Linear Programming

Linear Programming

Prepared by David K. J. MTETWA

African Virtual university

Universit Virtuelle Africaine
Universidade Virtual Africana

African Virtual University 

Notice
This document is published under the conditions of the Creative Commons
http://en.wikipedia.org/wiki/Creative_Commons
Attribution
http://creativecommons.org/licenses/by/2.5/
License (abbreviated cc-by), Version 2.5.

African Virtual University 

Table of Contents
I.

Linear programming_________________________________________ 3

II.

Prerequisite Course or Knowledge_ _____________________________ 3

III. Time_____________________________________________________ 3
IV. Materials__________________________________________________ 3
V.

Module Rationale_ __________________________________________ 4

VI. Content___________________________________________________ 5

6.1 Overview ____________________________________________ 5

6.2 Outline_ _____________________________________________ 6
6.3 Graphic Organizer______________________________________ 7

VII. General Objective(s)_________________________________________ 8

VIII. Specific Learning Activities____________________________________ 9
IX. Teaching and Learning Activities_______________________________ 10
X.

Learning Activities__________________________________________ 16

XI. Compiled List of all Key Concepts (Glossary)_____________________ 56

XII. Compiled List of Compulsory Readings _________________________ 62
XIII. Compiled List of (Optional) Multimedia Resources_________________ 63
XIV. Compiled List of Useful Links_________________________________ 65
XV. Synthesis of the Module_____________________________________ 68
XVI. Formative and Summative Evaluation_ __________________________ 69
XVII. References_______________________________________________ 82
XVIII. File Structure_____________________________________________ 84
XIX. Main Author of the Module_ __________________________________ 85

African Virtual University 

I. Linear Programming
by David K J Mtetwa

II. 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.

III. 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.

IV. 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 students 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].

African Virtual University 

V. 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.

African Virtual University 

VI. Content
6.1 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.

African Virtual University 

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.

6.2 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
o
o
o
o
o
o

The Big M algorithm

The Simplex algorithm
degeneracy
efficiency
Notion of duality
the primal simplex
the dual simplex

Stability considerations for a solution: sensitivity analysis

o Marginal analysis
o Parametric analysis

African Virtual University 

6.3

Graphic Organiser

ACTIVITY 1
Formulation of the
linear programming

LP problem
in Standard
General LP
problems

UNIT 1
2-var PL
ACTIVITY 2
Geometrical Solution

3-var PL
Boundedness

MODULE

ACTIVITY 3
Optimality Condition

Convergence
Degeneracy
Simplex Method

UNIT 2

ACTIVITY 4
Algebraic Approach

Efficiency

Big M-Method
Primal Simplex
Duality
Dual simplex
Parametric Analysis

ACTIVITY 5
Sensitivity Analysis
Marginal Analysis

African Virtual University 

VII. General Objective(s)

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

African Virtual University 

VIII. Specific Learning Objectives

(Instructional Objectives)

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)

Problem Formulation

o


o

The learner should be able to identify/recognize

an optimization situation in real life decision ma
king activity.
The learner should be able to produce a
mathematical model using appropriate language.

Problem Formulation

o


o

o

o



o



o


o

The learner should be able to describe, explain,

and apply optimality conditions to specific
linear programming problems.
The learner should be able to describe the under
lying logic of the Simplex algorithm.
The learner should be able to relate the algebraic
solution with the geometric solution.
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.
The learner should be able to explain the
meaning of duality and describe its role in the
search for solutions of linear programming
problems.
The learner should be able to describe the purposes
of carrying out a sensitivity analysis for a given
linear programming solution.
The learner should be able to describe a procedure
for carrying out a sensitivity analysis to a given
optimal solution.

African Virtual University 10

IX. Teaching and Learning Activities

Pre-assessment
Title of pre-assessment: Basic Algebraic Ideas Test
Rationale: To check leaners 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:

(a) -5

4

(d) 8
-3

(b) (1, 2, 3) (c) A

2 4 2 2 0

3. The matrix 1 1 6 4 9 is:

0 0 0 0 0

(a) 35

(b) 25

(c) 53

(d) 52

4. A singular matrix is one that:

(a ) is single (b) is invertable (c) is non-invertable (d) has determinant 1?

