Duality

Every linear programming problem has a mirror image problem which is also a linear
programming problem. Using a set of rules, the mirror image problem, called the dual can be
obtained for a given linear programming problem, known as primal problem in this context.

Before writing the dual to an LPP, it is necessary to express it in standard form if it is not so.

By standard form it is meant that

(i) all the variables in the problem should be non-negative; and

(ii) all the constraints should be of ≤ type if the problem is of maximisation type, and of ≥
type if it of the minimisation nature.

the necessary changes may be made as follows:

When all variables in the LPP are not non-negative

Max: 𝑧 = 2𝑥1 + 3𝑥2 + 4𝑥3

s.t.

𝑥1 + 𝑥2 + 𝑥3 ≤ 90

5𝑥1 + 6𝑥2 + 7𝑥3 ≤ 800

𝑥1 ≥ 0; 𝑥3 ≥ 0

Ans.

Replacing 𝑥2 by (𝑥4 − 𝑥5 ), we get

Max: 𝑧 = 2𝑥1 + 3(𝑥4 − 𝑥5 ) + 4𝑥3

Max: 𝑧 = 2𝑥1 + 3𝑥4 − 3𝑥5 + 4𝑥3

s.t.

𝑥1 + 𝑥4 − 𝑥5 + 𝑥3 ≤ 90

5𝑥1 + 6𝑥4 − 6𝑥5 + 7𝑥3 ≤ 800

𝑥1 ≥ 0; 𝑥3 ≥ 0; 𝑥4 ≥ 0 ; 𝑥5 ≥ 0

Page 1 of 7
When all constraints are not in the 'right' direction

If a constraint involves an inequality in a direction opposite to the one desired, then it is

7𝑥1 − 5𝑥2 ≥ 4, and the desired inequality direction is ≤

then we replace this constraint as

−7𝑥1 + 5𝑥2 ≤ −4

On the other hand, if a constraint involves an equation, then, to obtain inequalities, it is

2𝑥1 + 𝑥2 = 50

2𝑥1 + 𝑥2 ≤ 50

2𝑥1 + 𝑥2 ≥ 50

2𝑥1 + 𝑥2 ≤ 50

−2𝑥1 − 𝑥2 ≤ −50

Writing the Dual

Primal
Maximise Z = cx

Subject to

ax ≤ b

x≥0

where, c= row matrix containing the coefficients in the objective function,

x = column matrix containing the decision variables

a = matrix containing the coefficients in the constraints

b = column matrix containing the RHS values of the constraints

Dual

Minimise G = b′ y

Page 2 of 7
Subject to

a′ y ≥ c′

y≥0

where, b′ = transpose of the b matrix of the primal problem,

a′ = transpose of the coefficients matrix of the primal problem,

c′= transpose of the matrix of the objective function coefficients of the primal problem,

y = column matrix of the dual variables.

Example 4.1
For the LPP given in Example 3.1 reproduced below, write the dual.

Maximise 𝑍 = 40𝑥1 + 35𝑥2

Subject to

2𝑥1 + 3𝑥2 ≤ 60

4𝑥1 + 3𝑥2 ≤ 96

𝑥1 , 𝑥2 ≥ 0

In accordance with above, its dual shall be

Minimise 𝐺 = 60𝑦1 + 96𝑦2

Subject to

2𝑦1 + 4𝑦2 ≥ 40

3𝑦1 + 3𝑦2 ≥ 35

𝑦1 , 𝑦2 ≥ 0

Primal: Dual:

Maximise 𝑍 = 40𝑥1 + 35𝑥2 Minimise 𝐺 = 60𝑦1 + 96𝑦2

Subject to Subject to

2𝑥1 + 3𝑥2 ≤ 60 2𝑦1 + 4𝑦2 ≥ 40

4𝑥1 + 3𝑥2 ≤ 96 3𝑦1 + 3𝑦2 ≥ 35

Page 3 of 7
 𝑥1 , 𝑥2 ≥ 0 𝑦1 , 𝑦2 ≥ 0

Example 4.2
Write the dual of the following LPP:

Minimise 𝑍 = 10𝑥1 + 20𝑥2

Subject to

3𝑥1 + 2𝑥2 ≥ 18

𝑥1 + 3𝑥2 ≥ 8

2𝑥1 − 𝑥2 ≤ 6

𝑥1 , 𝑥2 ≥ 0

Solution.
Here, the first two inequalities are in the right direction (being type with a minimisation type
of objective function) while the third one is not. Multiplying both sides by —1, this can be
written as

