LINEAR OPTIMIZATION

MODELS
ž Optimization problems:
¡ Can be used to support and improve managerial
decision making
¡ Maximize or minimize some function, called the
objective function, and have a set of
restrictions known as constraints
¡ Can be linear or nonlinear

2
ž Typical applications:
¡ A manufacturer wants to develop a production schedule
and an inventory policy that will satisfy demand in
future periods and at the same time minimize the total
production and inventory costs
¡ A financial analyst would like to establish an investment
portfolio from a variety of stock and bond investment
alternatives that maximizes the return on investment
¡ A marketing manager wants to determine how best to
allocate a fixed advertising budget among alternative
advertising media such as web, radio, television,
newspaper, and magazine that maximizes advertising
effectiveness
¡ A company had warehouses in a number of locations.
Given specific customer demands, the company would
like to determine how much each warehouse should
ship to each customer so that total transportation costs
are minimized

3
ž Linear optimization models are also known
as linear programming
ž Linear programming:
¡ A problem-solving approach developed to help
managers make better decisions
¡ Numerous applications in today’s competitive
business environment
¡ For instance, GE Capital uses linear
programming to help determine optimal lease
structuring

4
 Problem Formulation
Mathematical Model for the Par, Inc. Problem
ž Illustration:
¡ Par, Inc. is a small manufacturer of golf
equipment and supplies whose management
has decided to move into the market for
medium- and high-priced golf bags. Par’s
distributor is enthusiastic about the new
product line and has agreed to buy all the golf
bags Par produces over the next three months

6
ž Operations involved in manufacturing a golf
bag:
¡ Cutting and dyeing the material
¡ Sewing
¡ Finishing (inserting umbrella holder, club
separators, etc.)
¡ Inspection and packaging

7
ž Table 7.1: Production Requirements Per Golf
Bag

8
ž Par’s production is constrained by a limited
number of hours available in each
department during the next three months:
Department Number of hours
Cutting and Dyeing 630
Sewing 600
Finishing 708
Inspection and 135
Packaging

ž Required profit contribution:

¡ Standard bag: $10/unit
¡ Deluxe bag: $9/unit

9
ž Develop a mathematical model of the Par,
Inc. problem to determine the number of
standard bags and the number of deluxe
bags to produce to maximize total profit
contribution
ž Problem Formulation
¡ Problem formulation or modeling:
Process of translating the verbal
statement of a problem into a
mathematical statement (or model)

10
ž General guidelines for problem
formulation:
¢ Understand the problem thoroughly
¢ Describe the objective
¢ Describe each constraint
¢ Define the decision variables
¢ Write the objective in terms of the decision
variables
¢ Write the constraints in terms of the
decision variables
ž Describe each constraint
Constraint Constraint

1 Number of hours of cutting and dyeing time

used must be less than or equal to the
number of hours of cutting and dyeing time
available.
2 Number of hours of sewing time used must
be less than or equal to the number of hours
of sewing time available.
3 Number of hours of finishing time used must
be less than or equal to the number of hours
of finishing time available.
4 Number of hours of inspection and packaging
time used must be less than or equal to the
number of hours of inspection and packaging
time available.
12
ž Define the decision variables
¡ S = number of standard bags
¡ D = number of deluxe bags
ž Write the objective in terms of the decision
variables
¡ If Par makes $10 for every standard and $9 for
every deluxe bag,
Total profit contribution = 10S + 9D = Objective
function
¡ Objective: Max 10S + 9D

13
ž Write the constraints in terms of the
decision variables
¡ Constraint 1:
Hours of cutting and Hours of cutting and
≤
dyeing time used dyeing time available
7
S + 1D ≤ 630
10
¡ Constraint 2:
Hours of sewing Hours of sewing
≤
time used time available
1 5
S + D ≤ 600
2 6

14
¡ Constraint 3:
Hours of finishing Hours of finishing
≤
time used time available
2
1S + D ≤ 708
3
¡ Constraint 4:
Hours of inspection and Hours of inspection and
≤
packaging time used packaging time available
1 1
S + D ≤ 135
10 4
¡ Nonnegativity constraints—based on the fact that the
number of standard or deluxe bags produced cannot
be negative
S ≥ 0 and D ≥ 0 or S, D ≥ 0

