Chapter 2-3 Optimization and Linear Programming
Chapter 2-3 Optimization and Linear Programming
1. Introduction
. Linear programming (LP) is a mathematical tool for solving optimization problems with constraints. Dr.
George Dantzig, who is considered the father of LP, developed the simplex method for solving linear
programs in 1947. Since then, LP has been applied in agriculture, banking, education, forestry, military,
manufacturing, petroleum, telecommunication, transportation, and many other industries.
(1) Decision variable: An alternative available to the decision maker in terms of amount of input or
output.
(2) Objective function: A mathematical expression consisting of decision variables and coefficients that
represents the goal of a linear program.
(3) Constraint: A mathematical expression consisting of decision variables and coefficients that restricts
the alternatives available to the decision maker. There are two types of constraint: structural and non-
negativity.
. Example 2.1: Dallas Toy Company manufactures two types of toys: car and train. A toy car sells for $6
and uses 1 hour of machine time as well as 2 pounds of raw material. A toy train sells for $9 and requires
1 hour of machine time as well as 4 pounds of raw material. There are 8 hours of machine time and 24
pounds of raw material available per day. Set up a linear program that may be used to determine the
quantity of each type of toy to be made to maximize the company’s daily total revenue.
[Solution] The key information on the problem is summarized in the following table:
Z = f(x, y) = 6x + 9y
(3) Constraints:
(a) Structural:
(b) Non-negativity:
x 0 and y 0 or x, y 0
Maximize Z = 6x + 9y
subject to: x+ y 8
2x + 4y 24
x , y 0
. Example 2.2: A farmer in Justin, TX, raises pigs and he needs to mix a pig feed from two grains. Each
pound of Grain A costs $16 and contains 3 units of Vitamin C as well as 6 units of Vitamin D. Each
pound of Grain B costs $12 and contains 5 units on Vitamin C as well as 2 units of Vitamin D. In order
for the pigs to be healthy, the feed must contain at least 15 units of Vitamin C and at least 18 units of
Vitamin D. Develop an LP model that may be used to determine the amount of each grain to be mixed to
minimize the total cost of the pig feed.
[Solution] The key information on the problem is summarized in the following table:
(3) Constraints:
(c) Structural:
Vitamin C requirement: 3a + 5b 15
Vitamin D requirement: 6a + 2b 18
(d) Non-negativity:
a 0 and b 0 or a, b 0
subject to: 3a + 5b 15
6a + 2b 18
a , b 0
. Once a problem is formulated as a linear program, it can be solved by using (1) the graphical method
(including the iso-value line approach and the corner-point approach), (2) the simplex method, or (3)
computer software. Each of these is discussed below.
. The graphical method provides a visual portrayal of many of the important concepts of a linear program
on the two-dimensional plane. However, it is only applicable to problems that involve at most two
decision variables.
(3) (i) Plot at least two iso-profit (or iso-cost) lines to identify the direction in which the objective
function value is increasing (or decreasing), or
(ii) Find the coordinates of the corner points and compute the objective function value at each of
them.
(4) Determine the optimal solution and the optimal objective function value.
. Example 2.3: Consider the linear program formulated in Example 2.1, which is reproduced below:
Maximize Z = 6x + 9y
subject to: x+ y 8
2x + 4y 24
x , y 0
(1) Plot the constraints and graph the feasible region for the LP.
(2) Plot iso-revenue lines to determine the optimal product mix and the maximum daily revenue.
[Solution] (1) The graph for the problem is shown the next page.
(2) It is seen that the objective function will attain its maximum value at point B. Solving
the simultaneous equations of x + y = 8 and 2x + 4y = 24, we find that the optimal
solution is (x*, y*) = (4, 4) and the optimal objective function value is 6x * + 9y* = 6(4) +
9(4) = 60. In other words, the company should produce 4 toy cars and 4 toy trains to
maximize its daily total revenue at $60.