1
5. The matrix 0

1
(a) 0

4 2 2 0

1 6 4 9
0 0 0 0 has rank:

2 1 1 0

(b) 1 (c) 2

(d) 3

African Virtual University 11

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

African Virtual University 12

9. What is the transpose of matrix A?

A = 1
2

(a) 2
4

1 3

0 2
-1 5
1 1 3
0 2 4 1 2 4 5 3

-1 5 (b) 1 0 -1 (c) 2 -1 0 (d) 0 1 2

2 5 1
1 3 3 2 5 1 2 2

10. Which of the following shows a convex set?

(b)

(a)

(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)

(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

African Virtual University 13

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

17. The determinant of the following matrix 0 4 5 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?

African Virtual University 14

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 its 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` )

African Virtual University 15

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.

African Virtual University 16

X. 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
The learner should be able to identify/recognize an optimization situation.
The learner should be able to produce a mathematical model using appropriate
language.
The learner should be able distinguish and relate between the Standard and
the General form.
The learner should be able represent the mathematical model geometrically.
The learner should be able identify feasible regions, vertices, and convexes.
The learner should understand the concepts of systems of linear equations,
constraint, feasible solution and feasible region.
The learner should be able to interpret a real life problem and transform it
into a linear programming problem.
The learner should be able to check or verify the feasible solution.
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.
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

African Virtual University 17

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
Last visited: 10-02-07
A gentle approach to linear programming: chapters 1-7
http://www.sce.carleton.ca/faculty/chinneck/po/Chapter1.pdf
Last visited: 08-03-07

African Virtual University 18

List of relevant useful links

Linear Programming Formulation
http://people.brunel.ac.uk/~mastjjb/jeb/or/lpmore.html
Last visited: 15-02-07
Linear Programming Formulation
http://home.ubalt.edu/ntsbarsh/opre640a/partVIII.htm
Last visited: 15-02-07
An Introduction to Linear Programming and the Simplex Algorithm by Spyros
Reveliotis
http://www2.isye.gatech.edu/~spyros/linear programming/linear programming. html
Last visited: 14-02-07
OR-Notes J E Beasley
http://people.brunel.ac.uk/~mastjjb/jeb/or/twomines
Last visited: 16-02-07
Linear Programming
http://home.ubalt.edu/ntsbarsh/opre640a/rpcotdp#rpcotdp
Last visited: 21-02-07
Linear Programming Formulation
www.people.brunel.ac.uk/~mastjjb/jeb/lp.html
Last visited: 15-02-07

Detailed description of the activity

The learner will:
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.
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.
Attempt the exercise given: The learner refers to indicated pages of the recommended textbook .
Open the relevant links to get deeper understanding of processes of problem
formulation.

African Virtual University 19

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
Last visited: 11-02-07
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.

mis the number of types of raw materials .
bi is the amount of ith raw material , for i= 1, 2,,m
i is the unit market price of ith raw material, for i=1, 2,m at a given moment.

African Virtual University 20

aij is the amount of ith raw material required to produce one unit of jth product, for
j=1,2,.,n
j is the unit market price for the jth product, for j=1,2,.,n
x j is the number of jth product produced
m

pa
i

ij

is the cost of production of one of jth product,

i =1
m

a
i

ij

= 1 a1 j + 2 a2 j + ... + m amj

i =1

c j is net profit for the jth product

m

c j = j - pi aij
i =1

cj xj

is the total net profit for x j 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 , x2 ,...., xn ) = c1 x1 + c2 x2 + .. + cn xn

African Virtual University 21

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
am2
amn

Sign Restrictions
xj 0

Further reading
Read the following link:

An Introduction to Linear Programming and the Simplex Algorithm by Spyros

Reveliotis
http://www2.isye.gatech.edu/~spyros/linear programming/linear programming.html

African Virtual University 22

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.

constraint on available land.
Then x1 + x2 125
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 18x1 + 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 ) = 40000x1 + 25x2

subject to the constraints

(1)

African Virtual University 23

x1 + x2 125

2x2 + x2 150

3x1 + x2 180

and

x1 0, x2 0
(2)
Summary guidelines for linear programming problem formulation
a)
b)
c)
d)
e)
f)

Understand the problem thoroughly.

Describe the objective.
Describe each constraint.
Define the decision variables.
Write the objective in terms of the decision variables.
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

African Virtual University 24

The General Linear Programming Problem

Readings
Read page 6 in the textbook.
Foundations and Extensions by Robert J. Vanderbei.
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

African Virtual University 25

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(200x) + 30(140 y)

We rewrite the problem as:
Maximize

f (x, y) = 5000x + 4200 y

Such that
x + y 40
200x 6000
140y 4000
x 0, y 0

African Virtual University 26

Learning Activity 2 : A Geometrical Approach to Linear Programming

Specific Learning Objectives
By the end of this section the learner must:
Be able to identify and construct linear equations.
Be able to construct the wanted regions given a set of inequalities, i.e. the
feasible regions of each Linear Programming Problem.
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.

