Chapter One: Introduction to Operations Research

Operations Research (OR) started just before World War II in Britain with the establishment
of teams of scientists to study the strategic and tactical problems involved in military
operations. The objective was to find the most effective utilization of limited military
resources by the use of quantitative techniques. Following the war, numerous peacetime
applications emerged, leading to the use of OR and management science in many industries
and occupations.
Operations Research (OR) is the study of mathematical models for complex organizational
systems. Optimization is a branch of OR which uses mathematical techniques such as linear
and nonlinear programming to derive values for system variables that will optimize
performance. Management science or operations research uses a logical approach to problem
solving. This quantitative approach is widely employed in business. Areas of application
include forecasting, capital budgeting, capacity planning scheduling inventory management,
project management and production planning.

Significance of Operation Research

 It provides a tool for scientific analysis and provides solution for various business
problems.
 It enables proper deployment and optimum allocation of scarce resources.
 It helps in minimizing waiting and servicing costs.
 It enables the management to decide when to buy and how much to buy through the
technique of inventory planning.
 It helps in evaluating situations involving uncertainty.
 It enables experimentation with models, thus eliminating the cost of making errors
while experimenting with reality.
 It allows quick and inexpensive examination of large numbers of alternatives.
 In general OR facilitates and improves the decision making process.
Areas of Application
 Inventory control
 Facility design (distribution decision)
 Product mix determination
 Portfolio analysis
 Allocation of scarce resources

1
  Investment decisions
 Project management.
CHAPTER 2: LINEAR PROGRAMING (LP)
2.1. Basic Concepts in LP

Linear programming deals with the optimization (maximization or minimization) of a

function of variables known as objective function, subject to a set of linear equation 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.
 By linearity is meant a mathematical expression in which the expressions among the
variables are linear.
Definition:
LP is a mathematical modelling technique useful for economic allocation of “scarce” or
“limited” resources (such as labor, material, machine, time, warehouse, space, capital, etc.) to
several competing activities (such as products, services, jobs, new equipment, projects etc.)
on the basis of a given criterion of optimality. All organizations, big or small, have at their
disposal, men, machines, money and materials, the supply of which may be limited. Supply of
resources being limited, the management must find the best allocation of resources in order to
maximize the profit or minimize the loss or utilize the production capacity to the maximum
extent.

Steps in formulation of LP model:

1. Identify activities (key decision variables).

2. Identify the objective function as a linear function of its decision variables.
3. State all resource limitations as linear equation or inequalities of its decision variables.
4. Add non-negative constraints from the consideration that negative values of the
decision variables do not have any valid physical interpretation.
5. Use mathematical techniques to find all possible sets of values of the variables
(unknowns) satisfying all the function
6. Select the particular set of values of variables obtained in step five that leads to the
achievement of the objective function.

 The result of the first four steps is called linear programming. The set of solutions
obtained in step five is known as the set of feasible solutions and the solution finally
selected in step six is called optimum (best) solution of the LP problem.

 A typical linear programming has two step

 The objective function
 The constraints
 Technical constraint, and
 The non-negativity constraint

2
The objective function: is a mathematical representation of the overall goal of the
organization stated as a linear function of its decision variables (Xj) to optimize the criterion
of optimality.
 It is also called the measure of performance such as profit, cost, revenue, etc

The general form of the objective function is expressed as:

Optimize (Maximize or Minimize) Z=  C j X j
Where: Z is the measure of performance variable (profit/cost), which is the function of X1, X2,
…,Xn. (Quantities of activities)
Cj (C1, C2… Cn) (parameters or coefficients) represent the contribution of a unit of the
respective variables X1, X2, …,Xn to the measure of performance Z (the objective function).

The constraints: the constraints must be expressed linear equalities or inequalities interms of
decision variables.
The general form of the constraints functions are expressed as:

a ij xj (<,>, ≤, =, ≥) bi

a11X1+a12X2+…+a1nXn (<,>, ≤, =, ≥) b1
a21X1+a22X2+…+a2nXn (<,>, ≤, =, ≥) b2
. . .
. . .
. . .
Am1Xm+am2X2+…+amnXn (<,>, ≤, =, ≥) bn

 aij‟s are called technical coefficients and measure the per unit consumption of the
resources for executing one unit of unknown variable (activities) Xj.
 aij„s can be positive, negative or zero in the given constraints.

 The bi represents the total availability of the ith resource.

 It is assumed that bi ≥ 0 for all i. However, if any bi < 0, then both sides of the
constraint i can be multiplied by 1 to make bi > 0 and reverse the inequality of
the constraint.

Requirements for a linear programming problem: Generally speaking, LP can be used for
optimization problems if the following conditions are satisfied.

1. There must be a well defined objective function which is to be either maximized or

minimized and which can be expressed as a linear function of decision variables
2. There should be constraints on the amount or extent of attainment of the objective and
these constraints must be capable of being expressed as linear equations or inequalities
interms of variables.
3. There must be alternative courses of action. Fore e.g., a given product may be
processed by two different machines and problem may be as to how much of the
product to allocate which machine.
4. Decision variables should be interrelated and non-negative.
5. The resource should be in limited supply.

2.2. Assumptions of Linear Programming

3
 A linear programming model is based on the following assumptions:

1. Proportionality assumption: A basic assumption of LP is that proportionality exists

in the objective function and the constraints. It states that the contribution of each
activity to the value of the objective function Z is proportional to the level of the
activity Xj, as represented by the CjXj term in the objective function. Similarly, the
resource consumption of each activity in each functional constraint is proportional to
the level of the activity Xj, as represented by the aijXj term in the constraint.
What happens when the proportionality assumption does not hold? In most cases we
use nonlinear programming.

