Chapter4 AFM3C
Chapter4 AFM3C
The linear programs we solved in chapter 3 contain only two variables, x and y, so that we
could solve them graphically. In practice, linear programs can contain thousands of variables
and constraints.
Later in this chapter we’ll learn to solve linear programs with more than two variables using
the simplex algorithm, which is a numerical solution method that uses matrices and row
operations. However, in order to make the problems practical for learning purposes, our
problems will still have only several variables.
Now that we understand the main concepts behind linear programming, we can also consider
how linear programming is currently used in large scale real-world applications.
Linear programming is used in business and industry in production planning, transportation
and routing, and various types of scheduling. Airlines use linear programs to schedule their
flights, taking into account both scheduling aircraft and scheduling staff. Delivery services
use linear programs to schedule and route shipments to minimize shipment time or minimize
cost. Retailers use linear programs to determine how to order products from manufacturers
and organize deliveries with their stores. Manufacturing companies use linear programming
to plan and schedule production. Financial institutions use linear programming to determine
the mix of financial products they offer, or to schedule payments transferring funds between
institutions. Health care institutions use linear programming to ensure the proper supplies are
available when needed. And as we’ll see below, linear programming has also been used to
organize and coordinate life saving health care procedures.
In some of the applications, the techniques used are related to linear programming but are
more sophisticated than the methods we study in this class. One such technique is called
integer programming. In these situations, answers must be integers to make sense, and can
not be fractions. Problems where solutions must be integers are more difficult to solve than
the linear programs we’ve worked with. In fact, many of our problems have been very
carefully constructed for learning purposes so that the answers just happen to turn out to be
integers, but in the real world unless we specify that as a restriction, there is no guarantee that
a linear program will produce integer solutions. There are also related techniques that are
called non-linear programs, where the functions defining the objective function and/or some
or all of the constraints may be non-linear rather than straight lines.
Chapter 4: Linear Programming: The Simplex Method
Many large businesses that use linear programming and related methods have analysts on
their staff who can perform the analyses needed, including linear programming and other
mathematical techniques. Consulting firms specializing in use of such techniques also aid
businesses who need to apply these methods to their planning and scheduling processes.
When used in business, many different terms may be used to describe the use of techniques
such as linear programming as part of mathematical business models. Optimization,
operations research, business analytics, data science, industrial engineering hand management
science are among the terms used to describe mathematical modelling techniques that may
include linear programming and related met
In the rest of this section we’ll explore six real world applications, and investigate what they
are trying to accomplish using optimization, as well as what their constraints might represent.
AIRLINE SCHEDULING
Airlines use techniques that include and are related to linear programming to schedule their
aircrafts to flights on various routes, and to schedule crews to the flights. In addition, airlines
also use linear programming to determine ticket pricing for various types of seats and levels
of service or amenities, as well as the timing at which ticket prices change.
The process of scheduling aircraft and departure times on flight routes can be expressed as a
model that minimizes cost, of which the largest component is generally fuel costs.
Constraints involve considerations such as:
Each aircraft needs to complete a daily or weekly tour to return back to its point of
origin.
Scheduling sufficient flights to meet demand on each route.
Scheduling the right type and size of aircraft on each route to be appropriate for the
route and for the demand for number of passengers.
Aircraft must be compatible with the airports it departs from and arrives at – not all
airports can handle all types of planes.
A model to accomplish this could contain thousands of variables and constraints. Highly
trained analysts determine ways to translate all the constraints into mathematical inequalities
or equations to put into the model.
After aircraft are scheduled, crews need to be assigned to flights. Each flight needs a pilot, a
co-pilot, and flight attendants. Each crew member needs to complete a daily or weekly tour
to return back to his or her home base. Additional constraints on flight crew assignments take
into account factors such as:
Pilot and co-pilot qualifications to fly the particular type of aircraft they are assigned to
Flight crew have restrictions on the maximum amount of flying time per day and the
length of mandatory rest periods between flights or per day that must meet certain
minimum rest time regulations.
Numbers of crew members required for a particular type or size of aircraft.
Linear Programming: The Simplex Method: Chapter 4
When scheduling crews to flights, the objective function would seek to minimize total flight
crew costs, determined by the number of people on the crew and pay rates of the crew
members. However the cost for any particular route might not end up being the lowest
possible for that route, depending on tradeoffs to the total cost of shifting different crews to
different routes.
An airline can also use linear programming to revise schedules on short notice on an
emergency basis when there is a schedule disruption, such as due to weather. In this case the
considerations to be managed involve:
Getting aircrafts and crews back on schedule as quickly as possible
Moving aircraft from storm areas to areas with calm weather to keep the aircraft safe
from damage and ready to come back into service as quickly and conveniently as possible
Ensuring crews are available to operate the aircraft and that crews continue to meet
mandatory rest period requirements and regulations.
LOANS
A car manufacturer sells its cars though dealers. Dealers can offer loan financing to
customers who need to take out loans to purchase a car. Here we will consider how car
manufacturers can use linear programming to determine the specific characteristics of the
loan they offer to a customer who purchases a car. In a future chapter we will learn how to do
the financial calculations related to loans.
A customer who applies for a car loan fills out an application. This provides the car dealer
with information about that customer. In addition, the car dealer can access a credit bureau to
obtain information about a customer’s credit score.
Based on this information obtained about the customer, the car dealer offers a loan with
certain characteristics, such as interest rate, loan amount, and length of loan repayment
period.
Linear programming can be used as part of the process to determine the characteristics of the
loan offer. The linear program seeks to maximize the profitability of its portfolio of loans.
The constraints limit the risk that the customer will default and will not repay the loan. The
constraints also seek to minimize the risk of losing the loan customer if the conditions of the
loan are not favorable enough; otherwise the customer may find another lender, such as a
bank, which can offer a more favorable loan.
To start the process, sales forecasts are developed to determine demand to know how much of
each type of product to make.
There are often various manufacturing plants at which the products may be produced. The
appropriate ingredients need to be at the production facility to produce the products assigned
to that facility. Transportation costs must be considered, both for obtaining and delivering
ingredients to the correct facilities, and for transport of finished product to the sellers.
The linear program that monitors production planning and scheduling must be updated
frequently – daily or even twice each day – to take into account variations from a master plan.
In the last chapter, we used the geometrical method to solve linear programming problems,
but the geometrical approach will not work for problems that have more than two variables.
In real life situations, linear programming problems consist of literally thousands of variables
and are solved by computers. We can solve these problems algebraically, but that will not be
very efficient. Suppose we were given a problem with, say, 5 variables and 10 constraints.
By choosing all combinations of five equations with five unknowns, we could find all the
corner points, test them for feasibility, and come up with the solution, if it exists. But the
trouble is that even for a problem with so few variables, we will get more than 250 corner
points, and testing each point will be very tedious. So we need a method that has a
systematic algorithm and can be programmed for a computer. The method has to be efficient
enough so we wouldn't have to evaluate the objective function at each corner point. We have
just such a method, and it is called the simplex method.
The simplex method was developed during the Second World War by Dr. George Dantzig.
His linear programming models helped the Allied forces with transportation and scheduling
problems. In 1979, a Soviet scientist named Leonid Khachian developed a method called the
ellipsoid algorithm which was supposed to be revolutionary, but as it turned out it is not any
better than the simplex method. In 1984, Narendra Karmarkar, a research scientist at AT&T
Bell Laboratories developed Karmarkar's algorithm which has been proven to be four times
faster than the simplex method for certain problems. But the simplex method still works the
best for most problems.
The simplex method uses an approach that is very efficient. It does not compute the value of
the objective function at every point; instead, it begins with a corner point of the feasibility
region where all the main variables are zero and then systematically moves from corner point
to corner point, while improving the value of the objective function at each stage. The
process continues until the optimal solution is found.
To learn the simplex method, we try a rather unconventional approach. We first list the
algorithm, and then work a problem. We justify the reasoning behind each step during the
process. A thorough justification is beyond the scope of this course.