−2𝑥1 + 𝑥2 ≥ −6

Now, we can write the primal and dual as follows:

Primal Dual

Minimise 𝑍 = 10𝑥1 + 20𝑥2 Maximize 𝐺 = 18𝑦1 + 8𝑦2 − 6𝑦3

Subject to subject to

3𝑥1 + 2𝑥2 ≥ 18 3𝑦1 + 𝑦2 − 2𝑦3 ≤ 10

𝑥1 + 3𝑥2 ≥ 8 2𝑦1 + 3𝑦2 + 𝑦3 ≤ 20

−2𝑥1 + 𝑥2 ≥ −6

𝑥1 , 𝑥2 ≥ 0 𝑦1 , 𝑦2 , 𝑦3 ≥ 0

Dual

Page 4 of 7
Maximize 𝐺 = 18𝑦1 + 8𝑦2 − 6𝑦3

subject to

3𝑦1 + 𝑦2 − 2𝑦3 ≤ 10

2𝑦1 + 3𝑦2 + 𝑦3 ≤ 20

𝑦1 , 𝑦2 , 𝑦3 ≥ 0

Example 4.3
Obtain the dual of the LPP given here:

Maximise 𝑍 = 8𝑥1 + 10𝑥2 + 5𝑥3

Subject to

𝑥1 − 𝑥3 ≤ 4

2𝑥1 + 4𝑥2 ≤ 12

𝑥1 + 𝑥2 + 𝑥3 ≥ 2

3𝑥1 + 2𝑥2 − 𝑥3 = 8

𝑥1 , 𝑥2 , 𝑥3 ≥ 0

Dual
Min 𝐺 = 4𝑦1 + 12𝑦2 − 2𝑦3 + 8𝑦4 − 8𝑦5

Subject to

𝑦1 + 2𝑦2 − 𝑦3 + 3𝑦4 − 3𝑦5 ≥ 8

0𝑦1 + 4𝑦2 − 𝑦3 + 2𝑦4 − 2𝑦5 ≥ 10

−𝑦1 + 0𝑦2 − 𝑦3 − 𝑦4 + 𝑦5 ≥ 5

𝑦1 , 𝑦2 , 𝑦3 , 𝑦4 , 𝑦5 ≥ 0

Page 5 of 7
Another way

Minimise 𝐺 = 4𝑦1 + 12𝑦2 − 2𝑦3 + 8𝑦6

Subject to

𝑦1 + 2𝑦2 − 𝑦3 + 3𝑦6 ≥ 8

0𝑦1 + 4𝑦2 − 𝑦3 + 2𝑦6 ≥ 10

−𝑦1 + 0𝑦2 − 𝑦3 − 𝑦6 ≥ 5

𝑦1 , 𝑦2 , 𝑦3 ≥ 0, 𝑦6 unrestricted in sign

Example 4.4.
Obtain the dual of the following LPP:

Maximise 𝑍 = 3𝑥1 + 5𝑥2 + 7𝑥3

Subject to

𝑥1 + 𝑥2 + 3𝑥3 ≤ 10

4𝑥1 − 𝑥2 + 2𝑥3 ≥ 15

𝑥1 , 𝑥2 ≥ 0, 𝑥3 unrestricted in sign

Ans.

Page 6 of 7
 −4𝑥1 + 𝑥2 − 2𝑥3 ≤ −15

Primal
Maximise 𝑍 = 3𝑥1 + 5𝑥2 + 7𝑥4 − 7𝑥5

Subject to

𝑥1 + 𝑥2 + 3𝑥4 − 3𝑥5 ≤ 10

−4𝑥1 + 𝑥2 − 2𝑥4 + 2𝑥5 ≤ −15

𝑥1 , 𝑥2 , 𝑥4 , 𝑥5 ≥ 0

Dual

Minimize 𝐺 = 10𝑦1 − 15𝑦2

Subject to

𝑦1 − 4𝑦2 ≥ 3

𝑦1 + 𝑦2 ≥ 5

3𝑦1 − 2𝑦2 ≥ 7

−3𝑦1 + 2𝑦2 ≥ −7

𝑦1 , 𝑦2 ≥ 0

Alternatively

Minimize 𝐺 = 10𝑦1 − 15𝑦2

Subject to

𝑦1 − 4𝑦2 ≥ 3

𝑦1 + 𝑦2 ≥ 5

3𝑦1 − 2𝑦2 = 7

𝑦1 , 𝑦2 ≥ 0

Page 7 of 7