(3) The coordinates of the four corner points along with the corresponding objective
function values are calculated as follows:
4
A = (0, 6): Z = 6x + 9y = 6(0) + 9(6) = 54 (we need to solve the simultaneous equations
of x = 0 and 2x + 4y = 24 to find x = 0 and y = 6)
B = (4, 4): Z = 6x + 9y = 6(4) + 9(4) = 60 (we need to solve the simultaneous equations
of x + y = 8 and 2x + 4y = 24 to find x = 4 and y = 4)
C = (8, 0): Z = 6x + 9y = 6(8) + 9(0) = 48 (we need to solve the simultaneous equations
of y = 0 and x + y = 8 to find x = 8 and y = 0)
Since 60 > 54 > 48 > 0, the optimal solution is (x *, y*) = (4, 4) and the optimal objective
function value is Z* = 60. In other words, the company should produce 4 toy cars and 4
toy trains to maximize its daily total revenue at $60.
. Remark: It can be shown that the optimal solution to a linear program always occurs at one of the
5
corners of the feasible region. Consequently, it suffices to evaluate and compare the objective function
values at the corner points to solve the LP. This is the so-called corner-point approach.
. Example 2.4: Consider the linear program formulated in Example 2.2, which is reproduced below:
(1) Plot the constraints and graph the feasible region for the LP.
(2) Plot iso-cost lines to determine the optimal amounts of grains to mix and the minimum total cost of
the pig feed.
[Solution] (1) The graph for the problem is shown on the next page.
(2) It is seen that the objective function will attain its minimum value at point B. Solving
the simultaneous equations of 3a + 5b = 15 and 6a + 2b = 18, we find that the optimal
solution is (a*, b*) = (2.5, 1.5) and the optimal objective function value is Z * = 16a* +
12b* = 16(2.5) + 12(1.5) = 58. In other words, the farmer should mix 2.5 pounds of
Grains A and 1.5 pounds of Grain B to minimize the total pig feed cost at $58.
(3) The coordinates of the three corner points along with the corresponding objective
function values are calculated as follows:
A = (0, 9): Z = 16a + 12b = 16(0) + 12(9) = 108 (we need to solve the simultaneous
equations of a = 0 and 6a + 2b = 18 to find a = 0 and b = 9)
B = (2.5, 1.5): Z = 16a + 12b = 16(2.5) + 12(1.5) = 58 (we need to solve the
simultaneous equations of 3a + 5b = 15 and 6a + 2b = 18 to find a = 2.5 and b = 1.5)
C = (5, 0): Z = 16a + 12b = 16(5) + 12(0) = 80 (we need to solve the simultaneous
equations of b = 0 and 3a + 5b = 15 to find a = 5 and b = 0)
Since 58 < 80 < 108, the optimal solution is (a *, b*) = (2.5, 1.5) and the optimal
objective function value is Z * = 58. In other words, the farmer should mix 2.5 pounds of
Grain A with 1.5 pounds of Grain B to minimize the total pig feed cost at $58.
. In contrast to the graphical method, the simplex method can be used to solve linear programs involving
any number of decision variables. It consists of a series of simple but tedious tableau iterations. While
most people rely on computer when applying the simplex method, some familiarity with the manual
computations will be helpful in understanding the technique.
. In implementing the simplex algorithm, even the slightest calculating error can easily distort the final
result. Hence, it would be best to work with numbers in fractional form.
. In solving a linear program using the simplex method, the non-negativity constraint is not considered
explicitly. The steps involved for a maximization LP are:
(d) Arrange the objective coefficients and the constraint coefficients into a tableau.
(e) Compute the numbers in the “Zj” row. (This can be done by multiplying the numbers in the jth
column with their counterparts in the left-most column (“C_j”) and then adding up the
results.)
(f) Compute the numbers in the “Zj - Cj” row. (For each j, subtract Zj with the corresponding
entry on the Cj row)
If all the numbers in the “Zj - Cj” row are either positive (negative for a minimization problem) or zero,
the current solution is optimal; otherwise, the following steps should be followed before an improved
solution is derived:
(a) The column with the most negative (positive for a minimization problem) number in the “Z j - Cj” is
termed the pivot column. A tie can be broken in an arbitrary way.
