You are on page 1of 5

Introduction

On the 1st of October we learned about Dual Simplex Method.

We can solve a Linear Program by Primal Simplex algorithm, or also by Dual Simplex. (So it is useful
because it provides the decision maker with an alternative way of looking at a problem.)

Classification of Linear Programming Problems


1. We call normal maximum LP problem an LP problem in which all the variables are required to
be non-negative and all the constraints are less than or equal to (≤) form.
2. While, a normal minimum LP problem is an LP problem in which all the variables are required
to be non-negative and all the constraints are greater than or equal to (≥) form.

Non-Normal Linear Programming Problem


In case of maximization problem:
1. If the constraint is Less than or equal to (≤) type, then no change is required.
2. If the constraint is Greater than or equal to (≥) type, then convert the ‘≥’ type inequality into
‘≤’ type by multiplying it by ‘-1’.
3. Any = type constraint can be changed to two constraints with the same left hand side, one of
them will be ≤ type another is ≥ type. Then the ≥ type is changed according to Point 2.

In case of minimization problem:


1. If the constraint is Less than or equal to (≤) type, then convert the ‘≤’ type inequality into ‘≥’
type by multiplying it by ‘-1’.
2. If the constraint is Greater than or equal to (≥) type, then no change is required.
3. Any = type constraint can be changed to two constraints with the same left hand side, one of
them will be ≤ type another is ≥ type. Then the ≤ type is changed according to Point 2.

Feasibility of the Primal and the Dual Simplex Methods


Let us consider the following rules:

1. The table (tableau) is primal feasible if the basis solution is non-negative vector.
2. The table is dual feasible if the reduced cost is non-negative vector.
3. The table is optimal, if and only if the table is both primal and dual feasible.
4. The Primal Simplex Method keeps the primal feasibility and tries to reach dual feasibility.
5. The Dual Simplex Method keeps the dual feasibility and tries to reach the primal feasibility.

Algorithm and Example


Before we start solving the given problem, let us consider the following steps(rules) for the Dual
Simplex Method:

Step 1. Formulate the Problem:

a.) Formulate the mathematical model of the given problem.


b.) If the objective function is minimization type then change it into maximization type.
c.) Convert all ≥ constraint to ≤ constraint by multiplying by -1.
d.) Transform every ≤ constraint into an = constraint by adding a slack variable to every
constraint and assign a 0 cost coefficient in the objective function.

Step 2. Find out the Initial basic solution

Find the initial basic feasible solution by setting zero value to the decision variables, and use only the
slack variables to satisfy the constraints.

Step 3. Test for Optimality

a.) If all the components of X B ≥ 0, then current solution is an optimal solution, terminate the
process.
b.) If there exists negative value in X B , then select one such negative component. Its row can be
called as key row. (If we can not choose such a negative number, the LP is infeasible.)
Zj
c.) Find Ratio= for any negative value in Key Row where the column of this
¿ KeyRow j∨¿ ¿
number (with index j) is outside of the basis. Take the absolute values of these ratios.
d.) Find the minimum of the above ratios, the column where it is taken is called Key Column.
(Usually this rule is called Minimum Rule). The number in the key row and key column is the
pivot number.
e.) Perform a basis transformation to get a new tableau.
f.) Repeat Step 3.

Let us consider the following example:

Find the solution using dual-simplex method.

x 1+ x2 ≥1

2 x1 + x 2 +3 x3 ≥ 3

x 1+ 2 x 2 ≥ 5

4 x1 +2 x 2+ x 3 ≥7

x≥0
z=10 x 1+12 x 2+ 15 x 3 →min

Solution:

Step 1.: Convert all ≥ constraint to ≤ constraint by multiplying by -1

−x 1−x 2 ≤−1

−2 x1 −x2 −3 x 3 ≤−3

−x 1−2 x2 ≤−5

−4 x1 −2 x 2−x 3 ≤−7

x≥0
^z =−10 x 1−12 x2−15 x 3 →max
Step 2.: Transform every ≤ constraint into an = constraint by adding a slack variable to every
constraint and assign a 0 cost coefficient for the slack variables in the objective function.

−x 1−x 2+ e1 =−1

−2 x1 −x2 −3 x 3 +e 2=−3

−x 1−2 x2 +e 3=−5

−4 x1 −2 x 2−x 3 +e 4 =−7

x≥0,e≥0
^z =−10 x 1−12 x2−15 x 3 +0 e 1+ 0 e2 +0 e 3 +0 e 4 →max

Also, we can write the objective function as the following: ^z =−10 x 1−12 x2−15 x 3 →max

Step 3.: Create the tableau

B XB x1 x2 x3 e1 e2 e3 e4

e1 -1 -1 -1 0 1 0 0 0

e2 -3 -2 -1 -3 0 1 0 0

e3 -5 -1 -2 0 0 0 1 0

e4 -7 -4 -2 -1 0 0 0 1

Z 0 10 12 15 0 0 0 0

Step 4.: With the help of Minimum rule, we can find the pivot element:

1. Let us choose a negative X B row → let it be Row 3=e3 (blue color) with the value of -5
2. Find the column by using Minimum Rule:
10
x 1 colum n ' s ratio= =10
|−1|
12
x 2 colum n ' s ratio= =6
|−2|

3. The minimum is taken in column x 2 and the pivot element is -2 (orange color).

Step 5.: Create a new tableau:

 divide the pivot row with the pivot element

 eliminate the nonzero values in the column of the pivot value by standard basis
transformation: We subtract “lambda” times of the row of the pivot value from the other
rows, the appropriate lambda values are: ½ for the first row, also for the second row, 1 for
the fourth row, and -6 for the last row.
The next simplex tableau is:

B XB x1 x2 x3 e1 e2 e3 e4

e1 3 −1 0 0 1 0 −1 0
2 2 2
e2 −1 −3 0 -3 0 1 −1 0
2 2 2
x2 5 1 1 0 0 0 −1 0
2 2 2
e4 -2 -3 0 -1 0 0 -1 1

Z -30 4 0 15 0 0 6 0

Step 4.: With the help of Minimum rule, we can find the pivot element:

1. Let us choose a negative X B row → let it be Row 4 =e 4 (blue color) with the value of -2
2. Find the column by using Minimum Rule:
4
x 1 colum n ' s ratio= =1.33
|−3|
15
x 3 colum n' s ratio= =15
|−1|

3. The minimum is taken in column x 1 and the pivot element is -3 (orange color).

The next simplex tableau is:

B XB x1 x2 x3 e1 e2 e3 e4

e1 11 0 0 1 1 0 −1 −1
6 6 3 6
e2 1 0 0 −5 0 1 0 −1
2 2 2
x2 13 0 1 −1 0 0 −2 1
6 6 3 6
x1 2 1 0 1 0 0 1 −1
3 3 3 3
Z −98 0 0 41 0 0 14 4
3 3 3 3
We reached an optimal solution, because all the components of X B ≥ 0. So, we need to terminate
the process.

Optimal solution is arrived with value of variables as:

−98
Z max= ≈−3 2.666
3

( 23 , 136 ,0|161 , 12 , 0 ,0)


X opt =

You might also like