones, number of
x 0 1 2 3 4 5 6 7 8 9
salesmen and the corresponding profits.
f1(x) 30 45 60 70 79 90 98 105 100 90
Profits (in thousands of rupees) (Profit)
No. of salesmen Zone 1 Zone 2 Zone 3 This table gives the profit corresponding to
0 30 35 42 different number of salesmen allocated to zone 1.
1 45 45 54 F1(A) = f1(x).
2 60 52 60 Stage II (Zones 1 and 2)
3 70 64 70 If S denotes the total profits from each combination
4 79 72 82 we have the general form
5 90 82 95 S = f2(x) + F1(A – x)
6 98 93 102
where A is the total number of salesmen.
7 105 98 110
We have to maximize
8 100 109 110
F2(A) = Max {f2(x) + F1(A – x)}.
9 90 100 110
This equation can be applied to determine the
In this example the stages are three zones and optimal distribution of any number of salesmen
the number of salesmen is the state variable. (A = 9, 8, 7, 6, 5, ...) as shown in the following table.

Zone 1
0 1 2 3 4 5 6 7 8 9 No. of Salesmen
Zone 2
30 45 60 70 79 90 98 105 100 90 Profits
0 35 65 80 95 105 114 125 133 140 135 125
1 45 75 90 105 115 124 135 143 150 145
2 52 82 97 112 122 131 142 150 157
3 64 94 109 124 134 143 154 162
4 72 102 117 132 142 151 162
5 82 112 127 142 152 161
6 93 123 138 153 163
7 98 128 143 158
8 109 130 145
9 100 130
The maximum profit for each value of A are the profits for different combinations can be
represented in bold. found along the diagonal. The maximum
The above table shows the profits for all com- profits for various combinations are given
binations. For a given number of salesmen, below:

A 0 1 2 3 4 5 6 7 8 9
S 65 80 95 105 115 125 135 143 154 163
0+0 0+1 0+2 1+2 or 0+3 1+3 0+5 1+5 3+4 or 1+6 3+5 6+3

We have distribution of 9 salesmen in 3 zones.

F2 (A) = Max {f2(x) + F1 (A – x)} S = f3 (x) + F2 (A – x)
Now we go to the next stage and consider the F3 (A) = Max {f3 (x) + F2 (A – x)}
 180 Operations Research

Stage III

Zone 1 + Zone 2
0 1 2 3 4 5 6 7 8 9
Zone 3
65 80 95 105 115 125 135 143 154 163
0 42 107 122 137 147 157 167 177 185 196 205
1 54 119 134 149 159 169 179 189 197 208
2 60 125 140 155 165 175 185 195 203
3 70 135 150 165 175 185 195 205
4 82 147 162 177 187 197 207
5 95 160 175 190 200 210
6 102 167 182 197 207
7 110 175 190 205
8 110 175 190
9 110 175
A = (x1 + x2) + x3
The maximum profits for different values of A are
A 0 1 2 3 4 5 6 7 8 9
S 107 122 137 149 159 169 179 190 200 210
0+0 0+1 0+2 1+2 1+3 1+4 1+5 5+2 5+3 5+4

Thus for A = 9. We find that the profit is Find the mode of allocation of the six boxes to
maximum for x1 + x2 = 4 and x3 = 5. From stage II the stores so as to maximize the profit.
we find that for A = 4, the profit has max value Solution Let us take stage I as the allocation of
115 with x1 = 3 x2 = 1. Therefore the optimal the boxes to 1 store
distribution of salesmen is
Stage I (Store 1)
Zone 1 ¾¾¾ ®3 ü
ï
Zone 2 ¾¾¾ ® 1ý Max profit = Rs 2,10,000 X 0 1 2 3 4 5 6
Zone 3 ¾¾¾ ®5 ï f1(x) profit 0 4 6 7 7 7 7
þ
Example 15.2 (Production allocation problem) Stage II (Stores 1 and 2) Total profit P = F2 (A)
The owner of four fruit shops has purchased six = Max {f2(x) + F1 (A – x)} where A denotes the
boxes of apple. The quantity in demand and the total number of boxes allocated. Distribution of
profits are different at these stores. The following different number of boxes at stores 1 and 2 is as
table gives the total profit at each store for various follows:
numbers of boxes allotted. Store 1 0 1 2 3 4 5 6 No. of
Number of Stores boxes
boxes 1 2 3 4 Store 2 0 4 6 7 7 7 7 Profit
0 0 0 0 0 0 0 0 4 6 7 7 7 7
1 4 2 6 2 1 2 2 6 8 9 9 9
2 6 4 8 3 2 4 4 8 10 11 11
3 7 6 8 4 3 6 6 10 12 13
4 7 8 8 4 4 8 8 12 14
5 7 9 8 4 5 9 9 13
6 7 10 8 4 6 10 10
 Dynamic Programming 181

The maximum profits for different no of The maximum profits for different values of A
allocations to stores 1 and 2 are are

A 0 1 2 3 4 5 6 A 0 1 2 3 4 5 6
P 0 4 6 8 10 12 14 P 0 6 10 12 14 16 18
x1 + x2 0+0 0 +1 0 + 2 1 + 2 2 + 2 3 + 2 4 + 2 0+0 0+1 0+2 0+3 0+4 0+5 0+6
1+1 2+1 3 +1 4+1 1+2 1+3 1+4 1+5

We have F2(A) = Max [f2(x) + F1(A – x)] For four stores taken together, we find that the
max profit is 18, for 6 baskets.
Stage III (Stores 1, 2 and 3)
The optimal distributions are
P = F3 (A) = Max {f3(x) + F2 (A – x)}

Store 3 Store 1 + Store 2 Store 1 Store 2 Store 3 Store 4

0 1 2 3 4 5 6 2 3 1 0
0 4 6 8 10 12 14 1 4 1 0
0 0 0 4 6 8 10 12 14 2 2 2 0
1 6 6 10 12 14 16 18 1 3 2 0
2 8 8 12 14 16 18 2 2 1 1
3 8 8 12 14 16 1 3 1 1
4 8 8 12 14 2 1 2 1
5 8 8 12 1 2 2 1
6 8 8
Example 15.3 (Capital budgeting problem) A
The maximum profits for different allocations company produces automobile parts, bicycle parts,
for stores 1, 2 and 3 are and sewing machine parts. An amount of
A 0 1 2 3 4 5 6 Rs 20,000 has been allotted for expanding
P 0 6 10 12 14 16 18
production facilities. In autoparts and bicycle parts
sections production can be increased either by
0+0 1+0 1+1 1+2 1+3 1+4 1+5
2+1 2+2 2+3 2+4
addition of new machines or by replacing old
inefficient machines by automatic machines. In
Stage IV (Stores 1, 2, 3 and 4) sewing machine parts section, additional amount
P = F4(A) = Max { f4(x) + F3(A – x)} can be invested only by adding new machine. The
cost of addition and replacement of machines and
Store 1 + Store 2 + Store 3 the corresponding returns are given below: Select
Store 4 0 1 2 3 4 5 6
a set of expansion plans which may yield
0 6 10 12 14 16 18 maximum return.
0 0 0 6 10 12 14 16 18 Alternatives Auto parts Bicycle parts Sewing machine
1 2 2 8 12 14 16 18 Cost Return Cost Return Cost Return
2 3 3 9 13 15 17
No expansion 0 0 0 0 0 0
3 4 4 10 14 16
Add new 4000 8000 8000 12000 2000 8000
4 4 4 10 14
achines
5 4 4 10
6 4 4 Replacement 6000 10000 12000 18000 – –
 182 Operations Research

Solution State Alternatives Optimal Solution

Stage I Each department can be considered as a 1 2
stage. The capital amount is the state variable. Let x3 C31 = 0 C32 = 2 Return Decision
us take auto parts section as stage I. R31 = 0 R32 = 8 F3(A)

State Alternatives Optimal Solution 0 0+0=0 – 0 1

1 2 3 2 0+0=0 8+0=8 8 2
x1 C11 = 0 C12 = 4 C13 = 6 Return Decision 4 0+8=8 8+0=8 8 1, 2
R11 = 0 R12 = 8 R13 = 10 F1(A) 6 0 + 10 = 10 8 + 8 = 16 16 2
8 0 + 12 = 12 8 + 10 = 18 18 2
0 0 – – 0 1 10 0 + 12 = 12 8 + 12 = 20 20 2
2 0 – – 0 1 12 0 + 20 = 20 8 + 12 = 20 20 1, 2
4 0 8 – 8 2 14 0 + 22 = 22 8 + 20 = 28 28 2
6 0 8 10 10 3 16 0 + 26 = 26 8 + 22 = 30 30 2
8 0 8 10 10 3 18 0 + 28 = 28 8 + 26 = 34 34 2
20 0 + 28 = 28 8 + 28 = 36 36 2
10 0 8 10 10 3
12 0 8 10 10 3 For stage III, optimal decision is the 2 nd
14 0 8 10 10 3 alternative with maximum return of Rs 36,000
16 0 8 10 10 3 (C32 = 2). For the remaining amount 18 we find
18 0 8 10 10 3 that in stage II optimal decision is the 3 rd
20 0 8 10 10 3 alternative with maximum return of Rs 28,000
(C23 = 12). For the remaining amount 6 we find
Stage II We consider auto parts and bicycle parts that in stage I optimal decision is the 3 rd
F2(A) = Max {f2(x) + F1(A – x)} alternative with maximum return of Rs 10,000
(C13 = 6).
State Alternatives Optimal Solution Thus for the allocation of Rs 20,000, the optimal
1 2 3 policy of expansion is
x2 C21 = 0 C22 = 8 C23 = 12 Return Deci-
1. Replacing old machines in auto parts
R21 = 0 R22 = 12 R23 = 18 F2(A) sion
2. Replacing old machines in bicycle parts and
0 0+0 = 0 – – 0 1
3. Adding new machines in sewing machines
2 0+0 = 0 – – 0 1 parts
4 0+ 8 = 8 – – 8 1
The optimum returns is Rs 36,000.
6 0+ 10 = 10 – – 10 1
8 0+ 10 = 10 12 + 0 =12 – 12 2 Example 15.4 (Advertising media problem) A
10 0+ 10 = 10 12+ 0 = 12 – 12 2 company is interested in selecting the advertising
12 0+10 = 10 12 + 8 = 20 18 + 0 = 18 20 2
media for its product. The frequency of advertising
14 0+ 10 = 10 12 + 10 = 22 18 + 0 = 18 22 2
per week and the corresponding expected sales in
thousands of rupees are given below:
16 0+ 10 = 10 12 + 10 = 22 18+ 8 = 26 26 3
18 0 + 10 = 10 12 + 10 = 22 18 + 10 = 28 28 3 Sales (in thousands of rupees)
20 0 + 10 = 10 12 + 10 = 22 18 + 10 = 28 28 3 Frequency TV Radio Newspaper
per week
Stage III We consider all the three sections to- 1 220 150 100
gether for stage III. The computations are given 2 275 250 175
below. 3 325 300 225
F3(A) = Max {f3(x) + F2(A – x)} 4 350 320 250
 Dynamic Programming 183

The costs of advertising in newspaper, radio and Rs 4000 per week for advertisement. Determine the
television are Rs 500, Rs 1000 and Rs 2000 optimal combination of advertising media and
respectively per appearance. The budget provides frequency.

Solution Let us consider TV as stage I. The costs are C11 C12 C13 C14 for the frequencies and the
corresponding returns are R11 R12 R13 R14.

State Frequency
1 2 3 4
x1 C11 = 2000 C12 = 4000 C13 = 6000 C14 = 8000 Return Decision
R11 = 220 R12 = 275 R13 = 325 R14 = 350

500 – – – – 0 0
1000 – – – – 0 0
1500 – – – – 0 0
2000 220 – – – 220 1
2500 220 – – – 220 1
3000 220 – – – 220 1
3500 220 – – – 220 1
4000 220 275 – – 275 2

F1(A) = Max f1(x)

Now let us take TV and Radio
Stage II
F2(A) = Max { f2(x) + F1(A – x)}

State Frequency Optimal

0 1 2 3 4
x2 C21 = 1000 C22 = 2000 C23 = 3000 C24 = 4000 Return Decision
R21 = 150 R22 = 250 R23 = 300 R24 = 320
500 0 – – – – 0 0
1000 0 150 – – – 150 1
1500 0 150 – – – 150 1
2000 220 150 + 0 250 – – 250 2
2500 220 150 + 0 250 + 0 – – 250 2
3000 220 150 + 220 250 + 0 300 – 370 1
3500 220 150 + 220 250 + 0 300 + 0 – 370 1
4000 275 150 + 220 250 + 220 300 + 0 320 470 2
 184 Operations Research

Stage III
TV, Radio and Newspaper
F3(A) = Max {f3(x) + F2(A – x)}

State Frequency Optimal

0 1 2 3 4
x3 C31 = 500 C32 = 1000 C33 = 1500 C34 = 2000 Return Decision
R31 = 100 R32 = 175 R33 = 225 R34 = 250 F3(A)
500 0 100 – – – 100 1
1000 150 100 + 0 175 – – 175 2
1500 150 100 + 150 175 + 0 225 – 250 1
2000 250 100 + 150 175 + 150 225 + 0 250 325 2
2500 250 100 + 250 175 + 150 225 + 150 250 + 0 375 3
3000 370 100 + 250 175 + 250 225 + 150 250 + 150 425 2
3500 370 100 + 370 175 + 250 225 + 250 250 + 150 475 3
4000 470 100 + 370 175 + 370 225 + 250 250 + 250 545 2

Thus the maximum return is 5,45,000. The optimal policy is: Twice in Newspaper, once in Radio and
once in TV per week.
Example 15.5 (Cargo loading problem) A 5 60 120 – 120 2
truck can carry a total of 10 tons of a product. 6 60 120 180 180 3
Three types of product are available for transport. 7 60 120 180 180 3
Their weight and values are given below. One unit in each of A and B must be taken.
Determine the loading such that the total value Hence x1 cannot exceed 7 tons. We take C and B
of the cargo is maximum (at least one unit of for stage II
each type must be loaded). F2(A) = Max {f2(x) + F(A – x)}

Type Value (Rs) Weight (tons) Stage II

A 20 1 State No. of units Optimal
B 50 2 (Weight) 1 2 3
C 60 2 C21 = 50 C22 = 100 C23 = 150 Value Deci-
W=2 W=4 W=6 F2(A) sion
Solution We have to determine how many units
of A, B and C is to be loaded. We take the loading 4 50 + 60 100 – 110 1
of type C as stage I. 5 50 + 60 100 – 110 1
6 50 + 120 100 + 60 150 170 1
State x1 No. of units Optimal 7 50 + 120 100 + 60 150 170 1
(Weight) 1 2 3 8 50 + 180 100 + 120 150 + 60 230 1
C11 = 60 C12 = 120 C13 = 180 Value Deci-
W=2 W=4 W=6 F1(A) sion 9 50 + 180 100 + 120 150 + 60 230 1

2 60 – – 60 1
Stage III (C, B and A are taken)
3 60 – – 60 1
F3 (A) = Max {f3(x) + F2 (A – x)}
4 60 120 – 120 2
 Dynamic Programming 185

State No. of units Optimal

1 2 3 4 5 6
x3 C31 = 20 C32 = 40 C33 = 60 C34 = 80 C35 = 100 C36 = 120 F3(A) Decision
W=1 W=2 W=3 W=4 W=5 W=6
5 20 + 110 – – – – – 130 1
6 20 + 110 40 + 110 – – – – 150 2
7 20 + 170 40 + 110 60 + 110 – – – 190 1
8 20 + 170 40 + 170 60 + 110 80 + 110 – – 210 2
9 20 + 230 40 + 170 60 + 170 80 + 110 100 + 110 – 250 1
10 20 + 230 40 + 230 60 + 170 80 + 170 100 + 110 120 + 110 270 2

We find that the maximum value is Rs 270. = Max {x.[( c – x)/2]2}

The loading is A – 2 units, B – 1 unit, C – 3 1
units. Take u = x[(c – x)/2]2 = x(c – x)2
4
Weight 2 + 2 + 6 = 10 tons
du 1
= [x 2 (c – x) ( – 1) + (c – x)2]
15.3 APPLICATIONS TO dx 4
1
MATHEMATICAL PROBLEMS = [(c – x)2 – 2x (c – x)]
4
Example 15.6 (Optimal sub-division problem) 1
Divide a positive number c into n parts such that = [(c – x) (c – 3x)]
4
their product is maximum.
du
Solution The problem can be taken as = 0 fi x = c/3 or x = c
Maximize Z = y1 y2 y3 …yn where dx
For max u, x = c/3. \ Therefore c – x = 2c/3
y1 + y2 + y3 +… + yn = c (yi ≥ 0 i = 1, 2,…, n)
1
Each i can be considered as a stage. Here yi is Max u = c/3. 4 (c/3)2 = (c/3)3
continuous. Hence the optimal decisions at each 4
stage are obtained by the method of maxima and \ f3(c) = (c/3)3 and y1 = y2 = y3 = c/3.
minima of differential calculus. Assume that the optimal policy for n = m is
(c/m, c/m, …, c/m)
For n = 1 (one stage) y1 = c only one part
fm(c) = (c/m)m
\ f1(c) = c
For n = m + 1, fm + 1 (c) = Max{x fm(c – x)}
For n = 2 (two stage) take y1 = x, y2 = c – x
= Max [x. {(c – x)/m}m]
f2(c) = Max y1 y2 = Max {x.f1(c – x)}
u = x (c – x)m / mm
= Max{x (c – x)}
Now u = x (c – x) = cx – x2 du
= (1/mm) {x.m (c – x)m–1.
du dx
dx
= c – 2x = 0 fi x = c/2. Therefore, c – x = (– 1) + (c – x)m}
m m–1
c/2 when u is maximum = (1/m ) (c – x) [c – x – mx]
\ Max x(c – x) = c/2. c/2 = (c/2)2 = (1/mm) (c – x)m–1 [c – (m + 1)x]
\ f2(c) = (c/2)2. y1 = y2 = c/2 u is maximum when x = c/(m + 1)
For n = 3 (c – x)m/mm = (mc)m/(m + 1)mmm = cm/(m + 1)m
f3(c) = Max y1 y2 y3 \ fm + 1 (c) = (c/m + 1). cm/(m + 1)m
= Max {x.f2(c – x)} = cm+1/(m + 1)m + 1
 186 Operations Research

\ The subdivision is c/(m + 1), c/(m + 1), …. \ f3(b) = b1/3 + 2 b1/2/b1/6

c/(m + 1) and fm + 1 (c) = [c/(m + 1)]m + 1 = b1/3 + 2b1/3 = 3b1/3
This shows that the result is true for n = m + 1. The optimal policy is {b1/3 b1/3 b1/3} and f3 (b)
Also we have seen that it is true for n = 1, 2, 3. = 3b1/3
Hence by induction, it is true for any value of n. Now let us assume that the result is true for
\ The optimal subdivision is n=m
(c/n, c/n, …, c/n) and the product is (c/n)n. \ b = y1 y2 y3 … ym
Example 15.7 Factorize a positive number b into Optimal policy is {b1/m b1/m … b1/m} and fm(b)
n factors such that their sum is minimum. = mb1/m
We have to find y1, y2, y3, …, yn such that y1, y2, fm + 1(b) = min {x + fm(b/x)}
y3,…., yn = b and their sum = min {x + m (b/x)1/m}
Z = y1 + y2 + y3 + ... + yn is minimum. Take u = x + m . b1/m/x1/m
Solution One stage problem du
(n = 1) = 1 + m b1/m {(–1/m) x –1/m – 1}
y1 = b dx
= 1 – b1/m / x1 + 1/m = 1 – b1/m / x(m + 1) / m
\ f1(b) = Min { y1} = b du
y1 = b
= 0 fi x (m + 1)/m = b1/m or x = b1/ m + 1
Two stage problem dx
(n = 2) when u is minimum
b = y 1y 2 . \ fm + 1 (b) = b1/m + 1 + m . b1/ m/b1/m(m + 1)
y1 = x, y2 = b/x = b 1/m + 1 + m . b1/m–1/m(m + 1)
f2(b) = min (y1 + y2) = b1/m + 1 + m b1/m + 1
= min (x + f1(b/x)} = (m + 1) b1/m + 1
= min (x + b/x) \ The optimal policy is
u = x + b/x { b1/m + 1 , b1/m + 1,….,b1/m + 1}.
du Thus we find that the result is true for n = m + 1
= 1 – b/x2 = 0 also. We have seen that the result is true for
dx
fi x2 = b or x = b when u is minimum. n = 1, 2, 3 … Therefore by principle of induction,
\ f2(b) = min {x + b/x} we find that it is true for all values of n.
1 \ The optimal factorization is
= b + b = 2 b = 2b 2 b1/n b1/n …b1/n and z = n b1/n
The optimal policy is {b1/2 , b1/2} and f2(b)
Example 15.8 If p1 + p2 + p3 + … + pn = 1, show
= 2b1/2
that the sum
(n = 3)
z = p1log p1 + p2 log p2 + … + pn log pn (pi ≥ 0)
b = y1 y2 y3
is minimum when p1 = p2 = p3 = p3 … pn = 1/n.
f3(b) = min {x + f2 (b/x)}
Solution For n = 1, p1 = 1 and z = p1 log p1 f1 (1)
= min {x + 2.(b/x)1/2}
= 1 log 1 = 0 trivial case
u = x + 2.b1/2/x1/2
For n = 2, take p1 = x and p2 =1 – x
du z = f2(1) = min {p1log p1 + p2 log p2}
= 1+ 2b1/2(–1/2)x–3/2
dx = min {x log x + (1 – x) log (1 – x) }
when u is minimum Take u = x log x + (1 – x) log (1 – x)
du du
= 1 – b1/2 / x3/2 = 0 = x.1/x + log x + (1 – x) ( – 1) / (1 – x)
dx dx
fi x3/2 = b1/2 or x = b1/3 + ( – 1) log (1 – x)
= 1 + log x – 1 – log (1 – x)

du
dx
= 0 fi log x – log (1 – x) = 0
fi log x = log (1 – x)
fi x = 1 – x or x = ½ when u is minimum
1 1 1 1
\ f2(1) = log + log = 2(1/2 log )
2 2 2 2
1 1
\ p1 = p2 = and min z = 2(1/2 log )
2 2
Let us assume that the result is true for n = m.
\ p1 + p2 + p3 + … + pm = 1 and z = p1 log p1
+ p2 log p2 + … + pm log pm is minimum when p1
= p2 =… = pm = 1/m or fm(1) = m(1/m log 1/m).
Now fm + 1(1) = min {x log x + fm(1 – x)}
= min {x log x + m (1 – x)/m
log (1 – x)/m}
Take u = x log x + m . (1 – x)/m log (1 – x)/m
du
= x . 1/x + log x + (1 – x) ( – 1) / (1 – x)
dx
+ log (1 – x)/m (– 1)
= log x – log (1 – x)/m
du
= 0 fi log x = log (1 – x)/m
dx
fi x = 1/(m + 1) when u is minimum.
\fm + 1 (1) = 1/(m + 1).log 1/(m + 1) +
m[{1 – 1/(m + 1)} / m]log
{1 – 1/(m + 1)} / m
= 1/(m + 1) log 1/(m + 1) +
[m/(m + 1)] log 1/(m + 1)
= (m + 1) [ 1/(m + 1) log {1/(m +
1)}]
Therefore, the result is true for n = m + 1. But
we have seen that it is true for n = 1, 2. Hence by
induction principle the result is true for all n. Thus
the optimal policy (1/n 1/n 1/n … 1/n) and fn(1) =
n [1/n log (1/n)]
Example 15.9 Solve the following problem
using dynamic programming.
Minimize z = x 12 + x22 + x32 subject to the
constraints
x1 + x2 + x3 = 15, x1, x2, x3 ≥ 0
Solution
Stage I f1 (s1) = f1 (x1) = x12 = 152
Dynamic Programming 187
Stage II x1 + x2 = 15. Therefore, s2 = 15
f2 (s2) = min {x22 + (15 – x2)2}
Take u = x22 + (15 – x2)2
du
= 2x2 + 2(15 – x2)( – 1) = 4x2 – 30
1 dx
2 du
If u is minimum = 0 fi x2 = 15/2
dx
\ x1 = 15/2
f2(15) = (15/2)2 + (15/2)2 = 2 (15/2)2
Stage III x1 + x2 + x3 = 15
f3(s3) = Min {x32 + f2(15 – x3)}
= Min {x32 + 2 ((15 – x3)/2)2}
Take u = x32 + 2. ((15 – x3)/2)2
du
= 2x3 + 2. 2(15 – x3)/4. (– 1)
dx
= 2x3 – (15 – x3) = 3x3 – 15
du
If u is minimum = 0. \ x3 = 5
dx
\ f3 (s3) = f3 (15)
= 52 + 2 . (10/2)2 = 3 (52) = 75
x1 = x2 = x3 = 5
Min z = 75 when x1 = x2 = x3 = 5
15.4 SOLUTION OF LINEAR
PROGRAMMING PROBLEMS
Given the LPP Maximize z = c1x1 + c2x2 + c3x3 +
… + cn subject to the constraint
n
å aij x j £ b j i = 1, 2, 3, … m
j =1
xj ≥ 0 j = 1, 2, 3, … n
we consider it as n stage problem. Each j denotes
a stage. b1, b2 …bm are the resources constituting
the states. Let (B1j, B2j, … , Bmj) be the state of the
system at jth stage and fj (B1j, B2j, … , Bmj) be the
optimum value of the objective function at jth stage
for the stages B1j, B2j, … , Bmj.
We use computational procedure, starting from
stage 1. The general equation is
fj (B1j, B2j, …) = Max {cjxj + fj – 1
(B1j – 1, B2j – 1, … , Bmj – 1)}
 188 Operations Research

Example 15.10 Solve using dynamic pro- Example 15.11 Maximize z = 5x1 + 2x2
gramming Subject to x1 + 2x2 £ 43
Maximize z = 3x1 + 4x2 x1 £ 23
subject to the constraints x1, x2 ≥ 0
2x1 + 5x2 ≥ 120 Solution
2x1 + x2 £ 40 Stage I
x1 , x2 ≥ 0 f1 (B11 B21) = Max (5x1)
Solution Here we have two stages for two = 5 max (x1)
variables. = 5 min { 43 – 2x2, 23}
Stage I Stage II
f1 (B11 B21) = Max (c1 x1) f2 (B12, B22) = max [5 min {43 – 2x2, 23}
= Max (3x1) + 2x2]
= 3 Max(x1) ì5(43 - 2 x2 ) + 2 x2 if 43 - 2 x2 £ 23
From the constraints we find that = Max í
x1 £ (120 – 5x2 )/ 2 and î5(23) + 2 x2 if 43 - 2 x2 ³ 23
and x1 £ (40 – x2 )/ 2 ì 215 - 8 x2 if x2 ³ 10
= Max í
In order to satisfy both the constraints î115 + 2 x2 if x2 £ 10
x1 £ min {(120 – 5x2 )/ 2, (40 – x2 )/ 2} Both the values become equal when x2 = 10
\ f1 (B11 B21) = 3 min {(120 – 5x2 )/2, \ f2 (B12 B22) = (215 – 80)
(40 – x2 )/ 2} or (115 + 20) = 135
\ f2 (43 , 23) = 135 x2 * = 10
Stage II
x1 * = 43 – 2x2 = 23
f2 (B12 B22 ) = Max { f1(B11 B21 ) + 4x2}
\ Solution is x1 = 23, x2 = 10 z * = 135
= Max [3 min {(120 – 5x2)/2,
(40 – x2) / 2} + 4x2] Example 15.12 Maximize z = 3x1 + 2x2
subject to the constraints
ì3(120 - 5 x2 ) / 2 + 4 x2 if x1 + x2 £ 300
ï (120 - 5 x ) / 2 £ (40 - x ) / 2
ï 2 2 2x1 + 3x2 £ 800
= max í
ï 3(40 - x2 ) / 2 + 4 x2 if x1, x2 ≥ 0
ïî (120 - 5 x2 ) / 2 ³ (40 - x2 ) / 2 Solution We apply the backward procedure,
taking stage I with x2.
Now (120 – 5x2)/2 £ (40 – x2)/2 f2 (B12, B22) = max (2x2)
fi 120 – 5x2 £ 40 – x2 = 2 max (x2)
fi 80 £ 4x2 fi x2 ≥ 20 = 2 [min { 300 – x1, (800 – 2x1)/3}]
(120 – 5x2)/2 ≥ (40 – x2)/2
Stage II
fi x2 £ 20
f1 (B11 B21) = max [3x1 + 2x2]
When x2 = 20
= max [3x1 + 2 min{300 – x1, (800 – 2x1) / 3}]
3 (120 – 5x2) / 2 + 4x2
= 3 (40 – x2) / 2 + 4x2 = 110 ì3x1 + 2(300 - x1 ) if
ï 300 - x1 £ (800 - 2 x1 ) / 3
\ f2 (120, 40) = 110 x2* = 20 ï
= max í
x1* = (40 – x2 * )/ 2 = 10
ï3x1 + 2(800 - 2 x1 ) / 3 if
Solution is x1 = 10, x2 = 20, z* = 110 ïî 300 – x1 ³ (800 - 2 x1 ) / 3
 Dynamic Programming 189

ì x1 + 600 if 100 £ x1 Solution (Backward Procedure)

= max í Stage I
î (5 x1 + 1600) / 3 if 100 ³ x1 f2 (B12 B22) = max (9x2)
Both the values are equal if x1 = 100 = 9 max(x2)
\ f1 (B11 B21) = 100 + 600 or (500 + 1600) / 3 = 9 min {25 – 2x1, 11}
for x1 = 100 Stage II
f1 (300, 800) = 700. x1* = 100 f1 (B11, B21) = max {x1 + 9x2}
x2* = 300 – x1 * = 200 = max [x1 + 9 min {25 – 2x1, 11}]
\ The solution is z* = 700 x1 = 100 x2 = 200 ì x1 + 9(25 - 2 x1 ) if 25 - 2 x1 £ 11
= max í
Example 15.13 Maximize Z = x1 + 9x2 î x1 + 9.11 if 25 - 2 x1 ³ 11
subject to the constraints ì 225 - 17 x1 if x1 ³ 7
2x1 + x2 £ 25 = max í
î x1 + 99 if x1 £ 7
x2 £ 11 \ f1 (25, 11) = 106, x1* = 7
x1, x2 ≥ 0 x2 * = 11
\ The solution is x1 = 7, x2 = 11, z* = 106

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. A corporation runs three plants. Each No. of workers Job

plant is requested to submit its proposals 1 2 3
giving total cost C and revenue R for each 0 0 0 0
proposal. Budget is Rs 5 lakh for allocation
1 25 20 33
to all the three plants. The following table
2 42 38 43
gives the details:
3 55 54 47
Proposal Plant 1 Plant 2 Plant 3 4 63 65 50
C1 R1 C2 R2 C3 R3 5 69 73 52
6 74 80 53
1 0 0 0 0 0 0
2 1 5 2 8 1 3 Determine the optimal allotment.
3 2 6 3 9 – – 3. A contractor has to complete four projects.
4 – – 4 12 – – He has six foremen and each project is to be
allotted to at least one foreman. The
Determine the allocation so as to maximize following table gives the time required and
the revenue. the number of foremen to complete the
2. Six workers are to be allotted to 3 types of projects. Determine the minimum time
jobs. The following table gives the expected required for completing all the projects with
returns from the various allotments. six foremen.
 190 Operations Research

No. of foremen Project 7. An investor has Rs 6000 to invest. This amount

A B C D can be invested in any of the three ventures A,
B and C. But he must invest in units of Rs 1000.
1 15 17 19 21
The following table gives the returns from the
2 13 15 18 18
3 12 13 17 18
investment in the ventures.
4. (Cargo loading problem). A truck can carry Investment Return
a total of 10 tons of a product. Three types
A B C
of product are to be loaded. Their weights
and values are given below: 0 0 0 0
1 0.5 1.5 1.2
Product Weight (tons) Value 2 1.0 2.0 2.4
T1 2 65 3 3.0 2.2 2.5
T2 3 80 4 3.1 2.3 2.6
T3 1 30 5 3.2 2.4 2.7
6 3.3 2.5 2.8
Determine how many units of T1, T2 and T3
are to be loaded in order to maximize the Find the optimal investment policy.
total value. [At least one unit of each type 8. Apply dynamic programming to solve.
must be selected.] (a) Minimize z = x12 + x22 + x32
5. The costs of advertisement in newspaper subject to the constraints
radio and TV are Rs 5000, Rs 10,000 and
x1 x2 x3 = 24
Rs 20,000 respectively per appearance. The
company has a budget of Rs 40,000 for x1, x2, x3 ≥ 0.
advertisement. The following table gives the (b) Maximize z = x1 x2 x3
frequency and the corresponding sales: subject to the constraints
Frequency Sales x1 + x2 + x3 = 12
per month Newspaper Radio TV x1, x2, x3 ≥ 0.
1 125 180 300 (c) Maximize z = 2x1 + 3x2 + 4x3
2 225 290 350
3 260 340 450 subject to the constraints
4 300 370 500 x1 + x2 + x3 = 18
Determine the optimal combination of the x1, x2, x3 ≥ 0.
advertising media and frequency. (d) Minimize z = x12 + x22 + x32
6. (Salesmen Problem). A company has five subject to the constraints
salesmen, who are to be allocated to three x1 + x2 + x3 ≥ 12
marketing zones. The expected returns for
different number of salesmen in different x1, x2, x3 ≥ 0.
zones are given below. 9. Solve the following LPP using dynamic
No. of salesmen Zones programming
1 2 3 (a) Maximize z = 3x1 + 5x2
1 45 30 35 subject to the constraints
2 58 45 45 x1 £ 4
3 82 70 64 x2 £ 6
4 93 79 72
5 101 90 82
3x1 + 2x2 £ 18
x1, x2 ≥ 0
Determine the optimal allocation.
 Dynamic Programming 191