We start out with an example we solved in the last chapter by the graphical method. This will
provide us with some insight into the simplex method and at the same time give us the chance
to compare a few of the feasible solutions we obtained previously by the graphical method.
Linear Programming: The Simplex Method: Chapter 4
4. The most negative entry in the bottom row identifies the pivot
column.
7. When there are no more negative entries in the bottom row, we are
finished; otherwise, we start again from step 4.
Now, we use the simplex method to solve Example 1 solved geometrically in section 3.1.
Example 1 Niki holds two part-time jobs, Job I and Job II. She never wants to work more than a
total of 12 hours a week. She has determined that for every hour she works at Job I,
she needs 2 hours of preparation time, and for every hour she works at Job II, she
needs one hour of preparation time, and she cannot spend more than 16 hours for
preparation. If she makes $40 an hour at Job I, and $30 an hour at Job II, how many
hours should she work per week at each job to maximize her income?
Chapter 4: Linear Programming: The Simplex Method
Solution: In solving this problem, we will follow the algorithm listed above.
STEP 1. Set up the problem. Write the objective function and the constraints.
Since the simplex method is used for problems that consist of many variables, it is
not practical to use the variables x, y, z etc. We use symbols x1, x2, x3, and so on.
Let x1 = The number of hours per week Niki will work at Job I.
and x2 = The number of hours per week Niki will work at Job II.
It is customary to choose the variable that is to be maximized as Z.
The problem is formulated the same way as we did in the last chapter.
Maximize Z = 40x1 + 30x2
Subject to: x1 + x2 ≤ 12
2x1 + x2 ≤ 16
x1 ≥ 0; x2 ≥ 0
STEP 2. Convert the inequalities into equations. This is done by adding one
slack variable for each inequality.
For example to convert the inequality x1 + x2 ≤ 12 into an equation, we add a non-
negative variable y1, and we get
x1 + x2 + y1 = 12
Here the variable y1 picks up the slack, and it represents the amount by which x1 +
x2 falls short of 12. In this problem, if Niki works fewer than 12 hours, say 10, then
y1 is 2. Later when we read off the final solution from the simplex table, the values
of the slack variables will identify the unused amounts.
We rewrite the objective function Z = 40x1 + 30x2 as – 40x1 – 30x2 + Z = 0.
After adding the slack variables, our problem reads
Objective function: – 40x1 – 30x2 + Z = 0
Subject to constraints: x1 + x2 + y1 = 12
2x1 + x2 + y2 = 16
x1 ≥ 0; x2 ≥ 0
STEP 3. Construct the initial simplex tableau. Each inequality constraint appears
in its own row. (The non-negativity constraints do not appear as rows in the simplex
tableau.) Write the objective function as the bottom row.
Now that the inequalities are converted into equations, we can represent the problem
into an augmented matrix called the initial simplex tableau as follows.
x1 x2 y1 y2 Z C
1 1 1 0 0 12
2 1 0 1 0 16
–40 –30 0 0 1 0
Linear Programming: The Simplex Method: Chapter 4
Chapter 4: Linear Programming: The Simplex Method
Here the vertical line separates the left hand side of the equations from the right side.
The horizontal line separates the constraints from the objective function. The right
side of the equation is represented by the column C.
The reader needs to observe that the last four columns of this matrix look like the
final matrix for the solution of a system of equations. If we arbitrarily choose x 1 = 0
and x2 = 0, we get
which reads
y1 = 12 y2 = 16 Z =0
The solution obtained by arbitrarily assigning values to some variables and then solving
for the remaining variables is called the basic solution associated with the tableau. So
the above solution is the basic solution associated with the initial simplex tableau. We
can label the basic solution variable in the right of the last column as shown in the table
below.
x1 x2 y1 y2 Z
1 1 1 0 0 12 y1
2 1 0 1 0 16 y2
–40 –30 0 0 1 0 Z
STEP 4. The most negative entry in the bottom row identifies the pivot column.
