Linear Programming

AVU-PARTNER INSTITUTION
MODULE DEVELOPMENT TEMPLATE
Prepared by Thierry Karsenti, M.A., M.Ed., Ph.D.
With : Salomon Tchaméni Ngamo, M.A. & Toby Harper, M.A.
Version 19.0, Thursday December 21st, 2006
Module Development Template 1

C. TEMPLATE STRUCTURE
I. INTRODUCTION
1. TITLE OF MODULE
Linear Programming
2. PREREQUISITE COURSES OR KNOWLEDGE

The Prerequisite courses are Basic Mathematics and Linear Algebra, which are offered in
this degree program. A knowledge of linear independence, basis, matrix operation,
inverses, inequalities, vector spaces, convex sets, and graph plotting is essential. These
concepts and skills are generally covered in the pre-requisite course (or equivalents)
mentioned above and constitute important background knowledge required to undertake
this module. A basic understanding of these and related concepts, and reasonable
competence in related manipulative skills (such as matrix and graphical representations
and associated algebraic manipulations), are essential background knowledge for this
module. Familiarity with these basic concepts and skills, which are assumed in this
module, must be secured first before proceeding with the module.
3. TIME
The recommended total time for this module is at least 120 study hours, with Unit 1 taking
40 hours [20 hours for each of the 2 Activities], and Unit 2 taking 80 hours [20 hours for the
first Activity, 34 hours for the second Activity, and 20 hours for the third Activity], and the
remaining 6 hours to be allocated for the pre-assessment (2 hours) and summative (4
hours) evaluation activities.
4. MATERIAL
Students should have access to the core readings specified later. Also, they will need a
computer to gain full access to the core readings. Additionally, students should be able to
install the computer softwares wxMaxima and Graph and use them to practice algebraic
concepts. These should be regarded as learning materials to facilitate easier accessing
and processing of the core concepts and skills that constitute the course. The following are
materials necessary to engage with the module meaningfully and, hopefully, complete it
successsfully: The student’s edition of the module (print form); a computer with effective
internet connectivity and MicroSoft Office 2003 and above; a scientific or programmable
calculator; graph plotting materials; CDs with materials downloaded from sites
recommended in the module; CDs with mathematical software such as MathType or
WinShell, Graph, wxMaxima, and at least one linear programming software that is free-
downloadable, recommended readings from texts identified in the module. [The
recommended readings can also be in print form].

5. MODULE RATIONALE
The importance of linear programming derives in part from its many applications and in part
from the existence of good general-purpose techniques for finding optimal solutions. Linear
programming is useful for guiding quantitative-related decision making in business,
industrial engineering enterprises, and to a lesser extent activities within the social and life
sciences. The linear programming skills will help teachers in some aspects of their own
personal life management activities and in their professional practice.
This module acts as a smooth and non-intimidating entry into the mathematical worlds of
dynamic linear programming, networks, and operations research for the learner who will
develop some interest in majoring those fields. Also it:
(a) is important in and of itself as a degree level mathematics course because it
introduces to the mathematics student new mathematical content with a distinctive
style of mathematical thinking;
(b) beautifully integrates theoretical concepts with their practical applications – both of a
mathematical and everyday - life in nature;
(c) is necessary for the prospective teacher of science and mathematics because
modern day youth and school students are now pre-disposed to a range of career
interests, many of which would be facilitated by a preparation that involves dealing
with linear programming and optimisation that are covered in this module.

II. CONTENT
6. Overview
Overview: Prose description
This module introduces the learner to a particular mathematical approach to analysing real
life activity that focuses on making specific decisions in constrained situations. The
approach, called linear programming, is presented here with an emphasis on appreciation
of the style of thinking and interpretation of mathematical statements generated, rather than
on computational competency per se, which is left to appropriate and readily available ICT
software package routines.
The module begins with Unit One that consists of 2 main Activities. Activity 1, formulation
of a linear programming problem, is on a mathematical description of the problematic
situation under consideration, and Activity 2, the geometrical approach considers a visual
description of a plausable solution to the problem situation. Unit 1 therefore should move
the learner towards an appreciation of real-life activity situations that can be modelled as
linear programming problems.
With 3 main activities, Unit 2 considers computational algorithms for finding plausible
optimal solutions to the linear programming problem situations of the type formulated in
Unit 1. Activity 3 examines conditions for optimality of a solution, which is really about
recognising when one is moving towards and arrives at a candidate and best solution.
Activity 4 discusses the centre piece of computational algebraic methods of attack, the
famed Simplex algorithm. This module focuses on the logic of the algorithm and the useful
associated qualitative properties of duality, degeneracy, and efficiency. The final Activity
touches on the problem of stability of obtained optimal solutions in relation to variations in
specific input or output factors in the constraints and objective functions. This so called
post optimality or sensitivity analysis is presented here only at the level of appreciation of
the analytic strategies employed.

Flow of Learning
Unit 1
Identifying, describing, understanding, and appreciating the general linear

programming problem situation, and plausible solutions for it.
Unit 2
Computational strategies for seeking solutions of linear programming problems,

recognizing potential and best solutions, and efficiency considerations.
Content Outline:

Unit 1: The linear programming problem
 Formulation of a linear programing problem
o The general linear programming problem
o The standardized linear programming problem
 Geometrical interpretation of a solution of a linear programming problem

o Two dimensions
o More than 2 dimensions
Unit 2: Computational algorithms

 Searching for and recognising a potential solution: optimality conditions for the
objective function of a linear programming problem
o Boundedness
o Convergence
 Algebraic interpretation of the solution to a linear programming problem

o The Big M algorithm
o The Simplex algorithm
o degeneracy
o efficiency
o Notion of duality
o the primal simplex
o the dual simplex
 Stability considerations for a solution: sensitivity analysis

o Marginal analysis
o Parametric analysis

LP problem
7
in Standard
ACTIVITY 1 Form
Formulation of the
linear programming General LP
problems
UNIT 1
2-var PL
ACTIVITY 2
Geometrical Solution
3-var PL
Boundedness
MODULE ACTIVITY 3
Optimality Condition
Convergence
7. General Objective(s): for the whole module

Degeneracy
Simplex Method
Efficiency
UNIT 2
ACTIVITY 4 Big M-Method
Algebraic Approach
Primal Simplex
Duality
Dual simplex
Module Development Template

Graphic Organiser
Parametric Analysis
ACTIVITY 5
Sensitivity Analysis
Marginal Analysis
Upon completion of this module students should:
a) have a general appreciation of the types of problems which are amenable to
analysis using linear programming
b) be able to formulate linear programming problems and solve them using
geometrical and linear algebraic techniques.
c) be able to use mathematical software packages to solve linear programming
problems
d) be able to discuss some theoretical notions of linear algebra and geometry with
concrete/practical contexts.
e) have developed some familiarity with the language of operations research
f) have developed a sense of algorithmic thinking

8. Specific Learning Objectives (Instructional Objectives): separate objectives for
each unit
You should be able to:

1. Identify and effectively model suitable problems with linear programming.
2. Apply knowledge of inequalities to solving optimisation related problems.
You should secure your knowledge of school mathematics in:

1. Linear functions and simultaneous equations.
You should exploit ICT opportunities in:

1. Using graph drawing software to investigate linear functions.
2. Using Computer Algebra Systems (CAS) software to solve linear systems.
Unit Learning objective(s)

1. Problem Formulation o The learner should be able to
identify/recognize an optimization
situation in real life decision making
activity.
o The learner should be able to produce a
mathematical model using appropriate
language.
2. Computational Algorithms o The learner should be able to describe,
explain, and apply optimality conditions
to specific linear programming problems.
o The learner should be able to describe
the underlying logic of the Simplex
algorithm.
o The learner should be able to relate the
algebraic solution with the geometric
solution.
o The learner should be able to perform the
Simplex algorithm on specific problem
situations with an appropriate linear
programming software and interpret the
resulting solution.
o The learner should be able to explain the
meaning of duality and describe its role in
the search for solutions of linear
programming problems.
o The learner should be able to describe
the purposes of carrying out a sensitivity
analysis for a given linear programming
solution.
o The learner should be able to describe a
procedure for carrying out a sensitivity

analysis to a given optimal solution.

III. TEACHING AND LEARNING ACTIVITIES
9. PRE-ASSESSMENT
Title of pre-assessment: Basic Algebraic Ideas Test

Rationale: To check leaner’s familiarity with some concepts assumed in the module
QUESTIONS
1. Which of the following is linear:

(a) ax2+by = c (b) ax - by = c (c) ax + by2 = c (d) acosx + by = c
2. Which of the following does not typically denote a vector:

�4 �
� �
(a) -5 (b) (1, 2, 3) (c) A (d) �8 �
�
�-3�
�
2 4 2 2 0�
�
�
1 1 6 4 9�
3. The matrix � �is:
�
0 0 0 0 0�
� �
(a) 35 (b) 25 (c) 53 (d) 5×2
4. A singular matrix is one that:

(a) is single (b) is invertable (c) is non-invertable (d) has determinant 1?
2
� 4 2 2 0�
�
1 1 6 4 9�
� �
5. The matrix �
0 0 0 0 0 �has rank:
� �
1
� 2 1 1 0�
(a) 0 (b) 1 (c) 2 (d) 3

6. Consider the linear equation x + y ≤ 1. Which of the graphs satisfy this inequality?
N.B. The wanted region is shaded.

(a)
(b)
(c)
(d)

7. If ax + by ≤ c for some numbers a, b, c then ax + by + d = c for some positive
number d.
(a) TRUE (b) FALSE
8. Which of the following is not directly associated with the Gaussian elimination method?
(a) Reducing n x n matrix to echelon form
(b) Detemining the consistency of a system
(c) Finding lower or upper triangular matrix
(d) Using elementary operations to reduce a system of equations
9. What is the transpose of matrix A?
�4 1 3�
A=� �1 0 2� �
�
�2 -1 5 �
�
1 0 2� �
� 4 1 2� � 4 5 3� � 1 1 3�
� �
2 -1 5 �(b) �
(a) � � �
1 0 -1�(c) ��
0 1 2�
2 -1 0 �( d ) � �
�
4 1 3�
� � � �3 2 5� � � �1 2 2� � � 2 5 1�
� �
10. Which of the following shows a convex set?
(a) (b)
(c) (d)
N.B. An object is convex if for every pair of points within the object, every point on the
straight line segment that joins them is also within the object.
11. The sequence {1/n} is bounded below by:

(a) 1 (b) ½ (c) 0 (d) n where is very large?
12. At an optimal value of a function:

(a) The function attains a local maximum value only.
(b) The function attains a local minimum value only
(c) The function attains a maximum or minimum value.
(d) None of the above statements is true?
13. The following is an example of a discontinuous function.

(a) y = |x| (b) y = 1/x (c) y = 3 (d) y = e-x

14. A basis of a vector space
(a) Has linearly dependent vectors.
(b) Is a set of only unit vectors.
(c) Has basis vectors that span the whole vector space.
(d) Can include the zero vector.
15. Which of these sequences does not converge?
(a) {(sin (n))/n} (b) {ln (n) } (c) {1/n} (d) {(n+1)/n}
16. A function is said to be bounded if:
(a) It is defined in open interval.