(b) Maximize z = 5x1 + 10x2 ties of power failure for the systems are given
subject to the constraints below:
10x1 + 5x2 £ 250 Power cells Probability of power failure
4x1 + 10x2 £ 200 System 1 System 2 System 3
2x1 + 3x2 £ 900 1 0.50 0.60 0.40
x1, x2 ≥ 0 2 0.15 0.20 0.25
3 0.04 0.10 0.10
(c) Maximize z = 3x1 + 4x2 4 0.02 0.05 0.05
subject to the constraints 5 0.01 0.02 0.01
2x1 + x2 £ 6 Determine how many cells should be as-
2x1 + 3x2 £ 9 signed to each system so as to maximize the
x1, x2 ≥ 0 overall system reliability (least probability
(d) Maximize z = 4x1 + 3x2 of total power failure).
12. The following table gives the number of votes
subject to the constraints
(in thousands) gained by a political party in
2x1 + x2 £ 2 three districts corresponding to various
– x1 + x2 £ 1 allocation of six party workers in each district.
x1, x2 ≥ 0 Workers Districts
10. A student has five days at his disposal to D1 D2 D3
revise the subject before examination. The 0 0 0 0
course is divided into four sections. He may 1 25 20 33
study a section for one day, two days, three 2 42 38 43
days, etc. or not at all. The expected grade 3 55 54 47
points he will get for different alternative 4 63 65 50
arrangements are as follows: 5 69 73 52
6 74 80 53
Study Sections
I II III Determine the number of workers to be
assigned to each district so as to maximize
0 0 1 0
the number of votes gained.
1 1 1 1
2 1 3 3 13. A company has eight salesmen working in
3 3 4 3 three areas A1, A2, A3. The sales (in thousands
of rupees) by each salesman in the three areas
How should be distribute the available days are given below. Determine the optimum
to the different sections so that he may get allocation of salesmen in order to maximize
maximum grade points? the sales.
11. Consider the problem of designing electronic
devices to carry five power cells, each of Number of 0 1 2 3 4 5 6 7 8
salesmen
which must be located within three electronic
systems. If the power fails in one system then A1 15 22 30 38 45 48 54 60 65
A2 26 35 40 46 55 62 70 76 83
it will be powered on an auxiliary basis by
A3 30 38 44 50 60 65 72 80 85
the cells of the remaining systems. The
probability of power failure in a particular 14. World Health Council wants to determine the
system depends on the number of cells number of medical teams to be allocated
originally assigned to it. Estimated probabili- among three countries C1, C2 and C3. It has
 192 Operations Research

five medical teams. The measure of where x1 + x2 + x3 + … xn = c and

effectiveness used, is additional man years a1 £ a2 £ a3 £ … £ an
of life. The following table gives the 18. Minimize Z = x12 + x22 + x32
estimated additional man years of life (in
subject to the condition
thousands) for each country against each
possible allocation of medical teams. x1 + x2 + x3 ≥ 10
x1, x2, x3 ≥ 0
Number of Additional man-years of life 19. Find the maximum value of
Medical (in thousands)
teams Z = x12 + 2x22 + 4x3
C1 C2 C3 subject to the condition
0 0 0 0 x1 + 2x2 + x3 £ 8
1 45 20 50 x1, x2, x3 ≥ 0
2 70 45 70
20. Find the minimum value of Z = x12 + 2x22 + 4x3
3 90 75 80
4 105 110 100
subject to the condition
5 120 150 130 x1 + 2x2 + x3 ≥ 8
x1, x2, x3 ≥ 0
Determine how many teams are to be
21. Minimize Z = xyz
allocated to each country to get maximum
effectiveness. subject to the condition
x+y+z = 5
15. An investor wants to invest Rs 60,000 in any
of the three ventures A, B and C, in units of x1, x2, x3 ≥ 0
Rs 10,000. The following table shows the 22. Solve the LPP
return from the investment in the ventures Maximize Z = 8x + 7y subject to the
(in thousands of rupees) constraints
2x + y £ 8
Amount Return from venture 5x + 2y £ 15
invested A B C x, y ≥ 0
0 0 0 0 23. Solve the LPP
1 5 15 12
Maximize Z = 3x + y subject to the constraints
2 10 20 24
2x + y £ 6
3 30 22 25
4 31 23 26 x£2
5 32 24 27 y£4
6 33 25 28 x, y ≥ 0
24. Solve the LPP
Determine the optimal investment policy.
Maximize Z = 3x + 2y subject to the
16. Find the minimum value of Z = x12 + x22 + x32
constraints
+ … + xn2
subject to the condition – 2x + y £ 1
x1, x2, x3, … , xn = C (C π 0) x£2
17. Find the maximum value of x+y £ 3
Z = a1x1 + a2x2 + a3x3 + … + anxn x, y ≥ 0
 Dynamic Programming 193

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. Maximum revenue = 17 lakh (d) x1* = x2* = x3* = 4; Z* = 48

9. (a). x1* = 2, x2* = 6; Z* = 36
2, 4,1
Optimal allocation to plants ü (b) x1* = 18.75, x2* = 12.5; Z* = 218.75
ý 2, 3, 2 (c) x1* = 9/4, x2* = 3/2; Z* = 51/4
1, 2 and 3 þ
or 3, 2, 2 (d) x1* = 1/3, x2* = 4/3; Z* = 16/3
2. Maximum return = 129 10. Section I = 1 day, Section II = 0, Section III
= 2 days
Optimal allocation to jobs ü Required maximum grade points = 5
ý 2, 3, 1
1, 2 and 3 þ 11. System 1 = 3 cells, System 2 = 1 cell,
3. Table (left out in the question) System 3 = 1 cell
Minimum time required = 67 Smallest probability of total failure is 0.0096
No. of foremen: A B C D 12. D1 : 2, D2 : 3, D3 : 1 Total votes = 129
1 2 1 2 13. A1 : 3, A2 : 1, A3 : 4 Total sales = 133
or 2 1 1 2 14. C1 : 1, C2 : 3, C3 : 1 Total = 170
4. T1 – 3 units, T2 – 1 unit, T3 – 1 unit, value = 15. A : 3, B : 1, C : 2 Total return = 69
305 16. X1 = X2 = X3 = … = Xn = C1/n
5. News- Radio TV Total Min value : nc2/n
paper sales = 705 17. X1 = X2 = X3 = … = Xn – 1 = 0, Xn = C
Max value : anc
Frequency Æ 2 1 1
18. X1 = X2 = X3 = 10/3; Min Z = 100/3
6. Zone 1 = 3, Zone 2 = 2, Zone 3 = 0
Total returns = 177 19. X1 = 8, X2 = 0, X3 = 0; Max Z = 64
7. A – 3000, B – 1000, C – 2000 20. X1 = X2 = X3 = 2; Min Z = 20
Total returns = 6.9 21. X = 0, Y = 0, Z = 5; Min Z = 0
8. (a) x1* = x2* = x3* = 241/3; Z* = 3.242/3 22. X = 0, Y = 15/2; Max Z = 52.5
(b) x1* = x2* = x3* = 4; Z* = 64 23. X = 2, Y = 2; Max Z = 8
(c) x1* = 0, x2* = 0, x3* = 18; Z* = 72 24. X = 2, Y = 1; Max Z = 8
 16
Sequencing Problems

CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■

16.1 INTRODUCTION of completion of jobs is independent of the

sequence of jobs.
If there are n jobs to be performed, one at a time,
on each of m machines, the sequence of the 16.2 PROCESSING n JOBS
machines in which each job should be performed
THROUGH TWO MACHINES
is given and the time required by the jobs on each
machine is also given, then the problem is to find The characteristics of this problem are
the sequence (order) of the jobs which minimizes (i) there are only two machines, A and B
the total time taken from the starting of the first (ii) each job is processed in the order A, B
job on the first machine till the completion of the
last job on the last machine. This is referred to as (iii) the processing times A1 A2…An on A and B1
the sequencing problem. In the case when there B2…Bn on B are known.
are three jobs and three machines the total number
of possible sequences will be (3!)3 = 216. Hence Job Processing time
it requires lot of computational time. Thus the A B
sequencing technique becomes very useful in that 1 A1 B1
it reduces computational work and time. 2 A2 B2
Here, the processing times on different 3 A3 B3
machines are exactly known and are independent . . .
of the order of processing. The time taken by the . . .
jobs in moving from one machine to another is n An Bn
negligible. Once a job has begun on a machine, it
must be completed before another job can begin We have to determine the sequence of jobs
on the same machine. Only one job can be which minimizes the total time elapsed from the
processed on a given machine at a time. The order start of the first job to the completion of last job.
 Sequencing Problems 195

The step-by-step procedure due to S.M. Johnson Solution The given table is
and R. Bellman is as follows.
Job Machines
Step 1 Examine the columns (showing the pro-
cessing times) on machines A and B and select the A B
smallest among all of them: min {Ai, Bi} 1 3 8
2 12 10
Step 2 3 5 9
(i) If the minimum value belongs to column A 4 2 6
schedule this job first on machine A. Suppose 5 9 3
this value falls in column B then schedule 6 11 1
this job last on machine A.
We find that the smallest value of time is 1 unit
(ii) If there are two equal minimum values one for job 6 in the 2 nd column B. Therefore we
in each column, schedule the one in the first schedule job 6 as last.
column as the first job and the other in the
second column as the last.
6
(iii) If both equal minimum values are in the first
column select the job corresponding to the The reduced processing times are
smallest subscript first.
Job A B
(iv) If the equal values occur in the second
1 3 8
column select the job corresponding to the
2 12 10
largest subscript first. 3 5 9
Step 3 Delete the assigned job. Continue the pro- 4 2 6
cess of placing the jobs next to first or next to last 5 9 3
till all the jobs are assigned. The resulting sequence
Now the smallest value is 2 for job 4 on A.
will minimize the total time T.
Hence we schedule job 4 as the first on A. We
Note: have
Idle time for the machine A (M1)
= Total time – Time when the last job is 4 6
finished on A.
The reduced table is
Idle time for the machine B (M2)
= Time at which the first job is finished on Job A B
A + S [Time of starting of the jth job on 1 3 8
B – Time when (j – 1)th job is finished 2 12 10
on B] 3 5 9
5 9 3
Example 16.1 Find the sequence of jobs that
minimizes the total elapsed time to complete the Now there are two equal minimum values 3 one
following jobs on two machines. for job 1 on A and the other for job 5 on B. By the
rule job 1 is scheduled next to job 4 and job 5
Job 1 2 3 4 5 6 before job 6. Thus we get
Machine A 3 12 5 2 9 11
4 1 5 6
Machine B 8 10 9 6 3 1
 196 Operations Research

The reduced table is Now we have 8 tasks with the processing times
as shown below:
Job A B
Task B C D E F G H I
2 12 10
3 5 9 Machine 1 5 4 9 6 8 7 5 4
Machine 2 8 7 4 3 9 3 8 11
Here the minimum is 5 on A for job 3. Hence
job 3 is scheduled next to job 1 on A. Naturally In this table the minimum time is 3 on Machine
job 2 is assigned next to job 3. Thus we get the 2 corresponding to E and G. Since the processing
optimal sequence. time of E on Machine 1 is less than that of G on
Machine 1, E is placed as the last and G is
4 1 3 2 5 6 processed before E. Thus we have

The following table gives the details of time taken.

We have the following table for the remaining jobs
Job A B
Task B C D F H I
Time-in Time-out Time-in Time-out
Machine 1 5 4 9 8 5 4
4 0 2 2 8 Machine 2 8 7 4 9 8 11
1 2 5 8 16
3 5 10 16 25 Now the least time is 4 which corresponds to
2 10 22 25 35 tasks C and I on Machine 1 and D on Machine 2.
5 22 31 35 38 Therefore C will be taken next to A and I will be
6 31 42 42 43 taken up next to C and D at last before G. Thus
the sequence is
Idle time for machine A is 43 – 42 = 1 unit.
Idle time for B is 2 + (42 – 38) = 6 units.
Example 16.2 Determine the optimal sequenc- The reduced table is
ing to complete the following tasks on two ma-
Task B F H
chines.
Machine 1 5 8 5
Task A B C D E F G H I Machine 2 8 9 8
Machine 1 2 5 4 9 6 8 7 5 4 The smallest time is 5 on Machine 1 for B and
Machine 2 6 8 7 4 3 9 3 8 11 H. The corresponding times of these jobs on
Machine 2 are the same, we can select either of
Solution The given table is:
the jobs for 4th and 5th places. Finally F is assigned
Task A B C D E F G H I the 6th place. Thus the optimal sequence is
Machine 1 2 5 4 9 6 8 7 5 4
Machine 2 6 8 7 4 3 9 3 8 11
or
The smallest processing time is 2 corresponding
to the task A on machine Machine 1.
\ We assign the task A to Machine 1 first. The minimum time elapsed for completing all
the jobs.
 Sequencing Problems 197

Task Machine 1 Machine 2 The reduced set of processing times is

Time-in Time-out Time-in Time-out Job 2 3 4 5
A 0 2 2 8 Machine 1 12 15 6 10
C 2 6 8 15 Machine 2 10 10 6 12
I 6 10 15 26
B 10 15 26 34 Here we have two equal minimum values 6 for
H 15 20 34 42 the same job 4 on both Machine 1 and Machine 2.
F 20 28 42 51 We can put job 4 next to job 1 or before job 7. Let
D 28 37 51 55 us choose job 4 next to job 1.
G 37 44 55 58
E 44 50 58 61 1 4 7 6

Total time taken to complete all the jobs is Now we have the reduced table
61 hours. Job 2 3 5
Idle time for Machine 1 is 61 – 50 = 11 hours
Machine 1 12 15 10
Idle time for Machine 2 is 2 + 58 – 58 = 2 hours
Machine 2 10 10 12
Example 16.3 Determine the optimal
sequencing of the following 7 jobs on two Here we have 10 as the minimum value for job
machines Machine 1 and Machine 2. 5 on Machine 1 and job 2 and job 3 on Machine 2.
By the rule job 5 is processed next to job 4, job 2
Job 1 2 3 4 5 6 7 before job 7 and job 3 before job 2. This gives the
optimal sequence
Machine 1 3 12 15 6 10 11 9
Machine 2 8 10 10 6 12 1 3 1 4 5 3 2 7 6
Solution: The given table is The elapsed time according to this sequence is
given below:
Job 1 2 3 4 5 6 7
Job Machine 1 Machine 2 Idle time for
Machine 1 3 12 15 6 10 11 9 Time-in Time-out Time-in Time-out Machine 2
Machine 2 8 10 10 6 12 1 3
1 0 3 3 11 3
The least processing time is 1 on Machine 2 4 3 9 11 17 0
for job 6. Assign job 6, the last place. We get 5 9 19 19 31 2
3 19 34 34 44 3
6 2 34 46 46 56 2
7 46 55 56 59 0
Next we see that there are two equal values 3 6 55 66 66 67 7
one for job 1 on Machine 1 and the other for job 7
17 hrs
on Machine 2.
Total time taken is 67 hrs
Job 1 2 3 4 5 7 Idle time for Machine 1 is 67 – 66 = 1 hr
Machine 1 3 12 15 6 10 9 Idle time for Machine 2 is 17 hrs
Machine 2 8 10 10 6 12 3
16.3 PROCESSING n JOBS ON
Take job 1 as the first and job 7 as the last but
THREE MACHINES
one. Therefore we have
For processing n jobs on three machines say A, B
1 7 6 and C, the previous method can be extended.
 198 Operations Research

Extension is possible with either one or both of The minimum value is 8 for job 3 on G. Hence
the following conditions. If neither condition holds job 3 is processed first
good this method fails. The conditions are
3
(i) the minimum time on machine A must be
greater than or equal to the maximum time The reduced columns are
on B. Job G H
(ii) the minimum time on C must be greater than
or equal to the maximum time on B. 1 13 9
2 16 15
One or both of these conditions must hold in 4 10 9
order to apply the method. 5 15 9
The procedure is given by the following steps:
Here the minimum is 9 for 3 jobs 1, 4 and 5 on
(i) Introduce two dummy machines G and H
H (decreasing order of the processing times on G
with the corresponding processing times
are 15, 13, 10). We place job 4 at the last, job 1
given by
next and job 5 before job 1. Job 2 is processed
Gj = Aj + Bj j = 1, 2, 3, …, n
next to job 3. Thus we get the sequence
Hj = Bj + Cj j = 1, 2, 3, …, n
(The processing times of the corresponding 3 2 5 1 4
jobs are added)
(ii) Consider the machines G and H and apply The elapsed time according to this sequence is
the previous method taking the order GH. given below.
Example 16.4 Determine the sequence which Job Machine
minimizes the total time for processing five jobs
A B C
on three machines A, B and C. The following table
gives the processing times. Time- Time- Time- Time- Time- Time-
in out in out in out
Job 1 2 3 4 5 3 0 6 6 8 8 16
Machine A 8 10 6 7 11 2 6 16 16 22 22 31
Machine B 5 6 2 3 4 5 16 27 27 31 31 36
Machine C 4 9 8 6 5 1 27 35 35 40 40 44
4 35 42 42 45 45 51
Solution In the table we find that
min Aj = 6 min Cj = 4 Total time elapsed 51 hrs
max Bj = 6 Idle time for A = (45 – 42) + (51 – 45) = 9 hrs
Idle time for B = 6 + 8 + 5 + 4 + 2 + 6 = 31 hrs
Since min Aj ≥ max Bj condition (i) is satisfied.
Idle time for C = 8 + 6 + 0 + 4 + 1 = 19 hrs
Now take Gj = Aj + Bj ( j = 1, 2, 3, 4, 5)
Hj = Bj + Cj ( j = 1, 2, 3, 4, 5) Example 16.5 The processing times of six jobs
We have on three machines M1, M2, M3 are given below.

Job Machine Job M1 M2 M3

G H 1 3 8 13
2 12 6 14
1 13 9 3 5 4 9
2 16 15 4 2 6 12
3 8 10 5 9 3 8
4 10 9 6 11 1 13
5 15 9
Determine the optimal sequence of jobs.
 Sequencing Problems 199

Solution Here we find that minimum Aj = 2 4 3 1 6 2 5

minimum Cj = 8; maximum Bj = 8; condition (ii)
is satisfied. The elapsed time for completing all the jobs is
Take Gj = Aj + Bj ( j = 1, 2, 3, 4, 5, 6) given by
Hj = Bj + Cj Job M1 M2 M3
We have the table giving processing times on Time- Time- Time- Time- Time- Time-
G and H as given below: in out in out in out
4 0 2 2 8 8 20
Job 1 2 3 4 5 6 3 2 7 8 12 20 29
G 11 18 9 8 12 12 1 7 10 12 20 29 42
H 21 20 13 18 11 14 6 10 21 21 22 42 55
2 21 33 33 39 55 69
The minimum time is 8 for job 4 on G. Hence job 5 33 42 42 45 69 77
4 is processed first.
The total time taken to complete all the job is
4 77 hrs
The idle time for M1 is 77 – 42 = 35 hrs
The reduced table is The idle time for M2 is 2 + 0 + 0 + 11 + 11 + 3
+ 32 = 59 hrs
Job 1 2 3 5 6
The idle time for M3 is 8 hrs
G 11 18 9 12 12
H 21 20 13 11 14 16.4 PROCESSING n JOBS
THROUGH m MACHINES
Now the minimum is 9 for job 3 on G. Hence
job 3 is processed next to job 4. Let there be n jobs A1, A2, A3 … An to be processed
through m machines M1, M2…Mm in the order M1
4 3 M2 M3 … Mm. The optimal solution to this problem
The reduced table is can be found if either or both of the following
conditions are satisfied.
Job 1 2 5 6
ìi = 1, 2, 3, ..., n
G 11 18 12 12 (i) Minimum Ai1 ≥ max Aij í
H 21 20 11 14 î j = 2, 3...., m - 1
ìi = 1, 2, ..., n
(ii) Minimum Aim ≥ max Aij í
Here the minimum is 11 for job 1 on G and for î j = 2, 3, ..., m - 1
job 5 on H. Process job 1 next to job 3 and job 5 or minimum processing time on M1 or Mm must
last in the sequence. We get be not less than the maximum processing time on
4 3 1 5 any of the remaining m – 2 machines. The table is

The reduced table is Machines

M1 M2 M3 ... Mm
Job 2 6 A1 A11 A12 A13 ... A1m
A2 A21 A22 A23 ... A2m
G 18 12 Jobs A3 A31 A32 A33 ... A3m
H 20 14 . . . . . .
The minimum 12 is for job 6 on G. Therefore . . . . . .
. . . . . .
job 6 is processed next to 1. The optimal sequence An An1 An2 An3 ... Anm
becomes
 200 Operations Research

If the condition is satisfied then we can solve The new table is

the problem as follows.
Job G H
Step 1 Examine whether the condition is satisfied A 17 19
Step 2 Convert the m machine problem to a two B 21 25
machine problem. C 20 23
Introduce two dummy machines G and H with The minimum is 17 for job A on G. Therefore, A is
the processing times processed first.
Gi = Ai1 + Ai2 + Ai3 + …. Aim – 1 (i = 1, 2, 3,…, n)
Hi = Ai2 + Ai3 + Ai4 + …. Aim (i = 1, 2, 3,…, n) In the reduced table the minimum is 20 for C
That is processing time of a job on G, is the on G. Hence C is processed next to A. Thus the
sum of the processing times of the job on M1, sequence is
M2,…., Mm–1 and processing time of a job on H, is
the sum of the processing times of the job on M2,
Total elapsed time is given by the following
M3, … Mm.
table.
Step 3 The new processing times so obtained can
Job M1 M2 M3 M4 M5
be used for solving two machine – n job problem
Time- Time- Time- Time- Time- Time- Time- Time- Time- Time-
as before. in out in out in out in out in out

Example 16.6 Solve the following sequencing A 0 7 7 12 12 14 14 17 17 26

C 7 12 12 16 16 21 21 27 27 35
problem of four jobs on five machines B 12 18 18 24 24 28 28 33 35 45
D 18 26 26 29 29 32 33 35 45 51
Job Machines
M1 M2 M3 M4 M5
Minimum time required to complete all the jobs
is 51 hrs.
A 7 5 2 3 9 Idle time for M1 : 51 – 26 = 25 hrs
B 6 6 4 5 10 Idle time for M2 : 7 + 2 + 2 + (51 – 29) = 33 hrs
C 5 4 5 6 8
Idle time for M3 : 12 + 2 + 3 + 1 + (51 – 32)
D 8 3 3 2 6
= 37 hrs
Solution In the given table min Ai1 = 5 Idle time for M4 : 17 + 1 = 18 hrs
min Ai5 = 6 max Aij = 6 Example 16.7 Solve the following sequencing
\ The condition is satisfied. problem.
Take Gi = Ai1 + Ai2 + Ai3 + Ai4
Machines
Hi = Ai2 + Ai3 + Ai4 + Ai5
We have the new table for two machines G and M1 M2 M3 M4 M5 M6 M7
H. A 20 10 9 4 12 9 40
Jobs B 22 8 11 8 10 10 30
Job G H C 12 7 10 7 9 12 32
A 17 19 D 30 6 5 6 10 11 28
B 21 25
C 20 23 Solution In the table min Ai1 = 12 min Ai7 = 28
D 16 14 max Aij = 12
Condition is satisfied.
Minimum is 14 corresponding to the job D on
Take Gi = Ai1 + Ai2 + Ai3 + Ai4 + Ai5 + Ai6
H. Therefore D is selected for processing last.
Hi = Ai2 + Ai3 + Ai4 + Ai5 + Ai6 + Ai7
The table of processing times on G and H is
 Sequencing Problems 201

Job Machines Deleting C from the table, the minimum of the

G H
A 64
B 69
C 57
D 68
The minimum is 57 for job C on G. Hence
job C is to be processed first
remaining processing times is 64 for job A on G.
\ job A is processed next to C.
84
77
77 In the remaining table (after deleting C and A)
66 the minimum is 66 for D on H. Hence job D is
assigned the last place. Thus we get the sequence
The total elapsed time is given by

Job M1 M2 M3 M4 M5 M6 M7
Time- Time- Time- Time- Time- Time- Time- Time- Time- Time- Time- Time- Time- Time-
in out in out in out in out in out in out in out
C 0 12 12 19 19 29 29 36 36 45 45 57 57 89
A 12 32 32 42 42 51 51 55 55 67 67 76 89 129
B 32 54 54 62 62 73 73 81 81 91 91 101 129 159
D 54 84 84 90 90 95 95 101 101 111 111 122 159 187

The total time elapsed is 187 hrs. M4 : 29 + 15 + 18 + 14 + 86 = 162

Idle time for: M5 : 36 + 10 + 14 + 10 + 76 = 146
M1 : 187 – 84 = 103 M6 : 45 + 10 + 15 + 10 + 65 = 145
M2 : 12 + 13 + 12 + 22 + 97 = 156 M7 : 57
M3 : 19 + 13 + 11 + 17 + 92 = 152

EXERCISES
■ ■ ■ ■ ■ ■ ■

Solve the following sequencing problems: 3. Seven jobs on two machines

1. Five jobs on two machines:
Job 1 2 3 4 5 6 7
Job 1 2 3 4 5 M1 5 7 3 4 6 7 12
M1 5 1 9 3 10 M2 2 6 7 5 9 5 8
M2 2 6 7 8 4
4. Six jobs on three machines
2. Six jobs on two machines
Job 1 2 3 4 5 6
Job 1 2 3 4 5 6 M1 18 12 29 35 43 37
M1 5 9 4 7 8 6 M2 7 12 11 2 6 12
M2 7 4 8 3 9 5 M3 19 12 23 47 28 36
 202 Operations Research

5. Six jobs on three machines 11. Seven jobs on three machines

Job 1 2 3 4 5 6 Job A B C D E F G
M1 8 3 7 2 5 1 M1 3 8 7 4 9 8 7
M2 3 4 5 2 1 6 M2 4 3 2 5 1 4 3
M3 6 7 5 11 5 6 12
M3 8 7 6 9 10 9
12. Five jobs on three machines
6. Four jobs on six machines
Job 1 2 3 4 5
Job Machines
M1 3 9 6 5 4
M1 M2 M3 M4 M5 M6 M2 4 5 1 2 3
A 18 8 7 2 10 25 M3 8 9 5 7 10
B 17 6 9 6 8 19
C 11 5 8 5 7 15 13. Six jobs on three machines
D 20 4 3 4 8 12
Job 1 2 3 4 5 6
7. Five jobs on four machines
M1 12 8 7 11 10 5
Job Machines M2 3 4 2 5 5 4
M3 7 10 9 6 10 4
M1 M2 M3 M4
A 10 3 5 14 14. Four jobs on four machines
B 12 2 6 7
C 8 4 4 12 Job 1 2 3 4
D 15 1 7 8
M1 15 5 5 15
E 16 5 3 10
M2 12 2 10 12
8. Five jobs on two machines M3 16 2 4 16
M4 18 3 4 18
Job 1 2 3 4 5
M1 3 8 5 7 4 15. Seven jobs on three machines
M2 4 10 6 5 8
Job 1 2 3 4 5 6 7
9. Five jobs on two machines M1 12 6 5 11 5 7 6
M2 7 8 9 4 7 8 3
Job 1 2 3 4 5 M3 3 4 1 5 2 3 4
M1 5 4 8 7 6
M2 3 9 2 4 10 16. Five jobs on four machines

Job A B C D E
10. Eight jobs on two machines
M1 11 13 9 16 17
Job A B C D E F G H M2 4 3 5 2 6
M1 14 26 17 11 9 26 18 15 M3 6 7 5 8 4
M2 21 15 16 21 22 12 13 25 M4 15 8 13 9 11
 Sequencing Problems 203

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. 2 4 3 5 1 Idle time for M3 = 51

Idle time for M4 = 25
Total time = 30 units
8. 1 5 3 4 2
Idle time for M1 = 2
Idle time for M2 = 3 Total time = 37 units
Idle time for M1 = 10 units
2. 3 1 6 5 2 4 Idle time for M2 = 4 units
Total time = 42 units 9.
Idle time for M1 = 3
Idle time for M2 = 6 Total time = 32 units
Idle time for M1 = 2 units
3. 3 4 5 7 2 6 1 Idle time for M2 = 4 units
Total time = 46 units
10.
Idle time for M1 = 3
Idle time for M2 = 5 Total time = 154 units
Idle time for M1 = 18 units
4. 2 1 3 4 6 5 Idle time for M2 = 9 units
Total time = 209 units 11. A D G F B C E
Idle time for M1 = 34
Idle time for M2 = 159 Total time = 59 units
Idle time for M3 = 44 Idle time for M1 = 13 units
Idle time for M2 = 37 units
5. Idle time for M3 = 7 units
Total time = 53 units 12. 4 1 5 2 3
Idle time for M1 = 27
Total time = 46 units
Idle time for M2 = 32
Idle time for M1 = 19 units
Idle time for M3 = 4
Idle time for M2 = 31 units
6. C A B D Idle time for M3 = 7 units
Total time = 112 units 13. 3 2 5 4 1 6
Idle time for M1 = 46
Total time = 62 units
Idle time for M2 = 99
Idle time for M1 = 9 units
Idle time for M3 = 85
Idle time for M2 = 39 units
Idle time for M4 = 95
Idle time for M3 = 16 units
Idle time for M5 = 79
Idle time for M6 = 41 14. C B D A
7. C A E D B (or)
Total time = 76 units
Idle time for M1 = 15 C B A D
Idle time for M2 = 61 Total time = 86 units
 204 Operations Research

Idle time for M1 = 25 units Idle time for M1 = 7 units

Idle time for M2 = 74 units Idle time for M2 = 13 units
Idle time for M3 = 63 units Idle time for M3 = 37 units
Idle time for M4 = 25 units
16. C A E D B
15. 3 5 6 2 1 4 7 Total time = 83 units
(or) Idle time for M1 = 18 units
Idle time for M2 = 63 units
3 5 2 6 1 4 7 Idle time for M3 = 53 units
Total time = 59 units Idle time for M4 = 29 units

Queueing Theory
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
17.1 INTRODUCTION
In our everyday life people are seen waiting at a
railway ticket booking office, a doctor’s clinic, post
office, bank, petrol pump and many other places
for getting service. When there is too much
demand on the facilities available, the customers
have to wait and hence they form a queue or
waiting line. The problem is to reduce congestion
and delay in serving the customer. A unit that
requires some service to be performed, is called a
customer. The system which offers service is called
the service facility. Customers arrive at a service
counter and are attended by servers. After the
service is completed, the customer leaves the
system.
17.2 CHARACTERISTICS OF A
QUEUEING SYSTEM
A queueing system can be completely described
by the following characteristics.
(i) Input process This gives the mode of arrival
of the customers into the system. Generally the
customers arrive in a random manner. Hence the
distribution of inter-arrival time follows some
17
probability law. We assume that the customers
arrive in a ‘Poisson’ fashion. That is to say, the
rate of arrival is a poisson variate. Mean (average)
arrival rate is taken as l.
(ii) Queue discipline This is the law according
to which the customers are served. The simplest
rule is First-in First-out (FIFO) where the cus-
tomers are served according to the order of their
arrival. Generally this rule is followed almost
everywhere, like ration shops, ticket booking, etc.,
Another rule is Last-in First-out (LIFO). Here the
unit which arrives last is served first as in the case
of a passenger who gets into a crowded bus as the
last man, gets down first or an item which is
dumped into a godown, as the last one, is taken out
first. The third rule is Service in Random Order
(SIRO) where the customers are selected for
service at random irrespective of their arrival time.
(iii) Service Mechanism This is the facility
available in the system to serve the customers. In
certain cases one server many find it difficult when
there is a large number of customers in a shop. Then
the number of servers may be increased in order to
reduce the waiting time of the customers lest the
 206 Operations Research
business should be lost. The number of servers
(service channels) depends on the number of
customers and their waiting time in the system. In a
multichannel system (as in the case of railway or
airlines ticket booking) many customers are served
simultaneously. Also the customers may be served
in batches of fixed size as seen in cinema hall,
marriage dinner party, etc. It is called bulk bsrvice.
(iv) Capacity of the system This is the capacity
of the queue or waiting line. It is the number as to
how many customers can be there in the queue.
Generally a queue is of infinite capacity. In some
systems the customers are admitted into a waiting
room whose capacity is limited. When the wait-
ing line reaches a certain length no further cus-
tomers are allowed until space becomes available
by a service completion.
17.3 SYMBOLS AND
ASSUMPTIONS
The following symbols are generally used:
n – number of customers in the system
both waiting and in service
l – average number of customers arriving
per unit time (arrival rate)
m – average number of customers served
per unit time (service rate)
l In the notation M/M/1/•, M denotes poisson arrival
r – (traffic density)
m
c – number of servers
E(n) – average number of customers in the
system L(s).
E(m) – average number of customers in the
queue L(q)
Pn(t) – Probability that there are n customers
in the system at time t.
N(t) – Total number of arrivals up to time t.
We make the following assumptions:
(i) Probability that one arrival occurs during a
small interval Dt is
l(Dt) + 0(Dt)
(ii) Dt is very small such that the probability of
more than one arrival in Dt is of order Dt.
0(Dt) represents terms such that
0( Dt )
lt ®0
Dt ® 0 ( Dt )
(iii) In short (Dt)2 and higher powers of Dt are
negligible.
(iv) The number of arrivals in the intervals are
statistically independent. The process has
independent increments.
(v) Similarly probability that one service is
completed during Dt is
m(Dt) + 0(Dt)
0( Dt )
lt ®0
Dt ® 0 ( Dt )
17.4 TRANSIENT AND STEADY
STATES
A queueing system is said to be in transient state
if the operating characteristics like input, output,
mean queue length, etc. are dependent on time. If
these are independent of time then the system is
said to be in steady state. For example Pn(t) is a
function of time t in transient state. But Pn(t) is
constant for any value of t if it is in steady state.
•
17.5 MODEL I (M/M/1/•
QUEUEING MODEL)
rate and exponential service rate, 1 denotes the
number of servers (single server system) and • is
the capacity of the system. It is understood that
service is done on FIFO basis. Mean arrival rate
is l and mean service rate is m.
17.5.1 Formula for Steady State
Probability Pn
The probability that there are n customers at time
(t + Dt) is given by Pn(t + Dt). This probability is
the sum of the probabilities of four mutually
exclusive cases (service completion and departure
are the same)
Pn(t + Dt) = Pn(t)P (no arrival in Dt).
P(no departure in Dt)
+ Pn(t).P (one arrival in Dt) P(one departure
in Dt)
 Queueing Theory 207

