Ch06 Linear Programming

SUPPLEMENT TO
CHAPTER 6
Linear Programming
SUPPLEMENT OUTLINE Solved Problems, 22 Learning Objectives

Introduction and Linear Discussion and Review
After completing this supplement,
Programming Model, 2 Questions, 25
you should be able to:
Graphical Solution Method, 5 Problems, 25
Formulate a linear
LO 1
Computer Solutions, 14 Mini-Case: Airline Fuel
Management, 34 programming model from
Sensitivity Analysis, 17 the description of a problem.
Key Terms, 22 Mini-Case: Son Ltd., 35
LO 2 Solve two-variable linear
programming problems
using the graphical solution
method.
LO 3 Solve linear Programming
problems using Excel’s

Solver.
LO 4 Do sensitivity analysis on
the solution of a linear

programming problem.
2 PART THREE System Design
INTRODUCTION AND LINEAR

LO 1
PROGRAMMING MODEL
Linear programming is a powerful quantitative tool used to obtain optimal solutions to
problems that involve requirements or limitations for resources, such as available mate-
rial, budget, labour, and machine time. These problems are referred to as constrained
optimization problems. Linear programming has been successful in optimizing numerous
decision-making problems including:
• Determining optimal blends of animal feed and gasoline ingredients
• Developing optimal production schedules
• Identifying the optimal mix of products to produce in a factory
On average, costs have been reduced about 5 percent due to increased efficiency.
A Linear programming model is a mathematical representation of a constrained optimi-
zation problem. Three components provide the structure of a linear programming model:
1. Decision variables
2. Objective
3. Constraints
decision variables Choices Decision variables represent choices in terms of amounts of either inputs to use or
in terms of amounts of either outputs to produce. Each unknown quantity is assigned a decision variable (e.g., x1, x2).
inputs or outputs. A Linear programming model requires a single goal or objective. An objective is either
maximized and minimized. A maximized objective might involve profit, revenue, etc.
objective function Conversely, a minimized objective might involve cost, time, or distance. The objective
Mathematical expression function is a mathematical expression involving the decision variables that represent the
involving the decision variables objective, e.g., profit, cost.
representing the objective, e.g., Constraints represent the requirements or limitations that restrict the choices available.
profit, cost. The three types of constraints are less than or equal to (≤), greater than or equal to (≥),
and simply equal to (=). A ≤ constraint implies an upper limit on the amount of some re-
constraints Represent the source (e.g., machine hours, labour hours, material) available for use. A ≥ constraint
requirements or limitations that specifies a minimum that must be achieved in the final solution (e.g., the product must
restrict the available choices. contain at least 50 percent real orange juice). The = constraint is more restrictive in the
sense that it specifies exactly what a mathematical expression of decision variables should
equal (e.g., make 200 units of product A). A linear programming model can consist of one
or more constraints. The constraints of a given problem define the set of all feasible com-
feasible solution space The binations of values of decision variables; this set is referred to as the feasible solution
set of all feasible combinations space. A linear programming solution technique is designed to search the feasible solu-
of values of decision variables tion space for the combination of values of decision variables that will yield an optimum
as defined by the constraints. (i.e., best possible value) in terms of the objective function.
In a linear programming model, symbols (e.g., x1, x2) represent the decision variables
parameters Coefficients and numerical values, called parameters, represent their coefficients in the objective
of decision variables in function and constraints. The parameters are input to the model and have fixed values.
the objective function and Example S-1 illustrates the components of a linear programming model.
constraints. First, the model lists and defines the decision variables. In this case, they are quantities
of three different products that might be produced.
Example S-1 Decision variables x1 = Quantity of product 1 to produce

x2 = Quantity of product 2 to produce
x3 = Quantity of product 3 to produce
Maximize 5x1 + 8x2 + 4x3 (profit) (Objective function)
Subject to
labour 2x1 + 4x2 + 8x3 ≤ 250 hours
Material 7x1 + 6x2 + 5x3 ≤ 100 kg (Constraints)
Product 1 x1 ≥ 10 units
x1, x2, x3 ≥ 0 (Non-negativity constraints)
SUPPLEMENT TO CHAPTER 6 Linear Programming 3
Next, the model states the objective function. It includes every decision variable in the
model and the profit contribution (per unit) of each decision variable. Thus, product 1 has
a profit of $5 per unit. The profit from product 1 for a given solution x1 will be 5 times the
value of x1; the total profit from all products will be the sum of the individual product
profits. Thus, if x1 = 10, x2 = 0, and x3 = 6, the value of the objective function would be:
5(10) + 8(0) + 4(6) = 74
The objective function is followed by a list (in no particular order) of four constraints.
Each constraint has a right-side value (e.g., the labour constraint has a right-side value of
250) that indicates the amount of the constraint and a relation sign that indicates whether
that amount is a maximum (≤), a minimum (≥), or an equality (=). The left side of each
constraint consists of the variables and a coefficient for each variable that indicates how
much of the right-side quantity one unit of the decision variable represents. For instance, for
the labour constraint, one unit of product 1 will require two hours of labour. The sum of the
values on the left side of each constraint represents the amount of that constraint used by a
solution. Thus, if x1 = 10, x2 = 0, and x3 = 6, the amount of labour used would be:
2(10) + 4(0) + 8(6) = 68 hours
Because this amount does not exceed the quantity on the right-hand side of the con-
straint, it is feasible for the labour constraint.
Note that the third constraint refers to only a single variable; x1 must be at least 10 units.
X1's coefficient is, in effect, 1, although that is not shown.
Finally, there are the non-negativity constraints. These are listed on a single line but
they reflect that none of the three decision variables is allowed to have a negative value.
In order for a linear programming model to be used effectively, certain assumptions
must be satisfied. These are:
1. Linearity: the impact of decision variables is linear in constraints and in the objective
function.
2. Divisibility: non-integer values of decision variables are acceptable.
3. Certainty: values of parameters are known and are constant.
Formulating a Ratio and a Percentage Constraint

1. A ratio. For example, “the ratio of x1 to x2 must be at least 3 to 2.” To formulate this
requirement, begin by setting up the ratio:
x1 3
≥
x2 2
But this is not a linear constraint. Cross multiply to obtain:

2x1 ≥ 3x2
Although this is a linear constraint, it is not yet in the standard form: all variables in a
constraint must be on the left side of the inequality (or equality) sign, leaving only a con-
stant on the right side. To achieve this, we must subtract the variable amount that is on the
right side from both sides. That yields:
2x1 − 3x2 ≥ 0
[Note that the direction of the inequality remains the same.]
2. A percentage. For example, “x1 cannot be more than 20 percent of the mix.” Suppose
that the mix consists of variables x1, x2, and x3. In mathematical terms, this constraint
would be:
x1 ≤ .20(x1 + x2 + x3)
Again, all variables must appear on the left side of the constraint. To accomplish that,
we can expand the right side and then subtract the result from both sides. Thus,
x1 ≤ .20x1 + .20x2 + .20x3
Subtracting yields:
.80x1 − .20x2 − .20x3 ≤ 0
Example S-2 A small company that assembles computers is about to start production of two new
types of personal computers. Each type will require assembly time, inspection time, and
storage space. The amounts of each of these resources that can be devoted to the produc-
tion of these computers are limited. The manager of the firm would like to determine the
quantity of each computer to produce in order to maximize the profit generated by sales
of these computers. In order to develop a suitable model of the problem, the manager has
met with the design and manufacturing personnel. As a result of those meetings, the man-
ager has obtained the following information:
Type 1 Type 2
Profit per unit $60 $50
Assembly time per unit 4 hours 10 hours
Inspection time per unit 2 hours 1 hour
Storage space per unit 3 cubic feet 3 cubic feet
The manager also has acquired information on the availability of company resources.
These daily amounts are:
Resource Amount Available Per Day

Assembly time 100 worker hours
Inspection time 22 worker hours
Storage space 39 cubic feet
The manager met with the firm’s marketing manager and learned that demand for the
computers was such that whatever combination of these two types of computers is pro-
duced, all of the output can be sold at the determined prices.
In terms of meeting the assumptions, it would appear that the relationships are linear:
The contribution to profit per unit of each type of computer and the time and storage
space per unit of each type of computer is the same regardless of the quantity produced.
Therefore, the total impact of each type of computer on the profit and each constraint is a
linear function of the quantity of that variable. There may be a question of divisibility
because, presumably, only whole units of computers will be produced. However, because
this is a recurring process (i.e., the computers will be produced daily, a noninteger solu-
tion such as 3.5 computers per day will result in 7 computers every other day), this does
not seem to pose a problem. Regarding the question of certainty, the manager could be
questioned to determine whether the values shown for assembly times, etc., are known
with certainty. Later in this chapter supplement, under “Sensitivity Analysis,” the effect of
uncertainty in parameters is studied.
Because we have concluded that linear programming is appropriate, let us now turn
our attention to constructing a linear programming model of the problem. First, we must
define the decision variables. The decision variables are the quantity of each type of com-
puter to produce. Thus,
x1 = quantity of type 1 computer to produce per day
Next, we can formulate the objective function. The profit per unit of type 1 computer
is listed as $60, and the profit per unit of type 2 is $50. So the objective function is
Maximize Z = 60 x1 + 50 x2
where Z is the value of the objective function, given values of x1 and x2. (In practice,
the objective function often is written without the Z.)
Now, we formulate the constraints. There are three resources with limited availability:
assembly time, inspection time, and storage space. The fact that availability is limited
means that these constraints will all be ≤ constraints. Suppose we begin with the assem-
bly constraint. Type 1 computer requires 4 hours of assembly time per unit, whereas type
2 computer requires 10 hours of assembly time per unit. Therefore, with a limit of 100
hours available per day, the assembly constraint is:
4x1 + 10x2 ≤ 100 hours
Similarly, each unit of type 1 computer requires 2 hours of inspection time, and each
unit of type 2 computer requires 1 hour of inspection time. With 22 hours available per
day, the inspection constraint is:
2x1+ 1x2 ≤ 22 hours
(Note: The coefficient 1 for x2 need not be shown. Thus, an alternative form for
this constraint is: 2x1+ x2 ≤ 22.) The storage constraint is determined in a similar
manner:
3x1 + 3x2 ≤ 39 cubic feet
The non-negativity constraints are:
x1, x2 ≥ 0
In summary, the mathematical model of the computer problem is:
Maximize 60x1 + 50x2
Subject to
Assembly 4x1 + 10x2 ≤ 100 hours
Inspection 2x1 + 1x2 ≤ 22 hours
Storage 3x1 + 3x2 ≤ 39 cubic feet
x1, x2 ≥ 0
Once you have formulated a model, the next task is to solve it (i.e., find the best fea-
sible siolution). The following sections describe two approaches to problem solution:
graphical solution method and computer solutions.
GRAPHICAL SOLUTION METHOD

