You are on page 1of 18

Lecture 11:

Post-Optimal Analysis

September 23, 2009


Lecture 11

Today

• Dual-Simplex Algorithm

• Post-Optimal Analysis

Chapters 4.4 and 4.5.

IE 310/GE 330 1
Lecture 11

Dual Simplex Method

• The dual simplex method will be crucial in the post-optimal analysis

• It used when at the current basic solution, we have


• The z -coefficients (reduced costs) satisfy optimality condition
• But the basic solution is infeasible
• Technical detail:
all constraints in the problem have to be converted to ≤

IE 310/GE 330 2
Lecture 11

Example 4.4-1
minimize z = 3x1 + 2x2 + x3 minimize z = 3x1 + 2x2 + x3
subject to 3x1 + x2 + x3 ≥ 3 subject to −3x1 − x2 − x3 ≤ −3
−3x1 + 3x2 + x3 ≥ 6 3x1 − 3x2 − x3 ≤ −6
x1 + x2 + x3 ≤ 3 x1 + x2 + x3 ≤ 3
x1 , x 2 , x 3 ≥ 0 x1 , x 2 , x 3 ≥ 0
Introducing slack variables x4, x5, x6, we generate the initial simplex table
using the slacks as basic variables

Basic x1 x2 x3 x4 x5 x6 RHS
z −3 −2 −1 0 0 0 0
x4 −3 −1 −1 1 0 0 −3
x5 3 −3 −1 0 1 0 −6
x6 1 1 1 0 0 1 3
A basic variable with the most negative value (infeasible) has to leave
In this case x5 is leaving the basis

IE 310/GE 330 3
Lecture 11

The entering variable is determining by the absolute value ratio test for
the z -row entries and the x5-row entries, but only on the negative values:
(here the entries in x2 and x3 column)

−2 −1

minimum of ,
−3

−1

The minimum is attained in x2-column so x2 enters;


-3 is the pivot element (in x5-row and x2-column)
We carry out the regular “pivoting operations”, and obtain

Basic x1 x2 x3 x4 x5 x6 RHS
z −5 0 − 12 0 − 23 0 4
x4 −4 0 − 23 1 − 13 0 −1
1
x2 −1 1 3
0 − 13 0 2
2 1
x6 2 0 3
0 3
1 1

IE 310/GE 330 4
Lecture 11

In the next iteration x4 leaves and x2 enters and the pivoting element is − 23
Again, we carry out the regular “pivoting operations”, and obtain

Basic x1 x2 x3 x4 x5 x6 RHS
z −3 0 0 − 12 − 12 0 9
2
x3 6 0 1 − 32 1
2
0 3
2
1 1 3
x2 −3 1 0 2
−2 0 2
x6 −2 0 0 1 0 1 0

The table is optimal - why?

IE 310/GE 330 5
Lecture 11

Dual in Sensitivity and Post-Optimal Analysis


• Sensitivity analysis without the use of the dual problem
• Increase or decrease of available resources
LP Model: Change in the right-hand side - Di
• Increase or decrease in product prices/production costs
LP Model: Change in the objective coefficients - di
• Using dual simplex method we can determine an optimal so-
lution for the perturbed problem from the optimal table of the
original problem

• Using the dual problem/duality we can analyze additional changes


• Introducing new operations (resources)
LP Model: New constraints in the original (primal) problem
• Introducing new products (activities)
LP Model: New variables in the original (primal) problem

IE 310/GE 330 6
Lecture 11

We discuss this changes in the light of duality


• Changes affecting feasibility
LP Model: right-hand side change or a new constraint
How to recover optimal if the perturbation causes the change in
basic optimal solution?

• Changes affecting optimality


LP Model: objective coefficient or new variable
How to find new optimal?

IE 310/GE 330 7
Lecture 11

Changes affecting feasibility


TOYCO Revenue Maximization (Primal) Problem
maximize z = 3x1 + 2x2 + 5x3
subject to x1 + 2x2 + x3 ≤ 430 (machine 1)
3x1 + 2x3 ≤ 460 (machine 2)
x1 + 4x2 ≤ 420 (machine 3)
x1 , x 2 , x 3 ≥ 0
Current optimal: z = 1350 x1 = 0, x2 = 100, x3 = 230
Suppose TOYCO wants to increase its operations to 602, 644, and 588.
New TOYCO Constraints
subject to x1 + 2x2 + x3 ≤ 603
3x1 + 2x3 ≤ 644
x1 + 4x2 ≤ 588
x1 , x 2 , x 3 ≥ 0

How will this change affect the TOYCO’S optimal solution and revenue?