(b) It is limited above and below only.
(c) It is limited above only.
(d) It is limited below only?
1 2 3�
�
�
0 4 5�
17. The determinant of the following matrix � �is:
�
0 0 6�
� �
(a) 11 (b) 24 (c) 21 (d) 18
18. Maximising profit generation can be equivalent to minimising production costs:
(a) TRUE (b) FALSE?
19. A vector has:

(a) magnitude only.
(b) direction only.
(c) magnitude and direction only.
(d) magnitude, direction and field only?
20. A Subspace of vector space:

(a) Is also a vector space.
(b) Is not a vector space.
(c) Is not a linear space.
(d) Is half of a vector space?

ANSWER KEY
1.b ( a and c are parabolic functions, d has a periodic function cosx)
2.a (a has magnitude but no direction, c typically denotes a matrix or vector, b and d are
vectors )
3.a (the matrix has 3 rows and 5 columns and so it’s a 3 by 5 matrix)
4.c (a singular matrix is a matrix with determinant zero and so it non-invertable)
5.d (the rank of a matrix is the number of non-zero rows and here they are three)
6.a (the graph in d gives the equation x - y �1 , b has shaded the unwanted region and c
has shaded the whole x-y plane)
7.a (for equality to hold you add some positive number on the left hand side)
8.c (you use the leading entry in each row to eliminate the non-zero coefficients below it,
thus creating an upper triangular matrix)
9.b (for transpose the first column becomes first row, second column becomes second row
and third column becomes third row so answer is b)
10 .c (c is the only shape in which when two points in the region are joined by a line, all the
points on the line are also in the shape)
11. c (the sequence is a decreasing sequence which approaches zero)
12 c (the optimal point is the maximum or minimum point)
13.b (for b the function is continuous everywhere else except at zero hence discontinuous)
14.c (for a vector to be called a basis it should be linearly independent and should span or
generate the whole vector space)
15.b (a and c converge to zero, d converges to 1 )
16.b (a function is bounded if it is bounded above and below)
17 b (just multiply the terms in the diagonal) .
18.a ( if you minimize production costs then you will maximize profit)
19.c (a vector quantity has size and direction e.g. velocity, acceleration)
20.a (a subspace of a vector space is also a linear or vector space since it satisfies the
three main properties of a vector space` )

PEDAGOGICAL COMMENT FOR LEARNERS
If you get below 50% in this pre-assessment it might idicate that you have forgotten some
facts of linear algebra. These might include things like simple definitions, properties and
computational procedures for such objects like matrices, vectors, sets, linear systems of
equations, and real numbers. In that case you are encouraged to browse through the
module on basic mathematical ideas and linear algebra before proceeding. If you get more
than 50% you are also encouraged to review the same module as needed while proceeding
with this module. Some of these basic ideas will surface as assumed knowledge in one
form or another in this module.

10. LEARNING ACTIVITIES
[Note: all key concepts are defined in the glossary, which is given in section 11 below.
Where a definition is encountered in the learning activities, reference is made to the
glossary for its articulation, while the Learning Activities concentrate on developing the
concept or skill that is carried by the definition. This is a device to minimise repetition.]
UNIT 1: PROBLEM FORMULATION
Learning Activity # 1: Formulation of a Linear Programming Problem
Specific learning objectives
o The learner should be able to identify/recognize an optimization situation.

o The learner should be able to produce a mathematical model using appropriate
language.
o The learner should be able distinguish and relate between the Standard and the
General form.
o The learner should be able represent the mathematical model geometrically.
o The learner should be able identify feasible regions, vertices, and convexes.
o The learner should understand the concepts of systems of linear equations,
constraint, feasible solution and feasible region.
o The learner should be able to interpret a real life problem and transform it into a
linear programming problem.
o The learner should be able to check or verify the feasible solution.
o The learner should be able to express the system of linear equations graphically, i.e.
the learner should understand the geometry of the linear programming model.
o The learner should be able to resolve the linear programming problem by the
geometrical approach.
Summary of the learning activity
The subject of linear programming has its roots in the study of linear inequalities. In this unit
we give an introduction to linear programming starting with a simple real life problem which
the learner can easily relate with. In linear programming the objective is to maximize or
minimize some linear functions of quantities called decision variables. This can be done
algebraically or geometrically. However here we resolve the linear programming problem
geometrically. With specific examples you are going to see how problems in business,
physical, chemical and biological sciences, engineering, architecture, economics,
agriculture and management are formulated. Learning will involve a number of activities
such as reading about linear programming, optimization, operation research and the nature
of the objective functions and constraints. The learning activities could also involve
extensive use of mathematical software packages to solve the linear programming
problems.

Key Terms (refer to glossary)
Linear programming
Problem formulation
Objective function
Optimal solution
Constraints
Feasible solution
Basic solution
Basis
Basic variables
Dictionary
Nonbasic variables
Slack variable
Surplus variable
Artificial variable
Unbounded solution
List of required readings
 Linear Programming: Foundations and Extensions by Robert J. Vanderbei

www.princeton.edu/~rvdb/LPbook/onlinebook.pdf
Last visited: 15-02-07
 Lecture notes on Optimization by Pravin Varaiya

http://robotics.eecs.berkeley.edu/~varaiya/papers_ps.dir/NOO.pdf
 A gentle approach to linear programming: chapters 1-7

http://www.sce.carleton.ca/faculty/chinneck/po/Chapter1.pdf
List of relevant useful links
Linear Programming Formulation

http://people.brunel.ac.uk/~mastjjb/jeb/or/lpmore.html

http://home.ubalt.edu/ntsbarsh/opre640a/partVIII.htm
An Introduction to Linear Programming and the Simplex Algorithm by Spyros

Reveliotis
http://www2.isye.gatech.edu/~spyros/linear programming/linear programming.html
OR-Notes J E Beasley
http://people.brunel.ac.uk/~mastjjb/jeb/or/twomines

Linear Programming
http://home.ubalt.edu/ntsbarsh/opre640a/rpcotdp#rpcotdp

www.people.brunel.ac.uk/~mastjjb/jeb/lp.html
Detailed description of the activity
The learner will:

o Read the notes that are available in the recommended textbook in order to get a
general understanding of the formulation of a Linear Programming Problem.
o Discuss in groups a situation which can be interpreted as a Linear Programming
Problem and is formulation. Here one person describes a maximization/ minimization
situation and allows someone else to Formulate that situation.
o Attempt the exercise given: The learner refers to indicated pages of the recommended
textbook .
o Open the relevant links to get deeper understanding of processes of problem
formulation.
Learning activities
Problem: A manager is worried about what to do to maximise production levels.
Suppose that we have a commercial farmer who produces a variety of farm produce on a
quarterly basis which include maize, beef, beans, etc. This farmer can only produce a
certain number of products due to several reasons. The main limitations are the size of
his/her farm, the amount of inputs he/she needs, the market, climatic conditions, etc.
Despite all these setbacks the farmer hopes to get some profit after selling his/her produce.
This is a typical example of a linear programming problem situation. Before defining what
we mean, in general, by a linear programming problem, let us consider a few practical real-
world problems that serve to motivate and, at least vaguely, to define this subject.

Reading
Read the following textbook section:

Linear Programming: Foundations and Extensions by Robert J. Vanderbei pages 3-10
(Managing a production facility)
Read also the following link:

The Standard Form of the Linear Programming Problem
http://en.wikipedia.org/wiki/Linear_programming
Notes
The following may help you to understand the notations from the above readings:
n is the number or variety of products produced by the company.

m is the number of types of raw materials .
bi is the amount of ith raw material , for i= 1, 2,…,m
ri is the unit market price of ith raw material, for i=1, 2,…m at a given moment.
aij
is the amount of ith raw material required to produce one unit of jth product, for j=1,2,
….,n
sj
is the unit market price for the jth product, for j=1,2,….,n
xj
is the number of jth product produced
m
�p a
i =1
i ij
m
�r a i ij = r1a1 j + r 2 a2 j + ... + rm amj

is the cost of production of one of jth product, i =1
cj
is net profit for the jth product
m
c j = s j - �pi aij
i =1
cj xj
xj
is the total net profit for units of product j.
Discussion
In pairs narrate a situation in which Linear Programming is involved and then formulate that
problem into a Linear Programming Problem.
Are you able to visualize the Formulation?
The General Form of the Linear Programming Problem

The general form for a Linear Programming problem is as follows:
Objective Function:
Maximise/minimise
f ( x1 , x 2 , ...., x n ) = c1 x1 + c 2 x 2 + .. + c n x n
Where:
x is the decision variable
n is the number of decision variables in the objective function
cj is unit cost or unit profit of the jth decision variable
Such that:
Technological Constraints:
��
�
��
ai1 x1 + ai 2 x2 + L + ain xn �= �b
�
��
� �
for i = 1, 2,L , m
Where
�a11 � �a12 � �a1n �
� � � � � �
ai1 = �M� , ai 2 = �M � , L , ain = �M �
�am1 � �am 2 � �amn �
� � � � � �
Sign Restrictions
xj ≥ 0
Further reading
Read the following link:

An Introduction to Linear Programming and the Simplex Algorithm by Spyros Reveliotis

A Step-by-Step formulation of the linear programming problem
Question: What is linear programming problem formulation?
Answer: Problem formulation or modeling is the process of translating a verbal statement

of a problem into a mathematical statement.
Example of linear programming problem [adapted from Koshy (1979)]
A farmer grows tomatoes and peas on her 125 hectres piece of land. It costs $20 000
dollars grow a hectare of tomatoes and $10 000 to grow a hectare of peas. However , Agri
Bank gave her only $1 500 000 as loan. It takes 18 working-hours to grow an hectare of
tomatoes and 6 working-hours to grow a hectare of peas. She wants to devote 1080
working-hours for the whole job. If the profits from a hectare of tomatoes and a hectare of
peas are $40 000 and $25 000, respectively. How many hectares of each should she grow
to maximize her total profit? What is the maximum possible profit?
Steps to follow
Question: What is the problem of the farmer?
Answer: The problem of the farmer is that of maximising her profit having constraints
on available land, production costs, and the number of working hours
Question: What are the constraits?
Answer:
i) Let x1 and x2 be the decision variables of the number of hectares for beans and
peas respectively.
Then x1 + x2 ≤ 125 constraint on available land.
ii) Her total growing expenses are 20000x1 + 10000x2 ≤ 1500 000
Or 2x1 + x2 ≤ 150 constraint on expenses
iii) The total number of required working hours is 18x 1 + 6x2 ≤ 1080
Or 3x1 + x2 ≤ 180 constraint on working hours
iv) x1 ≥ 0, and x2 ≥ 0 non negativity constraints
Thus the problem is to maximise the profit function
f ( x1 , x2 ) = 40000 x1 + 25 x2
(1)
subject to the constraints

x1 + x2 �125 �
2 x2 + x2 �150 �
�
�
3 x1 + x2 �180 �
� (2)
and
�
x1 �0, x2 �0 � �
Summary guidelines for linear programming problem formulation
a) Understand the problem thoroughly.

b) Describe the objective.
c) Describe each constraint.
d) Define the decision variables.
e) Write the objective in terms of the decision variables.
f) Write the constraints in terms of the decision variables.
Exercise 1
1. Formulate a mathematical model of the problems the production manager and

comptroller are suppose to solve.
2. Give examples of four other fields where optimization plays a pivotal role.
3. State the three conditions that are necessary for linear programming problem to be
linear.
4. Go the page 8 (of textbook) of Linear Programming: Foundations and Extensions by
Robert J. Vanderbei and do questions 1.1 and 1.2
The General Linear Programming Problem
Readings
1. Read page 6 in the textbook.