15
Mathematical Model for the Par, Inc. Problem
Mathematical model: a set of mathematical
relationships
Max 10S + 9D
subject to(s.t)
7
S + 1D ≤ 630 Cutting and dyeing
10
1 5
S + D ≤ 600 Sewing
2 6
2
1S + D ≤ 708 Finishing
3
1 1
S + D ≤ 135 Inspection and packaging
10 4
S, D ≥ 0

16
Mathematical Model for the Par, Inc.
Problem (contd.)

ž This is a linear programming model (or linear

program) because the objective function and all
constraint functions are linear functions of the
decision variables

ž Linear function: Mathematical function in which

each variable appears in a separate term and is
raised to the first power

17
 The Geometry of the Par, Inc. Problem
Solving Linear Programs with Excel Solver
ž To find the optimal solution to the problem
modeled as a linear program:

¡ The optimal solution must have the highest

objective function value
¡ The optimal solution must be a feasible
solution—a setting of the decision variables
that satisfies all of the constraints of the
problem
¡ Search over the feasible region—a set of all
possible solutions
¡ Find the solution that gives the best objective
function value
19
The Geometry of the Par, Inc. Problem

ž When only two decision variables, the functions

of variables are linear
ž If constraints are inequalities, the constraint cuts
the space in two
¡ The line and the area on one side of the line is
the space the satisfies that constraint
¡ These subregions are called half spaces
¡ The intersection of the half spaces make up the
feasible region
21
• To solve the Par, Inc. problem, we must find
the point in the feasible region that results in
the highest possible objective function value.
• An objective function contour is a set of
points that yield a fixed value of the
objective function.
• By choosing a fixed value of the objective
function, we may plot contour lines of the
objective function over the feasible region.
• As we move away from the origin we see
higher values of the objective function.
23
ž Based on the geometry of Figure 7.2, to solve a
linear optimization problem we only have to
search the extreme points of the feasible
region to find the optimal solution
ž Extreme points are found where constraints
intersect on the boundary of the feasible region
Solving Linear Programs with Excel Solver
ž The first step is to construct the relevant what-if
model
ž A what-if model for optimization allows the user
to try different values of the decision variables
and see:
¡ Whether that trial solution is feasible
¡ The value of the objective function for that
trial solution
ž Convey to Excel Solver the structure of the linear
optimization model

25
26
27
ž The optimal solution:
¡ To make 540 Standard bags and 252
Deluxe bags for a profit of $7,668
¡ Using all the cutting and dyeing time as
well as all finishing time, from cells
B19:B22 compared to C19:C22
ž The results are consistent with the results
obtained in Figures 7.1 and 7.2

28
29
ž A binding constraint is one that holds as an
equality at the optimal solution
ž The slack value for each less-than-or-equal-
to constraint indicates the difference
between the left-hand and right-hand values
for a constraint
ž By adding a nonnegative slack variable, we
can make the constraint equality
 Problem Formulation
Solution for the M&D Chemicals Problem
ž Illustration:
¡ Production requirements for M&D
Chemicals:
¢ The combined production for products A and
B must total at least 350 gallons
¢ Separately a major customer’s order for 125
gallons of product A must also be satisfied
¡ Processing time:
¢ Product A: 2 hours/gallon
¢ Product B: 1 hour/gallon
¢ For the coming month, 600 hours of
processing time are available

32
 ¡ Production cost: Product A: $2/gallon;
Product B: $3/gallon
¡ Objective: Minimizing the total
production cost
Problem Formulation
ž To find the minimum-cost production
schedule:
¡ Define the decision variables and the
objective function
Let A = number of gallons of product A
B = number of gallons of product B
¡ Objective function = 2A + 3B
33
ž Linear program for the M&D Chemicals problem:
Min 2A + 3B
s.t.
1A ≥ 125 Demand for product A
1A + 1B ≥ 350 Total production
2A + 1B ≤ 600 Processing time
A, B ≥ 0
ž A surplus variable tells how much over the
right-hand side the left-hand side of a greater-
than-or-equal-to constraint is for a solution

34
35
36
Alternative Optimal Solutions
Infeasibility
Unbounded
Alternative Optimal Solutions