(b) Divide each number in the right-most or solution quantity (“SQ”) column by the corresponding
number in the pivot column that is positive. The row with the smallest ratio is termed the pivot
row. Note that the pivot row must contain a positive number.
(c) The number at the intersection of the pivot column and the pivot row is termed the pivot number.
The pivot number is always positive.
8
Change the VIS column by replacing the pivot row variable with the pivot column variable,
together with the new variable’s coefficient in Cj row updated in the leftmost column.
(a) Derive new numbers in the pivot row by dividing the entire row by the pivot number.
(b) Update each of the remaining rows by performing the following operation:
New number at (row i, column j) = Old number at (row i, column j) – Old number at (row i,
pivot column) * New number at (pivot row, column j)
Be aware of signs!
Carry out the same update for the SQ column as well.
(Optional: this step will not be tested either in homework assignments or the exams)
This helps checking whether you made a calculation mistake in the process.
(4) Go to (2)
. Example 2.5: Reconsider the LP set up in Example 2.1, which was solved by using the graphical method
in Example 2.3 and is reproduced below. Use the simplex method to solve it.
Maximize Z = 6x + 9y
subject to: x+ y < 8
2x + 4y < 24
x , y > 0
Z= 6x + 9y + 0u + 0v
x + y + u + 0v = 8
2x + 4y + 0u + v = 24
Cj | 6 9 0 0 |
-----------------------------------------------------------------------------------
VIS | x y u v | SQ
-----------------------------------------------------------------------------------
0 u | 1 1 1 0 | 8 8/1 = 8
0 v | 2 4 0 1 | 24 24/4 = 6
-----------------------------------------------------------------------------------
Zj | 0 0 0 0 | 0
9
Z j - Cj | -6 -9 0 0 |
The current solution (u, v) = (0, 0) is not optimal since -6 < 0 and -9 < 0. Given that -9 is the
most negative number in the “Z j - Cj” row and 6 < 8, we see that the pivot column is “y”
(“y” is the entering variable), the pivot row is “v” (“v” is the leaving variable), and the pivot
number is 4. The new simplex tableau follows:
Cj | 6 9 0 0 |
-----------------------------------------------------------------------------------
VIS | x y u v | SQ
-----------------------------------------------------------------------------------
0 u | 1/2 0 1 -1/4 | 2 2/(1/2) = 4
9 y | 1/2 1 0 1/4 | 6 6/(1/2) = 12
-----------------------------------------------------------------------------------
Zj | 9/2 9 0 9/4 | 54
Z j - Cj | -3/2 0 0 9/4 |
The current solution (u, y) = (2, 6) is not optimal since -3/2 < 0. Given that -3/2 is the only
negative number in the “Z j - Cj” row and 4 < 12, the pivot column is “x” (“x” is the entering
variable), the pivot row is “u” (“u” is the leaving variable), and the pivot number is 1/2. The
new simplex tableau follows:
Cj | 6 9 0 0 |
-----------------------------------------------------------------------------------
VIS | x y u v | SQ
-----------------------------------------------------------------------------------
6 x | 1 0 2 -1/2 | 4
9 y | 0 1 -1 1/2 | 4
-----------------------------------------------------------------------------------
Zj | 6 9 3 3/2 | 60
Z j - Cj | 0 0 3 3/2 |
The current solution (x, y) = (4, 4) is optimal since all the numbers in the "Z j - Cj" row are
positive or zero. In conclusion, the optimal solution is (x *, y*) = (4, 4) and the optimal
objective function value is Z * = 60. In other words, the company should produce 4 toy cars
and 4 toy trains to maximize its daily total revenue at $60. (Note: These are the same as
those found in Example 2.3.)
. Remark: Each simplex tableau solution corresponds to a corner point of the feasible solution space in the
graphical method. For instance, the sequence of solutions (x, y) obtained in Example 2.5 above is (0, 0) -
> (0, 6) -> (4, 4), which corresponds to the sequence of points D -> A -> B one on one in Example 2.3.