LO 2
The graphical solution method involves drawing the feasible solution space and contours graphical solution Solution
of the objective function on a paper, and finding the optimal solution by observation. This method that involves drawing
method only applies to two-variable linear programs. the feasible solution space
Consider the linear programming model described in Example S-2 above. and contours of the objective
function on paper, and
finding the optimal solution
Plotting the Constraints and Determining the by observation. This method
Feasible Solution Space applies only to two-variable
Begin by drawing x1 as x-axis and x2 as y-axis on a coordinate system. Then, represent the linear programs.
non-negativity constraints as the top-right quadrant, as in Figure 6S-1.
Figure 6S-1 Quantity of x2

type 2 computer
Graph showing the non-
negativity constraints
Area of
feasibility
Nonnegativity
x1 = 0
constraints
x2 = 0
0 x1 Quantity of
type 1 computer
An equality constraint is represented by a straight line on the coordinate system that is

simple to draw (see steps 2 and 3 below). The procedure for representing an inequality
constraint is more involved:
1. Replace the inequality sign with an equal sign. This transforms the constraint into an
equation that is represented by a straight line. Plot this line according to steps 2 and 3
below.
2. Determine where the line intersects each axis.
a. To find where it crosses the x2 axis, set x1 equal to zero and solve the equation for
the value of x2.
b. To find where it crosses the x1 axis, set x2 equal to zero and solve the equation for
the value of x1.
3. Mark these intersections on the axes, and connect them with a straight line. (Note: If a
constraint has only one variable, it will be a vertical line if the variable is x1, or a hori-
zontal line if the variable is x2.)
4. Indicate by shading (or by an arrow) whether the inequality is greater than or less than.
(A general rule to determine which side of the line satisfies the inequality is to pick a
point that is not on the line, such as the origin (0,0), plug it in the left side of the con-
straint and see whether the left side is greater than or less than the constraint’s right-
hand-side RHS.)
5. Repeat steps 1–4 for each constraint.
Consider the assembly time constraint:
4x1 + 10x2 ≤ 100
Removing the inequality portion of the constraint produces this equation:
4x1 + 10x2 = 100
Next, identify the points where the line representing the above equation intersects each
axis, as step 2 describes. Thus, with x2 = 0, we find
4x1 + 10(0) = 100
Solving, we find that 4x1 = 100, so x1 = 25. Similarly, we can solve the equation for x2
when x1 = 0:
4(0) + 10x2 = 100
Solving for x2, we find x2 = 10.
Thus, we have two points: (x1 = 0, x2 = 10), and (x1 = 25, x2 = 0). We can now add
this line to our graph of the non-negativity constraints by connecting these two points
(see Figure 6S-2).
x2 Figure 6S-2
Plot of the first constraint
(assembly time) as an equation
10
4x
1 +1
0x
2 =1
00
0 25 x1
Next, we must determine which side of the line represents points that are less than 100.
To do this, we can select a test point that is not on the line, and substitute the x1 and x2
values of that point into the left side of the equation of the line. If the result is less than
100, this tells us that all points on that side of the line are less than the value of the line
(e.g., 100). Conversely, if the result is greater than 100, this indicates that the other side of
the line (relative to the point chosen) represents the set of points that will yield values that
are less than 100. A relatively simple test point to use is the origin (i.e., x1 = 0, x2 = 0).
Substituting these values into the left side of the equation yields
4(0) + 10(0) = 0
Obviously 0 is less than 100. Hence, the side of the line containing the origin repre-
sents the “less than” area (i.e., the feasible region for the first constraint).
The feasible region for this constraint and the nonnegativity constraints then becomes
the shaded portion shown in Figure 6S-3.
For the sake of illustration, suppose we try one other point, say x1 = 10, x2 = 10. Sub-
stituting these values into the assembly equation yields
4(10) + 10(10) = 140
Clearly this is greater than 100. Therefore, all points on this side of the line are greater
than 100 (see Figure 6S-4).
Continuing with the problem, we add the two remaining constraints to the graph. For
the inspection constraint:
1. Convert the constraint into an equation by replacing the inequality sign with an equa
lity sign:
2x1 + 1x2 ≤ 22 becomes 2x1 + 1x2 = 22
x2
x2
10
Ass 10 • 10,10
emb 4x
ly ti 1 +1
me 0x
Feasible 2 =1
00
region
0 25 x1
0 10 25 x1
Figure 6S-3
The feasible region for the first constraint and the Figure 6S-4
non-negativity constraints The point (10, 10) is above the constraint line
2. Set x1 equal to zero and solve for x2:

2(0) + 1x2 = 22
Solving, we find x2 = 22. Thus, the line will intersect the x2 axis at 22.
3. Next, set x2 equal to zero and solve for x1:
2x1 + 1(0) = 22
Solving, we find x1 = 11. Thus, the other end of the line will intersect the x1 axis at 11.
4. Add the line to the graph and determine the side of the line that represents the con-
straint (see Figure 6S-5). Note that the area of feasibility for the inspection constraint
is also below the line. However, the area feasible to both inspection and assembly
constraints is the intersection of the area feasible to assembly and the area feasible to
inspection.
The storage constraint is handled in the same manner:
1. Convert it into equality:
3x1 + 3x2 = 39
3(0) + 3x2 = 39
Solving, we find x2 = 13. Thus, x2 = 13 when x1 = 0.
3x1 + 3(0) = 39
Solving, we find x1 = 13. Thus, x1 = 13 when x2 = 0.
4. Add the line to the graph and determine the side of the line that represents the con-
straint (again it is below the line because of ≤ constraint) (see Figure 6S-6).
The feasible solution space is the set of all points that satisfies all constraints. The
shaded area shown in Figure 6S-6 is the feasible solution space for our problem.
The next step is to determine which point in the feasible solution space will produce
the optimal value of the objective function. This determination is made by plotting objec-
tive function contours.
x2 x2
22 22
Ins
Ins
pe
pe
ct
Feasible for inspection

ctio
ion
but not for assembly St

n
13 or
ag
e
10 Feasible for assembly 10
but not for inspection
Feasible Ass
Ass emb
Feasible for emb solution ly
ly
both assembly space
and inspection
0 11 25 x1 0 11 13 25 x1
Figure 6S-5 Figure 6S-6

Partially completed graph, showing the assembly, Completed graph of the computer problem showing
inspection, and non-negativity constraints all the constraints and the feasible solution space
Plotting Objective Function Lines and

Determining the Optimal Point
Plotting an objective function line involves the same logic as plotting a constraint line:
determine where the line intersects each axis. But the objective function is not an equa-
tion. We can get around this by simply setting it equal to some number. Any number will
do, although one that is evenly divisible by both coefficients is desirable.
Recall that the objective function for the computer problem is 60x1 + 50x2. Suppose we
decide to set the objective function equal to 300. That is,
60x1 + 50x2 = 300
We can now plot the line on our graph. As before, we can determine the x1 and x2 inter-
cepts of the line by setting one of the two variables equal to zero and solving for the other,
and then reversing the process. Thus, with x1 = 0, we have
60(0) + 50x2 = 300
Solving, we find x2 = 6. Similarly, with x2 = 0, we have
60x1 + 50(0) = 300
Solving, we find x1 = 5. This line is plotted in Figure 6S-7.
The $300 profit line is an isoprofit line; every point on the line (i.e., every combination
of x1 and x2 that lies on the line) will provide a profit of $300. We can see from the graph
that many combinations are both on the $300 profit line and are within the feasible solu-
tion space. In fact, considering noninteger as well as integer solutions, the possibilities
are infinite.
Suppose we now consider another profit line, say the $600 line. To do this, we set the
objective function equal to this amount. Thus,
60x1 + 50x2 = 600
Solving for the x1 and x2 intercepts yields these two points:
x1 intercept x2 intercept
x1 = 10 x1 = 0
x2 = 0 x2 = 12
This line is plotted in Figure 6S-8, along with the previous $300 line for purpose of
comparison.
x2 x2
22 22
Ins
Ins
pe
pe
ct
ctio
ion
13 13
n
St
St
12
or
o
ra
ag
ge
10 10
e
6 6
Pr
Pr
Pr
Ass Ass
of
of
of
emb emb
it
it
it
ly ly
=
=
=
$3
$6
$3
00
00
00
0 5 11 13 25 x1 0 5 10 11 13 25 x1

Computer problem with $300 profit line added Computer problem with profit lines of $300 and $600
Two things are evident in Figure 6S-8 regarding the profit lines. One is that the $600
line is farther from the origin than the $300 line; the other is that the two lines are parallel.
The lines are parallel because they both have the same slope. The slope is not affected by
the right side of the equation. Rather, it is determined solely by the coefficients 60 and
50. Regardless of the number we select for the value of the objective function, the result-
ing line will be parallel to these two lines. Moreover, if the number is greater than 600,
the line will be even farther away from the origin than the $600 line. If the number is less
than 300, the line will be closer to the origin than the $300 line. And if the number is be-
tween 300 and 600, the line will fall between the $300 and $600 lines. This knowledge
will help us in determining the optimal solution.
Consider a third line, one with the profit equal to $900. Figure 6S-9 shows that line
along with the previous two profit lines. As expected, it is parallel to the other two, and
even farther away from the origin. However, the line does not touch the feasible solution
space at all. Consequently, there is no feasible combination of x1 and x2 that will yield that
amount of profit. Evidently, the maximum possible profit is an amount between $600 and
$900.
We could continue to select profit lines in this manner, and eventually, determine an
amount that would yield the greatest feasible profit. However, there is a much simpler
alternative. We can plot just one profit line, say the $300 line. We know that all other
profit lines will be parallel to it. Consequently, by moving this one line parallel to itself
we can represent other profit lines. We also know that as we move away from the origin,
the profits get larger. What we want to know is how far the line can be moved out from
the origin and still be touching the feasible solution space, and the values of the decision
variables at that point (i.e., the optimal solution). Locate this point on the graph by plac-
ing a ruler along the $300 line (or any other convenient line) and sliding it away from the
origin, being careful to keep it parallel to the line. This approach is illustrated in Figure
6S-10.
Once we have determined where the optimal solution is in the feasible solution space,
we must determine the values of the decision variables at that point. Note that the optimal
solution is at the intersection of the inspection boundary line and the storage boundary
line (see Figure 6S-10). In other words, the optimal combination of x1 and x2 must satisfy
both boundary (equality) conditions. We can determine those values by solving the two
equations simultaneously. The equations are
Inspection 2x1 + 1x2 = 22
Storage 3x1 + 3x2 = 39
x2 x2
22 22
Ins
pe
18
Ins
ct
ion
pe
ctio
St
n
13 or 13
12 ag (last line)
e
10 10
Optimal
$9
00
solution
6 6
Ass Ass
$6
$3
emb St emb
or
00
00
$3
ly ag ly
00
0 5 10 11 13 15 25 x1 0 5 11 13 25 x1