ž Where the optimal objective function contour

line coincides with one of the binding
constraint lines on the boundary of the feasible
region
ž In these situations, more than one solution
provides the optimal value for the objective
function

38
ž Illustration using the Par, Inc. problem

¡ Original objective function: 10S + 9D

¡ Assume the profit for the standard golf bag
decreased to $6.30.
¡ Revised objective function: 6.3S + 9D
¡ The optimal solution occurs at two extreme
points:
¢ Extreme point 4 (S = 300, D = 420) and

¢ Extreme point 3 (S = 540, D = 252)

39
40
Infeasibility
ž Means no solution to the linear
programming problem
¡ No points satisfy all the constraints and the
nonnegativity conditions simultaneously
ž Graphically, a feasible region does not exist
ž Infeasibility occurs because:
¡ Management’s expectations are too high
¡ Too many restrictions have been placed on the
problem

41
42
ž Interpretation of Infeasibility for the Par,
Inc. problem
ž Let the management know that the
resources available are not sufficient to
make 500 standard bags and 360 deluxe bags
ž Provide details to the management on:
¡ Minimum amounts of resources that must be
available
¡ The amounts currently available
¡ Additional amounts that would be required to
accomplish this level of production

43
44
ž An infeasible problem when solved in Excel
Solver:
¡ Will return a message indicating that no
feasible solutions exists—indicating no solution
to the linear programming problem will satisfy
all constraints
¡ Careful inspection is necessary to identify why
the problem is infeasible
¡ One of the approaches is to drop one or more
constraints and re-solve the problem
¡ If we find an optimal solution for this revised
problem, then the constraint(s) that were
omitted, in conjunction with the others, are
causing the problem to be infeasible
45
Unbounded
ž The situation in which the value of the
solution
¡ May be made infinitely large—for a
maximization linear programming
¡ May be made infinitely small—for a
minimization linear programming
¡ Without violating any of the constraints

46
ž Illustration:
Consider the following linear program with
two decision variables,
X and Y:
Max 20X + 10Y
s.t.
1X ≥2
1Y ≤ 5
X, Y ≥ 0

47
48
ž Solving an unbounded problem using Excel
Solver:
¡ Returns a message “Objective Cell
values do not converge”
ž In linear programming models of real
problems:
¡ The occurrence of an unbounded
solution means that the problem has
been improperly formulated

49
Interpreting Excel Solver Sensitivity Report
ž Sensitivity analysis: The study of how the
changes in the input parameters of an
optimization model affect the optimal solution
ž It helps in answering the questions:
¡ How will a change in a coefficient of the
objective function affect the optimal
solution?
¡ How will a change in the right-hand-side
value for a constraint affect the optimal
solution?
ž The shadow price for a constraint is the
change in the optimal objective function value
if the right-hand side of that constraint is
increased by one

51
Interpreting Excel Solver Sensitivity Report
ž Consider the M&D chemicals problem:
A = number of gallons of product A
B = number of gallons of product B
Min 2A + 3B
s.t.
1A ≥ 125 Demand for
product A
1A + 1B ≥ 350 Total production
2A + 1B ≤ 600 Processing time
A, B ≥ 0
52
53
ž Classical sensitivity analysis:
¡ Based on the assumption that only one piece
of input data has changed
¡ It is assumed that all other parameters remain
as stated in the original problem
ž When interested in what would happen if
two or more pieces of input data are
changed simultaneously:
¡ The easiest way to examine the effect of
simultaneous changes is to make the changes
and rerun the model

54
Investment Portfolio Selection
Transportation Planning
Advertising Campaign Planning
ž The general notation for linear programs
uses the letter x with a subscript
ž In the Par, Inc. problem the decision
variables could be denoted as:
¡ 𝑥! = number of standard bags
¡ 𝑥" = number of deluxe bags
ž Advantage: Formulating a mathematical
model for a problem that involves a large
number of decision variables is much easier
ž Disadvantage: Not being able to easily
identify what the decision variables actually
represent in the mathematical model
56
ž Par, Inc. model using the general notation:
Max 10𝑥! + 9𝑥"
s.t.
7
𝑥! + 1 𝑥" ≤ 630 Cutting and dyeing
10
1 5
𝑥! + 𝑥"≤ 600 Sewing
2 6
2
1 𝑥! + 𝑥"≤ 708 Finishing
3
1 1
𝑥! + 𝑥"≤ 135 Inspection and
10 4
packaging
𝑥!, 𝑥"≥ 0