+ Pn + 1(t).P (no arrival in Dt) P (one departure i.e. mPn+1 = (l + m)Pn – lPn–1
in Dt) (l + m ) l
+ Pn – 1(t).P (one arrival in Dt) P (no departure Pn+1 = Pn - Pn -1 n ≥ 1 (3)
m m
in Dt)
Also from (2) we have l
= Pn(t)[1 – l(Dt) + 0Dt][1 – m(Dt) + 0(Dt)]
0 = –lP0 + mP1 or P1 = P0
+ Pn(t)[l(Dt) + 0Dt][m(Dt) + 0(Dt)] m
i.e. P1 = rP0
+ Pn + 1(t)[1 – l(Dt) + 0Dt][m(Dt) + 0(Dt)]
Substituting n = 1, 2, 3,… successively in the
+ Pn – 1(t)[l(Dt) + 0Dt][1 – m(Dt) + 0(Dt)].
recurrence relation (3) with
(n ≥ 1)
i.e. Pn(t + Dt) = Pn(t)[1 – l(Dt) – m(Dt) + 0(Dt)] l
P1 = P
+ Pn(t)[0(Dt)] m 0
+ Pn + 1(t)[m(Dt) + 0(Dt)] we get the value of Pn for all n, in terms of P0
+ Pn – 1(t)[l(Dt) + 0(Dt)] For n = 1
= Pn(t) – (l + m)Pn(t)(Dt) + mPn+1(Dt) (l + m ) l
+ lPn–1(t)(Dt) + 0(Dt) P2 = P1 - P0
m m
\ Pn(t + Dt) – Pn(t) = –(l + m)Pn(t)(Dt)
+ mPn+1 (t)(Dt) + lPn–1(t)(Dt) + 0(Dt) (l + m ) l l
= P0 - P0
Pn (t + D t ) - Pn (t ) m m m
= –(l + m)Pn(t) + mPn+1(t)
Dt l é (l + m ) ù
0( D t ) = mê m - 1ú P0
+ lPn–1(t) + ( D t ) ë û
2
Taking limits as Dt Æ 0 we get æ lö
= ç ÷ P0
Pn¢(t) = –(l + m) Pn(t) + mPn+1(t) + lPn–1(t) (1) è mø
For n = 0, Pn–1(t) does not exist. We take this
case separately. \ P2 = r2P0
P0(t + Dt) = P0(t).P (no arrival in (Dt)) For n = 2
+ P1(t).P (no arrival in (Dt)). P(one departure in (l + m ) l
P3 = P2 - P1
(Dt)) m m
= P0(t)[1 – l(Dt) + 0(Dt)] 2
+ P1(t)[1 – l(Dt) + 0(Dt)][m(Dt) + 0(Dt)] æ (l + m ) ö æ l ö l l
= ç ÷ ç ÷ P0 - . P0
= P0(t) – lP0(t)(Dt) + mP1(Dt) + 0(Dt)] è m ø è mø m m
P0(t + Dt) – P0(t) = –lP0t(Dt) + mP1(t)(Dt)
2
+ 0(Dt) æ lö æ (l + m ) ö
= ç ÷ P0 ç - 1÷
P0 (t + Dt ) - P0 (t ) è mø è m ø
= - l P0t + m P1 (t )
Dt 3
0( D t ) æ lö
+ . = ç ÷ P0
(D t ) è mø
Taking limits as Dt Æ 0 we get
\ P3 = r2P0
P0 ¢(t ) = - l P0 (t ) + m P1 (t ) (2) Thus we get \ P4 = r4P0 P5 = r5P0 and so on
For steady state probability Pn(t) and P0(t) are \ Pn = rnP0
independent of time. Hence Now the total probability is 1
Pn¢(t) = P0¢(t) = 0 \ P0 + P1 + P3 + ... + Pn + ... to • = 1
Therefore deleting the time t, we get i.e. P0 + rP0 + r2P0 + r3P0 + ... to • = 1
0 = -(l + m ) Pn + m Pn +1 + l Pn -1 from (1) P0[1 + r + r2 + r3 + ... •] = 1
 208 Operations Research

æ 1 ö (iii) Average number of customers in the queue

P0 ç = 1 Þ P0 = 1 - r ( r < 1)
è 1 - r ÷ø
E(m)
m = n – 1 (number of customers in the queue
Thus we get the expression for the steady state excluding one customer in service)
l ¥
probability, as Pn = rn(1 – r) where r = m .
E(m) = å (n - 1) Pn
Note: The formula for P0 is true only when (r < 1) n =1
or l < m (condition for the existence of sum to ¥ ¥
infinity of a G.P.) = å n.Pn - å Pn
n =0 n =1
17.5.2 Important Formulae = E(n) – (r + r2 + r3 + ... + •)
(i) Probability that the number of customers = E(n) – (1 – P0)
in the system is greater than or equal to n r
¥ = - [1 - (1 - r )]
P(N ≥ n) = å Pn 1- r
N =n
¥ r r2 l2
= -r= or
= å p (1 - r )
n
1- r 1- r m(m - l )
N =n
= (1 – r)[rn + rn + 1 + rn+2 + ... to •] r2 l2
i.e. E (m) = =
= (1 – r)[1 + r + r2 + ... + rn–1 + rn 1 - r m (m - l )
+ rn+1 + ... to • – (1 + r + r2 + r3 [Also denoted by L(q)]
+ ... + rn –1)] (iv) Expected waiting time of a customer in the
é 1 1 - rn ù system (Ws)
= (1 - r) ê - ú The expected period of time a unit has to be
ë1 - r 1 - r û in the system from arrival till completion of
é 1 - (1 - r n ) ù service is given by
= (1 - r) ê ú
ë 1- r û Expected number of units in
= rn the system
Ws =
(ii) Average number of customers in the Arrival rate
system E ( n)
\ Ws =
E(n) = å n.Pn éë E ( X ) = å XP( X ) ùû l
l 1
¥ = =
= å n.r n
(1 - r) l (m - l ) m - l
n =0 (v) Expected waiting time of a customer in the
¥ queue (Wq)
= r(1 - r) å n r
n -1
The expected time a unit has to wait in the
n =1 queue before entering into service is given
= r(1 – r)[1 + 2r + 3r2 + ... to •] by subtracting the service time from the total
= r(1 – r)(1 – r)–2 (Binomial series) waiting time in the system.
r(1 - r) r l 1
= = or Wq = Ws -
(1 - r) 2 (1 - r) m-l m
r 1 1 l
\ E(n) = [Also denoted by L(s)] = - =
1- r m - l m m(m - l )
 Queueing Theory 209

l Service time = 36 min per unit

\ Wq = Service rate = 1/36 per min
m (m - l )
\ m = 1/36.
1
Note: (i) Service rate = m. Service time = m l 30 ´ 36 3
r= = =
1
m 60 ´ 24 ´ 1 4
(ii) Arrival rate = (i) Number of trains exceeds 10 fi n ≥ 11
inter arrival time
(iii) E(n) = l.Ws; E(m) = lWq. This is known Required probability = r11
as Little’s formula æ 3ö
11
= ç ÷
Example 17.1 A TV repair man finds that the è 4ø
time spent on repairing has an exponential (ii) Average number of trains in the yard = E(n)
distribution with mean 30 min per unit. The arrival
of TV sets is poisson with an average of 10 sets æ r ö 3/ 4
per day of 8 hours. What is his expected idle time çè 1 - r ÷ø = 1 / 4 = 3
per day? How many sets are there on the average?
Solution Example 17.3 In a store with one server,
1 9 customers arrive on an average of 5 minutes.
Service time per unit = 30 min = hr Service is done for 10 customers in 5 minutes.
2 Find
Service rate (per hr) = 2 i.e. m = 2
(i) the average number of customers in the
Arrival rate = 10 per 8 hr
system.
= 10/8 per hr
(ii) the average queue length.
\ l = 10/8 = 5/4
(iii) the average time a customer spends in the
P0 = Probability that there is no unit
store.
in the system
(v) the average time a customer waits before
l
= 1–r=1– being served.
m Solution 9 customers arrive in 5 min
= 1 – 5/8 = 3/8 \ Arrival rate = 9/5 per min
\ Idle time per day of 8 hrs = 3/8 ¥ 8 = 3 hrs \ l = 9/5
Average number of units in the system = E(n) 10 customers are served in 5 min
r l 5/4 \ Service rate = 10/5 per min
= = = = 5 / 3 sets
1- r m - l 2 - 5/4 \ m=2
(i) average number of customers in the system
Example 17.2 In a railway yard goods trains
= E(n)
arrive at the rate of 30 trains per day. Assuming
that the service time follows exponential l 9/5
= = =9
distribution with an average of 36 minutes, find m - l 2 - 9/5
(i) the probability that the number of trains in (ii) average queue length = E(m)
the yard exceeds 10.
(ii) the average number of trains in the yard. l2 l l
= = .
Solution m(m - l ) m (m - l )
Arrival rate = 30 trains per day (24 hr)
9/5
30 = .9 = 8.1
= per hour 2
24
30 (iii) average time a customer spends in the system
\ l= per min = Ws
60 ´ 24
 210 Operations Research

E ( n) 1 (ii) Average queue length is E(m).

= =
l m-l l2 1 / 81 1
E ( m) = = =
1 m ( m - l ) 1 / 3(1 / 3 - 1 / 9) 6
= = 5 min
2 - 9/5 (iii) Probability that the phone is in use, is the
(iv) average time a customer spends in the queue, probability that the system is busy which is
before entering into service = Wq given by the traffic density r.
l \ Fraction of time, the system is busy
= 1
m (m - l ) =r =
3
1 æ l ö (iv) Let l1 be the arrival rate for the average
= m . çè m - l ø÷
waiting time in the queue to be at least
= (1/2)9 = 9/2 min 4 minutes
[Wq = Total time in the system – service time l1
1 Wq =
= 5– = 9/2 min] m ( m - l1 )
2
l1 9l1
Example 17.4 In a telephone booth, the arrivals \ 4= =
1æ1 ö 1 - 3l1
follow poisson distribution with an average of
ç - l1 ÷ø
3è3
9 minutes between two consecutive arrivals. The
duration of a telephone call is exponential with an \ 4 – 12l1 = 9l1
average of 3 min. l1 = 4/21
(i) Find the probability that a person arriving at Thus a second booth will be installed if the
the booth has to wait. arrival rate is 4/21.
(ii) Find the average queue length. \ Required increase in the flow of arrivals
(iii) Find the fraction of the day, the phone will
4 1 12 - 7 5
be in use. \ = - = = per min
(iv) The company will install a second booth if a 21 9 63 63
customer has to wait for phone, for at least Example 17.5 In a carwash station cars arrive
4 minutes. If so, find the increase in the flow for service according to poisson distribution, with
of arrivals in order that another booth will mean 4 per hour. The average service time of a
be installed. car is 10 min.
Solution (i) Determine the probability that an arriving car
Inter-arrival time = 9 min has to wait.
\ Arrival rate l = 1/9 per min (ii) Find the average time a car stays in the
Duration of a call = 3 min station.
\ Service rate m = 1/3 per min (iii) If the parking space cannot hold more than
l 3 1 6 cars, find the probability that an arriving
r = = = car has to wait outside.
m 9 3
(i) A person has to wait if the system is not Solution
empty. Arrival rate = 4 per hour
\ Probability that an arrival has to wait is l = 4/60 per min = 1/15
given by 1 – P0 Service time = 10 min per car
\ service rate = 1/10 per min.
1 1
1 - P0 = 1 - (1 - r) = r = m=
3 10
 Queueing Theory 211

l 10 2 1
r= = = E(m) is to be reduced to .
m 15 3 2
(i) The probability that an arriving car has to
wait Take the service rate as m1
= 1– P0 = 1 – (1 – r) l 16
\ 1/2 = 1 1 = 1 1
=r m ( m - l ) m ( m - 4)
= 2/3 2
= m - 4m - 32 = 0
1 1
(ii) Average time a car stays in the station
1 1
1 m = 8 or m = – 4 (not possible)
= Ws = \ m1 = 8
m-l
\ The revised service rate = 8
1 1 150
= 1 = = = 30 min \ Service time = 1/8 hr = 15/2 min
1 5 5
- Decrease in service time = 10 – 15/2 = 5/2 min
10 15 150 Increase in cost = 5/2 ¥ 10 = Rs 25
(iii) If the parking space is full, then there are \ The budget required
6 cars waiting for service and one car in = Rs 100 + Rs 25
service. Therefore n = 7. The new arrival has = Rs 125
to wait outside if n ≥ 8. The required Therefore in order to reduce the queue (increase
probability is the service rate to 8) the clinic has to provide
8 Rs 125 as the cost per patient.
æ 2ö
P (n ³ 8) = r8 = ç ÷
è 3ø Example 17.7 Trucks arrive at a loading dock
Example 17.6 On an average 96 patients per day at an average rate of 4 trucks per hour. The loading
require the service of an emergency clinic which of a truck takes 10 minutes on the average by a
can handle only one patient at a time. It takes on crew of three loadmen. The operating cost of a
the average 10 minutes to give treatment to a truck is Rs 20 per hour and the loadmen are paid
patient. The cost of the treatment is Rs 100 per at Rs 6 each per hour. Is it advisable to add another
patient for 10 minutes. The cost increases at crew of three loadmen?
Rs 10 per minute of time reduced. How much Solution Arrival rate of trucks = 4 per hour
amount should be budgeted by the clinic to reduce l=4
the queue size to 1/2? Service time (loading time) = 10 min = 1/6 hour
Solution Average number of arrivals = 96 per \ Service rate = 6 per hour
day m=6
96 Loading crew cost = 3 ¥ 6 = Rs 18
= = 4 per hour Total time taken by truck
24
\ l=4 1
= Ws =
Service time = 10 min per patient = 1/6 hr m-l
\ service rate = 6 per hour 1 1
m=6 = = hour
6-4 2
Queue length = E(m) Cost of waiting time
l2 16 1
= = = (arrival rate) ¥ (hourly cost)
m ( m - l ) 6(6 - 4) 2
4 1
= = ¥ 4 ¥ 20 = Rs 40
3 2
 212 Operations Research

Total hourly cost PN (t + D t ) - PN (t )

= - m PN (t ) + l PN -1 (t )
= crew cost + cost of waiting time Dt
= 18 + 40 = Rs 58 0( D t )
If the crew becomes 6, then +
(D t )
Crew cost = 6 ¥ 6 = Rs 36
In the limit as (Dt) Æ 0
Service rate = 12 per hour
PN¢ (t) = –mPN(t) + lPN–1(t)
m = 12
For steady state probability PN¢ = 0
1 \ 0 = –mPN + lPN–1
Ws =
m-l mPN = lPN–1
1 1 l
= PN = P
=
12 - 4 8
hour m N -1
Cost of waiting time = 1/8 ¥ 4 ¥ 20 = 10 Also for n < N,
The new total hourly cost = 36 + 10 = Rs 46 l+m l
Pn = Pn - Pn from (3) of 17.5.1
Existing total cost = Rs 58 per hour m m
Revised total cost = Rs 46 per hour. l
Therefore if the crew becomes 6, the cost is and P1 = P
m 0
reduced from Rs 58 to Rs 46. Hence it is advisable
Thus for n £ N, we get
to add three more loadmen. n
æ lö
Pn = ç ÷ P0 or Pn = r P0 (0 £ n £ N)
n
17.6 MODEL 2—M/M/1/N è mø
(QUEUE WITH FINITE N

CAPACITY N) Now å Pn = 1 (Total probability)

n =0
In this model the capacity of the system is limited i.e., P0 + r P0 + r 2 P0 + ... + r N P0 = 1
to N customers only. Maximum number of
customers allowed in the system is N. If there are P0 [1 + r + r 2 + ... + r N ] = 1
N customers in the system then the probability of é 1 - r N +1 ù
a new arrival is zero. So long as n < N this system \ P0 ê ú =1
is the same as model 1. For example, in a doctor’s ë 1- r û
1- r
clinic the waiting hall for the patients has only a \ P0 = ( r ¹ 1)
limited number of seats, say 20. When the waiting 1 - r N +1
hall is full, a newcomer has to stand outside and Note: Here the sum of N + 1 terms of a GP is
he is not considered as a member of the system. taken. Hence r need not be less than 1. The number
The maximum value n can take, is N, whereas in of units allowed into the system is controlled by
the previous model n can vary from 0 to •. the queue length N – 1 and not by l and m.
1
17.6.1 Formula for Pn If r = 1, P0(N + 1) = 1. Therefore P0 =
N +1
PN (t + Dt) = PN(t).P (no departure in Dt) Therefore we get
+ PN–1(t)P (one arrival in Dt).P æ 1- r ö
(no departure in Dt) Pn = r n ç ÷ (n £ N ) ü r ¹ 1
= PN(t)[1 – m(Dt) + 0(Dt)] è 1 - r N +1 ø
ï
+ PN–1(t)[l(Dt) + 0(Dt)][1 – m(Dt) + 0(Dt)] 1- r ý
P0 = ï
= PN(t) – mPN(t)(Dt) + lPN–1(t)(Dt) + 0(Dt) 1 - r N +1 þ
PN(t + Dt) – PN(t) = –mPN(t)(Dt) + 1
lPN–1(t)(Dt) + 0(Dt) Pn = P0 = (r = 1)
N +1
 Queueing Theory 213

17.6.2 Important formulae N

(i) Average number of customers in the = E ( n) - å Pn

n =1
system = E(n) – (1 – P0)
Average number of customers in the system
is given by é 1- r ù
= E (n) - ê1 - N +1 ú
N N ë 1- r û
E(n) = å n.Pn = å n.r n P0 é r - r N +1 ù
n =0 n =0
= E ( n ) - ê N +1 ú
N N
d (rn ) ë1- r û
= P0 .r å n.r = P0 .r å
n -1

n =0 n =0 d r r [1 - ( N + 1) r N + N r N +1 ]
=
d N n (1 - r)(1 - r N +1 )
= P0 .r år
d r n =0 r(1 - r N )
-
d é 1 - r N +1 ù 1 - r N +1
= P0 .r ê ú
dr ë 1 - r û
r [1 - ( N + 1) r N + N r N +1
é (1 - r){- ( N + 1) r N } - ù
ê ú - (1 - r)(1 - r N )]
ê (1 - r N +1 )(-1) ú =
= P0 .r ê ú (1 - r)(1 - r N +1 )
ê (1 - r) 2 ú
ê ú r
ë û = [1 – (N + 1)rN
(1 - r)(1 - r N +1 )
r
= P0 . + NrN+1 – (1 – rN – r + rN + 1)]
(1 - r ) 2
r
[–(N + 1)rN(1 – r) + 1 – rN + 1] =
(1 r)(1 - r N +1 )
-
r
= P0 . [1 – (N + 1)rN + NrN+1] [r – NrN + (N – 1)rN + 1]
(1 - r)2
(1 - r) r r2
. =
= -
1 r N +1
-
(1 r) 2 (1 - r)(1 - r N +1 )
[1 – (N + 1)rN + NrN + 1] [1 – NrN – 1 + (N – 1)rN] (r π 1)
r N ( N - 1)
=
(1 - r)(1 - r N +1 ) E(m) = 2( N + 1) (r = 1)
[1 – (N + 1)rN + NrN + 1] (iii) Expected waiting time in the system
( r π 1)
N
E ( n)
1 where l1 = l(1 – PN)
E (n) = å n.
WS =
If r=1 l1
n =0 N +1
N ( N + 1)
1 N E (n)
= WS =
=
N +1 2 2 l (1 - PN )
(ii) Average number of customers in the queue (iv) Expected waiting time in the queue
N
E(m) = å (n - 1) Pn Wq =
E ( m)
=
E ( m)
n =1 l 1
l (1 - PN )
 214 Operations Research

1 20 minutes and the service time is 36 minutes per

or Wq = WS -
m into service without waiting. Also find the average
Example 17.8 At a railway station only one
train is handled at a time. The yard can
accommodate only two trains to wait. Arrival
rate of trains is 6 per hour and the railway station
can handle them at the rate of 12 per hour.
Find the steady state probabilities for the
various number of trains in the system. Also
find the average waiting time of a newly arriving
train.
Solution We have
l = 6 per hour
m = 12 per hour
6 1
r= = = 0.5
12 2
Maximum queue length is 2 and the maximum
number of trains in the system is N = 3.
1- r 1 - 0.5
P0 = = = 0.53
1 - r N +1 1 - (0.5) 4
P1 = P0r = (0.53) (0.5) = 0.27
P2 = P0r2 = (0.53) (0.5)2 = 0.13
P3 = P0r3 = (0.53) (0.5)3 = 0.07
These are the probabilities for the various
number of trains in the system
3 + 3(1.8)3 + 4(1.8)4]
Now E(n) = å n.Pn
n =0
= 0 + 1(0.27) + 2(0.13) + 3(0.07)
= 0.74
Average number of trains in the system = 0.74
Each train takes 1/12 hour for service.
\ Service time for one train
= 1/12 hour = 0.085 hour
\ A newly arriving train finds 0.74 trains in
the system with service time
= 0.085 hr each.
\ Expected waiting time
= (0.74) (0.085) hr
= 0.0629 hour = 3.8 minutes
Example 17.9 The capacity of a queueing
system is 4. Inter-arrival time of the units is
unit. Find the probability that a new arrival enters
number of units in the system.
Solution Given that the inter-arrival time is
20 min
Therefore the arrival rate l = 1/20 per min
Service time = 36 min
\ service rate = 1/36 per min
l
\ r= = 36/20 = 1.8
m
N=4
(i) A new arrival can enter into service without
waiting if the system is empty. The
probability is
1- r r -1
P0 = N +1
= N +1
1- r r -1
1.8 - 1
= = 0.04
(1.8)5 - 1
(ii) Average number of units in the system is
4
E(n) = å n. pn = P1 + 2 P2 + 3P3 + 4 P4
n =1
= P0[r + 2r2 + 3r3 + 4r4]
= (0.04) [(1.8) + 2(1.8)2
= (0.04)(67.77) = 2.71 (nearly 3)
\ Average number of units in the system is
3 (nearly)
Example 17.10 Patients arrive at a clinic at the
rate of 30 patients per hour. The waiting hall cannot
accommodate more than 14 patients. It takes
3 minutes on the average to examine a patient.
(i) Find that probability that an arriving patient
need not wait.
(ii) Find the probability that an arriving patient
finds a vacant seat in the hall.
(iii) What is the expected duration of time a
patient spends in the clinic?
Solution
Arrival rate = l = 30 / hour
Service rate = m = 20 / hour
 Queueing Theory 215

l 17.7 MODEL 3—(MULTICHANNEL

r= 30/20 = 1.5 (r > 1)
m •
SYSTEM) M/M/C/•
Capacity of the system = N = 14
(i) An arriving patient need not wait if the In this model, instead of a single service channel,
system is empty. there are C parallel channels (C servers serving C
customers simultaneously). It is assumed that
r -1
\ Required probability = P0 = N +1 customers arrive at an average rate of l and each
r -1 server has the same mean service rate m. If there
1.5 - 1 0.5 are more than C customers in the system, all the
= = 0.001 servers will be busy and the mean service rate is
(1.5) - 1 (1.5)15 - 1
15
C m. If there are n customers where n < C then the
(ii) Since the capacity of the system is 14, an
service rate is nm and C – n servers remain idle.
arriving patient finds a vacant seat if the
Thus we find.
number of patients in the system is less than
(i) m1 = m, m2 = 2m, m3 = 3m ..., mn = nm
14. Hence the required probability is given
1<n<C
by
(ii) If n > C then C customers are served and
P0 + P1 + P2 + ... + P13
(n – C ) customers will be in the queue and
which is equal to 1 – P14. Thus the required
mn = cm n ≥ C.
probability is
(iii) Arrival rate is the same ln = l for all n
é (0.5) ù
1 - P0 r14 = 1 - ê ú (1.5)
14
17.7.1 Expression for Pn
ë (1.5) - 1 û
15
The steady state probabilities are
= 1 – 0.29 = 0.71
l
(iii) Average time spent by a patient in the clinic P1 = P
m1 0
E ( n)
WS =
l[1 - PN ] l2
P2 = P
m2 m1 0
r
E(n) = l3
(1 - r)(1 - r N +1 ) P3 = P
m3 m2 m1 0
[1 – (N + 1)rN + NrN + 1] .
(1.5) .
= × .
(0.5)[(1.5)15 - 1]
[1 – 15(1.5)14 + 14(1.5)15] ln
Pn = P
(1.5) mn mn -1 ... m3 m2 m1 0
= [1 – 15(292) + 14(438)] Thus we have
(0.5)[(437)]
3 ´ 1753 l n P0
= = 12 Pn =
n m.(n - 1).m (n - 2) m ... 3m 2 m.m
437
PN = P14 = P0r14 = (0.001)(292) = 0.29 (1 £ n < C)
12 12 l n P0
\ Ws = (1.5)(0.71) = 1.065 = 11.27 hrs =
C m.C m...C m (C - 1) m (C - 2) m...2m.m
Note: We find that WS becomes large when r > 1 (n ≥ C)
 216 Operations Research

17.7.3 Important Formulae

l n P0
Pn = (1 £ n < C) (i) Probability that an arrival has to wait, is given
n ! mn
by P(n > C)
l n P0 ¥ ¥
rn P
Pn =
C n -C
C !m n (n ≥ C) P(n ≥ C) = å Pn = å C ! C n0-C
n =C n =C

l ¥ n -C
= r we get rC æ rö
P0 å ç ÷
Taking
m =
C ! n =C è C ø
rn
Pn = P0 (1 £ n £ C) rC æ C m ö
n! .P0
C ! çè C m - l ø÷
=
r n P0
= n -C (n ≥ C) (ii) Probability that an arrival enters into service
C C!
without waiting, is given by P(n < C)
17.7.2 Expression for P0 P(n < C) = 1 – P(n ≥ C)
Total probability = 1 rC æ C m ö
¥ 1- .P0
C ! çè C m - l ÷ø
å Pn =1
n =0 (iii) Average queue length E(m)
c -1 n ¥ n
r P
å rn ! P0 + å C n -C 0C !
¥

n =0 n =C
=1 E(m) = å (n - C ) Pn
n =C
é c -1 r n ¥ ù rn
êå + å C n -C C ! ú P0 = 1 ¥
ë n =0 n ! n =C û = åK PK + C where K = n – C
K =0
é - ¥ n - C ù
c 1
r n
r C
æ rö
êå + å çè ÷ø ú P0 = 1 ¥
rC + K P0
ëê n = 0 n ! C ! n = C C ûú = å K. C !C K
K =0
é ù
ê c -1 r n rC ú rC ¥
rK r
êå +
1
ú P0 = 1 = P0 å K. Put = x
ê n =0 n ! C ! æ - r ö ú
K
C! K =0 C C
êë çè 1 ÷
C ø úû ¥
rC
(assuming that
r
< 1) =
C!
P0 å K xK
C K =0
1 ¥
rC r
P0 =
é c -1 r n r Cm ù = P0 å K x K -1
êå + . ú C! C K =1
ë n = 0 n ! C ! (C m - l ) û
Note: This result is valid only if rC +1
= P0 [1 + 2 x + 3x 2 + ... to ¥ ]
r l C.C !
< 1, or < 1 ie l < C m
C Cm rC +1
In other words, mean arrival rate must be the less = P0 (1 - x) -2
C .C !
than the maximum service rate.
 Queueing Theory 217

Solution
rC +1 1 Arrival rate = 10 per hour
= P0 2
C.C ! æ rö l = 1/6 per min
çè 1 - ÷
Cø Service time = 4 min
Service rate m = 1/4 per min
rC +1 C2 r = 2/3 and C = 2
= P0
C.C ! (C - r ) 2 P0 =
1
é r2 2m ù
rC +1 P0 ê1 + r + . ú
= ë 2 2m - l û
(C - 1) !(C - r) 2
1
rC +1m 2 .P0 =
1
=
(C - 1) !(C m - l ) 2 2 1 4 2
1+ + .
(iv) Average number of customers in the system 3 2 9 æ 1 1ö
çè 2 - 6 ÷ø
l
E(n) = E (m) +
m 1 1
= =
2 1 2
rC +1 P0 1+ +
+r 3 3
(C - 1) !(C - r) 2
1
(v) Average waiting time of a customer in the P0 =
2
queue
l 2 1 1
E (m) rC m.P0 P1 = P0 = . =
Wq = = m 3 2 3
l (C - 1) !(C m - l ) 2
(i) Probability that an arrival has to wait
(vi) Average waiting time in the system = P(n ≥ 2)
1 = 1 – P0 – P1
WS = Wq +
m 1 1 1
(vii) Average number of idle servers = 1- - =
2 3 6
Number of servers remaining idle = C – (average (ii) Expected number of idle girls
number servers in service)
l
l = C-
C – [E(n) – E(m)] = C – r = C - m
m
= 2 – (2/3) = 4/3
Example 17.11 A supermarket has two sales Also, expected number of idle girls
girls. The service time for each customer is = 2.P0 + 1.P1
4 minutes on the average and the arrival rate is
10 per hour. Find æ 1ö æ 1ö 4
= 2 ç ÷ + 1ç ÷ =
(i) the probability that an arrival has to wait. è 2ø è 3ø 3
(ii) the expected percentage of idle time for each \ Probability that a girl is idle
girl.
(iii) the expected waiting time of a customer in Expected number of idle girls
=
the system. Total number of girls
 218 Operations Research

4/3 2 25 3 32 3 4
= = = 0.67 . = . =
2 3 =
3!(4 - 2) 2
23 6.4 23 23
\ Percentage of idle time for each girl
= 67%
(iii) Average queue length = E(m) 17.8 MODEL 4—(MULTICHANNEL
Expected waiting time in the queue SYSTEM WITH LIMITED
CAPACITY) M/M/C/N
E ( m)
Wq =
l In this model the maximum number of customers
is limited to N and there are C service channels.
rC m P0 Therefore we have
=
(C - 1) !(C m - l ) 2 ì l 0£ n£ N
ln = í
(2 / 3) 2 .1 / 4.1 / 2 1 î0 n > N
= =
1!(1 / 3) 2 2 ì n m 0£ n£ C
mn = í C m C £ n £ N
Average waiting time in the system î
æ 1ö
1
WS = Wq + = ç ÷ + 4 = 4.5 min ì rn
m è 2ø ï P0 1£ n £C
Example 17.12 There are 4 counters at a check ï n!
post for verifying the passport of the tourists. The ïï r n P0
Pn = í n -C
C £n£ N
arrival rate of the tourists is l and the service rate ï C !C
has an average l/2. Find the expected queue length ï0 n> N
at the check post. ï
Solution ïî
Given that arrival rate = l
Service rate = l/2 17.8.1 Expression for P0
l
\r = = 2 C=4 N C -1 N
l/2 We have å Pn = å Pn + å Pn =1
n =0 n =0 n =C
1
P0 = C -1 (total probability)
n
æ C
ö
å rn ! + rC ! çè C mC m- l ø÷ C -1 n N n
n =0
i.e. å rn ! P0 + å C ! rC n -C P0 = 1
n! n =C
1
=
r 2 r3 r 4 æ 4.l / 2 ö é C -1 r n rC N r n - C ù
1+ r + + + ç ÷ P0 ê å + å n -C ú = 1
2 6 4! è 4.l / 2 - l ø ë n = 0 n ! C ! n =C C û
1 n -C n -C

å æçè Cr ö÷ø
= = 3/23 N N
1 + 2 + 2 + 4 / 3 + 16 / 24(2 / 1) Now å Cr n -C =
Expected queue length n =C n =C

N -C
rC +1 P0 r æ rö æ Pö
2

E(m) = =1+ + ç ÷ + ... + ç ÷

(C - 1)(C - r) 2 C èCø èCø
 Queueing Theory 219

N - C +1 N -C
æ rö rC P0 æ r ö
1- ç ÷
èCø = ç ÷
C! èCø
å K x K -1
= k =0
r
1- æ r ö
C çè x = C , K = n - C ÷ø
[Sum of N – C + 1 terms of the GP]
rC P0 r d æ N - C K ö
= C ! C dx ç å x ÷
1
\ P0 = é N - C +1 ù è k =0 ø
æ rö
C -1 n c ê 1 - çè ÷ø ú
å rn ! + Cr ! C ! êê C r úú rC P0 r d æ 1 - x K +1 ö
C ! C dx çè 1 - x ÷ø
= K=N–C
n =0
ê 1- ú
êë C úû
rC P0 r
ær ö = .
çè C ¹ 1÷ø
C! C
é (1 - x)( K + 1) x K (-1) - (1 - x K +1 )(-1) ù
r ê ú
Here need not be less than 1. ë (1 - x)2 û
C
rC P0 r
r E(m) =
If r = C then =1 C! C
C
n -C
é 1 - x K +1 - (1 - x)( K + 1) x K ù
ê ú
å æçè C ö÷ø
N
r
= N -C +1 ë (1 - x)2 û
n =C
r
1 where x = and K = N – C
P0 = C -1 n C
C
å rn ! + rC ! ( N - C + 1) (ii) Expected number of customers in the system
n =0
l
E(n) = E (m) + (1 - PN )
17.8.2 Important Formula m
(i) Expected number of customers in the queue (iii) Expected waiting time in the system
N E (n)
E(m) = å (n - C ) Pn WS =
l (1 - PN )
n =C
(iv) Expected waiting time in the queue
N
r n P0
= å (n - C )
C ! C n -C Wq = Ws -
1
=
E ( m)
n =C m l (1 - PN )
n -C Example 17.13 There are three stalls at an
rC P0
å (n - C ) æçè C ö÷ø
N
r
= automobile inspection centre. The centre can
C! n =C accommodate seven cars at the maximum (three
n - C -1 in service and four in waiting). On the average
rC P0 æ r ö N æ rö
ç ÷ å (n - C ) çè ÷ø
one car arrives every minute and the service time
=
C ! è C ø n =C C is 6 minutes. Find (i) the average number of cars
 220 Operations Research

in the system (ii) average waiting time in the r

system. x= = 2 K = N -C = 4
C
Solution Given
l = 1 per min 63 1 é1 - 25 - (1 - 2)(5)24 ù
m = 1/6 per min \E(m) = . .2. ê ú
C=3
3! 1141 ë (1 - 2)2 û
N=7
216 1 (1 - 32 + 80)
l = . .2.
r= =6 6 1141 1
m
1 72.49
P0 = = = 3.09
é 1141
æ rö ù
5

2 ê 1 - çè ÷ø ú = 3.09 = 3 cars (nearly)

r n r3 ê
å n ! + 3! ê 3r úú \ Expected number of cars in the system
n =0
ê 1- 3 ú l
ëê ûú E(n) = E (m) + (1 - PN )
m
1 1
P0 = = = 3.09 + 6(1 – PN)
36 216 æ 1 - 32 ö 1141
1+ 6 + + ç ÷
2 6 è 1- 2 ø r N P0 67 1
Now PN = N -C = = = 0.5
(i) Expected number of cars in the queue C!C 6.3 4
1141
C
r r \ E(n) = 3.09 + 6 (1 – 0.5) = 6 cars (nearly)
E(m) = .P0 .
C! C (ii) Average waiting time in the system
é 1 - x K +1 - (1 - x)( K + 1) x K ù E (n) 6
ê ú WS = = = 12 min
(1 - x)2 l (1 - PN ) 1(1 - 0.5)
ë û

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Arrivals at a telephone booth are poisson with
an average inter-arrival time of 10 minutes.
The length of a phone call is exponential with
mean 3 minutes.
(i) Find the probability that a person
arriving at the booth will have to wait.
(ii) Find the average length of the queue.
(iii) Find the total time a person spends in
the booth.
2. In a single server queueing system the arrival
rate is 25 per hour. What must be the service
rate in order that a customer has to wait for
not more than 10 minutes to enter into service.
3. In a repair shop having a single mechanic,
customers arrive at the rate of 4 per hour. It
takes the mechanic 6 minutes on the average
to inspect one item. Find the
(i) proportion of time the mechanic is idle.
(ii) probability that there is at least one
customer.
(iii) average number of customers in the
shop.
(iv) average time spent by a customer in the
shop.
4. Customers arrive at a box office window with
a single server, at a mean rate of 30 per hour.