Computer problem with profit lines of $300, $600, Finding the optimal solution to the computer
and $900 problem
The idea behind solving two simultaneous equations is to algebraically eliminate one
of the unknown variables (i.e., to obtain an equation with a single unknown). This can be
accomplished by multiplying the coefficients of one of the equations by a fixed amount
and then adding (or subtracting) the modified equation from the other. (Occasionally, it is
easier to multiply each equation by a fixed amount.) For example, we can eliminate x2 by
multiplying the inspection equation by 3 and then subtracting the storage equation from
the modified inspection equation. Thus,
3(2x1 + 1x2 = 22) becomes 6x1 + 3x2 = 66
Subtracting the storage equation from this produces
6x1 + 3x2 = 66
− (3x1 + 3x2 = 39)
3x1 + 0x2 = 27
Solving the resulting equation yields x1 = 9. The value of x2 can be found by substitut-
ing x1 = 9 into either of the original equations or the modified inspection equation. Sup-
pose we use the original inspection equation. We have
2(9) + 1x2 = 22
Solving, we find x2 = 4.
Hence, the optimal solution to the computer problem is to produce 9 type 1 computers
and 4 type 2 computers per day. We can substitute these values into the objective function
to find the optimal profit:
$60(9) + $50(4) = $740
Hence, the last profit line—the one that would last touch the feasible solution space as
we moved away from the origin parallel to the $300 profit line—would be the line where
profit is $740.
In this problem, the optimal values for both decision variables are integers. This will
not always be the case for the optimal solution to the linear programming model; one or
both of the decision variables may turn out to be non-integer. In some situations
non-integer values would be of little consequence. This would be true if the decision vari-
ables were measured on a continuous scale, such as the amount of fuel oil needed or if the
profit contribution per unit was small, as with the number of nails or ball bearings to
make. In some cases, the answer can simply be rounded down (maximization problems)
or up (minimization problems) with very little impact on the objective function. Here, we
assume that non-integer answers are acceptable as such.
Let’s review the procedure for finding the optimal solution using the graphical solution
method:
1. Draw the x1 and x2 axes.
2. Graph the constraints. Identify the feasible solution space.
3. Set the objective function equal to some number so that it crosses the feasible solution
space. Plot the line. Slide it in parallel in the direction of increasing objective function until
just before the line leaves the feasible solution space. This point is the optimal solution.
4. Determine which two constraints intersect there. Solve their equations simultaneously
to obtain the values of the decision variables at the optimum.
5. Substitute the values obtained in the previous step into the objective function to deter-
mine the value of the objective function at the optimum.
Redundant Constraints redundant constraint
In some cases, a constraint does not form a unique boundary of the feasible solution A constraint that does not
space. Such a constraint is called a redundant constraint. Two such constraints (x2 ≤ 25 form a unique boundary of the
and 2x1 + 3x2 ≤ 60) are illustrated in Figure 6S-11. Note that a constraint is redundant if feasible solution space.
its removal would not alter the feasible solution space.
Figure 6S-11 x2 x2 = 25
25
Examples of redundant
constraints
20 Redundant
2x
1 +3
x
2 =6
0
2x 1
+ x2
=2
0
Feasible
solution
space
0 10 30 x1
When a problem has a redundant constraint, at least one of the other constraints in the
problem is more restrictive than the redundant constraint.
Solutions and Corner Points
The feasible solution space is a polygon, a convex shape with straight-line sides (convex
means that any point between two feasible points is also feasible). Moreover, the solution
to any problem will be at one of the corner points (intersections of constraints) of the
polygon (except if there are multiple optimal solutions; described below). It is possible to
determine the coordinates of each corner point of the feasible solution space, and use
those values to compute the value of the objective function at those points. Comparing the
values of the objective function at the corner points and identifying the best one (e.g., the
maximum value) is another way to identify the optimal solution. For problems that have
more than two decision variables, a variation of this approach (called Simplex method) is
used to find the optimal solution.
In some instances, the objective function lines will be parallel to one of the constraint
lines that form the boundary of the feasible solution space. When this happens, every
Figure 6S-12 x2
Some LP problems have

multiple optimal solutions
Optimal line
segment
O
bj
ec
tiv
e
fu
nc
tio
n
0 x1
combination of x1 and x2 on the segment of the constraint line that touches the feasible
solution space represents an optimal solution. Hence, there are multiple optimal solutions multiple optimal
to the problem. Figure 6S-12 illustrates an objective function line that is parallel to a con- solutions When the objective
straint line. function line is parallel to
a constraint line that forms
Minimization a boundary of the feasible
Graphical solution method for minimization problems is quite similar to maximization solution space, all points on
problems. There are, however, two important differences. One is that at least one of the that boundary are optimal.
constraints will be of the ≥ variety. This causes the feasible solution space to be away
from the origin. The other difference is that the optimal corner point is the one closest to
the origin. We find the optimal corner point by sliding the objective function line (which
is an isocost line) toward the origin instead of away from it.
Solve the following problem using graphical solution method. Example S-3
Minimize Z = 8x1 + 12x2
Subject to 5x1 + 2x2 ≥ 20
4x1 + 3x2 ≥ 24
x2 ≥ 2
x1, x2 ≥ 0
1. Plot the constraints (shown in Figure 6S-13). Solution
a. Change constraints to equalities.
b. For each constraint, set x1 = 0 and solve for x2, then set x2 = 0 and solve for x1.
c. Graph each constraint. Note that x2 = 2 is a horizontal line parallel to the x1 axis and
2 units above it.
2. Shade the feasible solution space (see Figure 6S-13).
3. Plot the objective function.
a. Select a value for the objective function that causes it to cross the feasible solution
space. Try 8 × 12 = 96.
b. Graph the line 8x1 + 12x2 = 96 (see Figure 6S-14).
x2 x2
12 12
10 10
5x 1
Objective
Feasible function
+2 2
8 solution 8
x =
space
8x
20
6 6 1 +1
2x
4x 1
2 =$
4 4 96
+
3x 2
x2 = 2
=
24
2 2
Optimum
0 0
2 4 6 8 10 12 14 x1 2 4 6 8 10 12 14 x1

The constraints define the feasible solution space for The optimum is the last point the objective function
Example S-3 touches as it is moved toward the origin (Example 6S-3)
4. Slide the objective function toward the origin, being careful to keep it parallel to the
original line.
5. The optimum (last feasible point) is shown in Figure 6S-14. The x2 coordinate (x2 = 2)
can be determined by inspection of the graph. Note that the optimal point is at the
intersection of the line x2 = 2 and the line 4x1 + 3x2 = 24. Substituting the value of
x2 = 2 into this equation will yield the value of x1 at the intersection:
4x1 + 3(2) = 24 → x1 = 4.5
Thus, the optimum is x1 = 4.5 and x2 = 2.
6. Calculate the optimal objective value:
8x1 + 12x2 = 8(4.5) + 12(2) = 60
Slack and Surplus