2. Additivity assumption: States that every function in a linear programming model

(whether the objective function or the left-hand side of the a functional constraint)is
the sum of the individual contributions of the respective activities.
What happens when the additivity assumption does not hold? In most cases we use
nonlinear programming.

3. Divisibility assumption: States that decision variables in linear programming model

are allowed to have any values, including non-integer values, which satisfy the
functional and non-negativity constraints. Thus, since each decision variable
represents the level of some activity, it is being assumed that the activities can be run
at fractional levels.
In certain situations, the divisibility assumption does not hold because some of or all
the decision variables must be restricted to integer values. For this restrictions integer
programming model will be used.

4. Certainity assumption: This assumption states that the various parameters (namely,
the objective function coefficients, the coefficients in the functional constraints aij and
resource values in the constraints bi are certainly and precisely known and that their
values do not change with time. However, in real applications, the certainity
assumption is seldom satisfied precisely. For this reason it is usually important to
conduct sensitivity analysis after a solution is found that is optimal under the assumed
parameter values.

5. Finiteness: An LP model assumes that a finite (limited) number of choices

(alternatives) are available to the decision-maker and that the decision variables are
interrelated and non-negative. The non-negativity condition shows that LP deals with
real-life situations as it is not possible to produce/use negative quantities.
6. Optimality: In LP, the optimal solution always occurs at the corner point of the set of
feasible solutions.

Important Definitions

 Solution: The set of values of decision variables Xi (i=1, 2, ….,n)which satisfy the
constraints of an LP problem is said to constitute solution to that LP problem
 Feasible solutions: The set of values of decision variables Xi (i=1, 2… n) which
satisfy all the constraints and non-negativity conditions of an LP simultaneously.
 Infeasible solution: The set of values of decision variables which do not satisfy all the
constraints and non-negativity conditions of an LP simultaneously.

4
  Optimum solution: A feasible solution which optimizes (maximizes or minimizes)
the objective function value of the given LP problem is called an optimum basic
feasible solution.
 Unbounded solution: A solution which can increase or decrease the value of the
objective function of LP problem indefinitely is called unbounded solution.

2.3. Methods of Solving Linear Programming Problems

A linear programming problem can be solved by graphic method or by applying algebraic
method, called the simplex method.

1. Graphic method: it can be used when only two variables are involved. This method
consists of the following steps:
i. Represent the given problem in mathematical form.
ii. Plot each of the constraint on the graph.
iii. Identify the feasible region (or solution space) that satisfies all the constraints
simultaneously. Any point on or within the shaded region represents a feasible
solution to the given problem.
iv. Locate the solution points (identify the corner points from the solution space).
v. Determine the optimal solution. A solution that optimizes the objective function.
vi. Interpret the results
a) Maximization Problem: This is the case of Maximize Z with inequalities of constraints in
< form. Consider two models of color TV sets; Model A and B, are produced by a company
to maximize profit. The profit realized is $300 from A and $250 from set B. The limitations
are
a. Availability of only 40 hrs of labor each day in the production department,
b. A daily availability of only 45 hrs on machine time, and
c. Ability to sale 12 set of model A.
How many sets of each model will be produced each day so that the total profit will be as
large as possible?
Resources used per unit
Constraints Maximum Available
Model A Model B Hours
(x1) (x2)
Labor Hours 2 1 40

Machine Hours 1 3 45

Marketing Hours 1 0 12

Profit $300 $250

5
Solution:
1. Formulation of mathematical model of LPP
Max Z=300X1 +250X2
St:
2X1 +X2< 40
X1 +3X2< 45 LP Model
X1 < 12
X1, X2 > 0
2. Convert constraints inequalities into equalities
2X1 + X2 = 40
X1 + 3X2 = 45
X1 = 12
3. Draw the graph by finding out the x– and y–intercepts
2X1 +X2 = 40 ==> (0, 40) and (20, 0)
X1 +3X2 = 45 ==> (0, 15) and (45, 0)
X1 = 12 ==> (12, 0)
X1 , X2 = 0
2X1 +X2 = 40

X2
X1=0
40 X1=12

B
X1 +X2 = 45

15

Feasible C (12, 11)

Region X2=0
X1
D
A 12 20 45

4. Identify the feasible area of the solution which satisfies all constrains. The shaded region
in the above graph satisfies all the constraints and it is called Feasible Region.
5. Identify the corner points in the feasible region. Referring to the above graph, the corner
points are in this case are:
A (0, 0), B (0, 15), C (12, 11) and D (12, 0)
6. Identify the optimal point.

Corners Coordinates Max Z = 300 X1 +250X2

A (0, 0) $0

B (0, 15) $3750

C (12, 11) $ 6350 (Optimal)

D (12, 0) $3600

6
 7. Interpret the result. Accordingly, the highlighted result in the table above implies that 12
units of Model A and 11 units of Model B TV sets should be produced so that the total
profit will be $6350.
b) Minimization Problem: In this case, we deal with Minimize Z with inequalities of
constraints in > form. Suppose that a machine shop has two different types of machines;
Machine 1 and Machine 2, which can be used to make a single product. These machines vary
in the amount of product produced per hr., in the amount of labor used and in the cost of
operation. Assume that at least a certain amount of product must be produced and that we
would like to utilize at least the regular labor force. How much should we utilize on each
machine in order to utilize total costs and still meets the requirement?