Foundations and Extensions by Robert J. Vanderbei.
2. Follow this internet link for a detailed definition:

http://www.it.uu.se/edu/course/homepage/opt1/ht06/Lectures/fundamentat_thm-2up.pdf
Discussion
Find colleagues to talk this through with.

Do you now understand the formulation of a linear programming problem? If not consult
your tutor or your colleagues who have grasped the main idea.
Note
The following is a summary for the formulation of a Linear Programming Problem

points to take note of:
 Read the whole problem situation.

 Identify and define your unknown quantities (variables and constants).
 Express the objective function and the constraints in terms of variables.
Answers to Exercise 1
Question 1
For the answer to this question check “Managing a Production Facility” from page 3 to 5 of
Linear Programming: Foundations and Extensions
Question 2
Some of the major application areas to which linear programming can be applied are:
Blending
Production planning
Oil refinery management
Distribution
Financial and economic planning
Manpower planning
Blast furnace burdening
Farm planning
Question 3
Conditions for a mathematical model to be a linear program (linear programming) are:
 all variables continuous (i.e. can take fractional values)

 a single objective (minimise or maximise)
 the objective and constraints are linear i.e. any term is either a constant or a
constant multiplied by an unknown.

Question 4
1.1
Let x and y be our variables for time given to produce bands and coils respectively
.
Time constraint
x + y ≤ 40
Tonnage constraints
200x ≤ 6000
140y ≤ 4000
Objective function
f ( x, y ) = 25(200 x ) + 30(140 y )
We rewrite the problem as:
Maximize
f ( x, y ) = 5000 x + 4200 y
Such that
x + y ≤ 40
200x ≤ 6000
140y ≤ 4000
x ≥ 0, y ≥ 0

Learning Activity # 2: A Geometrical Approach to Linear Programming
Specific Learning Objectives
By the end of this section the learner must:

o Be able to identify and construct linear equations.
o Be able to construct the wanted regions given a set of inequalities, i.e. the feasible
regions of each Linear Programming Problem.
o Be able to solve Linear Programming Problems using the geometric approach.
Summary of the Learning Activity
In this activity we introduce the resolution of linear programming problems by geometry.

Here we look at notions like feasible region, feasible solutions, optimal solution and
convexity. Geometrically, linear constraints define a convex polyhedron, which is called the
feasible region. Since the objective function is also linear, hence a convex function. The
linearity of the objective function also implies that an optimal solution can only occur at a
boundary point of the feasible region, unless the objective function is constant, when any
point is a global minimum.
Note: There are two situations in which no optimal solution can be found. First, if the
constraints contradict each other (for instance, x ≥ 2 and x ≤ 1) then the feasible region is
empty and there can be no optimal solution, since there are no solutions at all. In this case,
the linear programming is said to be infeasible.
Feasible region
Feasible solutions
Optimal solution and convexity
Feasible production set
Production possibility set or opportunity set
Extreme points
Infeasible problem
Hyperplane
Halfspace
Convex polyhedral set
Convex polyhedral cone
List of Required Readings

 Lecture notes on optimization by Pravin Varaiya

List of Relevant Useful Links
Linear Programming: A Geometric Approach

http://www.wiley.com/college/sc/sullivan/CH03.pdf
Graphical Solution Method

Last visited: 21/02/07
Linear Programming: Geometric Approach

www.math.tamu.edu/~janice.epstein/141/notes/Ch3.pdf
Detailed Description of the Activity
The learner will:
o Read from the link given in order to have an understanding of the resolution of a
Linear Programming problem by geometry. This will give the learner a general picture
on the geometry approach.
o Analyze the example given in the link. This will help the learner to understand the
steps and procedures followed in the resolution.
o Do the exercise from the textbook Linear Programming: Foundations and Extensions
by Robert J. Vanderbei.
Learning Activities
A man walking along a line-segmented boundary of a field
Reading
Linear Programming: Foundations and Extensions by Robert J. Vanderbei page 22

section 5. www.princeton.edu/~rvdb/LPbook/onlinebook.pdf

Further reading
Read the following link on the Geometrical Approach

Linear Programming: A Geometric Approach page 171-176
This link will help you to understand how to graph inequalities and identify the wanted
regions which are called feasible regions in linear programming problems. The feasible
region contains all possible solutions and from these we find the optimal solution.
NOTE
When the number of variables in a linear programming problem is two or three the feasible
solutions can be determined graphically by drawing the graphs of inequalities of the
constraints.
The following steps are essential when one is solving linear programming problems
geometrically.
Step 1: Graph the constraints

Step 2: Identify the feasible region
Step 3: Graph the objective function (twice)
Step 4: Find the cornerpoints to the region of feasible solutions
Step 5: Evaluate the objective function at all the feasible corner points by:
Step 6: Identifying the optimal point
Step 7: Finding the coordinates of the optimal point
Step 8: Evaluating the objective function at the optimal solution.
Example
Lets us try to follow the steps using the following problem.
Maximize:
f ( x, y ) = 5 x + 7 y
subject to:
x + y - 7 �0
2 x - 3 y + 6 �0
x �0, y �0
1. Is the problem an linear programming?
Yes, if and only if: All variables have power of 1, and they are added or subtracted
(not divided or multiplied). The constraint must be of the following forms ( ≤, ≥, or =,
that is, the linear programming-constraints are always closed), and the objective
must be either maximization or minimization.

Answer Yes
2. Can I use the graphical method?
Yes, if the number of decision variables is either 1 or 2.
3. Use Graph Paper. Graph the constraints:
x + y - 7 �0
2 x - 3 y + 6 �0
x �0, y �0
4. Shade the unwanted region of each constraint inquality.
5. Throw away the sides that are not feasible.
After all the constraints are graphed, you should have a non-empty (convex) feasible
region, unless the problem is infeasible.

6. Create (at least) two iso-value lines from the objective function, by setting the
objective function to any two distinct numbers. Graph the resulting lines. By moving
these lines parallel, you will find the optimal corner (extreme point), if it does exist.
In general, if the feasible region is within the first quadrant of the coordinate system
then, for the maximization problems you are moving the iso-value objective function
parallel to itself far away from the origin point (0, 0), while having at least a common
point with the feasible region. However, for minimization problems the opposite is
true, that is, you are moving the iso-value objective parallel to itself closer to the
origin point, while having at least a common point with the feasible region. The
common point provides the optimal solution.
Exercise 2
Read Linear Programming: Foundations and Extension by Robert J. Vanderbei page 24
Work through numbers 2.1, 2.2, 2.3, 2.5 and 2.10.
Note: You are only required to construct the feasible regions
Further reading
Now go to the following links and read more on resolution by geometry:
Linear Programming: A Geometric Approach page 171-176

This is a maximizing problem in which the profit is being maximized.
Linear Programming: Geometric Approach

Answers to Excercise 2
For answers to the given problems check on page 449 of your main book, Linear
Programming: Foundations and Extensions by Robert J. Vanderbei

UNIT 2: RESOLUTION AND ANALYSIS
Learning Activity # 3: Optimality Conditions
After studying this section, you should:

o be able to interpret the derivatives and existence of optimality
o be able to prove the theorems of optimality
o be able to explain and prove the The Fundamental Theorem of Linear Programming
o be able to check optimality conditions for solving a Linear Programming problem.
o be able to carry out a sensitivity analysis for a given optimal solution.
In this unit we want to consider the computational aspects of the linear programming
problem and do a sensitivity analysis used to determine how stable the optimal solution is
to changes in some related variables. Here the resolution is done algebraically, i.e. using
the notions of duality, simplex method, the method of Big M, and the mutual primal
algorithm. Therefore the learner needs knowledge from unit 1 in order to understand unit 2
because the transformation of the linear programming problem into standard form helps in
the resolution of the linear programming problem algebraically. In this unit we do some
analysis using the derivatives and existence conditions of optimality.
Basic variables
Nonbasic variables
Augmented matrix
Pivot
Pivot column
Test ratio
Tableau
Cycling


Optimality Conditions for Constrained Optimization

http://ocw.mit.edu/NR/rdonlyres/Sloan-school-of-Management/15-
084JSpring2004/7240EF84-B20D-419F-B1C0-2DAF3277F5C4/0/lec6_constr_opt.pdf
Optimality Conditions
http://www.math.mtu.edu/
%7Emsgocken/ma5630spring2003/lectures/lag1/lag1/node1.html
Alternate Optimal Solutions, Degeneracy, Unboudedness, Infeasibility
http://mat.gsia.cmu.edu/QUANT/notes/node63.html#SECTION008300000000000000
00
Learning Activities
“How do you I know I have arrived at the best position?”
Optimality Conditions
In simple terms, optimality conditions are certain conditions which should tell you that you
have reached the optimal solution. For the Maximization Problem if all nonbasic variables
have negative or zero coefficients in the objective function, an optimal solution has been
obtained. If you were to substitute one of the nonbasic variables with a non-negative value
(since the variables are non-negative), then the value of the final solution is reduced.
Hence you are certain that the optimal solution has been obtained. Similarly, for the
Minimization Problem if all non-basic variables have positive or zero coefficients in the
objective function, an optimal solution has been obtained.
Reading
For Conditions of Optimality for a Linear Programming Problem and

the Fundamental Theorem of Linear Programming, read
Linear Programming: Foundations and Extensions by Robert J. Vanderbei

Further Reading

http://engr.smu.edu/~barr/ip/ch1/node7.html
http://www.maths.abdn.ac.uk/~igc/tch/mx3503/notes/node67.htm
Exercise 3
Look at the example from Linear Programming: Foundations and Extension by Robert J.
Vanderbei, page 13, whose solution is given on pages 15 and 16.
Now do this Exercise adapted from Wagner (1975)
Test for optimality conditions on the following linear programming probem
Maximize 4 x1 + 5 x2 + 9 x3 + 11x4
Subject to
x1 + x2 + x3 + x4 �15
7 x1 + 5 x2 + 3x3 + 2 x4 �120
3 x1 + 5 x2 + 10 x3 + 15 x4 �100
x1 �0, x2 �0, x3 �0, x4 �0
Solution
Let x0 be the value of the objective function add the slack variables x5 , x6 , x7 to your
constraints so that you have equality. Then write the system as
x0 - 4 x1 - 5 x2 - 9 x3 - 11x4 = 0
x1 + x2 + x3 + x4 + x5 = 15
7 x1 + 5 x2 + 3x3 + 2 x4 + x6 = 120
3 x1 + 5x2 + 10 x3 + 15 x4 + x7 = 100
After going through four iterations of the Simplex Method you get to a stage where the
nonbasic variables x2 , x4 , x5 or x7 can only have value zero. That is to say we have
695 3 11 13 5
x0 = - x2 - x4 - x5 - x7
7 7 7 7 7
This is the optimality condition which allows us to terminate the iteration.