If a constraint forms the optimal corner point of the feasible solution space, it is called a
binding constraint binding constraint. A binding constraint limits the value of the objective function; if the
A constraint that forms the constraint could be relaxed, an improved solution would be possible. For constraints that
optimal corner point of the are not binding, making them less restrictive will have no impact on the optimal solution.
feasible solution space. If the optimal values of the decision variables are substituted into the left side of a
binding constraint, the resulting value will exactly equal the right-side value of the con-
straint. However, this is different for a non-binding constraint. If the left side is greater
surplus When the optimal than the right side, we say that there is surplus; if the left side is smaller than the right
values of decision variables are side, we say that there is slack. Slack can occur only in a ≤ constraint; it is the amount by
substituted into the left side of which the left side is less than the right side when the optimal values of the decision vari-
a ≥ constraint, surplus is the ables are substituted into the left side. And surplus can occur only in a ≥ constraint; it is
amount the left side exceeds the amount by which the left side exceeds the right side of the constraint when the opti-
the right side. mal values of the decision variables are substituted into the left side.
For example, suppose the optimal values for a problem are x1 = 10 and x2 = 20. If one
slack When the optimal of the constraints is
values of decision variables are
3x1 + 2x2 ≤ 100,
substituted into the left side
of a ≤ constraint, slack is the substituting the optimal values into the left side yields
amount the left side is less than
3(10) + 2(20) = 70
the right side.
Because the constraint is ≤, the difference between the values of 100 and 70 (i.e., 30)
is the slack. Suppose the optimal values had been x1 = 20 and x2 = 20. Substituting these
values into the left side of the constraint would yield 3(20) + 2(20) = 100. Because the
left side equals the right side, this is a binding constraint; slack is equal to zero.
Now consider the constraint:
4x1 + x2 ≥ 50
Suppose the optimal values are x1 = 10 and x2 = 15; substituting these into the left side
yields
4(10) + 15 = 55
Because this is a ≥ constraint, the difference between 55 and 50 (i.e., 5) is the surplus.
If the optimal values had been x1 = 12 and x2 = 2, substitution would result in the left side
of 50. Hence, the constraint would be a binding constraint, and there would be no surplus
(i.e., surplus would be zero).
COMPUTER SOLUTIONS
LO 3
For linear programming problems with more than two decision variables, graphical solu-
tion method cannot be used. Instead a linear programming (LP) software must be used.
There are various LP software available that can solve very large problems with thou-
sands of variables and thousands of constraints (e.g., CPLEX). Here we will be using a
software that comes as an Add-in to Excel, thus making it widely available. It is called
Solver.
We will use the computer problem of Example S-2 to illustrate, using Excel’s Solver.
We repeat its linear programming formulation here for ease of reference.
Maximize 60x1 + 50x2 where x1 = the number of type 1 computers
x2 = the number of type 2 computers
Subject to
Assembly 4x1+ 10x2 ≤ 100 hours
Inspection 2x1 + 1x2 ≤ 22 hours
Storage 3x1+ 3x2 ≤ 39 cubic feet
x1, x2 ≥ 0
To use Solver:
1. First, enter the problem in an Excel worksheet, as shown in Figure 6S-15. What is
not obvious from the figure is the need to enter a formula for each cell where there is a
zero (Solver automatically inserts the zero after you input the formula). The formulas are
for the value of the objective function and the constraints. Before you enter the formulas,
designate the cells where you want the optimal values of x1 and x2 to appear (highlighted
in yellow). Here, cells B4 and C4 are used for decision variables. To enter a formula, click
on the cell that the formula will belong to, and then enter the formula, starting with an
equal sign. We want the optimal value of the objective function to appear in cell D5.
Therefore, in D5, enter the formula
=SUMPRODUCT(B4:C4,B5:C5)
Figure 6S-15
Excel worksheet for the
Computer problem
This is equivalent to = B4*B5 + C4*C5, but will be easier to use for larger problems.
The constraint formulas, in cells D7, D8, and D9, are
for D7: = SUMPRODUCT(B4:C4,B7:C7)
2. Now, click on Data on the top of the worksheet, and in that menu, click on Solver
(on the top right). If Solver is not there, you will need to add it in (In Excel 2010, click
File, then Options, next Add-ins; then click Go. . . in the bottom; a small window opens;
tick Solver add-in, and finally click OK). The Solver Parameters menu will appear (Fig-
ure 6S-16 is the filled-in version). Begin by entering the coordinate of the objective
(profit) function in front of Set Objective:. Alternatively, you can click and then use
the mouse (cursor) to select the cell. In this case, Solver will put $ signs in front of the
row number and column label ($D$5).
Click on the radio button for Max if it isn’t already highlighted. The By Changing Variable
Cells are the cells where you want the optimal values of the decision variables to appear. Here,
they are cells B4 and C4. We indicate this by the range B4:C4 (Solver will add the $ signs).
Next, add the constraints by clicking on Add. When the "Add Constraint" menu ap-
pears, enter the cell that contains the formula for the left side of the first constraint under
"Cell Reference", next select the appropriate inequality sign in the middle, and then enter
the cell that has the right-side amount under the "Constraint". After you have entered the
first constraint, click on Add again, and then enter the next constraint. (Note that con-
straints can be entered in any order.) After the last constraint, click OK rather than Add in
the Constraints window, and you will return to the Solver menu. Tick “Make Unconstrained
Variables Non-Negative,” if it is not ticked already. Select as "Select a Solving Method:"
the “Simplex LP” (this is for linear programs; Solver can also solve other optimization
problems). Finally, click Solve.
3. The Solver Results menu will then appear, indicating that a solution has been found,
or that an error has occurred. If there has been an error, go back to the Solver Parameters
menu and try to find the source of the error. For example, check to see that your constraints
Figure 6S-16
Solver Parameters menu for the
Computer problem (filled in)
FIGURE 6S-17
Excel worksheet solution to the
Computer problem
refer to the correct changing cells, and that the inequality directions are correct. Make the
corrections and click on Solve. Assuming everything is correct, in the Reports box, high-
light both Answer and Sensitivity (by clicking on them), and then click on OK.
4. Solver will incorporate the optimal values of the decision variables and the objective
function on your worksheet (see Figure 6S-17). We can see that the optimal values are
Type 1 = 9 units and Type 2 = 4 units, and the total profit is 740. The Answer report will
also show the optimal values of the Objective Cell and Variable Cells (middle part of
Figure 6S-18), and some information on the constraints (lower part of Figure 6S-18). Of
particular interest here is the indication of which constraints have slack and how much
slack. We can see that the constraint entered in cell D7 (Assembly) has a slack of 24, and
that the constraints entered in cells D8 (Inspection) and D9 (Storage) have slack of zero,
indicating that they are binding constraints.
SENSITIVITY ANALYSIS
LO 4
Sensitivity analysis is assessing the impact of potential changes of the parameters (the nu- sensitivity
merical values) of a linear programming model on its optimal solution. Such changes may analysis Assessing the
occur due to forces beyond a manager’s control or a manager may be contemplating mak- impact of potential changes
ing the changes, say, to increase profits or reduce costs. Also, if a parameter is originally of the parameters (numerical
only roughly estimated, the sensitivity analysis will show if its value has a significant ef- values) of an LP model on its
fect on the optimal solution. If so, further effort should be exerted to obtain a more accu- optimal solution.
rate estimate.
There are three types of parameters:
1. Objective function coefficients
2. Right-side values of constraints
3. Left-side coefficients
We will consider the first two of these here (changes to left-side constraint coefficients
are difficult to study). We begin with changes to objective function coefficients.
Objective Function Coefficient Changes
A change in the value of an objective function coefficient can cause a change in the opti-
mal solution. In the graphical solution method, this means a change to another corner
Figure 6S-18
Answer Report for the
Computer problem
point of the feasible solution space. However, there is a range of values for which the
optimal values of the decision variables will not change. For example, in the Computer
problem, if the profit on type 1 computers increased from $60 per unit to, say, $65 per
unit, the optimal solution would still be to produce 9 units of Type 1 and 4 units of Type 2
computers. Similarly, if the profit per unit on type 1 computers decreased from $60 to,
say, $55 per unit, producing 9 of Type 1 and 4 units of Type 2 would still be optimal.
These sorts of changes are not uncommon; they may be the result of things such as price
changes in raw materials, price discounts, cost reductions in production, and so on. Obvi-
ously, when a change does occur in the value of an objective function coefficient, it can
be helpful for a manager to know if that change will affect the optimal values of the deci-
range of optimality Range
sion variables. The manager can quickly determine this by referring to that coefficient’s
of values for an objective
range of optimality, which is the range of values of an objective function coefficient over
function coefficient over which
which the optimal values of the decision variables will not change.
the optimal solution values of
the decision variables remain Using Graphical Solution Method The range of optimality for Type 1 coefficient in the
the same. Computer problem is 50 to 100. This is so because this corner point of the feasible solu-
tion space will remain optimal as long as the slope of the objective function line (60/50)
stays within the slope of the constraint lines that are binding at the optimal solution. For
the Computer example, the binding constraints are the Inspection (with slope of 2/1) and
Storage (with slope of 3/3) (See Figure 6S-10). Note that currently we have
3/3 < 60/50 < 2/1. Let C1 = coefficient of Type 1 computer in the objective function. Note
that if C1 is changed, to hold the same relationships, i.e., 3/3 < C1/50 < 2/1, we should
have 50 ≤ C1 ≤ 100. If C1 exceeds 100, the optimal solution will move to the neighbor
corner point x1 = 11, x2 = 0, and if C1 goes below 50, the optimal solution will move to the
neighbour corner point, which is the intersection of the Inspection and Assembly con-
straint lines (x1 = 7.5, x2 = 7).
Instead, suppose the coefficient of Type 2 computers were to change. Its range of opti-
mality is 30 to 60. This is so because of a similar argument as above. Let C2 = coefficient
of Type 2 computer in the objective function. If C2 is changed, to hold the relationships
between the slope of the objective function and the slope of the binding constraints, we
should have 3/3 < 60/C2 < 2/1, or 30 ≤ C2 ≤ 60.
In summary, as long as the value of the change doesn’t take an objective coefficient
outside this range, 9 and 4 will still be the optimal values. Note, however, even for
changes that are within the range of optimality, the optimal value of the objective coeffi-
cients will change. If the Type 1 coefficient increased from $60 to $61, and 9 units of
Type 1 is still optimum, profit would increase by $9: 9 units times $1 per unit. Thus, for a
change that is within the range of optimality, a revised value of the objective function
must be determined.
Now, let’s see how we can determine the range of optimality using Excel's Solver.
Using Solver The range of optimality of the objective coefficients can be determined
from the Sensitivity Report (see Figure 6S-19). It shows the value of the objective coeffi
cients that was used in the problem for each type of computer (i.e., 60 and 50), and the
allowable increase and allowable decrease for each coefficient. By subtracting the allow-
able decrease from the original value of the coefficient, and adding the allowable increase
to the original value of the coefficient, we can obtain the range of optimality for each
coefficient. Thus, for Type 1:
60 − 10 = 50 and 60 + 40 = 100
Figure 6S-19
Sensitivity Report for the
Computer problem
Hence, the optimality range for the Type 1 coefficient is 50 to 100. For Type 2:
50 − 20 = 30 and 50 + 10 = 60
Hence, the optimality range for the Type 2 coefficient is 30 to 60.
Range for simultaneous objective coefficient changes To determine if simultaneous

changes to two or more objective coefficients will result in a change to the optimal solu-
tion, divide each coefficient’s change by the allowable change, given by its range of opti-
mality, in the same direction. Thus, if the change is a decrease, divide that amount by the
allowable decrease. Treat all resulting fractions as positive. Sum the fractions. If the sum
does not exceed 1.00, then simultaneous changes are within the range of optimality and
will not result in any change to the optimal values of the decision variables. For example,
suppose the objective coefficient of Type 1 computers was $75 (instead of $60) and the
objective coefficient of Type 2 computers was $40 (instead of $50). Recall the upper limit
of range of optimality of Type 1 objective coefficient was $100 and the lower limit of
range of optimality of Type 2 objective coefficient was $30. Then |(75 – 60)/
(100 – 60)| + |(40 – 50)/(30 – 50)| = .375 + .5 < 1 → no change in the optimal solution.
However, if the objective coefficient of Type 1 computers was $85, we will have |(85 –
60)/(100 – 60)| + |(40 – 50)/(30 – 50)| = .625 + .5 > 1 → change in the optimal solution
(to x1 = 11, x2 = 0).
Changes in the Right-Hand-Side (RHS) Value of a Constraint

In considering right-hand-side changes, it is important to know if a particular constraint
is binding at an optimal solution. Recall that a constraint is binding if substituting the
optimal values of the decision variables into the left side of the constraint results in a
value that is equal to the RHS value. In other words, that constraint stops the objective
function from achieving a better value (e.g., a greater profit or a lower cost). Each con-
shadow price Amount straint has a corresponding shadow price, which is the amount by which the value of
by which the value of the the objective function would change if there were a one-unit change in the RHS value
objective function would of that constraint. If a constraint is non-binding, its shadow price is zero, meaning that
change as a result of a one-unit increasing or decreasing its RHS value by one unit will have no impact on the value of
change in the RHS value of a the objective function. Non-binding constraints have either slack (if the constraint is ≤)
constraint. or surplus (if the constraint is ≥). Suppose a constraint has 10 units of slack at the opti-
mal solution, which means 10 units that are unused. If we were to increase or decrease
the constraint’s RHS value by one unit, the only effect would be to increase or decrease
its slack by one unit. But there is no profit associated with slack, so the value of the
objective function wouldn’t change. On the other hand, if the change is to the RHS
value of a binding constraint, this will cause the optimal values of the decision vari-
ables to change, and hence cause the value of the objective function to change. For ex-
ample, in the Computer problem, the inspection constraint is a binding constraint and
has a shadow price of 10 (this can be derived using Figure 6S-10; if the RHS of Inspec-
tion is changed from 22 to 23, the optimal solution will change to x1 = 10, x2 = 3, and
the objective function value will change to 60(10) + 50(3) = 750, $10 more than the
current objective value of 740. Similarly, if there was one hour less of inspection time,
total profit would decrease by $10).
In general, multiplying the amount of change in the RHS value of a constraint by the
constraint’s shadow price will indicate the change’s impact on the optimal value of the
range of feasibility Range objective function. However, this is true only over a limited range called the range of
of values for its RHS of a feasibility. In this range, the value of the shadow price remains constant. Hence, as long
constraint over which its as a change in the RHS value of a constraint is within its range of feasibility, the shadow
shadow price remains the price will remain the same, and one can readily determine the impact on the objective
same. function. However, if a change goes beyond the range of feasibility, a new solution and
shadow prices would have to be determined.
Let’s see how to determine the range of feasibility from Excel Solver.
Using Solver In the Sensitivity Report there is a table in the bottom labelled “Constraints”
(see Figure 6S-19). The table shows the shadow price for each constraint, its R.H. Side
value, and the allowable increase and allowable decrease. Adding the allowable increase
to the RHS value and subtracting the allowable decrease from the RHS value will produce
the range of feasibility for that constraint. For example, for the inspection constraint the
range of feasibility would be
22 + 4 = 26; 22 − 4 = 18
Hence, the range of feasibility for Inspection is 18 to 26 hours. Similarly, for the stor-
age constraint, the range of feasibility is
39 − 6 = 33 to 39 + 4.5 = 43.5
The range of feasibility for the assembly constraint is a little different; the assembly
constraint is non-binding (note the shadow price of 0) while the other two are binding
(note their non-zero shadow prices). The assembly constraint has a slack of 24 (the
difference between its RHS value of 100 and its final value of 76). With its slack of 24,
its RHS value could decrease by as much as 24 (to 76) before it would become binding.
Conversely, increasing its right-hand side value will produce only more slack. Thus,
no amount of increase in the RHS value of Storage will make it binding, so there is no
upper limit on its allowable increase. Solver indicates this by the large value (1E + 30)
shown for the allowable increase. So Storage's range of feasibility has a lower limit of 76
and no upper limit.
Range for simultaneous RHS changes If there are changes to more than one constraint’s
RHS value, analyze these in the same way as simultaneous changes to the objective func-
tion coefficients. That is, if the change is an increase, divide that amount by that con-
straint’s allowable increase; if the change is a decrease, divide the decrease by the
allowable decrease. Treat all resulting fractions as positive. Sum the fractions. As long as
the sum does not exceed 1.00, the shadow prices won’t change. For example, suppose the
RHS of Inspection changed to 20 (instead of 22) and the RHS of Storage changed to
41(instead of 39). Then |(20 – 22)/(18 – 22)| + |(41 – 39)/(43.5 – 39)| = .5 + .44 < 1 → no
change in the shadow prices. However, if the RHS of Storage changes to 42, we will have
|(20 – 22)/(18 – 22)| + |(42 – 39)/(43.5 – 39)| = .5 + .67 > 1 → change in the shadow
prices.
Table 6S-1 summarizes the impacts of changes that fall within either the range of
optimality or the range of feasibility.
Table 6s-1
CHANGES TO OBJECTIVE FUNCTION COEFFICIENTS THAT ARE WITHIN THE RANGE Summary of the impact of
OF OPTIMALITY
changes within their respective
ranges
Component Result
Values of decision variables No change