The time required to serve a customer is
90 seconds on the average. Find the average
waiting time of a customer and the average
number of customers in the system.
5. Customers arrive at a bank cash counter in a
poisson manner at an average rate of 30 per
hour. The cashier takes on an average
1.5 minutes for each cash cheque. Calculate
(i) the percentage of time the cashier is busy.
(ii) the average waiting time of a customer
in the bank.
6. Trucks arrive at a factory for collecting
furnished goods, at the rate of 10 per hour
and the loading rate is 15 per hour.
Transporters complain that their trucks have
to wait for nearly 2 hours at the factory.
Examine whether their complaint is justified.
What is the probability that the loaders are
idle.
7. Customers arrive at a one — window drive -
in bank at the rate of 10 per hour. Average
service time is 5 minutes. The space in front
of the window can accommodate a maximum
of 3 cars including the one under service.
(i) Find the probability that an arriving car
can directly drive to the space in front
of the window for service.
(ii) What is the probability that an arriving
can will have to wait outside the
indicated space?
(iii) How long is an arriving customer
expected to wait before starting service?
8. Letters arrive at the rate of 5 per hour, to a
typist’s table during an 8 hour day. The typing
rate is on the average 8 letters per hour. If
the time of the typist is valued at Rs 10 per
hour determine
(i) equipment utilization
(ii) the period of time, an arriving letter has
to wait
(iii) average cost due to waiting and
operating the typewriter.
9. In a public telephone both the arrivals are on
the average 15 per hour. A call on the average
takes 3 minutes. If there is only one phone in
the booth. Find
Queueing Theory 221
(i) the queue length at any time.
(ii) the proportion of time, the booth is idle.
10. In a petrol pump cars arrive at an average
rate of 10 per hour. If there is only one serving
unit and the service time has an average of
3 minutes, find
(i) the average number of cars in the system.
(ii) the average waiting time in the queue.
(iii) the probability that there are two cars in
the system.
11. At a one-man saloon customers arrive at the
rate of 5 per hour and the hair cutting time is
on the average 10 minutes per customer. Find
(i) the average number of customers in the
saloon.
(ii) the average number of customers waiting.
(iii) the probability that a new arrival does
not wait for service.
12. A single server coffee shop has space to
accommodate only 10 customers. If a new
arrival finds that the shop is full he goes to
another shop. The arrival rate is 10 per hour
and the service time is 5 minutes per
customer Find P0 and Pn.
13. Customers arrive at a photocopier shop at an
average rate of 5 per hour and they are served
at 6 per hour. There are only 5 seats available
for the customers. Find the average number
of customers and the average time a customer
spends at the shop.
14. An insurance company has three claim
adjusters. People with claims arrive at the
office at the rate of 20 per 8 hour day. The
mean service time is 40 minutes per claimant.
(i) How many hours per week an adjuster
is busy?
(ii) How much time a claimant spends in the
office?
15. Given an average arrival rate of 20 per hour
there are two options for a customer: A single
channel with service rate 22 customers per
hour or a two-server channel with service rate
of 11 customers per hour. Determine which
is a better option.
16. A tax consulting firm has 4 service counters
On the average 80 persons arrive during a
 222 Operations Research
day of 8 hours. The average service time per
person is 20 minutes. Find
(i) the average number of customers in the
system.
(ii) the average time a customer spends in
the system.
(iii) the expected number of idle servers.
(iv) the number of hours a server spends with
customers during one week.
17. A barber shop has two barbers and three
chairs for waiting customers. If a customer
finds that there is no vacant chair he will
leave the shop. Customers arrive at the rate
of 5 per hour and the service time is
15 minutes per customer. Find the expected
number of customers in the shop. Also find
the percentage of idle time for each server.
18. In a supermarket the average arrival rate of a
customer is 10 per 30 minutes. On the
average it takes 2.5 minutes to serve a
customer. Find the probability that there are
more than 6 customers in the queue.
19. In a telephone booth people arrive at the rate
of 15 per hour. The duration of a call is
3 minutes on the average. Find
(i) The expected number of customers in the
booth.
(ii) The proportion of time, the booth is idle.
20. At a telephone booth, the arrivals follow
poison distribution with an inter arrival time
of 12 min. The duration of a call is 4 minutes
on the average
(i) What is the probability that a fresh
arrival will not have to wait for the
phone?
(ii) What is the probability that an arrival
has to wait for more than 10 min?
(iii) What is the average length of the queue?
21. People arrive at a theatre ticket booth at the
rate of 25 per hour. Service time for a
customer has an average of 2 min. Find the
following:
(i) Mean number in the waiting line
(ii) Mean waiting time
(iii) Utilisation factor
22. In a single server bank, working from 7 am
to 1 pm, the average number of customers is
54 per day. Average service time is 5 min
per customer. Find the
(i) Average number of customers in the
system
(ii) Average number in the queue
(iii) Average waiting time.
23. In a single server tools store operators arrive
at the rate of 10 per hour. Service time is
3 minutes per operator. The rate of production
is 100 units per day of 8 hours. If the cost
of product is Rs 200, find the loss due to
waiting.
24. In a telephone exchange there are two
operators. Calls arrive at the rate of 15 per
hour and the average length of a call is
5 minutes. Find the
(i) probability that a customer has to wait.
(ii) average waiting time of a customer.
25. A petrol station has two pumps. Cars arrive
at the rate of 10 per hour and the average
service time is 4 minutes per car. Find
(i) the probability that a customer has to
wait.
(ii) the percentage of idle time.
26. There are four counters on the frontier of a
country to check passports and other
documents of tourists. The arrival rate is l
and the service rate is l/2. What is the
average queue length at each counter?
27. The mean arrival rate at a railway reser-
vation centre is 24 per hour and there are
3 counters
(i) Find the probability that there are no cus-
tomers if the mean service rate is
10/hr.
(ii) the average time spent by a customer in
the system if there are 5 counters
28. In a tax consulting company there are four
service counters. On the average 80 persons
arrive during an 8 hour service day. The
average service time is 20 minutes per
customer. Find the
(i) average number of customers in the
system.
 Queueing Theory 223

(ii) average waiting time of a customer in (iii) number of hours , each consultant spends
the system. with the customers, per week.

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. (i) 0.3 (ii) 9/70 (iii) 30/7 For two server channel Wq = 0.43 hrs
1 Hence two server channel is better.
2. m ≥ 16. (i) 7 (ii) 40 min (iii) 1 (iv) 33 hrs
2
3. (i) 3/5 (ii) 21/25 (iii) 2/3 (iv) 10 min 17. E(n) = 3, Percentage of idle time = 37%
4. Wq = 9/2min, E(n) = 3 18. Required probability = 0.3348
5. (i) 75% (ii) Ws = 6min 19. Average number of customers = 4
6. Ws = 12 min. Complaint is not justified. Proportion of idle time = 1/4
Probability that the loaders are idle = 1/3 20. (i) 0.67 (ii) 0.63 (iii) 1.5
7. (i) 91/216 (ii) 125/216 (iii) 25 min 21. (i) 4 (ii) 9.6 min (iii) 0.833
8. (i) 62.5% of time (ii) 12.5 min (iii) Rs 133.30 22. (i) 3 (ii) 2.25 (iii) 20 min
23. Average waiting time is 2/5 hour. Loss due
1 to waiting
9. (i) Lq = 2 (ii)
4 = (2 / 5) (100 / 8) 200 = Rs 1000
10. (i) 1 (ii) 3 min (iii) 1/8 24. (i) 0.48 (ii) 3.2 min
11. (i) 5 (ii) 4 (iii) 1/6 25. (i) 0.167 (ii) 67%
12. P0 = 1/6, Pn = 1/6(5/6)n 26. Lq = 4/23
13. E(n) = 5, Ws = 6 27. (i) P0 = 0.019 (ii) 6 min
14. (i) 22 hrs during a 5-day week (ii) 50min 28. (i) 6.61 (ii) 0.66 hours
15. For a single channel, Wq = 0.45 hr (iii) 33.3 hours

Inventory Models
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
18.1 INTRODUCTION
Inventory is the stock of raw materials and goods
required for production in a factory or finished
goods for sales. If an order for a product is received
we should have sufficient stock of materials
required for manufacturing the item in order to
avoid delay in production and supply. Also there
should not be over stock of materials and goods
as it involves storage cost and wastage in storing.
Therefore inventory control is essential to promote
business. Maintaining inventory helps to run the
business smoothly and efficiently and also to
provide adequate service to the customer.
Inventory control is very useful to reduce the cost
of transportation and storage. The basic questions
are
(i) When should an order for materials be
placed?
(ii) How much to be produced during each
period or how much to be ordered each time?
Inventory control helps us to solve these
problems to a great extent.
Also inventory control is necessary to ensure
adequate and prompt supply of items to the
18
customers and avoid shortage of stock and
consequent loss.
18.2 COSTS INVOLVED IN
INVENTORY CONTROL
Various costs incurred in the maintenance of
inventory are as follows:
(i) Holding cost (Storage cost) This is the
cost associated with holding (carrying) the
inventory in the godown. It includes rent for
godown, interest for the money locked up,
insurance premium, salaries for the
watchmen, deterioration (damage) cost etc.
Carrying cost is calculated in two ways.
(a) Carrying cost = (Holding cost per unit
for unit time) × (Average number of units
in stock)
(b) Carrying cost = (Cost of storing stock
of one rupee worth) × (Rupee value of
the units stored)
Carrying cost (storage cost) is denoted by C1.
(ii) Set-up cost (Ordering cost) Ordering cost
is associated with the cost of placing orders
for procurement of materials or finished

goods from suppliers. In the case of
production, setup cost includes the cost
involved in setting up the machines for
production run. Ordering cost includes the
cost of stationery, postage, telephones,
travelling expenses, handling of materials etc.
Ordering cost = (Cost per order or per set up)
× (Number of orders or production runs)
It is denoted by CS
(iii) Production cost (Purchase cost)
Production cost per unit item depends upon
the length of production runs. For long pro-
duction runs it is lower. Also purchase cost
is less for orders of large quantity due to
quantity discounts or price breaks.
(iv) Shortage cost (Stock out cost) If the
inventory on hand is not sufficient to meet
with the demand of materials or finished
goods, then it results in shortage of supply.
There are two cases: (i) Orders are kept
pending and as soon as stock is replenished
back orders are supplied. (ii) Unfilled
demand (order) results in lost sales and
cancellation of order. We consider only the
first case. Shortage cost per unit, per unit
time of delay is denoted by C2.
(v) Total inventory cost The total inventory
cost is denoted by C.
C = ordering cost + Holding cost + Shortage
cost. If price discounts are available on
purchase cost then purchase cost is also to
be included in the total cost.
18.3 CHARACTERISTICS OF
INVENTORY SYSTEM
(i) Demand The nature of demand must be
known for inventory control. Demand may
be deterministic or probabilistic. In the
deterministic case the quantity of material
required over a period of time is certain and
known before hand. The demand may be
uniform in many cases. But in the
probabilistic case the nature of demand is
uncertain and is given by some probability
distribution.
Inventory Models 225
(ii) Order cycle The period of time between
two consecutive placement of orders is called
order cycle.
(iii) Lead time When an order is placed, the
supplier takes some time to supply the item.
Lead time is the time between the placement
of an order and the receipt of the material.
(iv) Replenishment of stock Replenishment of
stock may occur instantaneously without
lead time or uniform replenishment may
occur when the goods are manufactured by
the factory itself.
(v) Planning horizon (period) The duration
of time over which a particular inventory
level will be maintained is called planning
horizon.
18.4 DETERMINISTIC
INVENTORY MODELS
WITH NO SHORTAGE
18.4.1 Economic Lot Size Model
with Constant Demand
In this model demand is assumed to be constant
or completely pre-determined. Our problem is to
determine the quantity to be ordered in the most
economical manner. This quantity is called
Economic Order Quantity (EOQ). We make the
following assumptions:
(i) Demand is known and is uniform. Q denotes
the lot size in each production run.
(ii) D denotes the total number of units produced
or supplied per unit period.
(iii) Shortages are not allowed. As soon as the
level of inventory reaches zero, the inventory
is replenished.
(iv) Production or supply of materials is
instantaneous. In other words, replenishment
of inventory is instantaneous without delay.
(v) Lead time is zero.
(vi) Set up cost per production run is CS.
(vii) Holding cost (storage cost) is C1. If the
holding cost is expressed as a percentage
I of the unit cost C of the commodity then
C1 = IC.
 226 Operations Research

Formula for EOQ (Wilson Harris Thus the total annual inventory cost is given by
Formula) CA = Total holding cost + Total set-up cost
Let us assume that the total period (say one year) 1 D
is divided into n parts each of length t. Let Q be = QC1 + CS
the quantity produced during each interval (run). 2 Q
Therefore the total quantity produced during n runs We want to determine Q such that C A is
is D = nQ. Also nt = 1. The following figure minimum. By the principle of maxima minima the
illustrates the model. dC A d 2C A
condition is = 0 and >0
dQ dQ 2
1 D
Now CA = QC1 + CS
2 Q
dC A 1  D 
= C1 +  − 2  CS
dQ 2  Q 
C1 DCS  d 2C A 
− 2  obviously 2
> 0 
2 Q  dQ 
If the demand rate is R then a quantity Q = Rt
must be produced in each run. The stock in a small dC A
= 0 fi C1Q2 = 2DCS
time interval dt is Rt dt. The total stock during the dQ
period t is given by
t 2 DCS
t2 1 1 \ Q=
ò Rt dt = R
2
= Rt.t = Q.t
2 2
C1
0 The optimum quantity is denoted by Q* which
1 / 2 Qt 1 is called Economic Order Quantity (EOQ)
\ Average stock = = Q.
t 2
2 DCS
1 \ EOQ = Q* =
[Area of the triangle OAP = Qt C1
2
1 This is called the Wilson Harris formula for
\ Total stock = Qt EOQ.
2
\ Average stock = 1/2 Q] Note: If the holding cost is expressed in terms of
1 the value of the stock held then C1 = IC and the
Holding cost per run = Q.C1.t value of Q* = Q*C.
2
1 1 2 DCS 2 DCS C
= Q.C1. = .C =
2 n IC I
1 1 The optimum annual inventory cost is
Annual holding cost = QC1 .n 1 D
2 n CA* = Q*C1 + * CS
(since there are n runs) 2 Q
1 1 2 DCS D
= QC1 = .C1 + CS
2 2 C1 2 DCS
Set-up cost per run = CS
Annual set-up cost = n .CS C1
D 1 1
= CS (Q D = nQ) = 2 DCS C1 + 2 DCS C1
Q 2 2
 Inventory Models 227

= 2 DCS C1 = 800 units

D 3200
Optimal period of one run (order cycle) is (ii) Number of orders n = * = =4
Q 800
Q*
t* = (iii) Time between two consecutive orders
D
D Q* 800 1
Number of runs per year is n = * = t* = = = = 3 months
Q D 3200 4
(iv) Total annual cost
Example 18.1 A factory requires 3600 kg of raw
material for producing an item per year. The cost = material cost + inventory cost
of placing an order is Rs 36 and the holding cost =D×C+ 2 DCS C1
of the stock is Rs 2.50 per kg per year. Determine
the EOQ. = 3200 × 6 + 2 × 3200 × 150 × 1.5
= 19200 + 1200
Solution We are given that = Rs 20400
D = 3600 kg per year
CS = Rs 36 per order Example 18.3 A manufacturer has to supply
C1 = Rs 2.50 per kg per year 12000 units of a product per year to his customer.
Shortages are not permitted and there is no lead
2 DCS
\ Now EOQ = Q* = time. The inventory holding cost is Rs 0.20 per
C1 unit per month and the set-up cost per run is
2 × 3600 × 36 Rs 350. Determine
= = 322 kg (i) the economic lot size
2.5
(ii) the period of one run
Q* 322
Period of one run t* = = (iii) the minimum annual inventory cost
D 3600
= 0.089 year Solution We have D = 12000 units per year
= 32 days = 1000 units per month
CS = Rs 350
D
Number of runs per year n = = 11.4 C1 = Rs 0.20 per month
Q*
2 DCS 2 × 1000 × 350
Example 18.2 The annual demand for an item (i) Q* = =
C1 0.20
is 3200 units. The unit cost is Rs 6. The inventory
= 1870 units
carrying cost is 25% per annum per unit. The cost
of one procurement is Rs 150. Determine Q* 1870
(ii) Period of one run t* = =
(i) EOQ D 1000
(ii) Number of orders per year = 1.87 months
(iii) Time between two consecutive orders = 56 days
(iv) Total annual cost (iii) Annual inventory cost
Solution Given D = 3200, C = 6, I = 0.25 CA = 2 DCS C1 × 12
C1 = IC = 1.5
= 2 × 1000 × 350 × (0.2) × 12
CS = 150
= 374 × 12 = Rs 4490
2 DCS 2 × 3200 × 150
(i) EOQ = Q* = = Example 18.4 Annual requirement of machine
C1 1.5
parts for a company is 9000 units and the cost of
 228 Operations Research

one unit is Rs 20. It costs Rs 15 for placing an Q*

order and the storage cost is 15% of the value per Length of a cycle is t* = = 1000/100
year. The company is at present placing orders D
every month. How much would it save if it places = 10 days
order as per the optimal period?
Solution
Given that annual requirement D = 9000
Ordering cost CS = Rs 15
Storage cost C1 = (15/100) × 20 = Rs 3 per year Lead time is 13 days. Hence an order is to be
2 DCS 2 × 9000 × 15 placed three days before the stock exhausts. There
\ EOQ = Q* = = must be sufficient stock for 3 days.
C1 3
\ Re order level is 3 × 100 = 300 units
= 300 units
Example 18.6 A company has to supply 100
Q* 300
t* = = year = 12 days units of an item to the customers every week. The
D 9000 item is purchased from the supplier at Rs 60 per
Annual inventory cost CA = 2 DCS C1 unit. The cost of ordering and procurement from
the supplier is Rs 150 per order. The storage cost
= 2 × 9000 × 15 × 3 = Rs 900
is 15% of the cost per year. Find the economic lot
\ The optimal period of one run is 12 days
size and the optimum cost per week.
and the minimum annual inventory cost is Rs 900.
The company is placing orders every month. Solution
\ t = 30 days We have D = 100 units per week
Ordering cost per year = 12 × 15 = 180 CS = Rs 150
Monthly requirement = 9000/12 C1 = 15% per year of the cost
= 750 units = 15/100 × 60
Average inventory = 1/2(750) = 375 = Rs 9 per unit per year (52 weeks)
Storage cost = 375 × 3 = Rs 1125
= Rs 9/52 per unit per week
\ Annual inventory cost = 1125 + 180
= Rs 1305 2 DCS 2 × 100 × 150
By ordering for 300 parts after every 12 days (i) \ Q* = =
the annual cost is Rs 900. Thus there is a saving C1 9 / 52
of Rs 405 per year. = 416 units
(ii) Optimum total cost
Example 18.5 The demand for an item is 100
units per day. For placing an order a cost of = Purchase cost + Inventory cost
Rs 400 is incurred. Holding cost is Rs 0.08 per = 100 × 60 + 2 DCS C1
day. If the lead time is 13 days determine the EOQ
and the re order point. = 6000 + 2 × 100 × 150 × 9 / 52
= 6000 + 72 = Rs 6072
Solution Given D = Rs 100 per day
CS = Rs 400 Example 18.7 The average daily requirement of
C1 = 0.08 per day an electrical appliance is 120 units and the firm
has 250 working days a year. The manufacturing
2 DCS 2 × 100 × 400 cost is Rs 0.50 per part. Insurance, tax, interest
\ Q* = =
C1 0.08 etc., come to 20% of the unit cost. The set-up cost
= 1000 units is Rs 50 per run. Determine the EOQ.
 Inventory Models 229

Solution demand rate (consumption rate) is r units per unit

Here D = 120 × 250 = 30000 units time (k > r). Each production run of length t is
CS = 50 divided into parts t1 and t2 such that
C1 = 20% of the cost per part (i) the inventory is building up at a constant rate
= 20/100 × 0.50 = 0.1 of (k – r) units per unit time during t1
2 DCS 2 × 30000 × 50 (ii) there is no production during t2 and the
EOQ = = inventory is decreasing at the rate of r units
C1 0.1
per unit time (t1 + t2 = t).
= 5477 parts
Example 18.8 The demand for a commodity is
600 units per year. The ordering cost is Rs 80. The
cost of the item is Rs 3 and the inventory holding
cost is 20% of the cost per year. If the lead time is
one year find
(i) The EOQ
(ii) The re-order point
(iii) Te minimum average annual cost
This model is illustrated in the above figure.
Solution At the end of time t1 let the inventory level be S.
D = 600 per year \ S = (k – r) t1. This inventory is used during the
CS = Rs 80 period t2 at the rate of r \ S = rt2.
C1 = IC = 20/100 × 3 = 0.60 per year
S S
2 DCS 2.600.80 \ t1 = and t2 =
(i) Q* = = = 400 units k−r r
C1 0.60 Let Q be the quantity produced per run, then
Q* Q = k t1
(ii) t* = = 400/600 = 2/3 year \ S = (k – r)t1 = kt1 – rt1 = Q – rt1
D
Lead time = 1 year Q-S
\ rt1 = Q – S fi t1 =
1 yr 1 yr
r
Q
Also t1 =
3 k
0 1/3 1/3 t 1/3 1/3 2t 2/3 3t Q Q-S Qr
Order should be placed after 1/3 year when \ = fi Q–S=
k r K
the stock level is 1/3 × 600 = 200 units
Q
\ A quantity of 400 units is to be ordered \ S = Q – r t1 = Q - r
when the inventory level is 200 units k
(iii) Minimum annual inventory cost æ k - rö
=ç
è k ÷ø
Q
= 2 DCS C1
Now, total inventory during time t
= 2.600.80(0.60) = Rs 240 = 1/2(t1 + t2)S = 1/2 St
18.4.2 Manufacturing Model 1 St S
Average inventory = =
In this model the assumptions are same as in model 2 t 2
1 except that the replenishment rate (manufactur- 1
Inventory holding cost = SC1
ing rate) is finite, say k units per unit time and 2
 230 Operations Research

1 Q k −r
= (k - r ) C1 2CS C1r 
2 k = 
 k 
CS
Set-up cost = n.Cs = 1 rC1  k − r 
t No. of runs per year n = *
=  
t 2CS  k 
S S Sk
Now t = t1 + t2 = = +
k-r r r (k - r ) Example 18.9 A contractor has to supply 10000
(k - r ) k Q bearings per day to an automobile manufacturer.
= Q. = He can produce 25000 bearings per day. The
k r (k - r ) r
holding cost is Rs 2 per year and the set-up cost is
Thus total inventory cost Rs 180. How frequently should the production run
1 æ k - rö r be made?
CA = ç
2è k ø
÷ QC1 + Q CS
Solution
dC A d 2C A We are given r = 10000, k = 25000
For CA to be minimum = 0 and >0
dQ dQ 2 CS = 180, CS = Rs 2 per year
dC A 1 k−r r = 0.0055 per day
Now =   C1 − 2 CS = 0
dQ 2  k  Q 2CS  kr 
EOQ = Q* =  
r 1k−r rC1  k − r 
⇒ 2
CS =   C1
Q 2 k  2.180 25000.10000
= .
C  kr  0.0055 15000
Q* = 2 S  
C1  k − r  = 33029 units
d 2C A Q* 33029
Obviously 2
> 0 \ CA minimum for Q* Period of one run t* = = = 3.3 days
dQ r 10000
2CS  kr  Example 18.10 A factory manufactures a
\ EOQ =  
C1  k − r  product at the rate of 100 units per day and the
daily demand is 40 units. The cost of one unit is
Q* Rs 5. Holding cost is 25% of the value per unit
\ Optimal period of a run t* =
r per year. Setup cost is Rs 150 per run. Determine
2CS  k  the economic lot size and the number of runs per
=   year. (Take 250 days for one year.)
rC1  k − r 
Total annual inventory cost Solution
Hence k = 100 per day
1k −r * r r = 40 per day
CA* =   Q C1 + * CS
2 k  Q C1 = 5 × 25/100 = 1.25 per year
1.25
1k−r 2CS  kr  = = 0.005 per day
=     C1 250
2 k  C1  k − r  CS = Rs 150
C1 (k − r ) 2CS  k 
+ r. CS Q* =  
2C1kr rC1  k − r 
 Inventory Models 231

During each run of 40 days, production

2.150  100.40 
=   continues for 20 days and there would be no
0.005  60  production for the remaining 20 days.
= 2000 units
Q*
18.4.3 Model with Price Breaks
2000
t* =
= = 50 days (Quantity Discounts)
r 40
250 In the previous models the cost of the material
Number of runs = =5 (purchase cost) was not affected by the order size
50 since it remains constant. Often we find that
Q* 2000 quantity discounts are offered to encourage the
t1 = = = 20 days
k 100 buyers to purchase more units of an item. There
(production period) may be quantity discount or price breaks. In such
\ t 2 = 30 days situations the variable purchase cost is also
included in determining the EOQ. Let us assume
Example 18.11 An item is produced at the rate that the cost per unit commodity is p1 if Q < q
of 50 units per day. The demand is at the rate of and p 2 if Q ≥ q where p1 > p2. Let the ordering
25 units per day. The set-up cost is Rs 100 and cost be CS and the holding cost be C1. If the
the holding cost is Rs 0.01 per unit per day. Find demand is D then the number of orders n is given
the EOQ and the minimum annual inventory cost. by n = D/Q
After how many days should production be The total annual cost is
stopped during each run?
D 1
Solution Given k = 50 units per day TC1 = Dp1 + CS + QC1 for Q < q
Q 2
r = 25 per day
Cs = Rs 100 D 1
TC2 = Dp2 + CS + QC1 for Q ≥ q
C1 = 0.01 per year Q 2
2CS  kr  Neglecting the price break, we get the EOQ
(i) EOQ = Q =
*
 
C1  k − r  2 DCS
Q* = ...(1)
C1
2 × 100  50 × 25 
= ×  = 1000 units Choose the quantity q1 such that
0.01  50 − 25 
TC1(Q*) = TC2(q1) ...(2)
k −r The curves representing TC1 and TC2 are given
Inventory cost = 2CS C1r  
 k  below:
25
= 2.100.(0.01)25.
50
= Rs 5 per day
(ii) Annual inventory cost CA* = 5 × 365
= Rs 1825
Q* 1000
(iii) t* = = = 40 days
r 25
Q* 1000
t1 = = = 20 days
k 50
\ t2 = 20 days
 232 Operations Research

The optimum quantity Qo depends on the value \ Q* < q < q1 (II zone)
of q (price break point). There are three zones. Qo = q = 15
0 < q < Q* (I zone); Q* < q < q1 (II zone) q ≥ q1 \ EOQ = 15
(III zone)
Equation (1) gives the value of Q* and using Example 18.13 The price break of a product is
(2) we can find q1. We can determine Qo as follows  Rs 10 q < 500
Qo = Q* if 0 < q < Q* (I zone) Unit cost =  Rs 9.25 q ≥ 500

= q if Q* < q < q1 (II zone) Monthly demand is 200 units. Storage cost is
= Q* if q ≥ q1 (III zone) 2% of the unit cost per month. Find the EOQ if
Example 18.12 The cost of a commodity is (i) ordering cost is Rs 350
Rs 2 for an order less than 15 and Re 1 for orders (ii) ordering cost is Rs 100
greater than or equal to 15. Holding cost per unit Solution
is Re 1 and setup cost is Rs 10 per run and the
requirement is 5 units per day. Find the EOQ. (i) Ordering cost CS = Rs 350
C1 = 10(0.02) = 0.2
Solution Given p1 = 2, p2 = 1
D = 200
D = 5, q = 15
CS = 10, C1 = 1 2 DCS 2.200.350
Q* = = = 837
Ignoring the price break we have C1 0.2
2 DCS q = 500 q < Q*
2.5.10
Q* = = = 10 units Hence Qo = Q* = 837
C1 1
(ii) CS = 100 C1 = 0.2 D = 200 q = 500
Since q = 15, we find that Q* < q
To find q1 p1 = 10 p1 = 9.25

D Q* 2 DCS 2.200.100
TC1(Q*) = Dp1 + CS + C1 Q* = = = 447
Q* 2 C1 0.2

5 × 10 10 \ Q* < q
= 10 + + .1 = 20 To find q1
10 2
D Q*
D
TC2(Q1) = Dp2 + CS + 1 .1
q TC1(Q* ) = Dp1 + CS + C1
q1 2 Q* 2

5.10 q1 200 447

= 5+ + = 200 × 10 + × 100 + × (0.2)
q1 2 447 2
Take *
TC1(Q ) = TC2(q1) = 2000 + 44.7 + 44.7 = 2089
D q
20 = 5 +
50 q1
+ TC2(q1) = Dp2 CS + 1 .C1
q1 2 q1 2
30q1 = 100 + q12 = 200(9.25) +
200 q
.100 + 1 . (0.185)
q12 – 30q1 + 100 = 0 fi q1 = 26.18 or 3.82 q1 2
By definition q1 is taken as the larger value [C1 = (9.25)(0.2) = 0.185]
\ q1 = 26.18 20000
= 1850 + + q1 (0.09)
Q* = 10 q = 15 q1 = 26.18 q1
 Inventory Models 233

20000 III Zone (q ≥ q1) Qo = Q*

Take 2089 = 1850 + + 0.09q1
q1
0.09q12 – 239q1 + 20000 = 0
239 ± 223
q1 = = 2567 or 88.9
0.18
Taking q1 = 2567 we find that q < q1
\ Q* < q < q1 (II zone)
Qo = q = 500
Note: The following figures illustrate the different
cases.
I Zone (0 < q < Q*) Qo = Q*

Example 18.14 The annual demand of a product

is 10000 units. Each unit costs Rs 100 for orders
below 200 units and Rs 95 for orders of 200 and
above. The annual inventory holding cost is 10%
of the value and the ordering cost is Rs 5 per order.
Find the EOQ.
Solution Given D = 10000 per year
CS = 5
10
C1 = 10% = 100 · = 10
100
p1 = 100, p2 = 95, q = 200
2 DCS 2.10000.5
Q* = = = 100 units
II Zone (Q* < q < q1) Qo = q C1 10
q = 200 \ Q* < q
To find q1
D Q*
TC1(Q*) = Dp1 + CS + C1
Q* 2
10000 100
= 10000 × 100 + ×5+ × 10
100 2
= 1001000
D q
TC2(q1) = Dp2 + CS + 1 C1
q1 2
10000 q
= 10000 × 95 + × 5 + 1 × 9.5
q1 2
 234 Operations Research