. Example 2.6: Reconsider the LP set up in Example 2.2, which was solved by using the graphical method
in Example 2.4 and is reproduced below. Use the simplex method to solve it
Z = 16a + 12b + 0c + Md + 0e + Mf
3a + 5b - c+ d+ 0e + 0f = 15
6a + 2b + 0c + 0d - e+ f = 18
Cj | 16 12 0 M 0 M |
-----------------------------------------------------------------------------------------------
VIS | a b c d e f | SQ
-----------------------------------------------------------------------------------------------
M d | 3 5 -1 1 0 0 | 15 15/3 = 5
M f | 6 2 0 0 -1 1 | 18 18/6 = 3
-----------------------------------------------------------------------------------------------
Zj | 9M 7M -M M -M M | 33M
Z j - Cj | 9M-16 7M-12 -M 0 -M 0 |
The current solution (a, b) = (0, 0) is not optimal since 9M - 16 > 0 and 7M - 12 > 0. Given
that 9M - 16 is the most positive number in the “Z j - Cj” row and 3 < 5, the pivot column is
“a” (“a” is the entering variable), the pivot row is “f” (“f” is the leaving variable), and the
pivot number is 6. The new simplex tableau follows:
Cj | 16 12 0 M 0 M |
---------------------------------------------------------------------------------------------------
VIS | a b c d e f | SQ
---------------------------------------------------------------------------------------------------
M d | 0 4 -1 1 1/2 -1/2 | 6 6/4 = 1.5
16 a | 1 1/3 0 0 -1/6 1/6 | 3 3/(1/3) = 9
---------------------------------------------------------------------------------------------------
Zj | 16 4M+16/3 -M M M/2-16/6 -M/2+16/6 | 6M+48
Z j - Cj | 0 4M-20/3 -M 0 M/2-16/6 -3M/2+16/6 |
The current solution (a, b) = (3, 0) is not optimal since 4M - 20/3 > 0 and M/2 - 16/6 > 0.
Given that 4M - 20/3 is the most positive number in the “Z j - Cj” row and 1.5 < 9, the pivot
column is “b” (“b” is the entering variable), the pivot row is “d” (“d” is the leaving
variable), and the pivot number is 4. The new simplex tableau follows:
Cj | 16 12 0 M 0 M |
----------------------------------------------------------------------------------------------------
VIS | a b c d e f | SQ
----------------------------------------------------------------------------------------------------
12 b | 0 1 -1/4 1/4 1/8 -1/8 | 1.5
16 a | 1 0 1/12 -1/12 -5/24 5/24 | 2.5
----------------------------------------------------------------------------------------------------
Zj | 16 12 -20/12 20/12 -44/24 44/24 | 58
Z j - Cj | 0 0 -20/12 -M+20/12 -44/24 -M+44/24 |
The current solution (a, b) = (2.5, 1.5) is optimal since all the numbers in the “Z j - Cj” row
are negative or zero. In conclusion, the optimal solution is (a *, b*) = (2.5, 1.5) and the
optimal objective function value is Z * = 58. In other words, the farmer should mix 2.5
pounds of Grains A with 1.5 pounds of Grain B to minimize the total pig feed cost at $58.
(Note: These are the same as those found in Example 2.4.)
. A large number of software packages are available for solving linear programs very efficiently. In what
follows, one of them are applied to the product-mix problem and the pig feed problem discussed
previously: Excel Solver.
. Before running any computer program to solve a linear program, it is important to ensure that the LP is in
standard form, i.e., only constants (no decision variables) are allowed to appear at the right-hand sides of
the constraints.
. Example 2.7: Reconsider the linear program developed in Example 2.1. Use Solver to solve the LP. Be
11
sure to include a copy of the Answer Report and interpret the key results.
