Mathematical Programming

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/337367019
MATHEMATICAL PROGRAMMING
Book · January 2020
CITATIONS READS
0 4,971
1 author:
S. M. Shahidul Islam
Hajee Mohammad Danesh Science and Technology University, Dinajpur, Bangladesh
23 PUBLICATIONS 45 CITATIONS
SEE PROFILE
All content following this page was uploaded by S. M. Shahidul Islam on 19 November 2019.
The user has requested enhancement of the downloaded file.

Mathematical Programming
This book consists of the preliminaries of mathematical programming,
convex sets, topics of linear programming, integer linear programming,
transportation problem, assignment problem and the basic of optimization
using calculus method. Every topic has been presented in a simple way. The
book becomes easier to understand for the readers, because, we follow the
method definition-example-theory-example-application in writing each
section. Students get here some done examples, mostly collected from the
question papers of different universities. Exercises are included for home
work. By studying the book, readers will learn the applications as well as the
basic understanding of mathematical programming.
S. M. Shahidul Islam
Dr. S. M. Shahidul Islam teaches Mathematics including
mathematical programming in many universities such as Including Theories and Applications
Hajee Mohammad Danesh Science and Technology
University, Asian University of Bangladesh, University
Brunei Darussalam, etc. His books “Business
Mathematics” & “Linear Programming” and many
journal papers are appreciated by the academicians.
Islam
MATHEMATICAL
PROGRAMMING
Including Theories and Applications
Dr. S. M. Shahidul Islam

Department of Mathematics
Hajee Mohammad Danesh Science and Technology University
Dinajpur, Bangladesh
LAP LAMBERT Academic Publishing

Imprint
Any brand names and product names mentioned in this book are subject
to trademark, brand or patent protection and are trademarks or registered
trademarks of their respective holders. The use of brand names, product
names, common names, trade names, product descriptions etc. even
without a particular marking in this work is in no way to be construed to
mean that such names may be regarded as unrestricted in respect of
trademark and brand protection legislation and could thus be used by
anyone.
Cover image: www.ingimage.com
Publisher:
LAP LAMBERT Academic Publishing
is a trademark of
International Book Market Service Ltd., member of OmniScriptum
Publishing Group
17 Meldrum Street, Beau Bassin 71504, Mauritius
Printed at:
KS OmniScriptum Publishing, Brivibas gatve 197, LV-1039 Riga,
Latvia, European Union
ISBN: 978-620-0-46432-3
Copyright © 2020 Dr. Shahidul Islam (sislam.math@gmail.com)

All rights reserved. No part of this publication may be reproduced, stored
in any electronic or mechanical form, including photocopy, recording or
otherwise, without the prior permission of the author.
Copyright © 2019 International Book Market Service Ltd., member of

OmniScriptum Publishing Group
Price: EUR (€) 108.90
ii
DEDICATED TO
Author’s wife
Mst. Bakul Nahar Parveen
&
Their affectionate daughter
Mst. Sumaiya Islam
iii
PREFACE
The book entitled Mathematical Programming has been
written with the blessings of Almighty Allah. The use of
mathematical programming has been increasing day by day in
many fields of science, socio-economics & commerce. Nowadays
mathematical programming is a compulsory course in many
disciplines. But the fact is that our students are always afraid of
mathematical programming as well as mathematics, and they need
proper guidance and dependable books. There are textbooks
available in the market on Mathematical Programming but most of
them are inadequate in meeting the requirements of the students. In
order to gear up the students in Mathematical Programming and
Operations Research we have attempted to write the book using
easy language and techniques.
The book consists of some preliminaries, topics of linear
programming, transportation problem and assignment problem. We
have included here a large number of examples, mostly collected
from question papers of different universities. We have verified
every questions and their solution minutely, and if there is any
error yet, we beg apology for that.
We are grateful and indebted to Professor Dr. A.A.K.
Majumdar, Ritsumeikan Asia-Pacific University, Japan; Professor
Dr. Md. Abdul Hoque, North South University, Bangladesh; and
many of our colleagues who encouraged and helped us in writing
the book. We are also grateful to the authors whose books we have
freely consulted in the preparation of the manuscript. Finally,
special recognition should go to the endless co-operation of
author’s wife Mst. Bakul Nahar Parveen without which the book
could not have seen the light of the day.
Constructive criticisms, corrections, and suggestions towards
the improvement of the book will be thankfully received.
Dr. S. M. Shahidul Islam
iv
CONTENTS
Page
Chapter 1 Setting the Scene 1 – 66
1.1 Linear programming 1
1.2 What is optimization 2
1.3 Summation symbol 3
1.4 General LP problem 3
1.5 Standard form of LP problem 5
1.6 Canonical form of LP problem 7
1.7 Advantages of LP 10
1.8 Limitations of LP 11
1.9 Some definitions 11
1.10 Some done examples 26
1.11 Exercises 57
Chapter 2 Convex Sets 67 – 102
2.1 Line segment 67
2.2 Convex set 68
2.3 Convex combination of a set of vectors 68
2.4 Extreme point 69
2.5 Convex cone 69
2.6 Convex hull 70
2.7 Convex polyhedron 70
2.8 Simplex 71
2.9 Hyper plane 71
2.10 Half space 73
2.11 Supporting hyper plane 76
2.12 Separating hyper plane 77
2.13 Convex function 77
2.14 Convex hull of a set 79
2.15 Closure, Interior and Boundary of a
convex set 80
2.17 Exercises 100
v
Chapter 3 Formulation and Graphical Methods 103 – 150
3.1 Formulation 103
3.2 Algorithm to formulate a linear
programming problem 103
3.3 Graphical method 107
3.4 Limitation of the graphical method 113
3.6 Exercises 145
Chapter 4 Simplex Methods 151 – 284
4.1 Simplex 151
4.2 Simplex method 151
4.3 Generating extreme point solution 154
4.4 Formation of an initial simplex table 165
4.5 Some definitions 169
4.6 Development of simplex algorithm 171
4.7 Simplex algorithm 175
4.8 Artificial basis technique 187
4.10 Exercises 277
Chapter 5 Duality in Linear Programming 285 – 422
5.1 Introduction 285
5.2 Different type of primal-dual problems 292
5.3 Primal-dual tables 306
5.4 Some theorems on duality 311
5.5 Complementary slackness conditions 347
5.6 Economic interpretation of primal-dual
problems 350
5.7 Shadow price 352
5.8 Dual simplex method 362
5.9 Dual simplex algorithm 362
5.11 Exercises 415
vi
Chapter 6 Sensitivity Analysis 423 – 486
6.1 Sensitivity analysis 423
6.2 An analytic example 423
6.3 Variation in the objective function
coefficients (c) 436
6.4 Variation in the right hand side
constants (b) 444
6.5 Variation in the coefficients matrix (A) 454
6.7 Exercises 483
Chapter 7 Integer Linear Programming 487 – 596
7.1 Integer linear programming 487
7.2 Some related definitions 488
7.3 Form of integer linear programming
problem (ILP) 493
7.4 Cutting plane for ILP problem 505
7.5 Algebraic description of cutting plane 506
7.6 Cutting plane algorithm for solving all
ILP problem 508
7.7 Cutting plane algorithm for solving
mixed ILP problem 528
7.8 Branch and bound algorithm for solving
ILP problem 538
7.9 Additive algorithm for solving 0–1
problem 545
7.11 Exercises 591
Chapter 8 Transportation Problem 597 – 646
8.2 Transportation problem 598
8.3 North-west corner rule 604
8.4 Least cost method 606
8.5 Vogel’s approximation method 606
vii
8.6 Optimality criterion 612
8.7 Loop 613
8.8 Generating optimal solution 614
8.9 Modified distribution (MODI) method 616
8.10 Degeneracy case 622
8.11 Multiple solutions 632
8.12 Total supply exceeds total demand
(unbalanced problem) 636
8.13 Maximization problem 640
8.14 Exercises 643
Chapter 9 Assignment Problem 647 – 670
9.2 Assignment problem 647
9.3 Algorithm of the Hungarian method 648
9.4 Dual of the assignment problem 653
9.5 Justification of Hungarian method 653
9.7 Exercises 668
Appendices 671 – 684
1 Classical Optimization 671
2 Model Exam Paper 679
3 Bibliography 683
viii
Chapter 1
Setting the Scene
Highlights
1.1 Linear programming 1.7 Advantages of LP
1.2 What is optimization 1.8 Limitations of LP
1.3 Summation symbol 1.9 Some definitions
1.4 General LP problem 1.10 Some done examples
1.5 Standard form of LPP 1.11 Exercises
1.6 Canonical form of LPP
1.1 Linear programming

Linear programming (LP) is a technique for determining an
optimum (maximizing profit or minimizing cost) schedule of
interdependent activities in view of the available resources.
Programming is just another word for “Planning” and refers to the
process of determining a particular plan of actions amongst several
alternatives. A linear programming problem with only two
variables presents a simple case, for which the solution can be
derived using a graphical method. We can solve any linear
programming problem by simplex method. A linear programming
problem may have a solution (a definite and unique solution or an
infinite number of optimal solutions or an unbounded solution) or
not.
Example 1.1
A company uses three different resources A, B and C to
manufacture two different products. 200 units of resource A, 120
units of B and 160 units of C are available. 1 unit of first product
requires 2 units of A, 2 units of B and 4 units of C, and 1 unit of the
Setting the Scene
second product requires 4 units of A and 2 units of B. It is known

that per unit first product earns a profit of 20 monetary units and
per unit second product earns a profit of 30 monetary units.
Company likes to know how many units of each product should be
manufactured for maximizing its total profit. This is an example of
the linear programming problem. Mathematical form of this
problem can be derived as follows.
Let x unit of first product and y unit of second product are to
be produced for maximizing total profit, then company’s total
profit z = 20x + 30y is to be maximized. As 1 unit of first product
requires 2, 2, 4 units of resources A, B, C respectively and 1 unit of
second product requires 4, 2 units of resources A, B respectively,
and 200 units of resource A, 120 units of B, 160 units of C are
available, so subject to the constraints are
2x + 4y < 200 or, x + 2y ≤ 100
2x + 2y < 120 or, x + y ≤ 60
4x + 0y < 160 or, x < 40
It is not possible for the manufacturer to produce negative
number of products, obviously x  0, y  0. Thus, we can summarize
the above linguistic linear programming problem as the
mathematical form below.
Maximize z = 20x + 30y
Subject to x + 2y  100, x + y  60, x  40
x  0, y  0 (non-negativity conditions)
1.2 What is optimization
The fundamental problem of optimization is to arrive at the best
possible decision under some given circumstances of some typical
problems, taken from a variety of practical field of interest. In
linear programming problem, we are supposed to optimize
(maximize or minimize) certain linear functions under certain
linear constraints.
2
1.3 Summation symbol

Greek capital letter ∑ (sigma) is usually used as the symbol of
summation. f(i) denotes a quantity whose value depends
on the value of i, but the expression
4
i
i 1
is read as ‘sigma i, i going from 1 to 4’ and means to insert 1 for i,
2 for i, 3 for i and 4 for i, and then sum the results. Therefore,
4
 i = 1 + 2 + 3 + 4 = 10.
i 1
In the above example, i under the sigma symbol is called the index
of the summation. Similarly,
5
i
i 1
2
= 1 + 4 + 9 + 16 + 25 = 55,
3
 (i
i 1
2
 1) = (12 – 1) + (22 – 1) + (32 – 1) = 0 + 3 + 8 = 11,
4
 p = p + p + p + p = 4p,
i 1
n
 p = np,
i 1
4
x
i 1
i = 𝑥1 + 𝑥2 + 𝑥3 + 𝑥4 .
1.4 General linear programming problem

General linear programming problem is to find a vector
(𝑥1 , 𝑥2 , …, 𝑥𝑗 , …, 𝑥𝑛 )
which minimizes (or maximizes) the linear function
𝑧 = 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛 (1)
subject to the linear constraint having m equations/ inequalities of
n variables as below
3
Setting the Scene
a11 x1  a12 x 2  ...  a1 j x j  ...  a1n x n  (or  or )b1 


a 21 x1  a 22 x 2  ...  a 2 j x j  ...  a 2 n x n  (or  or )b2 
 
 (2)
ai1 x1  ai 2 x 2  ...  aij x j  ...  ain x n  (or  or )bi 
 

a m1 x1  a m 2 x 2  ...  a mj x j  ...  a mn x n  (or  or )bm 
and 𝑥𝑗 ≥ 0; j = 1, 2, 3, . . ., n. (3)
where the 𝑎𝑖𝑗 , 𝑏𝑖 and 𝑐𝑗 are given constants and m  n. Constraint
equations will be multiplied by –1 to make all 𝑏𝑖 ≥ 0 in
minimization problem. The linear function 𝑧 = 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ +
𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛 which is to be minimized (or maximized) is
called objective function of the general LP problem. The
inequalities (or equations) in (2) are called constraints of the
general LP problem. The set of inequalities in (3) is usually known
as the set of non-negative restrictions. General LP problem can
be stated in many forms. Some common forms are as follows.
n
(a) Minimize (or, maximize) c
j 1
j xj
n
Subject to a
j 1
ij x j  (or  or )bi ; i = 1, 2, 3, . . ., m
𝑥𝑗 ≥ 0 ; j = 1, 2, 3, . . ., n
(b) Minimize (or, maximize) f(X) = C.X
Subject to A.X ≥ (or, = or,  ) b
X ≥0
where A = (𝑎𝑖𝑗 )𝑚×𝑛 is an m  n matrix, C = (𝑐1 , 𝑐2 , … , 𝑐𝑛 ) is
a row vector, X = (𝑥1 , 𝑥2 , … , 𝑥𝑛 ) is a column vector, b =
(𝑏1 , 𝑏2 , … , 𝑏𝑚 ) is a column vector and 0 is an n-
dimensional null column vector.
4
(c) Minimize (or, maximize) C.X