50000 19 Take q1 = 342.7

= 950000 + + q1
q1 4 Thus Q* < q < q1 (II zone)
50000 19 Qo = q = 100
Take 1001000 = 950000 + + q1 Monthly optimum inventory cost is
q1 4
19q12 – 204000q1 + 200000 = 0 500 100
TC2(100) = 500 × 95 + × 50 + × 19
q1 = 0.97 or 10736 100 2
Taking q1 = 10736 we get Q* < q < q1 = 47500 + 250 + 950
= Rs 48700
(II zone)
Qo = q = 200 18.4.4 Model with Shortages
Example 18.15 The monthly demand of an item This model is based on all the assumptions of the
is 500. The cost of the commodity is Rs 100 for I model except that the inventory system runs out
orders less than 100 and Rs 95 for orders above of stock for a certain period of time. In other words
100 units. The holding cost is 20% of the value shortages are allowed. Customers leave the orders
per month. The ordering cost is Rs 50. Find the with the supplier and this back order is supplied
EOQ and also the optimum monthly inventory as soon as the stock is replenished. It involves
cost. shortage cost. Shortage cost depends upon how
long the customer waits to receive supply. It is
Solution Given D = 500, CS = 50, expressed in rupees per unit of time of delay. Let
p1 = 100, p2 = 95 C2 be the shortage cost per unit time. The period
20 t of one run is divided into two parts t1 and t2.
C1 = 100 × = 20,
100 During the period t1 the items are drawn as per
q = 100 requirement from the stock and during t2 orders
2 DCS 2.500.50 for the items are accumulated but not supplied due
Q* = = = 50 to shortage of stock (t = t1 + t2).
C1 20
At the end of the period t1 a quantity Q is
Q* < q produced. This Q is divided into two parts Q1 and
D Q* Q2 such that Q1 is the quantity stored for current
TC1(Q*) = Dp1 + CS + C1 use and Q2 is exhausted to supply the back orders
Q* 2
This model is illustrated by the diagram given
500 50
= 50000 + × 50 + × 20 below:
50 2
= 50000 + 500 + 500 = 51000
D q
TC2(q1) = Dp2 + CS + 1 C1
q1 2
Q1 Q1
25000
= 47500 + + 10q1 t1 t2
q1 nt
O t 2t t1 t2
Taking TC1(Q*) = TC2 (q1)
25000
51000 = 47500 + + 10q1 1
q1 Total inventory over time t = Q1t1
2
10q12 – 3500q1 + 25000 = 0
1 / 2Q1t1
q1 = 342.7 or 7.3 Average inventory =
t
 Inventory Models 235

(1 / 2Q1t1 ) Also
Annual holding cost = × C1
t ∂C A Q12
1 = − C1
Total amount of shortage during t = Q2 t2 ∂Q 2Q 2
2
1  Q.2(Q − Q1 ) − (Q − Q1 ) 2  D
Annual shortage cost =
(1 / 2Q2 t2 )
× C2
+   C2 − 2 C S = 0
t
2  Q2  Q
D Q12 1
Set-up cost = nCS = CS i.e. − C1 + Q (Q − Q1 )C2 − (Q − Q1 ) 2 C2
Q 2 2
Total annual inventory cost − DCS = 0
(1 / 2Q1t1 ) (1 / 2Q2 t2 ) D 1 2 1
CA = C1 + C2 + C S − Q1 C1 + C2Q 2 − C2 QQ1 − Q 2 C2
t t Q 2 2
1 (t1 ) 1 (t2 ) D 1
= Q1C1 + Q2C2 + CS + QQ1C2 − Q12 C2 − DCS = 0
2 t 2 t Q 2
From similar triangles 1 2 2 2
t1 Q1 t Q Q C2 = Q1 C1 + Q1 C2 + DCS
= and 2 = 2 2 2 2
t Q t Q 2 2 2
Q C2 = Q1C1 + Q1 C2 + 2DCS
Qt Qt
\ t1 = 1 ; t2 = 2 2 DCS + Q12C1
Q Q Q2 = + Q12
C2

2 DCS + Q12C1
Q= + Q12 ...(2)
C2
QC2
From (1) Q1 =
C1 + C2
2 DCS Q 2 C22 C1
\ Q2 = +
1 Q12 1 ( Q − Q1 )
2
D C2 (C1 + C2 ) 2 C2
\ CA = C1 + C2 + CS
2 Q 2 Q Q Q 2 C22
(Q Q2 = Q – Q1) +
(C1 + C2 ) 2
If CA is to be minimum then
2 DCS Q 2 C22  C1 + C2 
∂C A ∂C A = +  
= =0 C2 (C1 + C2 ) 2  C2 
∂Q1 ∂Q
∂C A 1 2Q1 1 2(Q − Q1 ) 2 DCS Q 2 C2
= C1 + (−1)C2 = 0 = +
∂Q1 2 Q 2 Q C2 (C1 + C2 )
Q1 Q  C2  2 DCS
C1 + 1 C2 − C2 = 0 \ Q 2 1 −  =
Q Q  C1 + C2  C2
Q1 C2
(C1 + C2 ) = C2 fi Q1 = Q  C1  2 DCS
Q C1 + C2 Q2   =
...(1)  1
C + C 2  C2
 236 Operations Research

2 DCS C1 + C2 2 × 200 × 50(2 + 10)

\ Q2 = · =
C2 C1 2 × 10
2 DCS = 109.5 units
\ Q* = (C1 + C2 )
C1C2 Q* 109.5
Re-order cycle period t* = =
D 200
Q*C2
and Q*1 = = 1/2 day
C1 + C2
Example 18.17 The annual demand of an item
2 DCS (C1 + C2 )  C2  is 10000 units. The ordering cost is Rs 10. The
=   cost of the item is Rs 20. The holding cost is 20%
C1C2  C1 + C2 
of the value of the inventory per year. If the
2 DCS C2 shortage cost is 25% of the value per unit per year
=
C1 (C1 + C2 ) find the EOQ and the optimal annual inventory
cost.
Optimum annual cost is given by
2 Solution Given D = 10000 units
1 Q1* 1 (Q* − Q1* )2 D
CA* = C + C2 + * C S 20
2 Q * 1
2 Q *
Q CS = 10, C1 = × 20 = 4
100
On substitution and simplification, we get 25
C2 = × 20 = 5
2 DCS C1C2 100
CA* =
C1 + C2 2 DCS (C1 + C2 )
Q* =
Note: If shortages are not allowed then C2 = • C1C2
C2 Q * Q* 2 × 10000 × 10(9)
Q1* = = = = 300 units
C1 + C2 C1
+1
20
C2
2 DCS .(C1 C2 )
CA* =
2 DCS C1 + C2
= Q* =
C1 2 × 10000 × 10 × 20
= = Rs 666.7
which is the result obtained for Model 1. 9
Example 18.16 A commodity is to be supplied Example 18.18 A contractor supplies diesel
at the rate of 200 units per day. Ordering cost is engines to a truck manufacturer at the rate of
Rs 50 and the holding cost is Rs 2 per day. The 25 per day. The penalty for delay is Rs 10 per day
delay in supply induces a penalty of Rs 10 per per engine. The holding cost is Rs 16 per month.
unit per delay of one day. Find the optimal policy What should be the level of inventory at the
and the re-order cycle period. beginning of each month?
Solution Given D = 200 units per day Solution Here the fixed time interval is one
CS = 50, C1 = 2, C2 = 10 month.
t = 30 days. There is no set-up cost.
2 DCS (C1 + C2 ) 16
EOQ = Q* = Given C1 = per day C2 = 10
C1C2 30
 Inventory Models 237

Q = rt = 25 × 30 = 750 Similarly, the distribution of the shortage

QC2  10  inventory is given by
Q1* = = 750  
C1 + C2  16 / 30 + 10  0 D< y
G(y) = 
= 712 engines D − y D ≥ y
Example 18.19 The demand of an item is Assume that f (D) is the p.d.f. of demand. The
uniform at the rate of 20 units per month. The fixed expected inventory cost is given by
cost is Rs 10 per run. The production cost is Re 1 E[C(y)]= E (holding cost) + E (shortage cost)
per item and the holding cost is Re 0.25 per month. ∞ ∞
If the shortage cost is Rs 1.25 per item per month = C1 ∫ H ( y ) f ( D)dD + C2 ∫ G ( y ) f ( D)dD
determine the EOQ, the period of a run and 0 0
shortage period. y ∞
Solution Given D = r = 20; C1 = 0.25, = C1 ∫ ( y − D) f ( D)dD + C2 ∫ ( D − y ) f ( D)dD
C2 = 1.25, CS = 10 0 y

2 DCS (C1 + C2 ) ∂[ E{C ( y )}]

Q* = If E[C(y)] is to be minimum =0
C1C2 ∂y
2 × 20 × 10(0.25 + 1.25) ∂[ E{C ( y )}]
= = 44 =0
(0.25)(1.25) ∂y
y ∞
Q* 44
t* = = = 2.2 months (66 days) fi C1 ∫ f ( D)dD + C2 (−1) ∫ f ( D)dD = 0
D 20 0 y
Q*C2 1.25 y ∞
Q1 = = 44 × = 37
C1 + C2 1.5 fi C1 ∫ f ( D)dD − C2 ∫ f ( D)dD = 0
\ Q2 = 44 – 37 = 7 0 y

Q 7 ∞ y
t2 = 2 t = × 66 = 10.5 days f ( D )dD = 1 − ∫ f ( D)dD
Q 44
Now ∫
y 0

18.5 PROBABILISTIC MODEL y  y 

\ C1 ∫ f ( D)dD − C2 1 − ∫ f ( D)dD  = 0
In the deterministic models, the demand is known 0  0 
precisely. But in most practical situations the y
demand is probabilistic. We know only the (C1 + C2 ) ∫ f ( D)dD = C2
probability distribution of the future demand. This 0
probability distribution is determined from the data
y
collected from past experience. C2
In this probabilistic model, the demand D is
\ ∫ f ( D)dD = C1 + C2
0
probabilistic and the inventory on hand is y. There
is no setup cost and the lead time is zero. C1 is the C2
Thus we get P ( D ≤ y ) =
*
holding cost and C2 is the shortage cost. C1 + C2
The distribution of the inventory on hand is
This is the condition to be satisfied by y* and y*
given by
is determined accordingly.
y − D D < y In case the distribution of D is discrete then we
H(y) = 
0 D≥ y have
 238 Operations Research

E[C(y)] D 0 1 2 3 4 5
= C1 ∑ ( y − D) f ( D) + C2 ∑ ( D − y ) f ( D) P(D £ y) 0.1 0.3 0.55 0.75 0.9 1
*
If E[C(y )] is minimum then
C2 2 2
E[C(y* – 1)] ≥ E[C(y*)] and = = = 0.6
C1 + C2 1+ 2 3
E[C(y* + 1) ≥ E[C(y*)] From the table we find that
C2 P(D £ 2) = 0.55 P(D £ 3) = 0.75
P[D £ (y* – 1)] £ and
C1 + C2 P(D £ 2) < 0.6 < P(D £ 3)
C2 \ y* = 3
P[D £ y*] ≥
C1 + C2 Example 18.22 A bookseller purchases a weekly
C2 magazine at Rs 3 and sells at Rs 7 each. He cannot
\ P[D £ (y* – 1)] £ £ P[D £ y*] return unsold magazines. Weekly demand has the
C1 + C2
following distribution.
This is the condition to be satisfied by y*.
D 23 24 25 26 27 28 29 30 31 32
Example 18.20 The demand of an item is given
f (D) .01 .03 .06 .1 .2 .25 .15 .1 .05 .05
1 / 10 0 ≤ D ≤ 10
by the p.d.f. f(D) = 
0 D > 10 How many magazines should be ordered every
The holding cost is Re 1 per unit and shortage week?
cost is Rs 4 per unit. Find the EOQ.
Solution Given C1 = 3, C2 = 7 – 3 = 4
Solution Given C1 = 1, C2 = 4,
The cumulative distribution of demand is
C2 4 4
= = = 0.8
C1 + C2 1+ 4 5 D 23 24 25 26 27 28 29 30 31 32
P(D £ y) 0.01 0.04 0.10 0.20 0.40 0.65 0.80 0.90 0.95 1
1
f (D) =
10 C2 4 4
= = = 0.57
y *
y * C1 + C2 3 + 4 7
1 1
∫ ∫ 10 dD = 10 y
*
P[D £ y*] = f ( D)dD = From the table
0 0 P(D £ 27) = .4 P(D £ 28) = 0.65
* \ P(D £ 27) < .57 > P(D £ 28)
y
\ = 0.8 \ y* = 28
10
\ y* = 8 Note: If the cost, selling price and the salvage
value of unsold item are given then we use the
Example 18.21 The distribution of demand of formula
an item is given by Profit
P(D £ y*) =
D 0 1 2 3 4 5 Profit + loss
f (D) 0.1 0.2 0.25 0.2 0.15 0.1 Example 18.23 The sales of butter (demand for
butter) has the following probability distribution.
Holding cost is Re 1 and the shortage cost is
Rs 2. Find the EOQ. Demand (kg) 14 15 16 17 18 19 20

Solution Probability 0.02 0.08 0.1 0.4 0.2 0.15 0.05

Here C1 = 1 and C2 = 2 The cost of butter is Rs 80 per kg and the selling
The cumulative distribution of demand is price is Rs 120 per kg. Unsold stock is disposed

off at Rs 60 per kg at the end of the day. Determine
how much of butter is to be stocked at the
beginning of each day.
Solution The cumulative distribution of demand
is given by
D 14 15 16 17 18 19 20
Probability .02 .08 .1 .4 .2 .15 .05
P(D £ y) .02 .1 .2 .6 .8 .95 1
Profit per kg of butter = 120 – 80 = Rs 40
Salvage value = Rs 60
EXERCISES
■ ■ ■ ■ ■ ■ ■
1. A company purchases lubricants at the rate
of Rs 42 per tin. The requirement is 1800
per year. The cost of placing an order is
Rs 16 and the holding charges are 20 paise
per rupee per year. Find the EOQ.
2. Raw material costs Re 1 per unit and the
annual requirement is 2000 units. Ordering
cost is Rs 10 and carrying cost is 16% per
year per unit. Find the EOQ and optimum
inventory cost per year.
3. An aircraft company requires rivets at the
rate of 5000 kg per year. The cost of the rivet
is Rs 20 per kg and the ordering cost is
Rs 200. The holding cost is 10% per kg per
year. How frequently should orders be
placed? What quantity should be ordered
for?
4. A company has a demand of 12000 units of
an item per year. Set-up cost is Rs 400 and
the holding cost is 15 paise per unit per
month. Find the EOQ and the optimum
inventory cost per year.
5. The daily demand of an item is 30 units.
Holding cost is 5 paise per unit per day and
the setup cost is Rs 100. Find the economic
lot size, period of one run and the total
inventory cost.
Inventory Models 239
\ Loss per kg of unsold sold butter = Rs 20
Now, for y* the condition is
Profit
P(D £ y*) =
Profit + loss
40
= = 0.67
40 + 20
From the table
P(D £ 17) = 0.6 P(D £ 18) = 0.8
\ P(D £ 17) < 0.67 < P(D £ 18)
\ y* = 18
Every day 18 kg of butter is to be stocked.
6. A company has to supply 10000 auto parts
per day. They can produce 25000 parts per
day. Holding cost is 20 paise per unit per
year and the set-up cost is Rs 180 per run.
How frequently should production be made?
7. Given k = 24000, r = 12000, C1 =1.8,
CS = 400, find Q*.
8. An item is produced at the rate of 50 per
day. The demand is 25 items per day. The
setup cost is Rs 100 and the holding cost is
Re 0.01 per item per day. Find
(i) economic lot size
(ii) period of one run
(iii) minimum total inventory cost.
9. The production rate of tomato ketchup is 600
bottles per day and the demand is 150 bottles
per day. The cost per bottle is Rs 6. The set-
up cost is Rs 90 per run and the holding cost
is 20% of the cost per annum. Find the
economic lot size and the number of
production runs.
10. The annual sales of a particular grade of paint
is 30000 litres. The holding cost is Re 1 per
litre. The setup cost is Rs 80 per run.
Determine the economic lot size.
11. For a particular product the daily demand is
20 units, setup cost is Rs 100 and the holding
 240 Operations Research
cost is Rs 0.20 per day. The production rate
is 50 units per day. Determine the economic
lot size and the period of a run.
12. A shopkeeper purchases an item at the rate
of 50 units per month. The cost is Rs 6 for
orders less than 200 and Rs 5.70 for orders
not less than 200 units. The ordering cost is
Rs 10 and the holding cost is 20% of the
value per year. Find the EOQ.
13. Find the EOQ for a product for which the
price breaks are as follows:
Price (Rs) Quantity
10 q < 500
9 q ≥ 500
The monthly demand is 200 units. The
ordering cost is Rs 350 and the holding cost
is 2% of the cost per month.
14. The annual demand of an item is 10000 units.
The ordering cost is Rs 40 and the holding
cost is 20% of the cost per unit per year. The
price of the item is Rs 4 and there is a
discount of 10% for orders of 1500 units or
more. Find the EOQ.
15. Given the following data find the EOQ and
the reorder level for an item
Annual demand = 2500 units
Ordering cost = Rs 12.50
Holding cost = 25% of the average
inventory
No.of working days = 250 per year
Lead time = 3 days.
The cost of the item is Rs 4 per unit for orders
less than 200 and Rs 3 for orders not less
than 200.
16. The cost of a notebook is Rs 10 with a
discount of 5% for orders of 200 or more.
The demand is at the rate of 50 notebooks
per month and the ordering cost is Rs 10.
The holding cost is 20% of the value per year.
Determine the annual minimum inventory
cost and the number of orders per year.
17. A manufacturer has to supply 24000 units
of his product per year. Penalty for delay in
supply is Rs 0.20 per unit per month. The
holding cost is Rs 0.10 per unit per month
and the setup cost is Rs 350 per run. Find
the economic lot size.
18. The demand of an item is at the rate of 25
units per month. The fixed cost is Rs 15 per
run. The carrying cost is Rs 0.30 per item
per month. If the shortage cost is Rs 1.50
per item per month determine the economic
lot size and the period of a run.
19. A TV company requires speakers at the rate
of 8000 units per month. The setup cost is
Rs 12000 per run. The holding cost is
Rs 0.30 per unit per month. Find the EOQ
(i) when shortage are not allowed.
(ii) when the shortage cost is Rs 1.10 per
unit per month.
20. The demand for an item is 18000 units per
year. The holding cost is Rs 1.20 per unit
per year. The cost of production is Rs 400
and the shortage cost is Rs 5. Determine the
EOQ.
21. Find the EOQ if the demand is 16 units per
day, ordering cost is Rs 15, holding cost is
15 paise per unit per day and the shortage
cost is 75 paise per unit per day.
22. The annual demand of an automobile part is
6000 units. The setup cost per production
run is Rs 500 and the storage cost per unit
per year is Rs 8. If the shortage cost is Rs 20
per unit per year find
(i) economic lot size
(ii) period of one run
(iii) number of shortages per run.
23. A newspaper boy purchases newspapers at
Rs 2 per copy and sells them at Rs 3. The
demand for the newspaper has the following
probability distribution.
Demand 10 11 12 13 14 15 16
Probability 0.05 0.15 0.40 0.20 0.10 0.05 0.05
Unsold newspapers cannot be returned.
Determine how many copies should he
purchase in order to minimize his loss.
24. The demand for a food commodity has the
probability distribution given below:

Demand 10 11 12 13
Probability 0.10 0.15 0.20 0.25 0.30
The cost is Rs 30 each and the selling price
is Rs 50. Unsold items become useless.
Determine how many items should be
prepared.
25. Some of the spare parts of a car cost Rs 200
each. These parts can be ordered only
together with the car. Suppose there is a loss
of Rs 4800 for each spare part not in stock.
Also the probability distribution of their
demand is given by
Demand 0 1 2 3 4 5 6 or more
Probability 0.9 .04 .025 .02 .01 .005
Determine how many spare parts should be
procured.
26. The demand for an item has the p.d.f.
1 Determine the EOQ.
f (x) = (4000 £ x £ 5000)
1000
The holding cost is Re 1 and the shortage
cost is Rs 7. Find the EOQ
27. A bakery sells cake by weight. There is a
profit of Rs 4.50 per kg sold. Unsold cakes are
disposed off at a loss or Rs 1.50 per kg. The
distribution of demand is rectangular between
2000 and 3000 kg. Determine the optimum
quantity of cake to be prepared per day.
28. The demand for an item has rectangular
distribution
1 per year. The cost of the rivet is Rs 2 per kg.
f (x) = (5000 ≤ x ≤ 6000 )
1000
The holding cost is Rs 10 and the shortage
cost is Rs 30. Find the EOQ.
29. Find the EOQ, reorder level and the optimum
annual inventory cost given that the annual
demand is 1000 units, holding cost is 0.15
ordering cost is Rs 10, and lead time is 2 years.
30. Annual demand is 2500 units. The cost of
one unit is Rs 30. The ordering cost is
Rs 130 and the carrying cost is 10% per year.
Estimate the EOQ and also determine the
number of orders per year.
31. A factor has to supply a product at the rate
of 25 per day. The carrying cost is Rs 16 per
Inventory Models 241
14 month and the shortage cost is Rs 10 per unit
per day. The setup cost for production run is
Rs 10000. Determine how frequently should
the production be started and what should
be the level of inventory at the beginning of
each run.
32. An auto part has a demand of 9000 units per
year. The ordering cost is Rs 100 and the
holding cost is Rs 2.40 per unit per year. If
the production is instantaneous and no
shortages are allowed find the
(i) EOQ
(ii) Number of orders per year
(iii) Time between two consecutive orders
(iv) Annual inventory cost.
0
33. A shopkeeper purchases an item at Rs 40
per unit. Annual demand of the item is 2000
units. The cost of placing an order is Rs 15
and the holding cost is 20% per year.
34. The demand of an item is 400 units per week.
The cost of the item is Rs 50 per unit.
Ordering cost is Rs 75 and the holding cost
is 7.5% of the cost of the item per year. Find
the economic lot size and the optimum
inventory cost.
35. An item costs Rs 235 per ton. The monthly
requirement is 5 tons and the setup cost is
Rs 1000. The holding cost is 10% of the
value per year. Determine the EOQ.
36. A company uses rivets at the rate of 5000 kg
The cost of placing an order is Rs 20 and the
holding cost is 10% per year. How frequently
should the order for rivets be placed and for
how much?
37. The production rate of an item is 200 units
per day and the demand is 100 units per day.
The set-up cost is Rs 200 and the holding
cost of one unit is Rs 0.81 per day. Find the
economic lot size.
38. The cost of a hardware item is Rs 300 per
ton. Annual requirements are 120 tons and
the setup cost is Rs 240. If the inventory
carrying cost is 12% of the value determine
the period of one run.
 242 Operations Research

39. The demand for an item is 18000 units per
year. The setup cost is Rs 400 and the holding
cost is Rs 1.20. If the shortage cost is Rs 5
determine the optimal order quantity.
40. Determine the EOQ
(i) Annual demand: 240 units
Set-up cost: Rs 10
Holding cost: Rs 0.25 per month
Shortage cost: Rs 1.25 per month
(ii) Weekly demand: 16 units
Set-up cost: Rs 15
Holding cost: 15% per week
Purchase cost: Rs 8 per unit
Shortage cost: Rs 0.75 per week
(iii) Monthly demand: 25 units
Set-up cost: Rs 15
Holding cost: Rs 0.30 per month
Shortage cost: Rs 1.50 per month
41. The demand of an item has p.d.f
f(x) = 0.1 0 £ x £ 10
0 x > 10
The holding cost is Rs 0.50 per day
Shortage cost is Rs 4.5 per day. Find the
optimum inventory level.
42. Annual demand: 3600 units
Ordering cost: Rs 50
Holding cost: 20% of the unit cost
Price break O < q < 100 – Rs 20
Q > 100 – Rs 18

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. 83 23. y* = 13
2. Q* = 500; CA* = Rs 80 24. y* = 12
3. Q* = 1000; Order cycle = 2 months 12 days 25. y* = 2
4. Q* = 730; CA* = Rs 4157 26. y* = 4875
5. Q* = 346; Period of run = 11.5 days; 27. y* = 2750 kg
CA* = Rs 17.32 28. y* = 5750
6. Production run = 10.5 days 29. Q* = 365; Reorder level = 2000 units,
7. Q* = 3266 Rs 54.80
8. Q* = 1000; Period of run = 40; CA* = Rs 5 30. Q* = 466; 5.3 orders per year
9. Q* = 3309; No. of runs = 17 per year 31. 40 days; 943 units
10. Q* = 2191 32. Q* = 866; No. of orders = 10.4 / year; Time
11. Q* = 183; Period of run = 9 days between two orders = 35 days
12. Q0 = q = 200 Annual inventory cost = Rs 2080
13. Q0 = 837 33. 87 units
14. Q0 = 1500 34. Q* = 288; Inventory cost = Rs 65.80 per
15. Q0 = 250; Reorder level = 220 units week.
16. Q0 = 200; CA0 = Rs 494; 3 orders per year 35. Q* = 71.458 tons
17. Q* = 4583 36. t* = 73 days; Q* = 1000 kg
18. Q* = 17, t* = 20 days 37. Q* = 314
19. (i) 25298 (ii) 28540 38. t* = 4 months
20. Q* = 3857 39. Q* = 3856 units
21. Q* = 6 40. (i) 44 (ii) 32 (iii) 55
22. (i) Q* = 1024, (ii) t* = 2 months 41. 9
(iii) Q2* = 293 42. 134

Replacement Models
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
19.1 INTRODUCTION
Replacement becomes necessary when the
job-performing units such as men, machines,
equipment, parts etc., lose their efficiency
and effectiveness because of gradual deterioration
or sudden failure or breakdown. Planned
replacement of these items would reduce
maintenance cost and other overhead expenses.
When a machine loses its efficiency gradually the
maintenance becomes very expensive. Therefore
the problem is to determine the age at which it is
most economical to replace the item. On the other
hand, certain items such as bulbs, radio and
television parts fail suddenly without giving any
indication of failure and they become completely
useless. These items are to be replaced
immediately as and when they fail to function.
Thus there are two types of situations for
replacement.
(i) Replacement of items which deteriorate
and whose maintenance cost increases with
time.
(ii) Replacement of items which fail all on a
sudden.
19
19.2 REPLACEMENT OF ITEMS
WHICH DETERIORATE IN
EFFICIENCY WITH TIME
A machine loses efficiency with time and we have
to determine the best time at which we have to go
for a new one. In case of a vehicle the maintenance
cost is increasing as it is getting aged. These costs
increase day by day if we postpone the rep-
lacement.
First we consider the situation where the
maintenance costs increase with time assuming
that the value of money remains the same during
the period.
19.2.1 Formula for Finding the
Time for Replacement
Let C be the capital cost of the machine,
S(t) be the scrap value of the machine after t years,
f(t) be the operating cost or maintenance cost
at time t and n be the number of years. The
annual cost of the machine at time t is given by
C – S(t) + f(t). The total maintenance cost
n
during n years is given by å f (t ) (when t is
t =0
 244 Operations Research

n Year f(t) Sf(t) S(t) C – S(t) T TA

discrete) or ò f (t ) dt (when t is continuous). Thus
0 1 200 200 200 12000 12200 12200
we find that the total cost after n years is given by 2 500 700 200 12000 12700 6350
n 3 800 1500 200 12000 13500 4500
T = C - S (t ) + ò0 f (t ) dt 4 1200 2700 200 12000 14700 3675
The average annual cost is given by 5 1800 4500 200 12000 16500 3300
6 2500 7000 200 12000 19000 3167
TA = é C - S (t ) + ò f (t ) dt ù
1 n
7 3200 10200 200 12000 22200 3171
n ê
ë 0 ûú 8 4000 14200 200 12000 26200 3275
We have to determine the value of n which
would give minimum TA From the table we find that TA is minimum at
the end of the 6th year. Hence it is profitable to
1 n
Now TA = [ C - S (t ) ] + ò f (t ) dt replace the machine at the end of the 6th year.
n 0
dTA 1 Example 19.2 The maintenance cost and the
= - 2 [ C - S (t ) ] resale price of a truck are given below:
dn n
1 n 1 Year 1 2 3 4 5 6 7 8
- 2 ò f (t )dt + f (n) Mainte- 1000 1300 1700 2200 2900 3800 4800 6000
n 0 n
nance cost
dTA
TA is minimum fi =0 Resale 4000 2000 1200 600 500 400 400 400
dn price
[C - S (t ) 1 n
\ f ( n) - - ò f (t )dt = 0 The purchase price of the truck is Rs 8000.
n n 0 Determine the time at which it is profitable to
[C - S (t )] 1 n
f (n) = + ò f (t ) dt replace the truck.
n n 0 Solution Given
1é n ù C = 8000
= ê C - S (t ) + ò0 f (t ) dt ú
në û T = C – S(t) + Sf(t)
= TA T
Thus TA becomes the least at the time when TA =
n
TA = f (n)
Year f(t) Sf(t) S(t) C – S(t) T TA
Note: When t is discrete the tabular method is used
to find n for which TA is minimum. 1 1000 1000 4000 4000 5000 5000
2 1300 2300 2000 6000 8300 4150
Example 19.1 A machine costs Rs 12,200. The 3 1700 4000 1200 6800 10800 3600
scrap value is Rs 200. The maintenance costs of 4 2200 6200 600 7400 13600 3400
the machine are given below: 5 2900 9100 500 7500 16600 3320
Year 1 2 3 4 5 6 7 8 6 3800 12900 400 7600 20500 3417
7 4800 17700 400 7600 25300 3614
Mainte- 200 500 800 1200 1800 2500 3200 4000
8 6000 23700 400 7600 31300 3913
nance cost
From the table we find that the average annual
When should the machine be replaced?
cost is minimum at the end of the 5th year. Hence
Solution Given
it is profitable to replace the truck at the end of
C = 12,200
the 5th year.
S(t) = 200,
Example 19.3 A machine costs Rs 900. The
C - S (t ) + S f (t )
TA = annual operating cost is Rs 200 for the first year
n and is then increasing by Rs 2000 per year for
The average annual cost is given by the subsequent years. There is no scrap value.
following table. Determine the best age to replace the machine.
 Replacement Models 245

Solution Given Since TA is minimum at the end of the 3rd year,

C = 9000; S(t) = 0 the best time to replace the machine is at the end
Year f(t) Sf(t) S(t) C – S(t) T TA of the 3rd year.
1 200 200 0 9000 9200 9200 Example 19.4 There are two types of
2 2200 2400 0 9000 11400 5700 autorikshaws, type A and type B. The purchase
3 4200 6600 0 9000 15600 5200 prices of these two types are Rs 7000 and
4 6200 12800 0 9000 21800 5450 Rs 9000 respectively. The resale values and
5 8200 21000 0 9000 30800 6160
6 10200 31200 0 9000 40200 6700
the running costs for the two types are given
7 12200 43400 0 9000 52400 7486 below.

