Non-Standard form of LPP

Big M and Two‐Phase Methods

1
 Non-Standard form of LPP
• Simplex method is suitable for LPP in standard form
i.e. Maximization type objective
functional constraints ≤ type
non-negativity constraints
and RHS non-negative
• How to deal with the LPP in other (Non-standard) forms
- LPP with “≥” and “=” type constraints.
- Negative RHS
Difficulty: In identifying initial BFS 𝑥2
• Example:
3x1 + x2 = 3
Maximize 𝑍 4𝑥1 𝑥2
Subject to 3𝑥1 𝑥2 3 (3/5, 6/5)
4𝑥1 3𝑥2 6
𝑥1 2𝑥2 4
𝑥1 2𝑥2 4
𝑥1 , 𝑥2 0 0 (1, 0) 𝑥1
4𝑥1 3𝑥2 6
Identify feasible region? 2
 Approach for Non-Standard form of LPP
• Augmented form
𝑚𝑎𝑥𝑖𝑚𝑖𝑧𝑒 𝑍 4𝑥1 𝑥2
Subject to 3𝑥1 𝑥2 3
4𝑥1 3𝑥2 – 𝑥3 6
𝑥1 2𝑥2 𝑥4 4
𝑥1, 𝑥2, 𝑥3, 𝑥4 0

𝑥3 ∶ surplus variable, and 𝑥4 ∶ slack variable

But, the initial BFS is not available.

• Artificial‐Variable Techniques
1. Big‐M method
2. Two‐phase method
3
 Big M-method
 Introduce artificial variable to get initial BFS.
 Modified problem
Maximize 𝑍 4𝑥1 𝑥2 – M𝑥̅ 5 – M𝑥̅ 6
Subject to 3𝑥1 𝑥2 + 𝑥̅ 5 = 3
4𝑥1 3𝑥2 – 𝑥3 + 𝑥̅ 6 6
𝑥1 2𝑥2 𝑥4 4
𝑥1, … , 𝑥4 0 and 𝑥̅ 5, 𝑥̅ 6 ≥ 0
• For each artificial variable 𝑥̅ j subtract M𝑥̅ j from objective function
(add for minimum type problem), where M is a Large positive
number
• In order to get rid of the artificial variables in the final optimal
solution, we assign a very large penalty in the objective function
4
 Apply the procedure of Simplex method

Basis 𝑥1 𝑥2 x3 x4 x5 x6 RHS Ratio

𝑥̅ 5 3 1 0 0 1 0 3 1
Iteration 0 𝑥̅ 6 4 3 -1 0 0 1 6 3/2 Z row in not in proper
form (coefficient of basic
𝑥4 1 2 0 1 0 0 4 4 variable must be zero)
Z -4 -1 0 0 M M 0
Z -4-7M -1-4M M 0 0 0 -9M R0 → R0 – MR1 – MR2
𝑥1 1 1/3 0 0 1/3 0 1 3
Iteration 1 𝑥̅ 6 0 5/3 -1 0 -4/3 1 2 6/5
x4 0 5/3 0 1 -1/3 0 3 9/5
Z 0 (1-5M)/3 M 0 (4+7M)/3 0 4-2M
𝑥1 1 0 1/5 0 3/5 -1/5 3/5
Iteration 2 𝑥2 0 1 -3/5 0 -4/5 3/5 6/5
𝑥4 0 0 1 1 1 -1 1
Z 0 0 1/5 0 (5M+8)/5 (5M-1)/5 18/5

3 6
𝑶𝒑𝒕𝒊𝒎𝒂𝒍 𝒔𝒐𝒍𝒖𝒕𝒊𝒐𝒏 ∶ 𝑥1 ,𝑥 ,Z 18/5
5 2 5
5
 𝑥2

3x1 + x2 = 3
Path traced by Big M
(3/5, 6/5)

𝑥1 2𝑥2 4

0 (1, 0) 𝑥1
4𝑥1 3𝑥2 6

Note:
If the optimality condition is satisfied and
• No artificial variable remains in the basis current solution is optimal