Key Terms (refer to glossary)

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

African Virtual University 27

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
Last visited: 10-02-07

List of Relevant Useful Links

Linear Programming: A Geometric Approach
http://www.wiley.com/college/sc/sullivan/CH03.pdf
Last visited: 16-02-07

Graphical Solution Method

http://home.ubalt.edu/ntsbarsh/opre640a/rpcotdp#rpcotdp
Last visited: 21/02/07
Linear Programming: Geometric Approach
www.math.tamu.edu/~janice.epstein/141/notes/Ch3.pdf
Last visited: 16-02-07

Detailed Description of the Activity

The learner will:
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.
Analyze the example given in the link. This will help the learner to understand
the steps and procedures followed in the resolution.
Do the exercise from the textbook Linear Programming: Foundations and
Extensions by Robert J. Vanderbei.

African Virtual University 28

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 section5.
www.princeton.edu/~rvdb/LPbook/onlinebook.pdf
Last visited: 15-02-07

Further reading
Read the following link on the Geometrical Approach
Linear Programming: A Geometric Approach page 171-176
http://www.wiley.com/college/sc/sullivan/CH03.pdf
Last visited: 16-02-07
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:

African Virtual University 29

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.

f (x, y) = 5x + 7 y
Maximize:
subject to:

x+ y-70
2x - 3y + 6 0
x 0, y 0
1. Is the problem amenable to 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-70
2x - 3y + 6 0
x 0, y 0

African Virtual University 30

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.

African Virtual University 31

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
http://www.wiley.com/college/sc/sullivan/CH03.pdf
This is a maximizing problem in which the profit is being maximized.
Linear Programming: Geometric Approach
www.math.tamu.edu/~janice.epstein/141/notes/Ch3.pdf
Last visited: 16-02-07
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

African Virtual University 32

UNIT 2: RESOLUTION AND ANALYSIS

Learning Activity # 3: Optimality Conditions

Specific Learning Objectives
After studying this section, you should:
be able to interpret the derivatives and existence of optimality
be able to prove the theorems of optimality
be able to explain and prove the The Fundamental Theorem of Linear Pro-

gramming
be able to check optimality conditions for solving a Linear Programming
problem.
be able to carry out a sensitivity analysis for a given optimal solution.

Summary of the Learning Activity

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.

Key Terms (refer to glossary)

Basic variables
Nonbasic variables
Augmented matrix
Pivot
Pivot column
Test ratio
Tableau
Cycling

African Virtual University 33

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
Last visited: 10-02-07

List of Relevant Useful Links

Optimality Conditions for Constrained Optimization
http://ocw.mit.edu/NR/rdonlyres/Sloan-school-of-Management/15084JSpring2004/7240EF84-B20D-419F-B1C0-2DAF3277F5C4/0/lec6_
constr_opt.pdf
Last visited: 16-02-07
Optimality Conditions
http://www.math.mtu.edu/%7Emsgocken/ma5630spring2003/lectures/lag1/
lag1/node1.html
Last visited: 16-02-07

Alternate Optimal Solutions, Degeneracy, Unboudedness, Infeasibility

http://mat.gsia.cmu.edu/QUANT/notes/node63.html#SECTION0083000000
0000000000

African Virtual University 34

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 non- basic
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
www.princeton.edu/~rvdb/LPbook/onlinebook.pdf
Last visited: 15-02-07

Further Reading
http://engr.smu.edu/~barr/ip/ch1/node7.html
http://www.maths.abdn.ac.uk/~igc/tch/mx3503/notes/node67.htm

African Virtual University 35

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 4x1 + 5x2 + 9x3 + 11x4
Subject to

x1 + x2 + x3 + x4 15
7x1 + 5x2 + 3x3 + 2x4 120
3x1 + 5x2 + 10x3 + 15x4 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 - 4x1 - 5x2 - 9x3 - 11x4 = 0

x1 + x2 + x3 + x4 + x5 = 15
7x1 + 5x2 + 3x3 + 2x4 + x6 = 120
3x1 + 5x2 + 10x3 + 15x4 + 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

x0 =

695 3
11
13
5
- x2 - x4 - x5 - x7
7
7
7
7
7

This is the optimality condition which allows us to terminate the iteration.

