The Big- M Method for Solving LPM

 Big M Method is used to solve LPM involving a mixture of ‘≥’ , ‘≤’ and , ‘=’
constraints as well as minimization LPM.

 It is a modified version of the simplex method for maximization LPM, that first
finds a basic feasible solution by adding "artificial (A)" variables to the problem.
Artificial variables have no importance in a LMP. They are used only to facilitate
an initial basic feasible solution.

 Not to keep artificial variables in the solution the objective function of the original
LPM is modified using a very large number represented by ‘M’ to ensure that the
artificial variables are all equal to 0.

Step 1 : Make sure that all given constraints in any LPM have positive values
on the right hand side. If constraints have negative values multiply both sides
by -1 and reverse the inequality sign
Step 2: Standard form :
 To convert inequalities to equation introduce slack variables, surplus
variable and artificial variables depending on the inequality sign.
 Add Slack variable(si) to ith ‘≤’ constraint
 Add an artificial variable (Ai) to an ith ‘=’constraint
 Subtract a surplus variable(si) and add an artificial variable (Ai )to
an ith ‘≥’ constraint.

 The coefficients of the objective function for slack(s) and surplus(s)

variables will be 0.

 The coefficients of artificial variables in the objective function will be a

large negative or positive number M.

 -MAi for maximization LPM

 +MAi for a minimization LPM
Step 3: Initial tableau

 Initial basic feasible solution is obtained by setting the decision

and surplus value 0. The artificial variables and the slack
variables as basis.

 Across the top of the initial tableau list the decision variables first
then the slack/surplus variables followed by artificial variables.

 Z value for each column is obtained by multiplying each of the

numbers in the column by the respective row coefficient in column
C and adding up the result.

 C-Z value for each column is obtained by subtracting value of Z

from the C value in the top row.
Step .4 Apply the procedure outlined in the maximization problem to move to
another basic feasible solution with the following exceptions.

1) For a minimization problem select the most negative value in row C-Z
to determine the entering or incoming variable.

2) Since an artificial variable (Ai) value would not provide any information
about the LPM , it can be removed from the column altogether in
subsequent tableau once it leaves the basis.

Step 5: Determine whether the solution is optimal or not. The solution is said to
be optimal if
 there is no positive value in row C-Z ( for maximization problem
 there is no negative value in row C-Z for a minimization problem.
Example 1 :
Maximize Z = x1 – x2 + 3x3
Subject to x1 + x2 ≤ 20
x1 + x 3 = 5
x2 + x3 ≥ 10
x1, x2 ≥ 0

Example 2: Consider the example of mixing two types of food F1 and

F2 with the following LPM

Minimize Cost = 5X + 7Y
Subject to: 2X + Y ≥ 8 ---Vitamin A requirement
X + 2Y ≥ 10--- Vitamin B requirement
X≥0,Y≥0
Special cases

1. An optimal solution is said to be unbounded if the objective

function can be improved with out limit. If we select a new pivot
column and there are no positive entries in this column then the
LPPM has unbounded solutions.

2. A current value of zero in the quantity column of a simplex

tableau for a basic variable signals a condition known as
degeneracy. This condition occurs when there is a tie for the
lowest ratio.

In choosing the outgoing variable if there are two or more

candidates with smallest ratio choose either one and proceed with
the computation.
3. If a LPPM has no feasible solution it is referred as infeasibility. This can
occur if an optimal solution contains an artificial variable with positive value.

4. Multiple optimal solutions arise when a non-basic variable has a zero value in
the bottom row C-Z. It is possible to find an alternative optimal solution by
exchanging the non-basic variable that has a corresponding zero value in C- Z
row by a basic variable already in the solution with no change in the optimal
value of the objective function.

Example

Alternative solution x1 = 8, x2 = 2, s1 = 0, s2 =0 and z= 120

 Sensitivity Analysis (Post optimality Analysis)

 Sensitivity analysis enables managers to answer ‘What if...’ questions

concerning the impact of potential changes to model parameters or
constants.