Subject to x1P1 + x2P2 + . . . + xnPn ≥ (or, = or,  ) Po
X≥0
where Pj for j = 1, 2, . . ., n is the jth column of the matrix A
= (𝑎𝑖𝑗 ) and Po = b.
In some situations, it is convenient to define a new
unrestricted variable, say, 𝑥0 or z, which is equal to the value of the
objective function. Thus, the general LP problem can be
represented as below.
(d) Minimize (or, maximize) 𝑥0
n
Subject to 𝑥0 – c
j 1
j xj = 0
n
a
j 1
ij x j  (or  or )bi ; i = 1, 2, 3, . . ., m
𝑥𝑗 ≥ 0; j = 1, 2, 3, . . ., n
Example 1.2
Minimize z = 420𝑥1 + 300𝑥2
Subject to 4𝑥1 + 3𝑥2 ≥ 240
2𝑥1 + 𝑥2  100
𝑥1 , 𝑥2 ≥ 0
is a linear programming problem.
1.5 Standard form of LP problem
The characteristics of a standard LP problem are as below.
(i) The objective function is of the minimization or maximization
type.
(ii) All constraints are expressed in the form of equations, except
the non-negativity constraints which remain in inequality (≥)
form.
5
Setting the Scene
(a) If any constraint is of  type (or,  type) add slack

variable si (or, subtract surplus variable si) to convert it to
an equality, where si  0.
(b) If any constraint is of approximation type then introduce
slack and surplus variables at a time. As for example, 2x1
+ x2  3 is to be converted to 2x1 + x2 + s1 – s2 = 3 where
slack variable s1  0 and surplus variable s2  0.
(iii) The right hand side constant of each constraints must be non-
negative.
(iv) All decision variables are non-negative.
(a) If xi  0 then put xi = – xi/ where xi/  0
(b) If xi is unrestricted in sign then put xi = xi/ – xi// where xi/  0,
xi//  0.
(c) If xi  di (or xi  di) then put xi = xi/ + di (or, xi = di – xi/ )
where xi/  0.
Thus, the following LP problem is in the general standard form.
Minimize (or maximize)
𝑧 = 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛
Subject to
a11 x1  a12 x 2  ...  a1 j x j  ...  a1n x n  b1 

a 21 x1  a 22 x 2  ...  a 2 j x j  ...  a 2 n x n  b2 
 

ai1 x1  ai 2 x 2  ...  aij x j  ...  ain x n  bi 
 

a m1 x1  a m 2 x 2  ...  a mj x j  ...  a mn x n  bm 
xj ≥ 0; j = 1, 2, 3, . . ., n
Example 1.3
Minimize (or maximize) z = 420x1 + 300x2
6
Subject to 4x1 + 3x2 = 240

2x1 + x2 = 100
x1, x2 ≥ 0
is a standard linear programming problem.
Example 1.4
Convert the following LP problem into standard form.
Maximize
z = x1 + 4x2 + 3x3
Subject to
2x1 + 2x2 – 5x3  2
3x1 – x2 + 6x3  1
x1 + x2 + x3 = 4
x1  0, x2  0, x3 unrestricted.
Solution: Putting x2 = – x 2/ , x3 = x3/ – x3// ; x 2/  0, x3/  0, x3//  0 and
introducing slack variable s1  0 in 1st constraint and surplus
variable s2  0 in 2nd constraint, we get the problem as below.
Maximize
z = x1 – 4 x 2/ + 3 x3/ –3 x3// + 0 s1 + 0 s2
Subject to
2x1 – 2 x2/ – 5 x3/ + 5 x3// + s1 = 2
3x1 + x 2/ + 6 x3/ – 6 x3// – s2 = 1
x1 – x2/ + x3/ – x3// = 4
x1, x 2/ , x3/ , x3// , s1, s2  0.
This is the required standard form.
1.6 Canonical form of LP problem
Characteristics of the canonical form of a LP problem are given
below.
(i) Objective function is of minimization or maximization type
(ii) All constraints are expressed in the form of inequalities
connected by the sign ‘  ’ when the objective function is of
7
Setting the Scene
minimization type, or connected by the sign ‘  ’ when the

objective function is of maximization type.
(iii) All decision variables are non-negative.
(iv) Constraints are in ‘=’ type will be converted into two
inequalities; one is ‘  ’ type and the other is ‘  ’ type.
Thus, the general canonical LP problems are:
Minimize
𝑧 = 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛
Subject to
a11 x1  a12 x 2  ...  a1 j x j  ...  a1n x n  b1 

a 21 x1  a 22 x 2  ...  a 2 j x j  ...  a 2 n x n  b2 
 

ai1 x1  ai 2 x 2  ...  aij x j  ...  ain x n  bi 
 

a m1 x1  a m 2 x 2  ...  a mj x j  ...  a mn x n  bm 
xj ≥ 0; j = 1, 2, 3, . . ., n.
and
Maximize
𝑧 = 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛
Subject to
a11 x1  a12 x 2  ...  a1 j x j  ...  a1n x n  b1 

a 21 x1  a 22 x 2  ...  a 2 j x j  ...  a 2 n x n  b2 
 

ai1 x1  ai 2 x 2  ...  aij x j  ...  ain x n  bi 
 

a m1 x1  a m 2 x 2  ...  a mj x j  ...  a mn x n  bm 
xj ≥ 0; j = 1, 2, 3, . . ., n.
8
Example 1.5
Following LP problem is in canonical form.
Minimize z = 420x1 + 300x2
Subject to 4x1 + 3x2 ≥ 240
2x1 + x2 ≥ 100
x1, x2 ≥ 0
Example 1.6
Maximize z = 420x1 + 300x2
Subject to 4x1 + 3x2  240
2x1 + x2  100
x1, x2 ≥ 0
This is the canonical form of a LP problem.
Example 1.7
Transform the following linear programming problem to canonical
form.
Maximize z = x1 + 4x2 + 3x3
Subject to 2x1 + 2x2 – 5x3  2
3x1 – x2 + 6x3  1
x1 + x2 + x3 = 4
x1  3, x2  0 and x3 is unrestricted.
Solution: Given problem can be written as follows.
Maximize z = x1 + 4x2 + 3x3
3x1 – x2 + 6x3  1
x1 + x2 + x3  4
x1 + x2 + x3  4
Or, Maximize z = x1 + 4x2 + 3x3
–3x1 + x2 – 6x3  –1
–x1 – x2 – x3  – 4
x1 + x2 + x3  4
9
Setting the Scene
Putting x1 = x + 3, x2 = – x 2/ , x3 = x3/ – x3// ; x1/  0, x 2/  0, x3/  0,
/
1
x3//  0 in the last form, we get it as below.

Maximize z = x1/ – 4 x 2/ + 3 x3/ –3 x3// + 3
Subject to 2 x1/ – 2 x2/ – 5 x3/ + 5 x3//  – 4
– 3 x1/ – x 2/ – 6 x3/ + 6 x3//  8
– x1/ + x 2/ – x3/ + x3//  –1
x1/ – x 2/ + x3/ – x3//  1
x1/ , x 2/ , x3/ , x3//  0.
This is the required canonical form.
1.7 Advantages of linear programming
Followings are the main advantages of linear programming.
(i) It helps in attaining the optimum use of productive factors.
Linear programming indicates how a manager can utilize
productive factors most effectively by a better selection and
distribution of these elements. For example, more efficient
use of manpower and machines can be obtained by the use of
linear programming.
(ii) It improves the quality of decisions. The individual who
makes use of linear programming methods becomes more
objective than subjective. The individual having a clear
picture of the relationships within the basic equations,
inequalities or constraints can have a better idea about the
problem and its solution.
(iii) It can go a long way in improving the knowledge and skill of
tomorrow’s executives.
(iv) Although linear programming gives possible and practical
solutions, there might be other constraints operating outside
the problem, which must be taken into account, for example,
10
sales, demand, etc. Just because we can produce so many

units does not mean that they can be sold. Linear
programming method can handle such situations also because
it allows modification of its mathematical solutions.
(v) It highlights the bottleneck in the production processes.
When bottlenecks occur, some machines cannot meet
demand while others remain idle, at least part of the time.
Highlighting of bottlenecks is one of the most significant
advantages of linear programming.
1.8 Limitations of linear programming
Though having a wide field, linear programming has the following
limitations.
(i) For large problems having many limitations and constraints,
the computational difficulties are enormous, even when
assistance of large digital computers is available.
(ii) According to the linear programming problem, the solution
variables can take any value, whereas sometimes it happens
that some of the variables can have only integral values.
(iii) The model does not take into account the effect of time. The
operation research team must define the objective function
and constraints, which can change overnight due to internal
as well as external factors.
(iv) Sometimes, it is not possible to express both the objective
function and constraints in linear form.
1.9 Some definitions
Some important definitions related to the LP problem are discussed
below.
1.9.1 Objective function
Linear function 𝑧 = 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛
in the general linear programming problem which is to be
minimized (or maximized) is called the objective function. As for
11
Setting the Scene
example, 5x1 + 3x2 is the objective function of the following linear

programming problem.
2x1 + x2  6
x1, x2 ≥ 0
Note that the maximization of the objective function
𝑧 = 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛
and the minimization of the objective function
−𝑧 = −(𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑗 𝑥𝑗 + ⋯ + 𝑐𝑛 𝑥𝑛 )
under similar circumstances provide the same quantitative value.
In particular, if we solve the following two problems having same
constraints
2x1 + x2  6
x1, x2 ≥ 0
and
Maximize (–z) = – 5x1 – 3x2
2x1 + x2  6
x1, x2 ≥ 0,
we will get same solution to both problems.
Example 1.8
The following two linear programming problems
Subject to 2x1 + x2 ≥ 11
3x1 + 2x2  20
x1, x2 ≥ 0
and
12
Minimize (–z) = – 2x1 – 3x2

Subject to 2x1 + x2 ≥ 11
3x1 +2x2  20
x1, x2 ≥ 0
are equivalent to each other.
1.9.2 Constraints
The inequalities or equations in general linear programming
problem
a11 x1  a12 x 2  ...  a1 j x j  ...  a1n x n  (or  or )b1 

a 21 x1  a 22 x 2  ...  a 2 j x j  ...  a 2 n x n  (or  or )b2 
 

ai1 x1  ai 2 x 2  ...  aij x j  ...  ain x n  (or  or )bi 
 

a m1 x1  a m 2 x 2  ...  a mj x j  ...  a mn x n  (or  or )bm 
are called the constraints of the general linear programming
problem. Or, shortly A.X ≥ (or, = or,  ) b is called the constraints
of the general linear programming problem. As for example,
5x1 + 2x2 ≥ 15 and 7x1 + 3x2  6
are the constraints of the following linear programming problem
Maximize z = 4x1 + x2
7x1 + 3x2  6
x1, x2 ≥ 0.
1.9.3 Non-negativity conditions
Inequalities xj ≥ 0 (j = 1, 2, ..., n) are called the non-negativity
conditions to the general linear programming problem. As for
example, x1≥ 0 and x2 ≥ 0 are the non-negativity conditions to the
following linear programming problem
13
Setting the Scene
3x1 + x2  10
x1, x2 ≥ 0.
1.9.4 Solution
An n-tuple X = (x1, x2, ..., xn) of real numbers that satisfies all the
constraints of the general linear programming problem is called the
solution of the general linear programming problem. As for
example, X = (x1, x2) = (4, 0) is a solution to the following linear
programming problem
x1 + 2x2  7
x1, x2 ≥ 0.
1.9.5 Feasible solution
A feasible solution to a linear programming (LP) problem is a
solution, which satisfies the constraints (equality or inequality
constraints and the non-negativity constraints). As for example, X
= (1, 1, 0, 0) is a feasible solution of the following linear
programming problem.
Maximize z = x1 + 2x2 + x3 + 2x4
Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0.
1.9.6 Basic solution
A basic solution to a linear programming (LP) problem with m
constraints in n variables is a solution obtained by setting (n – m)
variables equal to zero and solving for the remaining m variables,
provided that the determinant of the coefficients of these m
variables is non-zero. These m variables are called basic variables.
Basic solution may or may not be feasible solution and conversely
feasible solution may or may not be a basic solution. As for
example, X = (0, 1, 0, – ½) is a basic non-feasible solution of the
following linear programming problem
14

Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0.
Note: We consider a system of m simultaneous linear equations
with n unknowns (m < n) AX = b, where A is an m  n matrix of
rank m, X and b are m  1 column matrices. Let B be any m  m sub-
matrix formed by m linearly independent columns of A. Then the
solution to the resulting system found by taking (n – m) variables
not associated with the columns of B equal to zero is called basic
solution of the given system of simultaneous linear equations. The
sub-matrix B is called the basis matrix of the system, the columns
of B are basis vectors and other than basis vectors are non-basis
vectors. The m variables associated with the columns of B are said
to be basic variables and the others are non-basic variables.
Example 1.9
Find all basic solutions to the following system of linear equations
4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8.
Solution: We can write the given system in the matrix form as
follows
 x1 
 
 4 2 3  8 x 
AX = b where, A =   , X =  2 
 3 5 4  6 x
 3
x 
 4
6
and b =   .
8
The column vectors of the coefficient matrix of A are
 4  2  3   8
a1 =   , a2 =   , a3 =   and a4 =  
 3 5  4   6
15
Setting the Scene
Since the rank of A is 2, the given system is consistent. We find

4
c 2 = 6 square sub matrices taking two at a time from a1, a2, a3, a4.
 4 2  4 3  4  8
B1=(a1,a2)=   , B2=(a1,a3)=   , B3=(a1,a4)=   ,
 3 5  3 4  3  6
 2 3  2  8  3  8
B4=(a2,a3)=   , B5=(a2,a4)=   , B6= (a3,a4)=   .
 5 4  5  6  4  6
Values of the determinants formed by above sub matrices are |B1| =
14, |B2| = 10, |B3| = 0, |B4| = – 7, |B5| = 28, |B6| = 14. Since |B3| = 0,
B3 is singular matrix. So, only five basic solutions will be found
corresponding to five basis matrices B1, B2, B4, B5 and B6.
 4 2
(i) For basis matrix B1= (a1, a2) =   , basic variables are x1, x2
 3 5
and non-basic variables are x3, x4. Putting x3 = x4 = 0, the system
reduce to
4x1 + 2x2 = 6
3x1 + 5x2 = 8.
Subtracting 2nd equation from 1st equation after multiplying first
one by 5 and 2nd one by 2, we get 14x1 = 14. So, x1 = 1.
Again subtracting 1st equation from 2nd equation after multiplying
first one by 3 and 2nd one by 4, we get 14x2 = 14. So, x2 = 1. Hence
the basic solution is X1 = (1, 1, 0, 0).
 4 3
(ii) For basis matrix B2 = (a1,a3) =   , basic variables are x1, x3
 3 4
reduce to
4x1 + 3x3 = 6
3x1 + 4x3 = 8.
After solving this system we get, x1 = 0, x3 = 2. Hence the basic
solution is X2 = (0, 0, 2, 0).
16
 2 3
(iii) For basis matrix B4 = (a2,a3)=   , basic variables are x2, x3
 5 4
reduce to
2x2 + 3x3 = 6
5x2 + 4x3 = 8
Solving this system we get, x2 = 0, x3 = 2. Hence the basic solution
is X3 = (0, 0, 2, 0).
 2  8
(iv) For basis matrix B5 = (a2, a4) =   , basic variables are
 5  6
x2, x4 and non-basic variables are x1, x3. Putting x1 = x3 = 0, the
system reduce to
2x2 – 8x4 = 6
5x2 – 6x4 = 8.
Solving this system we get, x2 = 1, x4 = –½. Hence the basic
solution is X4 = (0, 1, 0, – ½).
 3  8
(v) For basis matrix B6 = (a3, a4) =   , basic variables are x3,
 4  6
x4 and non-basic variables are x1, x2. Putting x1 = x2 = 0, the system
reduce to
3x3 – 8x4 = 6
4x3 – 6x4 = 8.
is X5 = (0, 0, 2, 0).
Note: The solutions X1 = (1,1,0,0), X2 = (0,0,2,0), X3 = (0,0,2,0), X5
= (0,0,2,0) are basic feasible solutions and X4 = (0,1,0,– ½) basic
non-feasible solution. Also (0,0,2,0) is degenerate basic feasible
solution, (1,1,0,0) is non-degenerate basic feasible solution and
(0,1,0,– ½) is non-degenerate basic solution.
17
Setting the Scene
1.9.7 Degenerate basic solution