African Virtual University 36

Learning Activity 4: Algebraic Approach

Specific Learning Objectives

After studying this section, you will:
Be able to interpret the derivatives and existence of optimality.
Be able to apply the Method (penalty) of big M and Dual simplex method.
Be able to differentiate the properties of duality, mutual primal algorithm, and
dual simplex algorithm.
Be able to follow the given algorithms step-by-step, i.e. mutual primal and
the dual simplex algorithms.
Be able to solve linear programming problems using appropriate mathematical
software (when available)

Summary of the Learning Activity

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.

Key Terms (refer to glossary)

Basis
Basic variables
Nonbasic variables
Slack variable
Surplus variable
Artificial variable
Basic solution
Degeneracy

African Virtual University 37

Primal problem
Dual problem
Weak duality
Strong duality
Primal feasibility
Dual feasibility
Primal feasibility
Dual feasibility

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
Last visited: 10-02-07

List of Relevant Useful Links

Simplex Method - Big M
http://www.math.uwo.ca/~heinicke/courses/236_03/bigM.pdf
Last visited: 15-02-07
Simplex Method Big M
http://www.computing.dcu.ie/~lkillen/teach/CA427Simplexbigmexample.
pdf
Last visited: 15-02-07
Dual Problem: Construction and Its Meaning
http://home.ubalt.edu/ntsbarsh/opre640a/rpcotdp#rpcotdp
Sensitivity Analysis for Linear Programming
http://mat.gsia.cmu.edu/QUANT/notes/node64.html
Last visited: 20/02/07

African Virtual University 38

Detailed Description of the Activity

The learner will:
Read the the textbook Linear Programming: Foundations and Extensions on
the Simplex Method pages 13-28, 29-43, and 89-107 of the textbook
Analyze the example that is provided in the link below:
- Simplex Method - Big M

http://www.math.uwo.ca/~heinicke/courses/236_03/bigM.pdf

Last visited: 15-02-07
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.

African Virtual University 39

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 finitedimensional 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: http://engr.smu.edu/~barr/ip/ch1/node6.html
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) = 170x + 225y

African Virtual University 40

Subject to:

x + y 300
2x + 3y 720
x 0, y 0
Step 1 Form the initial simplex tableau
Let us introduce two slack variables s1 0 and s2 0 then the first inqualities can
be written as equalities

x + y + s1 = 300
2x + 3y + 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

s1

-170

-225

s2

rhs

300

720

Step 2 Locate the pivot of the tableau

x

s1

-170

-225

s2

rhs

300

720

Pivot row

Pivot

Pivot column

African Virtual University 41

Step 3 Since the pivot is 3, we divide the second row by 3

x

s1

2/3

-225

-170

s2

1/3

rhs

300

240

Step 4 Adding -1 times the second to the first and 225 times the second row to the

third, we get
x

s1

1/3

2/3

-20

rhs

60

1/3

240

5400

s2

-1/3

75

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

s1

s2

rhs

-1

180

-2

120

60

55

57600

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 + 2s1 - s2
f = 57600 - 60s1 - 95s2

Maximum value

African Virtual University 42

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 - 2x2

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 + 3x1 + 2x2 = 0
x1 + x2 = 10

x1 - x3 = 4

Where f (x1 , x2 ) = -3x1 - 2x2 has been written as

x0 + 3x1 + 2x2 = 0 where x0 = f (x1 , x2 )

African Virtual University 43

Next, introduce artificial variables y1 and y2 , and let M =10 as our large integer, for
instance, giving

x0 + 3x1 + 2x2 + 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 - 17x1 - 8x2 + 10x3 = -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?

African Virtual University 44

Exercise 4
The following tableaus tables were obtained in the course of solving linear program
nonnegative 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).
z

x1

x2

s1

s2

rhs

20

-2

-1

-1

x1

x2

s1

s2

rhs

-1

20

-2

-2

c)

x1

x2

s1

s2

rhs

-2

-2

x1

x2

s1

s2

rhs

-1

-1

-1

a)

b)

d)

Hint: Answers to Excercise 4 can easily be obtained by refering to the definitions

of unboundedness, unique optimum solution, alternate optimum solution, and
degeneracy.