[Solution] Note that the LP is already is already in standard form. The Answer Report from Solver is
displayed below:
It is seen that the optimal solution is (x *, y*) = (4, 4) and the optimal objective function
value is Z* = 60. In other words, the company should produce 4 toy cars and 4 toy trains to
maximize its daily total revenue at $60.
. Example 2.8: Reconsider the linear program developed in Example 2.2. Use Solver to solve the LP. Be
sure to include a copy of the Answer Report and interpret the key results.
[Solution] Note that the LP is already in standard form. The Answer Report from Solver is as follows.
It is seen that the optimal solution is (a *, b*) = (2.5, 1.5) and the optimal objective function
value is Z* = 58. In other words, the farmer should mix 2.5 pounds of Grain A with 1.5
pounds of Grain B to minimize the total cost of the pig feed at $58.
. Example 2.9: Top Speed Bicycle Company manufactures a line of 10-speed bicycles. The firm has plants
in Baton Rouge and Omaha. Its major markets are located in New York, Chicago, and Los Angeles. The
unit transportation costs, the plant capacities, and the market sales requirements are summarized below:
NY CH LA Capacity
---------------------------------------------------------------------------
12
BR $2 $3 $5 20,000
OM 3 1 4 15,000
---------------------------------------------------------------------------
Requirement 12,000 7,000 16,000
[Solution] Let BN = Number of bicycles to be shipped from Baton Rouge to New York
BC = Number of bicycles to be shipped from Baton Rouge to Chicago
BL = Number of bicycles to be shipped from Baton Rouge to Los Angeles
ON = Number of bicycles to be shipped from Omaha to New York
OC = Number of bicycles to be shipped from Omaha to Chicago
OL = Number of bicycles to be shipped from Omaha to Los Angeles
. Example 2.10: Oriental Express is a fast-food outlet right next to the UTD campus and it is open 24
hours a day. Waiters report for duty at 3:00 A.M., 7:00 A.M., 11:00 A.M., 3:00 P.M., 7:00 P.M., or 11:00
P.M., and each works an 8-hour shift. The following table shows the minimum number of workers
needed during each of the six periods into which the day is divided:
Set up an LP that can be used to determine the minimum number of waiters required for one day's
operations of the restaurant.
[Solution] Let xi be the number of waiters reporting for duty at the beginning of period i, i = 1, 2, ..., 6.
An LP model for the scheduling problem follows:
Minimize Z = x1 + x2 + x3 + x4 + x5 + x6
subject to: x1 + x6 3
x1 + x2 8
x2 + x3 12
x3 + x4 9
x4 + x5 16
x5 + x6 4
13
x1 , x2 , x3 , x4 , x5 , x6 0
. Example 2.11: A small fabrication firm makes three basic types of components (A, B, and C) for use by
other companies. Each component is processed on three machines. The processing times and machine
capacities are shown in the following table.
Each component contains a different amount of each of two basic raw materials. Raw material 1 costs
$0.20 per ounce whereas raw material 2 costs $0.35 per ounce. There are 200,000 ounces of raw material
1 and 85,000 ounces of raw material 2 available. Relevant information on material requirements has been
collected and summarized below:
The unit selling prices of components A, B, and C are, respectively, $40, $28, and $24. Assume that the
company must make at least 1,200 units of component B, labor costs are negligible, and the objective is
to maximize the total profit. Formulate a linear programming model for the problem.
The unit profit of type A component is 40 - 0.20(32) - 0.35(12) = $29.4. Those for the other
two types of component can be found in the same way. An appropriate LP model follows:
. Example 2.12: Big Bucks Mutual Funds, Inc., located in Harrisburg, PA, just obtained $200,000 by
converting industrial bonds to cash and is now looking for other opportunities for these funds.
Considering Big Bucks' current investments, the firm's top financial analyst recommends that all new
investments should be made in the oil industry, steel industry, or government bonds. Specifically, the
analyst has identified five investment opportunities and projected their annual rates of return. The
investments and rates of return are shown in the following table:
----------------------------------------------------------
Atlantic Oil 7.3%
Pacific Oil 10.2%
Midwest Steel 6.1%
Harbor Steel 8.5%
Government bonds 9.6%
----------------------------------------------------------
(1) Neither industry (oil or steel) should receive more than 50% of the total amount of money available.