A basic solution to a linear programming problem is called
degenerate if the value of any basic variable is zero. As for
example, X = (0, 0, 2, 0) is a degenerate basic solution to the
following linear programming problem.
Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0
1.9.8 Non-degenerate basic solution
A basic solution to a linear programming problem is called non-
degenerate if the values of all basic variables are non-zero. As for
example, X = (1, 1, 0, 0) is a non-degenerate basic solution to the
Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0
1.9.9 Basic feasible solution
A basic feasible solution is a basic solution satisfying non-
negativity conditions to a linear programming problem. That is, a
basic solution which satisfies all subjective conditions (constraints
and non-negativity conditions) is called a basic feasible solution.
As for example, X = (1, 1, 0, 0) is a basic feasible solution to the
Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0
1.9.10 Degenerate basic feasible solution
A degenerate basic solution to a linear programming problem is
called degenerate basic feasible solution if it satisfies all
18
constraints and non-negativity conditions. As for example, X = (0,

0, 2, 0) is a degenerate basic feasible solution to the following
linear programming problem.
Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0
1.9.11 Non-degenerate basic feasible solution
A non-degenerate basic solution to a linear programming problem
is called non-degenerate basic feasible solution if it satisfies all
constraints and non-negativity conditions. As for example, X = (1,
1, 0, 0) is a non-degenerate basic feasible solution to the following
linear programming problem.
Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0
1.9.12 Optimum solution
Any feasible solution that optimizes (minimizes or maximizes) the
objective function of an LP problem is called an optimum solution
to that LP problem. The term “optimal solution” is also used for
“optimum solution”. As for example, X = (0, 0, 2, 0) is an optimum
solution to the following linear programming problem.
Subject to 4x1 + 2x2 + 3x3 – 8x4 = 6
3x1 + 5x2 + 4x3 – 6x4 = 8
x1, x2, x3, x4≥ 0
Note: Optimum solution must be a basic feasible solution. To
solve the above problem using Mathematica 5 software, we use the
code given below.
ConstrainedMax[x1 +2x2 + x3 +2x4, {4x1 + 2x2 + 3x3 – 8x4 = = 6,
3x1 + 5x2 + 4x3 – 6x4 = = 8}, {x1, x2, x3, x4}]
19
Setting the Scene
1.9.13 Slack variables

If the constraints of a general LP problem be
n
a
j 1
ij x j  bi ; i = 1, 2, 3, . . ., m,
then the non-negative variables xn+i satisfying

n
a
j 1
ij x j  x n i  bi ; i = 1, 2, 3, ..., m
are called slack variables. That is, when the constraints are
inequalities connected by the sign ‘  ’, in each inequality an extra
non-negative variable is to be added to the left hand side of the
inequality to convert it into an equation. The variables which are
added to convert the inequalities to equations are known as slack
variables. As for example, we add a non-negative variable x3 to the
left hand side of the inequality
2x1 + 3x2  4
to convert it an equation
2x1 + 3x2 + x3 = 4.
Thus x3 is a slack variable. This variable is called slack variable,
because
Slack = Requirement – Production.
1.9.14 Surplus variables
Let the constraints of a general LP problem be
n
a j 1
ij x j  bi ; i = 1, 2, 3, . . ., m,
then the non-negative variables x n i satisfying

n
a
j 1
ij x j  x n i  bi ; i = 1, 2, 3, ..., m
are called surplus variables. That is, when the constraints are
inequalities connected by the sign ‘  ’, in each inequality an extra
non-negative variable is to be subtracted from the left hand side of
20
the inequality to convert it into an equation. The variables which

are subtracted to convert the inequalities to equations are known as
surplus variables. As for example, we subtract a non-negative
variable x3 from the left hand side of the inequality
2x1 + 3x2  4
to convert it an equation
2x1 + 3x2 – x3 = 4.
Thus x3 is a surplus variable. x n i is called a surplus variable,
because
Surplus = Production – Requirement.
Example 1.10
Change the linear programming (LP) problem
Subject to 2x1 + x2  2
3x1 + 4x2 ≥ 12
x1, x2 ≥ 0
to standard form.
Solution: Adding slack variable s1 to L.H.S. of the first constraint
and subtracting surplus variable s2 from L.H.S. of the second
constraint, we can change the given LP problem to standard form
as given below.
Subject to 2x1 + x2 +s1 = 2
3x1 + 4x2 – s2 = 12
x1, x2, s1, s2 ≥ 0
Theorem 1.1
If a linear programming problem A.X = b, X ≥ 0 (where A is an m 
n matrix with rank m, X, b and 0 are m  1 column vectors and m <
n) has a feasible solution then it has at least one basic feasible
solution.
Proof: Let X = (x1, x2, ..., xn) be a feasible solution of the linear
programming problem A.X = b, X ≥ 0. Among the n components of
21
Setting the Scene
the feasible solution, suppose first k (k  n) components be positive

and the remaining (n–k) components be zero. That is, X = (x1, x2,
..., xk, 0, 0, ...,0) and if a1, a2, ..., ak be the column vectors of the
matrix A corresponding to the variables x1, x2, ..., xk then
k
x1a1 + x2a2 + ... + xk ak = b i.e, x a
i 1
i i =b (i)
Case – 1: If k  m and column vectors a1, a2, ..., ak are linearly

independent then it is clear that the feasible solution is a basic
feasible solution.
Case – 2: If k > m then the column vectors a1, a2, ..., ak are not
linearly independent and hence the solution X = (x1, x2, ..., xk, 0, 0,
..., 0) is not a basic feasible solution. So there exist scalar s1, s2, ...,
sk not all zero such that
k
s1a1 + s2a2 + ... + sk ak = 0; i.e, s a
i 1
i i =0 (ii)
Here, at least one si is positive (if necessary, multiply both sides of

s 
(ii) by –1 to make a si positive). Let p = Max i  , i = 1, 2, ..., k
i
 xi 
and then p > 0 as all xi > 0 and at least one si > 0 for i = 1, 2, ..., k.
1
Doing (i) – (ii), we have
p
k k
si s
i 1
( x i 
p
) a i = b or 
i 1
xi/ a i = b where xi/ = xi – i .
p
It implies that
 s s s 
X =  x1  1 , x 2  2 ,..., x k  k ,0,0,...,0 
/
 p p p 
= ( x1/ , x 2/ , ..., x k/ , 0, 0, ..., 0)
is a solution to the system of linear equations A.X = b.
22
si
Since p ≥ [equality holds for at least one i because p =
xi
s  s s
Max i  ], we have xi ≥ i  xi – i ≥ 0  xi/ ≥ 0 and for at
i
 xi  p p
/
least one i, xi = 0. Thus, X = ( x1 , x 2/ , ..., x k/ , 0, 0, ..., 0) with
/ /
maximum (k –1) positive components is a feasible solution of the

LP problem. Repeatedly applying this method, a feasible solution
with only m positive components will be obtained finally. And
then if the column vectors corresponding to the non-zero variables
are linearly independent, the final solution will be non-degenerate
basic feasible solution.
Case – 3: If k  m and column vectors a1, a2, ..., ak are not linearly
independent then the feasible solution is not a basic feasible
solution. Applying the above procedure given in case-2, the
number of non-zero variables can be reduced to r (r < m) of which
all the non-zero values are positive so that the associative column
vectors a1, a2, ..., ar are linearly independent and the solution is a
basic feasible solution.
Example 1.11
If x1 = 2, x2 = 3, x3 = 1 is a feasible solution to the system of linear
equations
2x1 + x2 + 4x3 = 11
3x1 + x2 + 5x3 = 14,
then reduce the above feasible solution to a basic feasible solution.
Solution: The matrix representation of the given system of linear
equations is
 x1 
 2 1 4    11 
A.X = b , where A =   , X =  x 2  and b =  
 3 1 5 x  14 
 3
23
Setting the Scene
Rank of the matrix A is 2 and hence the given equations are

linearly independent. The given solution (x1, x2, x3) = (2, 3, 1) is
feasible as all components are non-negative, but not basic feasible
because basic feasible solution to a system of two equations cannot
have more than 2 non-zero components.
Let the column vectors of the coefficient matrix of A are
 2  1  4
a1 =   , a2 =   and a3 =  
 3  1 5
and they are not linearly independent because more than two
vectors of two components cannot be independent. So, there exists
three scalars s1, s2, s3 not all zero such that
s1a1 + s2a2 + s3a3 = 0 (i)
 2  1  4  2s1  s 2  4s3   0 
 s1   + s2   + s3   = 0    =  
 3  1 5  3s1  s 2  5s3   0 
2s1  s 2  4s3  0 s s2 s
  1   3
3s1  s 2  5s3  0 5  4 12  10 2  3
s s s
 1  2  3   ( 1) (say)
1 2 1
 s1 = 1, s2 = 2, s3 = –1
Equation (i) implies
a1 + 2a2 – a3 = 0 (ii)
s 
Let p = Max i  for xi > 0
i
 xi 
s1 s 2 s 3 1 2 1 2
 p = Max ( , , )  p = Max ( , , )=
x1 x 2 x 3 2 3 1 3
Hence, the new feasible solution is
 s s s   1 2 1
X1 =  x1  1 , x2  2 , x3  3  =  2  2 ,3  2 ,1  2 
 p p p  3 3 3 
24
1 5
=( , 0, ).
2 2
We form the submatrix B1 by using the coefficients of positive
variables in the new solution, i.e., B1 is the matrix consisting of
first and third columns of the matrix A. Thus,
 2 4
B1 = (a1, a3) =   and |B1| = – 2  0.
 3 5
So, vectors a1, a3 are linearly independent and hence, X1= (1/2,
0, 5/2) is a basic feasible solution corresponding to the given
feasible solution.
Again we can rewrite equation (ii) as follows:
–a1 – 2a2 + a3 = 0 and so s1 = –1, s2 = –2, s3 = 1.
s 
i
 xi 
s s s 1  2 1
 p = Max ( 1 , 2 , 3 )  p = Max ( , , ) = 1.
x1 x 2 x 3 2 3 1
Hence we get another feasible solution as follows:
 s s s   1 2 1
X2 =  x1  1 , x2  2 , x3  3  =  2  , 3 , 1 
 p p p  1 1 1
= (3, 5, 0).
Submatrix formed by using the column vectors of A corresponding
to positive variables of the solution (3, 5, 0) is
 2 1
B2 = (a1, a2) =   and |B2| = –1  0.
 3 1
So, vectors a1, a2 are linearly independent and hence X2 = (3,
5, 0) is another basic feasible solution corresponding to the given
feasible solution. Therefore, (1/2, 0, 5/2) and (3, 5, 0) are required
basic feasible solutions corresponding to the given feasible
solution.
25
Setting the Scene
1.10 Some done examples

Following done examples can help the readers for better
understanding.
Example 1.12
Find the numerical values of
5 4
(i)  3i
i 1
2
and (ii)  (3  2 x)
x 1
5
Solution: (i)  3i
i 1
2
= 3.12 + 3.22 + 3.32 + 3.42 + 3.52
= 3 + 12 + 27 + 48 + 75
= 165
4
(ii)  (3  2 x) = (3 + 2.1) + (3 + 2.2) + (3 + 2.3) + (3 + 2.4)
x 1
= 5 + 7 + 9 + 11 = 32
Example 1.13
Write a linear programming (LP) problem consisting of two
variables.
Solution: Following LP problem consists of two variables namely
x and y.
Minimize z = 5x – 4y
Subject to 3x + 2y  10
6x – 3y  2
x, y  0
Example 1.14
Convert the following minimization LP problem to maximization
type.
Minimize z = 7x1 – 3x2 + 5x3
Subject to 2x1 – 3x2 + 2x3  – 5
5x1 + 2x2 + 3x3  2
3x1 + 2x3 = 20
26
x1, x2, x3  0
Solution: Minimization LP problem usually convert to
maximization LP problem just multiplying the objective function
by –1. Thus the required maximization LP problem is
Maximize (–z) = – 7x1 + 3x2 – 5x3
Subject to 2x1 – 3x2 + 2x3  – 5
5x1 + 2x2 + 3x3  2
3x1 + 2x3 = 20
x1, x2, x3  0.
Example 1.15
Transform the following linear programming problem to the
standard form.
Minimize 2x + y + 4z
Subject to –2x + 4y  5
x + 2y + z  5
2x + 3z  5
x, y, z  0
Solution: Adding slack variable u  0, v  0 to the left hand side of
the first and third constraints respectively and also subtracting
surplus variable w  0 from the left hand side of the second
constraint, we get the following standard LP problem of the given
problem.
Minimize 2x + y + 4z + 0u + 0v + 0w
Subject to –2x + 4y + u = 5
x + 2y + z – w = 5
2x + 3z + v = 5
x, y, z, u, v, w  0
Example 1.16
Make the standard form of the following LP problem.
Maximize z = 3x1 + 2x2 + 5x3
Subject to 2x1 – 3x2  3
x1 + 2x2 + 3x3  5
27
Setting the Scene
3x1 + 2x3  2
x1, x2, x3  0
Solution: Adding slack variables x4  0 and x5  0 to the left hand
sides of first and third constraints respectively and subtracting
surplus variable x6  0 from the left hand side of second constraint,
we get
Maximize z = 3x1 + 2x2 + 5x3 + 0x4 + 0x5 + 0x6
Subject to 2x1 – 3x2 + x4 = 3
x1 + 2x2 + 3x3 – x6 = 5
3x1 + 2x3 + x5 = 2
x1, x2, x3, x4, x5, x6  0.
Example 1.17
Transform the following maximization LP problem to
minimization type and then reduce it to standard form.
Subject to 2x1 – 3x2 + x3  – 3
x1 + 2x2 + 3x3  5
5x1 + 2x3 = 8
x1, x2, x3  0
Solution: Changing the sign of the objective function, we get the
minimization form of the given LP problem as below.
Minimize (–z) = – 2x1 – 2x2 – 5x3
Subject to 2x1 – 3x2 + x3  – 3
x1 + 2x2 + 3x3  5
5x1 + 2x3 = 8
x1, x2, x3  0
Adding slack variable x4  0 to the LHS of the first constraint and
subtracting surplus variable x5  0 from the
LHS of second constraint, we get
Minimize (–z) = – 2x1 – 2x2 – 5x3
Subject to 2x1 – 3x2 + x3 + x4 = – 3
28
x1 + 2x2 + 3x3 – x5 = 5
5x1 + 2x3 = 8
x1, x2, x3, x4, x5  0.
Multiplying first constraint by –1, we have
Minimize (–z) = – 2x1 – 2x2 – 5x3
Subject to –2x1 + 3x2 – x3 – x4 = 3
x1 + 2x2 + 3x3 – x5 = 5
5x1 + 2x3 = 8
x1, x2, x3, x4, x5  0.
Example 1.18
Reduce the linear programming (LP) problem given below to
standard form.
Subject to 2x1 – 5x2  5
3x1 + 2x2 + 3x3  – 3
5x1 + x2 + 2x3  10
x1, x2, x3  0
Solution: Adding slack variables x4  0 and x5  0 to the left hand
sides of first and third constraints respectively and subtracting
surplus variable x6  0 from the left hand side of second constraint,
we get
3x1 + 2x2 + 3x3 – x6 = – 3
5x1 + x3 + 2x5 = 10
x1, x2, x3, x4, x5, x6  0.
Multiplying second constraint by –1, we have
–3x1 – 2x2 – 3x3 + x6 = 3
5x1 + x3 + 2x5 = 10
29
Setting the Scene
x1, x2, x3, x4, x5, x6  0.