Value of objective function Will change
CHANGES TO RHS VALUES OF CONSTRAINTS THAT ARE WITHIN THE RANGE

OF FEASIBILITY
Component Result
Value of shadow price No change

Values of decision variables Will change
Value of objective function Will change
Key Terms binding constraint, 14 range of feasibility, 20

constraints, 2 range of optimality, 18
decision variables, 2 redundant constraint, 11
feasible solution space, 2 sensitivity analysis, 17
graphical solution, 4 shadow price, 20
multiple optimal solutions, 13 slack, 14
objective function, 2 surplus, 14
parameters, 2
Solved Problems
A construction firm specializes in building and selling single-family homes. The firm offers
two basic types of houses, model A house and model B. Model A houses require 4,000 labour
Problem 1 hours, 2 tonnes of stone, and 2,000 board feet of lumber. Model B houses require 10,000 labour
hours, 3 tonnes of stone, and 2,000 board feet of lumber. Due to long lead times for ordering
supplies and the scarcity of skilled and semiskilled workers in the area, the firm will be forced
to rely on its present resources for the upcoming building season. It has 400,000 hours of la-
bour, 150 tonnes of stone, and 200,000 board feet of lumber. What mix of Model A and B
houses should the firm construct if Model A yields a profit of $10,000 per unit and Model B
house yields a profit of $20,000 per unit? Assume that the firm will be able to sell all the units
it builds.
Solution a. Formulate the objective function and constraints:
Maximize Z = 10,000A + 20,000B
Subject to
Labour 4,000A + 10,000B ≤ 400,000 labour hours
Stone 2A + 3B ≤ 150 tonnes
Lumber 2,000A + 2,000B ≤ 200,000 board feet
A, B ≥ 0
b. Graph the constraints and objective function, and identify the optimum corner point (see
graph below). Note that the lumber constraint is redundant: it does not form a boundary of
the feasible solution space.
120
100
Number of units of model B
80 Redundant
60
Lu
m
50
be
Sto
r
40 ne Optimum
20 Lab
or
0
20 40 60 75 80 100 120
Number of units of model A
c. Determine the optimal quantities of models A and B, and calculate the resulting
profit. Because the optimal point is at the intersection of the stone and labour con-
straints, solve those two equations for their common point:
Labour 4,000A+ 10,000B = 400,000
−2,000 × (Stone 2A + 3B = 150)
4,000B = 100,000
B = 25
Substitute B = 25 in one of the equations, and solve for A:

2A + 3(25) = 150 A = 37.5
Z = 10,000(37.5) + 20,000(25) = $875,000
This LP model was solved by Excel’s Solver: Problem 2
Maximize 15x1 + 20x2 + 14x3 where x1 = quantity of product 1
x2 = quantity of product 2
x3 = quantity of product 3
Subject to
Labour 5x1 + 6x2 + 4x3 ≤ 210 hours
Material 10x1 + 8x2 + 5x3 ≤ 200 kg
Machine 4x1 + 2x2 + 5x3 ≤ 170 minutes
x1, x2, x3 ≥ 0
The following Sensitivity Report was obtained.
Variable Cells
Cell Name Final Reduced Objective Allowable Allowable
Value Cost Coefficient Increase Decrease
$B$4 Variables Prod 1 0 −10.6 15 10.6 1E + 30
$C$4 Variables Prod 2 5 0 20 2.4 10.6
$D$4 Variables Prod 3 32 0 14 36 1.5
Constraints
Cell Name Final Shadow Constraint Allowable Allowable
Value Price R.H. Side Increase Decrease
$E$7 Labour 158 0 210 1E + 30 52
$E$8 Material 200 2.4 200 70.90909091 30
$E$9 Machine 170 0.4 170 30 120
a. By how much would the profit per unit of product 1 have to increase in order for it to have a
positive value?
b. If the profit per unit of product 2 increased by $2 to $22, would the optimal production quan-
tities of products 2 and 3 change? Would the optimal value of the objective function
change?
c. If the available amount of labour decreased by 12 hours, would that cause a change in the
optimal values of the decision variables or the optimal value of the objective function?
d. If the available amount of material increased by 10 kg to 210 kg, how would that affect the
optimal value of the objective function?
e. If profit per unit on product 2 increased by $1 and profit per unit on product 3 decreased
by $.50, would these be within the range of simultaneous changes? Would the values of
the decision variables change? What would be the change in the value of the objective
function?
a. Note that the Final (i.e., optimal) Value of Product 1 is 0. The amount of increase in profit of Solution
product 1 would have to equal the Allowable Increase of $10.60.
b. No, because the change would be within its Allowable Increase of $2.4 (i.e., the range of
optimality), Yes, the objective function value would increase by an amount equal to the
optimal quantity of product 2 times its increased unit profit. Hence, it would increase
by 5($2) = $10.
c. No, labour has a slack of 52 hours. Consequently, the only effect would be to decrease the
slack to 40 hours.
d. The change is within the range of feasibility because 10 < 70.91 = Allowable Increase in

Material. The objective function value will increase by the amount of change multiplied by
material’s shadow price of $2.40. Hence, the objective function value would increase by
10($2.40) = $24.00.
e. To determine if the changes are within the range for simultaneous changes, we first calculate
the ratio of the amount of each change to the maximum allowable change in the same direc-
tion. For product 2, it is $1/$2.40 = .417; for product 3, it is −$.50/ − $1.50 = .333. Next, we
calculate the sum of these ratios: .417 + .333 = .750. Because this does not exceed 1.00, we
conclude that these changes are within the allowable range of optimality. This means that the
optimal values of the decision variables will not change. We can calculate the change to the
value of the objective function by multiplying each product’s optimal quantity by its changed
profit per unit: 5($1) + 32(−$.50) = −$11. Hence, with these changes, the value of the objec-
tive function would decrease by $11.
Problem 2 MHK manufactures appliances. The production planner needs to submit a production plan for
next year. The aggregate forecasts of demand for MHK appliances are as follows: Q1: 14,800,
Q2: 26,400, Q3: 35,000, and Q4: 19,200 units. The initial inventory is zero. It costs MHK $24
per appliance to hold it in inventory for a quarter. The regular workforce is 160 workers and
cannot be changed. The annual output per worker is approximately 480 appliances. A worker
costs $15 per hour (including fringe benefits) and works 480 hours a quarter. MHK can hire
temp workers to work during a second shift (a maximum of 160 temp workers can be used). The
hiring cost, including training, is estimated to be $1,800 each, and layoff cost is $1,200 each.
Temp wage rate and productivity are the same as permanent workers. Back-order cost is $150
per unit per quarter. Assume all unmet demand not met during a quarter is back-ordered; how-
ever there should be no back orders at the end of the fourth quarter. Ending inventory can be
zero. The objective is to minimize total labour, hire and lay off, inventory, and back-order costs.
Formulate this problem as a linear program in Excel and solve it using Excel’s Solver.
Solution
The output of 160 regular workers is 160 × (480 units per year/4 quarters in a year) = 19,200 units
per quarter. The labour cost per unit is $15 × (480 hours per quarter/120 units per quarter) = $60.
We introduce variables to represent the numbers of new hires and lay offs of temp workers dur-
ing each quarter. From these, the number of temp workers employed during each quarter is
determined, which in turn determines the number of units of product produced each quarter by
temp workers. The total permanent and temp production minus forecast demand determines
“Output – Forecast.” We also introduce variables to represent the ending inventory and back-
orders at the end of each quarter. The constraints are non-negativity for all variables and number
of temp workers during each quarter, and equations to set beginning inventory + (output –
forecast) − previous quarter’s back order equal to ending inventory – back-order for each quarter.
The latter constraints and the fact that both ending inventory and back order cost money would
force at most one of ending inventory and back order to be positive and equal to beginning
inventory + (output – forecast) − previous quarter’s back order. Choose Simplex LP and make
sure “Make Unconstrained variables Non-negative” is ticked. The optimal solution is given
below. The optimal number of temp workers 77.5 can be rounded up 78.
Quarter 1 2 3 4 Total
Forecast 14,800 26,400 35,000 19,200 95,400
Output
Perm. 19,200 19,200 19,200 19,200 76,800
Temp. hired
(workers) 0.0 77.5 0.0 0.0
Temp. laid-off
(workers) 0.0 0.0 0.0 77.5
Temp. (workers) 0 0.0 77.5 77.5 0.0 155
Temp. (units) 0 9,300 9,300 0
Output - Forecast 4,400 2,100 −6,500 0 0
Inventory
Beginning 0 4,400 6,500 0
Ending 4,400 6,500 0 0
Back order 0 0 0 0
Ending inv −
back order 4,400 6,500 0 0
Beg inv +
output-forecast
− previous
back order 4,400 6,500 0 0
Average inv 2,200 5,450 3,250 0 10,900
Costs
Perm. /unit @ 60 1,152,000 1,152,000 1,152,000 1,152,000 4,608,000
Temp. /unit @ 60 0 558,000 558,000 0 1,116,000
Hire /person @ $1,800 0 139,500 0 0 139,500
Lay off /person @ $1,200 0 0 0 93,000 93,000
Inventory /unit @ 24 52,800 130,800 78,000 0 261600
Back order /unit @ 150 0 0 0 0 0
Total 1,204,800 1,980,300 1,788,000 1,245,000 $6,218,100
1. What is linear programming? Why is it important? When should it be used? (LO1) Discussion and
2. What is a linear programming model? Briefly describe each of its three components. (LO1) Review Questions
3. What is meant by the term feasible solution space? What are its characteristics? (LO1 & 2)
4. What is the standard form of a constraint? (LO1)
5. What is an isocost line? An isoprofit line? (LO2)
6. What does sliding an objective function line toward the origin represent? Away from the ori-
gin? (LO2)
7. Explain the terms (a) redundant constraint (b) multiple optimal solutions. (LO2)
8. What is meant by (a) surplus (b) slack? (LO2)
9. Briefly explain these terms: (LO4)
a. Sensitivity analysis
b. Shadow price
c. Range of feasibility
d. Range of optimality
e. Range for simultaneous changes
1. Solve these problems using the graphical solution method and answer the questions that follow. Problems
Use simultaneous equations to determine the optimal values of the decision variables. (LO2)
a. Maximize Z = 4x1 + 3x2
Subject to
Material 6x1 + 4x2 ≤ 48 kg
Labour 4x1 + 8x2 ≤ 80 hr
x1, x2 ≥ 0
b. Maximize Z = 2x1 + 10x2
Subject to
R 10x1 + 4x2 ≥ 40
S 1x1 + 6x2 ≥ 24
T 1x1 + 2x2 ≤ 14
x1, x2 ≥ 0
c. Maximize Z = 6A + 3B (revenue)