IE 310/GE 330 8
Lecture 11

Optimal Table
Basic x1 x2 x3 x4 x5 x6 RHS
z 4 0 0 1 2 0 1350
x2 − 14 1 0 21 − 14 0 100
x3 32 0 1 0 12 0 230
x6 2 0 0 −2 1 1 20
The changes in the operations affect the right-hand side result in
Perturbed Table
Basic x1 x2 x3 x4 x5 x6 RHS
z 4 0 0 1 2 0 ?
x2 − 14 1 0 21 − 14 0 ?
x3 32 0 1 0 12 0 ?
x6 2 0 0 −2 1 1 ?

IE 310/GE 330 9
Lecture 11

Use column rule to determine the RHS data except for z -row
new solution = inverse optimal × new right-hand side
New solution
      
1
x2 2
− 14
0 602 140
1
      
 x3  =  0 0   644  =  322 
   
   2 
x6 −2 1 1 588 28

New solution is feasible, hence stays optimal.

Optimal operating point x1 = 0, x2 = 322 and x3 = 28


What is the new optimal value?

IE 310/GE 330 10
Lecture 11

New optimal value: substitute the new solution in the objective

z = 3x1 + 2x2 + 5x3 = 3 × 0 + 2 × 140 + 5 × 322 = 1890

IE 310/GE 330 11
Lecture 11

Alternative TOYCO plan

TOYCO Revenue Maximization (Primal) Problem


maximize z = 3x1 + 2x2 + 5x3 resource slack
subject to x1 + 2x2 + x3 ≤ 430 (machine 1) x4
3x1 + 2x3 ≤ 460 (machine 2) x5
x1 + 4x2 ≤ 420 (machine 3) x6
x1 , x 2 , x 3 ≥ 0
Current optimal: z = 1350 x1 = 0, x2 = 100, x3 = 230
As an alternative TOYCO considers shifting 20 units of time from machine
3 to machine 1, because at the current optimal there is a slack of 20 units
Optimal Table
Basic x1 x2 x3 x4 x5 x6 RHS
z 4 0 0 1 2 0 1350
x2 − 14 1 0 12 − 14 0 100
x3 32 0 1 0 12 0 230
x6 2 0 0 −2 1 1 20

IE 310/GE 330 12
Lecture 11

New TOYCO Constraints


subject to x1 + 2x2 + x3 ≤ 450
3x1 + 2x3 ≤ 460
x1 + 4x2 ≤ 400
x1 , x 2 , x 3 ≥ 0

How will this change affect the TOYCO’S optimal solution and cost?

IE 310/GE 330 13
Lecture 11

The changes in the operations affect the right-hand side result in


Perturbed Table
Basic x1 x2 x3 x4 x5 x6 RHS
z 4 0 0 1 2 0 ?
x2 − 14 1 0 21 − 14 0 ?
x3 32 0 1 0 12 0 ?
x6 2 0 0 −2 1 1 ?

Use column rule to determine the RHS data except for z -row
new solution = inverse optimal × new right-hand side
New solution
      
1
x2 2
− 14
0 450 110
1
      
 x3  =  0 0   460  =  230 
   
   2 
x6 −2 1 1 400 −40

New solution is not feasible.

IE 310/GE 330 14
Lecture 11

Corresponding operating point x1 = 0, x2 = 110 and x3 = 230


Corresponding cost: substitute the above values in the objective

z = 3x1 + 2x2 + 5x3 = 3 × 0 + 2 × 110 + 5 × 230 = 1370

The table corresponding to these changes

Basic x1 x2 x3 x4 x5 x6 RHS
z 4 0 0 1 2 0 1370
x2 − 14 1 0 12 − 14 0 110
x3 32 0 1 0 12 0 230
x6 2 0 0 −2 1 1 −40

What shall we do to find the new optimal solution?

IE 310/GE 330 15
Lecture 11

Apply the dual simplex method

Basic x1 x2 x3 x4 x5 x6 RHS
z 4 0 0 1 2 0 1370
x2 − 14 1 0 21 − 14 0 110
x3 32 0 1 0 21 0 230
x6 2 0 0 −2 1 1 −40 ← leaving the basis

Which nonbasic variable will enter the basis?

IE 310/GE 330 16
Lecture 11

x4 enters since the only negative entry in x6-row is at that place


Basic x1 x2 x3 x4 x5 x6 RHS
z 4 0 0 1 2 0 1370
x2 − 14 1 0 21 − 14 0 110
x3 32 0 1 0 12 0 230
x6 2 0 0 −2 1 1 −40 ← leaving the basis

Use −2 as pivoting element and perform the standard basis exchange


Basic x1 x2 x3 x4 x5 x6 RHS
z 5 0 0 0 52 12 1350
x2 14 1 0 0 0 14 100
x3 32 0 1 0 12 0 230
x4 −1 0 0 1 − 12 − 12 20

The new optimal value is worse than the old optimal value!
The revenue is reduced by 20. There is a slack of 20 on machine 1!!!

IE 310/GE 330 17

You might also like