This is the required standard form of the given LP problem.
Example 1.19
Convert the following linear programming problem into the
standard form:
3x1 – x2 + 6x3  2
x1 + x2 + 3x3 = 5
Solution: Putting x2 = – x 2/ , x3 = x3/ – x3// ; x 2/  0, x3/  0, x3//  0 and
introducing slack variable s1  0 in first constraint and surplus
variable s2  0 in second constraint, we get the problem as follows:
Maximize z = 5x1 – 4 x 2/ + 3 x3/ –3 x3// + 0.s1 + 0.s2
Subject to 2x1 – 2 x2/ – 5 x3/ + 5 x3// + s1 = 9
3x1 + x 2/ + 6 x3/ – 6 x3// – s2 = 2
x1 – x2/ + 3 x3/ –3 x3// = 5
x1, x 2/ , x3/ , x3// , s1, s2  0.
Example 1.20
Change the linear programming (LP) problem
9x1 – x2 + 6x3  5
x1 + 3x2 + x3 = 1
x1  2, x2  0, x3 unrestricted
to the standard form of a LP problem.
Solution: Putting x1 = x1/ + 2, x2 = – x 2/ , x3 = x3/ – x3// ; x1/  0, x 2/  0,
x3/  0, x3//  0 and adding slack variable s1  0 to the left hand side
30
of the first constraint and subtracting surplus variable s2  0 from

the left hand side of the second constraint, we get the problem as
follows:
Maximize z = 8( x1/ + 2) + 4(– x 2/ ) + 3( x3/ – x3// ) + 0.s1 + 0.s2
Subject to 6( x1/ + 2) + 2(– x2/ ) – 5( x3/ – x3// ) + s1 = 3
9( x1/ + 2) – (– x 2/ ) + 6( x3/ – x3// ) – s2 = 5
( x1/ + 2) + 3(– x2/ ) + ( x3/ – x3// ) = 1
x1/ , x 2/ , x3/ , x3// , s1, s2  0.
After simplifying, we get
Maximize z = 8 x1/ – 4 x 2/ + 3 x3/ –3 x3// + 0.s1 + 0.s2 + 16
Subject to 6 x1/ – 2 x2/ – 5 x3/ + 5 x3// + s1 = – 9
9 x1/ + x 2/ + 6 x3/ – 6 x3// – s2 = –12
x1/ – 3 x2/ + x3/ – x3// = – 1
x1/ , x 2/ , x3/ , x3// , s1, s2  0.
Multiplying all constraint equations by – 1, we get the following
form:
Maximize z = 8 x1/ – 4 x 2/ + 3 x3/ –3 x3// + 0.s1 + 0.s2 + 16
Subject to – 6 x1/ + 2 x2/ + 5 x3/ – 5 x3// – s1 = 9
– 9 x1/ – x 2/ – 6 x3/ + 6 x3// + s2 = 12
– x1/ + 3 x2/ – x3/ + x3// = 1
x1/ , x 2/ , x3/ , x3// , s1, s2  0.
This is the required standard form of the given LP problem.
Example 1.21
Transform the following linear programming (LP) problem to
standard form.
Minimize 5x + 2y + 4z
Subject to – 2x + 3y = 7
31
Setting the Scene
3x + 2y + z = 8
2x + 5z  11
x, y, z  0
Solution: Introducing both of the slack variable u  0, and the
surplus variable v  0 to the left hand side of the third constraint,
we have the following standard LP form of the given problem:
Minimize 5x + 2y + 4z + 0u + 0v
Subject to – 2x + 3y = 7
3x + 2y + z = 8
2x + 5z + u – v = 11
x, y, z, u, v  0.
Example 1.22
Reduce the following linear programming (LP) problem to the
standard form of an LP problem:
Minimize z = 5x1 + 4x2 + x3
9x1 – x2 + 6x3  5
2x1 + 3x2 + x3  1
Solution: Putting x1 = x1/ + 2, x2 = – x 2/ , x3 = x3/ – x3// ; x1/  0, x 2/  0,
x3/  0, x3//  0 and adding slack variable s1  0 to the left hand side
of the first constraint and subtracting surplus variable s2  0 from
left hand side of the second constraint and adding s3 – s4 (here, s3 
0 is slack variable and s4  0 is surplus variable) to the left hand
side of the third constraint, we get the problem as follows:
Minimize z = 5( x1/ +2)+ 4(– x 2/ )+( x3/ – x3// )
Subject to 6( x1/ + 2) + 2(– x2/ ) – 5( x3/ – x3// ) + s1 = 3
9( x1/ + 2) – (– x 2/ ) + 6( x3/ – x3// ) – s2 = 5
2( x1/ + 2) + 3(– x2/ ) + ( x3/ – x3// ) + (s3 – s4) = 1
x1/ , x 2/ , x3/ , x3// , s1, s2, s3, s4  0.
32
After simplifying, we get

Minimize z = 5 x1/ – 4 x 2/ + x3/ – x3// + 10
Subject to 6 x1/ – 2 x2/ – 5 x3/ + 5 x3// + s1 = – 9
9 x1/ + x 2/ + 6 x3/ – 6 x3// – s2 = –12
2 x1/ – 3 x2/ + x3/ – x3// + s3 – s4 = – 3
x1/ , x 2/ , x3/ , x3// , s1, s2, s3, s4  0.
Multiplying all constraint equations by –1, we get the following
form:
Minimize z = 5 x1/ – 4 x 2/ + x3/ – x3// + 10
Subject to – 6 x1/ + 2 x2/ + 5 x3/ – 5 x3// – s1 = 9
– 9 x1/ – x 2/ – 6 x3/ + 6 x3// + s2 = 12
–2 x1/ + 3 x2/ – x3/ + x3// – s3 + s4 = 3
x1/ , x 2/ , x3/ , x3// , s1, s2, s3, s4  0.
Example 1.23
Convert the following linear programming problem into the
standard form:
Minimize z = –3x1 + 4x2 –2x3 + 5x4
Subject to 4x1 – x2 + 2x3 – x4 = –2
x1 + x2 + 3x3 – x4  5
–2x1 + 3x2 – x3 + 2x4  1
x1, x2  0, x3  0, x4 unrestricted in sign.
Solution: Putting x3 = – x3/ , x4 = x 4/ – x 4// ; x3/  0, x 4/  0, x 4//  0 and
adding slack variable s1  0 to the left hand side of 2nd constraint
and subtracting surplus variable s2  0 from left hand side of 3rd
constraint, we get the problem as follows:
Minimize z = –3x1 + 4x2 –2(– x3/ ) + 5( x 4/ – x 4// ) + 0s1 + 0s2
Subject to 4x1 – x2 + 2(– x3/ ) – ( x 4/ – x 4// ) = –2
33
Setting the Scene
x1 + x2 + 3(– x3/ ) – ( x 4/ – x 4// ) + s1 = 5

–2x1 + 3x2 – (– x3/ ) + 2( x 4/ – x 4// ) – s2 = 1
x1, x2, x3/ , x 4/ , x 4// , s1, s2  0.
Multiplying first constraint by –1, we get
Minimize z = –3x1 + 4x2 +2 x3/ + 5 x 4/ – 5 x 4//
Subject to – 4x1 + x2 + 2 x3/ + x 4/ – x 4// = 2
x1 + x2 –3 x3/ – x 4/ + x 4// + s1 = 5
–2x1 + 3x2 + x3/ + 2 x 4/ – 2 x 4// – s2 = 1
x1, x2, x3/ , x 4/ , x 4// , s1, s2  0
Example 1.24
Transform the following linear programming (LP) problem to the
canonical form of an LP problem:
Subject to – 2x + 4y  3
x + 2y + z  4
2x + 3z  5
x, y, z  0.
Solution: Multiplying first and third constraints by –1, we get
Subject to 2x – 4y  – 3
x + 2y + z  4
– 2x – 3z  – 5
x, y, z  0
Example 1.25
Transform the linear programming problem
Maximize 2x + 3y + 4z
34
x + 2y + z  4
2x + 3z  5
x, y, z  0
to canonical form.
Solution: Multiplying second constraint by –1, we get
Maximize 2x + 3y + 4z
– x – 2y – z  – 4
2x + 3z  5
x, y, z  0
Example 1.26
Reduce the following LP problem into canonical form:
Minimize z = 2x1 + x2
–3x1 + x2  2
x1, x2  0.
Solution: We can convert the equality constraint 3x1 + 2x2 = 15
into two inequalities 3x1 + 2x2  15 and 3x1 + 2x2  15. Thus, the
given LP problem becomes
Subject to 3x1 + 2x2  15
3x1 + 2x2  15
–3x1 + x2  2
x1, x2  0
Multiplying second constraint by –1, we have
Subject to 3x1 + 2x2  15
–3x1 – 2x2  –15
–3x1 + x2  2
x1, x2  0.
This is the required canonical form of the given LP problem.
35
Setting the Scene
Example 1.27
Reduce the following LP problem into canonical form:
Maximize z = x1 + x2
Subject to x1 + 2x2  5
–3x1 + x2  3
x1  0, x2 is unrestricted in sign.
Solution: Taking x2 = x 2/ – x 2// ; x 2/ , x 2//  0, we have
Maximize z = x1 + x 2/ – x 2//
Subject to x1 + 2( x 2/ – x 2// )  5
–3x1 + x 2/ – x 2//  3
x1, x 2/ , x 2//  0
Multiplying second constraint by –1, we have the canonical form
of the given LP problem as below:
Maximize z = x1 + x 2/ – x 2//
Subject to x1 + 2 x 2/ – 2 x 2//  5
3x1 – x 2/ + x 2//  –3
x1, x 2/ , x 2//  0.
Example 1.28
Convert the following linear programming problem to canonical
form:
Minimize z = 2x1 + 3x2 + 4x3
3x1 – x2 + 6x3  1
x1 + x2 + x3 = 4
x1  –3, x2  0, x3 is unrestricted.
Solution: Replacing the equality constraint by two inequalities, we
have
3x1 – x2 + 6x3  1
36
x1 + x2 + x3  4
x1 + x2 + x3  4
x1  –3, x2  0, x3 unrestricted.
Because this is a minimization type LP problem, all constraints
must be greater equal (  ) type in the canonical form. So,
multiplying first and fourth constraints by –1, we have
Subject to –2x1– 2x2+ 5x3  –2
3x1 – x2 + 6x3  1
x1 + x2 + x3  4
– x1 – x2 – x3  – 4
x1  –3, x2  0, x3 unrestricted.
All variables must be non-negative. So we replace x1 by x1/ –3, x2
by – x 2/ and x3 by x3/ – x3// where x1/  0, x 2/  0, x3/  0, x3//  0,
and then we have
Minimize z = 2 x1/ – 3 x 2/ + 4 x3/ – 4 x3// – 6
Subject to – 2 x1/ + 2 x2/ + 5 x3/ – 5 x3//  – 8
3 x1/ + x 2/ + 6 x3/ – 6 x3//  10
x1/ – x 2/ + x3/ – x3//  7
– x1/ + x 2/ – x3/ + x3//  –7
x1/ , x 2/ , x3/ , x3//  0
Example 1.29
Convert the following LP problem in canonical form:
3x1 – x2 + 6x3  1
x1 + x2 + x3 = 4
37
Setting the Scene
Solution: Replacing the equality constraint by two inequalities, we

get
3x1 – x2 + 6x3  1
x1 + x2 + x3  4
x1 + x2 + x3  4
Because this is a maximization type LP problem, all constraints
must be less equal (  ) type in the canonical form. So, multiplying
second and third constraints by –1, we have
– 3x1 + x2 – 6x3  –1
– x1 – x2 – x3  – 4
x1 + x2 + x3  4
All variables must be non-negative. So, putting x1 = x1/ + 3, x2 = –
x 2/ , x3 = x3/ – x3// ; x1/  0, x 2/  0, x3/  0, x3//  0, we get the canonical
form of the given problem as below:
Maximize z = 2 x1/ – 3 x 2/ + 4 x3/ – 4 x3// + 6
Subject to 2 x1/ – 2 x2/ – 5 x3/ + 5 x3//  – 4
– 3 x1/ – x 2/ – 6 x3/ + 6 x3//  8
– x1/ + x 2/ – x3/ + x3//  –1
x1/ – x 2/ + x3/ – x3//  1
x1/ , x 2/ , x3/ , x3//  0.
Example 1.30
Find the basic feasible solutions to the system of linear equations
given below
38
4x1 + 2x2 – 3x3 = 1