African Virtual University 45

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 tailors problem
Uncontrollable Inputs

trousers

shirts

available

labour

50

raw material

60

net income

20

15

and its linear programming is:

Maximize

f (x1 , x2 ) = 20x1 + 15x2

subject to:
3x1 + 2x2 50

labour constraint

4x1 + 3x2 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 z1 = the dollar amount
payable to the tailor for every labour hour lost due to unforseen problems, and z2 =
the dollar amount payable to the tailor for every raw material unit lost.

African Virtual University 46

The insurance broker tries to minimize the total amount of $(50z1 +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 brokers problem is:

Minimize
f (z1 , z2 ) = 50z1 + 60z2
subject to:
3z1 + 4z2 20 Net income from a trouser
2z1 + 3z2 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 companys
problem is closely related to the tailors 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]
Last visited: 15-02-07

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

j =1

i =1

c j xj = bi yi

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.

African Virtual University 47

Note
The duality relationship can be summarised as follows:

Primal (Maximize)

Dual (Minimize)

Objective function
Right-hand side
jth column of coefficient
ith row of coefficients
jth variable nonnegative
jth variable unrestricted in sign
ith relation an inequality ( )
ith relation an equality

Right-hand side
Objective function
jth row of coefficient
ith column of coefficient
jth relation an inequality ( )
jth relation an equality
ith variable nonegative

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

4x1 + 5x2 + 9x3 + 11x4

Subject to:

x1 + x2 + x3 + x4 15
7x1 + 5x2 + 3x3 + 2x4 120
3x1 + 5x2 + 10x3 + 15x4 100

xj 0

For j =1, 2, 3, 4
The dual problem can easily be obtained using the duality relationships in the table
above as:
Minimize

15y1 + 120 y2 + 100 y3

Subject to

x1 + x2 + x3 + x4 15
7x1 + 5x2 + 3x3 + 2x4 120
3x1 + 5x2 + 10x3 + 15x4 100

xj 0

African Virtual University 48

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
Answers to Exercise 5
Question 1
Read the Tailors 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 - 2x2
Subject to

x1 + 2x2 - x3 + x4 0
4x1 + 3x2 + 4x3 - 2x4 3
-x1 - x2 + 2x3 + x4 = 1

x2 0, x3 0

We rewrite the problem as follows

African Virtual University 49

Maximize x1 - 2x2
Subject to

-x1 - 2x2 + x3 - x4 0
4x1 + 3x2 + 4x3 - 2x4 3
-x1 - x2 + 2x3 + x4 1
x1 + x2 - 2x3 - x4 -1
The dual of this problem is therefore:
Minimize 3y2 + y3 - y4
Subject to

- y1 + 4 y2 - y3 + y4 1
2 y1 + 3y2 - 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.

African Virtual University 50

Learning activity 5: Sensitivity Analysis

Specific Learning Objective
By the end of the activity the learner:
should be able to carry out a sensitivity analysis of a Linear Programming
solution.
Should be able to explain the reasons for carrying a sensitity analysis.

Summary of the Learning Activity

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.

Key Terms (refer to glossary)

Parametric analysis
Marginal analysis

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
Last visited: 10-02-07

African Virtual University 51

List of Relevant Useful Links

Sensitivity Analysis
http://www.jr2.ox.ac.uk/bandolier/booth/glossary/sensanal.html
Last visited: 15-02-07
Sensitivity Analysis
http://pespmc1.vub.ac.be/ASC/SENSIT_ANALY.html
Last visited: 15-02-07
Sensitivity Analysis
http://en.wikipedia.org/wiki/sensitivity_analysis
Last visited: 15-02-07
Management Science: Linear Programming Notes
http://www.strathcona.bham.ac.uk/Pdfs1%20management%20Course%20Y
ear%202/LINEAR%20PROGRAMM%20NOTES.PDF

Detailed Description of the Activity

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

African Virtual University 52

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)]

http://mat.gsia.cmu.edu/QUANT/notes/node64.html
Q. Where do I start in sensitivity analysis?
A. If we carry out the simplex method iterations on the initial tableau below

-170

s1

s2

RHS

300

720

-225

We get we get the final tableau called terminal tableau below

basis (identity matrix)

x1

x2

s1

s2

RHS

-1

180

-2

120

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.