 Managers use sensitivity analysis in order to determine how changes in

constraints, coefficients of objective function and coefficients of constraint
would affect the final (optimal) solution

 Sensitivity analysis uses the final simplex tableau containing the optimal
solution as the starting point.
1. Change in the RHS of a Constraint

 Shadow price: is a marginal value found in slack or surplus (s)column

in row Z of a simplex tableau. The shadow price indicates the impact
on the value of an objective function that would result from a one-unit
change (increase or decrease) in the right hand side of value of a
constraint,

 This interpretation is only valid, as long as the change in RHS constraint

(bi) is not large enough to make the current basic solution infeasible.
The range over which the right-hand side value of a constraint can
change is referred to as the Range of Feasibility.
Consider the following furniture company example
where, x1 : Number of tables produced x2 : number of chairs produced

Maximize Z= 90x1 + 25x2

8x1 +2x2 ≤ 400 Assembly department constraint
2x1 + x2 ≤ 120 finishing department constraint
x1 ≥ 0, x2 ≥ 0 non-negative constraints
basis C 90 25 0 0 Q

x1 x2 S1 S2

X1 90 1 0 1/4 -1/2 40

X2 25 0 1 -1/2 2 40

z 90 25 10 5 4600
c-z 0 0 -10 -5
The shadow prices are 10 and 5.
 If time in assembly department is increased by 0ne hour per day the profit
will increase by Birr10. Similarly a one hour increase in the finishing
department will increase profit By Birr 5

 A one hour decrease in assembly department will decrease profit by birr10

and; a one hour decrease in finishing department will decrease profit by
Birr 5.

 All coefficients in a simplex tableau are called substitution rates. They

indicate how much, and in what direction , a one- unit change in a
column variable will change the current quantity value(Q).

Example: In the furniture company the numbers in s1 ( assembly dept)

column indicate the following:

 a one labor –hour increase in assembly department will cause value of x1

to increase by ¼ and value of x2 to decrease by -1/2

 Range of Feasibility: Divide each value in a quantity column (Q)
by corresponding entry in a constraint (slack or surplus ) column.
The allowable decrease and increase values respectively, are the
smallest positive ratio and the negative ratio close to 0.

 Range of feasibility for maximization LPM is

Lower = Original _ Smallest positive
Limit amount ratio

Upper = Original + |Negative ratio|

amount Closest to 0

Limit
 For a minimization problem , the allowable decrease is the
negative ratio closest to zero and the allowable increase is the
smallest positive ratio
 Given amount of change ‘d’ in value of constraint within the range
of feasibility the corresponding change in optimal solution and
objective function can be obtained using the slack variable column
substitution rate as follows.

Revised (New) = Current ± d (substitution rate )

solution
Solution
Note:

 For an increase in the level of constraint beyond its upper limit , the
additional amount needed to achieve the upper limit (negative ratio
close to 0) is used to compute the revised solution. The slack variable
in the constraint will enter into solution replacing the variable whose
revised value will be equal to zero.

 For a decrease in the level of constraint beyond its lower limit, a new
simplex solution must be generated. The shadow prices and
substitution rates do not hold when the lower limit is exceeded.
Example 1: A ready –to-wear manufacturer is about to mass –produce
its three latest designs shirts: A, B, and C. Each design A shirt requires 6,
8 and 12 minutes in cutting, sewing and finishing departments; each
design B shirt requires 2, 10 and 9 minutes; while each design C shirt
requires 4, 4 and 3 minutes, respectively in each department. The
maximum time available in cutting, sewing and finishing departments,
respectively are 2400, 3200 and 3600 minutes per week. The profit on
each design A, B and C are birr 15, birr 6 and birr 5 respectively. The
final simplex tableau of the model is given in the next slide.

The LPM is:

Maximize profit z = 15x1 + 6x2 + 5x3 Birr
Subject to Cutting 6x1 + 2x2 + 4x3  2400 minutes
Sewing 8x1 +10x2 + 4x3  3200 minutes
Finishing 12x1 + 9x2 + 3x3  3600 minutes
x1, x2, x3 ≥0
 x1 : number of design A shirts
x2 ; number of design B shirts
x3 : number of design C shirts

Basis Q
x1 x2 x3 s1 s2 s3

x3 0 -1 1 2/
5 0 - 1/
5 240

s2 0 6 0 -4/5
1 -4/5 320

x1 1 1 0 -1/10 0 2/
15 240
i) What is the best product mix (optimal solution)? What is the optimal profit?
ii) Which department’s time is underutilized? By how many minutes?
iii) What are the shadow prices ? Which resource has the highest marginal
value?
iv) Over what range is each of the RHS values valid without any change on
shadow price.
v) Find the new solution assuming that the weekly time in finishing
department is increased by i) 1000 minutes ii) 1500 minutes
vi) If forced by circumstances to reduce the time available in cutting
department by i) 200 minutes ii) 500 minutes per week find the new
solution.
2. Change in an Objective Function Coefficients
Sensitivity analysis for an objective function coefficient
involves placing a range on the coefficient’s value. There are
two cases