6x1 + 4x2 – 5x3 = 1.
follows:
 x1 
 4 2  3    1
AX = b where, A =   , X =  x 2  and b =   .
 6 4  5 x   1
 3
The column vectors of the coefficient matrix A are
 4  2   3
a1 =   , a2 =   and a3 =   .
6  4   5
3
c 2 = 3 square sub matrices taking two at a time from a1, a2, a3 and
they are
 4 2  4  3
B1 = (a1, a2) =   , B2 = (a1, a3) =   and
 6 4  6  5
 2  3
B3 = (a2, a3) =   .
 4  5
Values of the determinants formed by the above sub matrices are
|B1| = 4, |B2| = –2, |B3| = 2.
Since all sub matrices are non-singular, we shall find three basic
solutions corresponding to three basis matrices B1, B2 and B3.
 4 2
(i) For basis matrix B1= (a1, a2) =   , x1 and x2 are basic
 6 4
variable, and x3 is non-basic variable. Putting x3 = 0, the system
reduce to
4x1 + 2x2 = 1
6x1 + 4x2 = 1.
We multiply first equation by 2 and then subtract second equation
from it. This implies 2x1 = 1. So, x1 = 1/2. Putting x1 = 1/2 in any
39
Setting the Scene
of the above equation, we have x2 = – 1/2. Hence the basic

solution is X1 = (½, –½, 0), which is not feasible as it contains
negative number.
 4  3
(ii) For basis matrix B2 = (a1, a3) =   , basic variables are
 6  5
x1, x3 and non-basic variable is x2. Putting x2 = 0, the system
reduce to
4x1 – 3x3 = 1
6x1 – 5x3 = 1
is X2 = (1, 0, 1) which is also feasible.
 2  3
(iii) For basis matrix B3 = (a2, a3) =   , basic variables are
 4  5
x2, x3 and non-basic variables are x1. Putting x1 = 0, the system
reduce to
2x2 – 3x3 = 1
4x2 – 5x3 = 1
Solving this system we get, x2 = –1, x3 = –1. Hence the basic
solution is X3 = (0, –1, –1) which is not feasible as it contains
negative numbers.
Therefore, X2 = (1, 0, 1) is the only one basic feasible solution to
the given system.
Example 1.31
Find all basic solutions to the simultaneous linear equations:
4x1 + 2x2 + 3x3 – 8x4 = 12
3x1 + 5x2 + 4x3 – 6x4 = 16.
follows:
40
 x1 
 
 4 2 3  8 x  12 
AX = b, where A =   , X =  2  and b =   .
 3 5 4  6 x
 3 16 
x 
 4
The column vectors of the coefficient matrix A are
 4  2  3   8
a1 =   , a2 =   , a3 =   and a4 =   .
 3 5  4   6
4
c 2 = 6 square sub matrices taking two at a time from a1, a2, a3, a4
and they are
 4 2  4 3
B1 = (a1, a2) =   , B2 = (a1, a3) =   , B3 = (a1,
 3 5  3 4
 4  8  2 3
a4) =   , B4 = (a2, a3) =   ,
 3  6  5 4
 2  8  3  8
B5 = (a2, a4) =   , B6 = (a3, a4) =   .
 5  6  4  6
Values of the determinants formed by the above sub matrices are
|B1| = 14, |B2| = 10, |B3| = 0, |B4| = – 7, |B5| = 28, |B6| = 14.
Since |B3| = 0, B3 is a singular matrix. So, only five basic solutions
will be found corresponding to five basis matrices B1, B2, B4, B5
and B6.
 4 2
(i) For basis matrix B1= (a1, a2) =   , basic variables are x1, x2
 3 5
and non-basic variables are x3, x4. Putting x3 = x4 = 0, the
system reduce to
4x1 + 2x2 = 12
3x1 + 5x2 = 16.
41
Setting the Scene
Subtracting double of the second equation from the multiplication

of the first equation with 5, we get 14x1 = 28. So, x1 = 2. Putting x1
= 2, we have x2 = 2. Thus, a basic solution is X1 = (2, 2, 0, 0).
 4 3
(ii) For basis matrix B2 = (a1, a3) =   , basic variables are x1,
 3 4
reduce to
4x1 + 3x3 = 12
3x1 + 4x3 = 16.
Solving this system we get, x1 = 0, x3 = 4. Thus, a second basic
solution is X2 = (0, 0, 4, 0).
 2 3
(iii) For basis matrix B4 = (a2, a3) =   , basic variables are x2,
 5 4
reduce to
2x2 + 3x3 = 12
5x2 + 4x3 = 16
Solving this system we get, x2 = 0, x3 = 4. Thus, a third basic
solution is X3 = (0, 0, 4, 0).
 2  8
(iv) For basis matrix B5 = (a2, a4) =   , basic variables are
 5  6
x2, x4 and non-basic variables are x1, x3. Putting x1 = x3 = 0, the
system reduce to
2x2 – 8x4 = 12
5x2 – 6x4 = 16
Solving this system we get, x2 = 2, x4 = –1. Thus, a fourth basic
solution is X4 = (0, 2, 0, –1).
 3  8
 4  6
and x4. Putting non-basic variables x1 = x2 = 0, the system reduce to
42
3x3 – 8x4 = 12
4x3 – 6x4 = 16.
Solving this system, we get x3 = 4, x4 = 0. Thus, a fifth basic
solution is X5 = (0, 0, 4, 0).
Note: Solutions X1 = (2, 2, 0, 0), X2 = (0, 0, 4, 0), X3 = (0, 0, 4, 0),
X5 = (0, 0, 4, 0) are basic feasible solution but X4 = (0, 2, 0, –1) is
not a basic feasible solution. (0, 0, 4, 0) is a degenerate basic
feasible solution, (2, 2, 0, 0) is a non-degenerate basic feasible
solution and (0, 2, 0, –1) is a non-degenerate basic solution.
Example 1.32
X/ = (x1, x2, x3) = (1, 3, 2) is a feasible solution to the system of
linear equations
2x1 + 4x2 – 2x3 = 10
10x1 + 3x2 + 7x3 = 33.
Reduce the feasible solution to a basic feasible solution.
follows.
 x1 
 2 4  2    10 
A.X = b where, A =   , X =  x 2  and b =  
10 3 7  x   33 
 3
Rank of A is 2 and hence the given equations are linearly
independent. The given solution X/ = (1, 3, 2) is feasible as all
components are non-negative, but it is not basic feasible because
basic feasible solution to a system of two equations cannot have
more than 2 non-zero components.
2  4   2
a1 =   , a2 =   and a3 =  
10   3  7 
43
Setting the Scene
vectors of two components cannot be independent. Therefore, there
exist three scalars s1, s2, s3 which all are not zero, so that
s1a1 + s2a2 + s3a3 = 0 (i)
2  4   2
 s1   + s2   + s3   = 0
10   3  7 
 2s1  4s 2  2s3   0 
   =  
10s1  3s 2  7 s3   0 
2s1  4s 2  2s3  0
 
10s1  3s 2  7 s3  0
s1 s2 s3
     (say)
28  6  20  14 6  40
s s s 1
 1  2  3    (say)
34  34  34 34
 s1 = –1, s2 = 1, s3 =1
So form (i) we get,
–a1 + a2 + a3 = 0 (ii)
s 
i
 xi 
s1 s 2 s 3 1 1 1 1
 p = max ( , , )  p = max ( , , )=
x1 x 2 x 3 1 3 2 2
Hence, we get new feasible solution as follows:
 s s s   1 1 1
X1 =  x1  1 , x2  2 , x3  3  = 1  1 , 3  1 , 2  1 
 p p p  2 2 2 
= (3, 1, 0).
Submatrix formed by taking the column vectors of A
corresponding to the positive variables in the new solution is B1 =
44
 2 4
(a1, a2) =   and |B1| = – 34  0. So, a1, a2 are linearly
10 3 
independent and hence X1 = (3, 1, 0) is the required basic feasible
solution corresponding to the given feasible solution.
Again we can write equation (ii) as follows:
a1 – a2 – a3 = 0 and so s1 = 1 , s2 = –1, s3 = –1.
s 
i
 xi 
s1 s 2 s 3 1 1 1
 p = max ( , , )  p = max ( , , )=1
x1 x 2 x 3 1 3 2
Hence, we get another feasible solution as follows:
 s s s   1 1 1
X2 =  x1  1 , x2  2 , x3  3  = 1  ,3  ,2  
 p p p  1 1 1 
= (0, 4, 3).
Submatrix formed by taking the column vectors of A
corresponding to the positive variables in the new solution is B2 =
 4  2
(a2, a3) =   and |B2| = 34  0. So, a2, a3 are linearly
3 7 
independent and hence X2 = (0, 4, 3) is another required basic
feasible solution corresponding to the given feasible solution.
Therefore, (3, 1, 0) and (0, 4, 3) are required basic feasible
solutions corresponding to the given feasible solution.
Example 1.33
If x1 = 1, x2 = 1, x3 = 1, x4 = 1 is a feasible solution to the system of
linear equations
2x1 + 2x2 – x3 + 4x4 = 7
x1 + 3x2 – 2x3 + 6x4 = 8.
45
Setting the Scene

follows: A.X = b ,
 x1 
 
 2 2 1 4 x  7
where A =   , X =  2  and b =   .
1 3  2 6 x
 3 8
x 
 4
Rank of A is 2 and hence the given equations are linearly
independent. The given solution (1, 1, 1, 1) is feasible as all
components are non-negative, but not basic feasible because basic
feasible solution to a system of two equations cannot have more
than 2 non-zero components.
 2  2  1  4
a1 =   , a2 =   , a3 =   and a4 =  
1  3   2 6
vectors having two components cannot be independent. So, there
exist four scalars s1, s2, s3, s4 all are not zero, such that
s1a1 + s2a2 + s3a3 + s4a4 = 0 (i)
 2  2  1  4
 s1   + s2   + s3   + s4   = 0
1  3   2 6
 2s1  2s 2  s3  4s 4   0 
   =  
 s1  3s 2  2s3  6s 4   0 
2s1  2s 2  s3  4s 4  0
 
s1  3s 2  2s3  6s 4  0
 2s1  2s 2  s3  4s 4  0
 
 2s1  6s 2  4s3  12s 4  0 [ Multiplyin g by  2]
Adding the two equations, we get
46
– 4s2 + 3s3 – 8s4 = 0.

This equation has many solutions and we consider a particular
solution s2 = 0, s3 = 8, s4 =3 and hence s1 = –2.
Branch –1
So form (i), we have
–2a1 + 8a3 + 3a4 = 0 (ii)
s 
i
 xi 
s s s s
 p = Max ( 1 , 2 , 3 , 4 )
x1 x 2 x 3 x 4
2 0 8 3
 p = Max ( , , , )=8
1 1 1 1
Hence, we get a new feasible solution as follows:
 s s s s 
X1 =  x1  1 , x2  2 , x3  3 , x4  4 
 p p p p
 2 0 8 3
= 1  , 1 , 1 , 1 
 8 8 8 8
5 5
= ( , 1, 0, ) which is not a basic solution.
4 8
Now the given system of linear equations becomes as follows:
2x1 + 2x2 + 4x4 = 7
x1 + 3x2 + 6x4 = 8
where x1 = 5/4, x2 = 1, x4 = 5/8 is a feasible solution.
Now we apply the above procedure again. There exist three scalars
r1, r2, r4 (all of them are not zero) such that
r1a1 + r2a2 + r4a4 = 0 (iii)
 2  2  4
 r1   + r2   + r4   = 0
1  3 6
47
Setting the Scene
 2r1  2r2  4r4   0 

   =  
1r  3 r2  6 r4  0
2r1  2r2  4r4  0

 
r1  3r2  6r4  0
r1 r r
  2  4   (say)
12  12 4  12 6  2
r1 r r 1
  2  4     (say)
0 8 4 4
 r1 = 0, r2 = 2, r4 = –1.
So form (iii) we get,
0a1 + 2a2 – a4 = 0 (iv)
r 
Let q = Max i  for xi > 0
i
 xi 
r1 r2 r4
 q = Max ( , , )
x1 x 2 x 4
0 2 1
 q = Max ( 5 , , 5 ) = 2.
4 1 8
 r r r 
(x1, x2, x4) =  x1  1 , x2  2 , x4  4 
 q q q
5 0 2 5 1 5 9
=   , 1 ,   = ( , 0, ).
4 2 2 8 2  4 8
That is, (x1, x2, x3, x4) = (5/4, 0, 0, 9/8) is a new feasible solution to
the given system.
Submatrix formed by taking column vectors of A corresponding to
positive variables of the new feasible solution is B1 = (a1, a4) =
48
 2 4
  and |B1| = 8  0. Thus, a1, a4 are linearly independent and
1 6
5 9
hence the solution (x1, x2, x3, x4) = ( , 0, 0, ) is a required basic
4 8
feasible solution corresponding to the given feasible solution.
Again we can write equation (iv) as follows:
0a1 – 2a2 + a4 = 0 and so r1 = 0 , r2 = –2, r4 = 1.
r 
Again let q = Max i  for xi > 0
i
 xi 
r1 r2 r4
 q = Max ( , , )
x1 x 2 x 4
0 2 1 8
 q = Max ( 5 , , 5 )= .
4 1 8 5
Hence, we get another feasible solution as follows:
 r r r 
(x1, x2, x4) =  x1  1 , x2  2 , x4  4 
 q q q
5 0 2 5 1 5 9
=   8 , 1  8 ,  8  = ( , , 0).
4 5 5 8 5 4 4
That is, (x1, x2, x3, x4) = (5/4, 9/4, 0, 0) is a new feasible solution to
the given system.
 2 2
  and |B2| = 4  0. Thus, a1, a2 are linearly independent and
1 3
hence the solution (x1, x2, x3, x4) = (5/4, 9/4, 0, 0) is another
required basic feasible solution corresponding to the given feasible
solution.
49
Setting the Scene
Branch –2
We can write equation (ii) as follows:
2a1 – 8a3 – 3a4 = 0 (v)
s 
i
 xi 
s1 s 2 s 3 s 4
 p = Max ( , , , )
x1 x 2 x 3 x 4
2 0 8 3
 p = Max ( , , , ) = 2.
1 1 1 1
 s s s s 
X2 =  x1  1 , x 2  2 , x3  3 , x 4  4 
 p p p p
 2 0 8  3
= 1  , 1  , 1  , 1 
 2 2 2 2 
5
= (0, 1, 5, ) which is not a basic solution.
2
Now the given system of linear equations becomes as follows:
2x2 – x3 + 4x4 = 7
3x2 – 2x3 + 6x4 = 8
where x2 = 1, x3 = 5, x4 = 5/2 is a feasible solution.
Now we apply the above procedure again. There exist three scalars
r2, r3, r4 (all of them are not zero) such that
r2a2 + r3a3 + r4a4 = 0 (vi)
 2  1  4
 r2   + r3   + r4   = 0
 3   2 6
 2r2  r3  4r4   0 
   =  
 3r2  2r3  6r4   0 
50
2r2  r3  4r4  0
 