African Virtual University 53

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 Rows1 through m to other rows; we never added or
subtracted Row0 to other rows).

s1

s2

RHS

3+

300

720

-170

-225

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) = 4x + 5y

subject to:
2x + 3y 12
x+ y5
x, y 0

The optimal tableau, after adding slacks s1 and s2 is

s1

s2

RHS

22

-2

-1

Now suppose we change the amount of the right-hand-side from 12 to (12+) in the
first constraint, the tableau changes to:

African Virtual University 54

s1

s2

RHS

22 +

-2

2+

-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 11
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
Last visited: 15-02-07
Sensitivity Analysis
http://en.wikipedia.org/wiki/sensitivity_analysis
Last visited: 15-02-07
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

African Virtual University 55

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
Last visited: 15-02-07

Further Activity
Optimization Methods Lecture 7 Sensitivity Analysis
http://ocw.mit.edu/NR/rdonlyres/Sloan-School-of-Management/15-093Fall2004/8842223F-9C8A-4B9E-83F8-DD28DE270140/0/lecture07.pdf
Management Science: Linear Programming Notes
http://www.strathcona.bham.ac.uk/Pdfs1%20management%20Course%20Y
ear%202/LINEAR%20PROGRAMM%20NOTES.PDF
Answers to Exercise 6
Page 449 of Foundations and Extensions by Robert J. Vanderbei

African Virtual University 56

XI. 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 X1 , X 2 S , belongs completely in S. Mathematical definition of convexity is expressed as:

S convex X 1 , X 2 S 2 ,x (0,1) : (1 - x) X 1 + xX 2 S
Convex sets

Nonconvex

Convex

African Virtual University 57

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 cx .
Strong duality is when the primal problem has an optimal solution and the

dual problem has also an optimal solution and cx = 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:aX=b

X:aX<b

African Virtual University 58

Half-planes
the feasible region of a linear inequality

For example the half-planes of the inequality a1 X1 + a2 X 2 b are the planes on

the left and right side of the equality a1 X1 + a2 X 2 b .

given a linear constraint:

a1 X1 + a2 X 2 + ... + an 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 setS) 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).

African Virtual University 59

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 organisations 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 column

Pivot row

Pivot

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

African Virtual University 60

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 nonbasic 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)
2)
3)
4)

The objective function is to be maximized.

All inequalities are of the type.
All right hand constants are non-negative.
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.

African Virtual University 61

Tableau
(pl. tableaux). A detached coefficient form of a system of equations, which can
change from x +Ay = b to x+Ay = 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.

African Virtual University 62

XII. Compiled List of Compulsory Readings

Linear Programming: Foundations and Extensions by Robert J. Vanderbei
www.princeton.edu/~rvdb/LPbook/onlinebook.pdf
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

African Virtual University 63

XIII. Compiled List of (Optional) Multimedia

Resources
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.

African Virtual University 64

The following sites enable you to practice the Linear Programming pivoting
without having to do the arithmetic!
1.
2.
3.
4.

Pivot Tool: http://www.sor.princeton.edu/~rvdb/JAVA/pivot/advanced.html

TUTOR: http://www.tutor.ms.unimelb.edu.au/
The Simplex Place: http://www.ifors.org/tutorial
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

African Virtual University 65

XIV. Compiled List of Useful Links

Linear Programming Formulation
http://people.brunel.ac.uk/~mastjjb/jeb/or/lpmore.html
Linear Programming Formulation
http://people.brunel.ac.uk/~mastjjb/jeb/or/lpmore.html
Last visited: 15-02-07
Linear Programming Formulation
http://home.ubalt.edu/ntsbarsh/opre640a/partVIII.htm
Last visited: 15-02-07
An Introduction to Linear Programming and the Simplex Algorithm by Spyros
Reveliotis http://www2.isye.gatech.edu/~spyros/linear programming/linear
programming.html
Last visited: 14-02-07
(1) Linear Programming
http://home.ubalt.edu/ntsbarsh/opre640a/rpcotdp#rpcotdp
Last visited: 21-02-07
(2) Linear Programming Formulation
www.people.brunel.ac.uk/~mastjjb/jeb/lp.html
Linear Programming: A Geometric Approach page 171-176
http://www.wiley.com/college/sc/sullivan/CH03.pdf
Optimality Conditions for Constrained Optimization
http://ocw.mit.edu/NR/rdonlyres/Sloan-school-of-Management/15084JSpring2004/7240EF84-B20D-419F-B1C0-2DAF3277F5C4/0/lec6_
constr_opt.pdf
Last visited: 16-02-07

