Linear Programming
Dr. T. T. Kachwala
Slide 2
Introduction to Linear Programming
Linear Programming is a problem solving approach to help
managers make decisions.
Linear programming is a method of mathematical programming
that involves optimization of a certain function called objective
function subject to defined constraints and restrictions.
Although it has very wide applications, the common type of
problems handled in linear programming call for determining a
product-mix that would maximize the total profit, given the profit
rates of the products involved and the resource requirements for
each of them, along with the amount of resources available.
Slide 3
Example of Product Mix Problem
M/s. P.M.S. Industries makes two kinds of leather purses for ladies.
Purse type ‘A’ is of high quality and purse type ‘B’ is of lower
quality. Contributions per unit were Rs.4 and Rs.3 for purse type
‘A’ and type ‘B’ respectively. Each purse of type ‘A’ requires two
hours of machine time per unit & that of type ‘B’ requires one hour
per unit. The company has 1000 hours per week of maximum
available machine time. Supply of leather is sufficient for 800
purses of both types combined per week. Each type requires the
same amount of leather. Purse type ‘A’ requires a fancy zip and 400
such zips are available per week, there are 700 zips for purse type
‘B’ available per week. Assuming no market or finance constraints
recommend an optimum product mix.
Example of Product Mix problem
1. Three types of data are defined in Product Mix Problem:
a. Requirement of resources per unit of product
b. Availability of resources, i.e., constraints of linear
programming problem
c. Objective function
2. The product mix problem is defined as follows; “How many units of
each type of product should be manufactured and sold, given the
requirement of resources within the constraints of maximum
availability such that the profit contribution is maximized.”
Slide 5
Introduction to Linear Programming
Thus, linear programming method is a technique for
choosing the best alternative from a set of feasible
alternatives, in situations where the objectives function, as
well as the constraints, are expressed as linear
mathematical functions.
Generally, the constraints in maximization problems are of
the `<’ type, and in the minimization problems of the `>’
type. But a given problem may involve a mix of
constraints involving the signs of ‘<’, ‘>’, and ‘=’. Usually
the decision variables are non-negative.
Example of Marketing Mix problem
Suppose a media specialist has to decide on allocation of advertisement in three
media vehicles. The unit costs of a message in the three media are Rs.1000,
Rs.750 & Rs. 500. The available budget is Rs.20,000 for a campaign period of one
year. The first medium is a monthly magazine and it is desired to advertise not more
than one insertion in one issue. Also at least six messages should appear in the
second medium. The number of messages in the third medium should strictly lie
between 4 & 8. The expected effective audience for unit message in the media
vehicle is shown below :
------------------------------------------------------------------------------
Vehicle : 1 2 3
------------------------------------------------------------------------------
Expected effective Audience : 80000 60000 45000
------------------------------------------------------------------------------
Draft this as an LP problem to find the optimal allocation that would maximize total
effective audience.
Example of Investment Mix problem
The Boffin Investment Trust wishes to invest Rs.10,00,000. There are five different
investment choices. The current returns on investments are as follows:
------------------------------------------------------------------------------------
Investment choice : L M N O P
------------------------------------------------------------------------------------
Annual yield % : 10 8 6 5 9
------------------------------------------------------------------------------------
Because of risk element involved, management restricts the investment in L to not
more than the combined total investment in N, O and P. Total investment in M & P
combined must be at least as large as that in N. Also management wishes to restrict
its investment in M to a level not exceeding that of O.
Construct an LP model to determine the optimum allocation of investment funds
amongst these 5 choices.
Formulation (Modeling) of Linear Programming Problem
Formulation (Modeling) of LPP means translating the
problem statements in to mathematical equations.
All solutions to linear programming problem starts with
formulation. It involves:
1. Defining the decision variables
2. Imposing the restrictions on the decision variables
as defined by the constraints
3. Defining the objective function in terms of the
decision variables
Significance of the title Linear Programming
The word Linear signifies that all equations (objective
function & constraints) are linear, i.e. all variables are
raised to power of index 1.
The word Programming is interpreted as:
1. A set of logical instructions (iteration – step by step
improvement) which are repeatedly applied to the
problem till we generate an optimum solution.
2. Choosing a course of action (LP is a Planning &
Scheduling tool)
Assumptions of Linear Programming
1. Proportionality: The contribution of each variable in the objective function on its usage of the
resources is directly proportional to the value of the variable. That is, if resource availability
increases by some percentage, then the output shall also increase by the same percentage.
2. Additivity: Sum of the resources used by different activities must be equal to the total quantity of
resources used by each activity for all the resources individually or collectively.
3. Divisibility: The variables are not restricted to integer values.
4. Certainty or Deterministic: Coefficients in the objective function and the constraints are
completely known and do not change during the period under study in all the problems that are
considered.
5. Finiteness: Variables and constraints are finite in number.
6. Optimality: In a linear programming problem we determine the decision variables so as to
optimise the objective function of the LPP.
7. The problem involves only one objective function namely profit maximisation or cost minimisation.
Slide 11
Graphical Solution to the LPP
Graphical Solution Procedure is a convenient way of identifying
an optimum extreme point for two variable problems.