(2) Government bonds should be at least 60% of the total steel industry investment.
(3) The investment in Pacific Oil cannot be more than 20% of the total oil industry investment.
Set up an LP model for the portfolio selection problem so that Big Bucks' projected return is maximized.
. Example 2.13: International Paper Mills, Inc., operates plants in Lewiston, ME, and Hanover, NH, in
addition to owning warehouses in Albany, NY, and Storrs, CT. Retail outlets are located in Boson, MA,
New York, NY, and Philadelphia, PA. Typically, finished products are transported from the plants to the
warehouses. The requirements at the retail outlets are met by shipments from the warehouses. Relevant
information about the supplies and demands has been summarized below:
15
The unit transportation costs (in dollars) for shipments from the two plants to the two warehouses and
from the two warehouses to the three retail outlets are as follows:
Lewiston 7 5 - - -
Hanover 3 4 - - -
Albany - - 8 5 7
Storrs - - 5 6 10
(2) How should the LP model formulated in (1) above be modified if it is possible to ship products from
the Hanover plant directly to the Philadelphia retail outlet at $9 per unit and from the Boston retail
outlet to the New York retail outlet at $ 4 per unit?
[Solution] (1) Let LA = Number of units of the product to be shipped from Lewiston to Albany
LS = Number of units of the product to be shipped from Lewiston to Storrs
HA = Number of units of the product to be shipped from Hanover to Albany
HS = Number of units of the product to be shipped from Hanover to Storrs
AB = Number of units of the product to be shipped from Albany to Boston
AN = Number of units of the product to be shipped from Albany to New York
AP = Number of units of the product to be shipped from Albany to Philadelphia
SB = Number of units of the product to be shipped from Storrs to Boston
SN = Number of units of the product to be shipped from Storrs to New York
SP = Number of units of the product to be shipped from Storrs to Philadelphia
LA + HA - AB - AN - AP = 0
LS + HS - SB - SN - SP = 0
AB + SB ³ 150
AN + SN ³ 100
AP + SP ³ 150
LA , LS , ... , SP ³ 0
(1) Plot the constraints and graph the feasible region for the LP.
(2) Plot iso-cost lines to determine the optimal solution and the minimum total cost.
(3) Use the corner-point approach to find the optimal solution as well as the minimum total cost.
(2) It is noted that the objective function will attain its minimum value at point D. Solving
the simultaneous equations of 10E + 15F = 90 and E - 3F = 0, we find that the optimal
solution is (E*, F*) = (6, 2) and the optimal objective function value is Z * = 1,200E* +
2,400F* = 1,200(6) + 2,400(2) = 12,000.
17
(3) The coordinates of the four corner points along with the corresponding objective
function values are calculated below:
A = (0, 6): Z = 1,200E + 2,400F = 1,200(0) + 2,400(6) = 14,400 (we need to solve the
simultaneous equations of E = 0 and -5E + 7.5F = 45 to find E = 0 and F = 6.)
C = (8, 8/3): Z = 1,200E + 2,400F = 1,200(8) + 2,400(8/3) = 16,000 (we need to solve
the simultaneous equations of E = 8 and E - 3F = 0 to find E = 8 and F = 8/3.)
D = (6, 2): Z = 1,200E + 2,400F = 1,200(6) + 2,400(2) = 12,000 (we need to solve the
simultaneous equations of E - 3F = 0 and 10E + 15F = 90 to find E = 6 and F = 2.)
Since 12,000 < 14,400 < 16,000 < 36,800 < 22,800, the optimal solution is (E *, F*) = (6,
2) and the optimal objective function value is Z * = 12,000.