3r2  3r3  6r4  0
r2 r3 r4
     (say)
 6  12 12  12  6  3
r2 r3 r 1
   4     (say)
6 0 3 3
 r2 = – 2, r3 = 0, r4 = 1.
So form (vi), we have
–2a2 + 0a3 + a4 = 0 (vii)
r 
Let q = Max i  for xi > 0
i
 xi 
r2 r3 r4
 q = Max ( , , )
x 2 x3 x 4
2 0 1 2
 q = Max ( , , 5 )=
1 5 2 5
Hence we get new feasible solution as follows:
 r r r 
(x2, x3, x4) =  x2  2 , x3  3 , x4  4 
 q q q
 2 0 5 1
= 1  2 , 5  2 ,  2  = (6, 5, 0).
 5 5 2 5
That is, (x1, x2, x3, x4) = (0, 6, 5, 0) is a new feasible solution of the
given system.
2 1
  and |B3| = –1  0. Thus, a2, a3 are linearly independent
 3  2
and hence the solution (x1, x2, x3, x4) = (0, 6, 5, 0) is a required
basic feasible solution corresponding to the given feasible solution.
51
Setting the Scene
Multiplying equation (vii) by –1, we have the new form of

equation (vii) as follows:
2a2 – 0a3 – a4 = 0 and so r2 = 2 , r3 = 0, r4 = –1.
r 
Again let q = Max i  for xi > 0
i
 xi 
r2 r3 r4
 q = Max ( , , )
x 2 x3 x 4
2 0 1
 q = Max ( , , 5 ) = 2.
1 5 2
 r r r 
(x2, x3, x4) =  x2  2 , x3  3 , x4  4 
 q q q
 2 0 5 1
= 1  , 5  ,   = (0, 5, 3).
 2 2 2 2 
That is, (x1, x2, x3, x4) = (0, 0, 5, 3) is a new feasible solution to the
given system.
positive variables in the new feasible solution is B4 = (a3, a4) =
 1 4
  and |B4| = 2  0. Thus, a3, a4 are linearly independent
  2 6
and hence the solution (x1, x2, x3, x4) = (0, 0, 5, 3) is a required
basic feasible solution corresponding to the given feasible solution.
5 9 5 9
We can conclude that (x1, x2, x3, x4) = ( , 0, 0, ), ( , , 0, 0),
4 8 4 4
(0, 6, 5, 0) and (0, 0, 5, 3) are basic feasible solutions to the given
system corresponding to the given feasible solution.
Example 1.34
Find the optimum solution to the linear programming (LP)
problem
52
Maximize z = 2x1 + 4x2 – x3 + 2x4

Subject to 2x1 + 6x2 + 2x3 + x4 = 3
6x1 + 4x2 + 4x3 + 6x4 = 2
x1, x2, x3, x4  0.
Solution: We can write the given constraint equations in the
matrix form as follows:
 x1 
 
 2 6 2 1 x   3
AX = b, where A =   , X =  2  and b =   .
 6 4 4 6 x
 3  2
x 
 4
The column vectors of the coefficient matrix of A are
 2 6  2 1
a1 =   , a2 =   , a3 =   and a4 =   .
6  4  4 6
4
c 2 = 6 square submatrices taking two at a time from a1, a2, a3, a4.
They are
 2 6  2 2
B1 = (a1, a2) =   , B2 = (a1, a3) =   ,
 6 4  6 4
 2 1  6 2
B3 = (a1, a4) =   , B4 = (a2, a3) =   ,
 6 6  4 4
6 1  2 1
B5 = (a2, a4) =   , B6 = (a3, a4) =   .
 4 6  4 6
Values of determinants formed by the above submatrices are
|B1| = –28, |B2| = – 4, |B3| = 6, |B4| = 16, |B5| = 32, |B6| = 8.
Since all sub matrices are non-singular, we shall get six basic
solutions corresponding to six basis matrices B1, B2, B3, B4, B5 and
B6 .
53
Setting the Scene
 2 6
(i) For basis matrix B1=(a1, a2)=   , basic variables are x1, x2
 6 4
and non-basic variables are x3, x4. Putting x3 = x4 = 0, the constraint
equations reduce to
2x1 + 6x2 = 3
6x1 + 4x2 = 2.
Solving the system of linear equations, we have x1 = 0, x2 = ½.
Hence, a basic solution is X1 = (0, ½, 0, 0) which is also feasible.
 2 2
(ii) For basis matrix B2 = (a1, a3) =   , basic variables are x1,
 6 4
x3 and non-basic variables are x2, x4. Putting x2 = x4 = 0, the
constraint equations reduce to
2x1 + 2x3 = 3
6x1 + 4x3 = 2.
Solving this system, we get x1 = –2, x3 = 7/2. Hence, a second
basic solution is X2 = (–2, 0, 7/2, 0) which is non-feasible.
 2 1
(iii) For basis matrix B3 = (a1, a4)=   , basic variables are x1,
 6 6
2x1 + x4 = 3
6x1 + 6x4 = 2.
Solving this system, we get x1 = 8/3, x4 =  7/3. Hence, a third
basic solution is X3 = (8/3, 0, 0, –7/3) which is non-feasible.
 6 2
(iv) For basis matrix B4 = (a2, a3) =   , basic variables are x2,
 4 4
6x2 + 2x3 = 3
4x2 + 4x3 = 2.
54
Solving this system, we have x2 = ½, x3 = 0. Hence, a fourth basic

solution is X4 = (0, ½, 0, 0) which is also feasible.
6 1
 4 6
x4 and non-basic variables are x1, x3. Considering non-basic
variables x1 = x3 = 0, the constraint equations reduce to
6x2 + x4 = 3
4x2 + 6x4 = 2.
Solving this system, we get x2 = ½, x4 = 0. Hence, a fifth basic
solution is X5 = (0, ½, 0, 0) which is also feasible.
 2 1
(vi) For basis matrix B6 = (a3, a4) =   , basic variables are x3,
 4 6
2x3 + x4 = 3
4x3 + 6x4 = 2.
Solving this system, we have x3 = 2, x4 = –1. Hence, a sixth basic
solution is X6 = (0, 0, 2, –1) which is not feasible.
Therefore, the given LP problem has only one basic feasible
solution, which is (0, ½, 0, 0) and the other solutions are not basic
feasible because they contains negative numbers. For
maximization type LP problem, a basic feasible solution which
maximizes the objective function, is an optimum solution. There is
only one basic feasible solution to this LP problem, so it is an
optimum solution. Thus, the required optimum solution is x1 = 0,
x2 = ½ , x3 = 0, x4 = 0 and zmax = 2.
Note: To solve this LP problem using Mathematica 5 software, we
use the code given below.
ConstrainedMax[2x1 + 4x2 – x3 + 2x4, {2x1 + 6x2 + 2x3 + x4 = = 3,
6x1 + 4x2 + 4x3 + 6x4 = = 2}, {x1, x2, x3, x4}]
55
Setting the Scene
Example 1.35
Reduce the LP problem into canonical form:
Maximize z = 2x1 – 3x2 + x3
x1 – x3  3
x1 + 2x2 = 1
x3  0, x1, x2 are unrestricted in sign.
Solution: Expressing 3rd constraint by using the symbols of "  "
and "  ", we have the given problem as follows:
x1 – x3  3
x1 + 2x2  1
x1 + 2x2  1
Multiplying 2nd and 3rd constraints by –1, we get
–x1 + x3  –3
–x1 – 2x2  –1
x1 + 2x2  1
Considering x1 = x1 – x1// , x2 = x 2/ – x 2// ; x1/  0, x1//  0, x 2/  0, x 2// 
/
0, we have
Maximize z = 2 x1/ –2 x1// – 3 x 2/ –3 x 2// + x3
Subject to 2 x1/ – 2 x1// + 3 x 2/ – 3 x 2//  5
– x1 + x1
/ //
+ x3  –3
– x1 + x1 – 2 x 2 + 2 x 2//
/ // /
 –1
x1 – x1 + 2 x 2 – 2 x 2
/ // / //
 1
x1 , x1 , x 2 , x 2 , x3  0.
/ // / //
This is the canonical form of the given LP problem.
56
1.11 Exercises
1.1 What do you mean by optimization?
1.2 Define with example objective function, constraints and non-
negativity condition of a linear programming problem.
1.3 Discuss the characteristics of the standard form and canonical
form of a linear programming problem.
1.4 What are the advantages and limitations of linear programming?
1.5 Define with examples feasible solution, basic solution, basic
feasible solution, non-degenerate basic feasible solution and
optimum solution of a linear programming problem.
1.6 Define slack and surplus variables with examples.
1.7 Find the numerical values of
5 n
(i) a
a 1
[Ans: 14] (ii)  a if
a 1
n is 6 [Ans: 20]
5 10
(iii)  3i
i 1
[Ans: 45] (iv) p
p 5
2
[Ans: 355]
1.8 Write the expanded form of

5
(i) x
a 1
a [Ans: x1 + x2 + x3 + x4 + x5]
n
(ii) a xi 1
i i  b [Ans: a1x1 + a2x2 + . . . +anxn = b]
1.9 Express using summation symbol:

6
(i) y1 – y2 + y3 – y4 + y5 – y6 [Ans:  ( )
i 1
i 1
yi ]
n
(ii) a1x1 + a2x2 + . . . + anxn = c [Ans: a x
i 1
i i c]
1.10 Convert the following LP problems into standard form:

i) Minimize z = 3x1 + 11x2
–3x1 + x2  3
57
Setting the Scene
x1, x2  0
ii) Maximize z = 3x1 + x2
–3x1 + 2x2  3
iii) Minimize z = 9x1 + 6x2
Subject to 5x1 + 3x2 + x3 = 10
–3x1 + x2 + 2x3  3
x1, x2, x3  0
iv) Maximize z = 2x1 + x2
–3x1 + x2  3
v) Maximize z = 3x1 + 6x2 + 2x3
–3x1 + x2 + 2x3  3
x1, x2, x3  0
vi) Maximize z = 3x1 + x2 + 3x3
Subject to 8x1 + 2x2 – x3  4
–3x1 + 2x2 + 2x3  3
x1, x2  0, x3 is unrestricted in sign.
vii) Maximize z = 3x1 + x2 + 3x3
–3x1 + 2x2 + 2x3  3
x1 + 5x2 – 3x3  7
viii) Maximize z = 2x1 + 6x2 + 5x3
–3x1 + x2 + 2x3  8
x1, x2, x3  0
ix) Minimize z = x1 + 7x2 + 5x3
–3x1 + 8x2 + 2x3  –3
58
x1, x2, x3  0
x) Minimize z = 10x1 + 11x2
–3x1 + x2  3
x1  0, x2  0
xi) Minimize z = 10x1 + 11x2
–3x1 + x2  3
xii) Maximize z = 3x1 + x2
–3x1 + 2x2  3
xiii) Maximize z = 3x1 + x2 + 3x3
–3x1 + 2x2 + 2x3  3
x1 + 5x2 – 3x3  7
x1  0, x2  0, x3 is unrestricted in sign.
xiv) Maximize z = 2x1 + 8x2 + 3x3
–3x1 + x2 + 2x3  3
3x1 + 5x2 – 3x3  6
x1  –2, x2  0, x3 is unrestricted in sign.
xv) Minimize z = x1 + 9x2 + 3x3
Subject to 7x1 + 2x2 – 9x3  38
–4x1 + 3x2 + 2x3  –4
5x1 + 5x2 – 3x3  6
xvi) Maximize z = 5x1 + 2x2 + 3x3 + x4
Subject to 5x1 + 4x2 – 9x3 + 4x4  10
–2x1 + 3x2 + 9x3 + 2x4  4
x1 + 5x2 – 2x3 + 3x4  56
x1, x2,x3, x4  0.
59
Setting the Scene
xvii) Maximize z = 7x1 + 2x2 + 3x3 + 2x4

Subject to 2x1 + 3x2 – 9x3 + 4x4 =10
2x1 + 3x2 – 9x3 + 2x4  4
3x1 + 5x2 – 2x3 + 2x4  25
x1, x2  0, x3  0, x4 unrestricted
xviii) Maximize z = 3x1 + 9x2 + 3x3 +3x4
Subject to 6x1 + 3x2 – 5x3 + 4x4 =12
3x1 + 8x2 – 9x3 + 2x4  40
3x1 + 5x2 – 2x3 + x4  19
xix) Minimize z = x1 + 9x2 + 3x3 + x4
Subject to 7x1 + 2x2 – 9x3 + 4x4  18
–4x1 + 3x2 + 2x3 + 2x4  –14
5x1 + 5x2 – 3x3 + 3x4  6
x1  0, x2  –2,x3  1, x4 is unrestricted in sign.
1.11 Convert the following linear programming problem into the
standard form:
9x1 – x2 + 6x3  5
x1 + 3x2 + x3  1
[Ans.: Maximize 8 x1 – 4 x 2/ + 3 x3/ –3 x3// + 16
/
Subject to – 6 x1/ + 2 x2/ + 5 x3/ – 5 x3// – s1 = 9

