UNIT IV: LINEAR PROGRAMMING AND ITS

APPLICATION

10. LINEAR PROGRAMMING: SIMPLEX AND DUALITY

11. THE TRANSPORTATION PROBLEM

12. THE ASSIGNMENT PROBLEM

LESSON 10​ LINEAR PROGRAMMING: SIMPLEX AND

DUALITY
10.0​OBJECTIVES​ 353
10.1​INTRODUCTION​ 353
10.1.1 Definition of Operations Research​ 354
10.1.2 Nature of Operations Research​ 354
10.1.3 Features of Operations Research​ 354
10.1.4 Methodology of Operations Research​ 355
10.2​LINEAR PROGRAMMING PROBLEM​ 357
10.2.1 Components of a LPP​ 358
10.2.2 Basic assumptions of a LPP Model​ 358
10.3​MATHEMATICAL FORMULATION OF LINEAR PROGRAMMING PROBLEM​ 359
10.4​GRAPHICAL METHOD​ 363
10.5​GENERAL LINEAR PROGRAMMING PROBLEM​ 369
10.6​CANONICAL FORM OF LINEAR PROGRAMMING PROBLEM​ 371
10.7​STANDARD FORM OF LINEAR PROGRAMMING PROBLEM​ 372
10.8​THEORY OF SIMPLEX METHOD​ 375
10.8.1 Some Important Definitions​ 375
10.8.2 Geometric Interpretation of Simplex Method​ 378
10.8.3 The Key Solution Concepts​ 379
10.9​THE SIMPLEX METHOD​ 380
 10.9.1 Conditions for Applying Simplex Method​ 381
10.9.2 The Simplex Algorithm​ 381
10.10 ARTIFICIAL VARIABLE TECHNIQUES​ 389
10.10.1 Big M Method​ 390
10.10.2 Two Phase Method​ 391
10.11 Special Cases in LPP​ 394
10.11.1 Multiple Optimal Solutions​ 394
10.11.2 Tie for entering basic variables​ 397
10.11.3 Degeneracy​ 397
10.11.4 No leaving variable (Unbounded Solution)​ 399
10.11.5 Infeasibility​ 401
10.11.6 Unrestricted (or unconstrained) variables​ 402
10.12 DUALITY IN LINEAR PROGRAMMING PROBLEM​ 406
10.12.1 Essence of Duality​ 407
10.12.2 Rules for converting Primal into its Dual​ 408
10.12.3 Relationship between Primal and Dual​ 413
10.12.4 Properties of Dual​ 413
10.13 DUALITY AND SIMPLEX METHOD​ 414
10.14 ECONOMIC INTERPRETATION​ 422
MISCLLANEOUS ILLUSTRATION​ 423
10.15 CONCLUSION​ 432
10.16 ANSWERS TO SELF-CHECK QUESTIONS​ 433
10.17 TERMINAL QUESTIONS​ 433
10.18 FURTHER READINGS​ 436
10.19 GLOSSARY OF THE TERMS USED​ 436
 10. Linear Programming: Simplex and Duality

We live in a world of shortages since our resources are limited. Thus there is always
a problem as to how to allocate the given resources in the best possible way. This
Lesson provides basic concepts of Operations Research. It covers some general
ideas on the subject, graphical method to solve the Linear Programming Problem
and the general concept of Linear Programming Problem. Once a LPP is formulated,
the next issue is to solve it for optimal values of decision variables, which leads to
maximization (say of profit) or minimization (e.g. cost). Simplex problem is a very
powerful tool for solving LPPs. It can handle any complexities in the LPPs
irrespective of the number and nature of the decision variables and constraints. The
application of Simplex Method allows not only the solution to the LPPs, but also
provides information which can be usefully employed by the manager in decision
making. Every LPP also has a mirror image problem which is also a linear
programming problem. This mirror LPP is called the Dual and the original problem is
called the Primal. The primal-dual relationship is very important for effective decision
making.

10.0​ Objectives

After going through this lesson you will be able to understand:

●​ Basics of Operations Research

●​ Graphical Method to solve the Linear Programming Problem
●​ Overview of Linear Programming Problem
●​ Basics of Simplex Algorithm and Duality Concepts
●​ How to apply simplex method in different types of LPP
●​ Primal-Dual Relationship
●​ Economic interpretation of Duality

10.1​ Introduction
Operations Research (OR) has become increasingly important in the face of
the fast moving technology and increasing complexities in business and
industry. It has become more difficult to allocate the available resources to the
various activities in a way that is more effective for the organization as a
whole. These kinds of problems can be solved using Operation Research. It
was first coined in 1940 in UK. In the beginning, it was used by military
services in World War II.

In India, Operation Research came into existence in 1949 with the opening of
an OR unit at the Regional research Laboratory at Hyderabad. In 1953 an OR

353
 unit was established in the Indian Statistical Institute, Calcutta, for the
application of OR methods in national planning and survey. In recent years
OR has had an increasingly great impact on the management of
organizations.

10.1.1 Definition of Operations Research

OR, in the most general sense, can be characterized as the application of

scientific methods, tools and techniques to problems involving the operations
of systems so as to provide those in control of the operations with optimum
solutions to the problems.

- Churchman, Ackoff, Arnoff

10.1.2 Nature of Operations Research

●​ OR involves research on operations.

●​ OR is applied to problems that concern how to conduct and coordinate the
operations (activities) within an organization.
●​ Research means that OR uses an approach that resembles the way research
is conducted in established scientific fields: observing and formulating
problems, gathering data, constructing model, solving and interpreting.

10.1.3 Features of Operations Research

Some significant features of OR are:

a.​ System (or executive) orientation: Operation research study is concern with
problems as a whole or as its system orientation. This means an activity by
any part of an organization has some effect on the activity of every other part.
The optimum operation of one part of a system may not affect the optimum
operation for some other part. Therefore to evaluate any decision, one must
identify all possible interactions and determine the impact on the organization
as a whole.

b.​ Inter-disciplinary Team Approach: OR is inter-disciplinary in nature and

requires a team approach to a solution of the problem. Since managerial
problems have economic, physical, psychological, biological, sociological and
engineering aspects, it requires a blend of people with expertise in the areas
of mathematics, statistics, engineering, economics, management, computer
science etc.

c.​ Scientific Approach: OR employs scientific method for the purpose of solving
problems. It is a formalized process of reasoning.

d.​ Uncovering new problems: Solution of an OR problem may uncover a number

of new problems. All these uncovered problems need not be solved at the
same time.

354
 e.​ Improvement in the Quality of Decision: OR gives bad answers to problems,
to which, otherwise, worse answers are given. This implies that by applying its
scientific approach, it can only improve the quality of solution but it may not be
able to give perfect solution.

f.​ Use of Computer: OR often requires a computer to solve the complex

mathematical model or to manipulate a large amount of data or to perform a
large number of computations that are involved.

g.​ Quantitative solution: OR provides the management with a quantitative basis

for decision making.

h.​ Human Factor: An OR study is incomplete without a study of human factors.

10.1.4 Methodology of Operations Research

The OR approach to problem solving consists of the following seven steps:

1.​ Formulating the Problem: The first step in OR is to develop a clear and
concise statement problem. After identification, formulation of the problem is
to be carried out. For this it is necessary to study comprehensively the
components of the problems
a.​ The environment
b.​ Decision maker,
c.​ The objective,
d.​ The alternative courses of action

2.​ Constructing a Mathematical Model: The next step is to build a suitable

mathematical model. A mathematical model should include mainly the
following three sets of elements
a.​ Decision variables and Parameters
b.​ Constraints or Restrictions, and
c.​ Objective function

3.​ Obtaining the Input Data: Once the mathematical model of the problem has
been formulated, the next step is to obtain the data to be used in the model as
input. Since the quality of data determines the quality of output, it is important
to obtain accurate and complete data

4.​ Deriving the solution from the Model: Having collected the input data, the next
step is to determine the values of decision variables that optimize the given
objective function. This deals with the mathematical calculations for obtaining
the solution to the model.

5.​ Validation of the Model: Validating a model requires determining if the model
can reliably predict the actual system’s performance. It also involves testing
the structural assumptions of the model to ascertain their validity. If during
validation, the solution cannot be implemented, one needs to find if some of

355
 the original constraints were incorrect and need to be modified. In such a
case, one must return to the problem formulation step and carefully make the
appropriate modifications to represent more realistic situation.

6.​ Establishing control over the solution: After testing the model and its solution,
the next step of the study is to establish control over the solution, by proper
feedback of the information on variables which deviated significantly. As soon
as one or more of the controlled variables change significantly, the solution
goes out of control. In such a situation the model may accordingly be
modified.

7.​ Implementation of the final results: Finally, the tested results of the model are
implemented to work. This involves a careful explanation of the solution to be
adopted and its relationship with the operating realities.

Self-Check Questions

Choose the correct alternative

1.​ Operations Research came into existence

a.​ In the year 1940
b.​ In the military context
c.​ During World War I
d.​ During World War II

2.​ Operations Research is

a.​ Applied decision theory
b.​ A scientific approach to problems solving for executive management
c.​ The science of use
d.​ All of the above

3.​ Operations Research approach is

a.​ Intuitive
b.​ Objective
c.​ Multi-disciplinary
d.​ All of the above

4.​ The scientific method in Operations Research consists of

a.​ Judgment Phase
b.​ Research Phase
c.​ Action Phase
d.​ All of the above

5.​ An optimization model in OR

a.​ Provides decision within its limited context,
b.​ Mathematically provides the best decision,
356
 c.​ Helps in evaluating various alternatives constantly,
d.​ All of the above

10.2​ Linear Programming Problem

Many business and economic situations are concerned with a problem of
planning activity. In each case, there are limited resources and the problem is
to make such a use of these resources so as to yield the maximum production
or to minimize the cost of production or to give the maximum profit. Such
problems are referred to as the problems of constrained optimization. Linear
programming (LP) is a technique for determining an optimum schedule of
interdependent activities in view of the available resources. Programming is
just another word for ‘planning’ and refers to the process of determining a
particular plan of action from amongst several alternatives. A manager can
use LP in following situations:

●​ How to allocate the advertising budget among various alternate advertising

media which have different degrees of effectiveness in reaching audience and
involve different costs?
●​ How should the HR manager of a hospital decide about the employment of
nurses that involve lowest cost and yet meets the requirements at different
times of the 24-hour-day?

LP deals with the optimization (maximization or minimization) of a function of

variables known as objective function, subject to a set of linear equations
and/or inequalities known as constraints. The objective function may be profit,
cost, production capacity or any other measure of effectiveness, which is to
be obtained in the best possible or optimal manner. The constraints may be
imposed by different resources such as market demand, production process
and equipment, storage capacity, raw material availability etc. Linearity means
a mathematical expression in which the expressions among the variables are
linear. Higher powers of the variables or other products do not appear in the
expressions for objective function as well as constraints

Few common areas of application of LP are as follows:

Table 10.1: Application Areas of LP

Industrial Application Misc. Applications Management Applications

-​ Product mix problems
-​ Blending problems -​ Media Selection problem -​ Diet Problem
-​ Production scheduling -​ Portfolio Selection problem -​ Agriculture Problem
problem -​ Profit planning problem -​ Flight Scheduling
-​ Trim Loss problem -​ Transportation problem Problem
-​ Assembly line balancing -​ Assignment problem -​ Environment Problem
-​ Make or by problem -​ Man power scheduling problem -​ Facilities location

357
10.2.1 Components of a LPP

A Linear Programming Problem (LPP) consists of three components, namely

the (i) decision variables (activities), (ii) the objective (goal), and (iii) the
constraints (restrictions).

(i)​ The decision variables refer to the activities that are competing one another
for sharing the resources available. These variables are inter-related in terms
of utilization of resources and need simultaneous solutions. All the decision
variables are considered as continuous, controllable and non-negative. The
non-negativity shows that LP deals with real-life situations for which negative
quantities are generally illogical.

(ii)​ A linear programming problem must have an objective which should be clearly
identifiable and measurable in quantitative terms. It could be of profit (sales)
maximization, cost (time) minimization, and so on. The relationship among the
variables representing objective must be linear.

(iii)​ There are always certain limitations (or constraints) on the use of resources,
such as labor, space, raw material, money, etc. that limit the degree to which
an objective can be achieved. Such constraints must be expressed as linear
inequalities or equalities in terms of decision variables.

10.2.2 Basic assumptions of a LPP Model

The following four basic assumptions are necessary for all linear programming
problems:

a.​ Certainty: In a LPP, it is assumed that all the parameters; such as availability
of resources, profit (or cost) contribution of a unit of decision variable and
consumption of resources by a unit decision variable must be known and
fixed. In other words, this assumption means that all the coefficients in the
objective function as well as in the constraints are completely known with
certainty and do not change during the period of study.

b.​ Divisibility (or continuity): This implies that solution values of the decision
variables and resources can take on any non-negative values, including
fractional values of the decision variables. For instance, it is possible to
produce 2.34 quintals of wheat or 12.12 thousand kilometers of cloth, so
these variables are divisible. But it is not possible to produce 2.6 refrigerators.
Such variables are not divisible and hence are to be assigned integer values.
When it is necessary to have integer variables, the integer programming
problem is considered to attain the desired values.

c.​ Proportionality: This requires the contribution of each decision variable in both
the objective function and the constraints to be directly proportional to the
value of the variable For example, if production of one unit of a particular
product uses 3 hours of a particular resource, then the production of 6 units of
that product uses 3 x 6, i.e., 18 hours of that resource.

358
 d.​ Additivity: The value of the objective function for the given values of decision
variables and the total sum of resources used, must be equal to the sum of
the contributions (profit or cost) earned from each decision variable and the
sum of the resources used by each decision variable respectively. For
example, the total profit earned by the sale of two products A and B must be
equal to the sum of the profits earned separately from A and B.

10.3​ Mathematical Formulation of Linear Programming Problem

First the given problem must be presented in linear programming form. This
requires defining the variables of the problem, establishing inter-relationships
between them and formulating the objective function and constraints.

The procedure for mathematical formulation of a linear programming problem

consists of the following steps:

Step 1: Study the given situation to find the key decisions to be made

Step 2: Identify the variables involved and designate them by symbols

xj (j=1,...,n)

Step 3: State the feasible alternatives: xj ≥ 0, for all j

Step 4: Identify the constraints in the problem and express them as linear
inequalities or equations, left hand side of which are linear functions
of the decision variables

Step 5: Identify the objective function and express it as a linear function of

the decision variables

Illustration 10.1 (Product Allocation Problem): A company has three

operational departments (weaving, processing and packing) with capacity to
produce three different types of clothes namely suitings, shirtings and
woollens yielding a profit of Rs. 2, Rs. 4 and Rs. 3 per meter respectively. One
meter of suiting requires 3 minutes in weaving, 2 minutes in processing and 1
minute in packing. Similarly one meter of shirting requires 4 minutes in
weaving, 1 minute in processing and 3 minutes in packing. One meter of
woolen requires 3 minutes in each department. In a week, total run time of
each department is 60, 40 and 80 hours for weaving, processing and packing
respectively.

Formulate the linear programming problem to find the product mix to

maximize the profit.

359
 Solution:​ The data of the problem is summarized below:
Departments Profit
Weaving Processing Packing (Rs. per
(in minutes) (in minutes) (in minutes) meter)
Suitings 3 2 1 2
Shirtings 4 1 3 4
Woollens 3 3 3 3
Availability (min) 60 × 60 40 × 60 80 × 60

Step 1:​ The key decision is to determine the weekly rate of production for
the three types of clothes

Step 2:​ Designate the weekly production of suiting, shirting and woolens by
x1 meters, x2 meters and x3 meters respectively.

Step 3:​ Since it is not possible to produce negative quantities, feasible

alternatives are sets of values of x1, x2 and x3 satisfying x1 ≥ 0, x2 ≥
0 and x3 ≥ 0.

Step 4:​ The constraints are the limited availability of three operational
departments. One meter of suiting requires 3 minutes of weaving.
The quantity being x1 meters, the requirement for suiting alone will
be 3x1 units. Similarly, x2 meters of shirting and x3 meters of woolen
will require 4x2 and 3x3 minutes respectively. Thus, the total
requirement of weaving will be 3x1 + 4x2 + 3x3, which should not
exceed the available 3600 minutes. So, the labor constraint
becomes

3x1 + 4x2 + 3x3 ≤ 3600.

​ ​ Similarly, the constraints for the processing department and

packing departments respectively are
2x1 + x2 + 3x3 ≤ 2400
and​ x1 + 3x2 + 3x3 ≤ 4800

Step 5:​ The objective is to maximize the total profit from sales. Assuming
that whatever is produced is sold in the market, the total profit (Z) is
given by the linear relation

Z = 2x1 + 4x2 + 3x3.

Thus the LPP can be put in the following mathematical format:

Find x1, x2 and x3 so as to

max​ Z = 2x1 + 4x2 + 3x3
s.t
3x1 + 4x2 + 3x3 ≤ 3600
​ ​ 2x1 + x2 + 3x3 ≤ 2400
x1 + 3x2 + 3x3 ≤ 4800
x1, x2, x3 ≥ 0

360
 Illustration 10.2: The manager of an oil refinery must decide on the optimal
mix of two possible blending processes of which the inputs and outputs per
production run are as follows:

The maximum amounts available of crude A and B are 200 units and 150
units respectively. Market requirements show that at least 100 units of
gasoline X and 80 units of gasoline Y must be produced. The profits per
production run from process 1 and process 2 are ` 300 and ` 400 respectively.
Formulate the process.
Input (Units) Output (Units)
Crude A Crude B Gasoline X Gasoline Y
1 5 3 5 8
2 4 5 4 4

Solution: Let x1 and x2 be the two blending processes. The LPP formulation
is
max ​ Z = 300x1 + 400x2
s.t.​ ​ 5x1 + 4x2 ≤ 200
​ ​ ​ ​ 3x1 + 5x2 ≤ 150
5x1 + 4x2 ≥ 100
8x1 + 4x2 ≥ 80​ ​ x1, x2 ≥ 0

Illustration 10.3: The ABC Electric Appliance Company produces two

products: Refrigerators and Ranges. Production takes place in two separate
departments. Refrigerators are produced in department I and Ranges in
department II. The company’s two products are produced and sold on a
weekly basis. The weekly production cannot exceed 25 Refrigerators in
department I and 35 Ranges in department II. Because of limited available
facilities in these two departments, the company regularly employs a total of
60 workers in two departments. A Refrigerator requires 2 Man-Week of labor
while a Range requires 1 Man-Week of labor. A refrigerator contributes a
profit of ` 60 and a Range contributes a profit of ` 40. How many units of
Refrigerators and Ranges should the company produce to realize a maximum
profit? Formulate the problem.

Solution: Let x1 and x2 units of Refrigerators and Ranges be produced. The

LPP formulation is
max​ Z = 60x1 + 40x2
s.t.​ ​ 2x1 + x2 ≤ 60
​ ​ ​ x1​ ≤ 25
x2​ ≤ 35​ ​ ​ ​ x1, x2 ≥ 0

Illustration 10.4: A furniture manufacturer company produces two types of

desks: Standard and Executive. These desks are sold to an office furniture
wholesaler and for all practical purposes; there is an unlimited market for any
mix of these desks, at least within the manufacturer’s production capacity.
Each desk has to go through four basic operations: cutting of the lumber,
joining of the pieces, pre-finishing and final finishing. Each unit of the standard
desk produced takes 48 minutes of cutting time, 2 hours of joining, 40 minutes
of pre finishing and 4 hours of final finishing. Each unit of the executive desk

361
 requires 72 minutes of cutting time, 3 hours of joining, 2 hours of pre finishing
and 5 hours 20 minutes of final finishing time. The daily capacity for each
operation amounts to 16 hours of cutting, 30 hours of joining, 16 hours of pre
finishing and 64 hours of final finishing time. The profit per unit produced is `
40 for the standard desk and ` 50 for the executive desk. Determine mix that
maximizes the revenue.

Solution: Let there be x1 units of Standard Desks and x2 units of Executive

Desks. The LPP formulation is
max ​ Z = 40x1 + 50x2
s.t.
​ ​ ​ 4/5x1 + 6/5x2 ≤ 16
​ ​ ​ 2x1 + 3x2 ≤ 30
2/3x1 + 2x2 ≤ 16
4x1 + 16/3x2 ≤ 64 ​​ x1, x2 ≥ 0

Illustration 10.5: A 24-Hours supermarket has the following minimal

requirement for cashiers:
Period 1 2 3 4 5 6
Time of Day 3-7 7-11 11-15 15-19 19-23 23-03
Min. Requirements 7 20 14 20 10 5

Period 1 follows immediately after period 6. A cashier works 8 consecutive

hours starting at the beginning of one of the six time periods. Determine a
daily employee worksheet which satisfies the requirement with the least
number of personnel. Formulate the problem as a LPP.

Solution: Let x1, x2, x3, x4, x5, x6 be the number of cashiers joining at the
beginning of the periods 1, 2, 3, 4, 5, 6. Then
max​ Z = x1 + x2 + x3 + x4 + x5 + x6
s.t.​ x1 ​ ​ ​ + x6 ​ ≥ 7
​ ​ ​ x1 + x2 ​ ​ ​ ≥ 20
​ ​ ​ x2 + x3 ​ ​ ​ ≥ 14
​ ​ ​ ​ x3 + x4 ​ ​ ≥ 20
​ ​ ​ ​ ​ x4 + x5 ​ ≥ 10
​ ​ ​ ​ ​ x5 + x6 ​ ≥ 5
x1 ,.., x6 ≥ 0

Illustration 10.6: The Agriculture Research Institute suggested to a farmer to

spread out at least 4800 kg of a special phosphate fertilizer and not least than
7200 kg of a special nitrogen fertilizer to raise productivity of crops in his
fields. There are two sources for obtaining these mixtures A and B. Both of
these are available in bags weighting 100 kg each and they cost ` 40 and ` 24
respectively. Mixture A contains phosphate and nitrogen equivalent of 20 kg
and 80 kg respectively while mixture b contains these ingredients equivalent
of 50 kg each.

Write this as LPP and determine how many bags of each type the farmer
should bring in order to obtain the required fertilizer at minimum cost.

362
 Solution: Let x1 be the number of type A and x2 be the number of type B
mixture. Then
min Z = 40x1 + 20x2
s.t.​ 20x1 + 50x2 ≥ 4800​
​ ​ ​ 80x1 + 50x2 ≥ 7200​ ​ x1, x2 ≥ 0

10.4​ Graphical Method

Once a problem is formulated as the mathematical model, the next step is to
solve the problem to get the optimal solution. A LPP with only two variables
presents a simple case, for which the solution can be derived using a
graphical method. Though in real life such small problems rarely encounter,
the graphical method provides a pictorial representation of the solution
process and a great deal of insight into the basic concepts used in solving
large LPPs.

Following are the steps to solve the problem by Graphical Method:

●​ Identify the problem- the decision variables, the objective function and the
constraints.
●​ Plot a graph that includes all the constraints and identify the feasible region.
●​ Obtain the point on the feasible region that optimizes the objective function-
this is the optimal solution.
●​ Interpret the result.

Illustration 10.7:​ max Z = 3x1 + 5x2

s.t. ​ x1 ≤ 4
​ ​​ ​ ​ ​ 2x2 ≤ 12
​ ​ ​ ​ ​ 3x1 + 2x2 ≤ 18​ ​ x1, x2 ≥ 0

Solution: The non-negativity x1, x2 ≥ 0, implies that values of the variables x1

and x2 can lie only in the first quadrant. The constraints are plotted with their
inequality sign changed into equality sign. The direction in which each
constraint holds good is then determined from the direction of the inequality
and is indicated by an arrow on its associated straight line. The constraint
conditions define the boundary of the region containing feasible solution.

For the problem, plot the 1st constraint by a line at x1 = 4, plot the 2nd
constraint by a line at 2x2 = 12 or x2 = 6 and for the 3rd constraint join the two
points (0,6) and (4,0). Now any point lying on or below the line satisfies the
three constraint equations. This is indicated by the direction of arrowheads as
shown in Figure 10.1.

The shaded area OACDE in the Figure satisfies all the three constraints and
also the non-negativity restrictions. This area is called the region of feasible
solutions or the feasible region. Any point in this shaded region is a feasible
solution to the given problem.

363
 To determine the optimal solution, the five vertices of the region OACDE are
O(0, 0), A(0, 6), C(2.5,6), D(4,3) and E(4,0).

Fig 10.1

Compute the value of z for the feasible solutions by substituting the

coordinates of O, A, C, D and E in the objective function equation

i.e. ​ Z0 = 0*3 + 0*5 = 0,

ZA = 0*3 + 6*5 = 30,

ZC = 2.5*3 + 6*5 = 36.5,

ZD = 4*3 + 3*5 = 27 and

ZE = 4*3 + 0*5 = 12

Now choose the one which gives a maximum value as optimal solution.

∴​ x1 = 2.5, x2 = 6 with max Z = 36.5

This will be interpreted as produce 2 units of Product A and 6 units of product

B to get maximum profit of ` 36.50.

Terminology

1.​ Solution: A solution is any specification of values for decision variables

(x1,…,xn)

2.​ Feasible Solution: Any solution to a general LPP which also satisfies the
non-negative restrictions of the problem is called a feasible solution to the
general LPP.

3.​ Infeasible Solution: A solution for which at least one constraint is violated is
called an infeasible solution

4.​ Feasible Region: Collection of all feasible solutions represents the feasible
region

364
 5.​ Optimal solution: The feasible solution that has the most favorable value of
the objective function is called an optimal solution. Most Favorable Solution
means largest value if objective function is to be maximized, or lowest value if
objective function is to be minimized

In the Example

1.​ Solution: Solution for this problem is O(0, 0), A(0, 6), B(0, 9), C(3,6), D(4.5, 4),
E(4, 0), F(2.5, 2.5), G(4, 0) and H(8, 0)

2.​ Feasible Solution: The solution O(0, 0), A(0, 6), C(2, 6), D(4, 3) and E(4, 0) is
a feasible solution as it satisfies all the constraints

3.​ Infeasible Solution: The solution B(0, 9), F(2.5, 2.5), G(4, 0) and H(8, 0) is
infeasible solution

4.​ Feasible Region: Collection of all feasible solutions i.e. the region OACDE
represents the feasible region

5.​ Optimal solution: The feasible solution C(2,6) has the maximum value of the
objective function, therefore is the optimal solution

Illustration 10.8
​ ​ max Z = 5x1+7x2
s.t.​ x1+ x2 ≤ 4
3x1+ 8x2 ≤ 24
10x1+ 7x2 ≤ 35​ ​ x1, x2 ≥ 0

Solution: The graph is shown in Figure 10.2.

Fig 10.2

The optimal solution is x1 =1.5, x2 =2.5, with the maximum Z value of 25.

Illustration 10.9:​ max Z = 2x1+2x2

s.t.​ x1 − x2 ≥ −1
​ ​ ​ ​ 0.5x1 + x2 ≤ 2,​ ​ x1, x2 ≥ 0

365
 Solution: The graph is shown in Figure 10.3.

Fig 10.3

From the graph it is clear that x1 = 0.5, x2 = 1.5, and Z = 4

Illustration 10.10:​ min Z = 4x1+ 3x2

s.t. x1 + 3x2 ≥ 9
2x1 + 3x2 ≥ 12
x1 + x2 ≥ 5, ​ ​ ​ x1, x2 ≥ 0

Solution: The graph is shown in Figure 10.4.

Fig 10.4

From the graph it is clear that x1 = 0, x2 = 6, and z = 18

Special Cases

1.​ Multiple Optimum Solutions: When the objective function is parallel to a

binding constraint (i.e., a constraint that is satisfied as an equation by the
optimal solution), the objective function will assume the same optimum value
at more than one solution point. For this reason they are called alternative
optima.

366
 Illustration 10.11:​ max Z = 3x1 + 2x2
s.t x1 ≤ 4
​ 2x2 ≤ 12
​ ​ 3x1 + 2x2 ≤ 18,​ ​ x1, x2 ≥ 0

Solution: The graph is shown in Figure 10.5.

Fig 10.5

From the graph it is clear that at x1 = 2, x2 = 6, and z = 18. Also at x1 = 4, x2 =

3, and z = 18. This implies the case of Multiple optimal solutions

2.​ Unbounded Solution: When the values of the decision variables may be
increased indefinitely without violating any of the constraints, the solution
space (feasible region) is unbounded. The value of objective function, in such
cases, may increase (for maximization) or decrease (for minimization)
indefinitely. Thus, both the solution space and the objective function value are
unbounded.

Illustration 10.12:​ min Z = 4x1+ 3x2

s.t., x1 + 3x2 ≥ 9
2x1 + 3x2 ≥ 12
x1 + x2 ≥ 5, ​ ​ ​ x1, x2 ≥ 0

Solution: The graph is shown in Figure 10.6.

367
 Fig 10.6

From the graph it is clear that values of the decision variables may be
increased indefinitely without violating any of the constraints, the solution
space (feasible region) is unbounded.

3.​ Infeasible Solution: When the constraints are not satisfied simultaneously,
the linear programming problem has no feasible solution. This situation can
never occur if all the constraints are of the ‘≤’ type.

Illustration 10.13: Solve the following LPP

max Z = x1 + x2
s.t.​ x1 + x2 ≤ 1
​ ​ ​ ​ -3x1 + x2 ≥ 3​​ ​ x1, x2 ≥ 0

Solution: The graph is shown in Figure 10.7.

Fig 10.7

From the graph it is clear that there is no feasible region and hence no
feasible solution.

4.​ Redundant Constraints: If a constraint when plotted does not form part of
the boundary making the feasible region of the problem, it is said to be
redundant constraint. Inclusion or exclusion of a redundant constraint does
not affect the optimal solution.

Illustration 10.14:
max Z = 40x1 + 35x2
s.t.​ 2x1 + x2 ≤ 60
​ ​ 4x1 + 3x2 ≤ 96​
4x1 +3.5 x2 ≤ 105 Redundant Constraint

​​ ​ x1, x2 ≥ 0

368
Self-Check Questions

Fill in the blanks

6.​ The relationship among the variables representing objective must be

___________.
7.​ The ___________ refer to the activities that are competing one another for
sharing the resources available.
8.​ A solution for which all the constraints are satisfied is called a __________
9.​ Collection of all feasible solutions represents the ____________
10.​When __________ the are not satisfied simultaneously, the linear
programming problem has no feasible solution

10.5​ General Linear Programming Problem

The general Linear Programming Problem (LPP) can be expressed as
follows:
Let Z be a linear function defined by
Z = c1x1 + c2x2 + …. + cnxn​ ​ ​ (10.1)
where cj’s are the constants.

Let (aij) be an m × n matrix and let (b1, …, bm) be a set of constants such that
​ ​ a11x1 + a12x2 + …. + a1nxn (≤, =, ≥) b1
​ ​ ​ ​ ​ ​ ​ ​ ​ ​ (10.2)
​ ​ am1x1 + am2x2 + …. + amnxn (≤, =, ≥) bm

and let​ ​ xj ≥ 0​​ ​ ​ ​ ​ ​ (10.3)

Then the problem of determining (x1, …, xn) which makes z a minimum (or
maximum) and which satisfies (10.2) and (10.3) is called a General Linear
Programming Problem.

For each constraint one and only one of the (≤, =, ≥) holds but the sign may
vary from one constraint to another. The general LPP can be stated in various
forms.

i.​ Compact Form of the general LPP

max (or min) z =

s.t

​ ​ ​ ​ ​ ​ ​ ​ ​ (10.4)

369
 and ​ ​ xj ≥ 0

where variables xj (j=1,…,n) are called decision variables,

cj, aij and bi (i=1,..,m) and (j=1,..,n) are the constants determined from
the statement of the problem

Constant cj represent the net unit contribution of decision variables xj

to the value of the objective function and are called objective
function coefficients

Constant bi denote the total availability of the ith resource and are
called stipulations and

Constant aij stands for the amount of resources and are called
structural coefficients

Also in general LPP m < n

ii.​ Matrix Form of the general LPP

max/min Z = cx
s. t.
​ Ax (≥, = or ≤) b​ ​ ​ ​ ​ ​ (10.5)
​ ​
and ​ x≥ 0

where

​ ​ A is a (m×n) matrix such that A = (aij)m×n =​ ​ ​

x is a (n×1) column vector such that x =

b is a (1×m) row vector such that b = ​

and ​

c is a (1×n) row vector such that c = ​

Here ​ ​ A is called the coefficient matrix

x is the decision vector,

b is the requirement vector,

370
 c is the cost (price or profit) vector of the LPP.

An element aij of the matrix indicates the amount of ith type of resource
necessary to manufacture one unit of product j. Hence, it may be said that
these elements represent the activity of the operational system.

10.6​ Canonical Form of Linear Programming Problem

After formulating the LPP, the next step is to obtain its solution. But before
using any method to obtain the solution, the problem must be available in a
particular form. Two forms are dealt with in this lesson are- the canonical form
and the standard form. The canonical form helps in dealing with duality theory
(discussed in this Section) and the standard form (discussed in next Section)
is used to develop the general procedure for solving any LPP.

The general formulation of LP discussed above can always be put in the

following form, called the canonical form:
max Z = cx
s.t.​ Ax ≤ b​​ ​ ​ ​ ​ ​ (10.6)
​ ​ ​ x≥ 0

The characteristics of this form are:

1.​ The objective form is of maximization type. In case the objective function is of
min type multiply it by (-1) to convert it to max type.

Example: ​ min Z = 3x1 + 5x2 + 3x3

is written as: max (-Z) = -3x1 - 5x2 - 3x3

2.​ All the constraint equations are of ≤ form except for the non-negative
constraints.
2.1 ​ A constraint equation of the type a11x1 + a12x2 + …. + a1nxn ≥ b1 is
changed to an ≤ inequality type by multiplying the constraint equation
by (-1) and changing the inequality.

Example:​ ​ 3x1 + 5x2 + 3x3 ≥ 7

is written as:​ ​ -3x1 - 5x2 - 3x3 ≤ -7

2.2​ A constraint equation of the type:

a11x1 + a12x2 + …. + a1nxn = b1
should be replaced by two weak inequalities ≥ and ≤ as
a11x1 + a12x2 + …. + a1nxn ≥ b1
and ​ a11x1 + a12x2 + …. + a1nxn ≤ b1
and the ≥ inequality must be changed to an ≤ inequality type as
discussed above.
371
 3.​ All the variables are non-negative.
A variable which is un-restricted in sign (i.e. positive, negative or zero) is
equivalent to the difference between two non-negative values.

Example: Let be un-restricted in sign, then:

,​ where ≥0

value of is positive, zero or negative depending upon whether is larger,

equal or smaller than .

10.7​ Standard Form of Linear Programming Problem

The general formulation of LP discussed above can always be put in the
following form, called the standard form:
max Z = cx
s.t.​ Ax = b​ ​ ​ ​ ​ ​ (10.7)
​ ​ ​ x ≥ 0​

The characteristics of this form are

1.​ All the constraints are expressed in the form of equations except for the
non-negative restrictions. The inequality constraints are changed to equality
constraints by adding or subtracting a no-negative variable from left hand side
of such constraints
2.​ The right hand side of each constraint equation is non-negative. If not, i.e.,
any b is negative, then multiply the equation by (-1) and change the
inequalities.
3.​ All the variables are non-negative. An unrestricted variable can be expressed
as a difference of two non-negative variables as explained earlier
4.​ The objective form is of maximization type, if not then convert it to
maximization type.

Any LPP can be put into standard form with the help of some elementary
transformations:

Standardization of the LP means to convert the inequalities of the constraints

into equations. The constraint equations are given as inequalities ≥, ≤, = or
‘-b’. Following are the ways to change these to equalities:

​ ​ Case I: Constraints of ‘≤’ Type

​ Let the constraints of a General LPP be:

372
 𝑛
∑ 𝑎𝑖𝑗𝑥𝑗 ≤ 𝑏𝑖 𝑖 = 1, 2, …, 𝑚
𝑗=1

𝑛
Then the non-negative variables xn+i which satisfy ∑ 𝑎𝑖𝑗𝑥𝑗 + 𝑥𝑛+𝑖 = 𝑏 are
𝑗=1 𝑖
called slack variables and are ≥ 0.

​ Example: ​ Consider the Problem (P1)

max Z = 40x1 + 35x2
s.t. ​ 2x1 + 3x2 ≤ 60
​ ​ ​ ​ 4x1 + 3x2 ≤ 96 x1, x2 ≥ 0

​ Consider the first inequality of the LP

​ ​ ​ 2x1 + 3x2 ≤ 60

​ To convert it into equation, add a variable x3 (as n = 3, ∴ n+1 = 4) on

the left hand side to get

​ ​ ​ 2x1 + 3x2 + x3 = 60

Variable x3 can vary from 0 to 60 (i.e. 0 ≤ x3 ≤60) depending upon the value of
2x1 + 3x2. The variable x3 is referred to as a slack variable because it takes up
any slack between the left hand side and the right hand side of the inequality
upon being converted into equation. Slack is the difference between the
amount available (bi) and the amount actually used. In the problem x3 is
indicating the amount of unused raw material.

Similarly x4 represents the labour hour not used. So the second constraint
equation

​ ​ ​ 2x1 + 3x2 ≤ 60

​ becomes

​ ​ ​ 2x1 + 3x2 + x3 = 60

All the slack variable (and other variables discussed next) must be added to
the objective function (i.e. z) equation. Therefore variables x3 and x4 must be
added in the z equation and for this the coefficients of the variables x3 and x4
must be determined. The coefficients represent the cost involved in not using
the raw material or labour hours. Thus it can be presumed that the unused
resources have no cost and therefore do not affect the profits. Thus the
coefficients assigned will be zero.

Upon the introduction of slack variables for the other functional constraints,
the original linear programming model (given by P1) can now be replaced by
the equivalent model (called the augmented form of the model) shown below:

373
 max Z = 40x1 + 35x2
s.t. ​ 2x1 + 3x2 + x3 = 60
​ ​ ​ ​ ​ 4x1 + 3x2 + x4 = 96 ​ ​ x1, x2, x3, x4 ≥ 0

This form is called the augmented form of the problem because the original
form has been augmented by some supplementary variables needed to apply
the simplex method.

Case II: Constraints of ‘≥’ Type

Let the constraints of a General LPP be:

​ ​ ; i = 1, 2…..m.

𝑛
Then the non-negative variables xn+i which satisfy ∑ 𝑎𝑖𝑗𝑥𝑗 − 𝑥𝑛+𝑖 = 𝑏 are
𝑗=1 𝑖
called surplus variables.

​ Example: Consider the following LP (P2)

min Z = 40x1 + 20x2
s.t.​ 20x1 + 50x2 ≥ 4800
​ ​ ​ 80x1 + 50x2 ≥ 7200​ ​ x1, x2 ≥ 0

To convert the inequality ≥ to equality, subtract the surplus x3. The surplus
represents the excess of what is generated (given by the left hand side of the
inequality) over the requirements (shown by right hand side value bi). In this
case finding the initial basic feasible solution is not possible as it will violate
the non-negativity constraint. Thus an artificial variable is x4 is also added to
the equation. The artificial variable is added only to get the starting solution.
So the constraints of the above problem change to
​ ​ ​ ​ 20x1 + 50x2 – x3 + x4 ​ = 4800
​ ​ ​ ​ 80x1 + 50x2 ​ ​ – x5 + x6 = 7200

Here variables x3 and x5 are the surplus variables and x4 and x6 are artificial
variables. Since artificial variables do not represent any quantity relating to the
decision problem, they must be driven out of the system and must not appear
in the final solution.

Thus an artificial problem is constructed by introducing the dummy variable,

i.e. the artificial variable into each constraint that needs one. Usual
non-negativity constraints are placed on these variables. The objective
function is modified to impose a penalty on having their value > 0. The
iteration forces the artificial variable to be 0 one at a time, until they all are
gone, after which the real problem is solved. Thus the augmented form of the
above problem becomes:
min​ z = 40x1 + 20x2 + x4 + x6
s.t.​ 20x1 + 50x2 – x3 + x4 ​ = 4800

374
​ ​ ​ 80x1 + 50x2 ​ ​ – x5 + x6 = 7200
​ ​ ​ ​ ​ ​ ​ ​ x1,…, x6 ≥ 0
​ Case III: Constraints of ‘=’ Type

Consider the following example where the constraints are already equations.

P3​ min Z = 40x1 + 20x2

s.t.
​ ​ ​ 20x1 + 50x2 = 48
​ ​ ​ 80x1 + 50x2 = 72​ ​ x1, x2 ≥ 0

In such a case also the artificial variable is added to get the initial basic
feasible solution.

Thus artificial variable is used when inequality is of form ≥ or =. This is

introduced just for the purpose of being the initial Basic Variable (BV) for that
equation.

Illustration 10.15: Write the following in augmented form

min Z = 40x1 + 20x2
s.t.​ 12x1 + 3x2 ≤ 48
​ ​ ​ 8x1 + 5x2 ≥ 55​ ​ x1, x2 ≥ 0

Solution: Convert the minimization type to maximization type. Thus the

Augmented Form is
max (-Z) = - 40x1 - 20x2 – x5
s.t.​ 12x1 + 3x2 +x3 ​ = 48
​ ​ ​ 8x1 + 5x2 -x4 + x5 = 55
​ ​ ​ ​ ​ x1,…, x5 ≥ 0

Illustration 10.16: Write the following in augmented form

max Z = 40x1 + 20x2
s.t.​ 12x1 + 3x2 ≤ 48
​ ​ ​ 8x1 + 5x2 ≤ - 55
x1 + 5x2 = 43
​ ​ ​ x1, x2 ≥ 0

Solution: Making right hand side positive and writing the Augmented Form
max Z = + 40x1 + 20x2 + x5 +x6
s.t.​ 12x1 + 3x2 +x3 ​ = 48
​ ​ ​ - 8x1 - 5x2 -x4 + x5 = 55
​ ​ ​ x1 + x2 ​ + x6 = 55
​ ​ ​ ​ ​ x1,…, x6 ≥ 0

10.8​ Theory of Simplex Method

10.8.1 Some Important Definitions

375
 Object Function: The linear function given by equation (10.1) which is to be
maximized or minimized is called objective function of the general LPP.

Constraints: The set of inequalities (10.2) are called the constraints of the
general LPP.

Non-negative restrictions: The set of inequation (10.3) is called the set of

non-negative restriction of the general LPP.

Example:​ max Z = 3x1 + 5x2

s.t. ​ x1 ≤ 4
​ ​ ​​ ​ 2x2 ≤ 12
​ ​ ​ ​ 3x1 + 2x2 ≤ 18​ ​ x1, x2 ≥ 0

Solution: A n–tuple (x1, x2,…,xn) of real numbers which satisfies the

constraints of a LPP is called a solution to be general LPP.

Basic Solution: Given a system of m simultaneous liner equations in n

unknowns (m < n)

Ax = b

Where A is an m x n matrix of rank m. We need to set n-m variables equal to

zero and solve for the remaining m variables. Such a solution is called a basic
solution to the given system of equations.

In other words, a solution obtained by setting any n variables (among m+n

variables) equal to zero and solving for remaining m variables is called a basic
solution. These m variables are called basic variables and the remaining n
variables are called non-basic variables

Example: Considering the augmented form for the Illustration 10.1, there are
only 3 equations and 5 variables. This cannot be solved. Therefore substitute
some variables as zero; the remaining variables are called basic variables.

Non-Basic Variables: are the variables that are set to zero.

​ Non-Basic Variables = Number of Variables – Number of Equations

​ ​ ​ = 5–3

= 2

​ ​ ∴ 2 Non-Basic Variables and 3 basic Variables

Properties of Basic Solution

1.​ Each variable is designated as either basic or non-basic variable.

2.​ No. of BV = No. of functional constraints

No. of NBV = Total no. of variables – No. of functional constraints.

376
3.​ Non-Basic Variables (NBV) are equal to zero.

4.​ Values of Basic Variables (BV) are obtained as the simultaneous solution of
system of equations.

5.​ Set of Basic Variable (BV) is referred to as Basis.

6.​ If the Basic Variable (BV) satisfies the non-negativity constraints, the basic
solution is a Basic Feasible Solution.

Basic Feasible Solution: A basic solution that satisfies the non-negativity

constraint (10.3) also is a basic feasible solution (BFS). All variables in a
(BFS) are ≥ 0.

Non-degenerate BFS: It is BFS in which all the m basic variables are positive
( > 0) and the remaining n variables are zero each.

Degenerate Solution: A basic solution to the system is called degenerate if

one or more of the basic variables vanish.

Associated cost vector: Let xB be a basic feasible solution to the LPP

max Z = cx
s.t. ​ Ax = b,
x>0
then the vector CB = (CB1, CB2, …, CBm) where CBi are components of c
associated with the basic variables, is called the cost vector associated with
the basic feasible solution xB.

The value of the objective function for the basic feasible solution xB is given
by

Z0 = CBxB
^
Improved Basic feasible solution: Let xB and 𝑥𝐵 be two feasible solutions to
^
the standard LPP. Then 𝑥𝐵 is said to be an improved basic feasible solution,
as compared to xB, if
^ ^
𝐶𝐵𝑥 ≥ 𝐶𝐵𝑥𝐵
𝐵

^ ^
Where 𝐶𝐵 is constituted of cost components corresponding to 𝑥𝐵.

Optimum Basic Feasible Solution: A basic feasible solution xB to the LPP

max Z = cx
s.t. ​ Ax = b,
x>0
is called an Optimal Basic Feasible Solution if Z0 = CBxB ≥ Z*.

377
Unbounded Solution: If the value of the objective function can be increased
or decreased indefinitely, the solution is called unbounded solution.

378
10.8.2 Geometric Interpretation of Simplex Method

The Simplex Method is an algebraic process. However, its underlying

concepts are geometric. To illustrate the general geometric concepts, consider
the following example. The graph for this example is given in Figure 10.8.

Fig 10.8

The five constraint boundaries and their points of intersection are highlighted
in this figure because they are the keys to the analysis. Here, each constraint
boundary is a line that forms the boundary of what is permitted by the
corresponding constraint. The points of intersection are the corner-point
solutions of the problem. The five that lie on the corners of the feasible
region—(0, 0), (0, 6), (2, 6), (4, 3), and (4, 0)—are the corner-point feasible
solutions (CPF solutions). The other three—(0, 9), (4, 6), and (6, 0)—are
called corner-point infeasible solutions.

In this example, each corner-point solution lies at the intersection of two

constraint boundaries. For any LP with n decision variables, two CPF
solutions are adjacent to each other if they share n - 1 constraint boundaries.
Such a line segment is referred to as an edge of the feasible region.

Since n = 2 in the example, two of its CPF solutions are adjacent if they share
one constraint boundary; for example, (0, 0) and (0, 6) are adjacent because
they share the x1 = 0 constraint boundary. Each CPF solution has two
adjacent CPF solutions, as enumerated in Table 10.2.
Table 10.2: The CPF and Adjacent CPF Solutions

379
 Adjacent CPF solution provides a very useful way of checking whether a CPF
solution is an optimal solution. Consider any linear programming problem that
possesses at least one optimal solution. If a CPF solution has no adjacent
CPF solutions that are better (as measured by Z), then it must be an optimal
solution.

Thus, for the example, (2, 6) must be optimal simply because its z = 36 is
larger than z for (0, 6) and z for (4, 3). This optimality test is the one used by
the simplex method for determining when an optimal solution has been
reached.

10.8.3 The Key Solution Concepts

Solution concept 1: The simplex method focuses solely on CPF solutions.

For any problem with at least one optimal solution, finding one requires only
finding a best CPF solution

Solution concept 2: The simplex method is an iterative algorithm

Solution concept 3: Whenever possible, the initialization of the simplex

method chooses the origin (all decision variables equal to zero) to be the
initial CPF solution.

Solution concept 4: Given a CPF solution, it is much quicker computationally

to gather information about its adjacent CPF solutions than about other CPF
solutions.

Therefore, each time the simplex method performs an iteration to move from
the current CPF solution to a better one, it always chooses a CPF solution
that is adjacent to the current one. No other CPF solutions are considered.
Consequently, the entire path followed to eventually reach an optimal solution
is along the edges of the feasible region.

Solution concept 5: After the current CPF solution is identified, the simplex
method examines each of the edges of the feasible region that emanate from
this CPF solution. The simplex method identifies the rate of improvement in z
that would be obtained by moving along the edge. Among the edges with a
positive rate of improvement in z, it then chooses to move along the one with
the largest rate of improvement in z. The iteration is completed by first solving
for the adjacent CPF solution at the other end of this one edge and then
relabeling this adjacent CPF solution as the current CPF solution for the
optimality test and (if needed) the next iteration.

At the first iteration of the example, moving from (0, 0) along the edge on the
x1 axis would give a rate of improvement in z of 3 (z increases by 3 per unit
increase in x1), whereas moving along the edge on the x2 axis would give a
rate of improvement in z of 5 (Z increases by 5 per unit increase in x2), so the
decision is made to move along the latter edge. At the second iteration, the
only edge emanating from (0, 6) that would yield a positive rate of
improvement in z is the edge leading to (2, 6), so the decision is made to
move next along this edge.

380
 Solution concept 6: A positive rate of improvement in z implies that the
adjacent CPF solution is better than the current CPF solution, whereas a
negative rate of improvement in z implies that the adjacent CPF solution is
worse. Therefore, the optimality test consists simply of checking whether any
of the edges give a positive rate of improvement in z. If none do, then the
current CPF solution is optimal.

In the example, moving along either edge from (2, 6) decreases Z. Since the
objective is to maximize Z, (2, 6) is optimal.

Self-Check Questions

Fill in the blanks

11.​The simplex method focuses solely on ___________ solutions

12.​A _____________________ in Z implies that the adjacent CPF solution is
better than the current CPF solution

10.9​ The Simplex Method

The graphical method is efficient in developing the conceptual framework
necessary for fully understanding the Linear Programming Process It suffers
from the great limitation that it can handle problem having only two decision
variables. However in real life situation problems with more than two decision
variables are encountered.

Solving the LPP means determining the set of non-negative values of

variables xj which will maximize z while satisfying the constraint equations.
The concept is simple but we have a set of m equations with (m+n) unknowns
and an infinite number of solutions are possible. There is a need for an
efficient and systematic procedure which will yield the desired solution in a
finite number of trials. An interactive procedure called Simplex Technique
helps to reach the optimum solution (if any) in a finite number of interactions.

The Simplex Method provides an efficient technique which can be applied for
solving LPPs of any magnitude.

The application of Simplex Algorithm for solving a LPP helps a manager to

answer following types of queries:

●​ What mix of products involved would yield the maximum profit

●​ Whether all resources would be fully utilized by the optimal mix, or are there
any unutilized resources and to what extend

●​ What mix of products would minimize the cost

381
 ●​ Does the problem have an alternate solution which is as good as the optimal
solution obtained? If yes, what is that solution?

●​ Does the problem have no solution that can meet all the requirements?

Simplex method, also called Simplex Technique was developed in 1947 by

G.B. Dantzig, an American Mathematician. It has the advantage of being
universal, i.e. any linear model for the solution exists, can be solved by it. In
principle, it consists of starting with a certain solution of which all that is basic
feasible, i.e. it satisfies the constraints as well as non-negativity conditions
(xj≥0, j = (1, 2,…, n).

The Simplex Algorithm is an iterative procedure for finding the optimal solution
to a LPP in a systematic manner. As discussed above, an optimal solution to
a problem, if exists, occurs at one of the corner points of the feasible region.
The simplex method, according to its iterative approach, results this optimal
solution from among the set of feasible solution to the problem. The algorithm
is indeed very efficient, because it considers only those feasible solutions
which are provided by the corner points, and that too not all of them.

The Simplex method always starts with the initial basic feasible solution, i.e.
origin, which is one of the corner points of the feasible region. This solution is
then tested i.e. it is ascertained whether improvement in the value of the
objective function is possible by moving to the next corner point of the feasible
region. If so, the solution at this point is obtained. This search for better corner
point is repeated, till after a finite number or trials, the optimal solution, if it
exists, is obtained.

10.9.1 Conditions for Applying Simplex Method

In order to apply the simplex method, the following conditions have to be

satisfied.

a.​ The right hand side of each of the constraint, i.e., bi should be non-negative. If
not then multiply both side by (-1) and change the inequality

b.​ Each of the decision variables of the problem should be non-negative

c.​ Objective function should be of maximization type. The objective function can
be of minimization type also. In that case the procedure reverses. Thus to
avoid confusion we consider here only maximization type of objective
function.

10.9.2 The Simplex Algorithm

Step 1:​ Check whether the objective function of the given LPP is of
maximization form. If it of minimization form, convert it to
maximization type by multiplying it by (-1):

​ ​ ​ ​ ​ i.e.​ min Z = max (-Z)

382
 Step 2: ​ Check whether all bi (i= 1, 2,..., m) are non-negative. If any one of bi
is negative, then multiply the corresponding constraint equation by
(-1) and change the inequality so as to get all bi non-negative.

Step 3: ​ Write the LPP in augmented form.

Step 4: ​ Initialization: Select the decision variables to be the initial non-basic

variable and slacks to be basic variables.

Step 5: ​ Compute

​ ​ ​ ​ Z = CB⋅ b​ ​ ​ ​ ​ ​ (10.8)

​ ​ ​ and​ ​​ ​ ​ ​ ​ (10.9)

​ ​ ​ ​ ​ ​ j = 1, 2, 3 ….n​ (10.10)

Step 6: ​ Optimality Test: The optimality test is based on the Solution

Concept 6.

The current basic feasible solution is optimal if and only if . If

solution is optimal, then stop, else go to step 7.

Note: If the problem is of minimization type, the optimal solution is

indicated when the values in 𝑗 rows are zero or positive.

Step 7: ​ Determining the direction of movement: In this step the Entering

Basic Variable is determined (based on Solution Concept 4) by
selecting the non–basic variable having the largest value.

EBV = max { }​ ​ ​ (10.11)

This column is called a pivot column. Put a box around it.

Step 8:​ Determining where to stop: This step determines how much to
increase the basic variable. When the value of non-basic variable is
increased from 0, it is converted to basic variable. This is called
entering basic variable as it enters the basis to improve the solution.
To determine the leaving basic variable, compute

​ ​ i=1, 2, …, m​ (10.12)

383
 and choose the minimum of them. Let the minimum of these ratios

be . Then the vector will leave the basis . The common

element , which is in the kth row and rth column is known as pivot
number and the kth row is called pivot row.

Step 9: ​ Solve for new Basic feasible solution (BFS):

Increasing one of the non-basic variables gives new BFS. Solve this
new BFS using elementary operations to get new solution. The
method for solving equation is based on Gauss Jordon (Gaussian
Elimination) Method. Now convert the pivot number to unity by
dividing its row by the leading element itself and all other elements
in its column to zero by making use of equation:

​ (10.13)

​​ and​ ​​ ​ ​ ​ ​ (10.14)

Step 10: ​ Go to step 2.

Illustration 10.17:​ max Z = 3x1 + 5x2

s.t.,​ x1​ ≤ 4​​
​ ​ ​ ​ x2 ≤ 12​
​ ​ ​ ​ 3x1 + 2x2 ≤ 18 ​ ​ x1, x2 ≥ 0

Solution:

Step 1:​ No need to change the objective as the problem is already of ‘max’
type

Step 2:​ No need to change right as all bi are positive

Step 3:​ Write the above LPP in augment form

max Z = 3x1 + 5x2
s.t.​ x1​ + x3​ = 4
​ ​ ​ ​ x2 ​ + x4​ = 12
​ ​ ​ ​ 3x1 + 2x2 ​ + x5 = 18 ​ x1,…, x5 ≥ 0

Here x1 and x2 are the decision variable and x3, x4 and x5 are slack
variables

Step 4: ​ Initialization: If a LP model has n variables and m constraints, then

m variables would be basic variables and n-m variables would be
non-basic. The basic variables form the basis.

384
 In the example, there are n = 5 variable and m = 3 constraints,
therefore m variables (slacks) are basic variables and n (5 – 3 ) = 2
are the non-basic variables. Thus the decision variables x1 and x2
are the non-basic variable and the slacks x3, x4, x5 are the basic
variable.

Construct the initial simplex tableau:

Table 10.3: Initial Tableau

cj 3 5 0 0 0
CB XB b a1 a2 a3 a4 a5
0 x3 4 1 0 1 0 0
0 x4 12 0 2 0 1 0
0 x5 18 3 2 0 0 1

Interpretation of the data in the above tableau (other simplex tableaus will
have similar interpretation)

i)​ The first row indicates the coefficients cj of the variable in the objective
function equation (10.1). These coefficients remain unchanged in the
subsequent tables. They represent the profit/cost per unit to the
objective function of each of the variables.

ii)​ Second row indicates the variables in the problem.

iii)​ The first column (CB column) represents the coefficients of the current
basic variables in the objective function equation. In the table this value
will be 0, 0, 0 as the cost of x3, x4, x5 are zero in the objective function
(i.e. z) equation.

Second column (XB) is the basic variables column. It represents the

basic variables of the current solution.

iv)​ The third column, the b column, is called quantity column. It indicates
the right hand value. In the example values 4, 12 and 18 will be
inserted in the column.

v)​ The coefficient matrix under non-basic variable x1, x2 represents their
coefficients aij in the constraints (10.2). Vector a1 represents the
coefficients of x1 i.e. 1, 0, 3 which are inserted vertically. Coefficients of
x2 are 0, 2, 2. Coefficients of x3 are 1, 0, 0, coefficients of x4 are 0, 1, 0
and coefficients of x5 are 0, 0, 1

Step 5: ​ Compute Z = CBXB

i.e.​​ Z = (0 * 4 + 0 * 12 + 0 * 18) = 0

Compute the net evaluation Δj using equations (10.10). The value is

inserted in the base row (i.e., the last row). To calculate this value
the formula is
Δ1 = c1 – CBa1 = 3 – (0 * 1 + 0 * 0 + 0 * 3) = 3

385
 Δ2 = c 2 – CBa2 = 5 – (0 * 0 + 0 * 2 + 0 * 2) = 5
Δ3 = c 3 – CBa3 = 0 – (0 * 1 + 0 * 0 + 0 * 0) = 0
Δ4 = c 4 – CBa4 = 0 – (0 * 10+ 0 * 1 + 0 * 0) = 0
Δ5 = c 5 – CBa5 = 0 – (0 * 0 + 0 * 0 + 0 * 1) = 0
This row determines whether or not the current solution is optimal.
Coefficients in this row represent the net profit (or net marginal)
improvement in the value of the objective function Z for each unit of
the respective column variable introduced into the solution. Δj
elements under slack variables (x3, x4, x5 in this case) are also
known as shadow price.

Step 6:​ Optimality Test: A positive Δj value indicates the amount by which
the profit will be increased if a unit of the corresponding variable is
introduced into the selection, a negative coefficient indicates the
amount by which the profit will be decreased if a unit of the
corresponding variable is introduced into the solution.

Since in the current solution we have two positive, this means that
the solution can be improved further, i.e. the solution is not optimal
and go to next step.

Step 7:​ Determining the entering basic variable: Each of the value in the Δj
row signifies the amount of increase in the objective function that
would occur if one unit of the variable were introduced in the
solution. Therefore select the variable that has the largest Δj value.

In the example, since Δ2 has the largest value, choose x2 to be the

entering variable and mark column a2 as the pivot column.

If more than one variable has the same maximum value of Δj, any
one of these variable may be selected arbitrarily as entering
variable.

Step 8:​ Determining the Leaving Basic Variable: To determine the leaving
basic variable divide elements under b column by the
corresponding elements of the pivot column (all the aj should be
positive only). Now the row having the minimum value of θ is
selected.

The θ column is also called replacement ratio column. These ratios

indicate the number of units of a variable (product) that can be
produced by trading all of the current level of basic variables.

In example θ values will be calculated as

θ1 value for 1 row will not be calculated as a1 is 0

θ2 = 12/2 = 6

θ3 = 18/2 = 9

386
 Leaving value ​ = min {6, 9} = 6

⇒ ​Variable x4 is the leaving basic variable. Therefore replace x4 by

x2 in the next table. Mark the x4 row as the pivot row. The
element which lies at the intersection of the pivot row and pivot
column is termed as pivot element.

Note that the next solution, which is to be obtained now, will be

feasible only if row containing minimum (non-negative) ratio is
selected as pivot row. In case a row containing higher
(non-negative) ratio is selected, the next solution will become
infeasible i.e., there will be some negative value in the b column.
Table10.4: Initial simplex tableau with all the computed values
cj 3 5 0 0 0
CB xB b a1 a2 a3 a4 a5 θ
0 x3 4 1 0 1 0 0 -

0 x4 12 0 2 0 1 0 6

0 x5 18 3 2 0 0 1 9

Z= 0
Δj 3 5 0 0 0

Step 9:​ Improving the Solution and construct the new tableau: In the new
tableau the variable x6 will be replaced by x2. To get the new basic
feasible solution,

a.​ Divide each element of the pivot row (including bi) by the pivot
element to get the corresponding values in the new table. The row
of values so derived is called the replacement row.

In the example the replacement row will be:

12/2​ 0​ 2/2​ 0​ 1/2​ 0
Or​ 6​ ​ 0​ 1​ 0​ 1/2​ 0

b.​ To make all the elements of the pivot column zero, perform the
following operation:

New row element = old row element – value of the × corresponding

​ ​ ​ ​ ​ ​ ​ pivot element replacement

387
 row values

​ For example to make the 3rd element of the pivot column zero,
preform the following operation:

​ ​ ​ ​ R3 (new table) = R3 (old table) – 2 * R2 (new table)

This gives 18​ 3​ 2​ 0​ 0​ 1​ → R3 (old table)

​ - 12​ 0​ 2​ 0​ 1​ 0​ →2*R2 (new table)
​ ​ 6​ 3​ 0​ 0​ -1 ​ 1​ → R3 (new table)

​ There will be no change in 1st Row of the old table as the element of
the pivot column is already zero.

​ Therefore the new tableau becomes

Table 10.5: Revised simplex tableau
cj 3 5 0 0 0
CB xB b a1 a2 a3 a4 a5 θ
0 x3 4 1 0 1 0 0 4
5 x2 6 0 1 0 1/2 0 -
0 x5 6 3 0 0 -1 1 2
Z= 30
Δj 3 0 0 -5/2 0

Step 10:​ Compute all Δj and go to step 2.

The next simplex tableau is:

Table 10.6: Optimal Solution simplex tableau

cj 3 5 0 0 0
CB xB b a1 a2 a3 a4 a5
0 x3 2 0 0 1 1/3 -1/3
5 x2 6 0 1 0 1/2 0
3 x1 2 1 0 0 -1/3 1/3
Z= 36
Δj 0 0 0 -3/2 -1

As all Δj ≤ 0, the solution is optimal.

∴​ x1 = 2, x2 = 6 and Z = 36

Illustration 10.18:​ min Z = x1 − 3x2 + 2x3

s.t.​ 3x1 − x2 + 2x3 ≤ 7​ ​
​ ​ ​ ​ 2x1 − 4x2 ≥ −12
​ ​ ​ ​ ​ − 4x1 + 3x2 + 8x3 ≤ 10​ x1, x2, x3 ≥ 0

Solution

Step 1:​ Convert the min type to max type

388
 max (- Z) = − x1 + 3x2 − 2x3

Step 2:​ Change the 2nd constraint equation to make the right hand side
positive

​ ​ ​ − 2x1 + 4x2 ≤ 12

Step 3:​ Write the above LPP in augment form

max (- Z) = − x1 + 3x2 − 2x3
s.t.​ 3x1 − x2 + 2x3 + x4 ​ = 7
​ ​ − 2x1 − 4x2 − ​ + x5 ​ = 12
​ ​ ​ − 4x1 + 3x2 + 8x3 ​ ​ + x6 = 10​ ​ x1,…, x6 ≥ 0

Step 4: ​ Initialization: The slacks x4, x5, x6 are the basic variable. Construct
the initial table

Step 5: ​ Compute Z = CBx and the net evaluation Δj using equations (10.10)
cj -1 3 -2 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x4 7 3 -1 2 1 0 0 -
0 x5 12 -2 4 0 0 1 0 12/4=3
0 x6 10 -4 3 8 0 0 1 10/3=3.3
3
Z= 0
Δj -1 3 -2 0 0 0

Step 6:​ Since in the current solution has a positive value of Δj, the solution
is not optimal. Go to next step.

Step 7:​ Select the variable that has the largest Δj value. As Δj is maximum
x2 is the entering basic variable. Mark a2 as the pivot column.

Step 8:​ Calculate θ (all aj > 0) and choose the row having the minimum
value of θ. Thus x6 is the leaving basic variable. Mark the x6 row as
the pivot row. The element which lies at the intersection of the pivot
row and pivot column is termed as pivot element.

Step 9:​ Improving the Solution and construct the new tableau: To get the
new basic feasible solution,

a.​ Divide Row 2 of the old table by 4 to make the pivot element 1.

3​ -1/2​ 1​ 0​ 0​ 1/4​ 1

b.​ To make all the elements of the pivot column zero, perform the
following operation:

​ ​ ​ ​ R1 (new table) = R1 (old table) – 1 * R3 (new table)

This gives ​ 7​ 3​ -1​ 2​ 1​ 0​ 0

​ ​ + 3​ 1/2​ 1​ 0​ 0​ 1/4 1

389
 ​ ​ ​ 10​ 5/2​ 0​ 2​ 1​ 1/4​ 0

and​ ​ R2 (new table) = R2 (old table) + 4 * R3 (new table)

This gives ​ 10​ -4​ 3​ 8​ 0​ 0​ 1

+ 9​ -3/2​ 3​ 0​ 0​ 3/4​ 0

​ ​ ​ 1​ -5/2​ 0​ 8​ 0​ -3/4​ 1

​ ​ Therefore the new tableau becomes

cj -1 3 -3 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x4 10 5/2 0 2 1 1/4 0 10*2/5=
4
3 x2 3 -1/2 1 0 0 1/4 0 -
0 x6 1 -5/2 0 8 0 -3/4 1 -
Z= -9
Δj 1/2 0 -2 0 -3/4 0

Step 10:​ Compute all Δj and go to step 2.

The next simplex tableau is:

cj -1 3 -3 0 0 0
CB XB b a1 a2 a3 a4 a5 a6
-1 x1 4 1 0 4/5 2/5 1/10 0
3 x2 5 0 0 2/5 1/5 3/10 14/5
0 x6 11 0 1 10 2/5 -1/2 3/5
Z= 11
Δj 0 0 -12/5 -1/5 4/5 0

As all Δj ≤ 0, the solution is optimal.

∴ ​ x1 = 4, x2 = 5 and Zmin = 11

​ ​ ​ Thus Zmax = -11

Self-Check Questions

State True / False

13.​The Simplex Algorithm is iterative procedure for finding solution to an LPP.

14.​To solve a LPP by Simplex Algorithm all the variable in it must be
non-negative
15.​In improving the solution, the pivot element may be positive, negative or zero
16.​Surplus variable cannot appear in the basis
17.​In solving a LPP, every constraint which involves equality should be replaced
by equivalent by a pair of inequalities.

390
10.10 Artificial Variable Techniques
In the above problems, the constraints were of ≤ type with non-negative right
hand side. There are, however, LPPs where the constraints are of ≥ or = type.
In such case surplus is subtracted to convert it into equality, as it represents
the excess of what is generated (given by left hand side of the value) over the
requirement (shown by the right hand side of the inequality). Since the surplus
is subtracted therefore the initial solution does not exist as it violates the
non-negativity constraint. Thus the artificial variable is added for obtaining an
initial solution. These are required for the constraints of ≥ or = type. Since
artificial variables do not represent any quantity relating to the decision
problem, they must be driven out of the system and must not show in the final
solution. To solve such types of LPPs there are two methods:
1.​ Big – M Method
2.​ Two Phase Method

10.10.1 Big M Method

The artificial variables are not real, these are introduced for the limited
purpose of obtaining an initial solution. Since the artificial variables do not
represent any quantity relating to the decision problem, they must be driven
out of the system first. This can be ensured by assigning extremely high cost
to them. Generally a value M is assigned to each artificial variable, where M
represents a number higher than any finite number. In case of minimize, add
(M) and in case of maximize, add (-M).

Illustration 10.19​ min Z = 40x1 + 24x2

s.t.​ 20x1 + 50x2 ≥ 4800
​ ​ ​ ​ ​ 80x1 + 50x2 ≥ 7200​ ​ x1, x2 ≥ 0

Step 1: ​ Convert the objective function to max type

Step 2: ​ All the bi are positive hence no change

Step 3:​ Write the LPP in augmented form

max (-Z) = ​ − 40x1 − 24x2 – x4 – x6
s.t.​​ 20x1 + 50x2 − x3 + x4​​ = 4800
​ ​ ​ ​ ​ 80x1 + 50x2 ​ − x5 + x6 = 7200
​ ​ ​ ​ ​ ​ ​ ​ ​ x1, …., x6 ≥ 0

cj -40 -24 0 -M 0 -M θ
CB XB b a1 a2 a3 a4 a5 a6
-M x4 4800 20 50 -1 1 0 0 96
-M x6 7200 80 50 0 0 -1 1 144
Z= -12000
M
Δj -40 -24 -M 0 -M 0
+100M +100M

391
 cj -40 -24 0 -M 0 -M θ
CB XB b a1 a2 a3 a4 a5 a6
-24 x2 96 2/5 1 -1/50 0 1/50 0 240
-M x6 2400 60 0 1 -1 -1 1 40
Z= -2304
-2400M
Δj -152/5 0 -12/25 12/2 -M 0
+60M +M 5
-2M

cj -40 -24 0 -M 0 -M θ
CB XB b a1 a2 a3 a4 a5 a6
-24 x2 0 1 -2/75 1/150 -1/150 -
80 2/75
-40 x1 1 0 1/60 -1/60 -1/60 1/60 2400
40
Z= 352
0
Δj 0 0 2/75 -M-2 -38/75 -M-38/
175 75

cj -40 -24 0 -M 0 -M
CB XB b a1 a2 a3 a4 a5 a6
-24 x2 144 8/5 1 0 0 -1/50 1/50
0 x3 2400 60 0 1 -1 -1 1
Z= -3456
Δj -8/5 0 0 -M -12/25 -M+12/25

The optimal solution for the problem is:

x1 = 0, x2 = 144 with Zmin = ` 3,456

Also x3 = 2400. This indicates the surplus phosphate ingredient obtained by

buying the least cost mix.

10.10.2 Two Phase Method

Difficulty with Big M method: When solving the LPP using computers, what
value should be given to M? Thus Two Phase Method is used.

This method separates the solution procedure into two phases. In Phase I all
the artificial variables are eliminated from the basis. If a feasible solution is
obtained in this phase, which has no artificial variable in basis in the final
table, then proceed to Phase II. In this phase, use the solution from the Phase
I as the initial basic feasible solution and use the Simplex Method to
determine the optimal solution.

The Algorithm

Phase I

392
 In this phase an initial BFS is found for the artificial LP. To do this an artificial
objective function is created which is the sum of all the artificial variables. This
problem is solved using the Simplex Algorithm.

