Chapter 2: Linear Programming

¨Introduction about LP
¨Problem Formulation
¨Solve LP using Graphical approach
¨Four Special Cases
¨Sensitivity Analysis
Notes: From now on, in every lecture, all students must participate
into the activities of the class. The results of these activities will be
scored and allocated to the assigments/homeworks at the end of
the semester.
 1. Introduction
¨ Many decisions in management are related with the
best usage resources of organizations.
¨Manager makes Decisions in order to satisfy Objectives,
Goals of organizations.
¨Resources: Materials, Machines, Man, Money, Time,
Space.
¨Linear Programming (LP) is a mathematical method that
helps managers to make decision related with
Resources Allocation. (references about Nobel laureate:
Kantorovich)
¨Extensively using computer. 3
 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 s
solution that maximize profits (or minimize lost/cost)
in given constraints.
Form of constraint functions could be:
¨Inequality (form  or )
¨Equality 4
¨All Objective function and Constraint functions are
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
¨ x2 = number of item B produced each week

Max z = 3x1 + 2x2 (objective function)

(Weekly profit = weekly revenue – weekly raw material costs – the weekly variable costs = 3 x1 + 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.
 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
7 2 x2  12

Number of batches of product 2

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. 7 2 x2  12

Number of batches of product 2

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
Z  20  3 x1  5 x2 44
Max Profit: Z=3*2+5*6=3
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 equality 77
I  ( x1  2, x 2  6)
at the optimum solution. 66
55
44 Z  3x1  5 x2
Other constraints are non-
33
binding
22
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

 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  188
9
Z  18  3x1  2 x2

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 3
Feasible segment line is
region Optimal, Z=18
X1  4 2
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 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 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
 Summary

¨The Geometry of the Feasible Region

¨ Graphing the constraints
¨Finding an optimal solution
¨ Graphical method
¨ Searching all the extreme points

¨Sensitivity Analysis
¨ Changing the RHS
¨ Can we change the objective’s coefficients ?