Subject to
Material 20A + 6B ≤ 600 kg
Machinery 25A + 20B ≤ 1,000 hr
Labour 20A + 30B ≤ 1,200 hr
A, B ≥ 0
i. What are the optimal values of the decision variables and Z?
ii. Do any constraints have slack? If yes, which one(s) and how much slack does each have?
iii. Do any constraints have surplus? If yes, which one(s) and how much surplus does each
have?
iv. Are any constraints redundant? If yes, which one(s)? Explain briefly.
2. Solve these problems using the graphical solution method and then answer the questions that follow.
Use simultaneous equations to determine the optimal values of the decision variables. (LO2)
a. Minimize Z = 1.80S + 2.20T
Subject to
Potassium 5S + 8T ≥ 200 grams
Carbohydrate 15S + 6T ≥ 240 grams
Protein 4S + 12T ≥ 180 grams
T T ≥ 10 grams
S, T ≥ 0
b. Minimize Z = 2x1 + 3x2
Subject to
D 4x1 + 2x2 ≥ 20
E 2x1 + 6x2 ≥ 18
F 1x1 + 2x2 ≤ 12
x1, x2 ≥ 0
i. What are the optimal values of the decision variables and Z?
ii. Do any constraints have slack? If yes, which one(s) and how much slack does each have?
iii. Do any constraints have surplus? If yes, which one(s) and how much surplus does each
have?
iv. Are any constraints redundant? If yes, which one(s)? Explain briefly.
3. An appliance manufacturer produces two models of microwave ovens: H and W. Both models
require fabrication and assembly work; each H uses four hours of fabrication and two hours of
assembly, and each W uses two hours of fabrication and six hours of assembly. There are 600
fabrication worker hours available this week and 480 worker hours of assembly. Each H contrib-
utes $40 to profit and each W contributes $30 to profit. Formulate this problem as a linear
program and use the graphical solution method to determine the quantities of H and W that will
maximize total profit. (LO1 & 2)
4. A small grocery store is preparing for the holiday season. The owner must decide how many
bags of deluxe mix and how many bags of standard mix of Peanut/Raisin to prepare. The deluxe
mix has 2/3 pound (lb) raisins and 1/3 lb peanuts, and the standard mix has 1/2 lb raisins and 1/2 lb
peanuts per bag. The store has 90 lb of raisins and 60 lb of peanuts to work with. (LO1 & 2)
Peanuts cost $.60 per lb and raisins cost $1.50 per lb. The deluxe mix will sell for $2.90 per
bag, and the standard mix will sell for $2.55 per bag. The owner estimates that no more than
110 bags of any type of mix can be sold. Formulate this problem as a linear program in order to
maximize total profit and use the graphical solution method to solve it.
5. A retired couple supplement their income by making pie crusts, which they sell to a local gro-
cery store. During the month of September, they produce regular and sweet crusts. The regular
crusts are sold for $1.20 per unit, and the sweet crusts are sold for $1.50 per unit. The couple is
able to sell all of the crusts they produce. Flour and sugar are purchased once each month. For
the month of September, they have 1,200 cups of sugar and 2,100 cups of flour. Each regular
crust requires 11/2 cups of sugar and 3 cups of flour, and each sweet crust requires 2 cups of
sugar and 3 cups of flour. (LO1 & 2)
a. Formulate an LP model to determine the number of regular and sweet crusts that will
maximize revenues if the couple working together can make a crust of either type in six
minutes. They plan to work no more than 60 hours.
b. Solve this problem using graphical solution method. Determine the optimal solution, and
the amounts of sugar, flour, and time that will be unused.
6. Solve each of these LP problems by Excel’s Solver. (LO3)
a. Maximize 4x1 + 2x2 + 5x3
Subject to
1x1 + 2x2 + 1x3 ≤ 25
1x1 + 4x2 + 2x3 ≤ 40
3x1 + 3x2 + 1x3 ≤ 30
x1, x2, x3 ≥ 0
b. Maximize 10x1 + 6x2 + 3x3
Subject to
1x1 + 1x2 + 2x3 ≤ 25
2x1 + 1x2 + 4x3 ≤ 40
1x1 + 2x2 + 3x3 ≤ 40
x1, x2, x3 ≥ 0
7. For Problem 6a, determine the following: (LO4)
a. The range of feasibility for each constraint’s RHS.
b. The range of optimality for the coefficients in the objective function.
8. For Problem 6b: (LO4)
a. Find the range of feasibility for each constraint’s RHS, and interpret your answers.
b. Determine the range of optimality for each coefficient in the objective function and inter-
pret your results.
9. A small company makes three similar products, which all follow the same three-step pro-
cess, consisting of milling, inspection, and drilling. Product A requires 12 minutes of mill-
ing, 5 minutes for inspection, and 10 minutes of drilling per unit; product B requires 10
minutes of milling, 4 minutes for inspection, and 8 minutes of drilling per unit; and product
C requires 8 minutes of milling, 4 minutes for inspection, and 16 minutes of drilling. The
company has 20 hours available during the next period for milling, 15 hours for inspection,
and 24 hours for drilling. Product A contributes $2.40 per unit to profit, B contributes
$2.50 per unit, and C contributes $3.00 per unit. Formulate an LP to determine the mix of
products in order to maximize total contribution to profits for the period. Solve it using
Excel’s Solver. Then, find the range of optimality for the profit coefficient of each vari-
able. (LO1 & 3 & 4)
10. A company makes four juice products using orange, grapefruit, and pineapple juices.
(LO1 & 3)
Product Price/Litre
Orange juice $1.00
Grapefruit juice .90
Pineapple juice .80
All-in-One 1.10
The All-in-One juice has equal parts of orange, grapefruit, and pineapple juice. Each product is
produced in a one-litre size container. On hand are 1,600 litres of orange juice, 1,200 litres of
grapefruit juice, and 800 litres of pineapple juice. The cost per litre is $0.50 for orange juice,
$0.40 for grapefruit juice, and $0.35 for pineapple juice. In addition, the manager wants grape-
fruit juice containers to be no more than 30 percent of all containers produced. She also wants
the ratio of the number of containers of orange juice to the number of containers of pineapple
juice to be at least 7 to 5. Formulate a linear programming model to determine how many con-
tainers of each product to produce to maximize total profit, and solve it using Excel’s Solver.
11. A wood products manufacturer uses idle time at the end of each week to make goods for stock.
Currently, two products are produced for stock: a chopping board and a knife holder. Both
items require three operations: cutting, gluing, and finishing. The manager has collected the
following data on these products: (LO1 & 2)
Time per Unit (Minutes)

Item Profit/Unit Cutting Gluing Finishing
Chopping board $2 1.4 5 12
Knife holder $6 0.8 13 3
The manager has also determined that 56 minutes are available for cutting, 650 minutes for
gluing, and 360 minutes for finishing.
a. Formulate this problem as an LP and solve it using the graphical solution method.
b. Which resources are not completely used by your optimal solution? How much of each
resource is unused?
12. The manager of the deli section of a grocery store has just learned that she has 112 kg of may-
onnaise, of which 70 kg is approaching its expiration date and must be used. To use up the
mayonnaise, the manager has decided to prepare two items: a ham spread and a deli spread.
Each container of ham spread will require 1.4 kg of mayonnaise, and each container of the deli
spread will require 1.0 kg of mayonnaise. The manager has received an order for 10 containers
of ham spread and 8 containers of the deli spread. In addition, the manager has decided to have
at least 10 more containers of each spread available for sale. Both spreads will cost $3 per
container to make, but ham spread sells for $5 per container and deli spread sells for $7 per
container. (LO1 & 2)
a. Formulate an LP that will minimize total cost. Solve it using the graphical solution method.
b. Formulate an LP that will maximize total profit. Solve it using the graphical solution method.
13. A manager wants to know how many units of each product to produce on a daily basis in order
to achieve the highest contribution to profit. Production requirements for the products are
shown in the following table. (LO1 & 3)
Material 1 Material 2 Labour

Product (kg) (kg) (hours)
A 2 3 3.2
B 1 5 1.5
C 6 — 2.0
Material 1 costs $5 per kg, material 2 costs $4 per kg, and labour costs $10 an hour. Product A
sells for $80 a unit, product B sells for $90 a unit, and product C sells for $70 a unit. Available
resources each day are 200 kg of material 1; 300 kg of material 2; and 150 hours of labour.
The manager must satisfy certain output requirements: The output of product A should not
be more than one-third of the total number of units produced; the ratio of units of product A
to units of product B should be 3 to 2; and there is a standing order for 5 units of product A
each day. Formulate a linear programming model for this problem, and then solve it using
Excel’s Solver.
14. A chocolate maker has contracted to operate a small candy counter in a fashionable store. To
start with, the selection of offerings will be intentionally limited. The counter will offer a regu-
lar mix of candy made up of equal parts of cashew, raisin, caramel, and chocolate, and a de-
luxe mix that is one-half cashew and one-half chocolates which will be sold in one-pound
boxes. In addition, the candy counter will offer individual one-pound boxes of cashew, raisin,
caramel, and chocolate. (LO1 & 3)
A major attraction of the candy counter is that all candies are made fresh at the counter. How-
ever, storage space for supplies and ingredients is limited. Bins are available that can hold the
amounts shown in the following table:
Capacity
Ingredient (pounds)
Cashews 120
Raisins 200
Caramels 100
Chocolates 160
In order to present a good image and to encourage purchases, the counter will make at least
20 boxes of each type of product each day. Any leftover boxes at the end of the day will be
given to a nearby nursing home for goodwill. The profit per box for various items has been
determined as follows:
Item Profit per Box

Regular $.80
Deluxe .90
Cashews .70
Raisins .60
Caramels .50
Chocolates .75
a. Formulate an LP model in order to maximize total profit.