Type A
Year 1 2 3 4 5 6 7 8
Running cost 1100 1300 1500 1900 2400 2900 3500 4100
Resale price 3100 1600 850 475 300 300 300 300
Type B
Year 1 2 3 4 5 6 7 8
Running cost 1300 1600 1900 2500 3200 4100 5100 6200
Resale price 4100 2100 1100 600 400 400 400 400
Determine which type of autorickshaw is to be purchased.
Solution type B should be replaced at the end of the 5th
year. Also the annual average cost of A is less than
For type A, C = Rs 7000
that of B uniformly.
Year f(t) Sf(t) S(t) C – S(t) T TA Hence it is advisable to purchase the
1 1100 1100 3100 3900 5000 5000 autorikshaw of type A.
2 1300 2400 1600 5400 7800 3900
Example 19.5 Machine A costs Rs 8000.
3 1500 3900 850 6150 10050 3350
4 1900 5800 475 6525 12325 3081
Annual operating cost is Rs 1000 for the first
5 2400 8200 300 6700 14900 2980
year and is increasing at the rate of Rs 2000 per
6 2900 11100 300 6700 17800 2967 year afterwards. The machine has no resale
7 3500 14600 300 6700 21300 3043 value. Another machine B costs Rs 10,000.
8 4100 18700 300 6700 25400 3175 The annual operating cost is Rs 400 for the
first year and is increasing at the rate of Rs 800
For type B, C = Rs 9000 every year with no resale value. If the machine
Year f(t) Sf(t) S(t) C – S(t) T TA A is one year old, is it advisable to replace A with
1 1300 1300 4100 4900 6200 6200 B? If so find the appropriate time for replacement.
2 1600 2900 2100 6900 9800 4900 Solution
3 1900 4800 1100 7900 12700 4233 For machine A, C = 8000
4 2500 7300 600 8400 15700 3925
Year f(t) Sf(t) S(t) C – S(t) T TA
5 3200 10500 400 8600 19100 3820
6 4100 14600 400 8600 23200 3867 1 1000 1000 0 8000 9000 9000
7 5100 19700 400 8600 28300 4043 2 3000 4000 0 8000 12000 6000
8 6200 25900 400 8600 34500 4313 3 5000 9000 0 8000 17000 5667
4 7000 16000 0 8000 24000 6000
From the above tables we find that type A 5 9000 25000 0 8000 33000 6600
should be replaced at the end of the 6th year and 6 11000 36000 0 8000 44000 7333
 246 Operations Research

For machine B, C = 10000 Machine B, S(t) = 0

Year f(t) Sf(t) S(t) C – S(t) T TA Year f(t) Sf(t) C – S(t) T TA
1 400 400 0 10000 10400 10400 1 2000 2000 50000 52000 52000
2 1200 1600 0 10000 11600 5800 2 6000 8000 50000 58000 29000
3 2000 3600 0 10000 13600 4533 3 10000 18000 50000 68000 22667
4 2800 6400 0 10000 16400 4100 4 14000 32000 50000 82000 20500
5 18000 50000 50000 100000 20000
5 3600 10000 0 10000 20000 4000
6 22000 72000 50000 122000 20333
6 4400 14400 0 10000 24400 4067
7 26000 98000 50000 148000 21143
7 5200 19600 0 10000 29600 4229
We find that the average annual cost of A is
From the tables we find that the best age for lowest in the 3rd year (Rs 26000) and that of B is
replacement of machine A is 3 years and the time lowest in the 5th year (Rs 20000). Hence machine
for replacement of B is 5 years. A should be replaced by machine B.
Now assume that at present machine A is one To find the time of replacement
year old. Hence its cost is Rs 9000 now, Rs 6000 Machine A should be replaced by machine B at
after one year, Rs 5667 after two years Rs 6000 the time when the total cost for the next year
after three years, Rs 6600 after four years and so exceeds the lowest average annual cost (Rs 20000)
on. Comparing with the average total cost of B we of machine B.
find that the average total cost for B does not Year Difference between the total costs
exceed that of one year old A in the 3rd year. Hence 1 46000
machine A should be replaced with machine B after 2 57000 – 46000 = 11000
3 years. 3 78000 – 57000 = 21000
4 109000 – 78000 = 31000
Example 19.6 Machine A costs Rs 45000 and
the running cost is Rs 1000 for the first year Thus the total cost of machine A during the third
increasing by Rs 10000 per year afterwards. year (Rs 21000) is more than the lowest average
Another machine B costs Rs 50000 and operating annual cost of machine B.
cost is Rs 2000 for the first year increasing by Therefore A should be replaced by B after two
Rs 4000 per year subsequently. If we have machine years.
A now, should we replace it with B? If so find the 19.2.2 Replacement of Items
best time for replacement. Whose Maintenance
Solution For machine A, C = 45000 and for Costs Increase with Time
machine B, C = 50000. The following tables show and the Value of Money
the average annual costs of the two machines, Also Changes with Time
taking the resale value to be zero. As the value of money changes with time we have
Machine A, S(t) = 0 to calculate the present worth of the money we
Year f(t) Sf(t) C – S(t) T TA gain or spend a few years hence. Suppose the initial
1 1000 1000 45000 46000 46000
cost of a machine is C and the maintenance costs
2 11000 12000 45000 57000 28500 be Cj( j = 1, 2, 3 ...) for the jth year. The value of
3 21000 33000 45000 78000 26000 money is decreasing at the rate of r% per year.
4 31000 64000 45000 109000 27250 Thus one rupee invested today worths (1 + r) rupee
5 41000 105000 45000 150000 30000 next year, (1 + r)2 after two years, ... (1 + r)n after
6 51000 156000 45000 201000 33500 n years. In other words ... (1 + r)–n is the present
7 61000 217000 45000 262000 37429 worth of one rupee spent after n years. Denoting
 Replacement Models 247

(1 + r)–1 by v we say that vn is the present worth C

factor (p.w.f) of one rupee spent after n years. Year Cn vn–1 Cnvn–1 + å Cnvn–1 S vn R
1
n = is called discount rate or depreciation 1 60000 1 60000 560000 1 560000
1+ r 2 60000 0.9091 54546 614546 1.9091 321904
value per unit money. Let Vn be the p.w.f. of all 3 60000 0.8264 49584 664130 2.7355 242782
the future costs. 4 60000 0.7513 45078 709208 3.4868 203398
5 60000 0.6830 40980 750188 4.1698 179910
C + C1 + C2 v + ... + Cn v n -1 6 80000 0.6209 49672 799860 4.7907 166961
Vn =
1 - vn 7 100000 0.5645 56450 856310 5.3552 159903
8. 120000 0.5132 61584 917894 5.8684 156413
The value of Vn should be minimum if we want 9 140000 0.4665 65310 983204 6.3349 155204
to replace after n – 1 years. The condition is given 10 160000 0.4241 67856 1051060 6.7590 155505
by
From the table we find that C9 < R < C 10
Cn -1 C [140000 < 155204 < 160000]
< Vn < n
1- v 1- v Therefore the computer must be replaced after
i.e. Cn–1 < (1 – n)Vn < Cn 9 years.
i.e. Cn–1 < Example 19.8 A machine costs Rs 250000. The
n -1 running and maintenance costs are Rs 120000 per
(C + C1 + C2 v + ... + Cn v )
(1 - v) < Cn year for the first five years and increasing there
1 - vn after by Rs 20000 per year. If the money is worth
10% per year and there is no salvage value for the
C + C1 + C2 v + ... + Cn v n -1
i.e. Cn – 1 < < Cn machine determine the best period for replacing
1 + v + v 2 + ... + v n -1 the machine.
C + C1 + C2 v + ... + Cn v n -1 Solution Given
Denoting by R
1 + v + v 2 + ... + v n -1 v=
1
=
1
=
1
= 0.9091
we have Cn–1 < R < Cn. R is called weighted 1 + r 1 + 0.1 1.1
average cost. Table for calculating the average cost is given
Note: If A and B are two machines with weighted below.
average costs R1 and R2 then we say that machine Year Cn vn–1 Cnvn–1 C + SCnvn–1 S vn R
A is better than B if R1 < R2
1 120000 1 120000 370000 1 370000
Example 19.7 A computer costs Rs 5 lakhs. 2 120000 0.9091 109092 479092 1.9091 250952
Running and maintenance costs are Rs 60000 for 3 120000 0.8264 99168 578260 2.7355 211391
each of the first five years, increasing by Rs 20000 4 120000 0.7513 90156 668416 3.4868 191699
per year in subsequent years. Assuming that the 5 120000 0.6830 81960 750376 4.1698 179955
value of money is decreasing at 10% per year and 6 140000 0.6209 86926 837302 4.7907 174777
that there is no salvage value for the computer, 7 160000 0.5645 90320 927622 5.3552 173219
find the optimal period for replacing the computer. 8 180000 0.5132 92376 1019998 5.8684 173812
Solution Given that money is worth 10% per 9 200000 0.4665 93300 1113298 6.3349 175740
year the discount factor 10 220000 0.4241 93302 1206600 6.7590 178518

1 1 Here C7 < R < C8 [160000 < 173219 < 180000]

v= = = 0.9091 \ The machine should be replaced after 7 years.
1 + 0.1 1.1
The average cost is given by the following table Example 19.9 There are two machines A and B.
C = 500000 A costs Rs 2500. Its running cost is Rs 400 per year
for the first five years and is increasing by Rs 100
 248 Operations Research

per year afterwards. B costs Rs 1250 with running The table shows that B should be replaced after
cost Rs 600 per year for 6 years increasing by 8 years.
Rs 100 per year thereafter. If money is worth 10% The weighted average cost of A in
per year and both machines have no scrap value, 9 years (Rs 875.86) is more than that of B in
determine which machine should be purchased. 8 years (Rs 840.11). Hence it is advisable to
Solution Let us calculate the average cost for A. purchase B.
Year Cn vn–1 Cnvn–1 C + SCnvn–1 S vn R Example1 9.10 A scooter costs Rs 6000. The
1 400 1 400 2900 1 2900 running cost and the salvage cost are given below.
2 400 0.9091 363.64 3263.64 1.9091 1709.45 If the interest rate is 10% per year find when the
3 400 0.8264 330.56 3594.20 2.7355 1313.84 scooter should be replaced.
4 400 0.7513 300.52 3894.72 3.4868 1116.93
5 400 0.6830 273.20 4167.92 4.1698 999.50 Year 1 2 3 4 5 6 7
6 500 0.6209 310.45 4478.37 4.7907 948.23 Running cost 1200 1400 1600 1800 2000 2400 3000
7 600 0.5645 338.70 4817.07 5.3552 899.40 Salvage value 4000 2666 2000 1500 1000 600 600
8 700 0.5132 359.24 5176.31 5.8684 881.92
9 800 0.4665 373.20 5549.51 6.3349 875.86 Solution The salvage value is to be taken to
10 900 0.4241 381.69 5931.20 6.7590 877.35 account only at the end of the year. Hence we
We find that A should be replaced after 9 years. multiply C n by v n–1 and S n by v n and obtain
Average cost for machine B the total cost at the end of (n – 1) th year as
C + SCnvn – 1 – Snvn. For the first year we find C +
Year Cn vn–1 Cnvn–1 C + SCnvn–1 Svn R
C, v0 – S1v, for the second year we find C + (C1 +
1 600 1 600 1850 1 1850 C2v) – S2V2 and so on.
2 600 0. 9091 545.46 2395.46 1.9091 1254.75
At the end of the nth year the average cost
3 600 0.8264 495.84 2891.30 2.7355 1056.95
4 600 0.7513 450.78 3342.08 3.4868 958.49 becomes
5
6
600 0.6830
600 0.6209
409.80
372.54
3751.88
4124.42
4.1698 899.77
4.7907 860.92 C+ å Cn vn -1 - Sn vn
(n =1, 2, 3,….)
7
8
700 0.5645
800 0.5132
395.15
410.56
4519.57
4930.13
5.3552 843.96
5.8684 840.11
å vn -1
9 900 0.4665 419.85 5349.98 6.3349 844.52 The following table gives the average cost at
10 1000 0.4241 424.10 5774.08 6.7590 854.28 the end of each year.

C = 6000
n–1 n–1 n–1
Year Cn v Cnv SCnv Sn vn Snvn SCnvn–1 C + SCnvn–1 Svn–1 R
– Snvn – Snvn
1 1200 1 1200 1200 4000 0.9091 3636.4 – 2436.4 3563.6 1 3563.6
2 1400 0.9091 1272.7 2472.7 2666 0.8264 2203.2 269.5 6269.5 1.9091 3284.0
3 1600 0.8264 1322.2 3794.9 2000 0.7513 1502.6 2292.3 8292.3 2.7355 3031.3
4 1800 0.7513 1352.3 5147.2 1500 0.6830 1024.5 4122.7 10122.7 3.4868 2903.1
5 2000 0.6830 1366.0 6513.2 1000 0.6209 620.9 5892.3 11892.3 4.1698 2852.0
6 2400 0.6209 1490.2 8003.4 600 0.5645 338.7 7664.7 13664.7 4.7907 2852.3
7 3000 0.5645 1693.5 9696.9 600 0.5132 307.9 9389 15389 5.3552 2873.6

C6 < R < C7 [2400 < 2852.3 < 3000] Therefore it is advisable to replace the scooter
at the end of 6th year.
 Replacement Models 249

19.3 REPLACEMENT OF ITEMS t -1

NC1
THAT FAIL COMPLETELY
tf (t - 1) - å f (i) < C2
i =1
t -1
There are many situations in which items do not
deteriorate with time but fail all on a sudden < t f (t ) - å f (i )
i =1
completely. It may not be possible to predict the
N .C1
time of failure. Hence we make use of the
probability distribution of the failure time which
\ tf (t ) - å f (i) > C2
is obtained from past experience. N .C1
Two types of replacement policies are followed i.e. tf (t ) > C + å f (i)
2
in such situations. They are
(i) Individual replacement policy NC1 + C2 å f (i )
i.e. C2 f (t ) >
(ii) Group replacement policy t
Under individual replacement policy, an item C (t )
(or equipment) is replaced immediately after its > (Average cost)
t
failure. Thus group replacement must be made at the
Under group replacement policy, all the items end of period t if the cost of individual replacement
are collectively replaced after a specific period of for the tth period is greater than the average cost
time irrespective of whether they have failed or till the end of t periods. If the cost of individual
not. If any item fails before this time it will also replacement at the end of (t – 1)th period is less
be replaced immediately. The optimal period of than the average cost for t periods group
replacement is determined by calculating the replacement should not be done. In general we take
minimum total cost. This total cost is calculated one period of time as one year.
using
(i) Probability of failure at time t Example 19.11 The probability distribution of
(ii) Number of items failing during time t the failure time of a certain type of electric bulb is
(iii) Cost of group replacement given below:
(iv) Cost of individual replacement Week 1 2 3 4 5 6 7 8
If p(x) is the probability that an item will fail at Prob. of 0.05 0.13 0.25 0.43 0.68 0.88 0.96 1.0
age x then its average age at failure is failure
k
E(x) = å x p ( x) The cost of individual replacement is Rs 4 per
bulb. The cost of group replacement is Re 1 per
x =1
bulb. If there are 1000 bulbs in use find the optimal
If there are N items in the group then the number replacement policy under
N (i) individual replacement
of failures per unit time is . Suppose C1 is
E ( x) (ii) group replacement
the cost of replacing per unit under group Solution
replacement and C2 be the cost of individual (i) We calculate the probability of failure during
replacement. We wish to take a decision about each week as follows:
group replacement. Let f (i) denote the number of
Week 1 2 3 4 5 6 7 8
failures and C(t) denote the total cost at time t.
t -1 Prob. of 0.05 0.08 0.12 0.18 0.25 0.20 0.08 0.04
Then C (t ) = NC1 + C2 å f (i ) . The optimum failure
i =1 Expected life
value of t must satisfy the inequality, E(x) = Sx p(x)
 250 Operations Research

= 1(0.05) + 2 (0.08) + 3 (0.12) N5 = Np5 + N1 p4 + N2 p3 + N3 p4 + N4 p1

+ 4(0.18) + 5 (0.25) + 6 (0.20) = 1000 (0.25) + 50 (0.18) + 82 (0.12)
+ 7(0.08) + 8 (0.04) + 128 (0.08) + 199 (0.05)
= 4.62 = 289
Average number of failures per week N6 = Np6 + N1 p5 + N2 p4 + N3 p3
1000 + N4 p2 + N5 p1
= = 216 = 1000 (0.20) + 50 (0.25) + 82 (0.18) +
4.62
Cost of individual replacement per week is 128 (0.12) + 199 (0.08) + 289 (0.05)
Rs 216 × 4 = Rs 864 = 272
(ii) Let Ni denote the total number of replace- N7 = Np7 + N1 p6 + N2 p5 + N3 p4 + N4 p3
ments made at the end of the i th week + N5 p2 + N6 p1
N = 1000 = 1000 (0.08) + 50 (0.20) + 82 (0.25)
N1 = Np1 = 1000 (0.05) = 50 + 128 (0.18) + 199 (0.12)
N2 = Np2 + N1p1 + 289 (0.08) + 272 (0.05)
= 1000 (0.08) + 50 (0.05) = 82 = 194
N3 = Np3 + N1p2 + N2 p1 N8 = Np8 + N1 p7 + N2 p6 + N3 p5 + N4 p4
= 1000 (0.12) + 50 (0.08) + 82 (0.05) + N5 p3 + N6 p2 + N7 p1
= 128 = 1000 (0.04) + 50 (0.08) + 82 (0.20)
N4 = Np4 + N1 p3 + N2 p2 + N3 p1 + 128 (0.25) + 199 (0.18)
= 1000 (0.18) + 50 (0.12) + 82 (0.08) + 289 (0.12) + 272 (0.08) + 194 (0.05)
+ 128 (0.05) = 195
= 199 Average cost at every week end is given by

End of Total cost of group Average cost

week replacement per week
1 1000 + 50 × 4 = 1200 1200
2 1000 + (50 + 82)4 = 1528 764
3 1000 + (50 + 82 + 128)4 = 2040 680 æÆ
4 1000 + (50 + 82 + 128 + 199)4 = 2836 709
5 1000 + (50 + 82 + 128 + 199 + 289)4 = 3992 798
6 1000 + (50 + 82 + 128 + 199 + 289 + 272)4 = 5080 847

We find that the weekly average cost is least at Week 1 2 3 4 5 6 7 8 9 10

the end of the 3rd week. Hence group replacement Prob. of .03 .04 .05 .06 .07 .08 0.09 0.16 0.20 0.22
is to be made at the end of 3rd week. failure
Note: Comparing the two types of replacement
The cost of repairing a broken machine is
we find that the total cost of individual replacement
Rs 200. Preventive maintenance service is done
for 3 weeks, is 3 × 864 = Rs 2592.
for all the 30 machines collectively at Rs 15 per
Under group replacement the total cost of
machine at the end of a period T. Find T so as to
replacement for 3 weeks is Rs 2040 only.
minimize the cost of maintenance.
Hence group replacement at the end of every
Solution Given N = 30. Let Ni denote the number
3 weeks is less expensive than individual
of service jobs done at the end of the ith week
replacement. N1 = Np1 = 30 (0.03) = 0.9 @ 1
Example 19.12 The following table gives the N2 = Np2 + N1p1 = 30(04) + 1 (0.03)
probability distribution of the failure time of a = 1.23 @ 1
machine. N3 = Np3 + N1 p2 + N2 p1
 Replacement Models 251

= 30 (0.05) + 1 (0.04) + 1 (0.03) N8 = Np8 + N1 p7 + N2 p6 + N3 p5 + N4 p4

= 1.57 @ + N5 p3 + N6 p2 + N7 p1
N4 = Np4 + N1p3 + N2 p2 + N3 p1 = 30 (0.16) + 1 (0.09) + 1 (0.08)
= 30 (0.06) + 1(0.05) + 1 (0.04) + 2 (0.07) + 2 (0.06) + 2 (0.05)
+ 2 (0.03)
+ 3 (0.04) + 3 (0.03)
= 1.95 @ 2
N5 = Np5 + N1 p4 + N2 p3 + N3 p4 + N4 p1 = 5.54 @ 6
= 30 (0.07) + 1 (0.06) + 1 (0.05) N9 = Np9 + N1 p8 + N2 p7 + N3 p6 + N4 p5
+ 2 (0.04) + 2 (0.03) + N5 p4 + N6 p3 + N7 p2 + N8 p1
= 2.35 @ 2 = 30 (0.20) + 1 (0.16) + 1 (0.09)
N6 = Np6 + N1 p5 + N2 p4 + N3 p3 + 2 (0.08) + 2 (0.07) + 2 (0.06)
+ N4 p2 + N5 p1 + 3 (0.05) + 3 (0.04) + 6 (.03)
= 30 (0.08) + 1 (0.07) + 1 (0.06) + 2 (0.05) = 7.12 @ 7
+ 2 (0.04) + 2(0.03) N10 = Np10 + N1 p9 + N2 p8 + N3 p7 + N4 p6
= 2.77 @ 3
+ N5 p5 + N6 p4 + Np4 + N8 p + N9 p1
N7 = Np7 + N1 p6 + N2 p5 + N3 p4 + N4 p3
+ N5 p2 + N6 p1 = 30 (0.22) + 1 (0.20) + 1 (0.16)
= 30 (0.09) + 1 (0.08) + 1 (0.07) + 2(0.06) + 2 (0.09) + 2 (0.08) + 2 (0.07)
+ 2 (0.05) + 2 (0.04) + 3 (0.03) + 3 (0.06) + 3 (0.05) + 6 (0.04)
= 3.24 @ 3 + 7 (0.03)
= 8.22 @ 8
End of week Total cost of maintenance Average cost (Rs)
1 (30 × 15) + (1)200 = 650 650
2 (30 × 15) + (1 + 1)200 = 850 425
3 (30 × 15) + (1 + 1 + 2)200 = 1250 417
4 (30 × 15) + (1 + 1 + 2 + 2)200 = 1650 412
5 (30 × 15) + (1 + 1 + 2 + 2 + 2)200 = 2050 410 æÆ
6 (30 × 15) + (1 + 1 + 2 + 2 + 2 + 3)200 = 2650 442
7 (30 × 15) + (1 + 1 + 2 + 2 + 2 + 3 + 3)200 = 3250 464
8 (30 × 15) + (1 + 1 + 2 + 2 + 2 + 3 + 3 + 6)200 = 4450 556
9 (30 × 15) + (1 + 1 + 2 + 2 + 2 + 3 + 3 + 6 + 7)200 = 5850 650
10 (30 × 15) + (1 + 1 + 2 + 2 + 2 + 3 + 3 + 6 + 7 + 8)200 = 7450 745
The average cost is minimum at the end of 5 weeks.
Hence group service should be done at the end of the 5th week.

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. The cost of a machine is Rs 6000. The resale Determine the best time for replacing the
value and maintenance costs every year are machine.
given below. 2. A machine costs Rs 6100. The scrap value is
Rs 100. The maintenance costs are given below:
Year 1 2 3 4 5 6
Year 1 2 3 4 5 6 7 8
Maintenance cost 1000 1200 1400 1800 2300 2800
Maintenance 100 250 400 600 900 1200 1600 2000
Resale value 3000 1500 750 325 200 200 cost
 252 Operations Research

When should the machine be replaced? 8. The following failure rates have been
3. The purchase price of a taxi is Rs 60000. Its observed for an electric bulb
resale value decreases by Rs 6000 per year. End of 1 2 3 4 5 6 7 8
The operating cost every year is given below: week
Year 1 2 3 4 5 Prob. of 0.10 0.15 0.20 0.35 0.65 0.90 0.95 1.00
Maintenance 1000 12000 15000 18000 20000 failure
cost The cost of replacing an individual bulb
After 5 years the operating cost is Rs 6000k is Rs 3 and the cost of group replacement
(k = 6, 7, 8, 9, 10,…) every year. What is the is Re 1 per bulb. Discuss whether group
best replacement policy? replacement is better than individual
4. A vehicle costs Rs 50000. Running costs and replacement, if there are 1000 bulbs.
resale values every year are given below: 9. Determine the optimal group replacement
Year 1 2 3 4 5 6 7 policy if there are totally 1000 bulbs and the
Running 5000 6000 7000 9000 12500 16000 18000
cost of individual replacement is Rs 5 and
cost the cost of group replacement is Rs 2 per
Resale 30000 15000 7500 3750 2000 2000 2000 bulb. The failure rate is given by the
value following table.
Thereafter running cost increases by Rs 2000, End of week 1 2 3 4 5 6
but resale value remains constant at Rs 2000. Prob. of failure 0.10 0.25 0.50 0.85 0.97 1.0
At what time is a replacement due?
5. The cost of a machine is Rs 5000. The Find the optimal group replacement policy.
maintenance cost during the nth year is given 10. A computer contains 10000 resistors. The
by cn = 500(n – 1) where n = 1, 2, 3,… cost of replacing a resistor is Rs 2. If all the
Suppose the discount rate is 5% per year. resistors are replaced at the same time it costs
After how many years will it be economical only Rs 0.75 per resistor. The percentage of
to replace the machine? resistors surviving at the end of each month
6. An equipment costs Rs 3000. The running is given by the following table:
costs are given below
End of month 1 2 3 4 5 6
Year 1 2 3 4 5 6 7 Percentage of survival 96 90 65 35 20 0
Running
cost 500 600 800 1000 1300 1600 2000
What is the optimal replacement plan?
11. The cost of a machine is Rs 60000. The
If money is worth 10% per year determine following table gives the data on running the
the optimal period of replacing the machine.
equipment.
Year 1 2 3 4 5
7. The cost of a machine A is Rs 5000.
The running costs are Rs 800 per year Resale value 42000 30000 20400 14400 9650
for each of the first five years, increasing Cost of spares 4000 4270 4880 5700 6800
by Rs 200 per year thereafter. Another Cost of labour 14000 16000 18000 21000 25000
machine B costs Rs 2500 but has Determine the optimum period of
running costs Rs 1200 per year for six replacement.
years, increasing by Rs 200 per year 12. Purchase price of machine A is Rs 200000.
thereafter. If money is worth 10% per year The particulars on running cost and resale
determine which machine should be value are given below:
purchased.
 Replacement Models 253

Year 1 2 3 4 5 6 7 16. A truck is priced at Rs 60000 and the

Running 30000 38000 46000 58000 72000 90000 110000 running costs are Rs 6000 for each of the
cost first four years, increasing by Rs 2000 per
Resale 100000 50000 25000 12000 8000 8000 8000 year during the subsequent years. If money
value is worth 10% per year determine when the
truck is to be replaced assuming that it has
(i) What is the optimum period of
no resale value.
replacement? 17. If it is required to get a return of 10% per
(ii) When the machine A is two years old, a annum on the investment, which of the
new model machine B is available. The following plans is preferable?
optimum period of replacement for B is
4 years with an average cost of Rs 72000. Plan A Plan B
Should A be replaced with B? If so (Rs) (Rs)
when? Cost 200000 250000
13. For a machine the following data are Scrap value after 15 years 150000 180000
available: Annual revenue 25000 30000

Year 1 2 3 4 5 6 18. There are 10000 resistors in a computer. The

cost of individual replacement of a resistor
Cost of spares 200 400 700 1000 1400 1600
Maintenance cost 1200 1200 1400 1600 2000 2600
is Re 1. For group replacement the cost is
Loss due to 600 800 700 1000 1200 1600
Rs 0.35 only. The percentage of survival of
breakdown the resistors at the end of the month t is given
Resale value 6000 3000 1500 800 400 400 below:
Month 0 1 2 3 4 5 6
If the present cost of the machine is Rs 12000,
find the optimum period of replacement. Percentage of 100 97 90 70 30 15 0
14. A company is considering the purchase of a survival
new machine at a cost of Rs 15000. The Determine the optimal replacement plan.
expected life of the machine is 8 years. The 19. The following failure rates have been
salvage value at the end of the life will be observed for an electric bulb.
Rs 3000. The annual running cost is Rs 7000. End of 1 2 3 4 5 6 7 8 9 10 11
Assuming the interest rate of 5%, calculate week
the present worth of the future cost after Prob. of 0.01 0.04 0.09 0.16 0.26 0.41 0.61 0.76 0.87 0.95 1.0
8 years. The presently owned machine has failure
an annual operating cost of Rs 5000. Its
Individual replacement costs Rs 1.25 and
maintenance cost is Rs 1500 in the second
group replacement costs Rs 0.50 per bulb.
year and increases at Rs 500 in the Determine the optimal group replacement
subsequent years. Compare the two machines policy.
to decide whether the new machine is to be 20. The following failure rates have been
purchased. observed for a certain type of fuse.
15. A manual stamper costs Rs 1000 and is
expected to last two years with an operating End of week 1 2 3 4 5
cost of Rs 4000 per year. An automatic Percentage of failure 5 15 35 57 100
stamper has a cost of Rs 3000 and expected There are 1000 fuses. Cost of individual
life of four years with an operating cost of replacement is Rs 5 and that of group
Rs 3000 per year. If money is worth 10% per replacement is Rs 1.25 per fuse. At what
annum determine which stamper should be intervals should the group replacement be
purchased. made?
 254 Operations Research
ANSWERS
■ ■ ■ ■ ■ ■ ■
1. At the end of 5th year.
2. At the end of 6th year.
3. At the end of first year.
4. At the end of 6th year.
5. At the end of 5th year.
6. At the end of 4th year.
7. It is advisable to purchase machine B.
8. Group replacement after 3 weeks is
advisable.
9. Individual replacement is cheaper Rs 1500
per week.
Group replacement may be done after
3 weeks. Total cost = Rs 4705.
10. Optimal plan is group replacement after
2 years.
11. At the end of the 4th year.
12. (i) At the end of the 5th year.
(ii) A should be replaced with B, when it is
four years old.
13. After 5 years.
14. The present worth of the new machine after
8 years is Rs 58212. The present worth of
the old machine after 8 years is Rs 44103.20.
Hence the new machine need not be
purchased.
15. Purchase an automatic stamper.
16. After 9 years.
17. Plan A.
18. Replace every 3 months.
19. After 6 weeks.
20. After every 2 weeks.

Project Scheduling
(PERT and CPM)
CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
20.1 INTRODUCTION
A project is a collection of inter-related activities
(tasks) which must be completed in a specified
time according to a specified sequence. A project
requires resources such as workers, materials,
money and facilities. Some examples of project
are: construction of a bridge, highway power plant,
maintenance of oil refinery or an aeroplane,
marketing a new product etc. Our aim is to develop
a sequence of activities of the project so that
the project completion time and cost are properly
balanced and kept at the optimum level. There
are two techniques used for project scheduling.
They are
(i) Critical Path Method (CPM) and
(ii) Programme Evaluation and Review
Technique (PERT)
These two techniques are applied using network
diagrams.
20.2 NETWORK DIAGRAMS
All the techniques of project completion start with
the representation of the project as a network of
20
activities. Networks are diagrams representing a
sequence of activities involved in the project
work. Network helps us to minimize the trouble
spots, delays etc., in the various stages of the
project.
20.2.1 Activity
Activity is a part of the project, which takes time,
effort, money and other resources. In the network
an activity is represented by an arrow whose tail
represents the start and head represents the end of
the activity.
20.2.2 Event (Node)
The beginning point and the end point of an
activity are called events (nodes). It is represented
by a number in a circle. The number of the head
node is always greater than that of the tail node. A
network is a combination of activities and events
involved in a project.
20.2.3 Network Construction
From the given set of activities we decide the
beginning and the ending activities. According to
the precedence order the activities are put in a
logical sequence. If an activity A is to be completed
 256 Operations Research

before starting another activity B then we write Example 20.1 Draw a network diagram for the
A < B. A is preceding B or B is succeeding A. Two following set of activities:
activities A and B can also be performed
A < B, C; B < D, E; C < D; D < F; E < G; F < G
simultaneously. Suppose A, B < C and B < E then
it is represented by introducing a dummy activity Solution
B1 as follows:
A C 5
3
E
C C1 E1
B1
1 2 4 6 7 8
A B D F G
B E
The network diagram of the project having the In this diagram C < D. Hence we introduce
activities A, B, C, D, E, F, G where A < B, C; B, C a dummy activity C1. Similarly, E < G. Therefore
< D; B < E, F; D, F < G is given by another dummy activity E1 is introduced.
D
Example 20.2 Draw a network diagram for the
C following set of activities:
B1 G
F
A < B, C; B < D, E; C < E; E < F; D, F < G;
A G<H
B E
Solution
There is no predecessor for A and hence A is
the beginning activity. B1 is a dummy activity.
There is no successor for E and G. Hence E and G
are the final activities.
20.2.4 Fulkerson’s Rule
After the network is drawn with the given activities
in their order we have to assign numbers to the (B1 is dummy)
events (nodes). The numbering is to be done
according to Fulkerson’s rule which gives the Example 20.3 Draw a network for the following
following steps: set of activities
(i) The initial node which has only outgoing Activity A B C D E F G H I
arrows is assigned (1). Immediate – – – A B,C A C D,E,F D
(ii) Delete all the arrows going out from (1). We predecessor
get a set of initial nodes. Assign numbers (2), Solution
(3), (4) etc. to these nodes.
G
(iii) Follow this method to the nodes (2) (3) (4)... 2 7
C1
and assign successive numbers to the newly
obtained nodes. C E H
(iv) Finally we are left with a single node with 3 6 I
1 B
no arrow going out. Assign the last number
F D1
to this end node. A D 5
(v) Always the numbers of the head node must 4
be greater than that of the tail node of each
activity. (C1 and D1 are dummy)
 Project Scheduling (PERT and CPM)
Note: The time taken for completing the activity
(i, j) is given by tij. In the network diagram tij is
marked by the side of the arrow representing the
activity (i,j).
20.3 CRITICAL PATH METHOD
20.3.1 Time Calculation
There are some terms which are associated with
the activities of a project. The Earliest Starting
Time of the activity (i, j) is denoted by ESi and the
Earliest Finishing Time of the activity (i, j) is
denoted by ESj. For the activity (i, j), ESj = ESi + tij.
If more than one activity end at the node (j) then
ESj = max {ESi + tij}.
i
tij
( j)
For the initial activity ES1 = 0. If (n) is the last
node then ES n represents the time taken to
complete the whole project. Also, ESn = EFn.
Starting from the last node we proceed
backward to calculate the latest times. The Latest
Starting Time of the activity (i, j) is denoted by
LSi and the Latest Finishing Time of the activity
(i, j) is denoted by LSj. For the activity (i, j),
LSi = LSj – tij . If more than one activity start from
(i) then LSi = min {LSj – tij}
j and form the critical path by double line.
tij
(i)
For the last node (n) ESn = LSn. For the activities
(i, j) and (j, k) the earliest starting time ESj of (j, k)
is the same as the earliest finishing time EFj of the
activity (i, j). Similarly, LSj = LFj.
An activity is called critical activity if a delay
in starting that activity will cause further delay in
completing the project. For a critical activity (i, j)
the following equalities hold
(i) ESi = LSi
257
(ii) ESj = LSj
(iii) ESj = ESi = LSj – LSi = tij
Total float for an activity (i, j) is the difference
between the maximum time available to finish the
activity and the time required to complete it.
[Total float is simply called as Float.] Total float
= LSj – ESi – tij
Free float is the time by which an activity can
be delayed beyond its earliest finish time without
affecting the earliest start time of a succeeding
activity
Free float = Total float – (LSj – ESj)
Independent Float = Total float – (LSi – ESi)
The path formed by all the critical activities is
called the critical path.
20.3.2 Step-by-Step Procedure
to Find the Critical Path
1. Draw the network diagram of the project and
indicate the time for each activity.
2. Calculate the earliest starting time at each
node.
3. Calculate the latest starting time at each node.
4. Prepare a table giving the time tij, ESi, ESj,
LSi, and LSj for each activity.
5. Find the difference between the earliest time
ESi and the latest time LSi. Note down the
difference (float) LSi – ESi, (LFj – EFj) for
each activity.
6. Choose the critical activities with zero float
7. The sum Stij of the critical activities gives
the time of completion of the project. (ESn
also denotes the time of completion of the
project.)
Example 20.4 A project consists of a series of
jobs A, B, C, D, E, F, G, H, I such that A < D, E; B,
D < F; C < G; B < H; F, G < I
The time of completion of each job is given
below:
Job A B C D E F G H I
Time 23 8 20 16 24 18 19 4 10
(days)
 258 Operations Research