Resource Used Minimum Required Hours

Items
Machine 1 (X1) Machine 2 (X2)

Product produced/hr 20 15 100

Labor/hr 2 3 15
Operation Cost $25 $30

Solution:
1. The LP Model:
Min.Z  25 X 130 X 2
St :
20 X 115 X 2 100
LP Model
2 X 13 X 2 15
X1, X 2  0
Dear student, can you graph the above constraints? See step 2 below.
2. The Graph of Constraint Equations:
20X1 +15X2=100 ==> (0, 20/3) and (5, 0)
2X1 + 3X2=15 ==> (0, 5) and (7.5, 0)
X1 , X2 = 0

X2
X1 =0
A (0, 20/3)

Feasible Region

B (2.5, 3.33)
X2 =0

X1
5 C (7.5, 0)

7
3. The Corners and Feasible Solution:

Corners Coordinates Min Z= 25 X1 + 30X2

A (0, 20/3) 200

B (2.5, 3.33) 162.5 (Optimal)
C (7.5, 0) 187.5

Since our objective is to minimize cost, the minimum amount (162.5) will be selected.
X1 = 2.5
X2 = 3.33 and
Min Z= 162.5
Note:
- In maximization problems, our point of interest is looking the furthest point from the
origin (Maximum value of Z).
- In minimization problems, our point of interest is looking the point nearest to the origin
(Minimum value of Z).
Example3. A firm manufactures two products A and B on which the profits earned per unit
are Birr 3 and Birr 4 respectively. Each product is processed on two machines M1 and M2.
Product A requires one minute of processing time on M1 and two minutes on M2, while B
requires one minute on M1 and one minute on M2. Machine M1 is available for not more than
7hrs. 30 minutes, while machine M2 is available for 10hrs during any working day. Find the
number of units of products A and B to be manufactured to get maximum profit.

Solution
Formulation of LPP model
Let x1 and x2 denote the number of units of products A and B to be produced per day.
Objective is to maximize the profit.
i.e. Maximize Z = 3x1 + 4x2
Constraints are on the time available for machines M1 and M2
i.e., for machine M1, 1x1 + 1x2 ≤ 450
for machine M2, 2x1 + 1x2 ≤ 600 Ans: x1=0, x2=450 and Zmax =Birr 1800
Where x1, x2 ≥ 0.

Example 4: The standard weight of a special purpose brick is 5kg and it contains two basic
ingredients B1 and B2. B1 costs Birr 5/kg and B2 costs Birr 8/kg. Strength considerations
dictate that the brick contains not more than 4kg of B1 and a minimum of 2kg of B2. Since the
demand for the product is likely to be related to the price of the brick, find graphically the
minimum cost of the brick satisfying the above conditions.

Solution
Formulation of LPP model
Let the quantities in kg of ingredients B1 and B2 to be used to make the bricks be x1
and x2 resp.
Objective is to minimize the cost of the brick.
i.e., Minimize Z = 5x1 + 8x2 Ans: x1=3kg, x2= 2kg, Zmin=Birr 31

8
 Constraints are
On the quantity of the ingredient B1: x1 ≤ 4,
On the quantity of the ingredient B2: x2 ≥ 2,
On the weight of the brick : x1 + x2 = 5
Where x1, x2 ≥ 0.

2.4. Some Special Cases of Linear Programming

There are certain special cases in linear programming problem. Some of these cases are
resulted from violations of the basic assumptions of LP. These include:

1. Multiple optimal solutions: This is the situation when more than one optimal solution
arises. This will happen whenever the objective function equation represents a line parallel
to some edge of the bounded solution space.
Example: Solve the following problems graphically

Max Z = 110x1+110x2,
Subject to:
x1 + x2 ≤ 9,
x1 ≥ 2,
20x1 + 50x2 ≤ 360,
x2 ≥ 3
x1, x2 ≥ 0.

2. Unbounded solution: it exists when an LP problem has no limit on the constraints i.e., the
feasible region is not bounded in any respect. Theoretically, the value of the decision
variable increases infinitely without violating the feasibility and value of the objective
function can increase to infinity. Mostly this situation occurs when the assumption of
finiteness is violating.

Example: Solve the following problems graphically

Max Z = 30x1+50x2,
Subject to:
3x1 +x2 ≥ 9,
x1 + 2x2 ≥12,
x1 + 2x2 ≥9,
x1, x2 ≥ 0.
3. Infeasible solution: infeasibility is a condition that exists when there is no solution to
an LP problem that satisfies all the constraints and non-negativity restrictions. It
means that the constraints in the problem are conflicting and inconsistent. Infeasibility
depends solely on the constraints and has nothing to do with the objective function

Example: Solve the following problems graphically

Max Z = 3x1+2x2,
Subject to:
-2x1 + 3x2 ≤ 9,
3x1 - 2x2 ≤ -20,
x1, x2 ≥ 0.

9
Binding and Non-Binding Constraints (Resources)

Once the optimal solution to an LPP is obtained, we may classify the constraints as being
binding and or non-binding. A constraint is termed as binding if the left hand side and the
right hand side of the constraint function s are equal when the optimal values of the decision
variables are substituted in to the constraint. On the other hand, if the substitutions of the
value of the decision variables do not lead to equality between the left and the right hand side
of the constraint, it said to be non-binding.

Redundant constraint(s): If the constraint, when plotted does not form part of the boundary
marking the feasible region of the problem, it is said to be redundant. The inclusion or the
exclusion of a redundant constraint obviously does not affect the optimal solution to the
problem.