African Virtual University 66

Mathematical approach of checking conditions of optimality

Optimality Conditions
http://www.math.mtu.edu/%7Emsgocken/ma5630spring2003/lectures/lag1/
lag1/node1.html
Last visited: 16-02-07
Alternate Optimal Solutions, Degeneracy, Unboudedness, Infeasibility
http://mat.gsia.cmu.edu/QUANT/notes/node63.html#SECTION008300000
00000000000
Degeneracy and unboundedness are well explained with aid of simple
examples.
The Fundamental Theorem of Linear Programming
http://www.it.uu.se/edu/course/homepage/opt1/ht06/Lectures/fundamentat_thm-2up.pdf
http://engr.smu.edu/~barr/ip/ch1/node7.html
http://www.maths.abdn.ac.uk/~igc/tch/mx3503/notes/node67.htm
These links provides theorems of linear programming and the relevant
proofs
Simplex Method - Big M
http://www.math.uwo.ca/~heinicke/courses/236_03/bigM.pdf
Last visited: 15-02-07
Simplex Method Big M
http://www.computing.dcu.ie/~lkillen/teach/CA427Simplexbigmexample.pdf
Last visited: 15-02-07
The algorithm of the B-M method is examplified
Dual Problem: Construction and Its Meaning
http://home.ubalt.edu/ntsbarsh/opre640a/rpcotdp#rpcotdp
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.
Sensitivity Analysis for Linear Programming

http://mat.gsia.cmu.edu/QUANT/notes/node64.html

Last visited: 20-02-07

This link gives a good exaplanation on Why sensetivity analysis?.
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.

African Virtual University 67

Duality in Linear Programming

http://www.web.mit.edu/15.053/www/AMP-Chapter-04.pdf
Explains the duality thoerem and duality properties
Basic Solutions
http://engr.smu.edu/~barr/ip/ch1/node6.html
Linear Programming: Geometric Approach
www.math.tamu.edu/~janice.epstein/141/notes/Ch3.pdf
Last visited date 16-02-07

African Virtual University 68

XV. 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.

African Virtual University 69

XVI. 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 students 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 students 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].

African Virtual University 70

Question 1 tests the students understanding of a linear programming problem situation through the students ability to formulate the linear programming problem
properly (i.e. writing the problem in mathematical language).
Questions 2 and 3 test the students 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 students capability to solve the linear programming problem
using the Simplex Method.
Questions 5 and 6 test the students 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 students capability to investigate the feasibility and boundedness
characteristics of a linear programming problem.
Question 9 tests the students capability to solve the linear programming problem
using the Dual Simplex method.
Question 10 tests the students capability to solve a linear programming problem
using the Simplex Method and then carry out a sensitivity analysis on the solution
obtained.

African Virtual University 71

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. Lets 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

220

160

370

10

300

Standard (x)

Precision (y)

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.

African Virtual University 72

Question 4
Use the Simplex method to solve the following linear programming problem

Maximize 4x1 + 5x2 + 9x3 + 11x4

Subject to

x1 + x2 + x3 + x4 15
7x1 + 5x2 + 3x3 + 2x4 120

3x1 + 5x2 + 10x3 + 15x4 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 + x 225

5x + 4x 1000

x + 2x 150

x,x 0

Question 6
Maximize
Subject to

-3x1 - 2x2

x1 + x2 = 10
x1 4

x1 0, x2 0

African Virtual University 73

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 4x1 + 5x2 + 9x3 + 11x4
Subject to

x1 + x2 + x3 + x4 15

7x1 + 5x2 + 3x3 + 2x4 120

3x1 + 5x2 + 10x3 + 15x4 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

subject to

x1 4x2

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

African Virtual University 74

Question 10
Consider the Linear Programing problem below
Maximize z = 15x1 + 10x2
Subject to

x1 2

x2 3
x +x 4

1
2

The optimal solution is x1= 2, x2 = 2, z = 50

Suppose that the third constraint is changed to x1+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.

African Virtual University 75

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 = 15x1 + 10x2 (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
subject to:

2x + 3y

x + y 50
4x + 2 y 220
2x + 4 y 160
2x + 10 y 370
5x + 6 y 300
x, y 0

African Virtual University 76

[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 - 4x1 - 5x2 - 9x3 - 11x4 = 0

x1 + x2 + x3 + x4 + x5 = 15
7x1 + 5x2 + 3x3 + 2x4 + x6 = 120

3x1 + 5x2 + 10x3 + 15x4 + x7 = 100

African Virtual University 77

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:

x0 =

695 3
11
13
5
- x2 - x4 - x5 - x7
7
7
7
7
7

Question 5
Standard form:
Maximize 3x + 4x
1

Subject to

2x + 3x + s = 600

x + x + s = 225

5x + 4x + s = 1000

x + 2x - s = 150

x ,x ,s ,s ,s ,s 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 ) and give it a large negative
1
coefficient in the objective function, to penalize it:
Maximize 3x + 4x
1