Learning Activity # 4: Algebraic Approach
After studying this section, you will:

o Be able to interpret the derivatives and existence of optimality.
o Be able to apply the Method (penalty) of big M and Dual simplex method.
o Be able to differentiate the properties of duality, mutual primal algorithm, and dual
simplex algorithm.
o Be able to follow the given algorithms step-by-step, i.e. mutual primal and the dual
simplex algorithms.
o Be able to solve linear programming problems using appropriate mathematical
software (when available)
In this section we will learn how to solve the linear programming problem algebraically. The
main algorithm we shall follow is the Simplex Algorithm. The simplex algorithm, developed
by George Dantzig (1947) solves linear programming problems by constructing an
admissible solution at a vertex of the polyhedron and then walking along edges of the
polyhedron to vertices with successively higher values of the objective function until the
optimum is reached. Although this algorithm is quite efficient in practice and can be
guaranteed to find the global optimum if certain precautions against cycling are taken, it
has poor worst-case behavior: it is possible to construct a linear programming problem for
which the simplex method takes a number of steps exponential in the problem size.
Actually, for some time it was not known whether the linear programming problem was
solvable in polynomial time.
Basis
Basic variables
Nonbasic variables
Slack variable
Surplus variable
Artificial variable
Basic solution
Degeneracy
Primal problem
Dual problem
Weak duality
Strong duality
Primal feasibility
Dual feasibility
Primal feasibility
Dual feasibility



Simplex Method - Big M

http://www.math.uwo.ca/~heinicke/courses/236_03/bigM.pdf
Simplex Method – Big M

http://www.computing.dcu.ie/~lkillen/teach/CA427Simplexbigmexample.pdf
Dual Problem: Construction and Its Meaning

Sensitivity Analysis for Linear Programming

http://mat.gsia.cmu.edu/QUANT/notes/node64.html
Last visited: 20/02/07
The learner will:

o Read the the textbook Linear Programming: Foundations and Extensions on the
Simplex Method pages 13-28, 29-43, and 89-107 of the textbook
o Analyze the example that is provided in the link below:
Simplex Method - Big M
o Do the exercise from the textbook Linear Programming: Foundations and Extensions
on the Simplex Method page 13 of the textbook numbers 2.1, 2.2, 2.3 and 2.4.
In addition to reading about the logic of the Simplex Algorithm and practicing applying it to
specific linear programming problem situations, in this Learning Activity the learner will be
encouraged to also peform those Simplex computational procedures using appropriate
software that is accessible to the learner.

Learning Activities
A woman, baby at the back, pondering over the easiest way of crossing a fast flowing river.
The Simplex Method
The methods of linear programming are based on the theories of matrices and finite-
dimensional vector spaces and the simplex method for solving linear programming
problems is built around the basic solution of a set of simultaneous linear equations. For
more details about the basic solutions and vector spaces read your module on Linear
algebra or try the well illustrated example by Richard S. Barr on the following link:
When you are solving linear programming problems using the Simplex algorithm the main
steps to remember are:
1. Set up the intial simplex tableau
2. Locate the pivot of the tableau
3. If the pivot is 1, then go to step 4 ; otherwise divide the pivotal row by the pivot to get
1 in the pivotal position
4. Convert the remaining entries of the pivotal column into zeros by using the
elementary row operations (which you have learnt in the linear algebra module)
5. Repeat steps 2 to 4 until a tableau with non-negative indicators is obtained. This is
the final tableau we need in this algorithm and it is called the terminal simplex.
6. The optimal solution and the maximum value of the objective function can be read
from that terminal tableau.
Let us go through the steps using the following example adapted from by Koshy (1979)
Maximize
f ( x, y) = 170 x + 225 y
Subject to:
x + y �300
2 x + 3 y �720
x �0, y �0
Step 1 Form the initial simplex tableau

Let us introduce two slack variables s 1 ≥ 0 and s2 ≥ 0 then the first inqualities can be
written as equalities
x + y + s1 = 300
2 x + 3 y + s2 = 720
We can now have the system of equations
x + y + s1 = 300
2x + 3y + s2 = 720
-170x - 225y + f = 0
Forming the initial simplex tableau
x y s1 s2 f rhs
1 1 1 0 0 300
720
2 3 0 1 0
-170 -225 0 0 1 0
Step 2 Locate the pivot of the tableau
x y s1 s2 f rhs
1 1 1 0 0 300
720
2 3 0 1 0 Pivot row
-170 -225 0 0 1 0 Pivot
Pivot column
Step 3 Since the pivot is 3, we divide the second row by 3
x y s1 s2 f rhs
1 1 1 0 0 300
2/3 1 0 1/3 0 240
-170 -225 0 0 1 0
Step 4 Adding -1 times the second to the first and 225 times the second row to the third,
we get
x y s1 s2 f rhs
1/3 0 0 -1/3 0 60
2/3 Template1
Module Development 0 1/3 0 240 38
-20 0 0 75 1 54000
New pivot
Step 5 The new pivot is 1/3; we divide the first row by 1/3 and then add -2/3 times the
resulting first row to the second and 20 times the first row to the third row.
x y s1 s2 f rhs
1 0 3 -1 0 180
0 1 -2 1 0 120
0 0 60 55 1 57600
Maximum value
Step 6 This tableau is the augmented matrix of the system
x + 3s1 - s2 = 180
y - 2s1 + s2 = 120
60s1 + 95s2 = 57 600
That is
x = 180 - 3s1 + s2
y = 120 + 2 s1 - s2
f = 57600 - 60 s1 - 95s2
f has maximum value of 57 600 when s1 = 0 and s2 = 0, and this yields x =180 and y=120
Note
The values of f, x and y can be easily read from the terminal simplex, thus the optimal
solution and the maximum value of the objective function can be read from this terminal
tableau.

The Big M method
In the case where we cannot find an initial basic feasible solution you could to use the Big
M method where you add a non-negative large integer M > 0.
To illustrate the use of the Big M method, let us follow the example adapted from Wagner
(1975)
Maximize f ( x1 , x2 ) = -3x1 - 2 x2
Subject to
x1 + x2 = 10
x1 �4
x1 �0, x2 �0
Then after adding a surplus variable x3 in the inequality above, you can write the model as
x0 + 3 x1 + 2 x2 = 0
x1 + x2 = 10
x1 - x3 = 4
Where f ( x1 , x2 ) = -3x1 - 2 x2 has been written as x0 + 3 x1 + 2 x2 = 0 where x0 = f ( x1 , x2 )
Next, introduce artificial variables y1 and y2 , and let M =10 as our large integer, for
instance, giving
x0 + 3 x1 + 2 x2 + 10 y1 + 10 y2 = 0
x1 + x2 + y1 = 10
x1 - x3 + y2 = 4
To initiate the algorithm, you have to subtract (M =10) times row 2 and (M =10) times row 3
from row 1 to eliminate y1 and y2 :
x0 - 17 x1 - 8 x2 + 10 x3 = -140
x1 + x2 + y1 = 10
x1 - x3 + y2 = 4
From here you can then proceed using the Simplex method.
You can verify that x1 = 4 and x2 = 6 are optimal.
Further reading
Follow this link.

The Simplex Method:
http://www-fp.mcs.anl.gov/otc/Guide/CaseStudies/simplex/feasible.html

Further Activity
The following link http://www-fp.mcs.anl.gov/otc/Guide/CaseStudies/simplex/feasible.html

explains the simplex method in a user friendly way, go through the activities in it and then
answer the following questions:
1. State the nature of a standard maximising problem

2. What happens in the initial simplex tableau stage?
Exercise 4
The following tableaus tables were obtained in the course of solving linear program
non-negative variables x1 and x2 and 2 inquality constraints, the objective function, z is
maximisation. Slack variables s1 and s2 were added. In each case, indicate whether the
linear program:
(i) is unbounded
(ii) has a unique optimum solution
(iii) has an alternate optimum solution
(iv) is degenerate (in this case, indicate whether any of the above holds).
a)
z x1 x2 s1 s2 rhs
1 0 3 2 0 20
0 1 -2 -1 0 4
0 0 -1 0 1 2
b)
z x1 x2 s1 s2 rhs
1 0 -1 0 2 20
0 0 0 1 -2 5
0 1 -2 0 3 6
c)
z x1 x2 s1 s2 rhs
1 2 0 0 1 8
0 3 1 0 -2 4
0 -2 0 1 1 0
d)
z x1 x2 s1 s2 rhs
1 0 0 2 0 5
0 0 -1 1 1 4
0 1 1 -1 0 4

Hint: Answers to Excercise 4 can easily be obtained by refering to the definitions of
“unboundedness,” “unique optimum solution,” “alternate optimum solution,” and
“degeneracy.”
Formulation of the Dual Problem
Every linear programming has another linear programming called its dual, which shares the
same data and is derived through rational arguments. In this context the original linear
programming is called the primal linear programming. Variables in the dual problem are
different from those in the primal, each dual variable is associated with a primal constraint,
it is the marginal value or Langrange multiplier corresponding to that constraint.
The problem faced by the Production Manager as the Optimist and the Comptroller as the
Persimist in Linear Programming: Foundations and Extensions by Robert J. Vanderbei,
chapter 1, identifies a situation for a dual problem. The Resource Allocation Problem in
chapter 5 pages 73-78 of same book is a good illustration of the formulation of a dual
problem from its primal problem.
Example [adapted from Arsham (2007)

[http://home.ubalt.edu/ntsbarsh/opre640a/partVIII.htm)]
Let us illustrate the formulation of the dual problem using a tailor’s problem
Uncontrollable Inputs
trousers shirts available
labour 3 2 50
raw material 4 3 60
net income 20 15
and its linear programming is:
Maximize
f ( x1 , x2 ) = 20 x1 + 15x2
subject to:
3 x1 + 2 x2 �50 labour constraint
4 x1 + 3 x2 �60 material constraint
x1 �0, x2 �0
Where x1 and x2 are the number of trousers and shirts to make.
Supose the tailor wishes to buy insurance for his net income. Let z 1 = the dollar amount
payable to the tailor for every labour hour lost due to unforseen problems, and z 2 = the
dollar amount payable to the tailor for every raw material unit lost.