Case 1: Changes of a variable that is not currently in the

solution mix ( non-basic variable). The range over which a
non-basic variable’s objective function coefficient can change
without causing the variable to enter the solution mix is called
its range of insignificance and is given by,

Range of insignificance ≤ original coefficient + |c-z|

Case 2: The range over which the objective function coefficient of a basic
variable that is in a solution can change without changing the optimal
values of the decision variables is called the range of optimality.
 The range of optimality is obtained dividing the values in row c-z by
corresponding values in row of the variable in question and choosing
the smallest positive ratio and the negative ratio closest to 0.
Range of optimality
Lower limit = original coefficient - |negative ratio|
Upper limit = original coefficient + positive ratio.

 As long as the actual value of the objective function coefficient is

within the range of optimality, the current basic feasible solution will
remain optimal.
Note: The change in basic variable within the range of optimality would
change the optimal value of the objective function. However the change
in non-basic variable within the range of significance will not change
the optimal value of the objective function
Example: Consider the ready –to –wear manufacturer LPM.
i. If the manager decides to increase the price of design B shirts to
Birr 10 what will be the implication of this change to the current
optimal solution and to the total profit?

ii. Find the range of optimality for x1 and x3? Find revised profit if
profit on each design C shirt is increased by i) Birr 2 ii) Birr 6
EXERCISE

An electronics firm produces x1 transistors, x2 resistor and x3 electronic tubes.

Each product must pass through three services: engineering service, direct
labor and administration service. Given the time (in hours) needed to each
product and the available time in each service the analyst of the firm
formulated the LPP given below.

Maximize profit z= 10x 1+ 6x2 +4x3

Subject to, 2x1+ 3x2 +4x3 ≤ 26 (engineering time constraint)
x1+ 3x2 +2x3 ≤ 22 (direct labor time constraint)
6x1+ 3x2 +4x3 ≤33 (administration time constraint)
x1, x2, x3 ≥0
The analyst also came up with the following partial final simplex tableau

C Q
x1 x2 x3 s1 s2 s3
Basis

s1 0 0 8/5 1 -4/5 -1/5 1

x2 0 1 8/15 0 2/5 -1/15 7

x1 1 0 2/5 0 -1/5 1/5 2

Complete the final simplex tableau and answer the questions in the next
slide.
i. How many of each type of electronic product should be produced
to maximize profit?

ii. How much is the maximum profit ?

iii. Which service hour is underutilized? by how many hours?

iv. If the manager of the firm is considering scheduling over time

which service would be the best choice for overtime? Why?

v. If the direct labor time is increased by 2 ½ hours at the same cost

as the previous time what impact would this have on profit?

vi. Find the range of feasibility for the direct labor time constraint.
 Duality in Linear Programming

 Every linear programming problem has an associated linear programming

problem called the dual problem. Referring to the original formulation of
the linear programming problem as the primal problem.

 A fundamental property of the primal–dual relationship is that the optimal

solution to either the primal or the dual problem also provides the optimal
solution to the other.

 In cases where the primal and the dual problems differ in terms of
computational difficulty, we can choose the easier problem to solve, The
dual of a maximization LPM a minimization LPM and vis versa.
Formulation of a dual problem
1. Identify the variables of dual problem which are same as the number of
constraints in primal problem.

2. The constraint values of the primal problem become the coefficient of

objective function of a dual problem,  and the coefficient of the
variables in the objective function of a primal problem become the
constraint value in the dual problem.

3. The first column in the constraint inequality of primal problem become

the first row in a dual problem and similarly the second column of
constraint become the second row in the dual problem and so on.

4. The directions of inequalities also changed, i.e. in the dual problem,

the sign is the reverse of a primal problem. Such that in maximization
primal problem, the inequality sign “≤” but in the dual problem, the
sign of inequality becomes “≥”.

Note: The dual of a dual problem is the primal problem.

 If the dual problem has an optimal solution, the primal problem has
an optimal solution and vice versa. Furthermore, the values of the
optimal solutions to the dual and primal problems are equal.