The most negative entry in the bottom row is –40; therefore the column 1 is identified.
x1 x2 y1 y2 Z
1 1 1 0 0 12 y1
2 1 0 1 0 16 y2
–40 –30 0 0 1 0 Z
Question Why do we choose the most negative entry in the bottom row?
Answer The most negative entry in the bottom row represents the largest coefficient in the
objective function – the coefficient whose entry will increase the value of the
objective function the quickest.
The simplex method begins at a corner point where all the main variables, the
variables that have symbols such as x1, x2, x3 etc., are zero. It then moves from a
corner point to the adjacent corner point always increasing the value of the objective
function. In the case of the objective function Z = 40x1+ 30x2, it will make more
Linear Programming: The Simplex Method: Chapter 4
sense to increase the value of x1 rather than x2. The variable x1 represents the
number of hours per week Niki works at Job I. Since Job I pays $40 per hour as
opposed to Job II which pays only $30, the variable x1 will increase the objective
function by $40 for a unit of increase in the variable x1.
Question Why do we find quotients, and why does the smallest quotient identify a row?
Answer When we choose the most negative entry in the bottom row, we are trying to increase
the value of the objective function by bringing in the variable x1. But we cannot
choose any value for x1. Can we let x1 = 100? Definitely not! That is because Niki
never wants to work for more than 12 hours at both jobs combined: x1 + x2 ≤ 12.
Can we let x1 = 12? Again, the answer is no because the preparation time for Job I is
two times the time spent on the job. Since Niki never wants to spend more than 16
hours for preparation, the maximum time she can work is 16 ÷ 2 = 8.
Now you see the purpose of computing the quotients; using the quotients to identify
the pivot element guarantees that we do not violate the constraints.
Answer As we have mentioned earlier, the simplex method begins with a corner point and
then moves to the next corner point always improving the value of the objective
function. The value of the objective function is improved by changing the number of
units of the variables. We may add the number of units of one variable, while
throwing away the units of another. Pivoting allows us to do just that.
The variable whose units are being added is called the entering variable, and the
variable whose units are being replaced is called the departing variable. The
entering variable in the above table is x1, and it was identified by the most negative
entry in the bottom row. The departing variable y2 was identified by the lowest of all
quotients.
Chapter 4: Linear Programming: The Simplex Method
STEP 6. Perform pivoting to make all other entries in this column zero.
In chapter 2, we used pivoting to obtain the row echelon form of an augmented
matrix. Pivoting is a process of obtaining a 1 in the location of the pivot element, and
then making all other entries zeros in that column. So now our job is to make our
pivot element a 1 by dividing the entire second row by 2. The result follows.
x1 x2 y1 y2 Z
1 1 1 0 0 12
1/2 0 1/2 0 8
–40 –30 0 0 1 0
To obtain a zero in the entry first above the pivot element, we multiply the second
row by –1 and add it to row 1. We get
x1 x2 y1 y2 Z
0 1/2 1 –1/2 0 4
1/2 0 1/2 0 8
–40 –30 0 0 1 0
To obtain a zero in the element below the pivot, we multiply the second row by 40
and add it to the last row.
x1 x2 y1 y2 Z
0 1/2 1 –1/2 0 4 y1
1/2 0 1/2 0 8 x1
0 –10 0 20 1 320 Z
We now determine the basic solution associated with this tableau. By arbitrarily
choosing x2 = 0 and y2 = 0, we obtain x1 = 8, y1 = 4, and z = 320. If we write the
augmented matrix, whose left side is a matrix with columns that have one 1 and all
other entries zeros, we get the following matrix stating the same thing.
STEP 7. When there are no more negative entries in the bottom row, we are
finished; otherwise, we start again from step 4.
Since there is still a negative entry, –10, in the bottom row, we need to begin, again,
from step 4. This time we will not repeat the details of every step, instead, we will
identify the column and row that give us the pivot element, and highlight the pivot
element. The result is as follows.