Find the critical path and the minimum time Job Time Earliest time Latest time Float
required to complete the project. Start Finish Start Finish
Solution The network of the given project is as ESi ESj LSj ESj
shown below:
(1, 2) 23 0 23 0 23 0Æ
(1, 3) 8 0 8 31 39 31
(1, 4) 20 0 20 18 38 18
(2, 3) 16 23 39 23 39 0Æ
(2, 7) 24 23 47 43 67 20
(3, 5) 0 39 39 57 57 18
(3, 6) 18 39 57 39 57 0Æ
(4, 5) 19 20 39 38 57 18
(5, 6) 0 39 39 57 57 18
(5, 7) 4 39 43 63 67 24
(6, 7) 10 57 67 57 67 0Æ

D1 and D2 are dummy activities with tij = 0. From the table we find that the critical activities
The earliest and latest starting times at the nodes are (1, 2), (2, 3), (3, 6) and (6, 7). Hence the critical
are calculated as given below. path is 1 – 2 – 3 – 6 – 7. The minimum time
ES1 = 0 ES2 = ES1+t12 = 0+23 = 23 of completion of the project is 23 + 16 + 18 +
10 = 67 days.
ES3 = Max [ES1 + t13, ES2 + t23]
= Max [0 + 8, 23 + 16] = 39 Example 20.5 A project consists of jobs A, B,
ES4 = ES1 + t14 = 0 + 20 = 20 C, D, E, F, G, H, I such that A < D; A < E; B < F;
ES5 = Max [ES4 + t45, ES3 + t35] D < F; C < G; C < H; F < I; G < I
= Max [20 + 19, 39 + 0] = 39 The time taken for each job is given below:
ES6 = Max [ES3 + t36, ES5 + T56] Job A B C D E F G H I
= Max [39 + 18, 39 + 0] = 57 Time 8 10 8 10 16 17 18 14 9
ES7 = Max [ES2 + t27, ES6 + T67, ES5 = t57] (days)
= Max [23 + 24, 57 + 10, 39 + 4] = 67
Now, Draw the network diagram. Find the critical
LS7 = ES7 = 67 path and minimum time of completion of the
LS6 = LS7 – t67 = 67 – 10 = 57 project.
LS5 = Min [LS6 – t56’ LS7 – t57] Solution The network diagram is given below:
= Min [57 – 0, 67 – 4] = 57
LS4 = LS5 – t45 = 57 – 19 = 38
LS3 = Min [LS6 – t36’ LS5 – t35]
= Min [57 – 18, 57 – 0] = 39
LS2 = Min [LS3 – t23, LS7 – t27]
= Min [39 –16, 67 – 24] = 23
LS1 = Min [LS4 – t14, LS3 – t13, LS2 – t12]
= Min [38 –20, 39 – 8, 23 – 23] = 0
To find the critical nodes we prepare the table
as given below:
(H1 is a dummy activity)
[EFj = ESi + tij; LSi = LFj – tij = LSj – tij The earliest and latest times are calculated as
Float = LSi – ESi = LSj – ESj] follows:
 Project Scheduling (PERT and CPM) 259

ES1 = 0 ES2 = ES1 + t12 = 0 + 8 = 8 The table giving the critical activities is
ES3 = Max [ES1 + t13, ES2 + t23]
Job Time Earliest time Latest time Float
= Max [0 + 10, 8 + 10] = 18
ES4 = ES1 + t14 = 0 + 8 = 8 Start Finish Start Finish
ES5 = Max [ES2 + t25, ES4 + t45] ESi ESj LSi LSj
= Max [8 + 16, 8 + 14] = 24 (1, 2) 8 0 8 0 8 0 Æ
ES6 = Max [ES4 + t46, ES3 + t36] (1, 3) 10 0 10 8 18 8
= Max [8 + 18, 18 + 17] = 35 (1, 4) 8 0 8 9 17 9
ES7 = Max [ES5 + t57, ES6 + t67] (2, 3) 10 8 18 8 18 0 Æ
= Max [24 + 0, 35 + 9] = 44 (2, 5) 16 8 24 28 44 20
ES7 = ES7 = 44 (3, 6) 17 18 35 18 35 0 Æ
LS6 = LS7 – t67 = 44 – 9 = 35 (4, 6) 18 8 26 17 35 9
(4, 5) 14 8 22 30 44 22
LS5 = LS7 – t57 = 44 – 0 = 44
(5, 7) 0 24 24 44 44 20
LS4 = Min [LS6 – t46, LS5 – t45] (6, 7) 9 35 44 35 44 0 Æ
= Min [35 – 18, 44 – 14] = 17
LS3 = LS6 – t36 = 35 – 17 = 18 From the table we find that the critical path is
LS2 = Min [LS5 – t25, LS3 – t23] 1 – 2 – 3 – 6 – 7.
= Min [44 – 16, 18 – 10] = 8 Minimum time of completion is 44 units.
LS1 = Min [LS4 – t14’ LS3 – t13, LS2 – t12] Example 20.6 Find the critical path of the
= Min [17 – 8, 18 – 10, 8 – 8] = 8 project with the following activities

Job (1, 2) (1, 3) (2, 4) (3, 4) (3, 5) (4, 9) (5, 6) (5, 7) (6, 8) (7, 8) (8, 10) (9, 10)
Time 4 1 1 1 6 5 4 8 1 2 5 7

Solution The network diagram of the project is ES9 = ES4 + t49 = 10

given below: ES10 = Max [10 + 7, 17 + 5] = 22
Again,
LS10 = 22
LS9 = LS10 – t9,10 = 15
LS8 = LS10 – t8,10 = 17
LS7 = LS8 – t7,8 = 15
LS6 = LS8 – t68 = 16
ES1 = 0 LS5 = Min [16 – 4, 15 – 8] = 7
ES2 = ES1 + t12 = 4 LS4 = LS9 – t49 = 10
ES3 = ES1 + t13 = 1
LS3 = Min [10 – 1, 7 – 6] = 1
ES4 = Max [4 + 1, 1 + 1] = 5
LS2 = LS4 – t24 = 9
ES5 = ES3 + t35 = 7
LS1 = Min [9 – 4, 1 – 1] = 0
ES6 = ES5 + t56 = 11
ES7 = ES5 + t57 = 15 The following table gives the slack time for
ES8 = Max [11 + 1, 15 + 2] = 17 each activity.
 260 Operations Research

Job Time Earliest time Latest time Float ES9 = Max [23 + 4, 19 + 1] = 27
ESi ESj Start Finish ES10 = Max [15 + 4, 27 + 7] = 34
LSi LSj LS10 = 34
(1, 2) 4 0 4 5 9 5 LS9 = 34 – 7 = 27
(1, 3) 1 0 1 0 1 0 Æ LS8 = 27 – 1 = 26
(2, 4) 1 4 5 9 10 5
(3, 4) 1 1 2 9 10 8 LS7 = 27 – 4 = 23
(3, 5) 6 1 7 1 7 0 Æ LS6 = Min [30, 15] = 15
(4, 9) 5 5 10 10 15 5
LS5 = 26 – 5 = 21
(5, 6) 4 7 11 12 16 5
(5, 7) 8 7 15 7 15 0 LS4 = 15 – 3 = 12
(6, 8) 1 11 12 16 17 5 LS3 = Min [16, 9] = 9
(7, 8) 2 15 17 15 17 0 Æ
(8, 10) 5 17 22 17 22 0 Æ LS2 = Min [2, 9] = 2
(9, 10) 7 10 17 15 22 5 LS1 = 2–2=0
The critical path is Job Time Earliest time Latest time Float
1 – 3 – 5 – 7 – 8 – 10
Time of completion is 22 days. Start- Finish- Start- Finish-
ESi ESj LSi LSj
Example 20.7 Find the critical path of a project
having the tasks as given below: (1, 2) 2 0 2 0 2 0 Æ
(2, 3) 7 2 9 2 9 0 Æ
Job Time Job Time (2, 4) 3 2 5 9 12 7
(1, 2) 2 (5, 8) 5 (3, 4) 3 9 12 9 12 0 Æ
(3, 5) 5 9 14 16 21 7
(2, 3) 7 (6, 7) 8 (4, 6) 3 12 15 12 15 0 Æ
(2, 4) 3 (6, 10) 4 (5, 8) 5 14 19 21 26 7
(3, 4) 3 (7, 9) 4 (6, 7) 8 15 23 15 23 0 Æ
(6, 10) 4 15 19 30 34 15
(3, 5) 5 (8, 9) 1
(7, 9) 4 23 27 23 27 0 Æ
(4, 6) 3 (9, 10) 7 (8, 9) 1 19 20 26 27 7
(9, 10) 7 27 34 27 34 0 Æ
Solution The network diagram is
The critical activities are (1, 2), (2, 3), (3, 4),
4 10 (4, 6), (6, 7), (7, 9), (9, 10). The critical path is 1 –
3 2 – 3 – 4 – 6 – 7 – 9 – 10
4 6 7
3 3 8 Time of completion is 34 days
7 14
Example 20.8 Find the critical path of the
2 7 5 5 5 1
1 2 3 8 9 project represented by the following network:
3 3 5 5 7
ES1 = 0
ES2 = 0+2=2
ES3 = 2+7=9 4 5 2 4
ES4 = Max [5, 12] = 12
ES5 = 9 + 5 = 14
1 1 2 5 4 4 6
ES6 = 12 + 3 = 15
ES7 = 15 + 8 = 23 Solution The earliest and latest times are
ES8 = 14 + 5 = 19 calculated as follows:
 Project Scheduling (PERT and CPM) 261

ES1 = 0; ES2 = 1; ES3 = 5 (i) Draw the network diagram

ES4 = 6; ES5 = Max [8, 11] = 11 (ii) Determine the critical path
ES6 = Max [13, 10] = 13 (iii) Find the total float, free float and the
independent float of the activities
ES7 = Max [16, 17] = 17
Solution The network diagram for the given
LS7 = 17; LS6 = 13 project is given below:
LS5 = Min [12, 11] = 11
LS4 = Min [6, 9] = 6; LS3 = 8
4
LS2 = Min [4, 1] = 1; LS1 = 0 J
13 H 6
Job Time Earliest time Latest time Float 10 I
Start Finish Start Finish LSi–ESi G 3
2
ESi ESj LSi LSj 6 K 9 L M
1 2 6 9 10
(1, 2) 1 0 1 0 1 0 Æ A B 7 3 5
4 C
(2, 3) 4 1 5 4 8 3 5 F
(2, 4) 5 1 6 1 6 0 Æ 10
2 D
(3, 5) 3 5 8 8 11 3 8
(4, 5) 5 6 11 6 11 0 Æ 7 E 4
(4, 6) 4 6 10 9 13 3
(5, 6) 2 11 13 11 13 0 Æ The earliest and latest times are calculated as
(5, 7) 5 11 16 12 17 1 follows:
(6, 7) 4 13 17 13 17 0 Æ ES1 = 0; ES2 = 0 + 6 = 6
ES3 = 0 + 2 = 2
The critical path is 1 – 2 – 4 – 5 – 6 – 7. Time of ES4 = Max [ 0 + 13, 2 + 10] = 13
completion is 17 units of time. ES5 = 6 + 4 = 10
Example 20.9 A company plans the following ES6 = Max [ 6 + 9, 10 + 7] = 17
activities promoting its business: ES7 = 6 + 2 = 8
Activity Description Time Preceding ES8 = 8 + 4 = 12
(Weeks) activity ES9 = Max [13 + 6, 17 + 3] = 20
ES10 = Max [12 + 10, 20 + 5] = 25
A Organize sales office 6 -
LS10 = 25
B Hire salesmen 4 A
C Train salesmen 7 B
LS9 = 25 – 5 = 20
D Select advertising agency 2 A
LS8 = 25 – 10 = 15
E Plan advertising campaign 4 D LS7 = 15 – 4 = 11
F Conduct campaign 10 E LS6 = 20 – 3 = 17
G Design package 2 - LS5 = 17 – 7 = 10
H Set up packing facilities 10 G LS4 = 20 – 6 = 14
I Packing initial stocks 6 H, J LS3 = 14 – 10 = 4
J Place order with manufacturer 13 - LS2 = Min [ 10 – 4, 17 – 9, 11 – 2] = 6
K Select distributors 9 A LS1 = Min [ 6 – 6, 4 – 2, 14 – 13] = 0
L Sell to distributors 3 C, K The following table gives the critical path and
M Transport the stocks 5 I, L the float times:
 262 Operations Research

Job (i, j) tij Earliest time Latest time Total float Free float Independent
ESi ESj LSi LSj LSi – ESi float
(1, 2) 6 0 6 0 6 0 Æ 0 0
(1, 3) 2 0 2 2 4 2 0 2
(1, 4) 13 0 13 1 14 1 0 1
(2, 5) 4 6 10 6 10 0 Æ 0 0
(2, 6) 9 6 15 8 17 2 2 2
(2, 7) 2 6 8 9 11 3 0 3
(3, 4) 10 2 12 4 14 2 1 0
(4, 9) 6 13 19 14 20 1 1 0
(5, 6) 7 10 17 10 17 0 Æ 0 0
(6, 9) 3 17 20 17 20 0 Æ 0 0
(7, 8) 4 8 12 11 15 3 0 0
(8, 10) 10 12 22 15 25 3 3 0
(9, 10) 5 20 25 20 25 0 Æ 0 0

The critical path is 1 – 2 – 5 – 6 – 9 – 10. The
time of completion is 25 weeks.
Notes:
(i) The difference between slack time and float
time is that slack is used for nodes whereas
float is applied for activities.
(ii) Total float of an activity (i, j) is LSi – ESi
(Slack at i) or LSj – ESj (Slack at j)
(iii) Free float = Total float – Slack time at j.
(iv) Independent float = Total float – Slack time
at i
(v) Negative value of a float is considered as
zero.
20.4 PERT CALCULATIONS
20.4.1 CPM and PERT
The following facts help to distinguish CPM and
PERT calculations.
(i) In CPM the time of an activity is determinis-
tic. But in PERT the time for each activity is
probabilistic. We are given three different
time estimates to = optimistic time (shortest
possible time to complete the activity), tm =
most likely time (normal time the activity
would take in general), tp = pessimistic time
(longest time taken for the activity if things
go wrong).We find the expected time te for
each activity using the formula
1
te =
6
(to + 4tm + t p ) times of an activity, we find the earliest expected
Variance of the distribution of time for each
activity is calculated using the formula
2
æ t p - to ö
s ij2 =ç
è 6 ÷ø
(ii) In CPM we find the minimum time required
to complete the project without considering
the cost involved. But in PERT the cost
involved in each activity is also taken into
account. The normal time required for each
activity can be crashed to the least time
required at an increased cost. We crash the
activities in a systematic way and find the
least possible time to complete the project at
an optimum cost.
(iii) In CPM, the time calculations are activity
oriented. But in PERT each node represents
an event and the calculations are based on
these events.
(iv) In CPM the time of completion of the project
is of deterministic nature where as in PERT
we can find only the probability of
completion of the project within a stipulated
time.
20.4.2 Time Calculations and
Critical Path
Instead of calculating the starting and finishing
 Project Scheduling (PERT and CPM) 263

time E(mi) of the event (i) and the latest expected Time calculations
time E(Li) of the event (i) E(mi) is found as in the E(m1) = 0
case of ESi in CPM. E ( mi ) = Max [ E ( mi ) + tij ]
i E(m2) = 0 + 3 = 3
Similarly E ( Li ) = Min [ E ( L j ) - tij ] . If E(mi) = E(m3) = 0 + 5 = 5
j
E(Li) then (i) is a critical node. The path joining E(m4) = 0 + 8 = 8
the critical nodes is the critical path. E(m5) = 3 + 7 = 10
Example 20.10 A project is represented by the E(m6) = Max [5 + 9, 3 + 10] = 14
network given below and has the following data:
E(m7) = Max [10 + 8, 14 + 8, 8 + 11] = 22
Task A B C D E F G H I
E(L7) = 22; E(L6) = 22 – 8 = 14
Optimistic time to 2 3 5 5 8 7 9 3 6 E(L5) = 22 – 8 = 14
Most likely time tm 3 5 8 7 10 9 11 8 8
Pessimistic time tp 4 7 11 9 12 11 13 13 10
E(L4) = 22 – 11 = 11
E(L3) = 14 – 9 = 5
3 F E(L2) = Min [14 – 10, 14 –7] = 4
9 I
B 6 7 E(L1) = Min [11 – 8, 5 – 5, 4 – 3] = 0
E 8
5 10 D 8 H
G
3 2 7
5 11 Event E(mi) E(Li) Slack
A
1 4 1 0 0 0 Æ
8 C 2 3 4 1
Determine the following: 3 5 5 0 Æ
(i) Expected task times, te 4 8 11 3
(ii) Variance of the tasks, s 2 5 10 14 4
(iii) Critical path 6 14 14 0 Æ
Solution We calculate the expected time and the 7 22 22 0 Æ
variance of each task as given below:
2 The critical path is 1 – 3 – 6 – 7. The expected
1 æ t p - to ö time of completion is 22 units.
te = (to + 4tm + t p ) s 2 = ç
6 è 6 ÷ø Example 20.11 Find the critical path of the
project given by the following network
Task t0 tm tp te s2
diagram
(1, 2) 2 3 4 3 0.111
(1, 3) 3 5 7 5 0.444
(1, 4) 5 8 11 8 1
(2, 5) 5 7 9 7 0.444
(2, 6) 8 10 12 10 0.444
(3, 6) 7 9 11 9 0.444
(4, 7) 9 11 13 11 0.444
(5, 7) 3 8 13 8 2.778
(6, 7) 6 8 10 8 0.444
 264 Operations Research

Job (1, 2) (1, 3) (1, 5) (2, 3) (2, 4) (2, 5) (3, 4) (3, 7) (4, 6) (4, 7) (5, 6) (6, 7)
to 2 2 5 6 9 5 4 6 12 7 3 10
tm 4 5 8 7 10 9 8 9 13 10 6 13
tp 6 8 11 8 11 13 12 12 14 13 9 16

Solution The time calculations are given The expected time of completion of the project
below: is 45 units.
1 20.4.3 Probability of Completion
te = (to + 4tm + t p )
6 Within a Specified Time
Job to tm tp te In PERT since the time estimates are based on
(1, 2) 2 4 6 4 probability, the project may not be completed
(1, 3) 2 5 8 5 exactly in time. We can calculate the probability
(1, 5) 5 8 11 8 of completing the project within any scheduled
(2, 3) 6 7 8 7 time. It is assumed that the probability distribution
(2, 4) 9 10 11 10 of the activity times is normal distribution. The
(2, 5) 5 9 13 9 mean of the distribution is T (expected time of
(3, 4) 4 8 12 8 completion) and the variance is s 2 = Ss ij2 where
(3, 7) 6 9 12 9
s ij2 are the variances of the critical activities. We
(4, 6) 12 13 14 13
(4, 7) 7 10 13 10 can calculate the probability of completing the
(5, 6) 3 6 9 6 project in a given time t as follows:
(6, 7) 10 13 16 13
t -T
Define Z1 = . Now Z is a standard
E( m1) = 0; E( m2) = 4 s2
E( m3) = Max [4 + 7, 0 + 5] = 11 normal variate and hence we can use the table of
E( m4) = Max [4 + 10, 11 + 8] = 19 area under normal curve to find P [Z £ Z1]. This is
E( m5) = Max [0 + 8, 4 + 9] = 13 the required probability.
E( m6) = Max [19 + 13, 13 + 6] = 32 Example 20.12 A project consists of the
E( m7) = Max [19 + 10, 11 + 9, 32 + 13] = 45 following activities and time estimates.
E( L7) = 45; E(L6) = 32; E(L5) = 26 Activity Estimated duration in weeks
E( L4) = Min [45 – 10, 32 – 13] = 19 Optimistic Most likely Pessimistic
E( L3) = Min [45 – 9, 19 – 8] = 11
(1, 2) 1 1 7
E( L2) = Min [11 – 7, 19 – 10, 26 – 6] = 4 (1, 3) 1 4 7
E( L1) = Min [4 – 4, 11 – 5, 26 – 8] = 0 (1, 4) 2 2 8
(2, 5) 1 1 1
Event E(mi) E(Li) Slack
(3, 5) 2 5 14
1 0 0 0 Æ (4, 6) 2 5 8
2 4 4 0 Æ (5, 6) 3 6 15
3 11 11 0 Æ
4 19 19 0 Æ (i) Draw the network.
5 13 26 13 (ii) Find the expected time and variance for each
6 32 32 0 Æ activity.
7 45 45 0 Æ (iii) What is the probability that the project will
be completed 4 weeks earlier than the
The critical path is 1 – 2 – 3 – 4 – 6 – 7 expected time.
 Project Scheduling (PERT and CPM) 265

(iv) What is the probability that the project will t - 17

be completed in 19 weeks. We have Z =
Solution The network is given below: 3
Given t = 17 – 4 = 13
2
2 1 13 - 17
\ Z1 = = -1.33
1 3 5 6 3
4 6 7
3 5
4
The expected time and variance of each activity
is given below:
Activity to tm tp te s2
(1, 2) 1 1 7 2 1
Z1 0 1.33
(1, 3) 1 4 7 4 1 –1.33
(1, 4) 2 2 8 3 1
(2, 5) 1 1 1 1 0 From the table of area under normal curve we
(3, 5) 2 5 14 6 4 have P [Z £ 1.33] = 0.91 = 0.5 + 0.41
(4, 6) 2 5 8 5 1 \ P [Z £ – 1.33] = 0.5 – 0.41 = 0.09
(5, 6) 3 6 15 7 4
(iv) To find the probability that the project would
The following are the time calculations be completed in 19 weeks:
E(m1) = 0; E(m2) = 2; E(m3) = 4
19 - 17
E(m4) = 3; E(m5) = Max (2 + 1, 4 + 6) = 10 We have Z 2 = = 0.67
E(m6) = Max {10 + 7, 3 + 5} = 17 3
E(L6) = 17; E(L5) = 10; E(L4) = 12
E(L3) = 4; E(L2) = 9;
E(L1) = Min {9 – 2, 4 – 4, 12 – 3} = 0
The critical events are given in the following
table:
Event E( mi) E( Li) Slack
1 0 0 0 Æ 0 Z2
2 2 9 7 0.67
3 4 4 0 Æ
4 3 12 9 From the table we find that
5 10 10 0 Æ \ P [Z £ 0.67] = 0.75
6 17 17 0 Æ
Thus the project would be completed 4 weeks
The critical events are 1, 3, 5, 6. Hence the earlier with a probability of 0.09 and in 19 weeks
critical path is 1 – 3 – 5 – 6. with a probability of 0.75.
Expected time of completion of the project is
17 weeks \ T = 17 Example 20.13 For the project represented by
s 2 = s1,3
2 2
+ s3,5 2
+ s5,6 =1+4+4=9 the following network, find the probability that
\s=3 the project will be completed
(iii) To find the probability that the project would (i) two weeks earlier than expected.
be completed 4 weeks earlier: (ii) two weeks later than expected.
 266 Operations Research

11-14-17 10-13-16 Event E(mi) E(Li) Slack

3 5 7 4-7-10
-11

1 0 0 0 Æ
7-9

6-9-12 9 10 2 7 7 0 Æ
2-4-6 0-0-0 6-7.5-12 8-13-18
3 11 11 0 Æ
1 2 4 6 8
6-7-8 8-12-16 3-4-5 5-9-13 4 19 23 4
5 25 25 0 Æ
Solution The time estimates to, tm, and tp are 6 23 27 4
given in the network diagram. The expected time 7 38 38 0 Æ
and variance of each activity is given by the 8 33 36 3
following table:
9 45 45 0 Æ
Activity to tm tp te Variance s 2 10 58 58 0 Æ

(1, 2) 6 7 8 7 0.111 The critical path is 1 – 2 – 3 – 5 – 7 – 9 – 10

(1, 3) 7 9 11 9 0.444 Expected time of completion of the project is
(2, 3) 2 4 6 4 0.444 58 weeks.
(2, 4) 8 12 16 12 1.778 s 2 = s 212 + s 22,3 + s 23,5 + s 25,7 + s 27,9 + s 29,10
(3, 4) 0 0 0 0 0 = 0.111 + 0.444 + 1 + 1 + 1 + 2.778 = 6.333
(3, 5) 11 14 17 14 1 (i) T = 58 weeks. Given t = 58 – 2 = 56 weeks.
(4, 6) 3 4 5 4 0.111
(5, 7) 10 13 16 13 1 s= 6.333 = 2.52
(5, 8) 6 7.5 12 8 1
t -T 56 - 58
(6, 8) 5 9 13 9 1.778 Z1 = = = - 0.79
(7, 9) 4 7 10 7 1 s 2.52
(8, 9) 6 9 12 9 1
(9, 10) 8 13 18 13 2.778
Now let us calculate the earliest and latest times
at each node.
E( m1) = 0; E( m2) = 7
E( m3) = Max [9, 11] = 11
E( m4) = Max [11 + 0, 7 + 12] = 19 –0.79 0 0.79
E( m5) = 25; E( m6) = 23 From the table we find that
E( m7) = 38 P [Z £ 0.79] = 0.7852 = 0.5 + 0.2852
E( m8) = Max [33, 32] = 33 P [Z £ –0.79] = 0.50 – 0.2852 = 0.2148
Probability that the project would be
E( m9) = Max [45, 42] = 45 completed two weeks earlier than expected,
E( m10) = 58 is 0.2148
E( L10) = 58; E( L9) = 45; E( L8) = 36 (ii) T = 58 weeks. Given t = 58 + 2 = 60 weeks.
t -T 60 - 58
E( L7) = 38; E( L6) = 27 Z2 = = = 0.79
E( L5) = Min [25, 28] = 25 s 2.52
From the table, P [Z £ 0.79] = 0.7852
E( L4) = 23; E( L3) = Min [11, 23] = 11 Probability that the project would be completed
E( L2) = Min [11, 7] = 7 two weeks later than expected, is 0.7852.
E( L1) = Min [0, 2] = 0 Example 20.14 The activities and their estimates
To find the critical path: of time of a project are given below:
 Project Scheduling (PERT and CPM) 267

A B C D E F G H I J K s 2 = s 21,4 + s 24,5 + s 25,6 + s 6,7

2

= 16 + 4 + 1 + 1.36 = 22.36
to 3 2 6 2 5 3 3 1 4 1 2
tm 6 5 12 5 11 6 9 4 19 3 4 \ s = 4.73
tp 12 14 30 8 17 15 27 7 28 8 12 Now T = 32.5 days. Given t = 35 days.
Precedence relationships are t -T 35 - 32.5
Z1 = = = 0.53
A < D, I; B < G, F; D < G, F; C < E; E < H, K s 4.73
F < H, K; G, H < J.
What is the probability that the project will be
completed in 35 days?
Solution The network diagram is given below

18 7
I J
2 K
6 3.5
A 6.5 D G 11 4 5
0 0.53
5 H
B F
1 6 3 5
7 From the table we find that
14 C E 11 P [Z £ 0.53] = 0.702
4 \ Probability that the project would be
completed in 35 days, is 0.702.
The expected time and variance of each job is
20.4.4 Cost Considerations in
given by the following table:
PERT
Job to tm tp te Variance
In PERT calculations the cost involved in
A (1, 2) 3 6 12 6.5 2.25 completing the project is also considered. There
B (1, 3) 2 5 14 6 4 are two kinds of costs, Direct cost and Indirect
C (1, 4) 6 12 30 14 16 cost. Direct costs are the costs associated with each
D (2, 3) 2 5 8 5 1 activity such as machine cost, labour cost, etc.
I (2, 7) 4 19 28 18 16
Direct cost varies inversely as the duration of the
F (3, 5) 3 6 15 7 4
G (3, 6) 3 9 27 11 16 activity. It increases when the time of completion
E (4, 5) 5 11 17 11 4 of the job is to be reduced (crashed) since more
H (5, 6) 1 4 7 4 1 machines and more labour, become necessary to
K (5, 7) 2 4 12 5 2.78 complete the job in less time. Indirect costs are
J (6, 7) 1 3 8 3.5 1.36 the costs due to management services, rentals, cost
of security, establishment charges and similar over-
Now E(m1) = 0; E(m2) = 6.5
head expenditures. When the duration of the
E(m3) = 11.5; E(m4) = 14 project is shortened, indirect cost decreases.
E(m5) = 25; E(m6) = 29 Therefore there is some optimum project duration
E(m7) = 32.5; E(L7) = 32.5 which is a balance between the direct costs
E(L6) = 29; E(L5) = 25; increasing with reducing of project duration and
E(L4) = 14; E(L3) = 18; the indirect costs increasing with the lengthening
E(L2) = 13; E(L1) = 0 of the duration. This method of finding the
The critical path is 1– 4 – 5 – 6 – 7 optimum duration is called Least Cost Schedule.
The expected time of completion of the project For each activity a normal time and normal cost
is 32.5 days is known. Also the crash time and the corres-
 268 Operations Research

ponding crash cost are also given. In order to The critical path is 1 – 2 – 5 (Longest path)
reduce the duration we crash the critical activities Time of completion = 18 days
one by one and obtain the optimum duration. Direct cost (Normal cost) = Rs 580
Indirect cost = 18 × 70 = Rs 1260
Step-by-step procedure Total cost = 580 + 1260 = Rs 1840
(i) Draw the network diagram.
(ii) Determine the critical path, normal duration Crashing First we crash the critical activity
and total normal cost of the project. (1, 2) having the least cost slope 50. The new
(iii) Find the cost slope of each activity using the network is given below:
formula 10
2 5
æ Crash cost – Normal cost ö 6
Cost slope = çè ÷
Normal time – Crash time ø 2
3
(iv) Crash the critical activity having the least
1
slope. 4 3 4
5
(v) Calculate the revised total cost [Revised
direct cost + new indirect cost] for the The critical path remains the same with revised
reduced duration. duration 16 days
(vi) Determine the new critical path and repeat Direct cost = Rs 680 (580 + 100)
the process of crashing systematically till an Indirect cost = 16 × 70 = Rs 1120
optimum duration is obtained. Total cost = Rs 1800
Example 20.15 For the project represented by Next we crash the critical activity (2, 5)
the following network and the table showing time The new network is given below
and cost, find the optimum duration and cost. 5
2 5
Activity Normal Crash DT DC DC
6 2
Time Cost Time Cost DT
(days) (Rs) (days) (Rs) 3

(1, 2) 8 100 6 200 2 100 50

(1, 3) 4 150 2 350 2 200 100 1 3 4
4 5
(2, 4) 2 50 1 90 1 40 40
(2, 5) 10 100 5 400 5 300 60 The new critical path (longest path) is 1 – 3 –
(3, 4) 5 100 1 200 4 100 25 4 – 5 and the duration becomes
(4, 5) 3 80 1 100 2 20 10 12 days direct cost = Rs 980 (680 + 300)
Indirect cost = Rs 70 per day. Indirect cost = 12 × 70 = Rs 840
Total cost = Rs 1820
Next we crash the critical activity (4, 5) [which
has the least slope 10] by two days. The network
becomes
5
2 5
Solution We find the critical path first.
6 2
E(m1) = 0; E(m2) = 8; E(m3) = 4
1
E(m4) = 10; E(m5) = 18; E(L5) = 18
E(L4) = 15; E(L3) = 10; E(L2) = 8
E(L1) = 0 1 3 4
4 5
 Project Scheduling (PERT and CPM) 269

The critical path becomes 1 – 2 – 5 with duration And the duration is 4 days. There is no indirect
11 days. cost. Hence the total cost is Rs 740.
Direct cost = Rs 1000 (980 + 20) Crashing The critical activities (4, 5) and
Indirect cost = Rs 11 × 70 = Rs 770 (5, 6) have the least slope 30. Hence we crash these
Total cost = Rs 1770 two activities by 4 days and 2 days respectively.
The critical activities (1, 2) and (2, 5) have been The network becomes
crashed already. Hence no more crashing can be 3
done. 6 7
Now we find that the activity (4, 5) can be 3
1 2 4 5 6
extended by one day without affecting the project 5 5 2 6
duration. This will save the crash cost of the Now the critical path (longest path) is 1 – 2 –
activity (4, 5) by one day. Thus we get the network 3 – 5 – 6 with duration 21 days
5
Total cost becomes 740 + 120 + 60 = Rs 920.
2 5 The critical activity (2, 3) has the least slope.
Hence we crash (2, 3) by one day. The new
6 2 network is
2
3
6 7
1 3 4 2
4 5
1 2 4 5 6
\ optimum duration = 11 days 5 5 2 6
Total cost = 1770 – 10 = Rs 1760 The critical path remains the same 1 – 2 – 3 –
Example 20.16 Find the optimum cost schedule 5 – 6 with duration 20 days. The total cost is
for the project with the following data Rs 960. Next we crash the critical activity (3, 5)
by 3 days. The revised net work is
Activity Normal Crash DT DC DC
3
Time Cost Time Cost DT
(days) (Rs) (days) (Rs) 6 4
2
(1, 2) 5 100 3 220 2 120 60
(1, 3) 6 120 3 195 3 75 25 1 2 4 5 6
5 5 2 6
(2, 3) 3 60 2 100 1 40 40 Now the critical path is 1 – 2 – 4 – 5 – 6 with
(2, 4) 5 80 3 170 2 90 45 duration 18 days. The total cost is Rs (960 + 150) =
(3, 5) 7 100 4 250 3 150 50
Rs 1110
(4, 5) 6 120 2 240 4 120 30
(5, 6) 8 160 6 220 2 60 30
The activity (2, 4) with slope 45 is to be crashed
next. The network becomes
740 Indirect cost is zero
3
Solution The network of the project is
6 4
2
3
7
6 1 2 4 5 6
3 5 3 2 6
1 2 4 5 6 Now the critical path 1 – 2 – 3 – 5 – 6 and the
5 5 6 8 duration is 17 days. The total cost is Rs 1200. Next
It is easily seen that the critical path (longest we crash the remaining critical activity (1, 2) by
path) is 1 – 2 – 4 – 5 – 6. 2 days. The new network is
 270 Operations Research