57
Investment Portfolio Selection
ž Portfolio selection problems involve
situations in which a financial manager must
select specific investments—for example,
stocks and bonds—from a variety of
investment alternatives
ž Objective: Maximization of expected return
or minimization of risk
ž Constraints: Restrictions on the type of
permissible investments, state laws,
company policy, and so on
58
illustration
ž Welte Mutual Funds, Inc., located in New York City, is
looking for investment opportunities for $100,000
ž The firm’s top financial analyst identified five
investment opportunities and projected their annual
rates of return

59
ž Welte investment guidelines:
¡ Neither industry (oil or steel) should
receive more than $50,000
¡ Amount invested in government bonds
should be at least 25 percent of the steel
industry investments
¡ The investment in Pacific Oil, the high-
return but high-risk investment, cannot
be more than 60 percent of the total oil
industry investment

60
ž Define the following decision variables:
𝑋!= dollars invested in Atlantic Oil
𝑋"= dollars invested in Pacific Oil
𝑋#= dollars invested in Midwest Steel
𝑋$= dollars invested in Huber Steel
𝑋% = dollars invested in government bonds
ž Specify the objective: Maximizing return
Max 0.073𝑋! + 0.103 𝑋" + 0.064 𝑋# + 0.075 𝑋$ + 0.045
𝑋%

61
ž Define the constraints:
Constraint 1: 𝑋! + 𝑋" + 𝑋# + 𝑋$ + 𝑋% = 100,000
Constraint 2: 𝑋! + 𝑋" ≤ 50,000
𝑋# + 𝑋$ ≤ 50,000
Constraint 3: 𝑋% ≥ 0.25(𝑋# + 𝑋$ )
Constraint 4: 𝑋" ≤ 0.60(𝑋! + 𝑋" )
Nonnegativity constraints: 𝑋! , 𝑋" , 𝑋# , 𝑋$ , 𝑋% ≥ 0

62
ž Linear programming model for the Welte
Mutual Funds investment problem:
Max 0.073𝑋! + 0.103 𝑋" + 0.064 𝑋# + 0.075 𝑋$ + 0.045
𝑋%
s.t.
𝑋! + 𝑋" + 𝑋# + 𝑋$ + 𝑋% = 100,000 Available funds
𝑋! + 𝑋" ≤ 50,000 Oil industry
maximum
𝑋# + 𝑋$ ≤ 50,000 Steel industry maximum
𝑋% ≥ 0.25(𝑋# + 𝑋$ ) Government bonds
minimum
𝑋" ≤ 0.60(𝑋! + 𝑋" ) Pacific Oil restriction
𝑋! , 𝑋" , 𝑋# , 𝑋$ , 𝑋% ≥ 0
63
64
Transportation Planning
ž Transportation problem arises in planning
for the distribution of goods and services
from several supply locations to several
demand locations
ž Quantity of goods available at each supply
location (origin) is limited
ž Quantity of goods needed at each of several
demand locations (destinations) is known
ž Objective: Minimize the cost of shipping
goods from the origins to the destinations
65
ž Illustration using Foster Generators problem
ž Involves the transportation of a product
from three plants to four distribution
centers
ž To determine how much of its production
should be shipped from each plant to each
distribution center

66
ž Production capacities over the next three-month
planning period for one type of generator:

ž The three-month forecast of demand for the

distribution centers:

67
68
ž Objective is to determine:
¡ Routes to be used
¡ Quantity to be shipped via each route
¡ Minimum total transportation cost
ž Let xij = number of units shipped from origin
i to destination j
where i = 1, 2, . . . , m and j = 1, 2, . . . , n

69
70
ž Supply constraints
x11 + x12 + x13 + x14 ≤ 5000 Cleveland supply
x21 + x22 + x23 + x24 ≤ 6000 Bedford supply
x31 + x32 + x33 + x34 ≤ 2500 York supply
ž Demand constraints
x11 + x21 + x31 = 6000 Boston demand
x12 + x22 + x32 = 4000 Chicago demand
x13 + x23 + x33 = 2000 St. Louis demand
x14 + x24 + x34 = 1500 Lexington demand