The insurance broker tries to minimize the total amount of $(50z 1 +60z2) payable to the
tailor by the insurance company. However, the tailor will set the constraints (conditions) by
insisting that the insurance company cover all his loss (that is, his net income) since he
cannot make the products. Therefore the broker’s problem is:
Minimize
f ( z1 , z2 ) = 50 z1 + 60 z2
subject to:
3 z1 + 4 z2 �20 Net income from a trouser
2 z1 + 3 z2 �15 Net income from a shirt
z1 �0, z2 �0
If we work out the solutions of this problem you will see that the insurance company’s
problem is closely related to the tailor’s problem.
Reading
Now read the following text:

Linear Programming: Foundations and Extensions by Robert J. Vanderbei, Chapter 5,
pages 55 – 79. [www.princeton.edu/~rvdb/LPbook/onlinebook.pdf]
Dual Theorem
a) In the event that both primal and dual problems possess feasible solutions, then the
primal problem has an optimal solution x *j for j= 1, 2, …., n, the dual problem has an optimal
solution y*i , for i = 1, 2, ….., m, and
n m
�c j x*j = �bi yi*

j =1 i =1
b) If either the primal or dual problem posses a feasible solution with a finite optimal
objective-function value, then the other problem possesses a feasible solution with the
optimal objective-function value.
Note
The duality relationship can be summarised as follows:
Primal (Maximize) Dual (Minimize)

Objective function Right-hand side
Right-hand side Objective function
jth column of coefficient jth row of coefficient

ith row of coefficients ith column of coefficient
th
j variable nonnegative jth relation an inequality (≥ )
jth variable unrestricted in jth relation an equality
sign
ith relation an inequality (≤ ) ith variable nonegative
ith relation an equality
ith variable unrestricted in sign
Example [adapted from Wagner (1975)]
Here is a didactic example to illustrate the formation of a dual problem from its primal.
Maximize
4 x1 + 5 x2 + 9 x3 + 11x4
Subject to:
x1 + x2 + x3 + x4 �15
7 x1 + 5 x2 + 3 x3 + 2 x4 �120
3 x1 + 5 x2 + 10 x3 + 15 x4 �100
x j �0
For j =1, 2, 3, 4
The dual problem can easily be obtained using the duality relationships in the table above
as:
Minimize
15 y1 + 120 y2 + 100 y3
Subject to
y1 + 7 y2 + 3 y3 �4
y1 + 5 y2 + 5 y3 �5
y1 + 3 y2 + 10 y3 �9
y1 + 2 y2 + 15 y3 �11
y1 �0, y2 �0, y3 �0

Exercise 5
1. In your own words explain what you understand by a dual problem in linear
programming.
2. State the dual theorem and then prove it.
3. Do questions 5.1 , 5.2 and 5.16 on page 79 of Linear Programming : Foundations
and Extensions by Vanderbei.
Further Reading
You can read the following link for more examples of formulating a dual problem in linear
programming:
http://www.maths.abdn.ac.uk/~igc/tch/mx3503/notes/node66.html
Duality in linear programming, Optimality Conditions

http://www-personal.umich.edu/~murty/310/310-2slides8.pdf -
Duality and OptimalityConditions

http://www.caam.rice.edu/~yzhang/caam378/Notes/chap5.pdf
Question 1
Read the Tailor’s Problem or the Resource Allocation Problem and then explain them in
your words to your colleague.
Question 2
Weak and strong duality theorems combined together.
Question 3
For question 5.1 we are given the problem as follows
Maximize x1 - 2 x2
Subject to
x1 + 2 x2 - x3 + x4 �0
4 x1 + 3 x2 + 4 x3 - 2 x4 �3
- x1 - x2 + 2 x3 + x4 = 1
x2 �0, x3 �0
We rewrite the problem as follows

Maximize x1 - 2 x2
Subject to
- x1 - 2 x2 + x3 - x4 �0
4 x1 + 3 x2 + 4 x3 - 2 x4 �3
- x1 - x2 + 2 x3 + x4 �1
x1 + x2 - 2 x3 - x4 �-1
The dual of this problem is therefore:
Minimize 3y2 + y3 - y4
Subject to
- y1 + 4 y2 - y3 + y4 �1
2 y1 + 3 y2 - y3 + y4 �-2
y1 + 4 y2 + 2 y3 - 2 y4 �0
- y1 - 2 y2 + y3 - y4 �0
5.2 To solve this problem refer to theorem 5.2 on page 60 of Linear Programming:
Foundations and Extensions. Find the dual of the problem as illustrated in the problem
above.

Learning activity # 5: Sensitivity Analysis
Specific Learning Objective
By the end of the activity the learner:

o should be able to carry out a sensitivity analysis of a Linear Programming solution.
o Should be able to explain the reasons for carrying a sensitity analysis.
In this activity we are going to answer some of the post-optimality questions such as
1) If the profit contribution of a particular basic activity decreases, does the current
solution remain optimal?
2) What happens if resource availability is curtailed?
3) What happens if a new activity is added?
4) How far the input parameter values vary without causing violent changes in an
optimal solution.
We are also going to do marginal and parametric analysis and learn what is involved in
each of the analyses. Learners should to do some sensitivity analysis on some optimal
solutions they have found in activity 4.
Parametric analysis
Marginal analysis
List of Required Readings:


http://www.jr2.ox.ac.uk/bandolier/booth/glossary/sensanal.html
http://pespmc1.vub.ac.be/ASC/SENSIT_ANALY.html

http://en.wikipedia.org/wiki/sensitivity_analysis
Management Science: Linear Programming Notes

http://www.strathcona.bham.ac.uk/Pdfs/1%20management%20Course%20Year
%202/LINEAR%20PROGRAMM%20NOTES.PDF
The learner will read on the definition of sensitivity analysis and why it is carried out.
They will refer mainly to the recommended text: Linear Programming: Foundations and
Extensions by Robert J. Vanderbei, which explains the purpose of sensitivity analysis
well. The learner is encouraged to discuss with colleagues and perfom some practical
computaions associated with sensitivity analysis using appropriate (and accessible)
software packages. Interpreting the results of running a sensitivity analysis is also an
important exercise at it has a direct bearing on eventual decision making processes. A
number of links that discuss this issue of sensitivity analyses and provide examples of
software based computer runs are supplied for the learner in this Learning Activity. The
learner will increase her/his appreciation and understanding if he/she also checks on
these links.
Learning Activities
Consider the problem of the production manager that we discussed in

Activity One. Even under normal business situations, various factors
influence the production, or, eventually, profit levels. Suppose the
manager has been advised by her planning committee that has solved
the associated linear progamming problem to maintain purchase orders
of some input materials and supplies to some warehouses at certain
specific levels. Those levels would be corresponding to the optimal
solution (production or profits) under those circumstances. Now
consider the case of an inflationary environment in a country with a
poorly performing economy. The influencing factors will obviously vary in their impact to
the production process depending on the availability and costs of the input materials and
output distribution costs. The variations are often non-uniform across factors and
unpredicable in both time period and size. Under those circumstances, the manager would
be very interested in knowing how much changes in individual or combinations of input
factors impact on the current optimum solution. This is where sensitivity analysis, which
looks at that problem, becomes useful. Technically, it analyses changes in the coefficients
of the objective function (the unit costs) or the right hand sides of the constraints (the
requirements) affect the optimal solution.

Discussion [adapted from Trick (2007)]
Q. Where do I start in sensitivity analysis?

A. If we carry out the simplex method iterations on the initial tableau below
f x y s1 s2 RHS
0 3 1 1 0 300
0 1 720
0 2 3
1 -170 -225 0 0 0
We get we get the final tableau called terminal tableau below
basis (identity matrix)
x y f s1 s2 RHS
1 0 0 3 -1 180
0 1 0 -2 1 120
0 0 1 60 55 57600
Row 0
It has all non-negative values in Row 0 (which we will often refer to as the cost row), all
non-negative right-hand-side values, and a basis matrix embedded.
Sensitivity analysis starts at this point to determine the effects of small changes in the
optimal solution. We will try to determine how that change affects the final tableau, and try
to re-form the final tableau accordingly.
Changing the values of the decision variables
The first change we will consider is that of changing a cost value, that is, changing the
values of x and y by a small amount  in the original problem. If we carry out the calculation
to find the optimal solution for the changed value of x in the original problem, we will get the
same final tableau as before, except that the corresponding cost entry would be  lower
(this is because we never did anything except add or subtract scalar multiples of Rows 1
through m to other rows; we never added or subtracted Row 0 to other rows).

f x y s1 s2 RHS
0 3+ 1 1 0 300
0 2 3 0 1 720
1 -170 -225 0 0 0
Carry out the calculation and then confirm the statement above.
Right-hand-side changes
For these types of changes, we concentrate on maximization problems with all ≤

constraints. Other cases are handled similarly.
Consider the following problem:
Maximize f ( x, y ) = 4 x + 5 y
subject to:
2 x + 3 y �12
x + y �5
x, y �0
The optimal tableau, after adding slacks s1 and s2 is
z x y s1 s2 RHS
1 0 0 1 2 22
0 0 1 1 -2 2
0 1 0 -1 3 3
Now suppose we change the amount of the right-hand-side from 12 to (12+) in the first
constraint, the tableau changes to:
z x y s1 s2 RHS
1 0 0 1 2 22 +
0 0 1 1 -2 2+
3 -
0 1 0 -1 3
This represents an optimal tableau as long as the right hand side is all non-negative. In
other words, we need  between -2 and 3 in order for the basis not to change. For any  in
that range, the optimal objective will be (22+). For example, with  equals 2, the new
objective is 24 with y = 4 and x = 1.
Similarly, if we change the right-hand-side of the second constraint from 5 to (5 + ) in the

original formulation, we get an objective of (22 + 2) in the final tableau, as long as
-1 ≤  ≤ 1

Basically sensitivity analysis tests what happens to the optimal solution if there are changes
in the supply of raw material or market values. It highlights changes in variables which
affect the optimal solution and those which do not.
Reading and Discussion
Read and discuss with a colleague
Linear Programming: Foundations and Extensions by Robert J. Vanderbei Chapter 7 pages

111-120: www.princeton.edu/~rvdb/LPbook/onlinebook.pdf
http://en.wikipedia.org/wiki/sensitivity_analysis
Use accessible software packages (such as Pivot Tool) to perform some sensitivity
analyses on any 2 optimum solutions to an linear programming problem that you have
previously obtained. Interpret the results of the analyses you have done and discuss your
interpretations with a colleague.
Pivot Tool: http://campuscgi.princeton.edu/~rvdb/JAVA/network/nettool/netsimp.html
Exercise 6
Linear Programming: Foundations and Extensions by Robert J. Vanderbei Chapter 7 page

120 # 7.1 and 7.2. www.princeton.edu/~rvdb/LPbook/onlinebook.pdf
Further Activity
Optimization Methods Lecture 7 Sensitivity Analysis

http://ocw.mit.edu/NR/rdonlyres/Sloan-School-of-Management/15-093Fall-2004/8842223F-
9C8A-4B9E-83F8-DD28DE270140/0/lecture07.pdf
Management Science: Linear Programming Notes

Page 449 of Foundations and Extensions by Robert J. Vanderbei

11. COMPILED LIST OF ALL KEY CONCEPTS (GLOSSARY)
Argumented matrix is formed by the coefficient of the dictionary of an linear programming

problem
Artificial Variable
A variable added to a linear program in phase 1 to aid finding a feasible solution.
Basic Solution
x of (Ax = b) is a basic solution if the n components of x can be partitioned into m "basic"
and n-m "non-basic" variables in such a way that:
 the m columns of A corresponding to the basic variables form a nonsingular basis
and,
 the value of each "non-basic" variable is 0.
The constraint matrix A has m rows (constraints) and n columns (variables).
Basis
The set of basic variables.
Basic Variables
A variable in the basic solution (value is not 0).
Constraints
A set of equalities and inequalities that the feasible solution must satisfy.
Convex, is a set of points S in the n-dim space, such that if the line segment connecting
any two points X 1 , X 2  S , belongs completely in S. Mathematical definition of convexity is
expressed as:
S convex � "X 1 , X 2 �S2 , "x �(0,1) : (1 - x) X 1 + xX 2 �S
Convex sets
Nonconvex convex
Cycling is when a sequence of pivots goes through the same tableaux and repeats itself
indefinitely.
Degeneracy occurs when some basic variable is at one of its bound values (canonically
zero). Without any given qualification, a (basic) solution is degenerate if one or more of its
basic values is zero (the canonical lower bound).