• At least one artificial variable appears in basis at zero level current

solution is optimal but degenerate solution

• At least one artificial variable in basis at non-zero level the problem has
no feasible solution
6
 Two-Phase Method

Maximize 𝑍 4𝑥1 𝑥2
subject to 3𝑥1 𝑥2 3
4𝑥1 3𝑥2 6
𝑥1 2𝑥2 4
𝑥1 , 𝑥2 0

• Modified problem
Maximize 𝑍 4𝑥1 𝑥2
subject to 3𝑥1 𝑥2 𝑥̅ 5 3
4𝑥1 3𝑥2 – 𝑥3 𝑥̅ 6 6
𝑥1 2𝑥2 𝑥4 4
𝑥1 , 𝑥2 , . . , 𝑥4 , 𝑥̅ 5 , 𝑥̅ 6 0
7
• Phase I: Problem
Minimize 𝑍’ 𝑥̅ 5 𝑥̅ 6 ⇒ 𝑀𝑎𝑥𝑖𝑚𝑖𝑧𝑒, 𝑍’ 𝑥̅ 5 𝑥6
𝑠𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜 3𝑥1 𝑥2 𝑥̅ 5 3
4𝑥1 3𝑥2 – 𝑥3 𝑥̅ 6 6
𝑥1 2𝑥2 𝑥4 4
𝑥1 , 𝑥2 , . . , 𝑥4 , 𝑥̅ 5 , 𝑥̅ 6 0

8
Apply simplex procedure for Phase I

Basis x1 x2 x3 x4 x5 x6 RHS Ratio

x5 3 1 0 0 1 0 3 1
Z row is not in proper form
Iteration 0 x6 4 3 -1 0 0 1 6 3/2 (coefficient of basic
variable must be zero)
x4 1 2 0 1 0 0 4 4
-Z’ 0 0 0 0 1 1 0
-Z’ –7 –4 1 0 0 0 -9 R0 → R0– R1– R2
x1 1 1/3 0 0 1/3 0 1 3
Iteration 1 x6 0 5/3 -1 0 -4/3 1 2 6/5
x4 0 5/3 0 1 -1/3 0 3 9/5
-Z’ 0 -5/3 1 0 7/3 0 -2
x1 1 0 1/5 0 3/5 -1/5 3/5
Iteration 2 x2 0 1 -3/5 0 -4/5 3/5 6/5 Indicates that
x4 0 0 1 1 1 -1 1 artificial
variables left
-Z’ 0 0 0 0 1 1 0
the basis.

• If minimum value of sum of artificial variables is positive, then LP has no feasible

solution. Otherwise, proceed for phase II
9
• Now, Phase II could be started from the final constraints row manipulation at
phase I and original objective function.

• Phase II

Basis x1 x2 x3 x4 RHS
x1 1 0 1/5 0 3/5
Iteration 0 x2 0 1 -3/5 0 6/5 Z row in not in proper
form (coefficient of basic
x4 0 0 1 1 1 variable must be zero)

𝐳 -4 -1 0 0 0
𝐳 0 0 1/5 0 18/5 𝑅 → 𝑅 + 4𝑅 𝑅

• Optimal solution ∶ 𝑥1 3/5, 𝑥2 6/5, 𝑧 18/5

10
Dealing with unrestricted variables

• Convert such variable as difference of two non-negative variables.

• Example:
• 𝑥𝑘 – unrestricted in sign
• 𝑥𝑘 = 𝑥 - 𝑥

• Assignment
𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒, 𝑧 𝑥1 2𝑥2 𝑥3
𝑠𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜, 2𝑥1 3𝑥2 4𝑥3 4
3𝑥1 5𝑥2 2𝑥3 7
𝑥1, 𝑥2 0 𝑎𝑛𝑑 𝑥3 𝑖𝑠 𝑢𝑛𝑟𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑

11