Chapter 2: Linear Programming
¨ Introduction to LP
¨ Problem Formulation
¨ Solve LP using Graphical approach
¨ Four Special Cases
¨ Sensitivity Analysis
LP PROBLEM
¨ Problem: determine some variables to Maximize or
Minimize usually Profit/ Cost, called Objective
function.
¨ Constraints: are functions show resources limitation
of companies/ organizations. The problem is to find
solution that maximize profits (or minimize lost/cost)
in given constraints. Form of constraint functions
could be:
¨ Inequality (form or )
¨ Equality
¨ All Objective function and Constraint functions are 2
linear functions.
An Example of an LP model
¨ ABC company manufactures two products: Item A and Item B
¨ Each Item A:
¨ Sell for $27 and uses $19 worth of raw materials.
¨ Increase ABC’s variable labor/overhead costs by $5.
¨ Requires 2 hours of finishing labor.
¨ Requires 1 hour of carpentry labor.
¨ Each Item B:
¨ Sell for $21 and used $9 worth of raw materials.
¨ Increases ABC’s variable labor/overhead costs by $10.
¨ Requires 1 hour of finishing labor.
¨ Requires 1 hour of carpentry labor.
¨ Each week ABC can obtain:
¨ All needed raw material.
¨ Only 100 finishing hours.
¨ Only 80 carpentry hours.
¨ Demand for the Item B is unlimited.
¨ At most 40 Item A are bought each week.
¨ ABC wants to maximize weekly profit (revenues – costs).
ABC’s LP model
¨ x1 = number of item A produced each week
¨ x = number of item B produced each week
2
Max z = 3x1 + 2x2 (objective function)
(Weekly profit = weekly revenue – weekly raw material costs – the weekly variable costs = 3x1 + 2x2)
Subject to (s.t.)
Each week, no more than 100 hours of finishing time may be used.
2 x1 + x2 ≤ 100 (finishing constraint)
Each week, no more than 80 hours of carpentry time may be used. (x1 + x2 ≤ 80)
x1 + x2 ≤ 80 (carpentry constraint)
Because of limited demand, at most 40 item A should be produced. (x1 ≤ 40)
x1 ≤ 40 (constraint on demand for Item A)
x1 ≥0 (sign restriction)
x2 ≥ 0 (sign restriction)
2. Formulating LP Problems
WYNDOR GLASS CO.
¨ Glass products : windows and glass doors.
¨ Plant 1: Aluminum frames and hardware : Product 1
¨ Plant 2: Wood frame: Product 2
¨ Plant 3: The glass and assembles the products: Product 1 & 2.
¨ Product 1: An 8-foot glass door with aluminum framing
¨ Product 2: A 4x6 foot double-hung wood-framed window
¨ WYNDOR GLASS CO problem: Determine what the production rates
should be for the two products in order to maximize their total profit
The production rate = The number of batches of the
products to be produced / week
WYNDOR GLASS CO Data
Formulation as a Linear Programming
Problem
¨ x1 = number of batches of product 1 produced per week
¨ x2 = number of batches of product 2 produced per week
¨ Z = total profit per week (in thousands of dollars) from producing these
two products
¨ The objective function is
Maximize profit Z = $3x1 + $5x2
Subject to the restrictions:
3x1 + 2x2 18
2x2 12
x1 4 Linear Programming
Problem
x1 0
x 0.
Practice 1: Leather Limited
¨ Leather Limited manufactures two types of
leather belts: the deluxe model and the regular
model.
¨ Each type requires 1 square yard of leather.
¨ A regular belt requires 1 hour of skilled labor and a
deluxe belt requires 2 hours of skilled labor.
¨ Each week, 40 square yards of leather and 60 hours of
skilled labor are available.
¨ Each regular belt contributes $3 profit and each deluxe
belt $4.
¨ Solve an LP to maximize profit.
Solution
¨ The decision variables are:
¨ x1 = number of deluxe belts produced weekly
¨ x2 = number of regular belts produced weekly
¨ The appropriate LP is:
max z = 4x1 + 3x2
s.t. x1 + x2 ≤ 40 (leather constraint)
2x1 +x2 ≤ 60 (labor constraint)
x1, x2 ≥ 0
Dorian Auto Case
¨ Dorian Auto manufactures luxury cars and trucks.
