ME/MF F344 – Engineering Optimization

Linear Programming Problems (LPP)

by
Siva PrasadAV S
BITS Pilani Course IC: Prof. AK Gupta
Pilani Campus
 LPP – Formulation

Reddy Mikks Company produces both interior and exterior paints from two raw materials
M1 and M2. The following table provides the basic idea of the problem:

A market survey indicates that the daily demand for interior paint cannot exceed that for
exterior paint by more than 1 ton. Also, the maximum daily demand for interior paint is 2
tons. Find the optimum (best) product mix of interior and exterior paints that maximizes
the total daily profit.

ME/MF F344 2
LPP – Formulation & Graphical Solution

Decision variables: Exterior paint → x1; Interior paint → x2

Objective function: Maximize Z = 5x1 + 4x2 in $1000

Constraints: Subject to 6x1 + 4x2 ≤ 24,

x1 + 2x2 ≤ 6,
x2 ≤ x1 + 1, x2
x2 ≤ 2, x1, x2 ≥ 0 (0,6)

Optimal point B (x1 = 3 ton, x2 = 1.5 ton)

Optimal value Z* = $21000. (0,3)
(0,2) D C
B
(0,1)
E
A
O (4,0) (6,0) x1

ME/MF F344 3
LPP – Formulation & Graphical Solution
x2
Important points: (0,6)

• Presence/absence of constraints 3 and 4 do not affect

the optimal solution.
(0,3)
(0,2) D C
• Multiple optimal solutions are possible when the slope B
of the objective function coincides with one of the (0,1)
E
constraints. A
O (4,0) (6,0) x1
• The feasible space or solution space is always convex.

• Optimum LP solution always occurs at a corner point

of the solution space.

ME/MF F344 4
Algebraic Solution – towards Simplex

Convert all the inequalities into equations by introducing slack/surplus variables, with
following conditions to satisfy:

1. All the constraints are equations with nonnegative right – hand side.
2. All the variables are nonnegative.
Let us take the Reddy Mikks problem, discarding the 3rd and 4th constraints.

Objective function: Maximize Z = 5x1 + 4x2

Constraints: Subject to 6x1 + 4x2 ≤ 24,

x1 + 2x2 ≤ 6, x1, x2 ≥ 0

In the standard LP form, Maximize Z = 5x1 + 4x2 + 0s1 + 0s2

Subject to 6x1 + 4x2 + s1 = 24,
x1 + 2x2 + s2 = 6, x1, x2, s1, s2 ≥ 0
Here s1 and s2 are called slack variables.

ME/MF F344 5
Algebraic Solution – by enumeration
Maximize Z = 5x1 + 4x2 + 0s1 + 0s2
Subject to 6x1 + 4x2 + s1 = 24,
x1 + 2x2 + s2 = 6,
x1, x2, s1, s2 ≥ 0
4 unknowns → x1, x2, s1, s2 and 2 equations.
Any two variables can take arbitrary values. Let us
take 0 as the arbitrary value. Further, there are 6
choices of selecting any two variables.

Let us list all the 6 choices. Call these variables as ‘non–basic’.

x2
S. No Non–Basic Basic Z
(0,6)
Variables Variables
1 x1, x2 (0,0) s1,s2 (24,6) 0 (O)
2 x1, s1 (0,0) x2,s2 (6,-6) Infeasible (0,3) C
3 x1, s2 (0,0) x2,s1 (3,12) 12 (C) B
(3,1.5)

4 x2, s1 (0,0) x1,s2 (4, 2) 20 (A)

5 x2, s2 (0,0) x1,s1 (6,-12) Infeasible A
O (4,0) (6,0) x1
6 s1, s2 (0,0) x1,x2 (3, 1.5) 21 (B)

ME/MF F344 6
Simplex Algorithm
Maximize Z –5x1 – 4x2 – 0s1 – 0s2 = 0
Subject to 6x1 + 4x2 + s1 = 24, S. No Non–Basic Basic Z
x1 + 2x2 + s2 = 6, Variables Variables
x1, x2, s1, s2 ≥ 0 1 x1, x2 (0,0) s1,s2 (24,6) 0 (O)

Simplex Table Pivot element

Identity columns x2
Entering variable (0,6)

x1 x2 s1 s2 RHS Ratio
(0,3) C
Z –5 –4 0 0 0
(3,1.5)
B
s1 6 4 1 0 24 4
Basic

s2 1 2 0 1 6 6
Leaving variable A
O (4,0) (6,0) x1
Each simplex table corresponds to a point in the graph, only when
the columns corresponding to the basic variables are identity.

ME/MF F344 7
Simplex Algorithm – Steps
1. Write the given LPP problem in the standard LP format.
2. Before constructing the simplex table, bring all the terms of the objective function to the
side of z.
3. Construct the simplex table with slack variables as basic and the remaining non–basic.
Ensure that the columns corresponding to basic variables have identity.