b. Solve it using Excel’s Solver.
15. Given the following linear programming model, solve it using Excel’s Solver, and then answer
the questions that follow. (LO4)
Maximize 12x1 + 18x2 + 15x3 where x1 = the quantity of product 1 to make etc.
Subject to
Material 2x1 + 2x2 + 4x3 ≤ 200 kg
Product 2 x2 ≤ 16 units
x1, x2, x3 ≥ 0
a. Are any constraints binding? If so, which one(s)?
b. If the profit on product 3 was changed to $22 a unit, what would the optimal values of the
decision variables be? The objective function? Explain.
c. If the profit on product 1 was changed to $22 a unit, what would the optimal values of the
decision variables be? The objective function? Explain.
d. If 10 hours less of labour time was available, what would the optimal values of the decision
variables be? The objective function? Explain.
e. If the manager decided that up to 20 units of product 2 could be produced (instead of 16),
how much additional profit would be generated?
f. If profit per unit on each product increased by $1, would the optimal values of the decision
variables change? Explain. What would the optimal value of the objective function be?
16. A garden supplies store prepares various grades of pine bark for mulch: nuggets (x1), mini-
nuggets (x2), and chips (x3). The process requires pine bark, machine time, labour time, and
storage space. The following LP model has been developed. (LO4)
Maximize 9x1 + 9x2 + 6x3 (profit)
Subject to
Bark 5x1 + 6x2 + 3x3 ≤ 600 kg
Storage 1x1 + 1x2 + 1x3 ≤ 150 bags
x1, x2, x3 ≥ 0
a. What is the shadow price of a kilogram of bark? Over what range is it appropriate?
b. What is the maximum price the store should pay for additional pine bark?
c. What is the shadow price of labour? Over what range is it in effect?
d. The manager obtained additional machine time through better scheduling. How much
additional machine time can be effectively used for this operation? Why?
e. If the manager can obtain either additional bark or additional storage space, which one
should she choose, and how much?
f. If a change in the chip operation increased the profit on chips from $6 per bag to $7 per
bag, would the optimal quantities change? Would the value of the objective function
change? If so, what would the new value(s) be?
g. If profits on chips increased to $7 per bag and profits on nuggets decreased by $.60 per
bag, would the optimal quantities change? Would the value of the objective function
change? If so, what would the new value(s) be?
h. If the amount of bark available decreased by 15 kg, machine time available decreased by 27
minutes, and storage capacity increased by 5 bags, would this fall in the range of feasibility
for simultaneous changes? If so, what would the value of the objective function be?
17. Linear programming has been used for determining a hospital or a ward’s mix of nursing
staff.1 Consider a ward of Fairbanks Memorial Hospital in Alaska. The hospital uses regis-
tered nurses (RN), licensed practical nurses (LPN), and nursing assistants (NA). The wage
rates per hour are $21, $13, and $11, respectively. The total patient care hours required in a
24-hour period is estimated to be 150 hours. The hourly capability coefficients (efficien-
cies) are 1.0, .90, and .75, respectively, i.e., 1 hour of work of an RN satisfies 1 hour of the
demand, 1 hour work of an LPN satisfies only .9 hour of the demand, and 1 hour work of a
NA satisfies only .75 hour of demand. RNs are qualified to do 100 percent, LPNs up to 85
percent, and NAs up to 50 percent of all tasks. Hospital policy is that at least 50 percent of
the staff is RNs. The number of LPNs and NAs to employ per day is limited to six. Each
RN, LPN, or NA works eight hours per day. Formulate a linear program to decide the num-
ber of RN, LPN, and NA to employ per day in order to minimize total wages subject to the
constraints given above. Solve it using Excel’s solver. (LO1 & 3)
18. Formulate Example 4 of Chapter 13, Aggregate Operations Planning, as a linear program
in Excel and solve it using Excel’s Solver. There is no need to consider the use of overtime.
(LO1 & 3)
*19. A composite wall has 100 m2 area and consists of three layers: plaster, fiberglass, and plywood.2
The thermal resistance of the wall should be at least 1.59 degrees Celsius per Watt. Plaster
should be between 2 and 10 cm thick. Fiberglass should be between 1 and 5 cm thick. Plywood
should be between 1 and 10 cm thick. The thermal resistance and cost of the material are given
in the table below. Formulate a linear program to determine the thickness of each layer in order
to minimize the total cost of the wall subject to meeting the minimum thermal resistance and
minimum/maximum thickness of each layer. Solve it using Excel’s Solver. (LO1 & 3)
Thermal Resistance Cost
Material (degrees Celsius per Watt per metre) ($/m3)
Plaster 3.57 300
Fiberglass 25 1,000
Plywood 7.69 200
20. A small oil refinery must decide how to mix the ingredients to produce gasoline in order to
minimize total ingredient cost subject to meeting demand for gasoline, not exceeding the sup-
ply of each ingredient, and meeting minimum octane number and volatility index, and maxi-
mum vapour pressure of the gasoline.3 The data are: (LO1 & 3)
1
S. K. Bordoloi and E. J. Weatherby, “Managerial Implications of Calculating Optimum Nurse Staffing in
Medical Units,” Health Care Management Review 24(4), Fall 1999, pp. 35–44.
2
Based on C. Onwubiko, Introduction to Engineering Design Optimization, New Jersey: Prentice Hall,
2000, p. 76.
3
Based on N. J. Driebeek, Applied Linear Programming, Massachusetts: Addison-Wesley, 1969, p. 28.
Availability Octane Vapour Volatility Cost

Ingredient (barrels) Number Pressure Index ($/barrel)
Butane 1,700 120 60 105 52
Light naphtha 1,000 84.5 18 30 64
Heavy naphtha 2,300 73 4 12 83
Catalytic naphtha 6,200 96 6.4 15 102
Catalytic reformate 4,000 99 2.5 3 110
Gasoline Requirements
Minimum Maximum Minimum
Demand Octane Vapour Volatility
(barrels) Number Pressure Index
10,000 95 11 18
Formulate this problem as a linear program and solve it using Excel’s Solver.
*21. The manager of an injection moulding shop is interested in using aggregate planning to deter-
mine the shop’s production and inventory levels during each of the next 17 months.4 The shop
uses 8 molding machines to produce approximately 150 different products, employing 15–25
workers. Because of the variety of products, standard (production) hour is used to measure
production and demand. The demands (in hundred standard hours) for the next 17 months are:
18, 24, 34, 30, 34, 29, 29, 20, 21, 22, 24, 19, 25, 25, 34, 33, and 37. Each worker provides 126
standard hours per month. Currently, 18 workers are employed. Hiring cost is estimated at
$470 per worker and lay off cost is estimated at $400 per worker. Labour cost is $13.50 per
standard hour and overtime labour cost is $23.50 per standard hour. End-of-the-month inven-
tories are desired to be close to the following levels (in hundreds of standard hours) for each
of the next 17 months: 51, 62, 62, 67, 65, 59, 48, 48, 47, 49, 44, 51, 53, 64, 67, 73, and 75.
The current inventory level is 5,400 standard hours. If there is excess inventory above the
desired inventory amount, $.80 per standard hour per month is charged as carrying cost, and
if there is “shortage” below the desired inventory amount, $3.60 per standard hour per month
is charged as shortage cost. The maximum number of workers cannot exceed 25. Formulate
this problem as a linear program in Excel and solve it using Excel’s Solver. (LO1 & 3)
22. In an analysis of a cattle feedlot, the following data about cattle feed, their nutrients, and
minimum daily requirements for fattening two-year-old cattle with average weight of 900
pounds were collected5 (LO1 & 3)
Feed- Calcium Phospho- Total Digestible Total Net Carotene

stuff/ % rous dry Protein Digestible Energy (mg/
units % Matter % Nutrients (Therms pound)
% % /100
pound)
Alfalfa 1.47 .24 90.5 10.5 50.3 40.6 8.2
Corn Silage .1 .06 27.4 1.2 18.1 18.1 5.8
Sorghum .08 .04 25.3 .8 15.2 12.2 3.3
Silage
Grain Corn .02 .28 87 6.8 82 82 1.3
Grain 0 0 89.1 8.8 80 77.9 0
Sorghum
Wheat .04 .39 89.5 11.1 80 80 .04
Cottonseed .24 .89 92.2 32.2 69.1 69.1 .1
Meal
4
N. K. Kwak and S. A. DeLurgio, “Optimal Production Scheduling and Inventory Control Under Non-stationary
Demands,” pp. 116–131, in Studies in Linear Programming, H. M. Salkin and J. Saha, Editors, Amsterdam:
North-Holland Publishing, 1975.
5
S. Ashour and C. Anderson, “Linear Programming Analysis of a Cattle Feedlot,” p. 191, in Studies in Linear
Programming, H. M. Salkin and J. Saha, Editors, Amsterdam: North-Holland Publishing, 1975.
Minimum
Daily
Require- .044 .044 20.7 1.53 15 14 50
ment pound pound pound pound pound Therms mg
The costs of feedstuff (cents/pound) are as follows:

Alfalfa 7.2
Corn 9.5
Silage
Sorghum 10
Silage
Grain Corn 8.6
Grain 7.2
Sorghum
Wheat 10.8
Cottonseed 20
Meal
Formulate a linear program to determine the amount of each ingredient to feed a two-year-old
cow per day in order to minimize the total feedstuff cost subject to meeting the minimum
daily requirements. Solve it using Excel’s Solver.
*23. The management of a tomato canning company has to decide the quantity of its products:
(whole tomato, tomato juice, and tomato paste cans), to make from the contracted tomatoes
that are expected soon. The tomatoes will have two grades: A and B; 600,000 pounds of grade
A and 2,400,000 pounds of grade B are expected. The purchase cost will be $.56 per pound of
grade A and $.31 per pound of grade B. The products are sold in cases of 6 cans. A case of
whole tomato cans sells for $4.00, a case of tomato juice cans sells for $4.50, and a case of
tomato paste cans sells for $3.80. The maximum demand for tomato juice is 200,000 cases
and for tomato paste is 300,000 cases (demand for whole tomato cans has no limit). A case of
whole tomato cans is made of 4.5 pounds of tomatoes, a case of juice cans is made of 5
pounds of tomatoes, and a case of tomato paste cans is made of 6 pounds of tomatoes. The
grades of tomatoes have been quantified on a nine-point standard scale as follows: Grade A is
9 whereas Grade B is 5. The minimum standard of whole tomato cans is 8 and of juice cans is
6. (LO1 & 3 & 4)
a. Formulate the above problem as a linear program in order to maximize the total profit
contribution, and solve it using Excel’s Solver.
b. i. What is the most profitable product? Why?
ii. There is an option to buy additional Grade A tomatoes from a different supplier. Should
the management buy more Grade A tomatoes at $.85 per pound? If so, how much should
be purchased? In what product should the additional Grade A tomatoes be used? (Hint: Use
Solver’s Sensitivity Report)
24. The number of employees required to work at the French fries station of a McDonald’s restau-
rant on a given Saturday is as follows6 (LO1 & 3)
Period FF Staff Required

11–noon 1
noon–1 pm 2
1–2 pm 2
2–3 pm 1
3–4 pm 1
4–5 pm 1
5–6 pm 1
6
M. Hojati and A. S. Patil, “An Integer Linear Programming-based Heuristic for Scheduling Heterogeneous
Part-time Service Employees,” European Journal of Operational Research, 209 (2011), pp. 37–50.
For simplicity, assume shifts are 3, 5, or 7 hours long, and can start at the beginning of each
hour starting at 11 am but have to end by 6 pm. The objective is to minimize total staff hours
used plus the number of shifts (staff) used. Formulate a linear program in Excel to model this
shift scheduling problem, and solve it using Excel's Solver.
25. A small refinery wants to optimize the blending of the inputs to gasoline production.7 The
inputs, their daily availability (in litres), and some of their important characteristics are shown
below: (LO1 & 3)
Inputs Availability Octane Reid Vapour

Rating Pressure
Reformate 15,572 98.9 7.66
Fluid Catalytic Cracker Gasoline (FCCG) 15,434 93.2 9.78
Isomerate 6,709 86.1 29.62
The output requirements are as follows:
Demand Octane Rating Reid Vapour