Solution of LPP graphically requires plotting all the constraints of
the problem simultaneously and identifying the feasible region for
the constraints. The feasible region for the problem represents the
region in which any point would satisfy all the constraints.
The optimal solution may be found by determining extreme points
of the feasible region (the variable and the corresponding
objective function value at each point) that optimizes ( maximizes
for a maximization problem and minimizes for a minimization
problem) the value of the objective function.
Slide 12
Graphical Solution to Linear Programming
Working Rules and Guidelines
1. Formulation of Linear programming problem: It involves:
a. defining the decision variables
b. imposing the restrictions on the decision variables as defined by the constraints
c. defining the objective function in terms of the decision variables.
2. Draw the graph corresponding to each equation of LPP.
3. Identify feasible solution area and the corresponding corner (extreme) points
4. Calculate the value of the objective function (Z or C) corresponding to each corner
(extreme) point and identify the corner (extreme) point corresponding to the
optimum value of the objective function. Such a corner point represents optimum
solution for the LPP.
Slack / Surplus Variable
Any unutilized (excess) capacity for ‘<’ constraint is referred to as the Slack
associated with the constraint. Similarly excess capacity corresponding to ‘>’ is
referred as Surplus.
For ‘<’ constraint, Slack Variables are added to the left hand side of the
inequality to represent excess capacity. Similarly for ‘>’ constraint, Surplus
Variables are subtracted from the left hand side of the inequality to convert the
constraint in to equality form.
Unutilized / Excess capacity makes no contribution to profit; thus slack / surplus
variables have coefficients of zero in the objective function.
Whenever a linear program is written in a form with all constraints expressed as
equalities, it is said to be written in Standard form.
Slide 14
Unique, Multiple, Infeasibility and Unbounded
Solutions
A problem may have a unique optimal solution, multiple optimal solutions, an unbounded
solution, or infeasible solution:
1. Unique optimal solutions means one optimum solution is available for LPP.
2. Multiple optimal solutions means more than one optimum solution is available for LPP
(Alternate optimum solutions).
3. Unbounded solution is present when the feasible region is unbounded from above and the
objective function is of maximization type, so that it is possible to increase the objective function
value indefinitely. Optimum solution is not available. A minimization problem with non-negative
variables will not have unbounded solution.
4. Infeasibility (no feasible solution) exists when there is no common point in the feasible areas for
the constraints of a problem. The feasible region is empty in such a case. Optimum solution is
not available for such a LPP.
Slide 15
Simplex (Mathematical) Solution to Linear Programming
Working Rules and Guidelines
Formulate the LP problem: (same as Graphical Solution)
1. identification of decision variables.
2. imposing the restrictions on the decision variables as defined
by the constraints of LPP.
3. defining the objective function in terms of the decision
variables.
Convert the inequalities into “=” sign:
If the inequalities is of “≤” type, then add a slack variable
‘S’.
If the inequalities is of “≥” type, then add an artificial
variable ‘A’ and subtract a surplus variable ‘S’.
If it is “=” sign then add artificial variable ‘A’.
This is done to form a unit matrix and thereby help to get an
Slide 16
Simplex (Mathematical) Solution to Linear Programming
Working Rules and Guidelines
Define the objective function (in terms of all the variables):
If the problem is of the minimization type, then the objective
function is given by C = c1 x1 + c2 x2 + ----, where c1, c2 --- are
the cost coefficients for the decision variables x 1, x2 ------.
However, if the problem is of the maximization type, then the
objective function is given by Z = p 1 x1 + p2 x2 + --, where p1,
p2 ----- are the profit coefficients for the decision variables x 1,x2
---.
Slide 17
Simplex (Mathematical) Solution to Linear Programming
Working Rules and Guidelines
“pi / ci” are referred as associate coefficients. On similar
lines the associate coefficient for Slack variable “S” is
zero & Artificial variable “A” is M. The associated
coefficient is indicated in the table below:
Variable Profit Cost
Xi pi ci
Si 0 0
Ai M M
Slide 18
Simplex (Mathematical) Solution to Linear Programming
Working Rules and Guidelines
Simplex table (Iterations):
1. Layout the initial Simplex table as per the standard format.
2. Calculate the net cell evaluation for all the variables listed
in the Simplex table. It will be observed that the net cell
evaluation for some of the variables are greater than zero.
This condition implies that the initial Simplex table is not
an optimum solution (Initial Simplex table is only a
reference table).
3. Modify the initial Simplex table through a process of iteration (step by
step improvement).
Slide 19
Simplex (Mathematical) Solution to Linear Programming
Working Rules and Guidelines
Simplex table (Iterations):
1. Modification of the initial Simplex table through a process of
iteration involves some mathematical calculations
(Replacement ratio, Gauss-Jordan rules for new coefficients
etc.).
2. The process continues till we reach the Optimum Simplex
Table (such that the net cell evaluation of all the variables in the
simplex table are ≤ 0).
Calculate the minimum cost / maximum profit using the objective
function for the values of the decision variables corresponding to
optimum solution.