3 If the indirect cost is Rs 100 per day, find the

least cost schedule (optimum duration).
6 4
2 Solution The network diagram of the project is
as shown below:
1 2 4 5 6
3 3 2 6 3 4
The new critical path is 1 – 3 – 5 – 6 having 5 3
duration 16 days. The total cost is Rs 1320. Next 4
we crash (1, 3) by 3 days. The network becomes 2

3 1 2 5
8 10
3 4
2 The critical path is 1 – 2 – 5 and the duration of
the project is 18 days.
1
3
2
3
4
2
5
6
6 Direct cost = Rs 580
The critical path (longest path) is 1 – 2 – 3 – Indirect cost = Rs 1800
5 – 6. The duration is 15 days. Total cost is Total cost = Rs 2380
Rs 1395. Thus the minimum project duration is Crashing The critical activity (1, 2) having
15 days and the total cost is Rs 1395. However we the least slope, is crashed by 2 days. The network
want to decrease the cost without increasing the becomes
duration. We find that by increasing the duration
of (1, 3) by 2 days we save (2 × 25) Rs 50. Also by 3
5
4
increasing the duration of (2, 4) by one day we 3
save Rs 45. The critical path remains the same. 4
2
Thus the revised schedule of activities is given by
1 2 5
the following network. 6 10
3 The critical path remains the same (1 – 2 – 5)
5 4 with duration 16 days.
2 Direct cost = Rs 680
1 2 4 5 6
Indirect cost = Rs 1600
3 4 2 6 Total cost = Rs 2280
The optimum cost of the project is 1395 – 50 – Next we crash the activity (2, 5) by 5 days. We
45 = Rs 1300 have,
The optimum duration is 15 days.
3 4
Example 20.17 The following table gives the 5 3
cost particulars of a project 4
2
Activity Normal Crash DT DC Slope
1 2 5
Time Cost Time Cost 6 5
(days) (Rs) (days) (Rs)
The new critical path is 1 – 3 – 4 – 5 having
(1, 2) 8 100 6 200 2 100 50 duration 12 days.
(1, 3) 4 150 2 350 2 200 100 Direct cost = Rs 980
(2, 4) 2 50 1 90 1 40 40 Indirect cost = Rs 1200
(2, 5) 10 100 5 100 5 300 60
Total cost = Rs 2180
(3, 4) 5 100 1 200 4 100 25
Now we crash (4, 5) by 2 days. We obtain the
(4, 5) 3 80 1 100 2 20 10
Rs 580
network
 Project Scheduling (PERT and CPM) 271

Total cost = Rs 2100

3 4 We can decrease the cost by Rs 10 by increasing
5 1
the duration of (4, 5) to 2
4 Thus the optimum cost is Rs 2090 with project
2
duration 11 days. The network schedule is
1 2 5
6 5 3 4
5 2
The critical path is 1 – 2 – 5 with duration 11 days.
4
Direct cost = Rs 1000 2
Indirect cost= Rs 1100
1 2 5
6 5

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. Construct the network for the projects whose activities and their relation ships are given below:
(a) A < B, C; C < G, F; D < G, F; E, F < H
(b) A < C; B < D, E; C, D < F; E < G; F<H
(c) A < C, D; B < E; C, E < F, G; D < H, G < I; H, I < J
(d) A < B, C; D < E; B, C, E < F; F < G; E < H; G, H < I
2. Find the critical path and time of completion of the projects described below:
(a) Activity (1, 2) (1, 3) (1, 4) (2, 5) (2, 6) (3, 5) (4, 6) (5, 6)
Time 4 2 10 12 12 6 8 9
(Days)

(b) Activity (1, 2) (1, 3) (2, 4) (3, 4) (3, 5) (4, 7) (5, 6) (5, 7) (6, 8) (7, 8)
Time 4 1 1 1 6 5 4 8 1 2
(Days)

(c) Activity (1, 2) (2, 3) (3, 4) (3, 7) (4, 5) (4, 7) (5, 6) (6, 7)
Time 3 4 4 4 2 2 3 2
(Days)

(d) Activity (1, 2) (1, 3) (2, 4) (3, 4) (3, 5) (4, 5) (4, 6) (5, 6)
Time 6 5 10 3 4 6 2 9
(Weeks)
(e) A < B, C, D, E; E < F; D < G; G, F < H; B < I
Activity A B C D E F G H I
Time
(Weeks) 1 4 2 3 2 3 2 1 3

(f) Job A B C D E F G H I J K
Time 13 8 10 9 11 10 8 6 7 14 18
A < B; B < C, E; C < D; E < F, H; D, F < G; H < I; G, I < J; J < K
 272 Operations Research

(g) Job A B C D E F G H I J K L
Time 30 7 10 14 10 7 21 7 12 15 30 15
A < B, C; B < D, G, K; C < D, G; D < E; E < F; F < H, I, L; G < L; H, I < J; K < L
(h) A < B, C; B < D, E; C < F, G; F < H; D < I; E, G < K; H < J
Activity A B C D E F G H I J K
Time 10 9 7 6 12 6 8 8 4 11 7
(weeks)

(i) 3. A project has the following data:

4 Job to tm tp
2 6

2 (1, 2) 3 4 5
1 (1, 3) 1 2 3
2 5 4 3
1 3 7 8 9 (2, 3) 6 8 10
1 8 (2, 4) 2 5 8
5 (2, 5) 3 5 7
4 5
3
(j) (3, 4) 6 9 12
(4, 7) 1 1 1
12
2 6 (5, 6) 2 5 8
6 13 16 (6, 7) 4 8 12
7
7 14 8 6 (i) Draw the network
1 3 5 8 9
12 (ii) Determine the critical path and the
13
9
4 7 project length
9

4. For the project having the following jobs and time estimates find the probability of completing the
project in 66 days.
Activity (1, 2) (2, 3) (2, 4) (3, 5) (4, 5) (4, 6) (5, 7) (6, 7) (7, 8) (8, 9)
to 9 10 5 3 4 10 8 6 12 7
tm 10 13 9.5 5 7 15 12 7.5 15 11
tp 11 16 11 13 10 26 22 12 24 15

5. For the project represented by the following

network find the probability of completing
the project in 26 days.

Find the probability that the project would

6. The network of a project is given below: be completed in 48 days.
 Project Scheduling (PERT and CPM) 273

7. A project has the following data: would be completed in 60 days?

Activity A B C D E F G H 8. Draw the network for the following data.
to 4 8 4 1 2 4 10 18 Activity A B C D E F G H I
tm 5 12 5 3 2 5 14 20
tp 6 16 12 5 2 6 18 34 to 3 14 11 2 2 10 3 4 1
tm 7 21 14 2 3.5 14 4 4.5 2
A < C; B < D; A, D < E; B < F; C, E, F < G; tp 11 28 17 2 8 21 5 8 4
G < H.
(i) Draw the network A < B, C, D; C < D; B < E; D, E < F; E < G;
(ii) Find the critical path and the expected E, F < H; G, H < I
time of completion of the project What is the probability that the project would
(iii) What is the probability that the project be completed in 56 days?

9. A project is represented by the following activities:

Activity (1, 2) (1, 3) (1, 4) (2, 5) (3, 6) (3, 7) (4, 6) (5, 8) (6, 9) (7, 8) (8, 9)
Time 2 2 1 4 8 5 3 1 5 4 3

Find the (a)

(i) Critical path
Activity Normal Crash
(ii) duration of the project
(iii) the free float and the independent float Time Cost Time Cost
for each activity (1, 2) 3 300 2 400
10. The following table gives the tasks and their (2, 3) 3 30 3 30
durations of a project. (2, 4) 7 420 5 580
(2, 5) 9 720 7 810
Task A B C D E F G H I (3, 5) 5 250 4 300
Time 23 8 20 16 24 18 19 4 10 (4, 5) 0 0 0 0
(5, 6) 6 320 4 410
A < D, E; B, D < F; C < G; B < H; F, G, H < I (6, 7) 4 400 3 470
Find the (6, 8) 13 780 10 900
(i) Critical path (7, 8) 10 1000 9 1200
(ii) duration of the project
(iii) the free float and the independent float Indirect cost = Rs 50 per week
for each activity (b)
11. Given Activity Normal Crash
Task A B C D E F G H I J Time Cost Time Cost
Time 14 22 10 16 12 10 6 8 24 16 (1, 2) 4 60 3 90
(1, 3) 2 38 1 60
A < B; B < C, D, E; C < F, G; F, G < H; (1, 4) 6 150 4 250
D, E, H < I; I < J (2, 4) 5 150 3 250
Find the critical path, project duration and (3, 4) 2 100 2 100
the free float for each task. (2, 5) 7 115 5 175
12. Find the least cost schedule for the following (4, 5) 4 100 2 240
projects by systematic crashing of the activities. Indirect cost = Rs 60 per day
 274 Operations Research

(c) 14. The following table gives the activities and

Activity Normal Crash their durations in their order, of a project.
Time Cost Time Cost
Draw the network and find the minimum time
required for completion of the project.
(1, 2) 6 300 4 400
Activity Preceding Duration
(2, 3) 8 400 5 550 Activity (in days)
(2, 4) 7 400 6 500
A – 9
(2, 5) 12 1000 8 1800
B – 4
(3, 5) 5 500 3 800 C – 7
(4, 5) 8 800 7 1000 F B, C 8
(5, 6) 6 700 5 800 E A 7
F C 5
Indirect cost = Rs 50 per day G E 10
(d) H E 8
Activity Normal Crash I D, F, H 6
Time Cost Time Cost 15. Draw the network and find the critical path
and also calculate the total float and free float
A 2 100 1 150
for each job in a project given by the
B 6 200 4 400 following table:
C 3 200 1 300
Job Precedence Duration
D 4 500 2 700
Activity (in hours)
E 2 400 1 550
A – 14
F 8 200 5 450
B A 22
G 5 300 3 500 C B 10
A < D; B < F; C < E; D < G; E < F D B 16
13. The following table gives the activities and E B 12
F C 10
their durations in a project.
G C 6
Activity Duration Immediate H F, G 8
(in weeks) Predecessor I D, E, H 24
A 4 E J I 16
B 2 A 16. Draw the network for the project represented
C 1 B by the following activities and find the
E 14 – critical path.
F 2 E
G 3 F Activity A B C D E F G H I J K
H 2 F Time 12 10 15 8 20 12 6 8 16 12 8
I 4 C, G, H (days)
J 12 I
A < C, D, I; B < G, F; D < G, F; F < H, K;
Draw the network and find the critical path.
G, H < J; I, J, K < E
17. Draw the network for the project whose activities are given below. Calculate the total float, free
float and independent float for each activity and also the project duration.
Activity (1, 2) (2, 3) (2, 4) (3, 5) (3, 6) (4, 5) (4, 7) (5, 8) (6, 8) (7, 8)
Time (weeks) 3 8 12 6 3 3 8 5 3 8
 Project Scheduling (PERT and CPM) 275

18. Find the critical path of the project having the following activities.
Activity (1, 2) (2, 3) (2, 4) (3, 5) (4, 6) (4, 7) (5, 8) (6, 7) (7, 8)
Time (days) 2 3 4 2 4 3 6 5 6

19. For the following project find the probability Activity Time estimates
of completion in 42 days. to tm tp

Activity to tp tm A 4 6 8
B 5 7 15
(1, 2) 6 10 8
C 4 8 12
(1, 3) 18 22 20
(1, 4) 26 40 33 D 15 20 25
(2, 5) 16 20 18 E 10 18 26
(2, 6) 15 25 20 F 8 9 16
(3, 6) 6 12 9 G 4 8 12
(4, 7) 8 12 10 H 1 2 3
(5, 7) 7 9 8 I 7 6 8
(6, 7) 3 5 4
A < B, C; B < D, E;
20. A project has the following data
C < F;
A < B, C; B, C < D
E < G;
B < E; D, E < F
D<G D, F < H;
G, H < I.
Activity Time estimates 22. Determine the critical path
to tm tp
A 1 4 7 Activity to tm tp
B 4 6 14 (1, 2) 3 4 5
C 3 3 3
(1, 3) 6 8 10
D 4 10 22
(1, 4) 2 3 4
E 3 8 13
F 2 5 14 (2, 5) 4 5 12
G 4 4 4 (3, 5) 5 7 9
(3, 6) 9 16 17
Find the expected time of completion. Also
find the probability that the project would (4, 7) 8 12 16
be completed in 30 days. (5, 9) 4 5 6
21. For the following data, draw the network, (6, 9) 7 8 15
calculate the expected time of completion and (7, 8) 8 11 14
find the probability of completing the project
(8, 9) 8 10 12
in 50 weeks.
 276 Operations Research

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. (a) (i) 1 – 3 – 5 – 9; Time = 15

B (j) 1 – 3 – 5 – 6 – 8 – 9; Time = 50
C 3. 1 – 2 – 3 – 4 – 7; Time = 22
G
A H
D F 3
9
E 2 4
(b) 8 5 1 7
G 1 2 8
5
E 5
6
5
D H
4. 0.1515
5. 0.963
B F
6. 0.898
C 7.
A
(c)
H

D
F
I
C

A G
E
B
(d) 1 – 2 – 3 – 4 – 5 – 6 – 7; Time = 57;
E H Probability = 0.7995
8.
G
D E1 F
C
A B1
B

2. (a) 1 – 2 – 5 – 6; Time = 25
(b) 1 – 3 – 5 – 7 – 8; Time = 17
(c) 1 – 2 – 3 – 4 – 5 – 6 – 7; Time = 18
(d) 1 – 2 – 4 – 5 – 6; Time = 31
(e) A – B – I; Time = 8 1 – 2 – 3 – 6 – 7 – 9 – 10 – 11; Time = 53;
(f) A – B – E – F – G – J – K; Time = 82 Probability = 0.8
(g) A – C – C1 – D – E – F – I Time = 98 9. 1 – 3 – 6 – 9; Duration = 15
– I1 – J; Free float is zero for all the activities except
(h) A – C – F – H – J; Time = 42 (5, 8), which has Free Float 4.
 Project Scheduling (PERT and CPM) 277

Independent float is zero for all activities (b) Normal duration is 13 days.
except (1, 4) with I.F. 1 and (3, 7) with I.F. 1. Total cost = Rs 1493
10. Critical path: 1 – 2 – 4 – 5
After crashing the critical activities, we
obtain the optimal duration 8 days
Optimal cost = Rs 1523
(c) Normal duration is 27 days.
Total cost = Rs 5450
Critical path: 1 – 2 – 4 – 5 – 6
After crashing the critical activities, we
obtain the optimal duration 22 days
Optimal cost = Rs 5700
(d) Normal duration is 14 days.
Total cost = Rs 1900
Critical path: B – F
After crashing the critical activities, we
Critical path: 1 – 2 – 6 – 7 – 8 obtain the optimal duration 10 days
Optimal cost = Rs 2300
Duration: 67
13. Network
Free Float: (3, 7) – 45; (4, 7) – 18; (5, 8) –
20; Zero for other activities
Independent Float: (2, 5) – 20; (3, 7) – 14;
Zero for other activities
11. 10
4
12 0 16
E1 J
14 2 22 3 E 16 24
1
A B D 8 I 9 Critical paths: E – A – B – C – I – J
10 Completion time: 37 weeks
C 8 H
6 14. Network
5 7
G
10 F1
F
0
6

Critical path: 1 – 2 – 3 – 5 – 6 – 7 – 8 – 9 – 10
Duration: 104
Free Float: (4, 8) – 16; (5, 7) – 4; Zero for
other activities
12. (a) Normal duration is 32 weeks.
Total cost = Rs 5820
Critical path: 1 – 2 – 5 – 6 – 7 – 8
After crashing the critical activities, we
obtain the optimal duration 29 weeks Critical path: A – E – H – I
Optimal cost = Rs 5805 Time of completion 30 days.
 278 Operations Research

15. Network 17. Network

Critical path: 1 – 2 – 4 – 7 – 8
Project duration: 31 weeks
Activity Total float Free float Independent
float
(1, 2) 0 0 0
Critical path: 1 – 2 – 3 – 4 – 7 – 8 – 9 – 10 (2, 3) 9 0 0
Time of completion is 104 hours (2, 4) 0 0 0
(3, 5) 9 1 0
Job Total Float
(3, 6) 14 0 0
(3, 5) 12 (4, 5) 8 0 0
(3, 8) 16 (4, 7) 0 0 0
(4, 6) 4 (5, 8) 8 8 0
(5, 8) 12 (6, 8) 14 14 0
(6, 7) 4 (7, 8) 0 0 0
For others 0 18. Critical path: 1 – 2 – 4 – 6 – 7 – 8
Free float is zero for all the activities. Project duration : 21 days
16. 19. Critical path: 1 – 4 – 7
Expected time of completion: 43 days
Required probability: 0.34
20. Expected time of completion: 28 days
Required probability: 0.69
21. Network

Expected time of completion: 47 weeks

Critical path: 1 – 2 – 3 – 4 – 5 – 6 – 7 Required probability = 0.8
(A – D – F – H – J – E) 22. Critical path: 1 – 4 – 7 – 8 – 9
Time of completion = 72 days Expected time of completion: 36 days
 Project Scheduling (PERT and CPM) 279

SHORT ANSWER QUESTIONS

■ ■ ■ ■ ■ ■ ■

1. Explain the method of numbering the events 7. Write the formula to find the latest finishing
of a network. time of an activity.
2. What is a dummy activity? 8. What are the differences between PERT and
3. Define: Total float, free float and indepen- CPM?
dent float. 9. Write the formula for finding the expected
4. Define critical path. duration of an activity.
5. Distinguish between float and slack. 10. If the SDs of six critical activities are 5, 1, 4,
6. If the duration of any activity is changed then 8, 9 and 3, what is s c.p.? (Ans: 14)
the project duration is affected. True or false?
(Ans: False)

CONCEPT REVIEW
■ ■ ■ ■ ■ ■ ■
21.1 INTRODUCTION
There are certain real world problems which are
very complicated in nature and it is not possible
to construct mathematical models for them. Such
problems can be solved by the method of
simulation. The method of simulation is applied
to problems which involve probability and
probability distribution. An analogous physical
model is constructed to represent the mathematical
and logical relationship among the variables of the
problem. Thus simulation is the process of
designing a model of a real system and conducting
experiments with this model to study the actual
system. For this purpose a series of random values
for the variables is generated using different
methods on a computer.
A children’s cycling park with various crossings
and signals is a simulated model of the traffic in a
city. Experiments are conducted in the laboratories
on simulated models and the behaviour of the real
system is predicted. Simulation experiments are
done only with the model. Hence there is no
interference on the system. With the help of
simulation we can spot the difficulties which may
21
Simulation
arise due to the change in equipment and process
of the real system. Also the simulation techniques
are easier to carry out compared to the much
complicated mathematical models. Moreover it is
advantageous to train people on simulated models
before putting their hands into the real system.
21.2 EVENT TYPE SIMULATION
In this type of simulation, the behaviour of the
system is observed for a long period of time and
relevant informations are collected based on these
observations.
Example 21.1 As an example let us consider a
queueing system with one server. We shall analyse
the system in order to evaluate the quality of
service by assessing the average waiting time per
customer and percentage of idle time of server.
There are two events, an arrival and a departure.
Denote the arrivals by a1, a2, a3, … and departures
by d1, d2, d3…. Assume that the inter arrival and
service time are 1.5 min and 4 min respectively.
We simulate this system for a period of 20 minutes.
A customer leaves the system as soon as his service
is over. Simulation starts with the arrival of the
 Simulation 281

first customer a 1 at time t = 0. His departure From this simulation table we get the average
time is t = 4. Next arrival a2 occurs at t = 1.5. waiting time of the customer as
Since the system is busy serving the first customer, 2.5 + 5 + 7.5 + 10 + 12.5 + 11 + 9.5 + 8 + 6.5 + 5
the second customer waits in the queue. At + 3.5 + 2 + 0.5
t = 3 third customer arrives (a3). At t = 4, first
14
customer leaves (d1). This procedure is repeated
83.5
till t = 20 min. The following table gives the results = = 6 min nearly
in detail. 14
Since the service facility is always busy, the
Time Event Customer Waiting time idle time of the server is zero.
(minutes) of the customer
21.3 MONTE–CARLO
0.0 a1 1 0 (c1)
1.5 a2 2 TECHNIQUE
3.0 a3 3
4.0 d1 1 4 – 1.5 = 2.5 (c2) This technique involves the selection of random
4.5 a4 4 observations in the simulation model. We assume
6 a5 5 a probability distribution giving the probability of
7.5 a6 6 the events and use random numbers for the
8.0 d2 2 8 – 3 = 5 (c3) cumulative density. These random numbers may
9.0 a7 7 be selected from the table of random numbers or
10.5 a8 8 generated using computer.
12.0 d3 3 12 – 4.5 = 7.5 (c4)
12.0 a9 9
21.3.1 Step-by-Step Procedure
13.5 a10 10 1. Decide the variables which influence the
15.0 a11 11 measure of effectiveness.
16.0 d4 4 16 – 6 = 10 (c5) 2. Determine the probability distribution for
16.5 a12 12 each variable.
18.0 a13 13 3. Choose a suitable set of random variables.
19.5 a14 14 4. Consider each random number as a decimal
20.0 d5 5 20 – 7.5 = 12.5 (c6) value of the cumulative probability.
5. Use the simulated values so generated into
Customers getting service after 20 min and their the formula derived from the chosen measure
waiting times during the period under simulation of effectiveness.
are given below. 6. Repeat steps 4 and 5 until the sample is
large enough to arrive at a satisfactory
Customer no. Waiting time (min)
decision.
7 20 – 9 = 11 Example 21.2 The sales (in thousands of rupees)
8 20 – 10.5 = 9.5 in a company follow Poisson distribution with
9 20 – 12 = 8
mean 5. Using the Monte–Carlo technique
10 20 – 13.5 = 6.5
estimate the sales for 10 days.
11 20 – 15 = 5
12 20 – 16.5 = 3.5 Solution The probability for k sales is given by
13 20 – 18 = 2 P(x = k) = e–5 5k / k! e–5 = 0.0067
14 20 – 19.5 = 0.5
The cumulative probabilities for the values of
Total = 46 k are given below:
 282 Operations Research

k Cumulative prob. k Cumulative prob. Arrival time Prob. Cumulative Interval

(min) prob.
0 0.0067 7 0.8666
1 0.0404 8 0.9319 0.5 0.02 0.02 00–01
2 0.1247 9 0.9682 1.0 0.06 0.08 02–07
3 0.2650 10 0.9763 1.5 0.10 0.18 08–17
4 0.4405 11 0.9845 2.0 0.25 0.43 18–42
5 0.6160 12 0.9999 2.5 0.20 0.63 43–62
6 0.7622 3.0 0.14 0.77 63–76
3.5 0.10 0.87 77–86
Now note down 10 two-digit random numbers 4.0 0.07 0.94 87–93
from the random number tables and read the 4.5 0.04 0.98 94–97
corresponding values of k. These will give the sales 5.0 0.02 1.00 98–99
for 10 days. The values are given below.

Random no. Sales Random no. Sales Service time Prob. Cumulative Interval
(min) prob.
48 05 64 06
11 02 79 07 0.5 0.12 0.12 00–11
53 05 16 03 1.0 0.21 0.33 12–32
20 03 57 05 1.5 0.36 0.69 33–68
17 03 2.0 0.19 0.88 69–87
81 07 2.5 0.07 0.95 88–94
3.0 0.05 1.00 95–99

Example Number 53 (5300) lies in the interval
(4405, 6160). Hence we take 05 corresponding to
k = 5.
Number 79 (7900) belongs to (7622, 8666).
Hence we take k = 7.
Example 21.3 At a sales depot the arrival of
customers and the service times follow the
following probability distributions.
Estimate the average waiting time and
percentage of idle time of the server, by simulation,
for 10 arrivals.
Solution 10 random numbers are chosen for
arrival and service each and are linked to
appropriate events. The results are given in the
following table.

Arrival Random Int Arr time Arrival time Random Service Service Waiting time
no. no. (min) (min) no. time
Start End Customer Server
1 93 4.0 4.0 78 2.0 4.0 6.0 – 4.0
2 22 2.0 6.0 76 2.0 6.0 8.0 – –
3 53 2.5 8.5 58 1.5 8.5 10.0 – 0.5
4 64 3.0 11.5 54 1.5 11.5 13.0 – 1.5
5 39 2.0 13.5 74 2.0 13.5 15.5 – 0.5
6 07 1.0 14.5 92 2.5 15.5 18.0 1.0 –
7 10 1.5 16.0 38 1.5 18.0 19.5 2.0 –
8 63 3.0 19.0 70 2.0 19.5 21.5 0.5 –
9 76 3.0 22.0 96 3.0 22.0 25.0 – 0.5
10 35 2.0 24.0 92 2.5 25.0 27.5 1.0 –
Total 4.5 7.0
 Simulation 283

From the above table we get, We find that the total demand for the 10 weeks
4.5 28
Average waiting time of a customer = is 28. Hence the average demand is = 2.8 cars
10 10
= 0.45 min per week.
7.0 Example 21.5 A company observes from past
Average idle time of the server = = 0.7 min experience that the demand for a special product
10
has the following probability distribution.
Example 21.4 A tourist car company finds that
during the past 200 days the demand for the car Daily demand 5 10 15 20 25 30
has the following frequency distribution.
Probability 0.01 0.20 0.15 0.50 0.12 0.02
Trips per week 0 1 2 3 4 5
Frequency 16 24 30 60 40 30 Simulate the demand for the next 10 days. Also
find the average demand.
Using random numbers simulate the demand Solution The probability distribution of the
for a period of 10 weeks. demand is given below:
Solution The probability distribution of the
demand (trips) is given by the following table: Demand Prob. Cumulative prob. Interval
5 0.01 0.01 00–00
Trips per Frequency Prob. Cumulative Interval 10 0.20 0.21 01–20
week prob. 15 0.15 0.36 21–35
0 16 0.08 0.08 00–07 20 0.50 0.86 36–85
1 24 0.12 0.20 08–19 25 0.12 0.98 86–97
2 30 0.15 0.35 20–34 30 0.02 1.00 98–99
3 60 0.30 0.65 35–64
4 40 0.20 0.85 65–84 The random numbers generated are:
5 30 0.15 1.00 85–99 69, 46, 16, 79, 64, 81, 11, 20, 60, 53.
Using these numbers, we simulate for 10 days.
Random numbers generated are 82, 95, 18, 96, The simulated demand is given by the following
20, 84, 56, 11, 52 and 03. Using these random table:
numbers we simulate for 10 weeks. The simulated
demand is shown in the following table: Day Random no. Demand
1 69 20
Week Random number Demand
2 46 20
1 82 4 3 16 10
2 95 5 4 79 20
3 18 1 5 64 20
4 96 5 6 81 20
5 20 2 7 11 10
6 84 4 8 20 10
7 56 3 9 60 20
8 11 1 10 53 20
9 52 3 Total 170
10 03 0
170
Total 28 Average demand = = 17 units
10
 284 Operations Research

EXERCISES
■ ■ ■ ■ ■ ■ ■

1. A bakery keeps stock plum cakes. Past 60 min. Estimate the average waiting time
experience shows that the daily demand of of the customer and the percentage of idle
the cake has the probability distribution given time for the server.
below: 3. In a shop the sales of a particular soap has
probability distribution P (x = r) = e–l lr/r!
Daily 0 10 20 30 40 50 (Poisson distribution) with l = 5. Given e–5
demand = 0.0067 and the random numbers 49, 58,
89, 15, 12, 94, 85, 34, 07 and 53 simulate
Prob. 0.01 0.20 0.15 0.50 0.12 0.02
the sales for 10 days.
4. In a machine shop the time of breakdown
Using the random numbers 25, 39, 65, 76, of machines and their repair times in
12, 05, 73, 89, 19, 49 simulate the demand hours have the following frequency
for the next 10 days. distributions.
2. The following probabilities are given for the
arrival and service times of the customers in Time between 10 11 12 13 14 15 16 17 18 19
a queuing system. breakdown

Arrival time Service time Frequency 4 10 14 16 12 6 4 3 3 1

Minutes Prob. Minutes Prob.
Repair time 8 9 10 11 12 13 14 15 16 17
2 0.15 1 0.10
4 0.23 3 0.22 Frequency 2 3 8 16 14 12 8 5 3 2
6 0.35 5 0.35
8 0.17 7 0.23 Use the random numbers 46, 64, 09, 48, 97,
10 0.10 9 0.10 22, 29, 40, 75, 10 and 09, 70, 41, 40, 37, 21,
44, 97, 51, 46 simulate the system and find
Using the random numbers 93, 14, 72, 10, out the average waiting time of the machine
21, 81, 87, 90, 38, 71, 63, 14, 53, 64, 42, 07, and the percentage of idle time for the
54 and 66 simulate the queue behaviour for mechanic.

ANSWERS
■ ■ ■ ■ ■ ■ ■

1. Day 1 2 3 4 5 6 7 8 9 10
Demand 20 30 30 30 10 10 30 40 10 30

Average demand = 24
 Simulation 285

2. Arrival Random Inter arrival Arrival Random Service Service Waiting time
no. no. time in mts time no. time in mts Start End Customer Server
1 93 10 10 71 7 10 17 – 10
2 14 2 12 63 5 17 22 5 –
3 72 6 18 14 3 22 25 4 –
4 10 2 20 53 5 25 30 5 –
5 21 4 24 64 5 30 35 6 –
6 81 8 32 42 5 35 40 3 –
7 87 8 40 07 1 40 41 – –
8 90 10 50 54 5 50 55 – 9
9 38 6 56 66 5 56 61 – 1
23 20

Average waiting time of the customer = 23/9 = 2.5 min

Idle time for the server = 20 min; Percentage = 20/60 ¥ 100 = 33%

3. Day 1 2 3 4 5 6 7 8 9 10
Random no. 49 58 89 15 12 94 85 34 7 53
Sales 5 5 8 3 2 9 7 4 2 5

4. Machine Random Inter Time Random Repair Repair Waiting time

no. no. arrival in hrs No. time in
Start End Machine Mechanic
time in hrs
hrs
1 46 13 13 9 10 13 23 - 13
2 64 14 27 70 13 27 40 - 4
3 9 11 38 41 12 40 52 2 -
4 48 13 51 40 12 52 64 1 -
5 97 18 69 37 11 69 80 - 5
6 22 12 81 21 11 81 92 - 1
7 29 12 93 44 12 93 105 - 1
8 40 13 106 97 17 106 113 - 1
9 75 14 120 51 12 120 132 - 7
10 10 11 131 46 12 132 144 1 -
4 32

Average waiting time of the machine = 4/10 = 24 min

Idle time for the mechanic = 32 hrs
Percentage = 32/144 × 100 = 22 %
 Index

Activity 255 Critical path 257

Advertising Media Problem 4
Algebraic Method in Game Theory 170
Artificial Variable 41
Assignment Problem 134
Balanced Problem 112
Basic cell 113
Basic feasible solution 24
Basic solution 23
Basic variables 23
Basis matrix 24
Bellman’s principle 178
Big M method 41
Blending problem 6
Bulk service 206
Canonical form 26
Capital budgeting problem 181
Cargo loading problem 184
Carrying cost 224
Column minima method 116
Conditional opportunity loss 154
Convex combination 13
Convex set 13
Cost slope 268
Critical activity 257
Critical node 263
Decision theory 149
Decision making under risk 152
Decision variables 3
Degeneracy in TP 113
Degenerate solution 30
Determinable game 160
Diet problem 5
Direct cost 267
Dominance property 163
Dual 59
Dual variables 120
Dynamic programming 178
Economic order quantity 225
Events 149
Event in PERT 255
Expected monetary value
criterion 152
Expected opportunity loss
criterion 154
Extreme point 14
Fair game 160
Feasible region 13
Feasible solution 24
Feasible solution of a TP 113

Float 257
Formulation 3
Free float 257
Fulkerson’s rule 256
Games with saddle point 160
Games without saddle point 161
Gomary’s constraint 90
Gomary’s cut 90
Graphical method for game theory 166
Holding cost 224
Hungarian method 135
Hurwicz criterion 150
Independent float 257
Indirect cost 267
Input process 205
Iteration 28
Laplace criterion 150
Lead time 225
Least cost schedule 267
Least cost method 117
Linear programming method
for games 171
Little’s formula 209
Maximax criterion 150
Maximin criterion 149
Maximization in AP 140
Matrix minima method 117
Minimax regret criterion 151
Mixed integer programming 89
Mixed strategy 159
Monte Carlo technique 280
Most likely time 262
Network diagram 255
Node 255
Non basic variable 23
North west corner rule 113
Optimal solution 24
Optimal strategy 159
Index 287
Optimal subdivision problem 184
Optimistic time 262
Order cycle 225
Ordering cost 224
Parametric programming 103
Pessimistic time 262
Pivot column 28
Pivot element 28
Pivot row 28
Planning horizon 225
Post optimality analysis 103
Present worth factor 247
Price break 231
Primal 59
Principle of duality 63
Production allocation problem 3
Production cost 225
Purchase cost 225
Pure integer programming 89
Pure strategy 159
Quantity discount 231
Queue discipline 205
Relative cost factor 27
Row minima method 115
Saddle point of the game 160
Salesman allocation problem 178
Savage criterion 151
Service mechanism 205
Setup cost 224
Sequencing problem 193
Shadow price 66
Shortage cost 225
Simplex algorithm 31
Simplex multipliers 120
Simplex table 27
Slack variable 25
Stage decision 178
Standard type of problem 59
State variable 178
Steady state 206
 288 Index

Stock out cost 225 Two person zero sum game 159
Strategy 149
Storage cost 224 Unbalanced AP 141
Surplus variable 25 Unbalanced TP 125

Total inventory cost 225 Value of the game 159

Total float 257 Vogel’s approximation method 117
Transient state 206
Transportation problem 112 Weighted average cost 247
Travelling salesman problem 142 Wilson – Harris formula 226