71
ž A 12-variable, 7-constraint linear programming
formulation of the Foster Generators transportation
problem:

72
73
Advertising Campaign Planning:
ž Designed to help marketing managers
allocate a fixed advertising budget to
various advertising media
ž Objective: Maximize reach, frequency, and
quality of exposure
ž Restrictions: Company policy, contract
requirements, and media availability

74
ž Illustration: Relax-and-Enjoy Lake
Development Corporation:
¡ Developing a lakeside community at a
privately owned lake
¡ Primary market includes all middle- and
upper-income families within approximately
100 miles of the development
¡ Employed the advertising firm of Boone,
Phillips, and Jackson (BP&J) to design the
promotional campaign

75
76
ž Problem Formulation:
¡ Budget: $30,000
¡ Restrictions imposed:
¢ At least 10 television commercials must
be used
¢ At least 50,000 potential customers must
be reached
¢ No more than $18,000 may be spent on
television advertisements
¡ The decision to be made is how many
times to use each medium
77
ž Define the decision variables:
¡ DTV = number of times daytime TV is used
¡ ETV = number of times evening TV is used
¡ DN = number of times daily newspaper is
used
¡ SN = number of times Sunday newspaper is
used
¡ R = number of times radio is used
ž Objective: Maximizing the total exposure
quality units for the overall
media selection plan

78
ž Linear programming model for the Relax-and-
Enjoy advertising campaign planning problem:
Max 65DTV + 90ETV + 40DN + 60SN + 20R Exposure quality

DTV ≤ 15
ETV ≤ 10 Availa-
DN ≤ 25 bility
of
SN ≤ 4
media
R ≤ 30

1500DTV + 3000ETV + 400DN + 1000SN +100R ≤ 30,000 Budget

79
ž Linear programming model for the Relax-
and-Enjoy advertising campaign planning
problem (contd.):
Televis
-ion
DTV + ETV ≥ 10
restric
1500DTV + 3000ETV ≤ 18,000
-tion

1000DTV + 2000ETV + 1500DN + 2500SN + 300R ≥ 50,000 Customers

reached
DTV, ETV, DN, SN, R ≥ 0

80
81
82
ž Illustration: Consider the Foster
Generators transportation problem
ž From Figure 7.14, the optimal solution:
x11 = 1000, x12 = 4000, x13 = 0, x14 = 0
x21 = 2500, x22 = 0, x23 = 2000, x24 = 1500
x31 = 2500, x32 = 0, x33 = 0, x34 = 0
ž Optimal cost: $39,500
ž For the revised model to be optimal, the
solution must give a total cost of $39,500

84
ž From Figure 7.14:
x13 = x14 = x22 = x32 = x33 = x34 = 0
ž If the sum of these variables is maximized
and if the optimal objective function value
of the revised problem is positive
ž A different feasible solution that is also
optimal is found

85
ž Revised model

86
87
ž In the original solution (Figure 7.14):
¡ Boston distribution center is sourced from all three
plants, whereas each of the other distribution centers is
sourced by one plant
¡ Hence, the manager in the Boston distribution center
has to deal with three different plant managers,
whereas each of the other distribution center managers
has only one plant manager
ž The alternative solution (Table 7.6) provides a more
balanced solution
¡ Managers in Boston and Chicago each deal with two
plants, and those in St. Louis and Lexington, which have
lower total volumes, deal with only one plant
¡ Because the alternative solution seems to be more
equitable, it might be preferred
ž Both the solutions give a total cost of $39,500
88
ž General approach to find an alternative optimal
solution to a linear program:
¡ Step 1: Solve the linear program
¡ Step 2: Make a new objective function to be
maximized; It is the sum of those variables that
were equal to zero in the solution from Step 1
¡ Step 3: Keep all the constraints from the original
problem; add a constraint that forces the original
objective function to be equal to the optimal
objective function value from Step 1
¡ Step 4: Solve the problem created in Steps 2 and
3; if the objective function value is positive, an
alternative optimal solution is found

89