– 9 x1/ – x 2/ – 6 x3/ + 6 x3// + s2 = 12
– x1/ + 3 x2/ – x3/ + x3// – s3 + s4 = 1
x1/ , x 2/ , x3/ , x3// , s1, s2, s3, s4  0.]
1.12 Convert the following LP problem into canonical form:
a. Maximize z = 2x1 + 5x2
–3x1 + 8x2  3
60
b. Minimize z = 3x1 + 4x2
–3x1 + 5x2 + 2x3  3
x1, x2, x3  0
c. Maximize z = 2x1 + 5x2
Subject to 3x1 + 3x2  –5
–3x1 + x2  3
d. Maximize z = 2x1 + 7x2 + 2x3
–3x1 + x2 + 2x3  3
x1, x2, x3  0
e. Maximize z = 3x1 + 7x2 + 3x3
–3x1 + 4x2 + 2x3  3
f. Maximize z = 13x1 + 10x2 + 3x3
Subject to 10x1 + 2x2 – x3  44
–3x1 + 2x2 + 8x3  3
9x1 + 5x2 – 3x3  7
g. Maximize z = 2x1 + 6x2 + 5x3
–3x1 + x2 + 2x3  8
x1, x2, x3  0
h. Minimize z = x1 + 7x2 + 5x3
–3x1 + 8x2 + 2x3  –3
x1, x2, x3  0
i. Minimize z = 10x1 + 11x2
–3x1 + x2  3
61
Setting the Scene
x1  0, x2  0
j. Minimize z = 10x1 + 11x2
–3x1 + x2  3
k. Maximize z = 3x1 + 7x2
–3x1 + 5x2  3
x1, x2 are unrestricted in sign.
l. Maximize z = 3x1 + x2 + 3x3
–3x1 + 2x2 + 2x3  3
x1 + 5x2 – 3x3  7
x1  0, x2  0, x3 is unrestricted in sign.
m. Maximize z = 7x1 + 8x2 + 13x3
Subject to 2x1 + 2x2 – 9x3  17
–5x1 + 3x2 + 2x3  3
3x1 + 5x2 – 3x3  6
n. Minimize z = x1 + 9x2 + 3x3
Subject to 7x1 + 2x2 – 9x3  38
–4x1 + 3x2 + 2x3  –4
5x1 + 5x2 – 3x3  6
o. Maximize z = 5x1 + 2x2 + 3x3 + x4
Subject to 5x1 + 4x2 – 9x3 + 4x4  10
–2x1 + 3x2 + 9x3 + 2x4  4
x1 + 5x2 – 2x3 + 3x4  56
x1, x2,x3, x4  0.
p. Maximize z = 7x1 + 2x2 + 3x3 + 2x4
Subject to 2x1 + 3x2 – 9x3 + 4x4 =13
2x1 + 3x2 – 9x3 + 2x4  4
3x1 + 5x2 – 2x3 + 2x4  25
62
q. Maximize z = 3x1 + 9x2 + 3x3 +3x4
Subject to 6x1 + 3x2 – 5x3 + 4x4 =12
3x1 + 8x2 – 9x3 + 2x4  40
3x1 + 5x2 – 2x3 + x4  19
x1  0, x3  3, x2 and x4 unrestricted
r. Minimize z = 7x1 + 9x2 + 3x3 + 7x4
Subject to 3x1 + 2x2 – 9x3 + 4x4  18
–2x1 + 3x2 + 4x3 + 2x4  –14
5x1 + 6x2 – 3x3 + 3x4  45
x1  0, x2  –2, x3  1, x4 is unrestricted.
s. Minimize z = 3x1 + 7x2 + 5x3 + 3x4 – 3x5

Subject to 4x1 + 3x2 + 4x3 + x4 – 3x5  24
3x1 + 3x2 + 2x3 + 2x4 – x5 = 22
2x1 + 8x2 + x3 + 3x4 – 3x5  –14
–3x1 + 5x2 + 2x3 + 2x4 – 7x5  –13
x1, x2  0, x3  3, x4  0, x5 unrestricted
1.13 Find all basic solutions to the system of linear equations:
a. x1 + 2x2 + x3 = 4 b. x1 + 2x2 – x3 = 4
2x1 + x2 + 5x3 = 5 2x1 + 2x2 + x3 = 4
c. 2x1 + 4x2 – x3 = 3 d. 3x1 + 4x2 – x3 = 6
2x1 + 4x2 + x3 = 6 2x1 + 4x2 + x3 = 4
e. 2x1 + 3x2 – x3 = 3 f. 2x1 – x2 + 3x3 + x4 = 6
4x1 + 4x2 + 2x3 = 12 4x1 – 2x2 – x3 + 2x4 = 10
g. x1 + 3x2 + 2x3 + 3x4 = 10 h. 2x1 + 3x2 + 6x3 + 4x4 = 1
2x1 – x2 + 4x3 + 6x4 = 16 3x1 – x2 + 9x3 + 6x4 = 1
i. 3x1 + 3x2 + 9x3 + 4x4 = 8 j. x1 + 3x2 + 4x3 – 4x4 = 4
3x1 – 2x2 + 9x3 + 6x4 = 4 x1 – 2x2 + 9x3 + 8x4 = 9
1.14 Find all basic feasible solutions of the system of linear
equations:
63
Setting the Scene
a. x1 + 2x2 + x3 = 4 b. 2x1 + 2x2 – x3 = – 4

2x1 + x2 + 5x3 = 5 2x1 + 2x2 + x3 = 4
c. 3x1 + 2x2 – x3 = 6 d. 3x1 + 4x2 – x3 = 8
x1 + 4x2 + x3 = 3 3x1 + 2x2 + x3 = 4
e. 2x1 + 4x2 + x3 = 3 f. 2x1 – x2 + 3x3 + x4 = 6
4x1 + 8x2 + 2x3 = 10 4x1 – 2x2 – x3 + 2x4 = 10
g. x1 + 3x2 + 2x3 + 3x4 = 10 h. 3x1 + 3x2 + 6x3 + 4x4 = 3
2x1 – x2 + 4x3 + 6x4 = 16 3x1 – 2x2 + 9x3 + 4x4 = 3
i. 2x1 + 4x2 + 7x3 + 4x4 = 8 j. 2x1 + 4x2 + 7x3 + 4x4 = 8
3x1 – 2x2 + 7x3 + 6x4 = 4 3x1 – 2x2 + 7x3 + 6x4 = 4
1.15 How many basic solutions are there in the following linearly
independent set of equations? Find all of them.
2x1 – x2 + 3x3 + x4 = 6
4x1 – 2x2 – x3 + 2x4 = 10
1.16 Show that x1 = 5, x2 = 0, x3 = – 1 is a basic solution to the
system of linear equations.
x1 + 2x2 + x3 = 4
2x1 + x2 + 5x3 = 5
Find the other basic solutions, if there is any.
1.17 Show that x1 = 2, x2 = –1, x3 = 0 is not a basic solution to the
3x1 – 2x2 + x3 = 8
9x1 – 6x2 + 4x3 = 24
Find all basic solutions, if there is any.
1.18 Show that x1 = 1, x2 = 1, x3 = 2 is a feasible solution of the
x1 + 2x2 + 3x3 = 9
2x1 – x2 + x3 = 3
Reduce the above feasible solution to a basic feasible
solution.
1.19 If x1 = 1, x2 = 1, x3 = 3 be a feasible solution to the system of
linear equations.
64
x1 + 2x2 + 3x3 = 12
2x1 – x2 + x3 = 4
1.20 If x1 = 1, x2 = 1, x3 = 1 be a feasible solution of the system of
linear equations.
3x1 + 2x2 + 3x3 = 8
2x1 – 3x2 + 3x3 = 2
1.21 If x1 = 2, x2 = 1, x3 = 1 be a feasible solution to the following

3x1 + 2x2 + x3 = 9
2x1 – 3x2 + 2x3 = 3
1.22 Show that x1 = 2, x2 = 1, x3 = 2 is a feasible solution to the
x1 + 2x2 + 3x3 = 10
2x1 – x2 + x3 = 5
1.23 Show that x1 = 1, x2 = 2, x3 = 1, x4 = 0 is a feasible solution to
the system of linear equations.
11x1 + 2x2 – 9x3 + 4x4 = 6
15x1 + 3x2 – 12x3 + 6x4 = 9
Reduce the feasible solution to basic feasible solutions and
also show that one of them is non-degenerate and the other is
degenerate.
1.24 If x1 = 1, x2 = 2, x3 = 1, x4 = 1 is a feasible solution to the
11x1 + 2x2 – 9x3 + 4x4 = 10
15x1 + 3x2 – 12x3 + 6x4 = 15
1.25 Find the optimum solution to the LP problem:
65
Setting the Scene

5x1 + x2 = 6
x1, x2  0.
1.26 Find the optimum solution to the following LP problem:
Minimize z = x1 + 2x2 + 3x3
Subject to x1 + 2x2 + x3 = 3
2x1 + x2 + 5x3 = 6
x1, x2, x3  0.
1.27 Find the optimum solution to the following LP problem:
2x1 + x2 + 5x3 = 3
x1, x2, x3  0.
1.28 Find the optimum solution to the LP problem

2x1 + 4x2 + x4 = 8
x1, x2, x3, x4  0.
1.29 Find the optimum solution to the linear programming (LP)
problem
Maximize z = 3x1 + 2x2 + x3
2x1 + x2 + 5x3 = 5
x1, x2, x3  0.
1.30 Find the optimum solution to the LP problem:
Minimize z = x1 – x2 + 2x3 + 3x4
Subject to 2x1 + x2 + 3x3 + 2x4= 11
3x1 – 3x2 + 5x3 + x4 = 17
x1, x2, x3, x4  0.
66
Appendix 1
Classical Optimization
10.1 Maxima and minima

A function f(x) is said to be maximum (minimum) at x = a if f(a) is
greater (less) than every other values assumed by f(x) in the
immediate neighborhood of x = a. These types of maxima/ minima
are local maxima/ minima. The maxima which is greater than all
other maxima is called global maxima, and the smallest minima is
the global minima.
Y
Maximum point y = f(x)
Minimum point
O X
Figure 10.1
dy
The gradient of a curve y = f(x) at all turning points is 0 (zero),
dx
dy
i.e., = 0 at all up peaks and down peaks. At the point of
dx
dy
maxima, the sign of changes from positive to negative when x
dx
dy
increases and at the minima, the sign of changes from negative
dx
to positive when x increases.
Appendices
10.2 Determination of optima of a function of one variable

If x = c be a point in the interval in which the function f x  is
defined, and if f / c   0 and f // c   0 , then f c  is maximum
or minimum accordingly f // c  is negative or positive. Otherwise,
if f / c   f // c   . . .  f n1 c   0 and f n c   0 , then
(i) when n is even, f c  is a maximum or a minimum according as
f n c  is negative or positive,
(ii) when n is odd, f c  is neither a maximum nor a minimum.
Example 10.1
Find for what values of x, the following expression is maximum
and minimum respectively: 15x4 + 8x3 – 18x2.
Find also the maximum and minimum values of the expression.
Solution: Let us consider,
f x   15x 4  8x 3  18x 2 .
 f / x  
d
(15 x 4  8 x 3  18 x 2 ) = 60x3 + 24x2 – 36x.
dx
And f // x  =
d
(60x3 + 24x2 – 36x) = 180x2 + 48x – 36.
dx
At the maximum or minimum, f / x  = 0. So, we have
60x3 + 24x2 – 36x = 0,
x(x +1)(5x – 3) = 0.
3
Implies, x = 0, x = –1 and x = .
5
When x = 0, f // 0  36, which is negative,
when x = –1, f // (1)  96, which is positive,
3  3  288
and when x = f //    , which is positive.
5, 5 5
672
Hence, the given expression is maximum at x = 0 and minimum at

3
x = –1 and x = .
5
3
Now, we have, f(0) = 0, f(–1) = –11 and f   = – 2.808. Thus,
5
the maximum value of the expression is 0 and the minimum values
are –11 and –2.808.
Example 10.2
Find maximum and minimum values of f(x) = x3 – 4x2 + 4x – 10.
Solution: Given that, f(x) = x3 – 4x2 + 4x – 10.
d 3
So, f /(x) = (x – 4x2 + 4x – 10) = 3x2 – 8x + 4,
dx
d
f //(x) = (3x2 – 8x + 4) = 6x – 8.
dx
When f x  is a maximum or a minimum, f / x  = 0, thus
3x2 – 8x + 4 = 0,
2
x = 2 and x = .
3
Now, when x = 2, f 2  4, which is positive.
//
2 //  2 
And when x = f    4, which is negative.
3 3
2
Hence, the given function is maximum at x = and minimum at x
3
= 2.
2 238
Here, f (2) = – 10 and f   = – = – 8.815.
3 27
Thus, the maximum value of the function is – 8.815 and the
minimum values is –10.
673
Appendices
10.3 Determination of optima of a multivariate function

For the function of two variables, if (a, b) be a point in the domain
in which the function f(x, y) is defined, and if
fx(a, b) = 0, fy(a, b) = 0 and
fxx(a, b).fyy(a, b) – [fxy(a, b)]2 > 0,
then f(a, b) is a maximum or a minimum according as
fxx(a, b) < or > 0 (and consequently fyy(a, b) < or > 0).
But if
fxx(a, b).fyy(a, b) – [fxy(a, b)]2 < 0,
then f(a, b) is neither a maximum nor a minimum.
And if
fxx(a, b).fyy(a, b) – [fxy(a, b)]2 = 0,
then further analysis is necessary.
In other word, we can say that f(a, b) is a minimum (or a
maximum) if fx(a, b) = 0, fy(a, b) = 0 and if the Hessian,
 2 f 2 f 
 2 
 x xy 
Hf =
 2 f 2 f 
 2 
 yx y  ( x , y ) ( a ,b )
is positive definite (negative definite). This is also applicable for
the functions of three or more variables.
Note: If all the principal minors are non-negative then the Hessian,
Hf is called positive definite and if all the principal minors have
opposite signs leading negative then the Hessian, Hf is called
negative definite.
Example 10.3
Examine for extreme values of the function:
f(x, y) = x2 + y2 + (x + y +1)2.
Solution: Given that f(x, y) = x2 + y2 + (x + y +1)2.
So, fx = 2x + 2(x + y +1) = 4x + 2y + 2,
fy = 2y + 2(x + y +1) = 2x + 4y + 2,
674
fxx = 4, fyy = 4 and fxy = 2.