¨ The company believes that its most likely customers are high-income women
and men.
¨ To reach these groups, Dorian Auto has embarked on an ambitious TV
advertising campaign and will purchase 1-mimute commercial spots on two
type of programs: comedy shows and football games
¨ Each comedy commercial is seen by 7 million high income women and 2
million high-income men and costs $50,000.
¨ Each football game is seen by 2 million high-income women and 12 million
high-income men and costs $100,000.
¨ Dorian Auto would like for commercials to be seen by at least 28 million high-
income women and 24 million high-income men.
¨ Use LP to determine how Dorian Auto can meet its advertising requirements at
minimum cost.
Dorian Auto Solution
¨ x1 = number of 1-minute comedy ads
¨ x2 = number of 1-minute football ads
¨ Objective functions is
min z = 50 x1 + 100x2
(to minimize total advertising cost.)
¨ Subject to:
¨ Commercials must reach at least 28 million high-income
women. (7x1 + 2x2 ≥ 28)
¨ Commercials must reach at least 24 million high-income
men. (2x1 + 12x2 ≥ 24)
¨ The sign restrictions are necessary, so x , x ≥ 0.
1 2
3. Graphical Solution
The graphical method works only when there are two
decision variables, but it provides valuable insight into how
larger problems are structured
¨ Graphical Representation of Constraints
• Isoprofit-line method
• Corner points method
Graphical Representation of Constraints
x2
Fig 1. Shaded area shows values of
9
( x1 , x2 ) allowed by
x1 0, x2 0, x1 4, 2 x2 12 8
Number of batches of product 2
7 2 x2 12
6
5
4 x1 4
3
2
1
0 1 2 3 4 5 6 7 x1
Number of batches of product 1
Graphical Representation of Constraints
x2
Fig 2. Shaded area shows the set of 9 3 x1 2 x2 18
permissible values of (x1, x2), called the 8
feasible region.
Number of batches of product 2
7 2 x2 12
6
5
4 x1 4
Feasible
Feasible
3 region
region
2
1
0 1 2 3 4 5 6 7 x1
Number of batches of product 1
3.1. Isoprofit- line method
Fig 3. The value of (x1 , x2 ) that x2
maximizes 3x1 5x2is (2, 6).
99
8
Z 36 3x1 5 x2 8
77
(2,6)
66
55
Z 20 3 x1 5 x2 44
33
Z 10 3 x1 5 x2 22
11
0 1 2 3 4 5 6 7 x1
Solving LP graphically by Isoprofit- line method
3 1 x2
x2 x1 Z Optimal solution
5 5
99
8
Z 36 3x1 5 x2 8 I ( x1 2, x 2 6)
77
66
55 Max Profit: Z=3*2+5*6=36
Z 20 3 x1 5 x2 44
33
Z 10 3 x1 5 x2 22
11
0 1 2 3 4 5 6 7 x1
Optimal Solution Structure
x2
99 Binding constraints
A constraint is said to be 88
binding if it holds with 77
I ( x1 2, x 2 6)
equality at the optimum 66
solution. 55
44 Z 3x1 5 x2
33
Other constraints are
22
non-binding 11
0 1 2 3 4 5 6 7 x1
Remarks:
Isoprofit lines (Z): parallel (same slope).
Optimal solution (any points): intersected (touched) by the feasible
region and isoprofit line that defines the largest Z-value.
Find the objective is to minimize Z: try Z-values tend to decrease
Find the objective is to maximize Z: try Z-values tend to increase
Activity 3
¨ Solve the ABC’s model by the graphical method (iso-line):
Max z = 3x1 + 2x2
(1)
Subject to (s.t.)
2 x1 + x2 ≤ 100
(2)
x1 + x2 ≤ 80 (3)
x1 ≤ 40 (4)
x1 ≥0
(5)
x ≥0 (6)
3.2. Corner-point method
Objective: Maximize Profit Z = 3X1 + 5X2
¨ The mathematical theory in LP shows that
the optimal solution must lie at one corner
point, or extreme point, of the feasible region
x2
• At A=(0,0): Profit = 0
• At B=(0,6): 99
88
Profit = 3(0) + 5(6) = 30 77
B=(0,6)
66 C=(2,6)
• At C=(2,6):
55
Profit = 3(2) + 5(6) = 36 44
33 D=(4,3)
• At D=(4,3): 22
11
E=(4,0)
Profit = 3(4) + 5(3) = 27
A=(0,0)
0 1 2 3 4 5 6 7 x1
• The optimal solution is C = (2,6)
which obtains profit of 36
4. Special cases in LP
¨ Infeasible solutions
¨ Unbounded solutions
¨ Redundant constraints
¨ Multiple optimal solutions
Case 1: Infeasible
Infeasible solutions occurred when:
• Have conflicting constraints ; or
• No solution satisfy all constraints; or
• Can not build the feasible solutions
region.
