Duality in Linear Programming

(LPP solved by Duality)

Ex. 1 .Write down the optimal solution of the following LPP by solving its dual:

Max = 3 − 2
s. t. ≤ 4
≤ 6
+ ≤ 5
≥ 1
and , ≥ 0.

Solution: The given LPP in standard primal form is

Max = 3 − 2
s. t. + 0. ≤ 4
0. + ≤ 6
+ ≤ 5
0. − ≤ −1
and , ≥ 0.

∴ its dual can be written as

Min = 4 + 6 + 5 −
s. t. + 0. + + 0. ≥ 3
0. + + − ≥ −2
and , , , ≥ 0.

Now, converting the objective function from minimization to maximization and introducing the
surplus variable to the first constraint and slack variable to the second constraint, the
dual LPP in standard form is

Max ′ = − 4 − 6 − 5 + + 0. + 0. , where ′ = −
s. t. + 0. + + 0. − = 3
0. − − + + = 2
and , , , , , ≥ 0.

Taking = 0, = 0, = 0, = 0, we get = 3, = 2, which is the starting B. F. S.

to the dual. Therefore, by using Simplex Method, we obtain the following simplex table:
−4 −6 −5 1 0 0 Mini Ratio
B w Y Y Y Y Y Y
Y −4 3 1 0 1 0 −1 0
Y 1 2 0 −1 −1 1 0 1
′ = c w = − 10 − 0 5 0 0 4 1

In this table, since − ≥ 0 for all , therefore, the solution under test is optimal. Hence the
optimal solution to the dual is
= 3, = 0, = 0, = 2 with Min = − Max ′ = 10 .
Now, from the above simplex table of the dual problem, we find that the net evaluations −
corresponding to the surplus variable and the slack variable are 4 and 1 respectively.
Therefore, by Rules of Duality,
= 4, = 1 with = 10 .
which is the required optimal solution to the given LPP .

Ex. 2. Solve the following LPP by converting it into dual:

Min = 15 + 10
s. t. 3 + 5 ≥ 5
5 + 2 ≥ 3
and , ≥ 0.

Solution: The given LPP is in standard primal form. Therefore, its dual is

Max = 5 + 3
s. t. 3 + 5 ≤ 15
5 + 2 ≤ 10
and , ≥ 0.

Introducing the slack variables and , the dual LPP in standard form is

Max = 5 + 3 + 0. + 0.
s. t. 3 + 5 + = 15
5 + 2 + = 10
and , , , ≥ 0.

Taking = 0, = 0, we get = 15, = 10, which is the starting B. F. S. to the dual.

Therefore, by using Simplex Method, we obtain the following simplex tables:
5 3 0 0 Mini Ratio
w
B w Y Y Y Y
Y
Y 0 15 3 5 1 0 5
Y 0 10 5 2 0 1 2 (Min)→
= c w = 0 − −5 −3 0 0
w
↑
Y
19
Y 0 9 0 1 − (Min)→
5

Y 5 2 1 0 5
= 10 − 0 −1 0 1
↑
Y 3 0 1 −

Y 5 1 0 −
= − 0 0
In the final simplex table, since − ≥ 0 for all , therefore, the solution under test is
optimal. Hence the optimal solution to the dual is

= , = with Max = .

Now, from the final simplex table of the dual, we find that the net evaluations −
corresponding to the slack variables and are and respectively.
Therefore, by Rules of Duality, the optimal solution to the given LPP is

= , = with = .

Ex. 3. Solve the following LPP. Write its dual and its solution.
Max = 30 + 23 + 29
s. t. 6 + 5 + 3 ≤ 26
4 + 2 + 5 ≤ 7
and , , ≥ 0.

Solution: Introducing the slack variables and , the given LPP in standard form can be
written as follows:
Max = 30 + 23 + 29 + 0. + 0.
s. t. 6 + 5 + 3 + = 26
4 + 2 + 5 + = 7
and , , , , ≥ 0.
Taking = 0, = 0, = 0, we get = 9, = 15, which is the starting B. F. S. to the
primal. Therefore, by using Simplex Method, we obtain the following simplex tables:

30 23 29 0 0 Mini Ratio
B x Y Y Y Y Y
Y 0 26 6 5 3 1 0

Y 0 7 4 2 5 0 1 (Min) →
= c x = 0 − − 30 − 23 − 29 0 0
↑
Y 0 0 2 − 1 −
Y 30 1 0 (Min) →
= − 0 −8 0
↑
Y 0 −4 0 − 1 −
Y 23 2 1 0

= − 16 0 0

In the final simplex table, since − ≥ 0 for all , therefore, the solution under test is
optimal. Hence the optimal solution to the given LPP is
 = 0, = , = 0 with = .

The given LPP is in standard primal form. Therefore, its dual is

Min = 26 + 7
s. t. 6 + 4 ≥ 30
5 + 2 ≥ 23
3 + 5 ≥ 29
and , ≥ 0.

Now, from the final simplex table of the primal problem, we find that the net evaluations −
corresponding to the slack variables and are 0 and respectively.
Therefore, by Rules of Duality, the optimal solution to the dual problem is
= 0, = with Min = .

Ex. 4. By solving the dual of the LPP

Min = −
s. t. 2 + ≥ 2
− − ≥ 1
and , ≥ 0

show that it has no feasible solution.

Solution: The given LPP is in standard primal form. Therefore, its dual is

Max = 2 +
s. t. 2 − ≤ 1
− ≤ −1
and , ≥ 0.

Introducing the slack variable to the first constraint and the surplus variable to the
second constraint, the dual LPP in standard form is

Max = 2 + + 0. + 0.
s. t. 2 − 5 + = 1
− + − = 1
and , , , ≥ 0.

∵ there does not exist a starting B. F. S., therefore, introducing the artificial variable to the
second constraint and assigning a very large negative price − M to this artificial variable, we
obtain the LPP as follows:

Max = 2 + + 0. + 0. − M
s. t. 2 − 5 + = 1
− + − + = 1
and , , , , ≥ 0.
Taking = 0, = 0, = 0, we get = 1, = 1, which is the starting B. F. S. to the dual.
Therefore, by using Simplex Method in the usual manner, we obtain the following simplex
tables:
3 −1 0 0 −M Mini Ratio
w
B w Y Y Y Y Y
Y
Y 0 1 2 −1 1 0 0 − ve
Y −M 1 −1 1 0 −1 1 1 (Min) →
= c w = −M − M− 2 −M− 1 0 M 0
↑
Y 0 2 1 0 1 −1 1
Y 1 1 −1 1 0 −1 1
= 1 − −3 0 0 −1 1+ M

In the last simplex table, since − = − 1 ≤ 0 and < 0, = 1, 2, therefore, the dual
problem has an unbounded solution.

Hence the given primal LPP has no feasible solution.

-------------------------------- x --------------------------------