Exercises: Solve the following LLPs using graphic method

1. a) Min Z = 2x1+x2,
b) Max Z = 2x1+3x2 Ans. a) x1=0, x2=2, Zmin=2, b) Unbounded
Subject to:
x1 - 3x2 ≤ 6,
2x1 + 4x2 ≥ 8,
x1 - 3x2 ≥ -6,
x1, x2 ≥ 0.

2. Max Z = 4x1+5x2, Ans. Unbounded Solution

Subject to:
x1 + x2 ≥ 1,
-2x1 + x2 ≤ 1,
4x1 - x2 ≥ 1
x1, x2 ≥ 0.

2.5. The Simplex Method

When a number of variables in a linear programming are more than two, graphic method
cannot be used because of the difficulty precisely representing the variables using more than a
two dimensional plane. That means it is impossible to represent these variables in the graph
and evaluate the corner points. In such cases, there is a comprehensive method of solving a
linear programming problem which is called the simplex method (simplex algorithm).

 Algorithm is a process where a systematic procedure is repeated (iterated) over and

over again until the desired result is obtained.

Simplex method (algorithm) is an iterative procedure that consists of moving from one basic
feasible to another in such a way that the values of the objective function doesn‟t decrease (in
case of maximization problem) and doesn‟t increase (in case of minimization problem).

 The process continues until an optimal solution is reached if it exists, otherwise, the
problem may be unbounded or not feasible and the simplex method can‟t find the
solution.

10
Conditions for applications of simplex method

There are some conditions that must be fulfilled for the application of simplex method. These
are:
1. The right hand side of each of the constraints bi should be non-negative.

 If the linear programming problem has a negative constraint, we should convert it to

positive by multiplying both sides by -1. Forexample, if one of the constraints is given by
2x1 - 4x2 ≥ -8, we cannot apply simplex method as it is. So, we should convert it to positive
by multiply both sides by -1. i.e., -2x1 +4x2 ≤ 8

2. Each of the decision variables of the problem should be non-negative. If one of the choice
variables is not feasible, simplex method cannot be applied. Therefore, feasibility is a
necessary condition for application of simplex method.

3. The inequality constraints of resources or any other activities must be converted in to

equations.

Since simplex method cannot be applied in the case of inequalities, it is a mandatory that
these inequalities should be converted in to equations. There are some variables that help us to
convert these inequalities in to equations. These variables are: Slack and surplus variables.
Slack variables: are variables that can change the less than type inequality in to equations.

 Slack variables represent unused amount of a given resource by the activities or idle
resources. For example, if one of the constraint is given as:
 x1 ≤ 4 we can convert this in to equations by introducing a slack variable.
 x1 +S1 =4. The values of S1 will range from zero to four. i.e., if x1=0, S1=4 meaning that
all the available resources is not used. If x1=4, S1=0 meaning that all the available
resource is fully utilized and there is no idle resource which is left over.

Surplus variables: if the constraint inequality is of greater than type, then we can convert it
into equation by subtracting some variable called surplus variable. Surplus variables represent
that the requirement of the resource is more than the availability.

 Surplus variables are the negative of the slack variables. For example, if one of the
constraint is given as:
x2 ≥ 8, this inequality can be converted to equation by introducing surplus variable as
x2 –S1= 8.

Note: Slack and surplus variables enter in to the objective function with zero coefficients.

Procedures of simplex method

Step 1: Express the problem in standard form.

The given problem is said to be expressed in standard form if the decision variables are non-
negative, R.H.S of the constraints are non-negative and the constraints are expressed as
equations. Non-negative slack variables are added to the L.H.S of the constraints to convert
them in to equations.

11
Example: Max Z = 3x1+4x2,
Subject to:
x1 + x2 ≤ 450,
2x1 +x2 ≤ 600,
Where x1, x2 ≥ 0.
Standard form of the above LP model is:
Max Z = 3x1+4x2 +0S1 +0S2,
Subject to:
x1 + x2 + S1=450,
2x1 + x2 +S2= 600,
Where x1, x2, S1, S2 ≥ 0

Step 2: Set up the initial solution

 In order to set up the initial solution, first we need to determine the basic and non-
basic variables.
 Slack variables at initial table become basic variables and decision variables are non-
basic variables. If we have n unknowns and m constraints, then m variables should be
basic and the rest (n-m) variables should be non-basic such that their value is zero.
 Basic variables are those variables whose values are obtained from the basic solution
where as non-basic variables are those whose values are assumed as zero in the basic
solution.

Example: The initial solution of the above standard form can be expressed in the
following simplex table
3 4 0 0
Contribution/unit Cj
Body matrix Identity matrix
CB BV x1 x2 s1 s2 b
0 s1 1 1 1 0 450
0 s2 2 1 0 1 600
Interpretation:
1. The first row indicates the coefficients Cj of the variables in the objective function and
they are unchanged in the subsequent tables.
2. The first column CB represents the coefficients of the current basic variables in the obj.
fun. The second column is the basis column and it represents the basic variables of the
current solution.
3. The body matrix (coefficient matrix) represents the coefficients of the constraints.
These coefficients represent the amount of resource required to make a unit of decision
variable.
4. The identity matrix represents the coefficients of slack variables in the constraints.
5. The b-column/solution values xB column indicates the quantities of the variable
resources or R.H.S values of the constraints or values of the basic variables, S1 and S2 in
the initial basic solution.