4. Identify the entering variable from the non – basic variable set as the one which has the
most negative coefficient in the objective function row. This decision is ensures that the
next solution is more optimal and hence called optimality criterion.
5. Take the ratios of the right hand side with the coefficients in the column of the entering
variable.
6. Choose the one with the least positive ratio to be the leaving variable. This decision is
ensures the constraints are satisfied and hence called feasibility criterion.
7. The element at the intersection of entering variable column and the leaving variable row
is called pivot element. Perform row operations to make the column corresponding to
the new basic variable identity column.

8. Repeat steps 4 to 7 until all the elements in the coefficients row are non–negative, which
indicates optimality or the best possible solution.
ME/MF F344 8
 Simplex Algorithm
x2
(0,6)
x1 x2 s1 s2 RHS Ratio
Z –5 –4 0 0 0
(0,3) C
s1 6 4 1 0 24 4
Basic

(3,1.5)
B
s2 1 2 0 1 6 6

R1→ R1 + 5/6R2; R2→ R2 /6; R3→ R3 – R2 /6 A

O (4,0) (6,0) x1
x1 x2 s1 s2 RHS Ratio
Z 0 –2/3 5/6 0 20 Optimality Reached.
x1 1 2/3 1/6 0 4 6 x1 x2 s1 s2 RHS Ratio
s2 0 4/3 –1/6 1 2 1.5 Z 0 0 3/4 1/2 21
x1 1 0 1/4 –1/2 3
R1→ R1 + 1/2R3; R2→ R2 – R3/2; R3→ 3/4R3 x2 0 1 –1/8 3/4 1.5

ME/MF F344 9
≥ & = Type Constraints
x2
Maximize Z = 5x1 + 4x2 D Optimal point: (0,6)
(0,6)
Optimal value Z* = 24
Subject to 6x1 + 4x2 = 24,
x1 + 2x2 ≥ 6, x1, x2 ≥ 0 Feasible space: line CD
(0,3) E
In the standard LP form, (3,1.5)
C
Maximize Z = 5x1 + 4x2 – 0S1
Subject to 6x1 + 4x2 = 24, B
A
x1 + 2x2 – S1 = 6, x1, x2, S1 ≥ 0 O (4,0) (6,0) x1
Here S1 is called surplus variable. Identity columns?

Initial Simplex Table:.

x1 x2 S1 RHS Ratio x1 x2 S1 RHS Ratio
Z –5 –4 0 0 Z –5 –4 0 0
x2 6 4 0 24 x1 6 4 0 24
S1 1 2 –1 6 x2 1 2 –1 6

ME/MF F344 10
 Artificial Variables – Big M Technique
x2
Maximize Z = 5x1 + 4x2 D Optimal point: (0,6)
(0,6)
Optimal value Z* = 24
Subject to 6x1 + 4x2 = 24,
x1 + 2x2 ≥ 6, x1, x2 ≥ 0 Feasible space: line CD
(0,3) E
In the standard LP form, (3,1.5)
C
Maximize Z = 5x1 + 4x2 – 0S1 – M (A1+A2)
Subject to 6x1 + 4x2 + A1= 24, B
x1 + 2x2 – S1 + A2 = 6, A
O (4,0) (6,0) x1
x1, x2, S1, A1, A2 ≥ 0
Here S1 is called surplus variable and A1, A2 are
artificial variables.
Initial Simplex Table:.
x1 x2 S1 A1 A2 RHS Ratio
Z –5 –4 0 M M 0 R1→ R1 – M (R2 + R3) to
get identity columns in
A1 6 4 0 1 0 24
the basic variable
A2 1 2 –1 0 1 6 columns
ME/MF F344 11
Artificial Variables – Big M Technique

x1 x2 S1 A1 A2 RHS Ratio x2
D
(0,6)
Z –5–7M –4–6M M 0 0 –30M
A1 6 4 0 1 0 24 4
A2 1 2 –1 0 1 6 6 (0,3) E
(3,1.5)
C
R1→ R1 + (7M+5)/6R2; R2→ R2 /6; R3→ R3 – R2 /6
B
x1 x2 S1 A1 A2 RHS Ratio O A
(4,0) (6,0) x1
Z 0 (–2–4M)/3 M (5+7M)/6 0 –2M +20
x1 1 2/3 0 1/6 0 4 6
A2 0 4/3 –1 –1/6 1 2 1.5

R1→ R1 + (2M+1)/2R3; R2→ R2 – R3 /2; R3→ 3/4R3

ME/MF F344 12
Artificial Variables – Big M Technique

x1 x2 S1 A1 A2 RHS Ratio x2
D
(0,6)
Z 0 0 –1/2 (3+4M)/4 (1+2M)/2 21
x1 1 0 1/2 1/4 –1/2 3 6
x2 0 1 –3/4 –1/8 3/4 1.5 – (0,3) E
(3,1.5)
R1→ R1 + R2; R2→ 2R2; R3→ R3 + 3/2 R2 C

B
x1 x2 S1 A1 A2 RHS Ratio O A x1
(4,0) (6,0)
Z 1 0 0 M+1 M 24
S1 2 0 1 1/2 –1 6
x2 3/2 1 0 1/4 0 6 Ignore the
negative ratio

Reached optimality. Optimal point (x1, x2) = (0,6) and the optimal value is Z* = 24.

ME/MF F344 13
THANK YOU

ME/MF F344 14