(daily litres) (min) Pressure (max)
Regular gas ≥ 25,000 ≥ 90 ≤ 21.18
Premium gas ≥ 10,000 ≥ 96 ≤ 21.18
At most, 38 percent of each type of gasoline can consist of FCCG. Regular gasoline sells for
$1.20 per litre and premium gasoline sells for $1.4 per litre.
a. Formulate a linear program to maximize daily revenue subject to the constraints given
above. Solve it using Excel’s Solver.
b. Is all FCCG used? Why not?
26. A farmer has two distant pieces of land and would like to grow wheat and Canola on them.
Because of differing soil conditions, there are differences in average yield and cost of grow-
ing on the two lands. The yields and costs are: (LO1 & 3)
Land 1 Land 2
Canola yield/acre 30 bushels 35 bushels
Cost/acre of canola $220 $250
Wheat yield/acre 45 bushels 40 bushels
Cost/acre of wheat $200 $180
Each land is 100 acres. The farmer has $45,000 budget to pay for the costs (before receiving any
revenues). Selling price of a bushel of canola is $14 and a bushel of wheat is $7. Formulate a linear
program to maximize the total profit for the farmer, and solve it using Excel’s Solver.
27. One of the products of Edwards Lifesciences (EL) is artificial heart valves made from the
heart valves of pigs.8 Different sizes of valves are required. However, the size of a pig’s
heart valve cannot be ascertained before the heart is purchased and opened. Therefore, EL
has a mismatch problem: shortages of some sizes and excess of others. A program was
established to document the size distribution of valves supplied by each supplier, and pur-
chases were made from those suppliers with the needed sizes. Linear programming was
used to determine the set of the suppliers that collectively satisfied EL’s demand. Suppose
EL purchases pig valves from three suppliers. The cost and size mix of the valves pur-
chased from each supplier are given in the table below. Each month EL places one order
with each supplier. Suppose next month, 250 large, 300 medium, and 100 small valves are
needed. Formulate an LP that can be used to minimize the cost of acquiring the needed
valves, and use Excel’s Solver to solve it. (LO1 & 3)
7
Based on K. Magoulas et al, “Instructions Are Given For Building Gasoline-blending LP,” Oils and Gas Journal,
86(27), July 1988, pp. 32–37.
8
S. S. Hilal and W. Erikson, “Matching Supplies to Save Lives: Linear Programming the Production of Heart
Valves,” Interfaces, 11(6), Dec. 1981, pp. 48–56.
Supplier Cost Per Valve ($) Percent Large Percent Medium Percent Small
1 10 30 50 20
2 14 10 60 30
3 12 20 55 25
*28. A Connecticut hospital employs 31 full-time orderlies.9 Each orderly works five days per week.
The demand for orderlies fluctuates with the day of the week, so the number required each day
depends on which day of the week it is: Saturday, 20; Sunday, 18; Monday, 18; Tuesday, 30;
Wednesday, 20; Thursday, 31; Friday, 15. The hospital wants to schedule the orderlies in order
to minimize the number of orderlies whose days off are not consecutive. Formulate a linear
program that will accomplish this goal, and solve it using Excel’s Solver. Hint: Define a variable
for staff having days off on each consecutive pairs of days, and orderlies having days off on each
day and another (non-specific) on-consecutive day. (LO1 & 3)
*29. A small manufacturer currently has an inventory of 120 20-ft rods, 160 15-ft rods, and 40 8-ft
rods.10 For production next week, the manufacturer needs at least 200 10-ft rods and 250 6-ft
rods. Each 20-ft rod is worth $10, each 15-ft rod is worth $8, and each 8-ft rod is worth $5.
Each cut costs $1. The manufacturer has the option to buy each 10-ft rod for $6 and each 6-ft
rod for $4. Formulate a linear program to minimize the total cost of meeting the demand, and
solve it using Excel’s Solver. (LO1 & 3)
30. The manager of a shipyard is wondering what type of business to accept in order to maximize
its total profit over next year. The possible jobs include ship repair and new ship construction.
The maximum number of ship repair is 30 and of new ship construction is 2. The profits are
$0.5 million for each ship repair and $7 million for each new ship. Each repair requires
60,000 hours of labour and each new ship requires 1.35 million hours of labour. The shipyard
has a maximum of 3 million hours of labour available next year. Each repair requires 10 days
in the dock and each new ship requires 140 days in the dock. The shipyard has a maximum of
350 dock days. Formulate a linear programming model to determine the number of each job
to accept in order to maximize total profit subject to the above capacity constraints. Solve it
using the graphical solution method. (LO1 & 2)
MINI-Case
Airline Fuel Management leg, the fuel burn-off in kilograms is expected to equal .0258
(duration × weight) + 4,254, where weight = deadweight of
L inear programming has been used for airline fuel manage-

ment. Given a sequence of flight legs, the problem is to deter-
mine the number of kilograms of jet fuel to buy in each stop in
81,522 kg + payload of 19,478 kg + weight of fuel on board,
which includes the initial fuel plus any fuel uplifted in Beirut
(to be determined). The tank capacity is 44,000 kg, and the
order to minimize the total purchase cost subject to restrictions maximum take off weight is 138,600 kg. The safety stock of
on tank capacity, maximum take off weight, and safety stock of fuel at the end of this leg is 7,508 kg. The price of jet fuel
fuel at the end of each leg. The problem is complicated by the in Beirut is $.7615 per gallon (a gallon of jet fuel weighs
fact that the weight of jet fuel itself causes more fuel burn-off. 3.1 kg). The price of jet fuel in Kuwait is $.7454 per
Consider the following small problem for the flight sequence gallon.
of an A310-200 of Middle East Airlines. The return leg from Kuwait to Beirut is expected to take 2.79
The jet has just landed in Beirut with 8,000 kg of fuel on hours. The fuel burn-off in kilograms is expected to equal .0277
board. The next leg is to Kuwait. The average flight time to (duration × weight) + 4,083. The payload will be approximately
Kuwait is 2.54 hours. The fuel burn-off has been shown to 8,278 kg. The safety fuel for this leg is 6,250 kg.
be approximately a linear function of the duration of flight The next leg is from Beirut to Copenhagen and is expected
multiplied by the take off weight of the airplane. On this to take 4.25 hours. The fuel burn-off in kilograms is expected
9
M. Rothstein, “Hospital Manpower Shift Scheduling by Mathematical Programming,” Health Services
Research 8(1), 1973, pp. 60–66.
10
Based on H. A. Eiselt and C-L Sandblom, Linear Programming and its Applications, Berlin: Springer-Verlag,
2007, p. 77.
to equal .02575(duration × weight) + 5,775. The payload will be The return leg from Jeddah to Beirut is expected to take
approximately 17,378 kg. The safety fuel for this leg is 5,450 kg. 2.37 hours. The fuel burn-off in kilograms is expected to equal
The price of jet fuel in Copenhagen is $.682 per gallon. .0252(duration × weight) + 4,084. The payload will be approxi-
The return leg from Copenhagen to Beirut is expected to take mately 2,778 kg. The safety fuel for this leg is 6,085 kg.
3.83 hours. The fuel burn-off in kilograms is expected to equal a. Formulate this problem as a linear program.
.02872(duration × weight) + 3,757. The payload will be approxi-
mately 19,678 kg. The safety fuel for this leg is 6,122 kg. b. Solve it using Excel’s Solver. Are the results what you
The next leg is from Beirut to Jeddah and is expected to take expect? Is there a direct way to determine the optimal solu-
2.46 hours. The fuel burn-off in kilograms is expected to equal . tion without using linear programming?
0254(duration × weight) + 4,009. The payload will be approxi- Source: P. P. Zouein et al., “A Multiple Period Capacitated Inventory
mately 21,878 kg. The safety fuel for this leg is 10,263 kg. The Model for Airline Fuel Management: A Case Study,” Journal of
price of jet fuel in Jeddah is $.7696 per gallon. Operational Research Society 53, 2002, pp. 379–386.
Mini-Case
Son Ltd. is primarily interested in maximizing total profit contribution.

Moreover, the manager wants to know how much labour will be
S on Ltd. manufactures a variety of chemical products used

by photoprocessors. Son was recently bought out by a con-
glomerate, and managers of the two organizations have been
needed, as well as the amount of each material to purchase.
Questions
working together to improve the efficiency of Son’s operations. Prepare a report that addresses the following issues:
Managers have been asked to adhere to weekly operating bud- 1. The optimal quantities of products and the necessary quan-
gets and to develop operating plans using quantitative methods tities of labour and materials.
whenever possible. The manager of one department has been 2. One proposal is to make equal amounts of the products.
given a weekly operating budget of $11,980 for production of What amount of each will maximize contribution, and what
three chemical products, which for convenience shall be referred quantities of labour and materials will be needed? How
to as Q, R, and W. The budget is intended to pay for direct labour much less will total profit contribution be if this proposal
and materials. Processing requirements for the three products, is adopted?
on a per-unit basis, are shown in the table below.
3. How would you formulate the constraint for material A
The company has a contractual obligation for 85 units of
if it was determined that there is a 5-percent waste factor
product R per week.
for material A?
Material A costs $4 per kg, as does material B. Labour costs
Labour Material A Material B
$8 an hour.
Product (hours) (kg) (kg)
Product Q sells for $122 a unit, product R sells for $115 a
unit, and product W sells for $76 a unit. Q 5 2 1
_
The manager is considering a number of different proposals R 4 2
regarding the quantity of each product to produce. The manager W 2 1/2 2

Ch06 Linear Programming

Uploaded by

Copyright:

Available Formats

Ch06 Linear Programming

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ch06 Linear Programming

Uploaded by

Copyright:

Available Formats

SUPPLEMENT TO

SUPPLEMENT OUTLINE Solved Problems, 22 Learning Objectives

problems using Excel’s

the solution of a linear

INTRODUCTION AND LINEAR

Example S-1 Decision variables x1 = Quantity of product 1 to produce

Formulating a Ratio and a Percentage Constraint

But this is not a linear constraint. Cross multiply to obtain:

Resource Amount Available Per Day

GRAPHICAL SOLUTION METHOD

Figure 6S-1 Quantity of x2

An equality constraint is represented by a straight line on the coordinate system that is

2. Set x1 equal to zero and solve for x2:

Feasible for inspection

but not for assembly St

Figure 6S-5 Figure 6S-6

Plotting Objective Function Lines and

Figure 6S-7 Figure 6S-8

Figure 6S-9 Figure 6S-10

Some LP problems have

Figure 6S-13 Figure 6S-14

Slack and Surplus

Range for simultaneous objective coefficient changes To determine if simultaneous

Changes in the Right-Hand-Side (RHS) Value of a Constraint

Values of decision variables No change

CHANGES TO RHS VALUES OF CONSTRAINTS THAT ARE WITHIN THE RANGE

Value of shadow price No change

Key Terms binding constraint, 14 range of feasibility, 20

Substitute B = 25 in one of the equations, and solve for A:

d. The change is within the range of feasibility because 10 < 70.91 = Allowable Increase in

Total 1,204,800 1,980,300 1,788,000 1,245,000 $6,218,100

c. Maximize Z = 6A + 3B (revenue)

Time per Unit (Minutes)

Material 1 Material 2 Labour

Item Profit per Box

a. Formulate an LP model in order to maximize total profit.

Availability Octane Vapour Volatility Cost

Feed- Calcium Phospho- Total Digestible Total Net Carotene

The costs of feedstuff (cents/pound) are as follows:

Period FF Staff Required

Inputs Availability Octane Reid Vapour

The output requirements are as follows:

Demand Octane Rating Reid Vapour

L inear programming has been used for airline fuel manage-

Son Ltd. is primarily interested in maximizing total profit contribution.

S on Ltd. manufactures a variety of chemical products used

You might also like

INTRODUCTION AND LINEAR