Dictionary is the systems of equations of the objective function and the constraint
equations after introducing slack variables
Dual degenerate is a solution where one of its non-basic variables has zero reduced cost.
In general, a solution is degenerate if it is not strictly complementary.
Duality:
Weak duality is say, if x, p are feasible solutions to primal and dual problems
respectively, then b�
p �c�
x.
Strong duality is when the primal problem has an optimal solution and the dual
problem has also an optimal solution and c� x* = b�
p * [where x* and p* are optimal
solutions of the primal and the dual problem respectively.]
Feasible. A point is feasible if it satisfies all constraints. The feasible region (or feasibility
region) is the set of all feasible points. A mathematical program is feasible if its feasible
region is not empty.
Primal feasiblity means having a feasible solution of the linear programming

problem
Dual feasiblity means having a feasible solution of the dual linear programming
problem
Feasible Solution
A solution vector, x, which satisfies the constraints.
a
X:aX>b
X X:aX=b
X:aX<b

Half-planes: the feasible region of a linear inequality

For example the half-planes of the inequality a1 X 1 + a 2 X 2  b are the planes on the left


and right side of the equality a1 X 1 + a 2 X 2  b .

given a linear constraint:

 
a1 X 1 + a 2 X 2 + ... + a n X n  = b ………….**

 
T
X 0 = X 01 , X 02 ,..., X 0 n
and a point satisfying Equation (**) as equality, we can perceive the solution space of
the equation:
a1 X 1 + a2 X 2 + ..... + an X n = b
Hyperplane. Is solution space of an n-var linear equation of the general form:
a1 X 1 + a2 X 2 + ..... + an X n = b
Isoprofit lines are lines for the objective function that are drawn to locate the optimal value
when using the geometric approach.
Linear programming problem is a set of (linear) inequalities (with a solution set S) and a
(linear) function (often cost or profit) whose value (within S) is to be maximized or
minimized.
Marginal analysis is concerned with the effects of small perturbations, maybe measurable
by derivatives.
Nonbasic Variables
A variable not in the basic solution (value = 0).
Objective Function
The function that is either being minimized or maximized. For example, it may represent the
cost that you are trying to minimize.
Operation research, sometimes known as management science, refers to the constant

scientific monitoring of an organisation’s ongoing activites, focussing on decision and
control problems that affect daily operations of the organisation. It involves the
representation of real world situations by mathematical models, together with the use of
quantitative methods (algorithms) for solving such models with a view to optimizing.
Optimal Solution (vector)

A vector x which is both feasible (satisfying the constraints) and optimal (obtaining the
largest or smallest objective value).

Optimal Solution is the maximun or minimun feaseble solution which is obtained at final
stage of the simplex method algorithm process.
Parametric analysis is concerned with larger changes in parameter values that affect the
data in the mathematical program, such as a cost coefficient or resource limit.
Phase I & Phase II

Phase I of a mathematical program is finding a feasible solution, and Phase II is entered
with a feasible solution to find an optimal solution.
Pivot. This is the algebra associated with an iteration of Gauss-Jordan elimination, using
the forward transformation. (refer to illustration below)
Pivot row
Pivot
Pivot column
Polytope is solution space (feasible region) of an n-var linear programming is geometrically

defined by the intersection of a number of half-spaces and/or hyperplanes equal to the
linear programming constraints, including the sign restrictions.
Polyhedron is a bounded polytope

Pricing. This is a tactic in the simplex method, by which each variable is evaluated for its
potential to improve the value of the objective function.
Primal degenerate is a pivot where the associated basic solution (x) does not change
(i.e., the non-basic variable enters the basis, but its level remains at the same bound
value, in which case no basic variable changes level).
Sensitivity analysis. The concern with how the solution changes if some changes are
made in either the data or in some of the solution values (by fixing their value).
Simplex method. An algorithm invented to solve a linear program by progressing from one
extreme point of the feasible polyhedron to an adjacent one. The method is an algorithm
strategy, where some of the tactics include pricing and pivot selection.
Simplex (pl. simplices). {x in R: Sum{xj} = 1}. For n=1, this is a point (x=1). For n=2, this is
a line segment, joining points (1,0) and (0,1). For n=3, this is a triangle, joining the vertices
(1,0,0), (0,1,0), and (0,0,1). This is sometimes called an n-simplex, denoted by Sn (note its
dimension is n-1). The open simplex excludes the axes: {x in Sn: x > 0}.

Slack variable. In an inequality constraint of the form g(x) ≤ b, the slack is b-g(x), which is
designated by the slack variable, s. Then, the original constraint is equivalent to the
defining equation, g(x) + s = b, plus s >= 0.
Slack Variable
A variable added to the problem to eliminate less-than constraints.
Standard Maximizing Problem is a linear programming problem

which satisfies all of the following 4 conditions:
1) The objective function is to be maximized.
2) All inequalities are of the type.
3) All right hand constants are non-negative.
4) All variables are non-negative.
[A Non-Standard Problem is simply a problem which is not standard, and hence fails to
satisfy at least one of [1] through [4] above.]
Surplus Variable
A variable added to the problem to eliminate greater-than constraints.
Tableau (pl. tableaux). A detached coefficient form of a system of equations, which can
change from x +Ay = b to x'+A'y' = b'. The primes denote changes caused by multiplying
the first equation system by the basis inverse (a sequence of pivots in the simplex
method).
Unbounded Solution
For some linear programs it is possible to make the objective arbitrarily small (without
bound). Such an linear programming is said to have an unbounded solution.

12. COMPILED LIST OF COMPULSORY READINGS (Complete reference +
abstract/rationale)
Linear Programming: Foundations and Extensions by Robert J. Vanderbei

All learners are required to read this book whenever he/she is starting a new unit or activity.
All other links are to help the learner to understand more about Linear Programming
problems

13. COMPILED LIST OF (OPTIONAL) MULTIMEDIA RESOURCES (Complete reference
+ abstract/rationale)
Reading # 1: Wolfram MathWorld (visited 03.11.06)
Complete reference : http://mathworld.wolfram.com
Abstract : Wolfram MathWorld is a specialised on-line mathematical encyclopedia.
Rationale: It provides the most detailed references to any mathematical topic. Students
should start by using the search facility for the module title. This will find a major article. At
any point students should search for key words that they need to understand. The entry
should be studied carefully and thoroughly.
Reading # 2: Wikipedia (visited 03.11.06)
Complete reference : http://en.wikipedia.org/wiki
Abstract : Wikipedia is an on-line encyclopedia. It is written by its own readers. It is

extremely up-to-date as entries are contunally revised. Also, it has proved to be extremely
accurate. The mathematics entries are very detailed.
Rationale: Students should use wikipedia in the same way as MathWorld. However, the
entries may be shorter and a little easier to use in the first instance. Thy will, however, not
be so detailed.
Reading # 3: MacTutor History of Mathematics (visited 03.11.06)
Complete reference :
http://www-history.mcs.standrews.ac.uk/Indexes
Abstract : The MacTutor Archive is the most comprehensive history of mathematics on the
internet. The resources are organsied by historical characters and by historical themes.
Rationale: Students should search the MacTutor archive for key words in the topics they
are studying (or by the module title itself). It is important to get an overview of where the
mathematics being studied fits in to the hostory of mathematics. When the student
completes the course and is teaching high school mathematics, the characters in the
history of mathematics will bring the subject to life for their students. Particularly, the role of
women in the history of mathematics should be studied to help students understand the
difficulties women have faced while still making an important contribution.. Equally, the role
of the African continent should be studied to share with students in schools: notably the
earliest number counting devices (e.g. the Ishango bone) and the role of Egyptian
mathematics should be studied.

The following sites enable you to practice the Linear Programming pivoting without
having to do the arithmetic!
1. Pivot Tool: http://www.sor.princeton.edu/~rvdb/JAVA/pivot/advanced.html
2. TUTOR: http://www.tutor.ms.unimelb.edu.au/
3. The Simplex Place: http://www.ifors.org/tutorial
4. Simplex Machine: http://www.ms.unimelb.edu.au/~moshe/lp/simplex7.html
5. linear programming-Explorer: http://www.maths.ed.ac.uk/linear programming-Explorer
6. Taha, H. (2003, 7th ed). TORA Solver: Operation Research: An Introduction
14. COMPILED LIST OF USEFUL LINKS (Name of page/site; URL;

Description/Rationale)
A. Linear Programming Formulation

B. Linear Programming Formulation

C. Linear Programming Formulation

http://home.ubalt.edu/ntsbarsh/opre640a/partVIII.htm
D. An Introduction to Linear Programming and the Simplex Algorithm by Spyros

Reveliotis http://www2.isye.gatech.edu/~spyros/linear programming/linear
programming.html
E. (1) Linear Programming

E. (2) Linear Programming Formulation

www.people.brunel.ac.uk/~mastjjb/jeb/lp.html
F. Linear Programming: A Geometric Approach page 171-176

G. Optimality Conditions for Constrained Optimization

http://ocw.mit.edu/NR/rdonlyres/Sloan-school-of-Management/15-
084JSpring2004/7240EF84-B20D-419F-B1C0-2DAF3277F5C4/0/lec6_constr_opt.pdf

Mathematical approach of checking conditions of optimality
H. Optimality Conditions
http://www.math.mtu.edu/%7Emsgocken/ma5630spring2003/lectures/lag1/lag1/node1.html
I. Alternate Optimal Solutions, Degeneracy, Unboudedness, Infeasibility

http://mat.gsia.cmu.edu/QUANT/notes/node63.html#SECTION00830000000000000000
Degeneracy and unboundedness are well explained with aid of simple examples.
J. The Fundamental Theorem of Linear Programming

http://www.it.uu.se/edu/course/homepage/opt1/ht06/Lectures/fundamentat_thm-2up.pdf
http://www.maths.abdn.ac.uk/~igc/tch/mx3503/notes/node67.htm
These links provides theorems of linear programming and the relevant proofs
K. Simplex Method - Big M

N. Simplex Method – Big M

http://www.computing.dcu.ie/~lkillen/teach/CA427Simplexbigmexample.pdf
The algorithm of the B-M method is examplified
O. Dual Problem: Construction and Its Meaning

This link takes the student step by step in the construction of the dual problem. The
association between the primal and dual problem is well illustrated with good examples.
P. Sensitivity Analysis for Linear Programming

This link gives a good exaplanation on “Why sensetivity analysis?”.
Q. Deterministic Modeling:Linear Optimization with Applications

http://www.mirrorservice.org/sites/home.ubalt.edu/ntsbarsh/Business-
stat/opre/partVIII.htm#ronline
This link explains the flow of the linear programming problem from formulation down to
sensetivity analysis. One problem is taken from formulation stage to sensetivity analysis
stage. Students are incouraged to visit this cite whenever they have problems of
understanding some of the linear programming technical terms and processes.
R. Duality in Linear Programming

http://www.web.mit.edu/15.053/www/AMP-Chapter-0 4.pdf
Explains the duality thoerem and duality properties