x2
Example:
99 3x1+5x2=50
Maximize Z = 3X1 + 5X2 88
77
(2,6)
Number of batches of product 2
St: 66
55
3X1 + 5X2 50 44
33
3X1 + 2X2 18 22
11
2X2 12
0 1 2 3 4 5 6 7 x1
X1 4 Number
Numberofofbatches
batchesofofproduct
product11
X1 0
X2 0
Case 2: Unbounded Solutions
¨ When value of Objective function
approached to infinity we said that the
problem is unbounded or missing one or
more constraints;
¨ The LP did not provide a finite solutions,
this implies the objective function
approaches to infinity without violating
any constraint.
Open ended problem
(4,∞ ), Z=∞
¨ Example: x2
Maximize Z = 3X1 + 5X2 9
8 (4,8), Z=52
St:
Number of batches of product 2
7
X1 0 6 (4,6), Z=42
5
X2 0 4 (4,4), Z=32
3
X1 4 2 (4,2), Z=22
1
0 1 2 3 4 5 6 7 x1
Number of batches of product 1
Case 3: Redundancy of Constraints
¨ A redundant constraint is a constraint that
will not affect to the solution space
¨ In reality, this will usually happens when
number of constraints and number of
variables are very large.
x2
Example:
9
Maximize Z = 3X1 + 5X2 8
7
(2,6)
Number of batches of product 2
St: 6
5
3X1 + 2X2 18 4
3 x1=7
2X2 12 2
1
X1 4
0 1 2 3 4 5 6 7 x1
X1 7 Number of batches of product 1
X1 0
X2 0
Case 4: Multiple solutions
¨ When objective function and one
constraint have the same slope we will
faced with the case multiple optimal
solutions
• Example: x2
Maximize Z = 3X1 + 2X2
St: 3 X1 +2X2 18 Z 18 3x1 2 x2 9
8
Number of batches of product 2
2 X2 12 7 3 x1 2 x2 18
6
Every point in the
X1 0 5
darker
4
X2 0 Feasible segment line is
3 region Optimal, Z=18
2
X1 4
1
0 1 2 3 4 5 6 7 x1
Number of batches of product 1
5. Sensitivity Analysis
Find sensitive parameters those x2
making small change in its value
(adding Δ in this case) changes the 9 3x1 2 x2 18
optimal solution 8 Z 36 3x1 5 x2
7
(2,6) 2 x2 12
6
Number of batches of product 2
5 x1 4
4
3
2
1
0 1 2 3 4 5 6 7 x1
Number of batches of product 1
Sensitivity Analysis on the Contsraints
¨ On constraint 1:
If we increase RHS by 1: 3X1 + 2X2 18+1
The new optimal solution is: X1= 7/3 , X2 =6 and the new
Z’ = 37 . The net change in profit is $1. This is called
shadow price (marginal value or dual price) associated
with constraint 1.
• On constraint 2:
If we increase RHS by 1: 2X2 13, X1= 5/3 ,
X2 =13/2
The net change in profit is $1.5. The shadow price associated
with constraint 2 is $1.5.
• On constrain 3:
X1 4 : not binding constraint x2
9 3x1 2 x2 18
8
Z 36 3x1 5 x2
7
(2,6) 2 x2 12
Number of batches of product 2
6
5 x1 4
4
3
Shadow price =0 2
1
0 1 2 3 4 5 6 7 x1
Number of batches of product 1
Remarks
- Consider binding constraints
- Slightly change RHS of each binding constraint
- Shadow price = Z(new)- Z(old):
- Shadow price > 0: if RHS changes, the optimal solution change,
that RHS is sensitive parameter
- Shadow price = 0: if RHS changes, the optimal solution does
not change. Thus that RHS is not sensitive parameter