Linear Programming: The Simplex Method: Chapter 4
x1 x2 y1 y2 Z
0 1 –1/2 0 4 y1 4 ÷ 1/2 = 8
1 1/2 0 1/2 0 8 x1 8 ÷ 1/2 = 16
0 –10 0 20 1 320 Z
We make the pivot element 1 by multiplying row 1 by 2, and we get
x1 x2 y1 y2 Z
0 2 –1 0 8
1 1/2 0 1/2 0 8
0 –10 0 20 1 320
Now to make all other entries as zeros in this column, we first multiply row 1 by –1/2
and add it to row 2, and then multiply row 1 by 10 and add it to the bottom row.
x1 x2 y1 y2 Z
0 1 2 –1 0 8 x2
1 0 –1 1 0 4 x1
0 0 20 10 1 400 Z
We no longer have negative entries in the bottom row, therefore we are finished.
Question Why are we finished when there are no negative entries in the bottom row?
Answer The answer lies in the bottom row. The bottom row corresponds to the equation:
0x1 + 0x2+ 20y1 + 10y2 + Z = 400 or
Z = 400 – 20y1 – 10y2
Since all variables are non-negative, the highest value Z can ever achieve is 400, and
that will happen only when y1 and y2 are zero.
In this section, we will solve the standard linear programming minimization problems using
the simplex method. Once again, we remind the reader that in the standard minimization
problems all constraints are of the form ax + by ≥ c.
The procedure to solve these problems was developed by Dr. John Von Neuman. It involves
solving an associated problem called the dual problem. To every minimization problem
there corresponds a dual problem. The solution of the dual problem is used to find the
solution of the original problem. The dual problem is a maximization problem, which we
learned to solve in the last section. We first solve the dual problem by the simplex method.
From the final simplex tableau, we then extract the solution to the original minimization problem.
Before we go any further, however, we first learn to convert a minimization problem into its
corresponding maximization problem called its dual.
Solution: To achieve our goal, we first express our problem as the following matrix.
1 2 40
1 1 30
12 16 0
Observe that this table looks like an initial simplex tableau without the slack variables.
Next, we write a matrix whose columns are the rows of this matrix, and the rows are the
columns. Such a matrix is called a transpose of the original matrix. We get:
1 1 12
2 1 16
40 30 0
The following maximization problem associated with the above matrix is called its dual.
Maximize Z = 40y1 + 30y2
Subject to: y1 + y2 ≤ 12
2y1 + y2 ≤ 16
y1 ≥ 0; y2 ≥ 0
Note that we have chosen the variables as y's, instead of x's, to distinguish the two problems.
Linear Programming: The Simplex Method: Chapter 4
Example 2 Solve graphically both the minimization problem and its dual maximization problem.
We have plotted the graph, shaded the feasibility region, and labeled the corner points.
The corner point (20, 10) gives the lowest value for the objective function and that
value is 400.
Now its dual is: Maximize Z = 40y1 + 30y2
Subject to: y1 + y2 ≤ 12
2y1 + y2 ≤ 16
y1 ≥ 0; y2 ≥ 0
We graph the inequalities:
Again, we have plotted the graph, shaded the feasibility region, and labeled the
corner points. The corner point (4, 8) gives the highest value for the objective
function, with a value of 400.
The reader may recognize that Example 2 above is the same as Example 1, in section 3.1. It is
also the same problem as Example 1 in section 4.1, where we solved it by the simplex method.
We observe that the minimum value of the minimization problem is the same as the
maximum value of the maximization problem; in Example 2 the minimum and maximum are
both 400. This is not a coincident. We state the duality principle.
Chapter 4: Linear Programming: The Simplex Method
Our next goal is to extract the solution for our minimization problem from the corresponding
dual. To do this, we solve the dual by the simplex method.
Example 3 Find the solution to the minimization problem in Example 1 by solving its dual using
the simplex method. We rewrite our problem.
Minimize Z = 12x1 + 16x2
Subject to: x1 + 2x2 ≥ 40
x1 + x2 ≥ 30
x1 ≥ 0; x2 ≥ 0
The minimization problem has a minimum value of 400 at the corner point (20, 10).
Chapter 4: Linear Programming: The Simplex Method