S. Basic Solutions
T. Linear Programming: Geometric Approach

Last visited date 16-02-07

15. SYNTHESIS OF THE MODULE
In this module you have been introduced to an area of mathematics called Linear
Programming that plays a central role in a wider field called Operations Research.
Operations Research deals with situations where one is interested in optimising (minimising
or maximising) a specific quantity that is often framed as a function of several influencing
factors (variables) with an overall goal of making decisions related to the behaviour of the
situation of interest. Linear programming offers a technique for resolving a class of
problematic situations of that nature.
To develop a feel for mathematics related to Linear Programming, you have learned the
following main ideas:
 Formulation of a linear programming problem. [Identify constraints of the situation

and the objective function to be optimised, then construct a system of inequality
relations involving them].
 Geometrical interpretation of a solution of a linear programming problem. [The

optimal solution is located at one corners where boundary lines of the feasible region
intersect, or along one of the boundary lines]
Optimality conditions for the objective function of a linear programming problem. [In the
step-by-step procedure for eliminating basic variables in the initial tableau constructed
from the objective function and its constraints, an indication that an optimal value has
been reached is that for a maximization problem, the coefficients of the basic feasible
variables in the objective function should be negative, meaning that the variables can no
longer contribute to increasing the value of the objective function. Similarly for a
minimization problem]
 Algebraic interpretation of the solution to a linear programming problem. [This is the

set of remaining nonbasic feasible variables after all basic feasible variables have
been eliminated in the step by step procedure from the initial tableau matrix
constructed from the objective function and its constraints.]
 Sensitivity analysis. [a solution is stable if small changes introduced into its
constraint variables do not produce large changes in the value of the objective
function associated with the solution].
In short, the module looked at formulation of the linear programming problem and its
solution using the geometric approach (graphical method) and the algebraic method
(simplex method). Stability of the solutions was checked using sensitivity analysis.
The authors of the module believe that the learner who has successfully developed a fairly
good understanding of the main concepts and mastery of the associated skills (at least as
indicated by the assessment activities in-built in the module) will not only be able to teach
the introductory levels of linear programming concepts at secondary school level, but will
also be in a position to proceed further within the field of Operations Research.

16. FORMATIVE AND SUMMATIVE EVALUATION
In this module evaluation of student learning comes in 2 components: the formative

(continuous or coursework) and the summative (cumulative) components. The
formative component is prompted by the various exercises appearing in all the learning
activities, with the evaluating process being excecuted by the student him/herself, that is
to say, they are self-evaluation learning tasks or exercises for student’s direct benefit.
The intent of that evaluation is to give the student a sense of the progress the student is
making in learning the material as he/she moves through the module. That evaluation
can also be recorded by the student [see provision for that on the student records Excel
spreadsheet in section 18 below.] Recording his/her progress is important for him/her to
build his/her learning profile for the material in the module and keep track of the
variations in levels of understanding or difficulty as the student moves from learning
activity to learning activity. This formative self-evaluation is best done using qualitative
categories such as poor, satisfactory, good, and very good. It is expected that the
student will conduct the self-evaluation process in as honest a manner as possible
because dishonest self-assessing will only serve to undermine the student to his/her
detriment.
The second component of the evaluation, the summative evaluation, is accomplished

through a formal written test consisting of 10 items that can be completed in 4 hours on
the average. The test will be administered in person by the course program providers at
a designated learning centre, where the students will also sit for the test in person.
Students will require some graph paper and graph plotting accessories, and a standard
scientific or programmable calculator, whichever they are comfortable with using. The
test will be marked or scored and evaluated by the course program providers and the
student’s performance in the test will be taken to represent the amount of cumulative
learning and understanding of the material in the module that the student has attained.
That assessed performance will in turn provide the final course mark or grade obtained
by the student for this course or module on linear programming.
A genuine attempt has been made to select or design test items that assess the extent
the student has grasped the essential features and characteristics of the general linear
programming problem described in Unit 1 box of the Flow of Learning chart in the earlier
overview section [see section 7], and developed associated competencies described in
the Unit 2 box of the same chart. A description of the test items and the related learning
objective each item is intended to assess, and the list of the actual test items now follow
below.
****The answers to the test items are also supplied at the end of the test item listing.****
,
Acknowledgement: [Some of the questions have been adaptated from selected exercises
in Wagner, H. M. (1975, 2nd ed). Principles of Principles of Operations Research.
Englewood Cliffs, New Jersey: Prentice-Hall].
Question 1 tests the student’s understanding of a linear programming problem situation

through the student’s ability to formulate the linear programming problem properly (i.e.
writing the problem in mathematical language).

Questions 2 and 3 test the student’s capability to write the problem in mathematical
language and solve it using the graphical (geometric) method. Graph paper is relevant
for accuracy of diagrams.
Question 4 tests the student’s capability to solve the linear programming problem using
the Simplex Method.
Questions 5 and 6 test the student’s capability to solve the linear programming problem
using the Big- M Method followed by the Simplex Method. The student should know
when it is appropriate to use the Big- M method.
Question 7 checks if the student can identify the primal and dual forms in a linear
programming situation and formulate the dual of a given primary linear programming
problem.
Question 8 tests the student’s capability to investigate the feasibility and boundedness
characteristics of a linear programming problem.
Question 9 tests the student’s capability to solve the linear programming problem using
the Dual Simplex method.
Question 10 tests the student’s capability to solve a linear programming problem using
the Simplex Method and then carry out a sensitivity analysis on the solution obtained.

Question 1
The Mufoya Bicycle Company produces two kind of bicycles by hand: Mountain Bike and
Street Racers. Mufoya wishes to find out the rate at which each type of bicycle should be
produced in order to maximize the profits on the sales of the bikes. Mufoya assumes that
the company can sell all the bicycles produced. The physical data on the production
process is available from the company engineer. A different team produces each kind of
bicycle, each team has a different maximum production rate: 2 mountain bikes per day and
3 racers per day respectively. Producing a bicycle of either type requires the same amount
of time on the metal finishing machine (a production bottle-neck), and this machine can
process at most a total of 4 bicycles per day, of either type. The company accountant
esimates that mountain bikes are currently generating a profit of $15 per bicycle and racers
are giving a profit of about $10 per bicycle. Formulate the Linear Programming Problem.
Questions 2 and 3
A manufacurer of electronic instruments produces two types of timers: a standard and a

precision model with net profits of $2 and $3 respectively. They are similar in design, each
taking about the same amount of time to assemble. Let us suppose that the manufacturer
wishes to maximize his net profit each day subject to the availability of resources and
marketing considerations. Let’s consider that his work force can produce no more than 50
instruments per day. Furthermore, suppose there are 4 main components in short supply
(a,b,c and d), and that they are used in different quantities for the types of timers as shown
below.
The Stock Constraints
Component Stock Standard (x) Precision (y)

a 220 4 2
b 160 2 4
c 370 2 10
d 300 5 6
That is to say each of the x standard timers uses 4 of component a, and each of the y
precision ones uses 2, and the factory cannot use more than what is in stock.
[A] Formulate the Linear Programming Problem.
[B] (i) Find the feasible region of the Linear Programming problem using the
geometrical method.
(ii) Find the value of the maximum profit.
Note: You need GRAPH PAPER for this question.

Question 4
Use the Simplex method to solve the following linear programming problem
Maximize 4 x1 + 5 x2 + 9 x3 + 11x4
Subject to
x1 + x2 + x3 + x4 �15
7 x1 + 5 x2 + 3x3 + 2 x4 �120
3 x1 + 5 x2 + 10 x3 + 15 x4 �100
x1 �0 , x2 �0 , x3 �0
Question 5
(a) Explain fully when the Big M method is used.
(b) Maximize 3x1 + 4x2
Subject to
2x1 + x2 ≤ 600
x1 + x2 ≤ 225
5x1 + 4x2 ≤ 1000
x1 + 2x2 ≥ 150
x1, x2 ≥ 0
Question 6
Maximize -3x1 - 2 x2
Subject to
x1 + x2 = 10
x1 �4
x1 �0, x2 �0
Question 7
(a) Explain what is meant by the primal and dual of a linear programming problem.
(b) Formulate the dual of the linear programming problem below
Maximize 4 x1 + 5 x2 + 9 x3 + 11x4

Subject to
x1 + x2 + x3 + x4 �15
7 x1 + 5 x2 + 3x3 + 2 x4 �120
3 x1 + 5 x2 + 10 x3 + 15 x4 �100
x1 �0 , x2 �0 , x3 �0
Question 8
Investigate the feasibility of the problem below
a) Maximize 5x1 + 4x2
subject to
x1 + x2 ≤ 2
−2x1 − 2x2 ≤ −9
x1, x2 ≤ 0.
b) Determine the boundedness of the problem below
Maximize x1 − 4x2
subject to
−2x1 + x2 ≤ −1
−x1 − 2x2 ≤ −2
x1, x2 ≥ 0.
Question 9
Minimize 2x1 + x3
Subject to
x1+ x2 - x3 ≥ 5
x1 - 2x2 + 4x3 ≥ 8
x1 ≥ 0, x2 ≥ 0, x3 ≥ 0
Question 10
Consider the Linear Programing problem below
Maximize z = 15 x1 + 10 x2
Subject to
.
x1 �2

x2 �3
x1 + x2 �4
The optimal solution is x1= 2, x2 = 2, z = 50
Suppose that the third constraint is changed to x 1+x2 ≤ 3. That is to say the new problem
becomes:
Maximize z =15x1+10x2
Subject to
x1 ≤ 2
x2 ≤ 3
x1+x2 ≤ 3
Investigate whether or not the solution to the Linear Programming problem is sensitive to
the change in the constraint.