Step 3: Perform optimality test

The test for optimality can be based on interms of the sign of the members of the index row.

12
 Contribution/unit Cj 3 4 0 0
Body matrix Identity matrix
CB BV x1 x2 s1 s2 b
0 S1 1 (1) 1 0 450key row
0 S2 2 1 0 1 600
Zj 0 0 0 0 0
∆j = Zj- Cj -3 -4 0 0
↑K
Where: Zj=∑CBiaij and Z=∑CBixBi
 Values in the Zj represent the contribution lost per unit of the variables. i.e., the amounts
by which contribution would be reduced if one of the corresponding variables x1, x2 …
was added.
 ∆j is also called the index row. This row determines whether or not the current solution
is optimal. Coefficients in this row represent the net profit (or net contribution or net
marginal improvement) in the values of the objective function Z for each unit of the
respective column variable introduced into the solution. For example, as x 1 increases by
one unit Z will increase by 3 units.
 A positive coefficient in the Zj-Cj row indicates the amount by which the objective
function will be decreased if a unit of the corresponding variable is introduced in to the
solution where as a negative coefficient indicates the amount by which the profit will be
increased if a unit of the corresponding variable is introduced in to the solution. These
coefficients or elements are known as shadow prices or accounting values or imputed
values of the resources.
 The solution is optimal if all the elements of the index row are +ve and zero. If at least
one of the elements of the index row is -ve, the given solution is not optimal and goes to
the next step. The next iteration will give as the next adjusted basic feasible solution.

Step 4: Iterate towards the optimal solution

In each iteration, the simplex method moves the current basic feasible solution to an improved
basic feasible solution. This is done by replacing one current basic variable by a new non-
basic variable as follows.

 Pivot column: The column which has the maximum -ve value of Zj-Cj, is the one which
should enter the solution (entering variable).

 Pivot row: The element lying at the intersection of key column and key row is called key
/pivot element and is enclosed in ( ). This pivot element corresponds to the leaving
variables. Leaving variable= (xbi /aij, aij) <0.

To find the new solution:

 If the key element (pivot number) is 1, then the row remains the same in the new
simplex table

 If the key element (pivot number) is other than 1, then divide each element in the pivot
row (including elements in xb column) by the pivot number, to find the new values of
the row in the new simplex table.

13
  The new values of the elements in the remaining rows for the new simplex table can
be obtained by performing elementary row operations on all rows so that all elements
except the key element in the key column are zero. i.e.,

No. in new row = No. in old row ± (No. above or below key element) (corresponding
No. in old row)

 If all ratios are negative or infinity, the solution is unbounded.

 The next solution will be feasible only if row containing minimum (non-negative)
ratio is selected as key row; in case a row containing higher (non-negative) ratio is
marked as key row, the next solution will become infeasible and there will be
some negative value in the b-column which is illogical.

Contribution/unit 3 4 0 0
Cj Body matrix Identity matrix
CB BV x1 x2 s1 s2 b
4 X2 1 1 1 0 450
0 S2 1 o -1 1 150
Zj 4 4 4 0 1800 2nd f. solution
∆j = Zj- Cj 1 0 4 0
↑Marginal loss value ↑ Shadow prices

Step 5: Perform optimality test for second feasible solution

If there is a negative number in the index row, then the given solution is not optimal. So we
proceed to the step 4 again. However, values in the index row are non-negative the given
solution is optimal. Hence the optimal solution is x1=0, x2= 450, Zmax=1800.

Remarks:

1. The optimal solution obtained above satisfies the non-negativity restrictions. It indicates
that no other set of values of x1 and x2 results in as high value of Z.

2. Slack variable s1 does not exist in the basis column. This means that the first resource is
fully utilized. However, s2 exists in the basis column and its value is 150. This means 150
units of the second resources remains unutilized.

3. Elements in the Zj- Cj row under slack variables indicate the shadow price (or accounting
values or imputed values) of resources.
 4 and 0 are called the shadow prices of the first and the second resources respectively.
 4 under s1 column indicate that if one more unit of the first resource is used, then the
value of Z will increase by 4 units. Therefore, if we want to increase Z, it should
increase the first resource.
 Capacity of the second resource has not been utilized fully; there will be no use
increasing it further.

14
Exercises: Solve the following problems by simplex method and interpret the results.

1. Max Z = 40x1+80x2
Subject to: Ans. x1=9 units, x2=10 units, Zmax=1160
2x1 + 3x2 ≤ 48,
x1 ≤ 15, s1=20, s2=0, s3=20
x1 ≤ 10,
x1, x2 ≥ 0.

2. Max Z = 2x1+5x2
Subject to: Ans. x1=0units, x2=380/9 units, x3=470/3 units, Zmax=3200/3
x1 + 4x2 ≤ 24, s1=0, s2=0, s3=1970/9
3x1 + x2 ≤ 21,
x1 + x2 ≤ 9,
x1, x2 ≥ 0.

Artificial Variables Technique

There are many linear programming problems where slack variables cannot provide a
solution. In these problems at least one of the constraints is of (≥) or (=) type. In such cases
we introduce another type of variables called artificial variables.

 Artificial variables are fictitious and have no any physical or economic meaning. They
assume the role of slack variables in the first iteration, only to be replaced at a later
iteration.

 Artificial variables are merely a device to get the starting basic feasible solution and then
the usual simplex procedure will be adopted until the optimal solution is obtained.

 Artificial variables are expressed by capital letter “A” and used for “≥” type of
inequalities and “=” equations.

 Artificial variables enter in to the objective function with “M” coefficient i.e., it inter in to
the objective function with –ve value (-M) for maximization and +ve value (+M) for
minimization, where M >0.

 M denotes some huge positive number (very huge penalty), for that firm manipulates the
artificial variables.

Steps of the big M-method algorithm

Step 1: Express the LP problem in the standard form by introducing slack variables.

 These variables are added to the L.H S of the constraints of (≤) and subtracted from
the constraints of (≥) type.
Step 2: Add non-negative artificial variable (A) corresponding to constraints having “≥” and
“=” equations.

Step 3: Set up the initial solution

15
  In order to set up the initial solution, first we need to determine the basic and non-
basic variables. In the initial solution, original variables have zero value, so they are
non-basic variables.

Step 4: Solve the modified LPP by the simplex method.

While making iterations, using the simplex method, one of the following three cases may
arise:

a. If no artificial variable remains in the basis and the optimality condition is satisfied,
then the solution is an optimal feasible solution.
b. If at least one artificial variable appears in the basis at zero level and optimality
condition is satisfied, then the solution is optimal feasible solution. But it is a
degenerate solution. The constraints are consistent though redundancy may exist in
them. Redundancy means that the problem has more than the required number of
constraints.
c. If at least one artificial variable appears in the basis at a non-zero level (with +ve
value in b-column) and the optimality condition is satisfied, then the original
problem has no feasible solution. Therefore, the original problem doesn‟t have
feasible solution because it contains artificial variable (a very large penalty M).
Remarks:

1. Slack variables are added to the constraints of (≤) type and subtracted from the
constraints of (≥) type.
2. Artificial variables are added to the constraints of (≥) and (=) type. Equality
constraints require neither slack nor surplus variables.
3. Variables, other than the artificial variables, once driven out in iteration, may re-inter
in a subsequent iteration.

Example:

Max Z = 3x1- x2
Subject to:
2x1 + x2 ≤ 2,
x1 + 3x2 ≥3,
x2 ≤ 4,
x1, x2 ≥ 0.

Step 1: Set up the problem in standard form

Max Z = 3x1- x2 +0s1+0s2+0s3-MA1,

Subject to:
2x1 + x2 + s1 = 2,
x1 + 3x2 - s2 +A1=3,
x2 + s3 = 4,
x1, x2, s1, s2, s3, A1≥ 0.
Step 2: Find initial basic feasible solution

Substituting x1=x2 =s2=0, then the initial b.f.s obtained is: s1=2, A1=3, s3=4, Z=-3M

16
Step 3: perform optimality test
Since the Zj- Cj is -ve under some variable columns the solution is not optimal and can be
improved

Step 4: Iterate in to the optimal solution

Successive iteration yields the following tables

Table 1.
3 -1 0 0 0 -M
Cj
CB BV x1 x2 s1 s2 S3 A1 b r
0 s1 2 1 1 0 0 0 2 2
-M A1 1 (3) 0 -1 0 1 3 1
0 S3 0 1 0 0 1 0 4 4
Zj=∑ CBaij -M -3M 0 M 0 -M -3M i.s
∆j = Zj- Cj -M-3 -3M+1 0 M 0 0
↑K
Table 2.
Cj 3 -1 0 0 0
CB BV x1 x2 s1 s2 S3 b r
0 s1 (5/3) 0 1 1/3 0 1 3/5
-1 x2 1/3 1 0 -1/3 0 1 3
0 S3 -1/3 0 0 1/3 1 3 -9
Zj=∑ CBaij -1/3 -1 0 1/3 0 -1 2nd b.f.s
∆j = Zj- Cj -10/3 0 0 1/3 0
↑K
Table 3.
Cj 3 -1 0 0 0
CB BV x1 x2 s1 s2 S3 b
3 x1 1 0 3/5 1/5 0 3/5
-1 x2 0 1 -1/5 -2/5 0 4/5
0 S3 0 0 1/5 2/5 1 16/5
Zj=∑ CBaij 3 -1 2 1 0 1
∆j = Zj- Cj 0 0 2 1 0

Therefore, optimal solution is given by:

X1=3/5, x2=4/5, Zmax=1
Exercise: solve the f.f. LPP
Max Z = x1+2x2+3x3- x4
Subject to:
x1+2x2+3x3=15
2x1+x2+5x3=20
x1+2x2+x3+ x4 =10
x1, x2, x3, x4 ≥0

Hint: the standard form is:

Max Z = x1+2x2+3x3- x4-MA1-MA2-MA3,
Subject to:
x1+2x2+3x3 +0x4+A1+0A2+0A3=15

17
 2x1+x2+5x3+0x4+0A1+A2+0A3=20
x1+2x2+x3+ x4 +0A1+0A2+A3=10
x1, x2, x3, x4 ,A1, A2, A3 ≥ 0

Special cases in the simplex method application

1. Tie Breaking in Simplex Method

While operating the simplex iteration, tie may occur and we may face indifference in
choosing the entry or the leaving variable. When tie may occur, the following are important to
do:

Tie in the choice of entering variables: the non-basic variable that enters the basis is the
one that gives the largest per unit improvement in the objective function.
 Variable having maximum –ve value in a maximization problem and the maximum +ve
value in the minimization problem in Zj- Cj row is the entering variable. A tie in the
entering variable exists when more than one variable has same largest (or smallest) value.
To break the tie any one of them is selected arbitrarily as the entering variable.

Tie in the choice of leaving variables: This is the case when the ratio is the same.
 This tie can be resolved by choosing one of the variables arbitrarily.

2. No leaving basic variable (unbounded objective function)

 This is the case when all elements of the pivot column are –ve or zero that no variable
qualifies to be a leaving basic variable.
 In this case no finite solution can be determined. Because one or more of the variables
can be increased indefinitely without violating the feasibility.
 The interpretation is that the objective function is not constrained or the value of the
objective function Z can increase indefinitely or the model is poorly constructed.

3. Multiple optimal solutions

 This is the case when the LPP have more than one optimal solution.
 In the simplex table if the Zj- Cj value is zero for a non-basic variable, it indicates that
the existence of more than one optimal solution.

4. Infeasible solution

 If all the constraints are not satisfied simultaneously, the model has no feasible
solution.
 In the case of big M method if at least one of the artificial variables has positive value
and the solution is optimal, then this is the indication of no feasible solution.

Minimization case of simplex method

For most part of finding solution for minimization problem using simplex method are
handled in the same fashion as maximization problem. The three key exceptions are:

 The coefficients of artificial variables (M) have positive sign in the objective function.

18
  The selection of pivot column (entering variables) is based on the largest positive
number (value) in the index (∆j) row.
 The solution is optimal when all values in the index (∆j) row are non positive.
The other alternative method of solving minimization problem is by converting it to
maximization. i.e., Min Z=Max (-Z)

Example
Min Z = 12x1+20x2
Subject to:
6x1 + 8x2 ≥100,
7x1 + 12x2 ≥120,
x1, x2 ≥ 0.

Standard form:
Min Z = 12x1+20x2 +0s1+0s2+MA1+MA2
Subject to:
6x1 + 8x2 -s1+A1=100,
7x1 + 12x2 –s2+A2=120,
x1, x2, s1, s2, A1, A2 ≥ 0.

Cj 12 20 0 0 M M
CB BV x1 x2 s1 s2 A1 A2 b r
M A1 6 8 -1 0 1 0 100 25/2
M A2 7 (12) 0 -1 0 1 120 10
Zj=∑ CBaij 13M 20M -M -M M M 220M ↑I.S
∆j = Zj- Cj 13M-12 20M-20 -M -M 0 0
↑K

Cj 12 20 0 0 M M
CB BV x1 x2 s1 s2 A1 A2 b r
M A1 4/3 0 -1 2/3 1 0 20 15
20 x2 7/12 1 0 -1/12 0 1 10 120/7
Zj=∑ CBaij -35/3+4/3M 20 -M -5/3+2/3M M M 200+20M
∆j = Zj- Cj 4/3M-1/3 0 -M 2/3M-5/3 0 0
↑K

Cj 12 20 0 0
CB BV x1 x2 s1 s2 b
12 x1 1 0 -3/4 1/2 15
20 x2 0 1 7/16 -3/4 5/4
Zj=∑ CBaij 12 20 -1/4 -9 205
∆j = Zj- Cj 0 0 -1/4 -9

2.6. Goal and Integer programming

19
Integer programming

In LPP all the decision variables were allowed to take any non-negative values as it is quite
possible and appropriate to have fractional values in many situations. However, there are
several conditions in business and industry that lead to planning models involving integer
valued variables. That is, many practical problems involve decision variables that must be
integer valued.
There are cases that the fractional values of variables may be meaningless in the context of
the actual decision problem. For example:
 In production, manufacturing is frequently scheduled interms of discrete quantities.
 Production of livestock involves discrete number of animals
Integer programming models require all of the assumptions implicit in linear programming
models except that certain specific variables must be integer valued.
 When all the variables are constrained to be integers, it is called a pure-integer
programming model, and in case only some of the variables are restricted to have integer
values, the problem is said to be a mixed-integer programming problem.
There are two methods used to solve IPP, namely
i. Gomory‟s fractional cut method
ii. Branch and bound method (search method)

i. Gomory’s Fractional Cut Method (Algorithm)

This method consists of first solving the IPP as an ordinary LPP by ignoring the restriction of
integer values and then introducing a new constraint to the problem such that the new set of
feasible solution includes all the original feasible integer solution, but does not include the
optimum non-integer solution initially found. This new constraint is called Fractional Cut or
Gomorian constraint. Then the revised problem is solved using simplex till an optimal integer
solution is obtained.

Gomory‟s Fractional Cut Method involves the following steps:

1. Solve the IPP as an ordinary LPP by ignoring the restriction of integer values
2. Test the integrality of the optimum solution.
 If all Xbi ≥ 0 and are integers, an optimum integer solution is obtained
 If at least one bi is not an integer then go to the next step
3. Select the source row
 If only one Xbi is non-integer, the row corresponding to non-integer solution will
be the source row

 If more than one Xbi is non-integer, choose the row having the largest fractional
part (fi) as a source row.
In this case write each Xbi as (Xbi = xbi +fi): Where xbi is integer part of Xbi
(solution) and fi is the positive fractional part of Xbi. 0< fi<1
 If there is a tie when two or more rows have the same positive larger fi), select
arbitrarily.
4. Find the new constraint (Gomorian constraint) from the source row.

20
 i. ∑aijXj = Xbi
ii. ∑((aij)+fi)Xj = xbi +fi
iii. ∑fiXj ≥ fi
iv. ∑(-fi)Xj ≤ -fi
v. ∑ (-fi)Xj + Gi = -fi Gomorian constraint and G is Gomorian slack.

5. Add the new (Gomorian) constraint at the bottom of the simplex table obtained in step
1 and find the new feasible solution.
 Note that Gomorian slack enter in to the objective function with zero coefficient
6. Test the integrality of the new solution
 Again if at least one Xbi (solution) is not an integer, go to step 3 and repeat the
procedure until an optimal integer solution is obtained.
Example: Find the optimal integer solution for the following LPP

Max Z = x1+x2
Subject to:
3x1 + 2x2 ≤ 5,
x2≤ 2,
x1, x2 ≥ 0 and are integers
Solution
1. Solve the IPP as an ordinary LPP by ignoring the restriction of integer values
Max Z = x1+x2 + 0s1+0s2
Subject to:
3x1 + 2x2 +s1= 5,
x2 +s2= 2,
x1, x2 , s1, s2 ≥ 0 and are integers
Cj 1 1 0 0
CB BV x1 x2 s1 s2 b r
0 s1 3 2 1 0 5
0 s2 0 1 0 1 2
Zj=∑ CBaij 0 0 0 0 0
∆j = Zj- Cj -1 -1 0 0

Cj 1 1 0 0
CB BV x1 x2 s1 s2 b r
1 x1 1 2/3 1/3 0 5/3
0 s2 0 1 0 1 2
Zj=∑ CBaij 1 2/3 1/3 0 5/3
∆j = Zj- Cj 0 -1/3 1/3 0

Cj 1 1 0 0
CB BV x1 x2 s1 s2 b r
1 x1 1 0 1/3 -2/3 1/3
1 x2 0 1 0 1 2
Zj=∑ CBaij 1 1 1/3 1/3 7/3
∆j = Zj- Cj 0 0 1/3 1/3

21
 2. Test the integrality of the optimum solution.
The solution is x1 =1/3, x2=2 and Z=7/3. Since x1 =1/3 is non-integer solution so
that the current feasible solution is not optimal integer solution.

3. Select the source row.  The row corresponding to non-integer solution.

x1 + 1/3s1- 2/3s2 = 1/3,

4. Find the new constraint (Gomorian constraint) from the source row.
x1 + 1/3s1- 2/3s2 = 1/3,
(1+0) x1 + (0+1/3) s1 + (-1+ 1/3) s2 = 0+1/3, and eliminate the integer part
1/3s1+1/3s2 ≥ 1/3,
-1/3s1-1/3s2 ≤ -1/3
-1/3s1-1/3s2 +G1= -1/3Gomorian constraint and G is Gomorian slack.

5. Add the new (Gomorian) constraint at the bottom of the simplex table obtained in
step 1 and find the new feasible solution.

Cj 1 1 0 0 0
CB BV x1 x2 s1 s2 G1 b
1 x1 1 0 1/3 -2/3 0 1/3
1 x2 0 1 0 1 0 2
0 G1 0 0 -1/3 -1/3 1 -1/3
Zj=∑ CBaij 1 1 1/3 1/3 0 7/3
∆j = Zj- Cj 0 0 1/3 1/3 0
The solution is not feasible because the solution (G1=-1/3) is –ve. Therefore, G1 will leave
the base.
 Select the new constraint as a pivot row. In order to select the entering variable,
use:
j
Max( , aij  0)
aij
 1  1 
j   
 Therefore, Max( , aij  0)  Max( 3 ,  3 )  (1,1)
aij  1   1 
  
 3  3 

Choose arbitrarily (Let s1 is chosen as entering variable)

Cj 1 1 0 0 0
CB BV x1 x2 s1 s2 G1 b
1 x1 1 0 0 -1 1 0
1 x2 0 1 0 1 0 2
0 s1 0 0 1 1 -3 1
Zj=∑ CBaij 1 1 0 0 1 2
∆j = Zj- Cj 0 0 0 0 1

22
Since all ∆j ≥ 0 and all Xbi ≥ 0 and are integers, the current feasible solution is optimal integer
solution.
Therefore, the solution is x1=0, x2=2 and Z=2

Exercises:
1. Max Z = 4x1+3x2
Subject to: Ans: x1=3, x2 =0 and Z=12
x1 + 2x2 ≤ 4,
2x1+x2 ≤ 6,
x1, x2 ≥ 0 and are integers

2. Max Z = 3x1+4x2
Subject to: Ans: x1=1, x2 =2 and Z=11
3x1 + 2x2 ≤ 8,
x1+4x2 ≤ 10,
x1, x2 ≥ 0 and are integers

ii. Branch and bound method (search method)

Goal Programming
Firms usually have more than one goal. For example, maximizing total profit, maximizing
market share, maintaining full employment, providing quality ecological management,
minimizing noise level in the neighborhood, and meeting numerous other non-economic
goals. It is not possible for LP to have multiple goals unless they are all measured in the same
units (such as dollars), a highly unusual situation. An important technique that has been
developed to supplement LP is called goal programming.

Goal programming “satisfices,”as opposed to LP, which tries to “optimize.” Satisfice means
coming as close as possible to reaching goals. The objective function is the main difference
between goal programming and LP. In goal programming, the purpose is to minimize
deviational variables, which are the only terms in the objective function.

Example of Goal Programming Harrison Electric Revisited

Goals Harrison‟s management wants to achieve, each equal in priority:
Goal 1: to produce as much profit above $30 as possible during the production period.
Goal 2: to fully utilize the available wiring department hours.
Goal 3: to avoid overtime in the assembly department.
Goal 4: to meet a contract requirement to produce at least seven ceiling fans.

23