. Example 2.15: Use the simplex method to solve the linear program below:
Cj | 12 6 0 0 -M 0 |
--------------------------------------------------------------------------------------------
VIS | y1 y2 y3 y4 y5 y6 | SQ
--------------------------------------------------------------------------------------------
0 y3 | 2 4 1 0 0 0 | 16 16/2 = 8
-M y5 | 5 3 0 -1 1 0 | 15 15/5 = 3
0 y6 | 1 0 0 0 0 1 | 5 5/1 = 5
--------------------------------------------------------------------------------------------
Zj | -5M -3M 0 M -M 0 | -15M
Z j - Cj |-5M-12 -3M-6 0 M 0 0 |
The current solution (y1, y2) = (0, 0) is not optimal since -5M - 12 < 0 and -3M - 6 < 0.
Given that -5M - 12 is the most negative number in the “Z j - Cj” row and 3 < 5 < 8, the
pivot column is “y1” (“y1“ is the entering variable), the pivot row is “y5” (“y5“ is the
leaving variable), and the pivot number is 5. The new simplex tableau is presented below:
Cj | 12 6 0 0 -M 0 |
--------------------------------------------------------------------------------------------
VIS | y1 y2 y3 y4 y5 y6 | SQ
--------------------------------------------------------------------------------------------
0 y3 | 0 14/5 1 2/5 -2/5 0 | 10 10/(2/5) = 25
12 y1 | 1 3/5 0 -1/5 1/5 0 | 3
0 y6 | 0 -3/5 0 1/5 -1/5 1 | 2 2/(1/5) = 10
--------------------------------------------------------------------------------------------
Zj | 12 36/5 0 -12/5 12/5 0 | 36
Z j - Cj | 0 6/5 0 -12/5 M+12/5 0 |
The current solution (y1, y2) = (3, 0) is not optimal since -12/5 < 0. Given that -12/5 is the
only negative number in the “Z j - Cj” row and 10 < 25, the pivot column is “y4” (“y4“ is the
entering variable), the pivot row is “y6” (“y6“ is the leaving variable), and the pivot number
is 1/5. The new simplex tableau follows:
Cj | 12 6 0 0 -M 0 |
--------------------------------------------------------------------------------------------
VIS | y1 y2 y3 y4 y5 y6 | SQ
--------------------------------------------------------------------------------------------
0 y3 | 0 4 1 0 0 -2 | 6 6/4 = 1.5
12 y1 | 1 0 0 0 0 1 | 5
0 y4 | 0 -3 0 1 -1 5 | 10
19
--------------------------------------------------------------------------------------------
Zj | 12 0 0 0 0 12 | 60
Z j - Cj | 0 -6 0 0 M 12 |
The current solution (y1, y2) = (5, 0) is not optimal since -6 < 0. Given that -6 is the only
negative number in the “Z j - Cj” row and 1.5 is the only positive ratio, the pivot column is
“y2” (y2 is the entering variable), the pivot row is “y3” (y3 is the leaving variable), and the
pivot number is 4. The new simplex tableau follows:
Cj | 12 6 0 0 -M 0 |
--------------------------------------------------------------------------------------------
VIS | y1 y2 y3 y4 y5 y6 | SQ
--------------------------------------------------------------------------------------------
6 y2 | 0 1 1/4 0 0 -1/2 | 3/2
12 y1 | 1 0 0 0 0 1 | 5
0 y4 | 0 -3 3/4 1 -1 7/2 | 29/2
--------------------------------------------------------------------------------------------
Zj | 12 6 3/2 0 0 9 | 69
Z j - Cj | 0 0 3/2 0 M 9 |
The current solution (y1, y2) = (5, 1.5) is optimal since all the numbers in the “Z j - Cj” row
are positive or zero. In conclusion, the optimal solution is (y1 *, y2*) = (5, 1.5) and the
optimal objective function value is Z * = 69.