Answers to questions
Question 1
The first step is to identify the variables. These are the values you can set or otherwise
control. The Mufoya variables are the production rates of the mountain bikes, x1 and
racers, x2 .
The objective function is as follows:
Maximize daily profit, i.e., maximize z = 15 x1 + 10 x2 (in $ per day)
The constraints are as follows:
Mountain bikes production limit x1 �2 (bikes per day)

Racers production limit x2 �3 (bikes per day)
Metal finishing machine production limit x1 + x2 �4 (bikes per day)
Questions 2 and 3
[A] Maximize 2x + 3y
subject to:
.
x + y �50
4 x + 2 y �220
2 x + 4 y �160
2 x + 10 y �370
5 x + 6 y �300
x, y �0
[B] Sketch:

Summary of optimal solution
Objectve Value = 130
x = 20.00
y = 30.00
Question 4
Let x0 be the value of the objective function add the slack variables x5 , x6 , x7 to your
constraints so that you have equality.
Then write the system as
x0 - 4 x1 - 5 x2 - 9 x3 - 11x4 = 0
x1 + x2 + x3 + x4 + x5 = 15
7 x1 + 5 x2 + 3x3 + 2 x4 + x6 = 120
3 x1 + 5x2 + 10 x3 + 15 x4 + x7 = 100
After going through four iterations of the Simplex Method you get to a stage where the non-
basic variables x2 , x4 , x5 or x7 can only have value zero.
That is to say we have:

695 3 11 13 5
x0 = - x2 - x4 - x5 - x7
7 7 7 7 7
Quesstion 5
Standard form:
Maximize 3x1 + 4x2
Subject to
2x1 + 3x2 + s1 = 600

x1 + x2 + s2 = 225
5x1 + 4x2 + s3 = 1000
x1 + 2x2 - s4 = 150
x1 , x2 , s1 , s2 , s3 , s4 ≥ 0
Not in canonical form because there is no basic variable in the fourth equation. Therefore
we add an artificial variable to that equation (r 1) and give it a large negative coefficient in
the objective function, to penalize it:
Maximize 3x1 + 4x2
Subject to
2x1 + x2 + s1 = 600
x1 + x2 + s2 = 225
5x1 + 4x2 +s3 = 1000
x1 + 2x2 - s4 + r1 = 150
x1 , x2 , s1 , s2 , s3 , s4 , r1 ≥ 0
x1 x2 s4 s1 s2 s3 r1 B
Z -3 -4 0 0 0 0 +M
s1 2 3 0 1 0 0 0 600
s2 1 1 0 0 1 0 0 225
s3 5 4 0 0 0 1 0 1000
r1 1 2 -1 0 0 0 1 150
Not in Canonical form because of +M entry on Z row for one basic variable (r 1). Pivot to
replace +M on Z row by zero - Z row – M*r1 row:

x1 x2 s4 s1 s2 s3 r1 b
x1 x2 s4 Z s1 (-3-M)s2 (-4-2M)
s3 Mr1 0b 0 0 0 -150M
Z -1/3 0 s1 0 2 4/3 03 0 0 M 1 800 0 0 0 600
s4 1/3 0 s2 1 1 2/3 01 0 0 -1 0 250 1 0 0 225
s2 1/3 0 s3 0 5 -1/3 14 0 0 0 0 25 0 1 0 1000
s3 7/3 0 r1 0 1 -4/3 02 1 -1 0 0 200 0 0 1 150
x2 2/3 1 0 1/3 0 0 0 200
x1 x2 s4 s1 s2 s3 r1 b
x1 x2 Z s4 -1 s10 s2-2 s03 r01 0b M 800
Z 0 s1 0 ½ 0 0 1 3/2 1 10 0M 0825 -3/2 375
s4 0 s2 0 ½ 1 0 1 ½ -1 0 0 1-1 0225 -½ 150
x1 1 s3 0 3 0 0 -1 2 3 00 00 175 -2 700
s3 0 x2 0 ½ 0 1 1 -½-7 0 1 00 025 ½ 75
x2 0 1 0 1 -2 0 0 250
Optimal tableau: Solution: x1* = 75 x2* = 150 Z* = 825

Question 6
Then after adding a surplus variable x3 in the inequality above, you can write the model as
x0 + 3 x1 + 2 x2 = 0
x1 + x2 = 10
x1 - x3 = 4
Next, introduce artificial variables y1 and y2 , and let M =10, giving
x0 + 3 x1 + 2 x2 + 10 y1 + 10 y2 = 0
x1 + x2 + y1 = 10
x1 - x3 + y2 = 4
To initiate the simplex algorithm, you have to subtract (M =10) times row 2 and (M =10)
times row 3 from row 1 to eliminate y1 and y2 :
x0 - 17 x1 - 8 x2 + 10 x3 = -140
x1 + x2 + y1 = 10
x1 - x3 + y2 = 4
The optimal solution x1 = 4 and x2 = 6
Question 7
Minimize 15 y1 + 120 y2 + 100 y3
Subject to.
y1 + 7 y2 + 3 y3 �4
y1 + 5 y2 + 5 y3 �5
y1 + 3 y2 + 10 y3 �9
y1 + 2 y2 + 15 y3 �11
y1 , y2 , y3 �0
Question 8
a) The second constraint implies that, x1 + x2 ≥ 4.5, which contradicts the first
constraint. If a problem has no feasible solution, then the problem itself is called
infeasible.

b) Here, we could set x2 to zero and let x1 be arbitrarily large. As long as x1 is greater
than 2, the solution will be feasible, and as it gets large the objective function does
too. Hence, the problem is unbounded.
Question 9
Question 10
You can solve the problem geometrically to find the optimal solution is x 1 = 2, x2 = 1, z = 40.
Since z and x2 changed when an original coefficient was changed, then we say that the L.P.
is sensitive.

17. REFERENCES
References
1. Vanderbei R. J. ( year). Linear Programming: Foundations and Extensions

All learners are required to read this book whenever he/she is starting a new unit or activity.
All other links are to help the learner to understand more about Linear Programming
problems
2. A gentle approach to linear programming

http://www.sce.carleton.ca/faculty/chinneck/po/Chapter1.pdf
3. J E Beasley , OR-Notes
http://people.brunel.ac.uk/~mastjjb/jeb/or/twomines
OR-NOTES provides a strong and easy to understand approach to the formulation of the
linear programming problems. The examples are well explained. Students are encouraged
to go through some of the worked examples to strengthen their problem linear
programming problem formulation techniques.
4. Varaiya, P. Lecture notes on Optimization

These notes go hand in hand with the main textbook
5. Management Science: Linear Programming Notes

These notes outline the problem formulation, graphical solution of linear programming
problems and the sensitivity analysis
6. Spyros,R. An Introduction to Linear Programming and the Simplex Algorithm

7. McCarl, B. and Spreen, T. H., 2002. DUALITY IN LINEAR PROGRAMMING.
8. Baker, S. L. (2006) Linear Programming I: Maximization
9. Linear Programming: A Geometric Approach

10. Basic Solutions

11. Arsham (2007)

http://home.ubalt.edu/ntsbarsh/opre640a/partVIII.htm)

The following references were also used by the authors
1. Wagner, H. M. (1975, 2nd ed). Principles of Principles of Operations Research.

Englewood Cliffs, New Jersey: Prentice-Hall.
2. Spencer, A. J. M., Parker, D. F., Berry, D. S., England, A. H., Faulkner, T. R., Green,
W. A., Holden, J. T., Middelton, D., & Rogers, T. G. (1977, reprint 1981). Engineering
Mathematics. London: Van Nostrand Reinhold Co Ltd.
3. Koshy, T. (1979). Finite Mathematics and Calculus (with application). Santa Monica,
CA: Goodyear Publishing Co. Inc
4. Taha, H. (2003, 7th ed). TORA Solver: Operation Research: An Introduction

University Name………………………………………………
Pre- Activity Activity Activity Activity Activit

Name Reg Number Gender Assessment 1 2 3 4 5
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Lecturer………………………………………. Signature………………
Coordinator……………………………. Signature……………….
Student Records

18. MAIN AUTHOR OF THE MODULE
This module was developed by David K J Mtetwa, BSc, MSc, MEd, PhD, Grad Cert Edu; in
collaboration with Admire Kurira, BSC, MSc; and Blessing Mufoya, BSc (Hons), MSc, who
are all mathematics teacher educators based in the Department of Science and
Mathematics Education at the University of Zimbabwe.
Dr Mtetwa was born and grew up in Eastern Zimbabwe, where he completed his primary
and secondary education. After obtaining his undergraduate degree majoring in
mathematics and physics at the University of Lesotho, Dr Mtetwa enrolled for a masters
degree in mathematics at a Canadian University. This was followed by a number of
teaching stints at a number of high schools and university colleges in Lesotho, Swaziland,
and Zimbabwe. His teaching certification training was completed at the University of
Zimbabwe in 1984, following which he crossed the Atlantic again, but his time to the United
States to undertake Masters and PhD degrees in mathematics education at the state
universities of New York and Virginia, that were completed in 1991. Dr Mtetwa has since
then been teaching and developing mathematics education courses for diploma and degree
levels at the University of Zimbabwe. He has supervised and graduated many post-
graduate students in the area of mathematics education. Dr Mtetwa is an active member of
various professional associations, including the influencial Southern African Association for
Research in Mathematics, Science, and Technology Education (SAARMSTE) and African
Commission on Mathematics Education (AFRCME).

Linear Programming

Uploaded by

Linear Programming

Uploaded by

AVU-PARTNER INSTITUTION

MODULE DEVELOPMENT TEMPLATE

Prepared by Thierry Karsenti, M.A., M.Ed., Ph.D.

With : Salomon Tchaméni Ngamo, M.A. & Toby Harper, M.A.

Version 19.0, Thursday December 21st, 2006

Module Development Template 1

2. PREREQUISITE COURSES OR KNOWLEDGE

Module Development Template 2

Module Development Template 3

Module Development Template 4

Identifying, describing, understanding, and appreciating the general linear

Computational strategies for seeking solutions of linear programming problems,

Module Development Template 5

 Geometrical interpretation of a solution of a linear programming problem

Unit 2: Computational algorithms

 Algebraic interpretation of the solution to a linear programming problem

 Stability considerations for a solution: sensitivity analysis

Module Development Template 6

7. General Objective(s): for the whole module

Module Development Template

Module Development Template 8

You should be able to:

You should secure your knowledge of school mathematics in:

You should exploit ICT opportunities in:

Unit Learning objective(s)

Module Development Template 9

Module Development Template 10

Title of pre-assessment: Basic Algebraic Ideas Test

1. Which of the following is linear:

2. Which of the following does not typically denote a vector:

(a) 35 (b) 25 (c) 53 (d) 5×2

4. A singular matrix is one that:

(a) 0 (b) 1 (c) 2 (d) 3

Module Development Template 11

Module Development Template 12

Module Development Template 13

(a) TRUE (b) FALSE

9. What is the transpose of matrix A?

11. The sequence {1/n} is bounded below by:

12. At an optimal value of a function:

13. The following is an example of a discontinuous function.

Module Development Template 14

16. A function is said to be bounded if:

(a) It is defined in open interval.

(a) 11 (b) 24 (c) 21 (d) 18

18. Maximising profit generation can be equivalent to minimising production costs:

(a) TRUE (b) FALSE?

19. A vector has:

20. A Subspace of vector space:

Module Development Template 15

1.b ( a and c are parabolic functions, d has a periodic function cosx)

4.c (a singular matrix is a matrix with determinant zero and so it non-invertable)

11. c (the sequence is a decreasing sequence which approaches zero)

12 c (the optimal point is the maximum or minimum point)

15.b (a and c converge to zero, d converges to 1 )

16.b (a function is bounded if it is bounded above and below)

17 b (just multiply the terms in the diagonal) .

Module Development Template 16

Module Development Template 17

UNIT 1: PROBLEM FORMULATION

Learning Activity # 1: Formulation of a Linear Programming Problem

Specific learning objectives

o The learner should be able to identify/recognize an optimization situation.

Summary of the learning activity

Module Development Template 18

List of required readings

 Linear Programming: Foundations and Extensions by Robert J. Vanderbei

 Lecture notes on Optimization by Pravin Varaiya

 A gentle approach to linear programming: chapters 1-7

List of relevant useful links

Linear Programming Formulation

Linear Programming Formulation

An Introduction to Linear Programming and the Simplex Algorithm by Spyros

Module Development Template 19

Linear Programming Formulation

Detailed description of the activity