The equations fx = 0 and fy = 0 are equivalent to
2x + y + 1 = 0,
x + 2y + 1 = 0.
By solving the system of linear equations above, we have
1 1
x =  and y =  .
3 3
1 1
The function may have an extreme value at (  ,  ).
3 3
2 2
Now, fxx.fyy – (fxy) = 4.4 – 2 = 12 > 0, also fxx = 4 > 0.
1 1
Therefore, f(x, y) is minimum at (  ,  ) and the minimum
3 3
1 1 1
value is f(  ,  ) = .
3 3 3
10.4 Constrained optimization with Lagrange multipliers

Differential calculus is also used to maximize or minimize a
function subject to constraints. Consider a function f(x, y) subject
to a constraint h(x, y) = k. We can form a Lagrange function F
considering Lagrange multiplier λ with the given function and
constraint as follows:
F(x, y, λ) = f(x, y) – λ[h(x, y) – k].
Since the constraint is always set equal to 0, the value of the new
function F(x, y, λ) is same as the original objective function f(x, y).
Now solving simultaneously
Fx(x, y, λ) = 0, Fy(x, y, λ) = 0 and Fλ(x, y, λ) = 0,
we shall get some critical points. At these critical points the
function will be maximum or minimum according the maximum or
minimum criterions of multivariate function.
Example 10.4
Determine whether the objective function
675
Appendices
f(x, y) = 4x – x2 + 2y subject to the constraint x + y = 1

has an extreme value or not.
Solution: We form the Lagrange function as follows.
F(x, y, λ) = 4x – x2 + 2y – λ(x + y – 1).
So, by taking derivatives, we have
Fx = 4 – 2x – λ, Fy = 2 – λ, Fλ = – (x + y – 1),
and Fxx = –2, Fxy = 0, Fxλ = –1, Fyx = 0, Fyy = 0, Fyλ = –1,
Fλx = –1, Fλy = –1, Fλλ = 0.
The equations Fx = 0, Fy = 0 and Fλ = 0 are equivalent to
4 – 2x – λ = 0, 2 – λ = 0, x + y – 1 = 0.
Solving this system, we have the candidate point (1, 0, 2) for
maximum or minimum. So, (1, 0, 2) as well as (1, 0) are critical
points of F(x, y, λ) and f(x, y) respectively. Now, at the candidate
point, Fxx = –2, Fxy = 0, Fxλ = –1, Fyx = 0, Fyy = 0, Fyλ = –1, Fλx = –
1, Fλy = –1, Fλλ = 0. So, the determinant formed by the Hessian
matrix is
 2 0 1
0 0 1 = 2 > 0
1 1 0
and leading element of principal diagonal Fxx = –2 < 0. Hence, the
candidate point (1, 0) maximizes the objective function and the
maximum value is f (1, 0) = 3.
Example 10.5
Production function of a firm is f(x, y) = 20x + 10y – 2x2, where x
means the number of labours and y means the unit of raw material.
If the cost of per unit labour is tk. 4, the cost of per unit raw
material is tk. 5 and the total spent is tk. 24, find the maximum
production of the firm.
Solution: Here,
Objective function: f(x, y) = 20x + 10y – 2x2,
Constraint: 4x + 5y = 24.
676
We construct the Lagrange function as follows:

F(x, y, λ) = 20x + 10y – 2x2 – λ(4x + 5y – 24).
Taking first order derivatives, we have
Fx = 20 – 4x – 4λ,
Fy = 10 – 5λ,
Fλ = – (4x + 5y – 24).
For optimum values, Fx = 0, Fy = 0 and Fλ = 0, and they are
equivalent to
20 – 4x – 4λ = 0,
10 – 5λ = 0,
4x + 5y – 24 = 0.
After solving the system, we have x = 3, y = 2.4, λ = 2.
Usually, we construct Hessian matrix as follows:
 Fxx Fxy Fx 
 
HF =  Fyx Fyy Fy  ,
F 
 x Fy F 
where, Fxx = – 4, Fxy = 0, Fxλ = – 4, Fyx = 0, Fyy = 0, Fyλ = – 5, Fλx =
– 4, Fλy = – 5 and Fλλ = 0.
So, the determinant formed by the Hessian is
4 0 4
0 0  5 = 100 > 0,
4 5 0
and the leading element of principal diagonal, Fxx = – 4 is negative.
Therefore, the production will be maximum when x = 3, y = 2.4
and the value of the maximum production = 20(3) + 10(2.4) – 2(3)2
= 66 units.
677
Appendices
Example 10.6
5 2
Total cost of producing x units is x + 175x + 125 and the selling
4
5
price per unit is 250 – x . What should be the output for a
2
maximum profit? And calculate that maximum profit.
Solution: Given that,
5
Cost function: c(x) = x 2 + 175x + 125,
4
5 5
Revenue function: r(x) = (250 – x )  x = 250x – x 2 ,
2 2
5 2 5 2
Profit function: p(x) = 250x – x – ( x + 175x + 125)
2 4
15 2
= – 125 + 75x – x
4
d d 15 2
Now, [p(x)] = (–125 + 75x – x )
dx dx 4
30
= 75 – x
4
The necessary and sufficient conditions for maximization are that
the first derivative of a profit function is equal to zero and the
second derivative is negative. So,
30
75 – x = 0 implies x = 10.
4
d2 30
And 2
[p(x)] = – , which is negative.
dx 4
Therefore, the profit is maximum at the output, x = 10 units.
Putting x = 10 in the profit function, we have
15
Maximum profit = – 125 + 75  10 –  (10) 2 = 250.
4
678
Appendix 2
Model Exam Paper
University Question 1
Subject: Mathematical Programming (Code 3778)
Time: 4 hours Total Marks: 100
Group A
(Answer any 10 questions.)
1. (a) Define convex set. [Sec. 2.2]
(b) Define hyperplane. [Sec. 2.9]
(c) What is the meaning of half space? [Sec. 2.10],
(d) Define convex hull with an example. [Sec. 2.6],
(e) What is the meaning of convex polyhedron? [Sec. 2.7]
(f) Define feasible solution. [Sec. 1.9]
(g) When do we introduce slack variable in a linear
programming problem? [Sec. 1.9]
(h) Define surplus variable. [Sec. 1.9]
(i) What is sensitivity analysis? [Sec. 6.1]
(j) State the canonical form of a LP problem. [Sec. 1.6]
(k) Define boundary of a convex set. [Sec. 2.15]
(l) Distinguish assignment problem from transportation
problem. [Sec. 9.2]
Group B
2. State and prove complementary slackness theorem in linear
programming. [Thm 5.7]
3. Show that the set of all optimal solution to a linear
programming problem is a convex set. [Thm 2.12]
679
Appendices
4. If the set of constraints T of a linear programming (LP)

problem is non-empty, closed and bounded then show that an
optimal solution to the LP problem exists and it attains at a
vertex of T. [Thm 2.10]
5. Show that the intersection of a family of convex sets is again a
convex set. [Thm 2.4]
6. Describe the role of a Pivot element in simplex tableau. [Sec.
4.5.5]
7. State and prove the basic duality theorem in linear
programming. [Thm 5.6]
8. Describe the simplex method for solving a linear programming
problem. [Sec. 4.2, Sec. 4.7]
9. State and prove mini-max theorem of the linear programming
problem. [Thm 4.4]
Group C
10. Show that S = {(x1, x2, x3): 2x1 + x2 – x3  1, x1 – 2x2 + x3  4}
is a convex set. [Ex. 2.22]
11. Solve graphically the following linear program:
Minimize z = 3x1 + 2x2 [Ex. 3.21]
Subject to x1 + 2x2  4
2x1 + x2  4
x1 + x2  5
x1, x2  0.
12. Solve the following linear programming problem using the
simplex method:
Maximize z = x1 + 4x2 + 5x3 [Ex. 4.60]
x1 + 2x2 + 3x3  14
3x1 + 2x2  14
x1, x2, x3  0.
680
13. Vitamin A and B are found in two different Foods F1 and F2.
One unit of Food F1 contains 2 units of vitamin A and 5 units
of vitamin B. One unit of Food F2 contains 4 units of vitamin A
and 2 units of vitamin B. One unit of Food F1 and F2 costs tk.5
and tk.3 respectively. The minimum daily requirements (for a
man) of vitamin A and B are 40 and 50 units respectively.
Consuming of excess vitamin is not harmful for health. Find
out the optimum mixture of Food F1 and F2 at minimum cost
which meets the daily minimum requirement of vitamin A and
B. Formulate this as linear programming problem and solve it
graphically. [Ex. 3.33]
14. Reduce the given LP problem into canonical forms:
Maximize z = 2x1 – 3x2 + x3 [Ex. 1.37]
x1 – x3  3
x1 + 2x2 = 1
x3  0, x1 and x2 are unrestricted in sign.
15. Construct the dual of the LP problem:
Minimize z = x1 + 2x2 + 3x3 [Ex. 5.18]
Subject to x1 – x2 + 2x3  4
x1 + x2 + 2x3  8
x2 – x3  2
x1, x2, x3  0.
16. By using Vogel’s approximation method, find the initial basic
feasible solution to the transportation problem represented by
the cost table below. [Ex. 8.1]
Dj Supply
1 2 3 4
Oi si
1 11 13 17 14 250
2 16 18 14 10 300
3 21 24 13 10 400
Demand
200 225 275 250
dj
681
Appendices
17. Solve the assignment problem represented by the following

matrix where each entity represents assigning cost of a person
to a job. [Ex. 9.1]
Jobs
1 2 3 4 5 6
1  9 22 58 11 19 27 
 
2  43 78 72 50 63 48 
3  41 28 91 37 45 33 
Persons  
4  74 42 27 49 39 32 
 
5  36 11 57 22 25 18 
6  3 56 53 31 17 28 
682
Appendix 3
Bibliography
Bhatti MA, “Practical Optimization Methods with Mathematica

Applications”, Springer-Verlag, New York, USA, 2000.
Charnes A and Cooper WW, “The Stepping Stone Method of
Explaining Linear Programming Calculations in Transportation
Problems”, Management Science, Vol. 1, No. 1, 1954.
Charnes A, Cooper WW and Ferguson RO, “Optimal estimation of
compensation by linear programming”, Management Science,
Vol. 1, pp 138-151, 1955.
Gupta PK and Hira DS, “Problems in Operations Research”, S
Chand & Company Ltd, New Dilhi 110055, India, 2006-2007.
Hillier FS and Lieberman GJ, “Introduction to Operations
Research”, McGraw-Hill, NY 10020, USA, 2001.
Houthakker HS, “On the Numerical Solution of Transportation
Problem”, Operations Research, Vol. 3, No. 2, 1955.
Islam SMS, “Business Mathematics”, Abir Publications, Dhaka,
Bangladesh, 2004.
Islam SMS, “Linear Programming”, Kabir Publications, Dhaka,
Bangladesh, 2008.
Islam SMS, Rahman AKMF and Majumdar AAK, “An Algorithm
for Generalized Assignment Problem with Employer and
Employees Satisfactions”, Journal of Mathematics and
Mathematical Science, Jahangirnagar University, Vol. 19, pp
43-52, 2004.
683
Appendices
Kuhn HW, “The Hungarian Method for the Assignment Problem”,

Naval Research Logistics Quarterly, Vol. 2, pp 83-97, 1955.
Majumdar AAK, “Lecturer Notes on Advanced Mathematical
Programming”, Jahangirnagar University, Dhaka, Bangladesh,
1996.
Reinfeld NV and Vogel WR, “Mathematical Programming”,
Prentice-Hall, Englewood, Cliffs, NJ, USA, 1958.
Sinha SM, “Mathematical Programming Theory and Methods”,
Elsevier, New Delhi, India, 2006.
Taha HA, “Operations Research: An Introduction”, Pearson
Prentice Hall, NJ 07458, USA, 2007.
684
View publication stats

Mathematical Programming

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mathematical Programming

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Book · January 2020

The user has requested enhancement of the downloaded file.

Dr. S. M. Shahidul Islam

LAP LAMBERT Academic Publishing

Cover image: www.ingimage.com

Copyright © 2020 Dr. Shahidul Islam (sislam.math@gmail.com)

Copyright © 2019 International Book Market Service Ltd., member of

Price: EUR (€) 108.90

Mst. Bakul Nahar Parveen

Mst. Sumaiya Islam

Dr. S. M. Shahidul Islam

1.1 Linear programming

second product requires 4 units of A and 2 units of B. It is known

1.3 Summation symbol

1.4 General linear programming problem

a11 x1  a12 x 2  ...  a1 j x j  ...  a1n x n  (or  or )b1 

(c) Minimize (or, maximize) C.X

(a) If any constraint is of  type (or,  type) add slack

Subject to 4x1 + 3x2 = 240

minimization type, or connected by the sign ‘  ’ when the

x3//  0 in the last form, we get it as below.

sales, demand, etc. Just because we can produce so many

example, 5x1 + 3x2 is the objective function of the following linear

Minimize (–z) = – 2x1 – 3x2

Maximize z = x1 + 2x2 + x3 + 2x4

Since the rank of A is 2, the given system is consistent. We find

1.9.7 Degenerate basic solution

constraints and non-negativity conditions. As for example, X = (0,

1.9.13 Slack variables

then the non-negative variables xn+i satisfying

then the non-negative variables x n i satisfying

the inequality to convert it into an equation. The variables which

the feasible solution, suppose first k (k  n) components be positive

Case – 1: If k  m and column vectors a1, a2, ..., ak are linearly

Here, at least one si is positive (if necessary, multiply both sides of

maximum (k –1) positive components is a feasible solution of the

Rank of the matrix A is 2 and hence the given equations are

1.10 Some done examples

x1, x2, x3, x4, x5, x6  0.

of the first constraint and subtracting surplus variable s2  0 from

After simplifying, we get

x1 + x2 + 3(– x3/ ) – ( x 4/ – x 4// ) + s1 = 5

Solution: Replacing the equality constraint by two inequalities, we

4x1 + 2x2 – 3x3 = 1

of the above equation, we have x2 = – 1/2. Hence the basic

Subtracting double of the second equation from the multiplication

Solution: We can write the given system in the matrix form as

– 4s2 + 3s3 – 8s4 = 0.

 2r1  2r2  4r4   0 

2r1  2r2  4r4  0

Multiplying equation (vii) by –1, we have the new form of

Maximize z = 2x1 + 4x2 – x3 + 2x4

Solving this system, we have x2 = ½, x3 = 0. Hence, a fourth basic

This is the canonical form of the given LP problem.

1.8 Write the expanded form of

1.9 Express using summation symbol:

1.10 Convert the following LP problems into standard form:

xvii) Maximize z = 7x1 + 2x2 + 3x3 + 2x4

Subject to – 6 x1/ + 2 x2/ + 5 x3/ – 5 x3// – s1 = 9

s. Minimize z = 3x1 + 7x2 + 5x3 + 3x4 – 3x5

a. x1 + 2x2 + x3 = 4 b. 2x1 + 2x2 – x3 = – 4

1.21 If x1 = 2, x2 = 1, x3 = 1 be a feasible solution to the following

Minimize z = 2x1 + 5x2