Phase II

The basic feasible solution found at the end of the Phase I is now used as a
starting solution for the original problem. In other words, the final table of
Phase I becomes the starting table of Phase II in which the artificial objective
function is replaced by the original objective function and the artificial variable
column(s) are deleted from the simplex tableau. Now find the solution for this
problem using simplex algorithm.

Illustration 10.20:​ min Z = 40x1 + 24x2

s.t.​ 20x1 + 50x2 ≥ 4800
​ ​ ​ ​ ​ 80x1 + 50x2 ≥ 7200​ ​ x1, x2 ≥ 0
Step 1: ​ Convert the objective function to max type

Step 2: ​ All the bi are positive hence no change

Step 3:​ Write the LPP in augmented form

max (-Z) = − 40x1 − 24x2 – x4 – x6
s.t.​​ 20x1 + 50x2 − x3 + x4​​ = 4800
​ ​ ​ ​ 80x1 + 50x2 ​ − x5 + x6 = 7200 ​ x1, …., x6 ≥ 0

Phase I
Step 4: ​ Modify the augmented form by assigning zero coefficients to each
decision variables (x1 and x2) and to the surplus variables (x3 and
x5) and assign a unit coefficient to each of the artificial variables
(since it is a max problem, the coefficient of each of the artificial
variable shall be -1). Thus the objective function for the Phase I
becomes
Max (-Z) = – x4 – x6

Step 5:​ Solve this artificial problem by applying the usual Simplex Method.
cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
-1 x4 4800 20 50 -1 1 0 0 240
-1 x6 7200 80 50 0 0 -1 1 90
Z= -12000
Δj 100 100 1 1 0 0

cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
-1 x4 3000 0 75/2 -1 1 1/4 -1/4 80
0 x1 90 1 5/8 0 0 -1/80 1/80 144
Z= -3000
Δj 0 75/2 0 0 1/4 -5/4

393
 cj 0 0 0 -1 0 -1
CB XB b a1 a2 a3 a4 a5 a6
0 x2 80 0 1 -2/75 2/75 1/150 -1/150
0 x1 40 1 0 1/60 -1/60 -1/60 -1/60
Z= 0
Δj 0 0 0 -1 0 -1

This is the optimal solution table for the Phase I. Now proceed to Phase II.
Phase II

Step 6: ​ Reproduce the Simplex Table after replacing cj row by the

respective coefficients from the objective functions of the original
problem, and deleting the columns a4 and a6

Step 7: ​ Solve this problem (tableau) using the usual Simplex Procedure
and obtain the optimal Solution.
cj -40 -24 0 0 θ
CB XB b a1 a2 a3 a5
-24 x2 80 0 1 -2/75 1/150 -
-40 x1 40 1 0 1/60 -1/60 2400
Z= -3520
Δj 0 0 2/75 -38/75
​ cj -40 -24 0 0
CB XB B a1 a2 a3 a5
-24 x2 144 8/5 1 0 -1/150
0 x3 2400 60 0 1 -1
Z= -3456
Δj -8/5 0 0 -12/25

The optimal solution for the problem is:

x1 = 0, x2 = 144 with Zmin = ` 3,456

Problems with mixed constraints

Illustration 10.21​ max Z = 2x1 + 4x2
s.t.​ 2x1 + x2 ≤ 18
​ ​ ​ ​ ​ 3x1 + 2x2 ≥ 30
​ ​ ​ ​ ​ x1 + 2x2 = 26​ ​ ​ x1, x2 ≥ 0

Augmented Form
max Z = 2x1 + 4x2 – x4 – x6
s.t.​ 2x1 + x2 + x3 ​ ​ ​ = 18
​ ​ ​ ​ 3x1 + 2x2 ​ – x4 + x5 ​ = 30
​ ​ ​ ​ x1 + 2x2 ​ ​ + x6 = 26​ ​ x1, …, x7 ≥ 0

Phase I
max Z = – x4 – x6
s.t.​ 2x1 + x2 + x3 ​ ​ ​ = 18
​ ​ ​ ​ 3x1 + 2x2 ​ – x4 + x5 ​ = 30

394
​ ​ ​ ​ x1 + 2x2 ​ ​ + x6 = 26​ ​ x1, …, x7 ≥ 0

cj 0 0 0 0 -1 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x3 18 2 1 1 0 0 0 18
-1 x5 30 3 2 0 -1 1 0 15
-1 x6 26 1 2 0 0 0 1 13
Z= -56
Δj 4 4 0 1 0 0

cj 0 0 0 0 -1 -1 θ
CB XB B a1 a2 a3 a4 a5 a6
0 x3 5 3/ 0 1 0 0 -1/2 10/3
2
-1 x5 4 2 0 0 -1 1 -1 2
0 x2 13 1/ 1 0 0 0 1/2 26
2
Z= -4
Δj 2 0 0 -1 0 -2
​

395
 cj 0 0 0 0 -1 -1
CB XB b a1 a2 a3 a4 a5 a6
0 x3 2 0 0 1 3/4 -3/4 1/4
0 x1 2 1 0 0 -1/2 1/2 -1/2
0 x2 12 0 1 0 1/4 -1/4 3/4
Z= 0
Δj 0 0 0 0 -1 -1

This is the optimal solution table for the Phase I LPP.

​ Phase II
cj 2 4 0 0
CB XB b a1 a2 a3 a4
0 x3 2 0 0 1 3/4
2 x1 2 1 0 0 -1/2
4 x2 12 0 1 0 1/4
Z= 52
Δj 0 0 0 0

The optimal solution for the problem is:

x1 = 2, x2 = 12 with Z = 52

10.11 Special Cases in LPP

10.11.1 Multiple Optimal Solutions

As seen in the graphical method an LPP may or may not be unique. This
happens when the objective function (cost-profit line) is parallel to one of the
constraint. In Simplex Method, notice that for all the basic variables, Δj = 0,
but for the non-basic variables, Δj ≠ 0. An optimal solution with Δj = 0 for basic
variables and Δj ≠ 0 for the non-basic variables indicates the solution is
optimal in the sense that no other solution to the given problem exists, which
yields an identical value of the objective function. When a non-basic variable
in an optimal solution has a zero value for Δj, then the solution is not unique,
and multiple solutions exist.

Illustration 10.22:​ max Z = 8x1 + 16x2

s.t​ x1 + x2 ≤ 200
​ ​ ​ ​ ​ x2 ≤ 125​
​ ​ ​ ​ 3x1 + 6x2 ≤ 900​ ​ x1, x2 ≥ 0

Solution: Augmented Form

​ ​ ​ max Z = 8x1 + 16x2
s.t​ x1 + x2 + x3 ​ ​ = 200
​ ​ ​ ​ x2 + x4 ​ =125​
​ ​ ​ 3x1 + 6x2 + x5 = 900​ ​ ​ x1, x2 ≥ 0

396
 cj 8 16 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 200 1 1 1 0 0 200
0 x4 125 0 1 0 1 0 125
0 x5 900 3 6 0 0 1 150
Z= 0
Δj 8 16 0 0 0

cj 8 16 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 75 1 0 1 -1 0 75
16 x2 125 0 1 0 1 0 -
0 x5 150 3 0 0 -6 1 50
Z= 200
0
Δj 8 0 0 -16 0

cj 8 16 0 0 0
CB XB b a1 a2 a3 a4 a5
0 x3 25 0 0 1 1 -1/3
16 x2 125 0 1 0 1 0
8 x1 50 1 0 0 -2 1/3
Z= 240
0
Δj 0 0 0 0 -8/3

All the Δj ≤ 0, this indicates that the solution is optimal solution with x1 = 50, x2
= 125 and Z = 2400.

Notice that in the above tableau, one of the non-basic variable x4 has Δj = 0.
This variable has a zero net contribution and therefore it can be included in
the basis without changing the value of the objective function. Now take x4 to
be the pivot column and perform the calculations.

cj 8 16 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 25 0 0 1 1 -1/3 25
16 x2 125 0 1 0 1 0 -
8 x1 50 1 0 0 -2 1/3 -
Z= 240
0
Δj 0 0 0 0 -8/3

cj 8 16 0 0 0
CB XB b a1 a2 a3 a4 a5
0 x4 25 0 0 1 1 -1/3
16 x2 100 0 1 -10 0 1/3
8 x1 100 1 0 2 0 -1/3
Z= 2400
Δj 0 0 0 0 -8/3

397
 Now the solution is optimal with x1= 100, x2=100 and Z = 2400. Again the
non-basic variable x3 =0 indicates that the problem has multiple optimal
solutions.

Illustration 10.23:​ max Z = 6x1 + 4x2

​ ​ ​ s.t.​ 2x1 + 3x2 ≤ 30
​ ​ ​ ​ ​ 3x1 + 2x2 ≤ 24​
​ ​ ​ ​ x1 + x2 ≥ 3​​ x1, x2 ≥ 0

Solution:

Augmented Form:
max Z = 6x1 + 4x2 – x6
​ ​ s.t.​ 2x1 + 3x2 + x3​​ ​ = 30
​ ​ ​ ​ 3x1 + 2x2 ​ + x4​ ​ = 24
​ ​ ​ x1 + x2 ​ – x5 +x6 ​ = 3 x1, …, x6 ≥ 0

Phase - I
max Z = – x6
​ ​ s.t.​ 2x1 + 3x2 + x3​​ ​ = 30
​ ​ ​ ​ 3x1 + 2x2 ​ + x4​ ​ = 24
​ ​ ​ x1 + x2 ​ – x5 +x6 ​ = 3 x1, …, x6 ≥ 0

cj 0 0 0 0 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x3 30 2 3 1 0 0 0 15
0 x4 24 3 2 0 1 0 0 8
-1 x6 3 1 1 0 0 -1 1 3
Z= -3
Δj 1 1 0 0 -1 0

cj 0 0 0 0 0 -1
CB XB b a1 a2 a3 a4 a5 a6
0 x3 24 0 1 1 0 2 -2
0 x4 15 0 -1 0 1 3 -3
0 x1 3 1 1 0 0 -1 1
Z= 0
Δj 0 0 0 0 0 -1

Optimal solution for phase I. Proceed to Phase II

Phase II
cj 6 4 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 24 0 1 1 0 2 12
0 x4 15 0 -1 0 1 3 5
6 x1 3 1 1 0 0 -1 -
Z= 18
Δj 0 0 0 0 6

cj 6 4 0 0 0

398
 CB XB b a1 a2 a3 a4 a5
0 x3 14 0 1 1 -2/3 0
0 x5 5 0 -1/3 0 1/3 0
6 x1 8 1 2/3 0 1/3 1
Z= 48
Δj 0 0 0 -2 0

Therefore, the solution is optimal with

​ ​ ​ Z = 48, ​ x1 = 8 and x2 = 0

As Δj of x2 (which is a non-basic variable) is zero, this indicates multiple

optimal solutions.

10.11.2 Tie for entering basic variables

The non-basic variable having the largest Δj value is chosen as an entering

basic variable. In case two or more non-basic variable have largest Δj,
selection of entering basic variable may be arbitrary, (though a wrong choice
may increase the number of iterations to reach the optimal solution).

Consider Illustration 10.20, the initial tableau shows that 2 variables x1 and x2
have the same largest value (=100) of Δj. Choose any of these to be the
entering basic variable.

10.11.3 Degeneracy

Degeneracy occurs in LPP when one or more of the basic variable assumes a
value of zero. This means that in the table there will be a tie between leaving
basic variable, i.e., two leaving basic variables have same minimum ratio.
Break the tie arbitrary.

Illustration 10.24​ max Z = 28x1 + 30x2

s.t.​ 6x1 + 3x2 ≤ 18
​ ​ ​ ​ ​ 3x1 + x2 ≤ 8
​ ​ ​ ​ ​ 4x1 + 5x2 ≤ 30​ x1, x2 ≥ 0

Augmented Form
max Z = 28x1 + 30x2
s.t.​ 6x1 + 3x2 + x3​​ = 18
​ ​ ​ ​ 3x1 + x2 + x4​ = 8
​ ​ ​ ​ 4x1 + 5x2 ​ + x5 = 30​ ​ ​ x1, …, x5 ≥ 0

cj 28 30 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 18 6 3 1 0 0 6

0 x4 8 3 1 0 1 0 8
0 x5 30 4 5 0 0 1 6
Z= 0
Δj 28 30 0 0 0

399
This is a degenerate solution as there are two leaving variables having
minimum ratio. Considering x3 as leaving basic variable we get the following
solution:

400
 cj 28 30 0 0 `0
CB XB b a1 a2 a3 a4 a5
30 x2 6 2 1 1/3 0 0
0 x4 2 1 0 -1/3 1 0
0 x5 0 -6 0 -5/3 0 1
Z= 180
Δj -32 0 -10 0 0

This is the optimal solution with Z= 180, x1 = 0 and x2 = 18. Here variable x5
and the solution is degenerate.

Now considering x5 as leaving basic variable we get the following solution:

cj 28 30 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 18 6 3 1 0 0 6
0 x4 8 3 1 0 1 0 8
0 x5 30 4 5 0 0 1 6
Z= 0
Δj 28 30 0 0 0
Initial Solution

cj 28 30 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 0 18/ 0 1 0 -3/5 0
5
0 x4 2 11/ 0 0 1 -1/5 10/11
5
30 x2 6 1 0 0 1/5 15/2
4/5
Z= 18
0
Δj 4 0 0 0 -6

cj 28 30 0 0 0
CB XB b a1 a2 a3 a4 a5
0 x1 0 1 0 5/18 0 -1/6
0 x4 2 0 0 -11/18 1 1/6
30 x2 6 0 1 - 2/9 0 1/3
Z= 18
0
Δj 0 0 -10/9 0 -16/
3
​
​ This is optimal with x1 = 0, x2 = 6 and Z = 180

Some observations follow:

●​ In the above tableau the variable x5 has a zero value although it is in the
basis. Thus the variable x5 and the solution are degenerate.
●​ Whenever there is tie in the replacement ratios for determining outgoing
variable, the tableau would give a degenerate solution.

401
 ●​ We know that successive simplex tableau represents improvements in the
value of the objective function (increases are observed for the
maximization and decreases for the minimization problems). However,
when the outgoing variable happens to be a degenerate variable, the
objective function value in the next tableau does not change.

A problem may temporarily be degenerate i.e., initially we may get a

degenerate solution, but later on degeneracy disappears.

Illustration 10.25:​ max Z = 5x1 + 2x2

s.t.​ 4x1 + 2x2 ≤ 16
​ ​ ​ ​ ​ 3x1 + x2 ≤ 9
​ ​ ​ ​ ​ 3x1 - x2 ≤ 9​ ​ x1, x2 ≥ 0

Augmented Form
max Z = 5x1 + 2x2
s.t.​ 4x1 + 2x2 + x3​ = 16
​ ​ ​ ​ 3x1 + x2 ​ + x4​ = 9
​ ​ 3x1 - x2 ​ + x5 = 9​ ​ x1, …, x5 ≥ 0

cj 5 2 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 16 4 2 1 0 0 4
0 x4 9 3 1 0 1 0 3
0 x5 9 3 -1 0 0 1 3
Z= 0
Δj 5 2 0 0 0

cj 5 2 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 4 0 2/3 1 -4/3 0 6
5 x1 3 1 1/3 0 1/3 0 9
0 x5 0 0 -2 0 -1 1 -
Z= 15
Δj 0 1/3 0 -5/3 0

Here x1 = 3 and x2 = 0 with Z = 15 is degenerate with the basic variable x5 =0

also begin degenerate. Also it is non-optimal. Thus the solution is to be
improved.
cj 5 2 0 0 0
CB XB b a1 a2 a3 a4 a5
2 x2 6 0 1 3/2 -2 0
5 x1 1 1 0 -1/2 1 0
0 x5 12 0 0 3 -5 1
Z= 17
Δj 0 0 -1/2 -1 0

Now this is non-degenerate optimal solution with x1 = 1, x2 = 6 and Z = 17.

402
10.11.4 No leaving variable (Unbounded Solution)

An LPP is said to have unbounded solution if its objective function value can
be increased (in case of a maximization problem) or can be decreased (in
case of a minimization problem) without limits. In simplex tableau, there may
be a situation where all the θ values are not non-negative, i.e., it is not
possible to find the leaving basic variable as every coefficient in the pivot
column is either negative or zero.. This indicates that the entering basic
variable could be increased indefinitely (Unbounded z). Such problem has an
unbounded solution.

Illustration 10.26:​ max Z = 10x1 + 20x2

​ ​ ​ ​ s.t.​ 2x1 + 4x2 ≥ 16
​ ​ ​ ​​ x1 + 5x2 ≥ 15​ x1, x2 ≥ 0

Solution:

Augmented Form
max Z = 10x1 + 20x2 – x4 - x6
s.t​ .​ 2x1 + 4x2 - x3 + x4​ = 16
​ ​ ​ ​ x1 + 5x2 ​ - x5 + x6 = 15​ ​ x1, …, x6 ≥ 0

Phase I
max Z = – x4 - x6
s.t.​​ 2x1 + 4x2 - x3 + x4​ = 16
​ ​ ​ ​ x1 + 5x2 ​ - x5 + x6 = 15​ ​ x1, …, x6 ≥ 0

cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
-1 x4 16 2 4 -1 1 0 0 4
-1 x6 15 1 5 0 0 -1 1 3
Z= -31
Δj 3 9 -1 0 -1 0

cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
-1 x4 4 6/5 0 -1 1 4/5 -4/5 10/3
0 x2 3 1/5 1 0 0 -1/5 1/5 15
Z= -4
Δj 6/5 0 -1 4/5 0 -9/5

cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x1 10/ 1 0 -5/6 5/6 2/3 -2/3 4
3
0 x2 7/3 0 1 1/6 -1/6 -1/3 1/3 3
Z= 0
Δj 0 0 0 -1 0 -1

This is the optimal solution for the Phase I. Thus proceed to Phase II.
403
 Phase II
cj 10 20 0 0 θ
CB XB b a1 a2 a3 a5
10 x1 10/3 1 0 -5/6 2/3 -
20 x2 7/3 0 1 1/6 -1/3 14
Z= 240/3
Δj 0 0 5 0

cj 0 0 0 0
CB XB B a1 a2 a3 a5
10 x1 15 1 5 0 -1
0 x3 14 0 6 1 -2
Z= 140
Δj 0 -30 0 10

It is observed form the table that the solution is not optimal and the entering
basic variable is x5. But the replacement ratio θ cannot be calculated since all
the elements of the pivot column are negative. This implies that x3 can be
increased indefinitely without driving one of basic variable is zero, or in other
words, the leaving basic variable cannot be determined and hence the
solution is unbounded.

10.11.5 Infeasibility

In case there is no feasible region, the problem is infeasible. In terms of

Simplex table, when in the final solution, an artificial variable is in the basis
but the solution is optimal then there is infeasible solution.

Illustration 10.27:​ max Z = 20x1 + 30x2

​ ​ ​ ​ s.t. ​ 2x1 + x2 ≤ 40
​ ​ 4x1 − x2 ≤ 20​
​ ​ ​ ​ ​ x1​ ≥ 30​ ​ ​ x1, x2 ≥ 0

Solution:

Augmented Form
max Z ​ = 20x1 + 30x2 – x6
s.t.​​ 2x1 + x2 + x3​ ​ = 40
​ ​ ​ ​ 4x1 − x2 + x4​ = 20
​ ​ ​ ​ ​ x1 ​ ​ − x5 + x6 = 30​ x1, …, x6 ≥ 0

Phase I
max Z ​ = – x6
s.t.​ 2x1 + x2 + x3​ ​ = 40
​ ​ ​ ​ 4x1 − x2 + x4​ = 20
​ ​ ​ x1 ​ ​ − x5 + x6 = 30​ x1, …, x6 ≥ 0

404
 cj 0 0 0 0 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x3 40 2 1 1 0 0 0 20
0 x4 20 4 -1 0 1 0 0 5
-1 x6 30 1 0 0 0 -1 1 30
Z= -30
Δj 1 0 0 0 -1 0

​ ​cj 0 0 0 0 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x3 30 0 3/2 1 -1/2 0 0 20
0 x1 5 1 -1/4 0 1/4 0 0 -
-1 x6 25 0 1/4 0 -1/4 -1 1 100
Z= -25
Δj 0 1/4 0 -1/4 -1 0

cj 0 0 0 0 0 -1
CB XB b a1 a2 a3 a4 a5 a6
0 x2 20 0 1 2/3 -1/3 0 0
0 x1 10 1 0 1/6 1/6 0 0
-1 x6 20 0 0 -1/6 -1/6 -1 1
Z= -20
Δj 0 0 -1/6 -1/6 -1 0

Here in the Phase I table Z = -20 and all Δj ≤ 0, but the artificial variable is still
in the basis. Therefore the problem has infeasible solution and we stop here.

10.11.6 Unrestricted (or unconstrained) variables

So far the problems solved had decision variables xj ≥ 0. There may be a

situation where a variable may be unrestricted in sign (i.e. it may have a
positive, zero, or negative value).

Illustration 10.28:​ max Z = 8x1 − 4x2

​ ​ ​ ​ s.t. ​ 4x1 + 5x2 ≤ 20
​ ​ − x1 + 3x2 ≥ − 23
x1 ≥ 0, x2 unrestricted in sign

Solution: Make the right hand side of the II constraint positive by multiplying
it by (-1) and change the inequality.

Since the use of Simplex Method requires that all the decision variables must
have non-negative values at each iteration, the unrestricted variable is
expressed as a difference of two non-negative variables. So here

​ ​

405
 Value of x2 will be positive, zero or negative depending upon is larger,
equal or smaller than

Thus the LPP becomes:

max X = 8x1 − 4 +4
​ ​ ​ ​ s.t. ​ 4x1 + 5 − 5 ≤ 20
​ x1 − 3 +3 ≤ 23​ x1, , ≥0
Augmented Form
max Z = 8x1 − 4 +4
​ ​ ​ ​ s.t. ​ 4x1 + 5 − 5 + x3 ​ ≤ 20
​ ​ ​ x1 − 3 +3 + x4​ ≤ 23​ x1, , , x3, x4 ≥ 0

cj 8 -4 4 0 0 θ
CB XB b a1 a3 a4

0 x3 20 4 5 -5 1 0 5
0 x4 23 1 -3 3 0 1 23
Z= 0
Δj 8 -4 4 0 0

cj 8 -4 4 0 0 θ
CB XB b a1 a3 a4

8 x1 5 1 5/4 -5/4 1/4 0 -

0 x4 18 0 -17/4 17/ -1/4 1 72/1
4 7
Z= 40
Δj 0 -14 14 -2 0

cj 8 -4 4 0 0 θ
CB XB b a1 𝑎2
'
𝑎2
" a3 a4

8 x1 175/17 1 0 0 3/17 5/17

4 72/17 0 -1 1 -1/17 4/17

Z= 1688/1
7
Δj 0 0 0 -1/17 -56/17

The solution is optimal with

​ ​ x1 = 175/17, = 0, = 72/17

⇒​ = 0 – 72/17 = -72/17

i.e.​ x2 = -72/17 and Z = 1688/17

Some important observations

406
1.​ It may be desired to convert a maximization problem into minimization
problem and even vice-versa.
2.​ Whatever to introduce stock, surplus or artificial variables depends on the
type of inequality not on the type of problem i.e. whatever maximization or
minimization.
3.​ Note that except the artificial variables, others, once driven out of the
solution, may re-enter in subsequent iterations.
4.​ In all the tables there must be an identity matrix of p × p, where p is the
number of constraints. The columns that constitute such an identity matrix
need not be adjacent.
5.​ The b-column (Quantity column) values can never be negative. Any
negative value is b-column during computations indicate error in
calculations.
6.​ Z-value in subsequent table will be at least (maximization problem) or at
most (minimization problem) equal to the value in the previous table.
7.​ If any solution violets any of the constants, there is error in computation.
The errors must be checked and corrected before proceeding further.
8.​ Sometimes, lower bounds may be specified in a LPP for example, it may
be stipulated that x1 cannot be less than C1 and x2 cannot be less than C2.
i.e. x1≥C1 and x2 ≥ C2. Such constants can be easily handled by
substituting x1=C1+ y1 and x2 = C2 +y2 in the model and then solving it in
terms of y1 and y2. (Please see Illustration 10.29).
9.​ When the objective function contains a constant term, the simplex method
may be applied by excluding the constant and the optimal solution may be
obtained. This constant is then added at the end to find the optimal value
of the objective function (Please see Illustration 10.29)
10.​The graphical method, the simplex technique and dual approach are just
different ways of solving a LPP.

Illustration 10.29: For a company engaged in the manufacturing of three

products X, Y and Z, the available data are given below. Determine the
product min to minimize the profit.
Operations​ Time in hours required for unit​ Total available hours for month
​ ​ ​ X​ Y​ Z
1​ ​ ​ 1​ 2​ 2​ ​ ​ 200
2​ ​ ​ 2​ 1​ 1​ ​ ​ 220
3​ ​ ​ 3​ 1​ 2​ ​ ​ 180

Profit / unit (Rs)​​ ​ 10​ 15​ 8

Minimum sales requirement​ 10​ 20​ 30
/month in units

Solution: Let x, y and w denote the number of units produced per month of
products X, Y and Z respectively objective is to maximize of monthly profit.
Thus
max Z = 10x + 15y + 8w
s.t.​ x + 2y + 2w ≤ 200

407
 2x + y + w ≤ 220
3x + y + 2w ≤ 180
Also​ ​ x ≥ 10
y ≥ 20
w ≥ 30
Solution of this problem will need introduction of 3 slacks, 3 surplus and 3
artificial variables. Therefore the problem will involve 12 variables and 6
constants. To reduce time and effort for the solution variables x, y, w having
lower bounds of 10, 20 and 30 respectively are substituted as follows:
​ ​ ​ ​ x = 10 + x1
y = 20 + x2
w = 30 + x3
​ ​ ​ where​​ x1, x2, x3 ≥ 0

Substituting these values in the LP Model we get:

max Z = 10(10 + x1) + 15(20 + x2) + 8(30 + x3)
s.t.​ (10 + x1) + 2(20 + x2) + 2(30 + x3) ≤ 200
2(10 + x1) + (20 + x2) + (30 + x3) ≤ 220
3(10 + x1) + (20 + x2) + 2(30 + x3) ≤ 180​x1, x2, x3 ≥ 0
Or
max Z = 10 x1 + 15x2 + 8x3 + 640
= Z’ – 640
s.t.​ x1 + 2x2 + 2x3 ≤ 90
2x1 + x2 + x3 ≤ 150
3x1 + x2 + 2x3 ≤ 70​ ​ x1, x2, x3 ≥ 0

Augmented Form
max Z = Z’ – 640
s.t.​ x1 + 2x2 + 2x3 + x4 ​ ​ = 90
2x1 + x2 + x3 + x5 ​ = 150
3x1 + x2 + 2x3 ​ + x6 = 70​​ x1,…, x6 ≥ 0

cj 10 15 8 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x4 1 2 2 1 0 0 45
90
0 x5 15 2 1 1 0 1 0 150
0
0 x6 3 1 2 0 0 1 70
70
Z’= 0
Δj 10 15 8 0 0 0

cj 10 15 8 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
15 x2 1/2 1 1 1/2 0 0 90
45

408
 0 x5 10 3/2 0 0 -1/2 1 0 70
5
0 x6 5/2 0 1 15/2 0 1 10
25
Z’= 67
5
Δj 5/2 0 -7 -15/2 0 0

cj 10 15 8 0 0 0
CB XB b a1 a2 a3 a4 a5 a6
15 x2 0 1 4/5 3/5 0 -1/5
40
0 x5 0 0 -3/5 -1/5 1 -3/5
90
10 x1 1 0 2/5 -1/5 0 2/5
10
Z’= 70
0
Δj 0 0 -8 -7 0 -1

Thus optimal solution is x1 = 10, x2 = 40, x3 =0 and Z’ 700

Thus ​ ​ ​ x = 10 + x1 = 10 + 10 = 20
y = 20 + x2 = 20 + 40 = 60
w = 30 + x3 = 30 + 0 = 30
​ ​ and​ ​ Z= 700 + 640 = ` 1340
This implies the optimal product mix is to produce 20 units of X, 60 units of Y
and 30 units of Z to get the maximum profit of ` 1340.

Self-Check Questions

State True / False

18.​If the outgoing variable does not correspond to the least non-negative
replacement ration, at least one basic variable would become negative in the
next iteration
19.​In the optimal solution, if a basic variable has a solution value equal to zero ,
then multiple solutions are indicated
20.​If a certain solution to an LPP is degenerate and non-optimal, the next
solution would necessarily be degenerate
21.​If the non-negative ratios are tied, then multiple solutions are indicated

10.12 Duality in Linear Programming Problem

In the previous lessons we have learned formulating LPP and solving them
using Graphical and Simplex Method. The simplex method also provides a
host of information which is useful from the management viewpoint.

409
 For every LPP, there is another LPP which is related to it and therefore, may
be derived from it. This original LPP is called the Primal and the associated
problem is called the Dual. Apart from mathematical connection between
primal and dual, the primal-dual relationship explains the economic
significance of the dual and helps a manager to answer following types of
queries:
●​ Is it advisable to produce all the items that can be produced in a given
situation? If some item is left out in the optimal product-mix, then under what
conditions will it be advisable to produce it?
●​ What is the marginal profitability of each of the resources of the firm? This
means by how much the profit will increase if more quantity of a particular
resource is added or how much reduction in profit will result if its availability is
reduced?

The primal can be rewritten by transposing the rows and columns of the
matrix A. Inverting the problem in this way results in a dual. The variables of
the dual are known as dual variables or shadow prices of the resources. A
solution to the dual may be found in a manner similar to that used for the
primal. The two problems have very closely related properties so that optimal
solution of the dual gives complete information about the optimal solution of
the primal and vice versa. Various useful aspects of this property are:
1)​ If the primal problem contains large number of constraints (rows) and a
smaller number of variables (columns), the computational procedure can be
considerably reduced by converting it into dual and then solving it.
2)​ It gives additional information as to how the optimal solution changes as a
result of the changes in the coefficients and the formulation of the problem.
This forms the basis for sensitivity analysis.
3)​ Calculation of the dual checks the accuracy of the primal solution.
4)​ Economic interpretation of the dual helps the management in making future
decisions.
5)​ Duality is used to solve LPP in which the initial solution is infeasible.
6)​ Solution of the dual can be used by the decision-maker for planning or
increasing resources.

10.12.1 Essence of Duality

Given the canonical form for the primal at the left its dual problem has the
form shown to the right.

Primal Problem​ ​ ​ ​ Dual Problem

​ ​ ​ ​

s.t.​ ​ i=1,2,…,m​ ​ s.t.​ ​ j=1,2,…,n

410
 ​ ​ ​ ​ j=1,2,…,n​ ​ ​ ​
i=1,2,…,m
Here wi are the dual variables and xj are the decision variables. Number of
constraints and interchanged in the two problems. The dual problem usually
uses exactly the same parameters as the primal problem, but in different
locations. Optimal solution of either problem reveals information about the
optimal solution of the other.

Matrix Notation

Let c and w = [w1,w2, …, wm] are row vectors and b and X are column vectors.
Then
Primal​ ​ ​ has​ ​ ​ Dual
max ZP= c x ​ ​ ​ ​ ​ min ZD = bT w
s.t.​ A x ≤ b​ ​ ​ ​ ​ s.t.​ AT w ≥ c
​ ​ x ≥ 0​ ​ ​ ​ ​ ​ w≥0
For example
max ZP = 3x1 +5x2​ ​ ​ min ZD = 4w1 + 12w2 + 18w3
s.t.​ x1 ​ ≤ 4​ ​ ​ s.t.​ w1​ + 3w3 ≥ 3
​ ​ 2x2 ≤ 12​ ​ ​ ​ 12w2 + 2w3 ≥ 5
​ ​ 3x 1 + 2x2 ≤ 18​ ​ ​ ​ ​ w1, w2, w3 ≥ 0
​ ​ x1, x2 ≥ 0

Where

​ ​ max ZP = [3 5] . ​ ​ ​ ​ min ZD = .

​ ​ s.t​ ​ ​ ​ s.t.​

​ ​ ​ ≥ 0​ ​ ​ ​ ​ ≥0
10.12.2 Rules for converting Primal into its Dual

Following are the steps to convert the primal into its dual

Step 1:​ Write the given LPP into canonical form:

411
 a.​ Change the objective function of minimization type to
maximization type

b.​ Change inequality ≥ to ≤

c.​ Change the equation type

a11x1 +a12x2 + ... + a1mxm = b

to its equivalent

a11x1 +a12x2 + ... + a1mxm ≥ b

a11x1 +a12x2 + ... + a1mxm ≤ b

and change inequality ≥ to ≤

Step 2:​ To obtain the dual of the primal

a.​ Transpose the rows and columns of constraint coefficients

b.​ Transpose the coefficients (c1,…,cn) of the objective function and

the right hand side constant (b1,…,bm)

c.​ Change the inequality from ≤ to ≥

d.​ Change the ‘max’ objective function to ‘min

Illustration 10.30:​ max Z = 3x1 + 5x2

s.t.​ x1​ ≤ 4​​
​ ​ ​ x2 ≤ 12
​ ​ ​ ​ ​ 3x1 + 2x2 ≤ 18 ​ ​ x1, x2 ≥ 0

Solution:

Step 1: ​ The given LPP is in canonical form

Step 2:​ To obtain the dual of the primal

a.​ Transpose the rows and columns of constraint coefficients

b.​ Transpose the coefficients (c1,…,cn) of the objective function and

the right hand side constant (b1,…,bm)
Now ​ c = [4 12 18] ​and ​ b = [3 5]
c.​ Change the inequality from ≤ to ≥
d.​ Change the ‘max’ objective function to ‘min
min ZD = 4w1 + 12w2 + 18w3

412
 Thus the dual is
min ZD = 4w1 + 12w2 + 18w3
s.t.​ w1 + 3w3 ≥ 3
12w2 + 2w3 ≥ 5
w1, w2, w3 ≥ 0

Illustration 10.31:​ max Z = 8x1 + 16x2

s.t ​ x1 + x2 ≤ 200
​ ​ ​ ​ x2 ≤ 125​
​ ​ ​ ​ 3x1 + 6x2 ≤ 900​ ​ x1, x2 ≥ 0
Solution: The given problem is in canonical form. Thus the dual is
min ZD = 200w1 + 125w2 + 900w3
s.t.​ w1 + 3w3 ≥ 8
w1 + w2 + 6w3 ≥ 16​ ​ w1, w2, w3 ≥ 0

Illustration 10.32:​ max Z = 28x1 + 30x2

s.t.​ 6x1 + 3x2 ≤ 18
​ ​ ​ ​ ​ 3x1 + x2 ≤ 8
​ ​ ​ ​ ​ 4x1 + 5x2 ≤ 30​ x1, x2 ≥ 0
Solution:​ The given problem is in canonical form. Thus the dual is
min ZD = 18w1 + 8w2 + 30w3
s.t.​ 6w1 + 3w2 + 4w3 ≥ 28
3w1 + w2 + 5w3 ≥ 30​ ​ w1, w2, w3 ≥ 0

Illustration 10.33:​ min Z = x1 − 3x2 + 3x3

s.t.​ 3x1 − x2 + 2x3 ≤ 7​ ​
​ ​ ​ 2x1 + 4x2 + 3x3 ≥ −12
​ ​ ​ ​ ​ − 4x1 + 3x2 + 3x3 ≤ 10​ x1, x2, x3 ≥ 0
Solution:​ The given problem is not in canonical form. Convert the ‘min’
type objective function to max type and the inequality of the second constraint
as ≤ by multiplying both the equations by (-1). Thus the primal in canonical
form is

max (−Z) = − x1 + 3x2 − 3x3

s.t.​ 3x1 − x2 + 2x3 ≤ 7​
​ ​ − 2x1 − 4x2 − 3x3 ≤ 12
​ ​ ​ − 4x1 + 3x2 + 3x3 ≤ 10​ ​ x1, x2, x3 ≥ 0

Now the Dual is

min ZD = 7w1 + 12w2 + 10w3
s.t.​ 3w1 − 2w2 − 4w3 ≥ − 1
​ ​ − w1 − 4w2 + 3w3 ≥ 3
2w1 − 3w2 + 3w3 ≥ − 3​ ​ w1, w2, w3 ≥ 0

Illustration 10.34:​ min Z ​= 40x1 + 24x2

s.t.​ 20x1 + 50x2 ≥ 4800
​ ​ ​ ​ ​ 80x1 + 50x2 ≥ 7200​ ​ x1, x2 ≥ 0
413
 Solution:​ Writing the given problem in canonical form
max (−Z) = − 40x1 − 24x2
s.t.​ − 20x1 − 50x2 ≤ − 4800
​ ​ ​ − 80x1 − 50x2 ≤ − 7200​ x1, x2 ≥ 0

​ ​ The dual is
min ZD = − 4800w1 − 7200w2
s.t.​ − 20w1 − 80w2 ≥ − 40
− 50w1 − 50w2 ≥ − 24​ ​ w1, w2 ≥ 0

Illustration 10.35:​ max Z = 2x1 + 4x2

s.t.​ 2x1 + x2 ≤ 18
​ ​ ​ ​ ​ 3x1 + 2x2 ≥ 30
​ ​ ​ ​ ​ x1 + 2x2 = 26​ ​ ​ x1, x2 ≥ 0

Solution:​ Writing the given problem in canonical form, the third constraint
equation will be replace by the pair
x1 + 2x2 ≥ 26
x1 + 2x2 ≤ 26

Therefore the primal is

max Z = 2x1 + 4x2
s.t.​ 2x1 + x2 ≤ 18
​ ​ ​ ​ − 3x1 − 2x2 ≤ − 30
− x1 − 2x2 ≤ − 26
x1 + 2x2 ≤ 26​ ​ x1, x2 ≥ 0

The dual is
min ZD = 18w1 − 30w2 + 26w3 − 26w4
s.t.​ 2w1 − 3w2 + w3 − w4 ≥ 2
w1 − 2w2 + 2w3 − w4 ≥ 4​ w1, w2, w3, w4 ≥ 0

Notice here that corresponding to an n-variable, m-constraint primal problem

there would be m-variable and n-constraint dual problem. In the example,
there are 2 variables and 3 constraints. Thus the dual should have 3 variables
and 2 constraints. But observed that the dual we obtained has 5 variables.
The seeming inconsistency can be resolved by expressing

Where is a variable unrestricted in sign. The variables w3 and w4, both are
non-negative, their differences could be greater than, less than or equal to
zero.

Thus the dual can be rewritten as

414
 min ZD = 18w1 − 30w2 + 26
s.t.​ 2w1 − 3w2 + ≥2
w1 − 2w2 + ≥ 4​​ w1, w2 ≥ 0, is unrestricted in sign

Illustration 10.36:​ max Z = 8x1 − 4x2

​ ​ ​ ​ s.t. ​ 4x1 + 5x2 ≤ 20
​ − x1 + 3x2 ≥ − 23​ x1 ≥ 0, x2 unrestricted in sign

Solution:​ Replace x2 by the difference of non-negative variables i.e.

​ ​ ​ Thus the primal is:

max Z = 8x1 −
​ ​ ​ ​ s.t. ​ 4x1 + ≤ 20

​ x1 − ≤ 23​ ​ x1, ≥0

The dual is
min zD = 20w1 + 30w2
s.t.​ 4w1 + w2 ≥ 8
5w1 − 3w2 ≥ −4
−5w1 + 3w2 ≥ 4​ ​ w1, w2 ≥ 0

The second constraint of the dual can be rewritten as

− 5w1 + 3w2 ≤ 4

Combining thus new constraint and the third constraint, we get

− 5w1 + 3w2 = 4

Thus the dual can be expressed as

min ZD = 20w1 + 30w2
s.t.​ 4w1 + w2 ≥ 8
−5w1 + 3w2 = 4​ ​ w1, w2 ≥ 0

Thus whenever constraint in a primal, involves an equality sign, its

corresponding dual variable shall be unrestricted in sign. Similarly an
unrestricted variable in the primal would imply that the corresponding
constraint shall bear an = sign.

Illustration 10.37:​ max Z = 3x1 + 4x2 + 7x3

​ ​ ​ ​ s.t. ​ x1 + x2 + x3 ≤ 10
​ ​ ​ ​ ​ 4x1 − x2 − x3 ≥ 15
​ ​ ​ ​ x1 + x2 + x3 = 7 x1, x2 ≥ 0, x3 unrestricted in sign

415
 Solution:​ Replacing x3 by the difference of non-negative variables i.e.

​ and the third constraint equation will be replace by the pair

x1 + x2 + x3 ≤ 7
x1 + x2 + x3 ≥ 7

​ ​ we get the primal in canonical form as

max Z = 3x1 + 4x2 +

​ ​ ​ ​ s.t. ​ x1 + x2 + ≤ 10
​ ​ ​ ​ ​ −4x1 + x2 + ≤ −15
x1 + x2 + ≤ 7

− x1 − x2 − ≤ − 7​ ​ x1, x2, ≥0
The dual is
min ZD = 10w1 − 15w2 + 7w3 − 7w4
s.t.​ w1 − 4w2 + w3 − w4 ≥ 3
w1 + w2 + w3 − w4 ≥ 4
w1 + w2 + w3 − w4 ≥ 7
− w1 − w2 − w3 + w4 ≥ −7​ ​ w1, w2, w3, w4 ≥ 0

Or​ min ZD = 10w1 − 15w2 + 7

s.t.​ w1 − 4w2 + ≥ 3
w1 + w2 + ≥ 4
w1 + w2 + ≥ 7
− w1 − w2 − ≥ −7 w1, w2≥ 0,
unrestricted

Or​ min ZD = 10w1 − 15w2 + 7

s.t.​ w1 − 4w2 + ≥ 3
w1 + w2 + ≥ 4
w1 + w2 + = 7​ w1, w2≥ 0,
unrestricted
10.12.3 Relationship between Primal and Dual

The symmetrical relationship between the primal and the dual is given in
Table 10.7.
Table 10.7: Primal-dual relationship

S.N Primal Dual

o

416
 1. If a primal has n variables and m Its dual has m variables and n
constraints constraints
2. Objective is to maximize Z Objective is to minimize Z
3. Cost vector is c and requirement Cost vector is bT and requirement vector
vector is b is cT
4. Coefficient Matrix A Coefficient Matrix AT
5. Constraints with sign ≤ Constraints with sign ≥
6. ith constraint an inequality ith variable wi ≥ 0
7. ith constraint an equality ith variable wi unrestricted in sign
8. jth variable xj ≥ 0 jth constraint an inequality
9. jth variable xj unrestricted in sign jth constraint an equality
10. ith slack variable positive ith variable zero
11. ith variable zero ith slack variable positive
12. Finite optimal solution Finite optimal solution with max zP=min
zD
13. Unbounded Solution No solution
14. No solution Unbounded solution

10.12.4 Properties of Dual

Theorem 1:​ Dual of dual of a given primal is the primal itself.

Theorem 2:​ If x0 is any feasible solution to the primal

max ZP= c x
s.t.​ A x ≤ b​​ ​ ​ ​ ​ (10.15)
​ ​ ​ ​ x≥0

​ ​ and w0 is any feasible solution to its dual

​ ​ ​ min ZD = bT w
​ ​ ​ s.t.​ AT w ≥ c​ ​ ​ ​ ​ (10.16)
​ ​ ​ ​ w≥0

then ​ cx0 ≤ bTw0.

​ ​ ​ This is weak duality theorem.

Theorem 3:​ If x0 is any feasible solution to the primal (10.15) and w0 is any
feasible solution to its dual (10.16) then x0 and w0 are optimal
solutions to the primal and the dual problems respectively.

Theorem 4:​ (Basic Theorem of Duality): Let a primal be given by (10.15) and
the associated dual be (10.16).

If x0 (w0) is an optimal solution to the primal (dual) then there

exists a feasible solution w0 (x0) to the dual (primal) such that

cx0 = bTw0

Theorem 5:​ (Fundamental Theorem of Duality): If the primal or the dual has
a bounded (finite) solution to an LPP, then the other problem
also possesses a finite optimum solution and the optimum
values of the objective function of the two problems are equal.

417
 Theorem 6:​ (Existence Theorem): If the primal (dual) has an unbounded
solution, then its dual (primal) has no feasible solution.

Self-Check Questions

State True / False

22.​The dual to a given LPP would have as many variables as the number of
constraints in the primal
23.​Corresponding to every unrestricted primal variable, an equality dual
constraint is obtained
24.​The dual to the dual of an LPP is the primal itself

10.13 Duality and Simplex Method

Since any LPP can be solved by using simplex method, the method is
applicable to both the primal as well as the dual. The fundamental theorem of
duality suggests that an optimal solution to the associated dual can be
obtained from that of its primal and vice versa. Consider the following
example

Example:

Primal​​ ​ ​ ​ ​ Dual
max ZP = 40x1 + 35x2 ​ ​ ​ min ZD = 60w1 + 96w2
s.t.​ 2x1 + 3x2 ≤ 60​ ​ ​ s.t.​ 2w1 + 4w2 ≥ 40
​ ​ 4x1 + 3x2 ≤ 96​ ​ ​ 3w1 + 3w2 ≥ 35
​ ​ ​ x1, x2 ≥ 0​ ​ ​ ​ w1, w2 ≥ 0​
Solve the primal using usual simplex method. The optimal solution to the
primal is
cj 40 35 0 0
CB XB b a1 a2 a3 a4
35 x2 8 0 1 2/3 -1/3
40 x1 18 1 0 -1/2 2/3
Z= 1000
Δj 0 0 -10/3 -25/3

Now consider the dual. Solving the dual using the II phase method, the
optimal solution to the dual is
cj -60 -96 0 0 θ
CB WB B y1 y2 y3 y5
-96 w2 25/3 0 1 -1/2 1/3
-60 w1 10/3 1 0 1/2 -2/3
Z= -1000
Δj 0 0 -18 -8

418
1.​ Correspondence between the variables of primal and dual

●​ The decision variable x1 in the primal corresponds to the surplus variable

w3 in the dual, while the decision variable x2 corresponds to the surplus
variable w5 in the dual

●​ Similarly the dual variable w1 and w2 corresponds to the slack variables x3

and x4 respectively of the primal

2.​ Comparison of the optimal solution to the primal and the dual

a.​ The objective function value of both the problems are the same

i.e.​Zmax = 1000 = ZD(Zmin)

b.​ The numerical value of each of the variables in the optimal solution to the
primal is equal to the value of its corresponding variable in the dual
contained in the Δj row.

Thus in the primal problem

x1 = 18, x2 = 8;

​ w1 = 10/3, w2 = 25/3 (Δj value, ignoring the sign, of the slack variables
in the optimal solution of the primal)

In the Dual problem

w1 = 10/3, w2 = 25/3;

​ x1 = 18, x2 = 8​ (Δj value, ignoring the sign, of the surplus

variables in the optimal solution of the dual)

The artificial variables are not considered because they do not

correspond to any variable in the primal and are introduced for a specific,
limited purpose only.

Clearly, if feasible solutions exist for both the primal and the dual then
both problems have optimal solutions of which objective function values
(optimal values) are same.

Also if one problem has unbounded solution the associated problem has
no feasible solution.

c.​ The optimal solution to the dual can be read from the optimal solution of
the primal and vice versa. Therefore there is no need to solve both the
primal and dual problems to obtain the optimal solution.

Illustration 10.38: A feed mixing operation can be described in terms of two

activities. The required mixture must contain four kinds of ingredients w, x, y
and z. Two basic feeds A and B, which contain the required ingredients are
available in the market. One kg of A contains 0.1 kg of W, 0.1 kg of y and 0.2

419
 kg of Z. likewise one kg of B contains 0.1 kg of x, 0.2 kg of y and 0.1 kg of Z.
the daily per head requirement is of at least 0.4 kg of W, 0.6 kg of x, 2 kg of y
and 1.8 kg of Z. Feed A can be brought for ` 0.07 per kg and B can be brought
for ` 0.05 per kg the availabilities, requirements and costs are summarized in
the table below.
Ingredient​ Feed A (kg)​ Feed B (kg)​ Requirement (kg)
W​ ​ 0.1​ ​ 0.0​ ​ ​ 0.4
X​ ​ 0.0​ ​ 0.1​ ​ ​ 0.6
Y​ ​ 0.1​ ​ 0.2​ ​ ​ 2.0
Z​ ​ 0.2​ ​ 0.1​ ​ ​ 1.8
Cost (`)​​ 0.007​ ​ 0.005

Determine the quantities of feeds A and B in the mixture so that the total lost
is minimum.

Solution: Let x1 and x2 be the extend of feeds A and B respectively. The

formulated LPP, P2, is given below

min Z = 0.07x1 + 0.05x2

s.t​ 0.1x1 ​ ≥ 0.4
​ ​ ​ ​ 0.10x2 ≥ 0.6
​ ​ ​ ​ 0.1x1 + 0.20x2 ≥ 2.0
​ ​ ​ ​ 0.2x1 + 0.10x2 ≥ 1.8​ ​ x1, x2 ≥ 0

Case I: Solving the Primal Problem P2

Augment Form:
max (-Z) = -0.07x1 - 0.05x2 – Mx4 – Mx6 – Mx8 – Mx10
s.t​ 0.1x1 ​ -x3 + x4 ​ ​ ​ ​ = 0.4
​ 0.10x2 ​ -x5 + x6 ​ ​ ​ = 0.6
​ ​ 0.1x1 + 0.20x2 ​ ​ -x7 + x8 ​ = 2.0
​ ​ 0.2x1 + 0.10x2 ​ ​ ​ ​ -x9 + x10 = 1.8
​ ​ ​ ​ ​ ​ ​ ​ ​ x1,…, x10 ≥ 0

cj -0.07 -0.05 0 -M 0 -M 0 -M 0 -M θ
CB XB b a1 a2 a3 a4 a5 a6 a7 a8 a9 a10
-M x4 0.4 0.1 0 -1 1 0 0 0 0 0 0 -
-M x6 0.6 0 0.1 0 0 -1 1 0 0 0 0 6
-M x8 2.0 0.1 0.2 0 0 0 0 -1 1 0 0 10
-M X1 1.8 0.2 0.1 0 0 0 0 0 0 -1 1 18
Z= -4.8M

420
 Δj -0.07 -0.05 -M 0 -M 0 -M 0 -M 0
+0.4 +0.4
M M

cj -0.07 -0.05 0 -M 0 -M 0 -M 0 -M θ
CB XB b a1 a2 a3 a4 a5 a6 a7 a8 a9 a10
-M x4 0.4 0.1 0 -1 1 0 0 0 0 0 0 -
0.0 x2 6.0 0 1 0 0 -10 10 0 0 0 0 -
5
-M x8 0.8 0.1 0 0 0 2 -2 -1 1 0 0 2/5
-M X1 1.2 0.2 0 0 0 1 -1 0 0 -1 1 6/5
Z= - 2.4M
Δj -0.07 0 -M 0 -0.05 0.5 -M 0 -M 0
+0.4 +3M -4M
M

Proceeding in the same way, the optimal solution is obtained after three more
iterations as follows

cj -0.07 -0.05 0 -M 0 -M 0 -M 0 -M θ
CB XB b a1 a2 a3 a4 a5 a6 a7 a8 a9 a10
0.0 x1 16/3 0.1 0 -1 1 0 0 0 0 0 0 -
7
0.0 x2 22/3 0 1 0 0 -10 10 0 0 0 0 -
5
0 x5 1/3 0.1 0 0 0 2 -2 -1 1 0 0 2/5
0 x1 1/3 0.2 0 0 0 1 -1 0 0 -1 1 6/5
Z= 0.74
Δj 0 0 0 -M 0 -M -0.1 -M +0.1 -0.3 -M+0.
3

Therefore the optimal quantities of feeds A and B in the mixture are 16/3 kg
and 22/3 kg respectively with total cost = ` 0.74.

Case II: Solving the Dual Problem

The Dual of the primal P2 would be as follows:

max ZD = 0.4w1 + 0.6w2 + 2.0w3 + 1.8w4
s.t​ 0.1w1 ​ ​ + 0.1w3 + 0.2w4 ≤ 0.07
0.1w2 + 0.2w3 + 0.1w4 ≤ 0.05,​ w1,…, w4 ≥ 0

Augmented Form:
max ZD = 0.4w1 + 0.6w2 + 2.0w3 + 1.8w4
s.t​ 0.1w1 ​ ​ + 0.1w3 + 0.2w4 + w5​ = 0.07
0.1w2 + 0.2w3 + 0.1w4 ​ ​ + w6 ​ = 0.05
w1,…, w6 ≥ 0

cj 0.4 0.6 2 1.8 0 0 θ

421
 CB W b y1 y2 y3 y4 y5 y6
B
0 w5 0.07 0.1 0 0.1 0.2 1 0 0.7
0 w6 0.05 0 0.1 0.2 0.1 0 1 0.2
5
Z= -4.8M
Δj 0.4 0.06 2 1.8 0 0

cj 0.4 0.6 2 1.8 0 0 θ

CB W b y1 y2 y3 y4 y5 y6
B
0 w5 0.45 0.1 -0.05 0 0.15 1 -0.5 0.3
0 w3 0.25 0 0.5 1 0.5 0 5 0.5
Z= -4.8M
Δj 0.4 -0.4 0 0.8 0 -10

cj 0.4 0.6 2 1.8 0 0

CB W b y1 y2 y3 y4 y5 y6
B
0 w5 0.45 2/3 -1/3 0 1 20/3 -10/
3
0 w3 0.25 -1/3 2/3 1 0 -10/ 20/3
3
Z= -4.8M
Δj -0.4/ -0.4/ 0 0 -16/ -22/
3 3 3 3

This is optimal solution with w1 = 0, w2 = 0, w3 = 0.1 and w4 = 0.3, with ZD = `

0.74

Optimal values of x1 and x2 are given under the slacks of the dual with
changed sign of coefficient in Δj row.

⇨​ Optimal value of x1 = 16/3 and x2 = 22/3

Illustration 10.39:​ Consider the following LPP. Solve it using simplex

method and determine the values of decision variables, dual variables and Z.
max Z = 8x1 + 16x2
s.t​ ​ x1 + x2 ≤ 200
​ ​ ​ ​ ​ ​ x2 ≤ 125​
​ ​ ​ ​ ​ 3x1 + 6x2 ≤ 900​ ​ ​ x1, x2 ≥ 0

Solution:​

The Augmented Form is

​ ​ ​ ​ max Z = 8x1 + 16x2
s.t​ x1 + x2 + x3 ​ ​ ≤ 200
​ ​ ​ ​ ​ x2 + x4 ≤ 125​
​ ​ ​ ​ ​ 3x1 + 6x2 ​ + x5 ​≤ 900​ ​ x1,…, x5 ≥ 0

cj 8 16 0 0 0 θ
CB XB b a1 a2 a3 a4 a5

422
 0 x3 200 1 1 1 0 0 200
0 x4 125 0 1 0 1 0 125
0 x5 900 3 6 0 0 1 150
Z= 0
Δj 8 16 0 0 0

cj 8 16 0 0 0 θ
CB XB b a1 a2 a3 a4 a5
0 x3 75 1 0 1 -1 0 75
16 x2 125 0 1 0 1 0 -
0 x5 150 3 0 0 -6 1 50
Z= 200
0
Δj 8 0 0 -16 0

cj 8 16 0 0 0
CB XB b a1 a2 a3 a4 a5
0 x3 25 0 0 1 1 -1/3
16 x2 125 0 1 0 1 0
8 x1 50 1 0 0 -2 1/3
Z= 240
0
Δj 0 0 0 0 -8/3

All the Δj ≤ 0, this indicates that the solution is optimal solution.

The values of decision variables are x1 = 50, x2 = 125 and Z = 2400.

From the optimal table the values of dual variables can also be written. Value
of w1 will be the Δj value (ignoring the sign) of the slack variable x3, value of w2
will be the Δj value (ignoring the sign) of the slack variable x4 and value of w3
will be the Δj value (ignoring the sign) of the slack variable x5. Thus

​ ​ ​ ​ w1 = 0, w2 = 0, w3 = 8/3 and ZD = -2400

Illustration 10.40: Consider the following LPP. Write its dual and solve it.
From the solution of dual, determine the values of dual variables and Z.
max Z = 3x1 + 4x2 + 7x3
​ ​ ​ s.t. ​ x1 + x2 + x3 ≤ 10
​ ​ ​ ​ 4x1 − x2 − x3 ≥ 15
​ ​ ​ ​ x1 + x2 + x3 = 7​ x1, x2 ≥ 0, x3 unrestricted in sign

Solution: The dual of the above LPP is:

min ZD = 10w1 − 15w2 + 7

s.t.​ w1 − 4w2 + ≥ 3
w1 + w2 + ≥ 4
w1 + w2 + = 7​w1, w2≥ 0, unrestricted

423
 Considering = w3 − w4 , where both w3 and w4 ≥ 0, the above dual can be
rewritten as
​ min ZD = 10w1 − 15w2 + 7w3 − 7w4
s.t.​ w1 − 4w2 + w3 − w4 ≥ 3
w1 + w2 + w3 − w4 ≥ 4
w1 + w2 + w3 − w4 = 7​ ​ w1, w2, w3, w4 ≥ 0

Writing this in augmented form,

max (-ZD) = −10w1 + 15w2 − 7w3 − 7w4 − w6 − w8 − w9
s.t.​ w1 − 4w2 + w3 − w4 − w5 + w6 = 3
w1 + w2 + w3 − w4 ​ ​ − w7 + w8 ≥ 4
w1 + w2 + w3 − w4 ​ ​ ​ + w9 = 7
​ ​ ​ ​ ​ ​ w1, …., w9 ≥ 0

Solving this LPP using the II Phase Method

cj 0 0 0 0 0 -1 0 -1 -1 θ
CB WB b y1 y2 y3 y4 y5 y6 y7 y8 y9
-1 w6 3 1 -4 1 -1 -1 1 0 0 0 3
-1 w8 4 1 1 1 -1 0 0 -1 1 0 4
-1 w9 7 1 1 1 -1 0 0 0 0 1 7
Z= -14
Δj 3 -2 3 -3 -1 0 -1 0 0

Solving the above initial table we get the optimal solution for phase I as
follows:
cj 0 0 0 0 0 -1 0 -1 -1 θ
CB WB b y1 y2 y3 y4 y5 y6 y7 y8 y9
0 w4 31/5 1 0 1 -1 -9/5 9/5 0 0 4/5 3
0 w2 4/5 0 1 0 0 -1/5 1/5 0 0 1/5 4
0 w8 3 0 0 0 0 -2 2 1 -1 1 7
Z= 0
Δj 0 0 0 0 0 -1 0 -1 -1

Proceeding to Phase II the optimal solution to the Problem is

cj -10 15 -7 7 0 0
CB WB b y1 y2 y3 y4 y5 y7
-7 w4 31/5 1 0 1 -1 -9/5 0
15 w2 4/5 0 1 0 0 -1/5 0
0 w8 3 0 0 0 0 -2 1
Z= -157/5
Δj -3 0 0 0 -48/5 0

Thus the optimal solution is w1 = 0, w2 = 4/5, w3 = 0 and w4 = 31/5, with ZD =

157/5

⇒​ = w3 − w4 = 0 - 31/5 = -31/5

424
 Illustration 10.41:​ Consider the following LPP. Solve it and interpret the
result.
max Z = 6x1 + x2
​ ​ ​ ​ s.t. ​ 2x1 + x2 ≥ 3
​ ​ ​ ​ ​ - x1 + x2 ≥ 0​ ​ x1, x2 ≥ 0

Solution:

Augmented Form
max Z = 6x1 + x2 – x4 – x6
​ ​ ​ s.t. ​ 2x1 + x2 – x3 + x4 ​ ​ =3
​ ​ ​ ​ – x1 + x2 ​ – x5 + x6 = 0​ ​ x1, ..., x6 ≥ 0
cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
-1 x4 3 2 1 -1 1 0 0 3/2
-1 x6 0 -1 1 0 0 -1 1 4
Z= -3
Δj 2 1 -1 0 -1 0

cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x1 3/2 1 1/2 -1/2 1/2 0 0 3
-1 x6 3/2 0 3/2 -1/2 1/2 -1 1 1
Z= -3/2
Δj 0 3/2 -1/2 -1/2 -1 0

cj 0 0 0 -1 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x1 1 1 0 -1/3 1/3 1/3 -1/3
0 x2 1 0 1 -1/3 1/3 -2/3 2/3
Z= 0
Δj 0 0 0 -1 0 -1

Phase I optimal solution. Proceeding to Phase II

cj 6 1 0 0 θ
CB XB b a1 a2 a3 a5
6 x1 1 1 0 -1/3 1/3
1 x2 1 0 1 -1/3 -2/3
Z= 7
Δj 0 0 5/3 -5/3

Since there is no leaving variable, the solution is unbounded.

Now consider the dual

min ZD = -3w1
​ ​ ​ s.t. ​ - 2w1 + w2 ≥ 6
​ ​ ​ ​ - w1 - w2 ≥ 1​​ w1,.., w6 ≥ 0

Augmented Form
min ZD = -3w1 – w4 – w6

425
 ​ ​ ​ s.t. ​ - 2w1 + w2 – w3 + w4​ ​ =6
​ ​ ​ ​ - w1 - w2 ​ – w5 – w6 = 1​ ​ w1,.., w6 ≥ 0

cj 0 0 0 -1 0 -1 θ
CB WB b y1 y2 y3 y4 y5 y6
-1 y4 6 -2 1 -1 1 0 0
-1 y6 1 -1 -1 0 0 -1 1
Z= -7
Δj -3 0 -1 0 -1 0

The solution is optimal but the artificial variables are still in the basis. Thus the
solution is infeasible.

Thus if the primal has unbounded solution, the dual will infeasible solution and
vice versa.

10.14 Economic Interpretation

Not only are the primal and the dual related mathematically, they are also
related in economic sense. Consider the following example:

Illustration 10.42: A company has two bottling plants, one located at Solan
and other located at Mohan Nagar. Each plant produces three drinks - Fruit
Juices, Coke and Pepsi named A, B and C respectively. The numbers of
bottles produced per day are as follows:
Solan​ ​ ​ ​ Mohan Nagar
A​ ​ 1,5000​ ​ ​ ​ 1,500​
B​ ​ 3,000​ ​ ​ ​ 1,000
C​ ​ 2,000​ ​ ​ ​ 5,000

A market survey indicates that during the month of April, there will be a
demand of 20,000 bottles of A, 40,000 bottles of B and 44,000 bottles of C.
The operating costs per day for plant at Solan and Mohan Nagar are 600 and
400 monetary units. For how many days each plant should be run in April so
as to minimize the production cost, while still meeting the market demand.

Solution: Let x1 and x2 are the number of days in April for which the bottling
plants at Solan and Mohan Nagar must run.

The formulation of LPP is

min ​ Z = 600x1 + 400x2
s.t.​ 1500x1+ 1500 x2 ≥ 20000​ ​ ​
​ 3000x1 + 1000x2 ≥ 40000
2000x1 + 5000x2 ≥ 44000​ ​ x1, x2 ≥ 0

Optimal Solution ⇒ x1 = 12 days and x2 = 4 days resulting z = 8800 monetary

units

Optimal Solution for dual

426
 ​ y1= 0

y2 = 11/65

y3 = 3/65

Wmax = 8800

Now the RHS of the constraints denote monetary units, the LHS must also be
expressed in monetary units.

In the first constraint first term 1500 y1 1500 is the number of bottles of A
produced per day.

Hence y1 must denote the cost of producing one bottle of A

​ y2 must denote the cost of producing one bottle of A

​ y3 must denote the cost of producing one bottle of A

●​ y1, y2 and y3 are dual variables or shadow prices. These are called shadow
price as these represent the true accounting value or imputed value of the
three drinks.

●​ The objective function of the dual is to maximize the total accounting values of
the drinks produced per month

●​ The two constraints ensure that the total accounting value of the daily output
of each plant must remain less than the daily cost of operating the plant.

Miscellaneous Illustration
Illustration 10.43: The owner of a Metro Sports wishes to determine how many
advertisements to place in the selected three monthly magazines A, B and C. His
objective is to advertise in such a way that total exposure to principal buyers of
expensive sports good is maximized. Percentage of readers for each magazine is
the number of advertisements placed multiplied by the number of principal
buyers. The following data may be used:
Magazine​ ​ ​ A​ ​ B​ ​ C
Readers​ ​ ​ 1 Lakh​ ​ 0.6 lakh​​ 0.4 lakh
Principal buyers​​ ​ 20%​ ​ 15%​ ​ 8%
Cost per advertisements​ 8000​ ​ 6000​ ​ 5000

The budgeted amount is at most ` 1 lakh for the advertisements. The owner has
already decided that magazine. A should have no more than 15 advertisements
and that B and C each have at least 80 advertisements. Formulate an LP model
for the problem.

Solution: The mathematical formulation is:

Decision variables: Let

427
 x1 = number of insertions in Magazine A

x2 = number of insertions in Magazine B

x3 = number of insertions in Magazine C

Objective function: Let Z be total exposure

​ ​ max Z = (20% × 1,00,000)x1 + (15% × 60,000)x2 + (8% × 40,000)x3
​ ​ ​ = 20,000x1 + 9,000x2 + 3,200x3
Constraints:
​ 8000x1 + 6000x2 + 5000x3 ≤ 1,00,000​ ​ (Budgeting Constraints)
​ ​ ​ x1 ≤ 15, x2 ≥ 8, x3 ≥ 8​ ​ (Advertisement Constraints)
x1, x2, x3 ≥ 0​ ​ ​ (Non-negative Constraints)
Thus LPP is
max Z​= 20,000x1 + 9,000x2 + 3,200x3
s.t.​ 8000x1 + 6000x2 + 5000x3 ≤ 1,00,000
​ ​ ​ ​ x1 ≤ 15, x2 ≥ 8, x3 ≥ 8
and​ ​ ​ x1, x2, x3 ≥ 0
Illustration 10.44: The super bazaar in a city daily needs between 32 and 40
workers depending on the time of day. The rush hours are between noon and
2 pm. The table below indicates the number of workers needed at various
when the bazaar is open.
Time period​ ​ ​ number of workers needed
9 am – 11 am​ ​ ​ ​ 32
11 am – 1 pm​ ​ ​ ​ 40
1 pm – 3 pm​ ​ ​ ​ 35
3 pm – 5 pm​ ​ ​ ​ 33

The super bazaar now employs 34 fulltime workers, but needs a few part-time
workers also. A part time worker must put in exactly 4 hours per day, but can
start any time between 9 am and 1 pm. Full time workers work from 9 am to 5
pm, but are allowed an hour for lunch (half of the full-time eat at 12 noon, the
other half at 1 pm. The management of the super bazaar limits part-time
hours to a maximum of 50% of the day’s total requirement.

Part-timers earn ` 48 per day on the average, while full-timers earn ` 140 per
day in salary and benefits on the average. The management wants to set a
schedule that would minimize total manpower costs. Formulate this problem
as an LPP model to minimize total daily manpower cost.

Solution: The key decision is to determine the number of full-time and

part-time workers needed in super bazaar.

Decision variables: Let y represent the number of full time workers starting at
9 am, 11am and 1pm respectively. (j=1, 2, 3)

Let Z be total daily manpower cost, then

​ ​ min Z ​= 140 y + 48 (x1 + x2 + x3)
​ ​ s.t.​ y + x1 ​ ​ ≥ 32​ ​ (9 am - 11 am need)

428
​ ​ ​ (1/2)y + x1 + x2​ ≥ 40​ ​ (11 am - 1 pm need)
​ ​ ​ (1/2)y + x2 + x3​ ≥ 35​ ​ (1 pm - 3 pm need)
​ ​ ​ y+​ + x3​ ≥ 33​ ​ (3 am - 5 pm need)
​ ​ ​ y ​ ​ ≤ 34​ ​ (Full-timers available)
​ ​ ​ 4 (x1 + x2 + x3) ≤ 0.50(32+40+35+33)
(Part timers’ hour cannot exceed 50% of total hours
required each day which is the sum of the workers
needed each hour)
x1, x2, x3 ≥ 0

Illustration 10.45: Air force is experimenting with three types of bombs P, Q

and R in which three kinds of explosive viz., A, B and C will be used. Taking
the various factors into consideration, it has been decided to use at most 600
kg of explosive A, at least 480 kg of explosive B ad exactly 540 kg of
explosive C. Bomb P requires 3, 2, 2 kg of A, B and C respectively. Bomb Q
requires 1, 4, 3 kg of A, B and C respectively. Bomb R requires 6, 2, 3 kg of A,
B and C respectively. Bomb P will give the equivalent of a 2 ton explosion,
Bomb Q will give the equivalent of a 3 ton explosion and Bomb R will give the
equivalent of a 4 ton explosion. Under what production schedule can the Air
Force make the biggest target? Formulate this as an LPP.

Solution: Let the Air Force experiment with x1 units of bomb P, x2 units of
bomb Q, and x3 units of bomb R. The mathematical formulation of the problem
is
​ ​ ​ max Z = 2x1 + 3x2 + 4x3
​ ​ ​ s.t. ​ 3x1 + x2 + 6x3 ≤ 600​ ​ (Explosion A constraint)
2x1 + 4x2 + 2x3 ≥ 480​ ​ (Explosion B constraint)
​ 2x1 + 3x2 + 3x3= 540​ ​ (Explosion C
constraint)
x1, x2, x3 ≥ 0

Illustration 10.46: A finished product must weigh exactly 150 kg. The two raw
materials used in manufacturing the products are A, with a cost of ` 2 per unit
and B with a cost of ` 8 per unit. At least 14 units of B and not more than 20
units of A must be used. Each unit of A and B weighs 5 and 10 grams
respectively. How much of each type of raw material should be used for each
unit of the final product in order to minimize the cost.

Solution: Let the product mix contain of x1 units of Product A and x2 units of
B.
​ ​ ​ min Z = 2x1 + 8x2
​ ​ ​ s.t. ​ 5x1 + 10x2 ​ =150
x1​ ​ ≤ 20
​ ​ ​ x2​ ≥ 14​ ​ x1, x2 ≥ 0

Augmented Form
​ ​ ​ max (-Z) = -2x1 - 8x2 - x3 - x6
​ ​ ​ s.t. ​ 5x1 + 10x2 + x3 ​ ​ = 150
x1​ ​ + x4 ​ = 20
429
 ​​ ​ ​ x2 ​ ​ − x5 + x6 = 14​ x1,…, x6 ≥ 0

Phase I
​ ​ ​ max (-Z) = - x3 - x6
​ ​ ​ s.t. ​ 5x1 + 10x2 + x3 ​ ​ = 150
x1​ ​ + x4 ​ = 120
​ ​​ ​ ​ x2 ​ ​ − x5 + x6 = 14​ x1,…, x6 ≥ 0

cj 0 0 -1 0 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
-1 x3 150 5 10 1 0 0 0 15
0 x4 20 1 0 0 1 0 0 -
-1 x6 14 0 1 0 0 -1 1 14
Z= -164
Δj 5 11 0 0 -1 0

cj 0 0 -1 0 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
-1 x3 10 5 0 1 0 10 -10 1
0 x4 20 1 0 0 1 0 0 -
0 x2 14 0 1 0 0 -1 1 -
Z= -10
Δj 5 0 0 0 10 -11

cj 0 0 -1 0 0 -1 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x5 1 1/2 0 1/10 0 1 -1
0 x4 20 1 0 0 1 0 0
0 x2 15 1/2 1 1/10 0 0 0
Z= 0
Δj 0 0 -1 0 0 -1

Optimal solution for Phase I

Phase II

cj -2 -8 0 0 θ
CB XB b a1 a2 a4 a5
0 x5 1 1/2 0 0 1 2
0 x4 20 1 0 1 0 20
-8 x2 15 1/2 1 0 0 30
Z= -120
Δj 2 0 0 0

cj -2 -8 0 0 θ
CB XB b a1 a2 a4 a5
-2 x1 2 1 0 0 2
0 x4 18 0 0 1 -2
-8 x2 14 0 1 0 -1
Z= -116

430
 Δj 0 0 0 -4

This is the optimal solution. Thus use 2 units of A and 14 units of B which will
give the cost of ` 116.

Illustration 10.47: Ashok chemicals company manufactures two chemicals A

and B which are sold to the manufacturer of soaps and detergents. On the
basis of the next month’s demand, the management has decided that the total
production for chemicals A and B should be at least 350 kg. Moreover, a
major customer’s order for 125 kg of product A must also be supplied. Product
A requires 2 hours of processing time per kg and product B requires 1 hour of
processing time per kg. For the coming month, 600 hours of processing time
is available.

The company wants to meet the above requirements at minimum total

production cost. The production costs are Rs 2 per kg for product A and Rs 3
per kg for product B.

Ashok chemicals company wants to determine its optimal product mix and the
total cost relevant thereto.

a.​ Formulate the above as an LPP

b.​ Solve the LPP using simplex method

c.​ Does the problem have multiple optimal solutions? Why?

Solution: Let x1 and x2 be the quantity in kg to be produced of the product A

and B respectively. Then
​ ​ min Z = 2x1 + 3x2​ ​ ​ Total cost
​ ​ s.t. ​ x1 + x2 ≥ 350​ ​ Total output of A and B
x1​ ≥ 125​ ​ Customer order for product A
​ 2x1 + x2 ≤ 600​ ​ Time available
x1, x2 ≥ 0

Augmented Form
​ ​ ​ max (-Z) = - 2x1 - 3x2 – x4 - x6
​ ​ ​ s.t. ​ x1 + x2 + x3 - x4 ​ ​ = 350
x1​ ​ + x5 - x6​ = 125
​ ​ 2x1 + x2​ ​ + x7 = 600
x1, …, x7 ≥ 0

Phase I​ ​ max (-Z) = – x4 - x6

​ ​ ​ s.t. ​ x1 + x2 + x3 - x4 ​ ​ = 350
x1​ ​ + x5 - x6​ = 125
​ ​ 2x1 + x2​ ​ + x7 = 600
x1, …, x7 ≥ 0

cj 0 0 0 -1 0 -1 0 θ
CB XB b a1 a2 a3 a4 a5 a6 a7
-1 x4 350 1 1 -1 1 0 0 0 350

431
 -1 x6 125 1 0 0 0 -1 1 0 125
0 x7 600 2 1 0 0 0 0 1 300
Z= -475
Δj 2 1 -1 0 -1 0 0

cj 0 0 0 -1 0 -1 0 θ
CB XB b a1 a2 a3 a4 a5 a6 a7
-1 x4 225 0 1 -1 1 1 -1 0 225
0 x1 125 1 0 0 0 -1 1 0 -
0 x7 350 0 1 0 0 2 -2 1 175
Z= -225
Δj 0 1 -1 0 1 -2 0

cj 0 0 0 -1 0 -1 0 θ
CB XB b a1 a2 a3 a4 a5 a6 a7
-1 x4 50 0 1/2 -1 1 0 0 -1/2 100
0 x1 300 1 1/2 0 0 0 0 1/2 600
0 x5 175 0 1/2 0 0 1 -1 1/2 350
Z= -50
Δj 0 1/2 -1 0 0 -1 1/2

cj 0 0 0 -1 0 -1 0
CB XB b a1 a2 a3 a4 a5 a6 a7
0 x2 100 0 1 -2 2 0 0 -1
0 x1 250 1 0 1 -1 0 0 1
0 x5 125 0 0 1 -1 1 -1 1
Z= 0
Δj 2 1 -1 0 -1 0 0

Thus Phase I optimal.

Phase II
cj 2 3 0 0 0 θ
CB XB b a1 a2 a3 a5 a7
3 x2 100 0 1 -2 0 -1 -
2 x1 250 1 0 1 0 1 250
0 x5 125 0 0 1 1 1 125
Z= 800
Δj 0 0 4 0 1

cj 2 3 0 0 0 θ
CB XB b a1 a2 a3 a5 a7
3 x2 350 0 1 0 2 1 -
2 x1 125 1 0 0 -1 0 25
0
0 x3 125 0 0 1 1 1 12
5
Z= 800
Δj 0 0 0 -4 -3

Thus the solution is optimal with x1 = 125 unit, x2 = 350 units and Z = 800. The
solution is unique optimal solution and no multiple optimal solutions exist,
because none of the non-basic variables have Δj ≠ 0.

432
 Illustration 10.48: A firm uses three machines in the manufacture of 3
products. Each unit of product A requires 3 hours on machine I, 2 hours on
machine II and 1 hour on machine III. Each unit of Product B requires 4 hours
on machine I, 1 hour on machine II and 3 hours on machine III, while each
unit of Product C requires 2 hours on each of the three machines. The
contribution margin of the three products is ` 30, ` 40 and ` 35 per unit
respectively. The machine hours available on 3 machines are 90, 54 and 93
respectively.

a.​ Formulate the above problem as LPP

b.​ Obtain optimal solution to the problem by using Simplex Algorithm. Which of
these products shall not be produced by the firm?

c.​ Calculate the percentage of capacity utilization in the optimal solution

d.​ What are the shadow prices of the machine hours

e.​ Is the optimal solution degenerate?

Solution:

i.​ Let x1, x2 and x3 be the output of product A, B and C respectively. The LPP
can be formulated as
​ ​ max Z = 30x1 + 40x2 + 35x3
​ ​ s.t. ​ 3x1 + 4x2 + 2x3 ≤ 90
2x1 + x2 + 2x3 ≤ 54
​ x1 + 3x2 + 2x3 ≤ 93​ ​ x1, x2, x3 ≥ 0

ii.​ Using Simplex Algorithm to solve the above LPP

​ Augmented Form
​ ​ max Z = 30x1 + 40x2 + 35x3
​ ​ s.t. ​ 3x1 + 4x2 + 2x3 + x3​​ = 90
2x1 + x2 + 2x3 + x4​ = 54
​ x1 + 3x2 + 2x3 ​ + x5 = 93​ x1, …, x6 ≥ 0

cj 30 40 35 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x5 90 3 4 2 1 0 0 45/2
0 x4 54 2 1 2 0 1 0 54
0 x2 93 1 3 2 0 0 1 31
Z= 0
Δj 30 40 35 0 0 0

cj 30 40 35 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
40 x2 45/2 3/4 1 1/2 1/4 0 0 45
0 x4 63/2 5/4 0 3/2 -1/4 1 0 21
0 x2 51/2 -5/4 0 1/2 -3/4 0 1 51
Z= 900
Δj 0 0 15 -10 0 0

433
 cj 30 40 35 0 0 0
CB XB b a1 a2 a3 a4 a5 a6
40 x2 12 1/3 1 0 1/3 -1/3 0
35 x3 21 5/6 0 1 -1/6 2/3 0
0 x2 15 -5/3 0 0 -2/3 -1/3 1
Z= 1215
Δj -25/2 0 0 -15/2 -10 0

The above Table shows that the optimal solution to the problem is x1 = 0, x2 =
12 and x3 = 21 with Z= 1215. This indicates that product A should not be
produced by the firm. This is because, as indicated by the Δj value (= -25/2)
each unit of x1 produced would reduce the profit by ` 25/2 (=12.50).

iii.​ The capacity available and the capacity used as per optimal solution is:
​ Machine ​ capacity​ ​ Capacity​ Capacity​ percentage
​ ​ ​ ​ ​ ​ Unused​​ utilized​ ​ utilization
​ I​ 90​ 0​ 0(=90-0)​ 100.00%
​ II​ 54​ 0​ 0​ 100.00%
​ ​ III​ 93​ 15​ 78​ 83.87%

​ ​ ​ (Solution values of the slack variables)

iv.​ Shadow prices of the machine hours are given by the Δj value (ignoring the
signs) of the slack variables
​ Machine I -​ ` 15/2 = ` 7.50 / hr
​ Machine II -​ ` 10 / hr
​ Machine III - ​Nil
v.​ The optimal solution is not degenerate because none of the basic variable has
a solution value equal to zero.

Illustration 10.49: A company makes three products X, Y and Z out of three

raw materials A, B and C. the number of units of raw materials required to
produce one unit of the product is given in the table below.
​ X​ Y​ Z
A​ 1​ 2​ 1
B​ 2​ 1​ 4
C​ 2​ 5​ 1

The unit profit contributions of the product X, Y and Z are ` 40, ` 25 and ` 50
respectively. The number of units of raw materials available is 36, 60 and 45
respectively.

i.​ Determine the product mix that will maximize the total profit.
ii.​ From the final simplex table, write the solution to the dual and give the
economic interpretation.

Solution:

i.​ Let the company produce x1, x2 and x3 units of the products X, Y and Z
respectively. The mathematical representation of the problem is
​ ​ max Z = 40x1 + 25x2 + 50x3

434
 ​ ​ s.t. ​ x1 + 2x2 + x3 ≤ 36
2x1 + x2 + 4x3 ≤ 60
​ ​ 2x1 + 5x2 + x3 ≤ 45​ ​ x1, x2, x3 ≥ 0

​ Augmented Form
​ ​ max Z = 40x1 + 25x2 + 50x3
​ ​ s.t. ​ x1 + 2x2 + x3 + x4​ ​ = 36
2x1 + x2 + 4x3 ​ + x5​ = 60
2x1 + 5x2 + x3 ​ + x6 = 45​
x1, …, x6 ≥ 0

cj 40 25 50 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x4 36 1 2 1 1 0 0 36
0 x5 60 2 1 4 0 1 0 15
0 x6 45 2 5 1 0 0 1 45
Z= 0
Δj 40 25 50 0 0 0

cj 40 25 50 0 0 0 θ
CB XB b a1 a2 a3 a4 a5 a6
0 x4 21 1/2 7/4 0 1 -1/4 0 42
0 x3 15 1/2 1/4 1 0 1/4 0 30
0 x6 30 3/2 19/ 0 0 25/2 1 20
4
Z= 75
0
Δj 15 25/ 0 0 -25/2 0
2

cj 40 25 50 0 0 0
CB XB b a1 a2 a3 a4 a5 a6
0 x4 11 0 1/6 0 1 -1/6 -1/3
0 x3 5 0 -4/3 1 0 1/3 -1/3
0 x6 20 1 19/ 0 0 -1/6 2/3
6
Z= 105
0
Δj 0 -35 0 0 -10 -10

Therefore optimal solution to the given primal is x1 = 20 units, x2 = 0 and x3 = 5

units with Z = ` 1050.

ii.​ The dual is:

min ZD = 36w1 + 60w2 + 45w3
s.t. ​ w1 + 2w2 + 2w3 ≥ 40
2w1 + w2 + 5w3 ≥ 25
​ ​ ​ w1 + 5w2 + w3 ≥ 50​ w1, w2, w3 ≥ 0

From the optimal table above, w1 =0 and w2 =10 with min ZD = ` 1050

Economics Interpretation

435
 Suppose the manager of the company wants to sell the three raw materials A,
B, and C instead of using them for making products X, Y and Z and then, by
selling the products earn a profit of ` 1050, per unit of raw materials of A, B
and C respectively.

Then the cost to the purchased of all the three raw materials will be
`(36w1+60w2+45w3). Of course the purchaser will like to set the selling prices
of A, B and C so that the total cost is minimum.

∴ Objective function will be: min 36w1+60w2+45w3

The optimal table above indicates that the marginal values of raw materials A,
B and C are ` 0, ` 10 and ` 10 per unit respectively. Thus if the manager sells
the raw materials A, B and C at price ` 0, ` 10 and ` 10 per unit respectively,
he will get the same contribution of ` 1050 which he is going to get if he uses
these resources for productively products X, Y and Z and then sells them.

Self-Check Questions

State True / False

25.​For an unbounded primal problem, the dual would be infeasible

26.​The net evaluations (ignoring the minus sign) corresponding to the
slack/surplus variables of a problem represent optimal values of the dual
variables.

10.15 Conclusion

Decision-making is an all-pervasive feature of management. A systematic

approach to decision-making calls for application of scientific methods.
Mathematical programming involves optimization of certain objective faction
subject to certain constraints. The LP method is a technique for choosing the
best alternatives from a set of feasible alternatives, in which the objective
function as well as constraints can be expressed as linear mathematical
function. In this Lesson you have learned different methods to solve the LPP:
The graphical method, simplex method and dual method. The simplex method
is a step-by-step and an efficient and reliable algorithm for solving linear
programming problems.

Although it has a useful geometric interpretation, the simplex method is an

algebraic procedure. At each iteration it moves from the current basic feasible
solution to a better, adjacent basic feasible solution by choosing both an
entering basic variable and a leaving basic variable and then using Gaussian
elimination to solve a system of linear equations. When the current solution

436
 has no adjacent BF solution that is better, the current solution is optimal and
the algorithm stops.

In this lesson the full algebraic form of the simplex method is presented to
convey its logic, and then the method is streamlined to provide a more
convenient tabular form. You have also learned to write the dual for a give
primal. The method of deriving a dual from the primal is explained and then
the method to solve the primal and/or the dual is also described. The
economic interpretation of the dual is also given.

10.16 Answers to Self-Check Questions

1.​ a
2.​ d
3.​ d
4.​ d
5.​ d
6.​ Linear
7.​ Decision variables
8.​ Feasible solution
9.​ Feasible region
10.​Constraints
11.​CPF
12.​Positive rate of improvement
13.​True
14.​True
15.​False
16.​False
17.​False
18.​True
19.​False
20.​False
21.​False
22.​True
23.​True
24.​True
25.​True
26.​True

10.17 Terminal Questions

1.​ What is Operations Research?
2.​ “Operations Research is an aid for the executive in making his decision based
on scientific method analysis.” Discuss the above statement in brief.

437
3.​ What is LPP? Explain the general form of LP.
4.​ What are the steps of graphical method to solve an LP?
5.​ A company is producing two types of LPG stoves-Supreme and Cook-n-Grill.
The material and labour requirement in respect of these and the total
availability of these resources are as follows:
Item​ Supreme​ Cook-n-Grill​ Total Availability
Material A​ 2 kg/unit​ 2.5 kg/unit​ 2000 kg
Material B​ 2 pieces/unit​ 3 pieces/unit​ 2400 pieces
Labour Grade I​ 3 hrs/unit​ 5 hrs/unit​ 6000 hrs
Labour Grade II​ 1 hr/unit​ 1 hr/unit​ 900 hrs

Further, according to present conditions, the company earns Rs 40 on the

sale of one unit of Supreme and ` 60 on one unit of Cook-n-Grill. The
company wants to decide about the number of units that it should produce
next month so that it can earn maximum profit. Formulate this as LPP.
6.​ A local travel agent is planning a charter trip to a major sea port. The eight
day and seven night package includes the fare for round trip, surface
transportation, board and lodging and selected tour options. The charter is
restricted to 200 persons and past experience indicates that there will not be
any problem for getting 200 clients. The problem for the travel agent is to
determine the number of Deluxe, Standard and Economy tour packages to
offer for this chapter. These three plans differ according to seating and service
for the flight, quality of accommodations, meal plans and tour options. The
table summarizes the estimated prices for three packages and the
corresponding expenses for the travel agent. The travel agent has hired an
aircraft for the flat fee of Rs 2,00,000 for the entire trip. In planning the trip, the
following consideration must be taken into account
●​ At least 10% of the packages must be of the Deluxe type.
●​ At least 35% but not more than 70% must be of the standard type.
●​ At least 30% must be of the economy type.
●​ The maximum number of Deluxe packages available in any aircraft is
restricted to 60.
●​ The Hotel desires that at least 120 of the tourists should be on the Deluxe
and Standard packages taken together.

Tour Plan Price Hotel Costs Meal and other

Expenses

Deluxe 10,000 2,000 4,750

Standard 7,000 2,200 2,500

Economy 6,500 1,900 2,200

The travel Agent wishes to determine the number of packages to offer in each
type so as to maximise the total profit.

438
 ●​ Formulate this as a linear programming
●​ Restate the above LPP in terms of two decision variables, taking
advantages of the fact that 200 packages will be sold
●​ Find the optimal solution using graphical method for the restated problem
and interpret your results.

7.​ A firm plans to purchase at least 200 quintals of scrap containing high quality
metal X and low quality metal Y. It decides that the scrap to be purchased
must contain at least 100 quintal of X-metal and not more than 35 quintals of
Y-metal. The firm can purchase the scrap from two suppliers, A and B, in
unlimited quantities. The percentage of X and Y metals in terms of weight in
the scraps supplied by A and B is given below:
​ ​ ​ Metals​ ​ Supplier A​ ​ Supplier B
​ ​ ​ X​ ​ 25%​​ ​ 75%
​ ​ ​ Y​ ​ 10%​​ ​ 20%

The price of A’s scrap is Rs 200 per quintal and that of B’s is Rs 400 per
quintal. Formulate and solve it graphically to determine the quantities that the
firm should buy from the two suppliers so as to minimize total purchase cost.
8.​ Define the following terms:

​ ​ a) Objective function, b) Decision variables, c) Optimal solution, d) Feasible

solution, e) unrestricted variables, f) Slack, ​ g) Surplus
9.​ If a LPP possesses a) Multiple optimal solution b) Infeasible solution
c) Unbounded Solution, how can you detect it in course of Simplex
computation?
10.​Explain the use of artificial variable in LP.
11.​Nooh’s Boats makes three different kinds of boats. All can be made profitably
in this company, but the company’s monthly production is constrained by the
limited amount of labour, wood and screws available each month. The director
will choose the combination of boats and that maximizes his revenue in view
of the information given in the following table:
​ ​ Input​ Row Boat​ Cance​ Keyak​ ​ Monthly Available
​ ​ Labour (Hrs)​ 12​ 7​ 9​ 1,260 hrs
​ ​ Wood (Board feet)​ 22 18 16​ 19,008 board feet
​ ​ Screws (Kg)​ 2​ 4​ 3​ 396 kg
​ ​ Selling price (Rs)​ 4000​ 2000​ 5000
a.​ Formulate the LPP
b.​ Solve it by Simplex.
c.​ How many boats of each type will be produced and what will be the
resulting revenue?
d.​ Which, if any, of the resources are not fully utilised? If so, how much of
spare capacity is left?
e.​ How much wood will be used to make all of the boats given in the optimum
solution?

439
 12.​Solve
​ min z = x1 - 3x2 + 2x3
s.t.​ 3x1 - x2 + 3x3 ≤ 7
- 2x1 + 4x2​ ≤ 12
- 4x1 + 3x2 + 8x3 ≤ 10 ​ x1, x2, x3 ≥ 0
13.​Solve
​ a​ max Z = 2x + 3y + 5z​ ​ b.​ max Z = 4x1 + 5x2 - 3x3 + 50
​ ​ s.t.​ x + y - z ≥ - 5​ ​ s.t.​ x1 + x2 + x3 = 10
​ ​ ​ - 6x + 7y - 9z ≤ 4​ ​ ​ x1 - x2 ≥ 1
x + y + 4z = 10​ ​ ​ 2x1 + 3x2 + x3 ≤ 40
​ ​ ​ x, y ≥ 0, z unrestricted in sign ​ ​ ​ x1, x2, x3≥0
Write the dual of the following LPPs and then solve

14.​ max Z = 5x1 + 6x2​ ​ ​ 15.​ min Z = 2x1 + 2x2

s.t. -x1 + 5x2 ≥ 3​ ​ ​ ​ s.t.​ 2x1 + 4x2 ≥ 1
x1 + 2x2 ≥ 5​x1, x2 ≥ 0​ ​ ​ x1 + 2x2 ≥ 1
​ ​ ​ ​ ​ ​ ​ ​ 2x1 + x2 ≥ 1 x1, x2 ≥ 0

16.​ max Z = 4x1 + 3x2 + 6x3 ​ ​ 17.​ min Z = 5x + 6y

s.t.​ x1 + x3 ≥ 2​ ​ ​ ​ s.t.​ x + y ≥ 2
x2 + x3 ≥ 5​ x1, x2 , x3 ≥ 0​ ​ ​ 4x + y ≥ 4 ​ x, y ≥ 0

18.​min Z = x1 + x2​ ​ ​ 19. max Z = - 3x1 - x2

s.t.​ 2x1 + x2 ≥ 2​ ​ ​ s.t. x1 + x2 ≥ 1
- x1 - x2 ≥ 1 x1, x2 ≥ 0​ ​ 2x1 + 3x2 ≥ 2 x1, x2 ≥ 0

10.18 Further Readings

1.​ F.S. Hillier and G.J Lieberman, Introduction to Operations Research, 7th Ed.,
Tata McGraw Hill Publishing Company Limited, New Delhi.
2.​ K. Swarup, [Link] and M. Mohan, Operations Research, 13th Ed., Sultan
Chand and Sons Education Publishers, New Delhi
3.​ N.D. Vohra, Quantitative Techniques in Management, 4th Ed., Tata McGraw
Hill Education Private Limited, New Delhi.

10.19 Glossary of the terms used

●​ Constraints: There are always certain limitations (or constraints) on the use of
resources, such as labor, space, raw material, money, etc. that limit the
degree to which an objective can be achieved. Such constraints must be
expressed as linear inequalities or equalities in terms of decision variables.
●​ Decision variables: Decision variables refer to the activities that are competing
one another for sharing the resources available.

440
●​ Dual: For every LPP, there is another LPP which is related to it and therefore,
may be derived from it. This original LPP is called the Primal and the
associated problem is called the Dual.
●​ Feasible Region: Collection of all feasible solutions represents the feasible
region.
●​ Feasible Solution: Any solution to a general LPP which also satisfies the
non-negative restrictions of the problem is called a feasible solution to the
general LPP.
●​ Infeasible Solution: ​A solution for which at least one constraint is violated is
called an infeasible solution.
●​ Linear programming: Linear Programming is a technique for determining an
optimum schedule of interdependent activities in view of the available
resources.
●​ Objective: A linear programming problem must have an objective which
should be clearly identifiable and measurable in quantitative terms. It could be
of profit (sales) maximization, cost (time) minimization, and so on. The
relationship among the variables representing objective must be linear.
●​ Optimal solution: The feasible solution that has the most favorable value of
the objective function is called an optimal solution. Most Favorable Solution
means largest value if objective function is to be maximized, or lowest value if
objective function is to be minimized.
●​ Redundant Constraints: If a constraint when plotted does not form part of the
boundary making the feasible region of the problem, it is said to be redundant
constraint. Inclusion or exclusion of a redundant constraint does not affect the
optimal solution.
●​ Shadow prices: The variables of the dual are known as dual variables or
shadow prices of the resources.
●​ Solution: A solution is any specification of values for decision variables
(x1,…,xn).
●​ Unbounded solution: When the values of the decision variables may be
increased indefinitely without violating any of the constraints, the solution
space (feasible region) is unbounded. The value of objective function, in such
cases, may increase (for maximization) or decrease (for minimization)
indefinitely. Thus, both the solution space and the objective function value are
unbounded.

441
MBM 106​ ​ Unit II, Lesson 5

LESSON 11 ​ THE TRANSPORTATION PROBLEM

11.0​OBJECTIVES​ 441
11.1​INTRODUCTION​ 441
11.2​THE TRANSPORTATION PROBLEM​ 443
11.3​THE TRANSPORTATION PROBLEM TABLE​ 443
11.4​EXISTENCE OF SOLUTION IN TRANSPORTATION PROBLEM​ 448
11.5​LOOPS IN TRANSPORTATION PROBLEM TABLE​ 448
11.6​SOLUTION TO THE TRANSPORTATION PROBLEM​ 449
11.6.1 Finding the initial BFS​ 450
11.6.2 The Modified Distribution (MODI) Method​ 458
11.7​SPECIAL CASES IN TRANSPORTATION PROBLEM​ 465
11.7.1 Unbalanced Transportation Problem​ 465
11.7.2 Different Production Costs​ 468
11.7.3 Prohibited Routes​ 468
11.7.4 Multiple Optimal Solutions​ 470
11.7.5 Degeneracy​ 472
11.7.6 Maximization Problem​ 477
MISCELLANEOUS ILLUSTRATIONS​ 480
11.8​CONCLUSION​ 485
11.9​ANSWERS TO SELF-CHECK QUESTIONS​ 486
11.10 TERMINAL QUESTIONS​ 486
11.11 FURTHER READINGS​ 488
11.12 GLOSSARY OF THE TERMS USED​ 488
MBM 106​ ​ Unit IV, Lesson 11

11. The Transportation Problem

The distribution of goods produced by a factory from various warehouses (or

sources) to different markets (or destinations) where they are required, causes
problems to almost every businesses. The Transportation Method is developed to
deal with the transportation of goods from different sources to different destinations,
given relevant data like available quantities at various sources, demand at each
destinations, the cost of shipping along each route and non-availability of certain
routes (if any).

11.0​ Objectives

After going through this lesson you will be able to understand:

●​ The concept of Transportation Problem

●​ Finding initial solution to Transportation Problem
●​ Determining the optimal solution to the Transportation Problem
●​ Variants of Transportation Problem

11.1​ Introduction
In the previous lesson, solutions to the general LPP have been discussed.
The Simplex Algorithm can be used to solve any LPP for which the solution
exists but computation by this method becomes more and more laborious as
the number of variables and constraints increase. Therefore to simplify the
calculations there is a special branch of LP that comprises problems of the
transportation type. The term ‘transportation’ is associated with such problems
principally because in studying efficient transportation routes, a special
procedure was used which is known as transportation method. The origin of
transpiration models dates back to 1941 by F.L. Hitchcock. This method can
be used for a variety of situation like scheduling, production, investment, plant
location, inventory control, etc.

The method allows managers to seek answers to questions like:

●​ What is the optimal way of shipping goods from various warehouses to
different markets so as to minimize the total cost involved in the shipping?
●​ How to handle a situation when some routes are not available or when some
units have to be necessarily transported from a particular source to market?
●​ How would the optimal shipping schedule change if some routes become
cheaper / costlier?

443
MBM 106​ ​ Unit IV, Lesson 11

Example

Suppose a manufacturer of refrigerators has three plants situated at places A,

B and C. Suppose further that his buyers are located in three regions X, Y and
Z where he has to supply them the goods. Also, assume that the demand in
the regions and the production in different plants per unit time period are
known and equal in aggregate and further that the cost of transporting one
refrigerator from each plant to each of the requirement centers is given and
constant. The manufacturer’s problem is to determine as to how he should
route his product from plants to the market places so that the total cost
involved in the transportation is minimized. In other words, he has to decide
as to how many refrigerators should he supplied from A to X, Y and Z; how
many from B to X, Y and Z; and how many from C to X, Y and Z, to achieve
the least cost (Figure 11.1).

Fig 11.1 Transportation Problem

The places where the goods originate from (plants in the example) are called
the sources or the origins and places where they are to be supplied are the
destinations. Thus the transportation problem can be stated as the problem of
manufacturer to decide as to how many units may be transported from
different origins to different destinations so that the total transportation cost is
minimum. The distinct feature of transportation problems is that sources and
destinations must be expressed in terms of only one kind of unit.

Assumptions in the Transportation Problem are:

1.​ Total quality of the item available at different sources is equal to the total
requirements at different destination.
2.​ Items can be transported conveniently from all sources to destinations.
3.​ The unit transportation cost of the items from all sources to destinations is
certainly and precisely known.
4.​ The transportation cost on a given route is directly proportional to the number
of units shipped on that route.
5.​ The objective is to minimize the total transpiration cost for the organization as
a whole and not for individual supply and distribution centre.

444
MBM 106​ ​ Unit IV, Lesson 11

11.2​ The Transportation Problem

The classical transportation problem can be stated mathematically as follows:

Suppose there are m sources and n destinations. Let si = quantity of product

available at source i, di = quantity of product required at destination j, cij = the
cost of transporting one unit of product from origin i to destination j and xij =
the quantity transported from origin i to destination j. The objective is to
determine the number of units to be transported from source i to destination j
so that the total transportation cost is minimum. In addition, the supply limits
at the sources and the demand requirements at the destinations must be
satisfied exactly.
Mathematically, find xij (i=1,…,m; j=1,..,n) in order to

Minimize​ Z= ​ ​ ​ ​ ​ (11.1)

Subject to​ ​ ​ ( i = 1, 2 ,.., m)​ ​ (11.2)

​ ​ ​ ​ ​ (j = 1, 2, …,n) ​ ​ (11.3)

​ ​ and​ ​ ​ xij > 0 (i = 1, …, m; j = 1, .., n)​ (11.4)

Here Z is the total cost.

The two sets of constraints will be consistent, i.e. the system will be balanced
if

​​ ​ ​ ​ (11.5)

which means that the total quantity available at the origins is precisely equal
to the total amount at the destinations?

Equity sign of the constraint causes one of the constraints to be redundant

(and hence can be deleted) so that the problem will have (m + n – 1)
constraints and (m×n) unknowns.

A transportation problem has a feasible solution only if the above restriction

(11.5) is satisfied. Thus equation (11.5) is a necessary as well as a sufficient
condition for a transportation problem to have a feasible solution.

11.3​ The Transportation Problem Table

445
MBM 106​ ​ Unit IV, Lesson 11

Since the TP is just a special case of general LPP the application of simplex
method would give an optimum solution to the problem. However, simplex like
method for solving such problems has been developed as it is simple to solve
it by transportation method than by the simplex method. The transportation
model can also be portrayed in a tabular form by means of a transportation
tableau, shown in the Table 11.1.
Table 11.1: The Transportation Tableau

Source Destination j Supply

i si
1 2 … n

1 c11 c12 … c1n s1

xij

2 c21 c22 … c2n s2

… … … … … ...

m cm1 cm1 … cmn sm

Demand d1 d2 … dn ∑si = ∑dj

, dj

The tableau can be thought of as a matrix within a matrix, of dimension, m ×

n. The cij is per unit cost of shipping a unit from the ith origin to jth destination.
xij is the amount shipped from ith source to jth destination. Right and bottom
sides of the transportation tableau shows, respectively, the amount of supplies
si available at source i and amount demanded dj at each destination j. The si’s
and dj’s represent the supply and demand constraints. The mn large squares
are called the cells. The cell located at the intersection of a row data column is
designated by its row and column headings. Thus the cell located at the
intersection of row A and column 3 is called cell (A,3). Unit costs are placed in
each cell.

The aggregate transportation cost is determined by multiplying the various xij’s

with corresponding cij’s and then adding them up all. The solution to the
transportation problem calls for determining values of xij’s as would yield the
minimum aggregate transportation cost. When a problem is solved, some of
the xij’s would assume positive values indicating utilized routes. The cells
containing xij values are called occupied or filled cells and each of them
represents the presence of a basic variable. For the remaining cells, called
the empty cells, xij’s would be zero. These are the routes that are not utilized
by the transportation pattern and their corresponding variables (xij’s) are
regarded to be non-basic.

Remember that the transportation costs are assumed to be linear in nature.

The number of constraints is a transportation tableau is m + n where m =

446
MBM 106​ ​ Unit IV, Lesson 11

number of rows and n = number of columns. The number of variables required

for forming a basis is (m + n – 1). This is because there are only (m + n -1)
independent variables in the solution basis. Considering the conditions of
feasibility and non-negativity, the number of basic variables representing
transportation routes utilized equals (m+n-1), and all other variables are
non-basic or zero, representing the unutilized routes.

Definitions

1.​ Feasible Solution:​ A feasible solution to a transportation problem is a

set of non-negative allocations xij that satisfy row
and column restrictions.

2.​ Basic Feasible Solution:​ A feasible solution to a transportation problem is

said to be a Basic Feasible Solution (BFS) if it
contains no more than (m + n -1) non-negative
allocations, where m is the number of rows and n is
the number of columns of the transportation
problem.

3.​ Optimal Solution:​ A feasible solution (not necessarily basic) that

minimizes (maximizes) the transportation cost
(profit) is called an optimal solution.

4.​ Non-degenerate Solution:​ A basic feasible solution to a m×n transportation

problem is said to be non-degenerate if

​ a.​ total number of non-negative allocations is

exactly (m + n – 1) and

​ b.​ these (m + n – 1) allocations are in

independent position.

5.​ Degenerate BFS:​ A BFS in which the total number of non-negative

allocations is less than (m+n–1) is called degenerate
BFS.

Formation of Transportation Problem

Illustration 11.1: ABC limited has three productions shops supplying a

product to five warehouses. The cost of production varies from shop to shop
and cost of transportation from one shop to a warehouse also varies. Each
shop has a specific production capacity and each warehouse has certain
amount of requirement. The costs of transportation are as given below:
Warehouse
Shop​ ​ I​ II​ III​ IV​ V​ Supply
A​ ​ 6​ 4​ 4​ 7​ 5​ 100
B​ ​ 5​ 6​ 7​ 4​ 8​ 125
C​ ​ 3​ 4​ 6​ 3​ 4​ 175
Demand 60 80 85 105 70​ 400

The cost of manufacturing of the product at different production shop is

447
MBM 106​ ​ Unit IV, Lesson 11

Shop​ ​ Variable Cost​ ​ Fixed Cost

A​ ​ ​ 14​ ​ 7000
B​ ​ ​ 16​ ​ 4000
C​ ​ ​ 15​ ​ 5000

Find the optimum quantity to be supplied from each shop to different

warehouse at minimum total cost.

Solution: In the cost matrix, the data pertaining to fixed costs is of no use.
Thus only including the costs plus the variable costs as shown below:
Table 11.2: The TP Tableau

Shop Warehouse si
​
I II III IV V

A 6+14=20 4+14=18 4+14=18 7+14=21 5+14=19 100

B 5+16=21 6+16=22 7+16=23 4+16=20 8+16=24 125

C 3+15=18 4+15=19 6+15=21 3+15=18 4+15=19 175

dj 60 80 85 70 70

Thus the final TP Table is

Table 11.3: The Starting TP Tableau

Shop Warehouse si
​
I II III IV V

A 20 18 18 21 19 100

B 21 22 23 20 24 125

C 18 19 21 18 19 175

dj 60 80 85 70 70

Illustration 11.2: ABC Company wishes to develop a monthly production

schedule for the next three months depending upon the sales commitments,
the company can either keep the production constant, allowing fluctuations in
inventory or inventories can be maintained at a constant level, with fluctuating
production.

Fluctuating production necessitate in working overtime, the cost of which is

estimated to be double the normal production cost of `12 per unit. Fluctuating
inventories result in inventory carrying cost of ` 2 per unit. If the company fails

448
MBM 106​ ​ Unit IV, Lesson 11

to fulfill its sale commitment; it incurs a shortage cost of ` 4 per unit per month.
The production capacities for the next three months are shown below:
Month​ ​ Production​ Capacity​ Sales
​ ​ ​ Regular (R)​ Overtime (O)
​ 1​ ​ 50​ ​ 30​ ​ 60
​ 2​ ​ 50​ ​ 0​ ​ 120
​ 3​ ​ 60​ ​ 50​ ​ 40

Determine the optimal production schedule.

Solution: The cost elements in each cell are determined as follows:

1.​ If items are produced in a month for sales during the same month, then there
will be no inventory carrying cost. Thus total cost will be either the normal
production cost or overtime production cost. Thus the cost elements for cell
(R1, 1) (R2, 2) (R3, 3) are Rs 12 each and for cells (O1, 1), (O2, 2) and (O3, 3)
are ` 24 each.
2.​ If items are produced in a month for sales during subsequent months in
addition to the production costs (normal or overtime), inventory carrying cost
at the rate of ` 2 per month will be incurred).

Cell Production Inventory Total

Cost Carrying Cost Cost
(`) (`) (`)

(R1,2) 12 2 14

(R1,3) 12 4 16

(O1,2) 24 2 26

(O1,3) 24 4 28

(R2,3) 12 2 14

3.​ If the company fails to fulfill its sales commitment, it incurs a shortage cost of `
4 per unit per month. In addition to the production costs (normal or overtime),
carrying (shortage cost at the rate of ` 2 per month will be incurred.

Cell Production Inventory Total

Cost Carrying Cost Cost
(`) (`) (`)

(R2,1) 12 4 16

(R3,2) 12 4 16

(R3,1) 12 8 20

(O3,2) 24 4 28

449
MBM 106​ ​ Unit IV, Lesson 11

(O3,1) 24 8 32

Using the above information, the given problem can be presented in tabular
form as follows:
Table 11.4: The TP Tableau
Month Total Cost (`) Demand
R1 O1 R2 R3 O3
M1 12 24 16 20 32 60
M2 14 26 12 16 28 120
M3 16 28 14 12 24 40
Supply 50 30 50 60 50

Self-Check Questions

Fill in the blank

1.​ A ______________ solution to a transportation problem is a set of

non-negative allocations xij that satisfy row and column restrictions.
2.​ A feasible solution to a transportation problem is said to be a ____________ if
it contains no more than (m + n -1) non-negative allocations.
3.​ The cells containing xij values are called _____________cells and each of
them represents the presence of a _________ variable.
4.​ Transportation costs are assumed to be _______________ in nature.

11.4​ Existence of solution in Transportation Problem

The general transportation problem always has a solution.

Theorem 11.1: Existence of feasible solution: A necessary and sufficient

condition for the existence of a feasible solution to the general transportation
problem that is

Total Supply = Total demand

i.e.​

Theorem 11.2: Existence of an optimum solution: There always exists an

optimum solution to a TP.

Theorem 11.3: The number of basic (decision) variables of the general

transportation problem at any stage of feasible solution must be (m+n -1).

450
MBM 106​ ​ Unit IV, Lesson 11

11.5​ Loops in Transportation Problem Table

Loop: In a transportation table, an ordered set of four or more cells is said to
form a Loop if

a.​ Any two adjacent cells in the ordered set lie either in the same row or in
the same column, and
b.​ Any three or more adjacent cells in the ordered set do not lie in the same
row or the same column.

The first cell of the set is considered to follow the last in the set, i.e. each cell
(except the last) must appear only once in the ordered set.

Remarks:
1.​ Every loop has an even number of cells.
2.​ The allocation are said to be in independent positions if it is not possible to
increase or decrease any independent individual allocation without changing
the positions of these allocations, or violating the rim conditions, i.e. a closed
loop cannot be formed through these allocations.
3.​ Each row and column in the transportation table should have only one (+) and
(-) sign. All cells that have a (+) or (-) signs, except the starting unoccupied
(non-basic) cell are occupied (basic) cells.
4.​ Closed loops may or may not be in square shape. Following are the possible
shapes of the loops.

Fig 11.2: Possible Shapes of Closed Loops

A feasible solution to a TP is basic if the corresponding cell in the

transportation table does not contain a loop.

Example: Consider the following two cases represented in Table 11.5(a) and
(b).

In Table 11.5(a) if we join the positive allocation by horizontal and vertical

lines, a closed loop is obtained. Here the ordered set of cells forming a loop is

L = {(4, 2), (2, 2), (2, 3), (1, 3), (1, 5), (4, 5)}

The loop in Table 11.5(b) is not allowed, because it does not satisfy the
conditions of a loop.

451
MBM 106​ ​ Unit IV, Lesson 11

Table 11.5: Closed Loop Examples

​​
(a)​ ​ ​ ​ (b)

11.6​ Solution to the Transportation Problem

The transportation problem solution steps are:

1.​ Formulate the given problem as a TP

2.​ Setup the TP in the tabular form known as transportation problem tableau.

3.​ Examine whether . If not introduce a dummy row or column

having all its cost elements as zero and supply or demand as the difference of
supply and demand (or rim requirements).

4.​ Find an initial BFS that must satisfy all the supply and demand constraints.

5.​ Examine the solution obtained in Step 4 for optimality.

6.​ If the solution is not optimum, modify the shipping schedule by including that
unoccupied cell whose inclusion may result in an improved solution.

7.​ Repeat Step 4 until no further improvement is possible.

Methods to find an initial BFS and the method for attaining optimum solution
are discussed in the following sections.

11.6.1 Finding the initial BFS

The first step in using the transportation method is to obtain a feasible

solution, namely, the one that satisfies the rim requirements. The initial
feasible solution can be obtained by several methods. The commonly used
are:
1.​ North-West Corner (NWC) Rule,
2.​ Least Cost Method (LCM) and
3.​ Vogel’s Approximation Method (VAM).

1.​ North-West Corner Rule

452
MBM 106​ ​ Unit IV, Lesson 11

The steps of the North-West Corner (NWC) Rule are as follows:

Step 1:​ Start with the north-west (upper-left) corner of the transportation
tableau and consider the cell in the first column and the first row.
Corresponding to this cell are the values s1 and d1 respectively in
row 1 and column 1. Compare the s1 with d1:

i)​ If s1 > d1, i.e. if the amount available at s1 is greater than the amount
required at d1, then set x11 = d1, and find the balance supply and
demand and proceed to cell (1,2), i.e. proceed vertically.

ii)​ If s1 = d1, set x11 = d1, compute the balance supply and demand and
proceed to cell (2,2), i.e. proceed diagonally. Also make a ε (ε→0)
allocation to the least cost cell in s1 or d1.

iii)​ If s1 < d1, set x11 = s1, and find the balance supply and demand and
proceed to cell (2,1), i.e. proceed horizontally.

Step 2:​ Continue in this manner, step by step, away from north-west corner
until, finally, a value is reached in the south-east corner.

Illustration 11.3: A firm owns facilities at seven places. It has manufacturing

plants at places A, B and C with daily output of 500, 300 and 200 units of an
item respectively. It has warehouses at places P, Q, R, and S with daily
requirements of 180, 150, 350 and 320 units respectively. Per unit shipping
charges on different routes are given below.
To:​ ​ P​ ​ Q​ ​ R​ ​ S
From A:​​ 12​ ​ 10​ ​ 12​ ​ 13
From B:​​ 07​ ​ 11​ ​ 08​ ​ 14
From C:​ 06​ ​ 16​ ​ 11​ ​ 07

The firm wants to send the output from various plants to warehouses involving
minimum transportation cost. How should it route the products so as to
achieve its objective?

Solution: Constructing the transportation table for the problem:

To si

P Q R S

A 12 10 12 13 500

From
B 7 11 8 14 300

C 6 16 11 7 200

dj 180 150 350 320 1000

Table 11.6: Transportation Tableau

453
MBM 106​ ​ Unit IV, Lesson 11

Step 1:​ Starting with the north-west (upper-left) corner, the first cell of the
transportation tableau and compare the s1 = 500 with d1= 180. As

​ ​ ​ ​ s1 (= 500) > d1 (= 180), ​ ​ set x11 = 180 (d1).

​ ​ Balance supply s1 = 500 – 180 = 320

​ ​ ​ and demand d1 = 180 – 180 = 0.

​ ​ Proceed to cell (1, 2)

Step 2a:​ Now comparing the s1 with d2

​ ​ ​ Again​ s1 (= 320) > d2 (= 150), ​ ​ set x12 = 150 (d2).

​ ​ Balance supply s1 = 320 – 150 = 170

​ ​ ​ and demand d2 = 150 – 150 = 0.

​ ​ Proceed to cell (1, 3)

Step 2b:​ Comparing the s1 with d3

​ ​ ​ This time ​ s1 (= 170) < d3 (= 350), ​ set x13 = 170 (s1).

​ ​ Balance supply s1 = 170 – 170 = 0

​ ​ ​ and demand d3 = 350 – 170 = 180.

​ ​ Proceed to cell (2, 3)

Step 2c:​ Comparing the s2 with d3

​ ​ ​ Again​ s2 (= 300) > d3 (= 180), ​ ​ set x23 = 180 (d3).

​ ​ Balance supply s2 = 300 – 180 = 120

​ ​ ​ and demand d3 = 180 – 180 = 0.

​ ​ Proceed to cell (2, 4)

Step 2d:​ Comparing the s2 with d4

​ ​ ​ ​ s2 (= 120) < d4 (= 320), ​ ​ set x24 = 120 (s2).

​ ​ Balance supply s2 = 120 – 120 = 0

​ ​ ​ and demand d4 = 320 – 120 = 200.

​ ​ Proceed to cell (3, 4)

Step 2e:​ Comparing the s3 with d4

454
MBM 106​ ​ Unit IV, Lesson 11

​ ​ ​ Again​ s3 (= 200) > d4 (= 200), ​ ​ set x34 = 200 (d4).

​ ​ Balance supply s3 = 200 – 200 = 0

​ ​ ​ and demand d3 = 200 – 200 = 0.

​ As we have reached the south-east corner and all the row and
column requirements are satisfied, we stop. This solution is a
feasible solution as all supply and demand constraints are fully
satisfied. Table 11.7 shows the initial solution.
Table 11.7: Initial Feasible Solution: NWC Method

P Q R S si
To
From

A 12 10 12 13 500
180 150 170

B 7 11 8 14 300
180 120

C 6 16 11 7 200
200

Deman 180 150 350 320 Z= 10,220

d

Total Cost (Z) = 12×180 + 10×150 + 12×170 + 8×180 + 14×120 +

7×200
= ` 10,220

This routing of the units meets all the row and column requirements and
involves a total of 6 (= 3 + 4 – 1) shipments as there are 6 occupied cells. It
involves a total cost of Rs 10,220.

The following points may be noticed:

1.​ The quantities allocated are placed at the right of the cell in bold and they
represent the values of the corresponding decision variables (xij). These cells
are called basic (allocated or occupied) cells. Cells without allocations are
called non-basic (vacant or un-occupied) cells. Values of the corresponding
variables are all zero in these cells.

2.​ Although the method is simple, but is inefficient as the method of allocation
does not take into account the transportation cost and therefore, may not yield
a good (i.e., most economical) initial solution.

Illustration 11.4: Consider the following TP, find its initial BFS using NWC
Rule

Distribution Center

455
MBM 106​ ​ Unit IV, Lesson 11

A B C D si

Plants 1 2 3 11 7 6

2 1 0 6 1 1

3 5 8 15 9 10

dj 7 5 3 2

Table 11.8

Solution:
Table 11.9 Initial Feasible Solution: NWC Method

Distribution Center

A B C D si

Plants 1 2 3 11 7 6/0
6

2 1 0 6 1 1/0
1

3 5 8 15 9 10/5/3/
5 3 2 2/0

dj 7/1/0 5/0 3/0 2/0

After allocating in cell (2, A) si = dj therefore move diagonally to the cell (3, B).
Here the number of allocations = 5 < (m+n-1), the initial solution is said to be
degenerate transportation solution.

2.​ Least Cost Method (LCM)

As the name suggests, this method entails making allocation at each step by
selecting from the routes (that is, combinations of sources and destinations)
available, the one with the minimum cost. To begin, choose the cell which has
the minimum cost, consider the supply available at the source (si) and the
demand at the destination (dj) corresponding to that, and allocate the lower of
the two (si or dj) to the cell. Then delete the row or column (or both if si = dj)
whichever is satisfied by this allocation. If the row is deleted, then the column
value (dj) is revised by subtracting the quantity assigned. Similarly, if the
column is deleted, then the row total is revised. Next consider the routes
remaining and again choose the one with the least cost, make assignment
and adjust the row/column total. Continue in this manner until all the units are
assigned.

It may be mentioned that at any stage, if there is a tie in the minimum cost, so
that two or more routes have the same least cost of shipping, then,

456
MBM 106​ ​ Unit IV, Lesson 11

conceptually, either of them may be selected. However, a better initial solution

is obtained if the route chosen is the one where largest quantity can be
assigned. Thus, if there are three cells for which the (least) cost value is
equal, then consider all of these one by one and determine the quantity (by
reference to the demand and supply quantities given) which can be
dispatched, and choose the cell with the largest quantity. If there is still a tie,
then either of them may be selected.

For Illustration 11.3, the initial solution by using least cost method is obtained
as shown in Table 11.10. To begin with, the lowest of all cost elements is 6, for
the cell CP, with corresponding supply and demand being 200 and 180 units
respectively. Accordingly, assign 180 to this cell, delete the column headed P
and adjust the supply in the row to 20. Of the remaining cost elements
(excluding those of the deleted column), the least is 7, corresponding to the
cell CS. The quantity assigned to this cell is 20, being lower of demand (=320)
and supply (=20). Delete the row C and adjust the demand for S at 300. The
least cost element in the remaining cells is 8, for BR. With the relevant supply
and demand values equal to 300 and 350 units respectively, assign 300 to the
cell BR, delete the row B and adjust the demand at R to 50. Now, with only
one supply source remaining, the amount would be transferred to the current
requirements at warehouses Q, R, and S: 150 to Q, 50 to R, and 300 to S.
Table 11.10: Obtaining Initial Feasible Solution: LC Method

P Q R S si

A 12 10 12 13 500/350/0
150 50 300

B 7 11 8 14 300/0
300

C 6 16 11 7 200/20/0
180 20

Demand 180/0 150/0 350/50/0 320/300/0

Total Cost Z = 10 × 150 + 12 × 50 + 13 × 300 + 08 × 300 + 06 × 180 + 07 × 20

= ` 9,620

The solution has six occupied cells and it is seen to involve a total cost of `
9,620.

Table 11.11: Initial Feasible Solution: Least Cost Method

P Q R S si

A 12 10 12 13 500
150 50 300

457
MBM 106​ ​ Unit IV, Lesson 11

B 7 11 8 14 300
300

C 6 16 11 7 200
180 20

Demand 180 150 350 320 Z = ` 9,620

Illustration 11.5: Consider the following TP, find its initial BFS using LCM

Distribution Center

A B C D si
Plants 1 2 3 11 7 6
2 1 0 6 1 1
3 5 8 15 9 10
dj 7 5 3 2
Solution:
Table 11.12: Initial Feasible Solution: LC Method

Distribution Center

A B C D si

Plants 1 2 3 11 7 6/0
6

2 1 0 6 1 1/0
1

3 5 8 15 9 10/9/5/0
1 4 3 2

dj 7/1/0 5/4/0 3/0 2/0 Z= `11,200

3.​ Vogel’s Approximation Method

Vogel’s Approximation Method, commonly known as VAM, is another

cost-based method used for obtaining initial solution to a transportation
problem. Actually, it uses cost differentials rather than absolute costs, to select
the appropriate cell to make allocations. VAM makes effective use of the cost
information and yields a better initial solution rather than obtained by other
methods. The steps are as follows:
Step 1:​ First, consider every row of the cost matrix individually and find the
difference between two least cost cells in it. Then, repeat the
exercise for every column. These differences are referred to as Row
Penalty and Column Penalty.
Step 2:​ Consider all cost differences and identify the row or column with the
largest value.

458
MBM 106​ ​ Unit IV, Lesson 11

Step 3:​ Focus on the cell with the smallest cost in the row or column so
chosen and make allocation in that. The number of units allocated
would be lower of the corresponding supply and demand, i.e.
allocated min (si, dj).
Step 4:​ Cross out the row or column which has been satisfied by the
allocation and adjust the quantity of demand/supply. Recalculate the
cost differences for the reduced matrix and proceed in the same
manner as discussed earlier.
Step 5:​ Continue with the process until all units have been assigned.

Note:​ In case of tie among the highest penalties, select the row or column
having minimum cost. In case of a tie in the minimum cost also, select the cell
which can have maximum allocation. If there is tie among maximum cells
also, select the cell arbitrarily for allocations. Following these rules yields the
best possible initial BFS and reduces the number of iterations required to
reach the optimal solution.

The initial solution to Illustration 11.3 using VAM is obtained in Table 11.13.

Step 1:​ Compute the row penalty and column penalty for each row and
column by computing cost differences between the pairs of least
cost cells for each row and column.
​ ​ ​ For ​ I row,​ Row Penalty 12 – 10 = 2
II row,​ Row Penalty 8 – 7 = 1
III row, Row Penalty 6 – 7 = 1
Similarly
For​ I Column, Column Penalty​​ 6–7=1
II Column, Column Penalty​ 11 – 10 = 1
III Column, Column Penalty​ 11 – 8 = 3
IV Column, Column Penalty​ 13 – 7 = 6
Table 11.13(a) Initial Solution by VAM

P Q R S si R.P
To .
From

A 12 10 12 13 500 2

B 7 11 8 14 300 1

C 6 16 11 7 200 1

Demand 180 150 350 320

C.P. 1 1 3 6↑

459
MBM 106​ ​ Unit IV, Lesson 11

Step 2:​ The largest penalty of these being 6 (= 13 – 7), i.e. choose the 4th
column.

Step 3:​ In this column the minimum cost is 7, i.e. in cell (3,4), therefore
allocation is done in this cell, i.e.
​ ​ ​ ​ x34 ​ = min (s3, d4)
​ ​ ​ ​ ​ = min (200, 320) = 200
​ ​ ​ ​ Recalculate the supply s3 = 200 – 200 = 0
​ ​ ​ ​ ​ and the demand d4 = 320 – 200 = 120
Step 4:​ As s3 = 0, i.e., for the 3rd row the supply is completely satisfied,
therefore cross the 3rd row.
Table 11.13(b) Initial Solution by VAM

P Q R S si R.P
To .
From

A 12 10 12 13 500 2

B 7 11 8 14 300 1

C 6 16 11 7 200/ 1
200 0

Demand 180 150 350 320/12

0

C.P. 1 1 3 6↑

Step 5:​ Calculating the penalties for the reduced matrix and proceeding as
above we get the following tables
Table 11.13(c) Initial Solution by VAM

P Q R S si R.P
To .
From

A 12 10 12 13 500 2

B 7 11 8 14 300/12 1
180 0

Deman 180/0 150 350 320/12

d 0

C.P. 5↑ 1 4 1

460
MBM 106​ ​ Unit IV, Lesson 11

Table 11.13(d) Initial Solution by VAM

Q R S si R.P
To .
From

A 10 12 13 500 2

B 11 8 14 300/120/ 3
120 0

Deman 150 350/23 320/12

d 0 0

C.P. 1 4↑ 1

Table 11.13(e) Initial Solution by VAM

Q R S si R.P
To .
From

A 10 12 13 500/280/120/ 2
150 230 120 0

Deman 150/0 350/230/ 320/120/

d 0 0

C.P.

Thus the final initial BFS is

Table 11.14: Obtained Initial Feasible Solution: VAM

P Q R S si
To
From

A 12 10 12 13 500
150 230 120

B 7 11 8 14 300
180 120

C 6 16 11 7 200
200

Demand 180 150 350 320 Z= ` 9,440

Total Cost Z = 10 × 150 + 12 × 230 + 13 × 120 + 7 × 180 + 8 × 120 + 7 × 200

461
MBM 106​ ​ Unit IV, Lesson 11

= ` 9,440
Notice that the initial solution has a total of six occupied cells. It involves a
total cost of ` 9,440. The following points may be noted.
a)​ Vogel’s approximation method is also called the Penalty Method because the
cost differences that it uses are nothing but the penalties of not using the
least-cost routes. Since the objective function is the minimization of the
transportation cost, in each iteration that route is selected which involves the
maximum penalty of not being used.
b)​ In operation of this method that when a row is deleted, the other row
differences do not change and the column cost differences need to be
recalculated, and if a column is deleted then the row difference require
recalculation. In the event of both row and column being deleted (when si = dj)
then all cost difference need to be calculated again.
c)​ The initial solution (to Illustration 11.3) obtained by VAM is found to be the
best of all as it involves the lowest total cost among all the three initial
solutions. Of course, it does not come as a rule, but usually VAM provides the
best initial solution. Therefore, this method is generally preferred in solving the
problems.

Illustration 11.6: Consider the following TP, find its initial BFS using VAM

Distribution Center

A B C D si

Plants 1 2 3 11 7 6

2 1 0 6 1 1

3 5 8 15 9 10

dj 7 5 3 2

Solution:
Table 11.15 Initial Feasible Solution: VAM

Distribution Center Row Penalty

A B C D si

Plants 1 2 3 11 7 6/0 1 1 5 - -
1 5 ←

2 1 0 6 1 1/0 1 - - - -
1

3 5 8 15 9 10/7/0 3 3 4 4 4
6 3 1

462
MBM 106​ ​ Unit IV, Lesson 11

dj 7/6/0 5/0 3/0 2/1/0 Z= `10200

Column 1 3 5 6↑
Penalty
3 5↑ 4 2

3 - 4 2

5 - 15↑ 9

5↑ - - 2

11.6.2 The Modified Distribution (MODI) Method

An optimality test is done to find whether the obtained feasible solution is
optimal or not. This test can be performed only on that feasible solution in
which:

a.​ Number of allocations is (m + n – 1), where m is the number of rows and n is

the number of columns. In Illustration 11.3, m=3 and n = 4 and the number of
allocations is 6 (=4 + 3 – 1) hence optimality test can be performed.

b.​ These (m + n – 1) allocations should be in independent positions. A simple

rule for allocations to be in independent positions is that it is impossible to
travel from any allocation back to itself by a series of horizontal and vertical
jumps from one occupied cell to another, without direct reversal of routes, i.e.
it is not possible that the cells form a closed loop.

Optimality test procedure involves examination of each vacant cell to find

whether or not making an allocation in it reduces the total transportation cost.
The common methods for this purpose are the stepping-stone method and the
modified distribution (MODI) method. Here only MODI method to determine
the optimal solution is discussed.

This method is also called u-v method. The MODI method is an efficient
method of testing the optimality of a transportation solution. It avoids
extensive scanning and reduces the number of steps required in the
evaluation of the empty cells. The method gives a straightforward
computational scheme whereby the opportunity cost of each of the empty
cells can be determined. The steps of the method are as follows:

Step 1:​ Consider the initial solution. Introduce the dual variables
corresponding to the supply and demand constraints. If there are m
sources and n destinations, there will be m + n dual variables. Let ui
(i=1, …,m) and vj (j =1, …,n) be the dual variables corresponding to
supply and demand constraints.

Step 2:​ Calculate the ui and vj values with the help of occupied cells only
using the formula

463
MBM 106​ ​ Unit IV, Lesson 11

​ ​ ​ cij = ui + vj​ ​ ​ ​ ​ ​ (11.6)

​ For this, assign a zero value arbitrarily to a row variable ui or column

variable vj. Generally, this value is assigned zero to the rows or
columns having maximum cells allocated. Now calculate the ui and vj
using equation (11.6) for occupied cells. Repeat the exercise for any
other occupied cells in the row (column) in consideration. Enter the
ui along the right of the matrix and vj across the bottom of the matrix.

Step 3:​ Having determined all ui and vj values, calculate net evaluations Δij
for each unoccupied cell using equation (11.7)

∆ij = ui + vj – cij​ ​ ​ ​ ​ (11.7)

Step 4:​ If all Δij ≤ 0, the solution is optimal and unique. Cell(s) having a zero
value for Δij implies that the given solution is optimal but is not
unique. There exists other solution(s) that would be as good as this
solution.

Step 5:​ Determine the entering basic variable by selecting the cell with the
largest Δij value, i.e.

​ ​ Entering basic variable = ​ for ​ (11.8)

Step 6:​ Determine the leaving basic variable by tracing a closed loop path in
the table. When we intend to assign some units to an unoccupied
cell, it calls for making adjustments in some other occupied cells
because such an assignment affects the row and column totals of
supply and demand respectively that need to be maintained for
feasible solution. In effect the assignment of one or more units in a
(unoccupied) cell produces a chain effect and a closed loop (path)
attempt to capture it.

​ Identify a closed loop which starts and ends at the entering

non-basic cell (r, s) connecting only the basic cells. When a closed
loop is to be drawn, start with the entering non-basic cell (r, s). Now
move clockwise (or anticlockwise), draw an arrow from this cell to an
occupied cell in the same row or column, as the case may be. After
that, move vertically (or horizontally) to another occupied cell and
draw an arrow. Again move horizontally (or vertically) to another
occupied cell and continue the procedure until reaching the cell (r, s).
In the process of moving from one occupied cell to another, (a) move
only horizontally or vertically, but never diagonally, and (b) step over
empty and, if needed, over occupied cells without changing them.
Thus, a closed loop would always have right-angled turn with
corners only on the occupied cells.

​ Having traced the path, place plus and minus signs alternatively in
the cells on each turn of the loop, beginning with a plus (+) sign in
the empty cell. An important restriction is that there must be exactly

464
MBM 106​ ​ Unit IV, Lesson 11

one cell with a plus sign and one cell with a minus sign in any row or
column in which the loop takes a turn. This restriction ensures that
the row and column requirements would not be violated when units
are shifted among cells.

Step 7:​ To improve the solution, choose

​ ​ θ = {minimum of allocations made in the cells having a (-) sign}

Add this value of θ to all cells having a (+) sign and subtract the
same from the cells having a (-) sign.

Go to Step 2.

Illustration 11.3 (Continued)

Step 1:​ Consider the initial solution (Table 11.6).

Step 2:​ As the 1st row has maximum cells allocated (3 cells are allocated),
take u1 = 0. Now using equation (11.6) we have:
From cell c11, u1 + v1 = 12 ⇒ 0 + v1 = 12 ⇒ v1 = 12
Similarly
From cell c12, u1 + v2 = 10 ⇒ 0 + v1 = 10 ⇒ v1 = 10
From cell c13, u1 + v3 = 12 ⇒ 0 + v3 = 12 ⇒ v3 = 12
From cell c23, u2 + v3 = 8 ⇒ u2 = – 4
From cell c24, u2 + v4 = 14 ⇒ v4 = 18
From cell c34, u3 + v4 = 12 ⇒ u3 = – 11
Table 11.16: Calculating dual variables

To P Q R S si ui
From

A 12 10 12 13 500 0
180 150 170

B 7 11 8 14 300 –4
180 120

C 6 16 11 7 200 – 11
200

dj 180 150 350 320

vj 12 10 12 18

Step 3:​ Calculating Δij for the unoccupied cells using equation (11.7):
∆14 = u1 + v4 – c14 = 0 + 18 – 13 = 5
∆21 = u2 + v1 – c21 = – 4 + 12 – 7 = 1
∆22 = u2 + v2 – c22 = – 4 + 10 – 11 = – 5
∆31 = u3 + v1 – c31 = – 11 + 12 – 6 = – 5
465
MBM 106​ ​ Unit IV, Lesson 11

∆32 = u3 + v2 – c32 = – 11 + 10 – 16 = – 17
∆33 = u3 + v3 – c33 = – 11 + 12 – 11 = – 10
These ∆ij’s are entered in bottom left corner of the cell
Table 11.17: Computation of ∆ij

P Q R S si ui
To
From

A 12 10 12 13 500 0
180 150 170
5

B 7 11 8 14 300 –4
180 120
1 -5

C 6 16 11 7 200 – 11
200
-5 -17 -10

dj 180 150 350 320

vj 12 10 12 18

Step 4:​ The solution is not optimal as there are two positive Δij go to next
step.

Step 5:​ From equation (11.8)

Entering basic variable = max{5,1}​ ​ for

​ ​ ​ ​ ​ =​ 5

​ This implies that cell (1,4) will enter.

Step 6:​ A closed loop starting from cell (1, 4) is formed to determine the
leaving variable
Table 11.18: Forming a closed loop

466
MBM 106​ ​ Unit IV, Lesson 11

Start from the cell (1, 4), moving clockwise to occupied cell in the
same column, i.e. to cell (2, 4) then to cell (2, 3), then cell (1, 3) and
then back to cell (1, 4).

Assign alternate (+) and (-) starting by assigning a (+) sign in the cell
(1, 4), i.e., the entering non-basic cell.

Step 7:​ Solution will be improved by choosing

​ ​ θ = min {170, 120} = 120

Adding and subtracting the value of θ we get the new solution as, go
to step 11.

467
MBM 106​ ​ Unit IV, Lesson 11

Table 11.19: Improved Solution

Now again as the solution is not optimal go to Step 11 and get the
following solution
Table 11.20: Optimal TP Tableau

To P Q R S si ui

From

A 12 10 12 13 500 0
150 230 120
-1

B 7 11 8 14 300 –4
180 120
-5 -5

C 6 16 11 7 200 –6
200
-1 -12 -5

dj 180 150 350 320 Z = 9440

vj 11 10 12 18

This solution is optimal as all Δij ≤ 0 with

x12 =150, x13 =230, x14 =120, x21 =180, x22 =120, x34 =200 and

Z = ` 9,440

Further, since all ∆ij, values are negative, the solution is unique. Also
notice the solution given in the above Table is the same as the initial
solution obtained using VAM. Hence, the solution obtained by VAM
is also optimal and needs no revision.

Illustration 11.7: Consider the following transportation problem

Table 11.21

468
MBM 106​ ​ Unit IV, Lesson 11

Availability

6 1 9 3 70

11 5 2 8 55

10 12 4 7 90

Requirements 85 35 50 45

Is the solution below to it an optimal solution? If not, modify it to obtain optimal

solution.
Table 11.22

Availability

6 1 9 3 70
50 20

11 5 2 8 55
55

10 12 4 7 90
30 35 25

Requirements 85 35 50 45

Solution:

Performing the optimality test using MODI Method

Table 11.23: Initial TP Tableau

Table 11.24: Improved TP Tableau

469
MBM 106​ ​ Unit IV, Lesson 11

Table 11.25: Improved TP Tableau

Table 11.26: Improved TP Tableau

Table 11.27: Optimal solution

ui

6 1 9 3 0
30 40
0 -11

11 5 2 8 4
5 50
-8 -1

10 12 4 7 4
85 5
-7 -2

470
MBM 106​ ​ Unit IV, Lesson 11

vj 6 1 -2 3

As all the Δij ≤ 0 the solution is optimal with Z = 1160

Self-Check Questions

Fill in the blank

5.​ Methods to find the initial solution to a TP are ___________.

6.​ VAM is better than NWC rule as it considers the _________ also.
7.​ MODI Method is called __________ method.
8.​ Differences between the two costs in VAM are called ____________.

11.7​ Special Cases in Transportation Problem

Following are some variants of transportation problem:

11.7.1 Unbalanced Transportation Problem

The methods we have discussed for solving a transportation problem require

that the aggregate supply (∑si) is equal to the aggregate demand (∑dj).
However, situations may arise when the two are unequal. The two such
possibilities are, first, when the aggregate supply exceeds the aggregate
demand (i.e. ∑si > ∑dj) and second, when aggregate supply falls short of the
aggregate demand (i.e. ∑si < ∑dj). Such problems are called unbalanced
transportation problems. Balancing must be done before they can be solved.

When the aggregate supply exceeds the aggregate demand, the excess
supply is assumed to go to inventory and costs nothing for shipping. A column
of slack variables is added to the transportation tableau which represents a
dummy destination with a requirement equal to the amount of excess supply
and the transportation costs equal to zero.

On the other hand, when the aggregate demand exceeds the aggregate
supply in a transportation problem, balance is restored by adding a dummy
origin. The row representing it is added with an assumed total availability
equal to the difference between the total demand and supply, and with each of
the cells having a zero unit cost. In some cases, however, when the penalty of
not satisfying the demand at a particular destination(s) is given, then such
penalty value should be considered instead of zero values.

Once the transportation problem is balanced, its solution proceeds in exactly

the same manner as discussed earlier.

Illustration 11.8: A product is produced for four factories A, B, C and D. The

471
MBM 106​ ​ Unit IV, Lesson 11

unit production costs in them are ` 2, ` 3, ` 1, and ` 5 respectively. Their

production capacities are: Factory A – 50 units, Factory B – 70 units, Factory
C – 30 units and Factory D – 50 units. These factories supply the product to
four stores, demand of which is 25, 35, 105 and 20 units respectively. Unit
transport cost in rupees from each factory to each store is given in table
below:
Stores
1 2 3 4
A 2 4 6 11
Factories
B 10 8 7 5
C 13 3 9 12
D 4 6 8 3

Table 11.28

Determine the extent of deliveries from each of the factories to each of the
stores so that the total production and transportation cost is minimum.

Solution: First compile a new table of unit costs which consists of both
production and transportation costs. Thus the new table is as given below:
Table 11.29

Stores Productio
n
Capacity
1 2 3 4
Factories
A 2+2 4+2 6+2 11 + 2 50
B 10 + 3 8+3 7+3 5+3 70
C 13 + 1 3+1 9+1 12 + 1 30
D 4+5 6+5 8+5 3+5 50
Demand 25 35 105 20

Table 11.30

Stores Productio
n
Capacity
1 2 3 4
Factories
A 4 6 8 13 50
B 13 11 10 8 70
C 14 4 10 13 30
D 9 11 13 8 50
Demand 25 35 105 20

Here​ ​ Total production capacity ​ = 200 units

Total Demand ​ ​ = 185 units

Surplus capacity ​ ​ = 15 units

The production capacity and demand are not balanced in this case and we
have a surplus of 15 units of the product. Therefore, create a dummy
472
MBM 106​ ​ Unit IV, Lesson 11

destination (5th store). The associated cost coefficients are taken zero.
Table 11.31: Starting TP Table
Stores Productio
n
Capacity
1 2 3 4 5
Factories
A 4 6 8 13 0 50
B 13 11 10 8 0 70
C 14 4 10 13 0 30
D 9 11 13 8 0 50
Demand 25 35 105 20 15

Table 11.32: Initial solution using VAM

Finding initial solution using VAM

1 2 3 4 5 si Row Penalties
A 4 6 8 13 0 50/25/ 4 2 2 2 5 - -
25 5 20 20/0 ←
B 13 11 10 8 0 70/0 8 2 2 2 2 2 2
70
C 14 4 10 13 0 30/0 4 6 - - - - -
30 ←
D 9 11 13 8 0 50/35/ 8 1 1 3 5 5 -
15 20 15 15/0 ← ←
dj 25/0 35/5/0 105/85 20/0 15/0
/
15/0
C 5 2 2 0 0
o 5 2 2 0 -
l 5↑ 5 2 0 -
u - 5↑ 2 0 -
m - - 2 0 -
n - - 3 0 -
P
- - -
e
n
a
l
t
i
e
s

Here number of allocations = 4 + 5 – 1 = 8. These allocations are at

independent position. Therefore optimality test can be performed.
Table 11.33: Optimal solution

1 2 3 4 5 ui
A 4 6 8 13 0 -5
25 5 20
-10 -5
B 13 11 10 8 0 -3
70
-7 -3 -3 -3
C 14 4 10 13 0 -7

473
MBM 106​ ​ Unit IV, Lesson 11

30
-12 -4 -12 -7
D 9 11 13 8 0 0
15 20 15
0 0
vj 4 6 8 13 0
25 5 20

Since all the Δij ≤ 0, the solution is optimal. As the Δij = 0, there exists
alternate optimal solutions.

The practical significance of demand being 15 units less than production

capacity is that the company may cut down the production of 15 units at the
factory where it is uneconomical, i.e. at factory D.

The optimum cost (transportation cost + production cost)

Z = 4 × 25 + 6×5 + 8×20 + 10×70 + 4×30 + 13×15 + 8×20 +0×15 = ` 1,465

11.7.2 Different Production Costs

Sometimes a particular product may be manufactured at different locations

and then transported from different production plants. The production cost
could be different in different units due to various reasons, e.g. new and more
automation will bring down the production cost. In such problems the
manufacturing cost is added to the transportation cost while finding the
optimal solution (Refer to Illustration 11.8). If the transportation problem, along
with the variable production costs, fixed costs at various production plants is
also given, the latter are neglected and no consideration is given to them
while solving.

11.7.3 Prohibited Routes

Sometimes in a given transportation problem some route(s) may not be

available. This could be due to a variety of reasons like the unfavorable
weather conditions on a particular route, strike on a particular route, etc. In
such situations, there is a restriction on the routes available for transportation.
To handle a situation of this type, assign a very large cost represented by M to
each of such routes which are not available. Then the problem is solved in the
usual way. The effect of adding a large cost element would be that such
routes would automatically be eliminated in the final solution.

Illustration 11.9: A company has factories at four different places, which

supply warehouses A, B, C, D and E. Monthly factory capacities are 200,175,
150 and 325 respectively. Monthly warehouse requirements are 110, 90, 120,
230 and 160 units respectively. Unit shipping costs are given in the table
below. The costs are in rupees. Shipment from 1 to B and from 4 to D is not
possible. Determine the optimum distribution to minimize shipping costs.
Table 11.34

474
MBM 106​ ​ Unit IV, Lesson 11

A B C D E
I 13 - 31 8 20
II 14 9 17 6 10
III 25 11 12 17 15
IV 10 21 13 - 17

Solution: To avoid allocations in cell (1,B) and (4, D) a very heavy penalty +M
is allocated to these cells.

Also here total capacity = 200 + 175 + 150 + 325 = 850 units

And total demand = 110 +90 +120 + 230 +160 = 710 units

As demand is less than the capacity, a dummy warehouse is created to

absorb this additional capacity of 140 units. Thus the starting table is:
Table 11.35: Starting Solution

A B C D E F si
I 13 M 31 8 20 0 200
II 14 9 17 6 10 0 175
III 25 11 12 17 15 0 150
IV 10 21 13 M 17 0 325
di 110 90 120 230 160 140

Using VAM the initial solution is

Table 11.36: Initial solution using VAM

A B C D E F si
I 13 M 31 8 20 0 200
200
II 14 9 17 6 10 0 175
30 145
III 25 11 12 17 15 0 150
10 140
IV 10 21 13 M 17 0 325
110 80 120 15
di 110 90 120 230 160 140

MODI is used to find the optimal solution. After a few iterations, we get the
optimal solution shown below:

Table 11.37: Optimal solution

A B C D E F si
I 13 M 31 8 20 0 200
200
II 14 9 17 6 10 0 175
30 145
III 25 11 12 17 15 0 150
90 45 15

475
MBM 106​ ​ Unit IV, Lesson 11

IV 10 21 13 M 17 0 325
110 75 140
di 110 90 120 230 160 140

11.7.4 Multiple Optimal Solutions

The optimal solution to a given problem may, or may not, be unique. Recall
that the solution to a transportation problem is optimal if all ∆ij ≤ 0. For a given
solution, if all the ∆ij values are negative, then it is unique. If, however, a cell(s)
has/have ∆ij = 0, then multiple optimal solutions are indicated so that there
exist transportation pattern(s) other than the one obtained which can satisfy
all the rim requirements for the same cost.

To obtain an alternate optimal solution, trace a closed loop beginning with a

cell having ∆ij = 0, and get the revised solution in the same way as a solution
is improved. It may be observed that this revised solution would also entail the
same total cost as before. Note that for every ‘zero’ value of ∆ij in the optimal
solution tableau, a revised solution can be obtained.

Illustration 11.10 Given the following transportation problem:

Table 11.38 Transportation Problem: Unbalanced Problem with Prohibited Routes

Market
Warehouse Supply
A B C

1 10 12 07 180

2 14 11 06 100

3 09 05 13 160

4 11 07 09 120

Demand 240 200 220

It is known that currently nothing can be sent from warehouse 1 to market A

and from warehouse 3 to market C. Solve the problem and determine the
least cost transportation schedule. Is the optimal solution obtained by you
unique? If not, what is/are the other optimal solution(s)?

Solution: The given transportation problem is unbalanced since aggregate

demand = 240 + 200 + 220 = 660 units while the aggregate supply = 180 +
100 + 160 + 120 = 560 units. Thus, a dummy warehouse (5) is introduced with
supply of 660 - 560 = 100 units, and zero transportation cost since no
penalties are provided for not satisfying the demand at each market. Further,
since routes 1-A and 3-C are given as prohibited, the cost element for each of
these is replaced by M. The information is presented in the Table below. VAM
is used to determine the initial solution.

476
MBM 106​ ​ Unit IV, Lesson 11

Table 11.39: Initial Feasible Solution by VAM

P Q R si Row Penalty
To
From

1 M 12 7 180/0 5 5 - - -
180

2 14 11 6 100/60/ 5 5 5 3 3
60 40 0 ← ←

3 9 5 M 160/0 4 4 4 4 -
160 ←

4 11 7 9 120/80/ 2 2 2 4 4
80 40 0

5 0 0 0 100/0 0 - - - -

100

Deman 240/14 200/40/ 220/40/

d 0 0 0

Colu 9↑ 5 6
mn
Pena 2 2 1
lty

2 2 3

2 2 -

3 4↑ -

Solving for optimality using MODI Method

Table 11.40: Optimal Solution

To P Q R si ui

From

1 M 12 7 180 15
180
15-M -1

2 14 11 6 100 14
60 40
-1

477
MBM 106​ ​ Unit IV, Lesson 11

3 9 5 M 160 9
160
0 1-M

4 11 7 9 120 11
80 40
-6

5 0 0 0 100 0
100
-4 -8

dj 240 200 220

vj 0 -4 -8

In the table all Δij ≤ 0, hence the solution is optimal with

x13 =180, x21 = 60, x23 = 40, x32 = 160, x41 = 80, x42 = 40, x51 = 100

and Z = ` 4,300

The solution given in Table 11.40 is not unique because in cell 3-A, ∆ij = 0. To
obtain an alternative optimal solution, draw a closed loop as shown in Table
11.41.
Table 11.41: Improved Solution

Thus the new solution is

Table 11.42: Optimal Solution

To P Q R si ui

From

478
MBM 106​ ​ Unit IV, Lesson 11

1 M 12 7 180 15
180
15-M -1

2 14 11 6 100 14
60 40
-1

3 9 5 M 160 9
80 80
1-M

4 11 7 9 120 11
120
-6

5 0 0 0 100 0
100
-4 -8

dj 240 200 220

vj 0 -4 -8

This solution is also optimal with Z = ` 4,300

11.7.5 Degeneracy
We have already seen that a basic feasible solution of a transportation
problem has m+n-1 basic variables, which means that the number of occupied
cells in such a solution is one less than the number of rows plus the number
of columns. It may happen sometimes that the number of occupied cells is
smaller than m+n-1. Such a solution is called a degenerate solution.

Degeneracy in a transportation problem can figure in two ways. Degeneracy

can arise in the first instance when an initial feasible solution is obtained (see
Illustration 11.10). Secondly, the problem may become degenerate when two
or more cells are vacated simultaneously in the process of transferring units
along the closed path. The difficulty, when a solution is degenerate, is that it
cannot be tested for optimality as all ui and vj values cannot be determined.

To overcome degeneracy, proceed by assigning an infinitely small amount,

close to zero, to one (or more, if the need be) empty cell and treat the cell as
an occupied cell. This amount is represented by a Greek letter ε (epsilon) and
is taken to be such an insignificant value that would not affect the total cost.
Thus, it is big enough to cause the particular route to which it is assigned to
be considered as a basic variable but not large enough to cause a change in
the total cost and the other non-zero amounts.

It is important to remember that a ε cannot be placed in any randomly

selected unoccupied cell. Instead, the cell chosen for inserting ε must be an

479
MBM 106​ ​ Unit IV, Lesson 11

independent cell, originating from which a closed loop cannot be traced.

Placing an ε in a cell that is not independent will fail to yield unique ui and vj
values and, consequently, would not resolve the problem of degeneracy. It
may further be noted that if a given problem requires two (or more) ε, then a
cell in which an epsilon has already been placed will be treated as occupied
while determining independence of cells for inserting an ε subsequently.

While a number of independent cells may be available in a given situation,

choose the one with the smallest unit cost value. It is not necessary to
evaluate each and every empty cell for independence to insert epsilon.
Instead, a cell with minimum cost may be examined whether it is independent
or not. If it is found to be independent, then place ε in it and if not then move
to another cell.

After placing ε in the required number of cells, we proceed to solve the

problem in the usual manner.

Illustration 11.11 Determine optimal solution to the problem given below.

Table 11.43

Market
Plant Supply
A B C D

1 6 4 9 1 40

2 20 6 11 3 40

3 7 11 0 14 50

4 7 1 12 6 90

Demand 90 30 50 30

Since the aggregate supply is 220 units and the aggregate demand is 200
units, we introduce a dummy market, M5, for an amount equal to 20(the
difference between the aggregate supply and demand), with all cost elements
equal to zero. The initial solution is given in Table 11.44.

It is obtained using VAM and its degenerate because it contains only seven
basic variables, since there are only 7 occupied cells and not 8 (= 4 + 5 - 1),
required for a non-degenerate solution in this case. Here empty cells P1M3,
P2M3, P3M1, P3M2, P3M4, P3M5, and P4M3 are independent, while others are
not. For removing degeneracy, let us place ε in the cell P3M5 as it has lowest
cost and then test it for optimality (Table 11.44).

The solution is found to be non-optimal. The improved solution is given in

Table 11.45.
Table 11.44: Initial Solution

480
MBM 106​ ​ Unit IV, Lesson 11

To A B C D E si
From

1 6 4 9 1 0 40
30 10

2 20 6 11 3 0 40
20 20

3 7 1 0 14 0 50
50 ε
4 7 1 12 6 0 90
60 30

dj 90 30 50 30 20 700

Table 11.45: Initial Feasible Solution: Non-optimal

Notice that in improving the solution, only ε moves, as this is the minimum
quantity in the cells bearing - sign on the closed loop. The optimality test,
MODI, suggests that the solution given in Table 11.46 is an optimal one.

481
MBM 106​ ​ Unit IV, Lesson 11

Table 11.46: Revised Solution: Optimal

To A B C D E si ui

From

1 6 4 9 1 0 40 0
30 10
-4 -10 -2

2 20 6 11 3 0 40 2
20 20
-12 -4 -10

3 7 1 0 14 0 50 1
ε 50
0 -12 -1

4 7 1 12 6 0 90 1
60 30
-12 -4 -1

dj 90 30 50 30 20 Z= 700

vj 6 0 -1 1 -2

The solution can be stated as:

x11 = 30, x14 = 10, x24 = 20, x33 = 50, x41 = 60, x42 = 30, all other xij = 0;

and Total cost = ` 700.

When the problem becomes degenerate at the solution-revision stage, ε is

placed in one (or more, if necessary) of the recently vacated cells, with the
minimum cost, and then proceed with the problem in the usual manner.

Illustration 11.12: Solve the following transportation problem.

Table 11.47: Transportation Problem

To
From Supply
A B C D

1 7 3 8 6 60

2 4 2 5 10 100

3 2 6 5 1 40

482
MBM 106​ ​ Unit IV, Lesson 11

Demand 20 50 50 80 200

Solution: The initial is as given below:

Table 11.48: Initial Basic Feasible Solution: NW Corner Rule

To A B C D si
From

1 7 3 8 6 60
20 40

2 4 2 5 10 100
10 50 40

3 2 6 5 1 40
40

dj 20 50 50 80 Z= ` 970

Table 11.49: Initial Basic Feasible Solution: NW Corner Rule

This revised solution is degenerate because there are only 5 occupied cell
and not 6. The two recently vacated cells are A2 and B1. Per unit cost of
these being ` 3 and ` 10, respectively, we may assign to the cell A2 to cope
with degeneracy. Thus, solution is also found to be non-optimal. Hence, we
move to the next iteration and further revise the solution as shown in Table
11.52.
Table 11.50: Revised Solution: Non-optimal

483
MBM 106​ ​ Unit IV, Lesson 11

Table 11.51: Revised Solution: Optimal

To A B C D si ui

From

1 7 3 8 6 60 0
20 40
-2 -2

2 4 2 5 10 100 -1
20 30 50
-5

3 2 6 5 1 40 -5
40
-2 -8 -4

dj 20 50 50 80 Z= ` 730

vj 5 3 6 6

The optimality test indicates that this solution is optimal. The optimal
transportation plan, then, is:

x12 = 20, x14 = 40, x21 = 20, x22 = 30, x23 = 50, x34 = 40, all other xij, = 0, for a
total cost of ` 730.

11.7.6 Maximization Problem

As previously mentioned, the classical transportation problem is one of the
minimization types. However, a transportation tableau may contain unit profits
instead of unit costs and the objective function be maximization of total profits.
To solve a maximization type of problem, the transportation method is not
applied directly. The given problem is first converted into an equivalent
minimization problem by subtracting the maximum profit from each element of
the given (profit) matrix. The procedure yields a loss matrix to which the

484
MBM 106​ ​ Unit IV, Lesson 11

transportation method is applied. The entries in the opportunity loss matrix

indicate how much each of the values is away from that constant (the highest
profit value). The minimization of the opportunity loss (‘distance’ from the
maximum value) automatically leads to maximization of total profit. However,
it should be remembered that while the transportation method is applied to the
opportunity loss matrix, the total profit value is obtained by using the
transportation pattern of the solution on the unit profit matrix, and not through
the opportunity loss matrix.

It may be noted, however, that if a maximization type of transportation

problem is unbalanced, then it should be balanced by introducing the
necessary dummy row/column for a source/destination, before converting it
into a minimization problem. Similarly, if such a problem has a prohibited
route, then the payoff element for such a route should be substituted by —M
before proceeding to convert to minimization type.

Illustration 11.13: Solve the following TP.

Table 11.52: Transportation Problem

To
From Supply
A B C D

1 12 18 6 25 200

2 8 7 10 18 500

3 14 3 11 20 300

Demand 180 320 100 400 1000

Solution: To solve this problem, first convert the unit profit matrix into
opportunity loss matrix by subtracting each of the value in it from the largest
value equal to 25. The information is presented in Table 11.53.

Table 11.53: Loss Matrix with initial solution

To A B C D si
From

1 13 7 19 0 200
200

2 17 18 15 17 500

485
MBM 106​ ​ Unit IV, Lesson 11

100 400

3 11 22 14 5 300
180 20 100

dj 180 320 100 400 Z=

`15,180

Table 11.54: Opportunity Loss Matrix: Initial Solution

Also given in the table is the initial feasible solution obtained using VAM. The
solution is tested for optimality using MODI method and is found to be non –
optimal. A closed loop is traced beginning with the cell which has the largest
positive opportunity cost. The revised solution is obtained (Table 11.55). This
solution is tested for optimality and is also found to be non – optimal. It is
further revised by making x23 a basic variable and is tested to be optimal
(Table 11.55). Accordingly, the maximum profit obtainable is ` 15,400.
Observe that the total profit at each stage is found with reference to the profit
matrix given in the problem.
Table 11.55: Revised Solution: Optimal

To A B C D si ui

From

1 13 7 19 0 200 0
200
-11 -15 -4

2 17 18 15 17 500 11
120 100 280
-4

3 11 22 14 5 300 9
180 120
-6 -1

486
MBM 106​ ​ Unit IV, Lesson 11

dj 180 320 100 400 Z= ` 15,400

vj 2 7 4 -4

Illustration 11.14: A company has three plants at locations A, B and C. which

produce the same product. It has to supply this to buyers located at D, E and
F. The weekly plant capacities for A, B and C are 100, 800 and 150 units
respectively, while the buyer requirements are 750, 200 and 500 units
respectively for D, E and F. The unit shipping costs (in `) are given here:

​ Buyer
Plant​ D​ E​ F
A​ 8​ 4 10
B​ 9​ 7​ 9
C​ 6​ 5​ 8
Table 11.56

Assume that the penalty for failing to supply buyer requirement is ` 4, ` 3 and `
3 per unit in respect of D, E and F respectively. Determine the optimal
distribution for the company so as to minimize the cost of transportation and
penalty payable.

Solution: This problem is an unbalanced problem thus adding a dummy plant

T with supply value of 400 units and cost elements equal to the penalty values
4, 3, and 3 in column headed D, E and F respectively.

This information is shown in the Table 11.57. The table also shown the initial
solution determined using VAM.
Table 11.57: Initial Solution

Buyer D E F si ui

Plant

A 8 4 10 100 0
100
-2 -4

B 9 7 9 800 11
600 100 100

C 6 5 8 150 15
150
-1 -2

T 4 3 3 400
400
-1 -2

dj 750 200 500 Z= ` 8,300

487
MBM 106​ ​ Unit IV, Lesson 11

Penalty
vj 6 4 6 cost
= ` 1,200

Upon testing, the solution is found to be optimal. It involves a cost equal to `

8300 and a penalty cost of ` 1200.

Miscellaneous Illustrations

Illustration 11.15: Consider the following transportation problem

Destination
Source​ P​ Q​ R​ S​ Supply
A​ 6​ 11​ 9​ 8​ 200
B​ 13​ 10​ 8​ 15​ 300
C​ 9​ 9​ 14​ 12​ 500
D​ 12​ 10​ 12​ 10​ 100
Demand​ 350​ 250​ 300​ 200
a.​ Find the optimal solution to this problem and determine the total cost of
transportation.
b.​ Is the solution unique? If no identify the alternate optima
c.​ If the transporter agrees to reduce the transportation charges on the route
B-S by 40%, how will it affect the optimal solution?
d.​ It is decided to send a total of 150 units on the route A-P. How will the
optimal solution and total cost be affected?
e.​ Trace the effect of sending 100 units on the route C-S.

Solution:

a.​ The initial solution obtained using VAM, is seen to be degenerate.

Accordingly a ε is placed in the cell A-R as it is an independent cell.
Table 11.58: Initial Solution

P Q R S si
A 6 11 9 8 200
200 ε
B 13 10 8 15 300
300
C 9 9 14 12 500
150 250
D 12 10 12 10 100
100
Demand 350 250 300 200

Solving for optimality

Table 11.59: Improved Solution

488
MBM 106​ ​ Unit IV, Lesson 11

Table 11.60: Optimal Solution

P Q R S ui
A 6 11 9 ε 8 0
100 100
-5
B 13 10 8 15 -1
300
-8 -5 -8
C 9 9 14 12 3
250 250
-2 -1
D 12 10 12 10 1
100
-4 -2 -1
vj 6 6 9 9 Z=
9300

b.​ The optimal solution obtained in the table is a unique since Δij is not equal
to 0.
c.​ A 40% reduction on the route BS implies that the unit cost on their route
will be revised to ` 9. In the optimal solution above, the Δij = -8 with a cost
value of ` 15 for this route would be revised to -2 with the unit cost of ` 9.
However, since Δij continues to be negative, no cost saving is possible by
supplying goods on this route. Hence, the optimal solution will not be
affected by this change.
d.​ The optimal solution calls for transferring 100 units on the Route AP. Now,
if it is desired to send 150 units on this route, we shall reduce the supply at
source A and the demand at destination P by 150 units and resolve the
problem. For this solution, we make the route AP prohibited since no more
units can be send on it.
Table 11.61: Optimal Solution
P Q R S ui
A M 11 9 8 0
ε 50
-6
B 13 10 8 15 -1
300
-9 -6 -8
C 9 9 14 12 4
200 250 50

489
MBM 106​ ​ Unit IV, Lesson 11

-1
D 12 10 12 10 2
100
-5 -3 -1
vj 5 5 9 8 Z=
9350

The solution is given in above table. This solution is optimal with a total
transportation cost of ` 9350.
e.​ To transfer 100 units on the route CS, draw a closed loop beginning with
this cell in Table 11.60 to see if it can indeed be done. The closed path is
CS AS AP CP CS. With given quantities, it is evident that it is
possible to do so.

The quantities of these cells would be revised to AS – 0, AP- 200, CP –

150, CS -100.

Further, since Δij = -1 for the cell CS, this transfer of 100 units will cause
the total cost to increase by 100 × 1 = `100.

Illustrations 11.16: A company has a manufacturing plant which has the

following regular production capacity during a quarter. Additional capacity can
be obtained by the use of overtime but it increases the production cost. The
forecast demand for the following quarter is also given below: the company
starts with an initial inventory from the past quarter and requires a final
inventory at the end of the quarter. Items which are produced but not
immediately used can be stored with a storage charges. Data is given in the
table below:
Production Capacity​ ​ Forecast Demand
Month​ ​ Regular​​ Overtime​ Month​ ​ Demand
Jan​ ​ 200​ ​ 44​ ​ Jan​ ​ 200
Feb ​ ​ 190​ ​ 38​ ​ Feb ​ ​ 150
March​ ​ 210​ ​ 42​ ​ March​ ​ 250
April​ ​ 200​ ​ 40​ ​ April​ ​ 250

Regular production cost ​ = ` 10 per unit

Overtime production cost ​ = ` 14 per unit
Initial Inventory​ ​ = 100 Units
Final Inventory ​ ​ =125 Units
Storage Cost on Items ​ = 10% per unit per month
production on regular basis
Storage cost on items ​ = 15% per unit per month
produced on overtime basis
Determine the production schedule in order to minimize the total cost of
production and storage.

Solution: To setup the transpiration table for the problem note the following.
1.​ If full overtime capacity is used, the availability will be 100 + 800 + 164 =
1064 units, whereas forecast demand plus final inventory is only 850+123
= 975 units. Hence a dummy destination (month) will be assumed to take

490
MBM 106​ ​ Unit IV, Lesson 11

up the balance 89 units. The cost in the associated dummy cells, will be
assumed to be zero.
2.​ Forecast demand for January is 200 units. Since initial inventory is 100
units, the effective demand for their month is 100 units only.
3.​ Items that are produced in January and are also used in January have unit
production cost of ` 10 and ` 14 for regular (R) and overtime production
(OT) respectively. Same is true for the months of February, March and
April.
4.​ Items that are produced in January but are used in February, will have to
be stored for a month, increasing the production–cum storage cost to
` 11 (= `10 + ` (10 ×10)/100) and ` 16 (= `14 + ` (14 ×15)/100)
respectively for regular and overtime production. If used in March, the
costs will increase to ` 12 and ` 18 and if used in April, they will become `
13 and ` 20 respectively.
5.​ Final inventory at the end of the quarter (i.e. April) is 125 units. The cost
coefficients in the cells will be same as in April.
6.​ Items produced in February cannot be supplied in January. Shortages are
not permitted as shortage cost or opportunity cost is not given. Hence M is
assigned to such cells. Same is true for other months.

This entire information can be put in tabular form as follows:

Table 11.62: Starting Table
Jan Feb Mar Apr Final Dm Production
Inv Capacity
Jan R 10 11 12 13 13 0 200
OT 14 16 18 20 20 0 44
Feb R M 10 11 12 12 0 190
OT M 14 16 18 18 0 38
Mar R M M 10 11 11 0 210
OT M M 14 16 16 0 42
Apr R M M M 10 10 0 200
OT M M M 14 14 0 40
Demand 200 150 250 250 125 89

Finding the initial solution using VAM we get.

Table 11.63: Initial Solution
Jan Feb Mar Apr Final Dm Production
Inv Capacity
Jan R 10 11 12 13 13 0 200
100 50 50
OT 14 16 18 20 20 0 44
44
Feb R M 10 11 12 12 0 190
112 40 38
OT M 14 16 18 18 0 38
38
Mar R M M 10 11 11 0 210
210
OT M M 14 16 16 0 42

491
MBM 106​ ​ Unit IV, Lesson 11

42
Apr R M M M 10 10 0 200
200
OT M M M 14 14 0 40
37 3
Demand 200 150 250 250 125 89

Solve this problem using usual transportation problem approach.

Illustration 11.17: A construction company is interested in taking loans from

banks for some of its projects P, Q R and T. The rates of interest and the
lending capacity differ from bank to bank. All these projects are to be
completed. The relevant details are provided in the following table. Assuming
the role of a consultant, advise this company as to how it should take the
loans so that the total interest payable will be the least. Are there alternate
optimal solutions? If so, indicate one such solution.
Interest rate in percentage for project max credit
(in thousands of `)
Bank​ ​ ​ ​ P​ Q​ R​ S​ T
Private Bank (Pvt)​ ​ 20​ 18​ 18​ 17​ 17​ any amount
National Bank (N)​ ​ 16​ 16​ 16​ 15​ 16​ 400
Cooperative Bank (C) ​ ​ 15​ 15​ 15​ 13​ 14​ 200
Amount Required​ ​ 200​ 150​ 200​ 125​ 75
(in thousands of `)

Solution: The total amount required for the five projects is ` 750 thousands.
Since Pvt Bank can lend any amount, allocate ` [750-(400+600)=100]
thousands to this bank. The balanced problem is shown below. The initial BFS
is found by VAM.
Table 11.63: Initial Solution

Projects Max Credit

Banks P Q R S T (in thousands
of `)
Pvt. 20 18 18 17 17 100
100
N 16 16 16 15 16 400
200 50 150
C 15 15 15 13 14 250
50 125 75
Amt reqd 200 150 200 125 75
(in thousands
of `)

Table 11.64: Optimal Solution

Projects ui
Banks P Q R S T
Pvt. 20 18 18 17 17 18
100
-2 0 -1 0
N 16 16 16 15 16 16
200 50 150
-1 -1
C 15 15 15 13 14 15
50 125 75

492
MBM 106​ ​ Unit IV, Lesson 11

0 0
dj 0 0 0 -2 -1 Z=` 1,16,250

The solution is optimal as all Δij ≤ 0.

Further, since some of the cells have Δ ij = 0, alternate optimal solutions exist.
To get one such solution, include cell (Pvt,T) as the basic cell and reallocate
as shown below.
Table 11.65: Initial optimal solution with re-allocation

Self-Check Questions

Fill in the blanks

9.​ A problem is unbalanced if _________________.

10.​A profit matrix can be converted to a loss matrix by ________________
11.​Degeneracy occurs when _____________________

State True/False

12.​TP can be degenerate at any stage.

13.​For a maximization transportation problem –M is added to the prohibited
routes.
14.​The transportation costs are assumed to be linear in nature.
15.​Closed loops can be square shaped only.

11.8​ Conclusion
In this Lesson a special type of linear programming problem, the
transportation problem is studied. The Simplex Algorithm can be used to solve
any LPP for which the solution exists, but as the number of variables and
constraints increase the computation by this method becomes more and more

493
MBM 106​ ​ Unit IV, Lesson 11

laborious. Therefore to simplify the calculations there is a special branch of LP

that comprises problems of the transportation type.

The Transportation Method is developed to deal with the transportation of

goods from different sources to different destinations, given relevant data like
available quantities at various sources, demand at each destinations, the cost
of shipping along each route and non-availability of certain routes (if any).
This method can be used for a variety of situation like scheduling, production,
investment, plant location, inventory control, etc.

11.9​ Answers to Self-Check Questions

1.​ Feasible
2.​ Basic Feasible solution
3.​ occupied or filled, basic
4.​ linear
5.​ NWC, LCM and VAM
6.​ Costs
7.​ u-v
8.​ Row and column penalties
9.​ Total of supply and demand is not equal
10.​Subtracting largest element from each element of the profit matrix
11.​Number of allocation <(m+n-1)
12.​True
13.​False
14.​True
15.​False

11.10 Terminal Questions

1.​ Describe Transportation Problem (TP) and give its mathematical formulation
2.​ How do you deal with the following type of TP a) Degenerate, b) Unbalanced
and c) Maximizing problem? Why does degeneracy in TP occur?
3.​ Three refineries with maximum daily capacities of 6 million, 5 million and 8
million litres of gasoline supply three destination areas with daily demands of
4 million, 8 million and 7 million. The gasoline is transported to the three
destinations centres through a network of pipe-lines. The transportation cost
is estimated based on the length of the pipeline at about 10 paise per 100
litres per km. The mileage table summarized below shows that refinery 1 is
not connected to distribution area 3. Find the optimum transportation cost of
distribution.

​ ​ ​ Destination Areas ​ 1​ 2​ 3
​ ​ ​ ​ ​ 1​ 120​ 180​ 0
​ ​ Refinery ​ 2​ 300​ 100​ 80
3​ 200​ 250​ 120

494
MBM 106​ ​ Unit IV, Lesson 11

4.​ A firm manufacturing a single product has three plants I, II and III. The three
plants have produced 60, 35, 40 units of the product respectively during this
month. The firm had commitment to sell 22 units to customer A, 45 units
to customer B, 20 units to customer C, 18 units to customer D and 30 units
to customer E. Find the minimum possible transportation cost of shifting
the manufactured product to the 5 customers. The net per unit cost of
transportation from three plants to 5 customers is given below:

​ ​ ​ Customer ​ A​ B​ C​ D​ E
​ ​ ​ ​ I​ 4​ 1​ 3​ 4​ 4
​ ​ ​ Plant​ II​ 2​ 3​ 2​ 2​ 3
​ ​ ​ ​ III​ 3​ 5​ 2​ 4​ 4
5.​ A company has factories A, B, C, which supply warehouses at W, X, Y, Z.
Monthly factory capacities are 160, 150, and 150 units respectively. Monthly
warehouse requirements are 80, 90, 110 and 160 units respectively. The unit
shipping costs (in Rs) are as follows:

To​ W X Y Z
​ ​ ​ ​ ​ ​ A​ 42 48 38 37
​ ​ ​ ​ ​ From​ B​ 40 49 52 51
​ ​ ​ ​ ​ ​ C​ 39 38 40 43

​ Determine the optimum distributions for this company to minimize shipping

costs.
6.​ Given the following data
​ ​ ​ ​ ​ Destination
​ ​ ​ ​ ​ 1​ 2​ 3​ Capacity
​ ​ ​ ​ I​ 2​ 2​ 3​ 10
​ ​ Source​​ II​ 4​ 1​ 2​ 15
​ ​ ​ ​ III​ 1​ 3​ ×​ 40
​ ​ ​ Demand​ 20​ 15​ 30

How many units should be transported from sources to destinations so that

the total cost of transporting all the units to their destinations is a minimum?
7.​ Consider the following TP
To

​ ​ ​ ​ E​ F​ G​ H​ J
​ ​ ​ A​ 8​ 10​ 12​ 17​ 15
​ From​ ​ B​ 15​ 13​ 18​ 11​ 9
​ ​ ​ C​ 14​ 20​ 6​ 10​ 13
​ ​ ​ D​ 13​ 19​ 7​ 6​ 12

The present allocation is as follows:

A to E : 90​ ​ A to F: 10​ ​ B to F: 150​ ​ C-F: 10
C to G: 50​ ​ C to J: 120​ ​ D to H: 210​ ​ D-J: 70

a.​ Check if this is optimal solution, if not, then find the optimal schedule?
b.​ If in the above problem, the transportation cost from A to G is reduced to
10, what will be new optimal schedule?

495
MBM 106​ ​ Unit IV, Lesson 11

8.​ A particular product is manufactured in factories A, B, C and D and is sold at

center 1, 2 and 3. The cost (in rupees) of product per unit and capacity (in kg)
per unit time of each plant is given below:
Factory​​ Cost(`)/unit​ ​ Capacity (kg) /unit
A​ ​ ​ 12​ ​ ​ 100
B​ ​ ​ 15​ ​ ​ 20
C​ ​ ​ 11​ ​ ​ 60
D​ ​ ​ 13​ ​ ​ 20

The sale price in rupees/unit and demand in kg/unit time are as follows:
Sales Centre​ ​ Sales price (`)/unit​ ​ Demand(kg) /unit
1​ ​ ​ ​ 15​ ​ ​ ​ 120
2​ ​ ​ ​ 14​ ​ ​ ​ 140
3​ ​ ​ ​ 16​ ​ ​ ​ 60

Find the optimum sales distribution.

11.11 Further Readings

1.​ F.S. Hillier and G.J Lieberman, Introduction to Operations Research, 7th Ed.,
Tata McGraw Hill Publishing Company Limited, New Delhi.
2.​ K. Swarup, [Link] and M. Mohan, Operations Research, 13th Ed., Sultan
Chand and Sons Education Publishers, New Delhi
3.​ N.D. Vohra, Quantitative Techniques in Management, 4th Ed., Tata McGraw
Hill Education Private Limited, New Delhi.

11.12 Glossary of the terms used

●​ Feasible Solution:​ A feasible solution to a transportation problem is a set of

non-negative allocations xij that satisfy row and column restrictions.
●​ Basic Feasible Solution:​ A feasible solution to a transportation problem is
said to be a Basic Feasible Solution (BFS) if it contains no more than (m + n
-1) non-negative allocations, where m is the number of rows and n is the
number of columns of the transportation problem.
●​ Optimal Solution:​ A feasible solution (not necessarily basic) that minimizes
(maximizes) the transportation cost (profit) is called an optimal solution.
●​ Non-degenerate Solution:​ A basic feasible solution to a m×n transportation
problem is said to be non-degenerate if
a.​ total number of non-negative allocations is exactly (m + n – 1) and
b.​ these (m + n – 1) allocations are in independent position.
●​ Degenerate BFS:​ A BFS in which the total number of non-negative
allocations is less than (m+n–1) is called degenerate BFS.
●​ Loop: ​In a transportation table, an ordered set of four or more cells is said to
form a Loop if

496
MBM 106​ ​ Unit IV, Lesson 11

a.​ Any two adjacent cells in the ordered set lie either in the same row or in
the same column, and
b.​ Any three or more adjacent cells in the ordered set do not lie in the same
row or the same column.

497
MBM 106​ ​ Unit IV, Lesson 12

LESSON 12​ THE ASSIGNMENT PROBLEM

12.0​OBJECTIVES​ 491
12.1​INTRODUCTION​ 491
12.2​THE ASSIGNMENT PROBLEM​ 492
12.2.1 Definition of the Assignment Problem​ 492
12.2.2 Mathematical Representation of the Assignment Problem​ 492
12.3​COMPARISON WITH THE TRANSPORTATION PROBLEM​ 493
12.4​SOLUTION OF THE ASSIGNMENT PROBLEM​ 494
12.5​SPECIAL CASES IN ASSIGNMENT PROBLEM​ 503
12.5.1 Unbalanced Assignment Problem (Non-square matrix)​ 503
12.5.2 Maximization problem​ 504
12.5.3 Restrictions on assignments​ 505
12.5.4 Alternate optimal solutions​ 505
MISCLLENEOUS ILLUSTRATIONS​ 505
12.6​CONCLUSION​ 505
12.7​ANSWERS TO SELF-CHECK QUESTIONS​ 505
12.8​TERMINAL QUESTIONS​ 505
12.9​FURTHER READINGS​ 505
12.10 GLOSSARY OF THE TERMS USED​ 505
MBM 106​ ​ Unit IV, Lesson 12

12. The Assignment Problem

In previous lessons we discussed the simplex and the transportation techniques for
solving linear programming problems. However, there are some special cases of
linear programming problems solutions of which can be obtained by special
techniques. They are easier to apply and greatly reduce the computational work
required by the simplex and the transportation techniques. This lesson deals with
one such special case - the assignment problem which finds many applications in
allocation and scheduling. Assignment is generally made on one-to-one basis and if
there are more jobs to do than can be done, one can decide which job to leave
undone or what resource to add.

12.0​ Objectives

After going through this lesson you will be able to understand:

●​ Rules to solve a Game

●​ Concept of Dominance

12.1​ Introduction
There are many situations where the assignment of people, machine etc.
may be called for, like, assigning salesmen to different regions, vehicles and
drivers to different routes, products to factories, jobs to machines, contract to
bidders and research problems to teams etc. Assignment is a problem
because people possess varying abilities for performing different jobs, and,
therefore the cost of performing those jobs by different people is different. In
Assignment Problem, the question is how the assignments should be made
in order that the cost (profit) involved is minimized (maximized). The Method
allows managers to seek answers to questions like:

●​ How to assign the given jobs to some workers on a one-to-one basics when
completion time of performance is given for each combination and it is
desired that the jobs are completed in least time or at the least cost?

●​ How to deal with situations when the number of jobs do not match with the
number of job performers, when some job(s) cannot be performed by, or is
not given to, a particular performer, or when a certain job has to be given to a
particular individual?

499
MBM 106​ ​ Unit IV, Lesson 12

12.2​ The Assignment Problem

The Assignment problem is a special type of LPP, where assignees are being
assigned to different tasks.

The assumptions for formulating an Assignment problem are

1.​ The number of assignees (resources) and the number of the tasks are the
same, say n
2.​ Each assignee is to be assigned to exactly one task
3.​ Each task is to be performed by exactly one assignee
4.​ Cost cij is associated with assignee i (i = 1, 2, …,n) in performing task j (j=1,
2, …,n)
5.​ The objective is to determine how all n assignments should be made to
minimize the total cost.

Any problem satisfying all these assumptions can be solved using the
algorithm designed for assignment problem.

12.2.1 Definition of the Assignment Problem

Given n resources and n jobs and given the effectiveness of each resource
for each job, the problem is to assign each resource to one and only one job
so as to optimize the given measure of effectiveness.

Table 12.1 represents the assignment of n resources (persons, machines,

etc.) to n jobs. cij is the cost of assigning ith resource to jth job and xij
represents the assignment of ith resource to jth job. If ith resource can be
assigned to jth job, xij = 1, otherwise zero. The objective is to make
assignments that minimize the total assignment cost or maximize the total
associated gain.
Table 12.1: Representation of AP

Jobs

Resource 1 2 … n
s

1 c11 c12 … c1n

2 c21 c22 … c2n

: : : : :

N Cn1 Cn2 … cnn

Thus an assignment problem can be represented by a matrix of order n.

500
MBM 106​ ​ Unit IV, Lesson 12

12.2.2 Mathematical Representation of the Assignment Problem

Let xij, denote the assignment of resource i to job j such that

xij = ​ 1, if the ith resource is assigned to jth job

​ 0, if the ith resource is not assigned to jth job

Then, the model is given by

𝑛 𝑛
∑ ∑ 𝑐𝑖𝑗𝑥𝑖𝑗
minimize Z = 𝑖=1 𝑗=1 (12.1)
𝑛
∑ 𝑥𝑖𝑗 = 1
s.t. 𝑗=1 i =1, 2,…, n
(12.2)
𝑛
∑ 𝑥𝑖𝑗 = 1
𝑖=1 j =1, 2,…, n
(12.3)

and xij = 0 or 1 (12.4)

​ ​ ​ ​ ​ ​ ​ ​

Constraint equation (12.2) means one job is assigned to the ith resource and
constraint equation (12.3) means one resource is assigned to the jth job.

The technique used for solving assignment problem makes use of two
theorems:

Theorem 12.1: (Reduction Theorem):​ In an assignment problem, if a

constant is added or subtracted to every element of a row (or column) in the
cost matrix [cij], then an assignment that minimizes the total cost on one
matrix also minimizes the total cost on the other matrix.

Theorem 12.2:​ If all cij ≥ 0 and we can find a set xij = x*ij such that
𝑛 𝑛
*
∑ ∑ 𝑐𝑖𝑗𝑥𝑖𝑗 = 0
𝑖=1 𝑗=1

then this solution is optimal.

12.3​ Comparison with the Transportation Problem

An assignment problem may be regarded as a special case of the
transportation problem. Referring to Table 12.2, resources represent the
‘sources’ and jobs represent the ‘destinations’. Number of sources is equal to

501
MBM 106​ ​ Unit IV, Lesson 12

the number of destinations, supply at each source is unity (si = 1 for all i) and
demand at each destination is also unity (dj = 1 for all j). The cost of
‘transporting’ (assigning) resource i to job j is cij and the number of units
allocated to a cell can be either one or zero, i.e. they are non-negative
quantities.
Table 12.2: Comparison with TP

Jobs

Resource 1 2 … N si
s

1 c11 c12 … c1n 1

2 c21 c22 … c2n 1

: : : : : :

n Cn1 Cn2 … cnn 1

dj 1 1 … 1

However the transportation algorithm is not very useful to solve this model
because of degeneracy. In this model, when an assignment is made, the row
as well as column requirements are satisfied simultaneously (row and
column conditions being always unity), resulting in degeneracy. Thus the
assignment problem is a completely degenerate form of the transportation
problem. In n × n problem, there will be n assignments instead of n + n - 1 or
2n - 1 which makes the computations quite cumbersome. However, the
special structure of the assignment model allows a more convenient and
simple method of solution.

Self-Check Questions

Fill in the blanks

1.​ AP is a _________________ form of TP.

2.​ For a assignment problem, number of rows __________ number of column.
3.​ Assignments are made on __________ basis.

12.4​ Solution of the Assignment Problem

An AP can be solved using complete enumeration method, transportation
method, simplex method and Hungarian method. In this lesson the
Hungarian method to solve AP is discussed.

502
MBM 106​ ​ Unit IV, Lesson 12

The Hungarian method (or the Reduced Matrix Method or the Flood’s
technique), suggested by Mr. Koning of Hungary, is used for solving
assignment problems since it is quite efficient and results in substantial time
saving over the other techniques. It is based on the concept of opportunity
cost. Opportunity costs show the relative penalties associated with resources
to a task as opposed to making assignment of the best (least) cost
assignment. It involves a rapid reduction of the original matrix and finding of
a set of n independent zeros, one in each row and column, which results in
an optimal solution. The method consists of the following steps:

Step 1:​ Determine the cost table from the given problem.

Step2:​ If number of assignees = number of tasks, go to step 3. If number of

assignees ≠ number of tasks, add a dummy column or a dummy
row, as the case may be, so that the cost table becomes a square
matrix with the cost of dummy row/column as zero.

Step 3:​ Subtract the smallest element of each row from all the elements of
the row.

Step 4:​ Modify the resulting matrix by subtracting the minimum element of
each column from all the elements of the respective column. Thus
obtain the first modified matrix.

Step 5:​ In the modified matrix, search for an optimal solution by making
zero assignments as follows:

a)​ Examine the first row

b)​ If there is only one zero in it, then enclose this zero in a box (□)
and cross (×) all other zeros in the column passing through the
enclosed zero.

c)​ Examine all rows and repeat step (b).

d)​ If any row has more than one zero, then do not touch that row and
pass onto the next row.

e)​ Repeat steps (a) to (d) for the column, starting from the first
column of the matrix.

Step 6:​ If number of assignments (□) = n (order of the cost matrix)

​ ​ An optimal solution is reached.

​ Else go to step 7.

Step 7:​ Draw minimum number of horizontal and vertical lines crossing all
zeros. This consists of the following sub steps:

a)​ Mark (√) the rows that do not have any enclosed zeros.

503
MBM 106​ ​ Unit IV, Lesson 12

b)​ Mark (√) the columns (not already marked) that have crossed
zeros in marked rows.

c)​ Mark (√) the rows (not already marked) that have enclosed zeros in
marked columns.

d)​ Repeat sub-steps (a) and (c) till no more rows or columns can be
marked.

e)​ Draw straight lines through all unmarked rows and marked
columns. This gives the minimum number of lines crossing all
zeros. If this number is equal to the order of the matrix, then it is an
optimal solution, otherwise go to step 7.

Step 8:​ Iterate towards the optimal solution. Examine the uncovered
elements. Select the smallest element and subtract it from all the
uncovered elements. Add this smallest element to every element
that lies at the intersection of two lines. Leave the remaining
elements of the matrix as such. This yields second basic feasible
solution.

Step 9:​ Repeat steps 5 through 9 successively until the number of lines
crossing all zeros becomes equal to the order of the matrix. In such
a case every row and column will have one assignment. This
indicates that an optimal solution has been obtained. The total cost
associated with this solution is obtained by adding the original costs
in the assigned cells.

Illustration 12.1: A machine tool company decides to make four

subassemblies through four contractors. Each contractor is to receive only
one subassembly. The cost of each subassembly is determined by the bids
submitted by each contractor and is shown in the table below in hundreds of
rupees.
Subordinates​ ​ Contractors

1 2 3 4

1 15 13 14 17

2 11 12 15 13

3 13 12 10 11

4 15 17 14 16

i.​ Show that the assignment model is a special case of the transportation
model.

i.​ Assign the different subassemblies to contractors so as to minimize the

504
MBM 106​ ​ Unit IV, Lesson 12

total cost.

Solution:

i.​ Comparing this model with the transportation model, we find that si = 1
and dj = 1. Thus, the assignment model can be represented as in Table
12.3.
Table 12.3: Comparison with TP

Contractors
Subassemblie
s 1 2 3 4 si

1 15 13 14 17 1

2 11 12 15 13 1

3 13 12 10 11 1

4 15 17 14 16 1

dj 1 1 1 1

Therefore, the assignment model is a special case of the transportation

model in which
a)​ All right-hand side constants in the constraints are unity i.e., si = 1, di =
1;
b)​ All coefficients of xij in the constraints are unity; and
c)​ m = n.

ii.​ Applying the Hungarian Algorithm for solving the assignment problem:

Step 1:​ The given cost table is as shown below:

Table 12.4: The Cost Matrix
Subordinates​ ​ Contractors

1 2 3 4

1 15 13 14 17

2 11 12 15 13

3 13 12 10 11

4 15 17 14 16

Step 2:​ As number of assignees = number of tasks, no need to add dummy

row/column

505
MBM 106​ ​ Unit IV, Lesson 12

Step 3:​ Subtract the smallest element of each row from all the elements of
the row.

In the given situation, the minimum element in first row is 13. So,
subtract 13 from all the elements of the first row. Similarly subtract
11, 10 and 14 from all the elements of row 2, 3 and 4 respectively
(Table 12.5).
Table 12.5: Table after row reduction
Contractors

Subassemblie 1 2 3 4
s

1 2 0 1 4

2 0 1 4 2

3 3 2 0 1

4 1 3 0 2

Step 4:​ Now perform column reduction. In Table 12.5, column 1, 2 and 3
have zero as the smallest value. Column 4 has 1 as the smallest
value, thus subtract this minimum element from all its elements.
Table 12.6 represents the first modified matrix.
Table 12.6: First modified matrix

1 2 3 4

1 2 0 1 3

2 0 1 4 1

3 3 2 0 0

4 1 3 0 1

Step 5: ​ Making zero assignments.

In the given problem, row I has a single zero in column 2. Make an

assignment there by enclosing this zero by a square (□). It means
subassembly 1 is assigned to contractor 2. Since contractor 2 has
been assigned subassembly 1 and as a contractor can be assigned
only one subassembly, any other zero in column 2 is crossed. Since
there is no other zero in this column, crossing is not required.

Next, row 2 has a single unmarked zero in column 1, make an

assignment.

Skip row 3 as it has more than one zero.

506
MBM 106​ ​ Unit IV, Lesson 12

Row 4 has a single unmarked zero in column 3, make an

assignment and cross the 2nd zero in column 3. Now, row 3 has a
single unmarked zero in column 4, make an assignment here. This
is shown in the Table below.
Table 12.7: Making zero Assignment (optimal solution)

Step 6:​ Check if optimal assignment can be made in the current solution.

Since, the number of assignments (4) = order of matrix (4), i.e.,

there is one assignment in each row and in each column, the
optimal assignment can be made in the current solution.

The optimal assignment policy is and the minimum cost is:

​ Subassembly​ ​ Contractor​ ​ Cost
​ 1​ 2​ 1300 (costs taken from
​ 2​ 1​ 1100 the original matrix)
​ 3​ 4​ 1100
​ 4​ 3​ 1400
​ ​ ​ Total​ 4900

Illustration 12.2: Four different jobs can be done on four different machines.
The set-up and take-down time costs are assumed to be prohibitively high for
changeovers. The matrix below gives the costs in rupees of producing jobs i
on machine j.
Table 12.8
Machines

Job M1 M2 M3 M4
s

J1 5 7 11 6

J2 8 5 9 6

J3 4 7 10 7

J4 10 4 8 3

How should the jobs be assigned to the various machines so that the total
cost is minimized?

Solution:

Determining the first modified matrix

507
MBM 106​ ​ Unit IV, Lesson 12

Table 12.9: Table after Row Reduction

M1 M2 M3 M4

J1 0 2 6 1

J2 3 0 4 1

J3 0 3 6 3

J4 7 1 5 0

Table 12.10: Table after Column Reduction

M1 M2 M3 M4

J1 0 2 2 1

J2 3 0 0 1

J3 0 3 2 3

J4 7 1 1 0

Make zero assignments by first moving row-wise (skipping 2nd row as it has
more than one zero) and then moving column wise.
Table 12.11: Making zero assignments

As number of assignments (3) ≠ order of the matrix (4), the solution is not
optimal. Thus improve the solution by drawing the minimum number of
horizontal and vertical lines to cover all zeros. Referring to Table 12.12:

a)​ Marking (√) the 3rd row as it does not have any enclosed zeros

b)​ 1st column has crossed zeros in marked rows. Thus mark (√) column 1.

c)​ The 1st row has enclosed zeros in marked columns. Thus row I is marked (√).

d)​ As all the rows or columns have been marked repetition is not needed.

508
MBM 106​ ​ Unit IV, Lesson 12

e)​ Straight lines are drawn through all unmarked rows and marked columns.
Table 12.12: Drawing minimum lines

The number of such lines is 3 which is less than the order of matrix (n = 4).
Hence optimal assignment is not possible in the current solution. Thus further
reduction of Table 12.12 is necessary.

The smallest uncovered (by the line) cost element is 1. Subtract this from all
the elements that do not have a line through them including itself. Add this
smallest element to every element that lies at the intersection of two lines.
Leave the remaining elements of the matrix unchanged. Proceeding in this
manner we get the following matrix:
Table 12.13: Second feasible solution

Still the solution is not optimal as number of assignments (3) ≠ order of the
matrix (4). Thus improving the solution we get the following table.

Now make the zero assignment. Since there is no row with exactly one
unmarked zero, we start considering the columns. We find there is no column
as well with single unmarked zero. Therefore, we make assignment arbitrarily
in any zero and cross the remaining zero in the same column as well as the
same row. Let us make assignment at zero in cell (J1, M1) arbitrarily and
cross the remaining zeros in column 1 and row 1. See now if there is exactly
one zero (unmarked) in any row or in any column. We find row 3 has a single
zero in column 3. We make assignment here in cell (J3, M3) and cross
remaining zeros in column 3. Next assignment is made in cell (J2, M2) in row
2 and the remaining zeros in column 2 are crossed. Last assignment is made
in cell (J4, M4). This is the case of Alternate (Multiple) optimal solution
(discussed in Section 12.5.4).
Table 12.14: Optimal solution

509
MBM 106​ ​ Unit IV, Lesson 12

As there is assignment in each row and in each column, optimal assignment

can be made in the current solution. Hence optimal assignment policy is

Job J1 should be assigned to machine M1,

J2 should be assigned to machine M2,

J3 should be assigned to machine M3,

J4 should be assigned to machine M4,

and optimum cost = ` (5 + 5 + 10 + 3) = ` 23.

Illustration 12.3: Five wagons are available at stations 1, 2, 3, 4 and 5.

These are required at five stations I, II, III, IV and V. The mileages between
various stations are given by the table below. How should the wagons be
transported so as to minimize the total mileage covered?
Table 12.15

I II III IV V

1 10 5 9 18 11

2 13 9 6 12 14

3 3 2 4 4 5

4 18 9 12 17 15

5 11 6 14 19 10

Solution:

Reduce the Matrix

Table 12.16: Table after row reduction

I II III IV V

1 5 0 4 13 6

510
MBM 106​ ​ Unit IV, Lesson 12

2 7 3 0 6 8

3 1 0 2 2 3

4 9 0 3 8 6

5 5 0 8 13 4

Make zero assignments and check for optimal assignments. Since the
number of allocations is only 3, hence optimal assignment is not possible in
the current solution.
Table 12.17: Improving the solution after column reduction

511
MBM 106​ ​ Unit IV, Lesson 12

Table 12.18 shows the improved solution.

Table 12.18: Second Feasible Solution

I II III IV V

1 3 0 3 10 2

2 6 4 0 4 5

3 0 1 2 0 0

4 7 0 2 5 2

5 3 0 7 10 0

Check for optimal assignments (Table 12.19). Here the minimum number of
lines crossing all zeros is 4. Hence optimal assignment is not possible in the
current solution.

Table 12.19: Drawing lines

Improving this solution, we get

Table 12.20: Third Feasible Solution

I II III IV V

1 1 0 1 8 0

2 6 6 0 4 5

3 0 3 2 0 0

4 5 0 0 3 0

5 3 2 7 10 0

512
MBM 106​ ​ Unit IV, Lesson 12

Checking optimality, the minimum number of lines crossing all zeros is 4.

Hence optimal assignment is not possible in the current solution. Thus
improving this solution we get:
Table 12.21: Drawing lines

The improved solution is:

Table 12.22: Forth Feasible Solution

I II III IV V

1 0 0 1 7 0

2 5 6 0 3 5

3 0 4 3 0 1

4 4 0 0 2 0

5 2 2 7 9 0

Proceeding as above, we find that there is an assignment in every row and

column. Hence Table 12.23 gives optimal assignment policy. The minimum
mileage associated with this solution is -

Zmin = 10 + 6 + 4 + 9 + 10 = 39

Table 12.23: optimal solution

Self-Check Questions

513
MBM 106​ ​ Unit IV, Lesson 12

Fill in the blanks

4.​ Hugarian method is also called ___________________ technique.

5.​ The solution is optimal if _____________ equals the order of the matrix.

12.5​ Special Cases in Assignment Problem

This section discusses the variations of the AP

12.5.1 Unbalanced Assignment Problem (Non-square matrix)

An AP is unbalanced when the number of resources is not equal to the

number of jobs. Since the Hungarian method of solution requires a square
matrix, fictitious resources or jobs may be added and zero costs be assigned
to the corresponding cells of the matrix. These cells are then treated the
same way as the real cost cells during the solution procedure.

Illustration 12.4: A company has 4 machines on which 3 jobs are to be

done. Each job can be assigned to one and only one machine. The cost of
each job on each machine is as given below. Determine the optimum
assignment.
​ ​ ​ ​ ​ ​ Machine
​ ​ ​ ​ Job​ A​ B​ C​ D
​ ​ ​ ​ 1​ 18​ 24​ 28​ 32
​ ​ ​ ​ 2​ 8​ 13​ 17​ 19
​ ​ ​ ​ 3​ 10​ 15​ 19​ 22

Solution: As m ≠ n, add a dummy row with cost = 0.

Table 12.24: Adding a dummy row

Jobs Machines

A B C D

1 18 24 28 32

2 8 13 17 19

3 10 15 19 22

4 0 0 0 0

Performing Row and Column reductions and improving the solution

Table 12.25: First Improved Solution

514
MBM 106​ ​ Unit IV, Lesson 12

Table 12.26: Second Improved Solution

Table 12.27: Optimal Solution

Optimal Assignment
​ ​ Cost
A-1​ ​ 18
B-2​ ​ 13
C-3​ ​ 19
D-4​ ​ 0
​ ​ Total 50

12.5.2 Maximization problem

Sometimes the assignment problem may deal with maximization of the

objective function. The maximization problem has to be changed to
minimization before the Hungarian method may be applied. This
transformation may be done in either of the following two ways:

i.​ By subtracting all the elements from the largest element of the matrix, or

ii.​ By multiplying the matrix elements by - 1.

Illustration 12.5: A company has a team of four salesmen and there are four
districts where the company wants to start its business. After taking into

515
MBM 106​ ​ Unit IV, Lesson 12

account the capabilities of salesmen and the nature of districts, the company
estimates that the profit per day in rupees for each salesman in each district
is as below.
Table 12.28: Profit Matrix

District

1 2 3 4

Salesman A 16 10 14 11

B 14 11 15 15

C 15 15 13 12

D 13 12 14 15

Find the assignment of salesmen to various districts which will yield

maximum profit.

Solution: As the given problem is of maximization type, it has to be changed

to minimization type before solving it by the Hungarian method. This is
achieved by subtracting all the elements of the matrix from the largest
element in it. Thus subtracting all the elements of profit the table from the
largest element 16, the equivalent ‘opportunity loss’ matrix is given as:
Table 12.29: Opportunity Cost matrix

1 2 3 4

A 0 6 2 5

B 2 5 1 1

C 1 1 3 4

D 3 4 2 1

Performing row and column reductions

Table 12.30: First modified matrix

1 2 3 4

A 0 5 1 4

B 1 4 0 0

C 0 0 2 3

516
MBM 106​ ​ Unit IV, Lesson 12

D 2 3 1 0

Performing the optimality test

Table 12.31: Optimal Solution

As there is one assignment in each row and in each column, optimal

assignment can be made in the current feasible solution. Assignment policy
shall be

Salesman A should be assigned district 1,

Salesman B should be assigned district 3,
Salesman C should be assigned district 2,
Salesman D should be assigned district 4.
Maximum profit per day = ` (16 + 15 + 15 + 15) = ` 61

12.5.3 Restrictions on assignments

Sometimes technical, space, legal or other restrictions do not permit the

assignment of a particular resource to a particular job. Such problems can be
solved by assigning a very heavy cost (infinite cost) to the corresponding cell.
Such a job will then be automatically excluded from further consideration
(making assignments).

Illustration 12.6: Four new machines M1, M2, M3 and M4 are to be installed in
a machine shop. There are five vacant places A, B, C, D and E available.
Because of limited space, machine M2 cannot be placed at C and M3 cannot
be placed at A. cij the assignment cost of machine i to place j in rupees is
shown below. Find the optimal assignment schedule.
Table 12.32

1 2 3 4 5

A 4 6 10 5 6

B 7 4 - 5 4

C - 6 9 6 2

D 9 3 7 2 3

517
MBM 106​ ​ Unit IV, Lesson 12

Solution: As the given matrix is non-square, add a dummy machine and

associate zero cost the corresponding cells. As machine M2 cannot be
placed at C and M3 cannot be placed at A, assign infinite cost (M) in cells
(M2, C) and (M3, A), resulting in the following matrix:
Table 12.33: Initial Table

1 2 3 4 5

A 4 6 10 5 6

B 7 4 M 5 4

C M 6 9 6 2

D 9 3 7 2 3

E 0 0 0 0 0

Performing row and column operations

Table 12.34: Table after reductions

1 2 3 4 5

A 0 2 6 1 2

B 3 0 M 1 0

C M 4 7 4 0

D 7 1 5 0 1

E 0 0 0 0 0

Check if optimal assignment can be made in the current feasible solution or

not
Table 12.35: Optimal Solution

As there is no row and no column without assignment, optimal assignment

can be made in the initial feasible solution. The optimal assignment of
various machines is as follows:

518
MBM 106​ ​ Unit IV, Lesson 12

Machine M1 to place A,
Machine M2 to place B,
Machine M3 to place E,
Machine M4 to place D,
and place C will remain vacant
Total assignment cost = ` (4 + 4 + 2 + 2) = ` 12

12.5.4 Alternate optimal solutions

Sometimes, it is possible to have two or more ways to strike off all zero
elements in the reduced matrix for a given problem. In such cases, there will
be alternate optimal solutions with the same cost. Alternate optimal solutions
offer a great flexibility to the management since it can select the one which is
most suitable to its requirement.

Consider Illustration 12.2, Table 12.14. In this example as there are multiple
zeros in rows and columns which implies there are alternate solutions.

Determining alternate solutions is further explained in the following

Illustration.

Miscellaneous Illustrations
Illustration 12.7: A fast-food chain wants to build four stores. In the past the
chain has used six different construction companies, and having been
satisfied with each, has invited them to bid for each job. The final bids (in
thousands of rupees) are shown in the following table:
Table 12.36

1 2 3 4 5 6

1 85.3 90 87.5 82.4 89.1 91.3

2 78.9 84.5 99.4 80.4 89.3 88.4

3 82.0 31.3 28.5 66.5 80.4 109.7

4 84.3 34.6 86.2 83.3 812.0 812.5

Since the fast-food chain wants to have each of the new stores ready as
quickly as possible, it will allot at most one job to a construction company.
What assignment will result in the minimum total cost?

Solution: As the problem involves a non-square 4 × 6 matrix, two dummy

stores, say D1 and D2 are added with zero cost for each associated cell. Also
all the elements are multiplied by 10 and the cost in hundreds of rupees is
given by the matrix given below.
Table 12.37: Initial Table

519
MBM 106​ ​ Unit IV, Lesson 12

1 2 3 4 5 6

1 853 900 875 824 891 913

2 789 845 994 804 893 884

3 820 313 285 665 804 1097

4 843 346 862 833 850 855

D1 0 0 0 0 0 0

D2 0 0 0 0 0 0

Performing row and column reductions yields

Table 12.38: First modified table

1 2 3 4 5 6

1 29 76 51 0 67 89

2 0 56 205 15 104 95

3 535 28 0 380 519 812

4 497 0 516 487 504 509

5 0 0 0 0 0 0

6 0 0 0 0 0 0

Proceeding as above, we find that there is an assignment in every row and

column Hence Table 12.38 gives optimal assignment policy as:
Store ​ ​ ​ Construction Company​ ​ ​ Cost
1​ 4​ 82400
2​ 1​ 78900
3​ 3​ 28500
4​ 2​ 34600
​ ​ Total​ 224400

Consider the last two rows. Since each row and column had more than one
zero assign zeros arbitrarily such that there is one-to-one correspondence
between the store and the construction company.

Illustration 12.8: A company has four territories open and four salesmen
available for assignment. The territories are not equally rich in their sales
potential. It is estimated that a salesman operating in each territory would
bring in the following annual sales:
Territory​ ​ I​ ​ II​ ​ III​ ​ IV
Annual Sales (`)​​ 60,000​ ​ 50,000​ ​ 40,000​ ​ 30,000

520
MBM 106​ ​ Unit IV, Lesson 12

The four salesmen are also considered to differ in ability; it is estimated that
working under the same conditions, their yearly sales would be
proportionately as follows:
Salesman​ ​ ​ A​ ​ B​ ​ C​ ​ D
Proportion​ ​ ​ 7​ ​ 5​ ​ 5​ ​ 4

If the criterion is maximum expected total sales, the intuitive answer is to

assign the best salesman to the richest territory, the next best salesman to
the second richest territory, and so on. Verify this answer by Assignment
Model.

Solution: The first step is to prepare the square matrix. The sum of sales of
four salesmen = 7 + 5 + 5 + 4 = 21.

Taking the sales of ` 10,000 as one unit of sale, the annual sales of the four
salesmen in four territories are:
7 7 7 7
For A: 21
×6; 21
×5; 21
×4; 21
×3;
5 5 5 5
For B: 21
×6; 21
×5; 21
×4; 21
×3;
5 5 5 5
For C: 21
×6; 21
×5; 21
×4; 21
×3;
4 4 4 4
For D: 21
×6; 21
×5; 21
×4; 21
×3;

For convenience, the sales will be considered for 21 years rather than
considering the annual sales of each salesman in each territory. This will
avoid the fractional term. The problem is to determine the assignments which
will make the total sales maximum.

The effectiveness matrix will be given as:

Table 12.39: Profit Matrix

Salesman Territories

I II III IV

A 42 35 28 21

B 30 25 20 15

C 30 25 20 15

D 24 20 16 12

Converting maximization type to minimization type by subtracting the highest

element, i.e., 42, we get
Table 12.40: Loss matrix

I II III IV

521
MBM 106​ ​ Unit IV, Lesson 12

A 0 7 14 21

B 12 17 22 27

C 12 17 22 27

D 18 22 26 30

Performing the row and column reductions,

Table 12.41: Modified Solution

I II III IV

A 0 3 6 9

B 0 1 2 3

C 0 1 2 3

D 0 0 0 0

Finding initial solution

Table 12.42: First Feasible Solution

Table 12.43: Second Feasible Solution

Table 12.45: Optimal Solution

522
MBM 106​ ​ Unit IV, Lesson 12

Maximum annual sales =​ ` (1/21)(42 + 25 + 20 + 12) ×10,000

​ ​ ​ ​ =​ ` 47, 142, 86

Illustration 12.9: A small garment making unit has five tailors stitching five
different types of garments. All the five tailors are capable of stitching all the
five types of garments. The output per day tailor and the profit (`) for each
type of garment are given below:
Table 12.44

Tailor Garments

I II III IV V

A 7 9 4 8 6

B 4 9 5 7 8

C 8 5 2 9 8

D 6 5 8 10 10

E 7 8 10 9 9

Profit(`)/Garment 2 3 2 3 4

a.​ Which type of garment should be assigned to which tailor in order to

maximize the profit, assuming that there are no other constraints
b.​ If tailor D is absent for a specific period and no other substitute tailor is
available, what should be the optimum assignment?

Solution:

a.​ Construct the profit matrix first by multiplying the profit per garment with
the number of garments stitched as shown below:
Table 12.45: Starting Solution

Garments

I II III IV V

A 14 27 8 24 24

523
MBM 106​ ​ Unit IV, Lesson 12

Tailor
B 8 27 10 21 32

C 16 15 4 27 32

D 12 15 16 30 40

E 14 24 20 27 36

This profit matrix is converted to loss matrix by subtracting all the cell values
from the highest cell value (40)
Table 12.46: Loss Matrix

I II III IV V

A 26 13 32 16 16

B 32 13 30 19 8

C 24 25 36 13 8

D 28 25 24 10 0

E 26 16 20 13 4

Performing row and column operations

Table 12.47: Modified Solution

I II III IV V

A 0 0 3 0 3

B 11 5 6 8 0

C 3 17 12 2 0

D 15 25 6 7 0

E 9 12 0 6 0

Performing optimality test

Table 12.48: Improved Solution

524
MBM 106​ ​ Unit IV, Lesson 12

Table 12.49: Second Improved Solution

Table 12.50: Optimal Solution

The solution is optimal

Tailor​ Garment​ ​ Profit (`)
A​ ​ 1​ ​ ​ 14
B​ ​ 2​ ​ ​ 27
C​ ​ 4​ ​ ​ 27
D​ ​ 5​ ​ ​ 40
E​ ​ 3​ ​ ​ 20
​ ​ ​ ​ Total 128

b.​ If tailor is absent, the profit earned by him will be zero and the profit
matrix becomes
Table 12.51: As Tailor D is absent

Garments

I II III IV V

525
MBM 106​ ​ Unit IV, Lesson 12

A 14 27 8 24 24

Tailor B 8 27 10 21 32

C 16 15 4 27 32

D 0 0 0 0 0

E 14 24 20 27 36

Converting this profit matrix to loss matrix by subtracting all the cell
values from the highest cell value (36)
Table 12.52: Loss Matrix

I II III IV V

A 22 9 28 12 12

B 28 9 26 15 4

C 20 21 32 9 4

D 36 36 36 36 36

E 22 12 16 9 0

Performing row and column operations

Table 12.53: Modified Matrix

I II III IV V

A 13 0 19 3 3

B 24 5 22 11 0

C 16 17 28 5 0

D 0 0 0 0 0

E 22 12 16 9 0

Performing optimality test we find that this is not optimality. Through various
iterations we get the optimal solution as
Table 12.54: Optimal Solution

526
MBM 106​ ​ Unit IV, Lesson 12

I II III IV V

A 2 0 8 0 8

B 8 0 6 3 0

C 3 15 15 0 3

D 0 11 0 8 16

E 6 7 0 1 0

The optimal assignment policy is

Tailor​ Garment​ ​ Profit (`)
A​ ​ 2​ ​ ​ 27
B​ ​ 5​ ​ ​ 32
C​ ​ 4​ ​ ​ 27
E​ ​ 3​ ​ ​ 20
​ ​ ​ ​ Total 106

Garment 1 remains unstitched since tailor D is absent.

Self-Check Questions

Fill in the blanks

6.​ A problem is unbalanced if _________________.

7.​ A profit matrix can be converted to loss matrix __________________.

12.6​ Conclusion
In this Lesson you have studied solving the assignment problem. Assignment
is a problem because people possess varying abilities for performing different
jobs and therefore, the costs of performing those jobs by different people are
different. Obviously, if all persons could do a job in same time or at the same
cost then it would not matter who among them is assigned the job. Thus, in
an assignment problem, the question is how the assignments should be
made in order that the total cost involved is minimized or the profit is
maximized.

The solution to the Assignment Problem in this lesson is discussed using

Hungarian Method. The method involves three steps: 1. Obtaining zeros in
every row and column; 2. Covering all zeros with minimum number of
horizontal and vertical lines to check whether assignment can be made, 3.
Making assignments at zero. Modifications are needed if the minimum
number of lines covering all zeros is smaller than the order of the matrix.

527
MBM 106​ ​ Unit IV, Lesson 12

12.7​ Answers to Self-Check Questions

1.​ Degenerate
2.​ =
3.​ one-to-one
4.​ Flood’s
5.​ Number of enclosed zeros
6.​ Number of rows are not equal to number of columns
7.​ Subtracting the highest value element from all the elements of the matrix

12.8​ Terminal Questions

1.​ Describe is Assignment Problem. How is it special case of TP? Explain.

2.​ Explain the methods to solve the AP
​ ​ Solve the following assignment problems
​ 3.​ a b c d e ​​ ​ 4 a b c d e
​ A 3 8 2 10 3​ ​ ​ A 3 9 2 3 7
​ B 8 7 2 9 7​ ​ ​ B 6 1 5 6 6
​ C 6 4 2 7 5​ ​ ​ C 9 4 7 10 3
​ D 8 4 2 3 5​ ​ ​ D 2 5 4 2 1
​ E 9 10 6 9 10​ ​ ​ E 9 6 2 4 6

5.​ ​ I II III IV V​ ​ 6.​ I II III IV

​ ​ 1​ 10 11 4 4 8​ ​ ​ 1​ 2 3 4 5
​ 2​ 7 11 10 14 12​ ​ ​ 2​ 4 5 6 7
​ 3​ 5 6 9 12 14​ ​ ​ 3​ 7 8 9 8
​ 4​ 13 15 11 10 7​ ​ ​ 4​ 3 5 8 4

7.​ A company plans to assign 5 salesmen to 5 districts in which it operates.

Estimates of sales revenue in thousands of rupees for each salesman in
different districts are given in the following table. What should be the
placement of the salesman if the objective is to maximize the expected sales
revenue?
​ District​ 1​ 2​ 3​ 4 ​ 5
​ ​ A​ 40​ 46​ 48​ 36​ 48
​ ​ B​ 48​ 32​ 36​ 29​ 44
​ ​ Salesman ​ C​ 49​ 35​ 41​ 38​ 45
​ ​ D​ 30​ 46​ 49​ 44​ 44
​ ​ E​ 37​ 41​ 48​ 43​ 47

12.9​ Further Readings

1.​ F.S. Hillier and G.J Lieberman, Introduction to Operations Research, 7th Ed.,
Tata McGraw Hill Publishing Company Limited, New Delhi.

528
MBM 106​ ​ Unit IV, Lesson 12

2.​ K. Swarup, [Link] and M. Mohan, Operations Research, 13th Ed., Sultan
Chand and Sons Education Publishers, New Delhi
3.​ N.D. Vohra, Quantitative Techniques in Management, 4th Ed., Tata McGraw
Hill Education Private Limited, New Delhi.

12.10​ Glossary of the terms used

●​ Assignment problem: is a special type of LPP, where assignees are being
assigned to different tasks

●​ Unbalanced AP: An AP is unbalanced when the number of resources is not

equal to the number of jobs.

529