Subject to

2x + x + s = 600

x + x + s = 225

5x + 4x +s = 1000

x + 2x - s + r = 150

x ,x ,s ,s ,s ,s ,r 0

1
1

African Virtual University 78

x1

x2

s4

s1

s2

s3

r1

-3

-4

+M

s1

600

s2

225

s3

0 1000

r1

-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*r row:
1

x1

x2

(-3 -M) (-4 -2 M) 0

-150M

s1

600

s2

225

s3

1000

r1

-1

150

-1

-2

3/2

-3/2 375

- 150

-2

700

75

-1/3

4/3

M 800

s4

1/3

2/3

-1 250

s2

1/3

-1/3 1

25

s3

7/3

-4/3 0

200

x2

2/3

1/3

200

1
2
3
2

s4

s1

s2

s3

r1

800

African Virtual University 79

M 825

s4

-1

-1 225

x1

-1

75

s3

-7

25

x2

-2

250

Optimal tableau: Solution: x * = 75

1

x * = 150
2

Z* = 825

Question 6
Then after adding a surplus variable x3 in the inequality above, you can write the
model as

x0 + 3x1 + 2x2 = 0
x1 + x2 = 10

x1 - x3 = 4

Next, introduce artificial variables y1 and y2 , and let M =10, giving

x0 + 3x1 + 2x2 + 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 - 17x1 - 8x2 + 10x3 = -140

x1 + x2 + y1 = 10

x1 - x3 + y2 = 4

The optimal solution x1 = 4 and x2 = 6

African Virtual University 80

Question 7
Minimize 15y1 + 120 y2 + 100 y3
Subject to.

y1 + 7 y2 + 3y3 4
y1 + 5y2 + 5y3 5
y1 + 3y2 + 10 y3 9

y1 + 2 y2 + 15y3 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

African Virtual University 81

Question 10
You can solve the problem geometrically to find the optimal solution is x1 = 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.

African Virtual University 82

XVII. References
Vanderbei R. J. ( year). Linear Programming: Foundations and Extensions
www.princeton.edu/~rvdb/LPbook/onlinebook.pdf
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
A gentle approach to linear programming
http://www.sce.carleton.ca/faculty/chinneck/po/Chapter1.pdf
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.
Varaiya, P. Lecture notes on Optimization
http://robotics.eecs.berkeley.edu/~varaiya/papers_ps.dir/NOO.pdf
These notes go hand in hand with the main textbook
Management Science: Linear Programming Notes
http://www.strathcona.bham.ac.uk/Pdfs1%20management%20Course%20Ye
%202/LINEAR%20PROGRAMM%20NOTES.PDF
These notes outline the problem formulation, graphical solution of linear pr
gramming problems and the sensitivity analysis
Spyros,R. An Introduction to Linear Programming and the Simplex Algorithm
http://www2.isye.gatech.edu/~spyros/linear programming/linear programming.
html
McCarl, B. and Spreen, T. H., 2002. DUALITY IN LINEAR PROGRAMMING.
Baker, S. L. (2006) Linear Programming I: Maximization
Linear Programming: A Geometric Approach
http://www.wiley.com/college/sc/sullivan/CH03.pdf
Basic Solutions
http://engr.smu.edu/~barr/ip/ch1/node6.html
Arsham (2007)
http://home.ubalt.edu/ntsbarsh/opre640a/partVIII.htm)

African Virtual University 83

The following references were also used by the authors

Wagner, H. M. (1975, 2nd ed). Principles of Principles of Operations Research.
Englewood Cliffs, New Jersey: Prentice-Hall.
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.
Koshy, T. (1979). Finite Mathematics and Calculus (with application). Santa
Monica, CA: Goodyear Publishing Co. Inc
Taha, H. (2003, 7th ed). TORA Solver: Operation Research: An Introduction

Signature.

Activity
1

Coordinator.

PreAssessment

Signature

Gender

Lecturer.

Reg Number

Activity
2

Activity
3

Activity
4

Activity
5

Activity
Ave
Exam

102

Date

Date

Grade

XVIII. File Structure

Module Development Template

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Name

University Name

Student Records

African Virtual University 84

African Virtual University 85

XIX. 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).