. Example 2.16: Use the simplex method to solve the following linear program:
Z = 10A + 20B + MC + 0D + ME + 0F
A+ B+ C + 0D + 0E + 0F = 50
A+ 0B + 0C - D + E + 0F = 20
0A + B+ 0C + 0D + 0E + F = 40
Cj | 10 20 M 0 M 0 |
-----------------------------------------------------------------------------------------------
VIS | A B C D E F | SQ
-----------------------------------------------------------------------------------------------
M C | 1 1 1 0 0 0 | 50 50/1 = 50
M E | 1 0 0 -1 1 0 | 20 20/1 = 20
0 F | 0 1 0 0 0 1 | 40
-----------------------------------------------------------------------------------------------
Zj | 2M M M -M M 0 | 70M
Z j - Cj | 2M-10 M-20 0 -M 0 0 |
The current solution (A, B) = (0, 0) is not optimal since 2M - 10 > 0 and M - 20 > 0. Given
that 2M - 10 is the most positive number in the “Z j - Cj” row and 20 < 50, the pivot column
is “A” (“A” is the entering variable), the pivot row is “E” (“E” is the leaving variable), and
the pivot number is 1. The new simplex tableau follows:
Cj | 10 20 M 0 M 0 |
20
-----------------------------------------------------------------------------------------------
VIS | A B C D E F | SQ
-----------------------------------------------------------------------------------------------
M C | 0 1 1 1 -1 0 | 30 30/1 = 30
10 A | 1 0 0 -1 1 0 | 20
0 F | 0 1 0 0 0 1 | 40
-----------------------------------------------------------------------------------------------
Zj | 10 M M M-10 -M+10 0 | 30M+200
Z j - Cj | 0 M-20 0 M-10 -2M+10 0 |
The current solution (A, B) = (20, 0) is not optimal since M - 20 > 0 and M - 10 > 0. Given
that M - 10 is the most positive number in the “Z j - Cj” row and 30 is the only positive ratio,
the pivot column is “D” (“D” is the entering variable), the pivot row is “C” (“C” is the
leaving variable), and the pivot number is 1. The new simplex tableau follows:
Cj | 10 20 M 0 M 0 |
-----------------------------------------------------------------------------------------------
VIS | A B C D E F | SQ
-----------------------------------------------------------------------------------------------
0 D | 0 1 1 1 -1 0 | 30
10 A | 1 1 1 0 0 0 | 50
0 F | 0 1 0 0 0 1 | 40
-----------------------------------------------------------------------------------------------
Zj | 10 10 10 0 0 0 | 500
Z j - Cj | 0 -10 10-M 0 -M 0 |
The current solution (A, B) = (50, 0) is optimal since all the numbers in the “Z j - Cj” row are
negative or zero. In conclusion, the optimal solution is (A *, B*) = (50, 0) and the optimal
objective function value is Z * = 500.
. Example 2.17: Run Solver to solve the LP for the transportation problem formulated in Example 2.9. Be
sure to include a copy of the Answer Report and interpret the key results.
Variable Cells
Cell Name Original Value Final Value Integer
$B$35 BN 0 12000 Contin
$C$35 BC 0 0 Contin
$D$35 BL 0 8000 Contin
$E$35 ON 0 0 Contin
$F$35 OC 0 7000 Contin
$G$35 OL 0 8000 Contin
Constraints
Cell Name Cell Value Formula Status Slack
$H$37 LHS 20000 $H$37<=$J$37 Binding 0
$H$38 LHS 15000 $H$38<=$J$38 Binding 0
$H$39 LHS 12000 $H$39>=$J$39 Binding 0
$H$40 LHS 7000 $H$40>=$J$40 Binding 0
$H$41 LHS 16000 $H$41>=$J$41 Binding 0
21
It is seen that the optimal solution is (BN *, BC*, BL*, ON*, OC*, OL*) = (12,000, 0, 8,000, 0,
7,000, 8,000) and the optimal objective function value is Z * = 103,000. In other words, the
optimal shipping schedule is as follows to minimize the total transportation cost at
$103,000:
From To Quantity
--------------------------------------------
BR NY 12,000
BR CH 0
BR LA 8,000
OM NY 0
OM CH 7,000
OM LA 8,000
--------------------------------------------
. There are four special types of linear programs and each of them is graphically illustrated below: