Appendix Linear Programming

Appendix 1
Appendix: Linear Programming

Solution Methods
Introduction Unbalanced Problems

Simplex Method Degeneracy
Simplex Maximization Solutions Assignment Method
Simplex Minimization Solutions Review and Discussion Questions
Interpreting Simplex Solutions
Problems
Postoptimality Analysis
Unusual Features of Some LP Problems Computer Solutions
Transportation Method Selected Bibliography
Characteristics of Transportation Problems Answers to Odd-Numbered Problems
Solution Procedures of the Transportation Method
Although the graphical solution method that was presented in Chapter 8 is a useful learn-
ing device for those with limited knowledge of linear programming (LP), other meth-
ods are used daily in the worlds of business and government to solve real LP problems.
Preeminent among these are the simplex, transportation, and assignment methods.
SIMPLEX METHOD
Simplex does not mean simple. Nevertheless, the simplex method is used term after
term by thousands of students just like you, and you can do it too. The main thing is
to keep your eye on the big picture: Understand the overall procedures of the simplex
method because it is easy to get bogged down in the nitty-gritty details and lose sight
of the general process. Pay attention to the small details; this is necessary. But also
constantly keep in mind the overall process.
The best way to learn the simplex method is to use it to work LP problems. “Ex-
perience is the best teacher” certainly applies here.
Simplex Maximization Solutions

Table 1 presents the steps in the simplex method. Read them carefully, and don’t worry
too much about being confused. Remember, get the big picture. Working through an
example will demonstrate the meaning of each of these steps. Example 1 uses the sim-
plex method to solve Problem LP-1 from Chapter 8 while methodically following the
steps of the simplex method. Work through every step in this example meticulously;
this is absolutely necessary to an understanding of the method. Notice that the step
numbers in this example correspond to the step numbers in Table 1. You will not need
a calculator, because calculators compute in decimals that require rounding or trun-
cating. These sources of error are unacceptable in the simplex method, and therefore
all calculations are carried out in fractions.
1
2 Appendix: Linear Programming Solution Methods
Table 1 Steps in the Simplex Solution Method
1. Formulate the objective and constraint functions.

2. Add slack variables to convert each constraint to an equality ().
3. Add artificial variables to constraints that were originally or to produce a starting solution.
4. Set up the first tableau, starting solution.
5. Check solution for optimality. If optimal, stop. If not, continue.
6. Select a variable to enter to improve the solution.
7. Select a variable to leave the solution.
8. Perform row operations to complete the new solution.
9. Return to Step 5 and continue until optimality is achieved.
Example 1 Simplex Solution of LP-1, a Maximization Problem
As part of its strategic planning process, the Precision Manufacturing Company must determine
the mix of its products to be manufactured next year. The company produces two principal prod-
uct lines for the commercial construction industry, a line of powerful portable circular saws and
a line of precision table saws. The two product lines share the same production capacity and are
sold through the same sales channels. Although some product variety does exist within each prod-
uct line, the average profit is $900 for each circular saw and $600 for each table saw. The pro-
duction capacity is constrained in two ways, fabrication and assembly capacity. A maximum of
4,000 hours of fabrication capacity is available per month, and each circular saw requires two
hours and each table saw requires one hour. A maximum of 5,000 hours of assembly capacity is
available per month, and each circular saw requires one hour and each table saw requires two
hours. The marketing department estimates that there is a maximum market demand next year of
a total of 3,500 saws per month for both product lines combined. How many circular saws and
how many table saws should be produced monthly next year to maximize profits?
1. Formulate the objective and constraint functions.
LP-1 was formulated in Chapter 8 as:
Max Z 900X1 600X2
2X1 X2 4,000 (fabrication—hours)
X1 2X2 5,000 (assembly—hours)
X1 X2 3,500 (market—saws)
where:
X1 number of circular saws to be manufactured per month
X2 number of table saws to be manufactured per month
2. Add slack variables to convert each constraint to an equality (ⴝ):
a. 2X1 X2 4,000 (fabrication—hours)
Note that the left-hand side of the expression is less than or equal to the right-hand side
(RHS). If the expression is to be an equality (), something must be added to the left-
hand side to increase its value up to the level of the RHS. We shall add a slack variable
S1 to take up the slack between the value of the left-hand side and the RHS. S1 will take
Appendix: Linear Programming Solution Methods 3
on the value of zero if the left-hand side exactly equals 4,000 and a value of 4,000 if X1
and X2 equal zero. When X1 and X2 take on values larger than zero, the value of S1 will
decrease accordingly so that the left-hand side of the expression exactly equals 4,000:
2X1 X2 S1 4,000
Note the subscript 1 in S1 denotes that S1 is the slack variable for the first constraint. When
we proceed with and complete the simplex solution, and S1 takes on some specific value
at the end, we shall automatically know that S1 belongs to the first constraint—unused fab-
rication hours. Similarly, S2 will belong to the second constraint—unused assembly hours.
b. The second constraint is converted to an equality by adding a slack variable S2 to the
left-hand side:
X1 2X2 S2 5,000 (assembly—hours)
c. The third constraint is converted to an equality by adding a slack variable S3 to the left-
hand side:
X1 X2 S3 3,500 (market—saws)
We now have this LP problem:
Max Z 900X1 600X2
2X1 X2 S1 4,000
X1 2X2 S2 5,000
X1 X2 S3 3,500
3. Add artificial variables to constraints that were originally 傶 or ⴝ to produce a starting
solution. Since the constraints in this problem were all , no artificial variables are required.
4. Set up the first tableau, starting solution.
A tableau simply means a table. Each solution will be a tableau.
a. First, let all variables appear in the objective function and constraint functions. This is
achieved by assigning zero coefficients to all variables not appearing in these expres-
sions:
Max Z 900X1 600X2 0S1 0S2 0S3
2X1 X2 S1 0S2 0S3 4,000
X1 2X2 0S1 S2 0S3 5,000
X1 X2 0S1 0S2 S3 3,500
b. Now put the problem above into the tableau format. The format is achieved by entering
the set of coefficients for all the variables into the framework below:
First Tableau
C 900 600 0 0 0
SOL RHS X1 X2 S1 S2 S3 ␾
0 S1 4,000 2 1 1 0 0
0 S2 5,000 1 2 0 1 0
0 S3 3,500 1 1 0 0 1
Z 0 0 0 0 0 0
(C ⴚ Z) 900 600 0 0 0
c. Constraints. The RHS value of each constraint is placed in the RHS column, and the
coefficients of the variables in the constraint are placed under the variable columns.
d. Objective function. The coefficients of the variables in the objective function are placed
in Row C above the appropriate variable columns.
e. SOL column. Note that the shaded matrix under the S1, S2, and S3 columns is an iden-
tity matrix, a NW to SE diagonal of ones with all other elements zero. The variables
over the columns of the identity matrix are always in the SOL column.
SOL S1 S2 S3
S1 1 0 0
S2 0 1 0
S3 0 0 1
This condition will always hold: A variable found in the SOL column will have a one
at the intersection of its row and column, and all other elements in its column will be
zero. This condition holds for all tableaus. The variables in the SOL column are referred
to as the basic variables for that particular tableau. The number of basic variables is al-
ways equal to the number of constraints. All of the variables not listed in the SOL col-
umn are referred to as the nonbasic variables for that particular tableau. Exactly which
variables are basic and which variables are nonbasic will change with each different so-
lution (tableau) as the simplex method proceeds.
f. C column. The zeros in the C column and opposite S1, S2, and S3 are taken from Row
C elements above S1, S2, and S3. This condition will always hold: The values in the C
column are the coefficients in the objective function that correspond to the variables
found in the SOL column of the tableau (i.e., the basic variables).
g. Row Z. The elements in Row Z are all computed. Since the elements in the C column
were zero, all the elements in Row Z were zero; however, the elements in the C column
will not always be zero. Each element in Row Z is computed as follows:
ZRHS (0)(4,000) (0)(5,000) (0)(3,500) 0
ZX1 (0)(2) (0)(1) (0)(1) 0
ZX2 (0)(1) (0)(2) (0)(1) 0
ZS1 (0)(1) (0)(0) (0)(0) 0
ZS2 (0)(0) (0)(1) (0)(0) 0
ZS3 (0)(0) (0)(0) (0)(1) 0
This step will become clearer in subsequent tableaus.
h. Row (C Z). These values are computed by subtracting each element in Row Z from
its counterpart in Row C:
(C Z)X1 900 0 900
(C Z)X2 600 0 600
(C Z)S1 00 0
(C Z)S2 00 0
(C Z)S3 00 0
i. The starting solution. This completes the explanation of the first tableau. The solution
of the first tableau is:
X1 0 circular saws to be manufactured per month
X2 0 table saws to be manufactured per month
S1 4,000 unused fabrication hours per month

S2 5,000 unused assembly hours per month
S3 3,500 unsatisfied market demand for saws per month
Z $0 profits per month
To read the solution from the simplex tableau you need only look at the SOL and RHS
columns. The variables in the SOL column (basic variables) are equal to the correspond-
ing values in the RHS column. All other variables (nonbasic variables) are equal to zero.
Therefore, from the SOL and RHS columns, S1 4,000, S2 5,000, S3 3,500, and
Z 0. Because X1 and X2 are not listed in the SOL column (i.e., they are nonbasic vari-
ables in this tableau), X1 0 and X2 0.
5. Check solution for optimality. If optimal, stop. If not, continue.
Solutions are optimal when all the values in Row (C Z) are either zero or negative. When
any of the values in this row are nonzero positive numbers, the solution can be improved by
continuing and the solution is not optimal. In our first tableau of LP-1, both X1 and X2 have
Row (C Z) elements that are nonzero and positive. It is therefore not optimal, and we must
continue to the next tableau.
6. Select a variable to enter to improve the solution.
Select the variable to enter that has the largest positive element in Row (C Z). X1 will en-
ter since its Row (C Z) element is the largest positive value in Row (C Z).
7. Select a variable to leave the solution:
First Tableau (Continued)

C 900 600 0 0 0
0 S1 4,000 2 1 1 0 0 4,000/2 2,000 , Leaving Variable
0 S2 5,000 1 2 0 1 0 5,000/1 5,000 (smallest nonnegative)
0 S3 3,500 1 1 0 0 1 3,500/1 3,500
Z 0 0 0 0 0 0
(C ⴚ Z) 900 600 0 0 0
+
Entering Variable
(largest positive)
The variable to leave this starting solution is determined by entering the column of the en-
tering variable, X1, dividing the number of each row into its RHS value, and recording this
value in the column to the right. The leaving variable has the smallest nonnegative value.
If zero values or ties occur in the column, do not worry. In the case of a tie, arbitrarily se-
lect one of the tying variables to leave, and in the case of a zero, let that variable leave. (Re-
member that division by zero is undefined.)
8. Perform row operations to complete the new solution.
This is perhaps the most confusing point in the simplex method. Table 2 lists the steps in
performing row operations. These steps will be followed to construct the second tableau.
a. Identify the pivot element. It is found at the intersection of the entering-variable column
(X1) and the leaving-variable row (S1): The pivot element is 2 .
b. Divide the pivot row, element by element, by the pivot element. Enter this new row in
the next tableau: The first constraint row (4,000 2 1 1 0 0) is divided by 2, and
the result is entered in the second tableau in the first-row position. The variable in the
SOL column in this row is the new entering variable X1, and its coefficient (900) in Row
C is entered in Column C.
Second Tableau
C 900 600 0 0 0
900 X1 2,000 1 /
12 /
12 0 0
0 S2
0 S3
Z
(C ⴚ Z)
c. Convert the other elements in the pivot column (X1) to zero. Transform the second row
in the first tableau (5,000 1 2 0 1 0) by multiplying the new row in the second
tableau (2,000 1 1/2 1/2 0 0) by 1 (the negative of the second row’s pivot col-
umn value) and adding this transitional row to the second row of the first tableau:
Multiply the row first entered in the second tableau by 1:
1(2,000) 1(1) 1(1/2) 1(1/2) 1(0) 1(0)
This gives values of
2,000 1 1/2 1/2 0 0
which are added to the row being transformed
5,000 1 2 0 1 0
to get values of
3,000 0 32
/ 1/2 1 0
The third row in the first tableau (3,500 1 1 0 0 1) is now transformed by mul-
tiplying the row first entered in the new tableau (2,000 1 1/2 1/2 0 0) by 1 (the
negative of the third row’s pivot column value) and adding this transitional row to the
third row of the first tableau:
Multiply the row first entered in the second tableau by 1:
1(2,000) 1(1) 1(1/2) 1(1/2) 1(0) 1(0)
2,000 1 1/2 1/2 0 0
which are added to the row being transformed
3,500 1 1 0 0 1
to get values of
1,500 0 12
/ 1/2 0 1
d. These rows are entered into the second tableau in the second- and third-row positions.
The variables in the SOL column for these rows do not change:
Second Tableau (Continued)
C 900 600 0 0 0
900 X1 2,000 1 /
12 / 12 0 0
0 S2 3,000 0 /
32 1/2 1 0
0 S3 1,500 0 1/2 1/2 0 1
Z 1,800,000 900 450 450 0 0
(C ⴚ Z) 0 150 450 0 0
e. Row Z for each column is computed by summing the products of values in Column C
and that column as follows:
ZRHS (900)(2,000) 0(3,000) 0(1,500) 1,800,000
ZX1 (900)(1) (0)(0) (0)(0) 900
ZX2 (900)(1/2) (0)(3/2) (0)(1/2) 450
ZS1 (900)(1/2) (0)(1/2) (0)(1/2) 450
ZS2 (900)(0) (0)(1) (0)(0) 0
ZS3 (900)(0) (0)(0) (0)(1) 0
Row (C Z) is again computed by subtracting each element in Row Z from its coun-
terpart in Row C.
f. The solution to the second tableau is:
X1 2,000 circular saws to be manufactured per month
X2 0 table saws to be manufactured per month
S1 0 unused fabrication hours per month
S2 3,000 unused assembly hours per month
S3 1,500 unsatisfied market demand for saws per month
Z $1,800,000 profits per month
The values of X1, S2, S3 (basic variables), and Z are found in column RHS of the sec-
ond tableau. X2 and S1 are not in the SOL column (they are nonbasic variables); there-
fore they both equal zero. Does this solution look familiar? It should, because it is iden-
tical to Point C in Example 8.4 of the graphical method.
9. Return to Step 5 and continue until optimality is achieved:
Steps 5, 6, and 7. Check solution for optimality. If optimal, stop. If not, continue. Se-
lect a variable to enter the solution. Select a variable to leave the solution:
Second Tableau (Continued)

C 900 600 0 0 0
900 X1 2,000 1 /
12 /
12 0 0 2,000/1/2 4,000
0 S2 3,000 0 3/2 1/2 1 0 3,000/3/2 2,000 , Leaving Variable
0 S3 1,500 0 1/2 1/2 0 1 1,500/1/2 3,000 (smallest nonnegative)
Z 1,800,000 900 450 450 0 0
(C ⴚ Z) 0 150 450 0 0
+
Entering Variable
(largest positive)
The second tableau is not optimal since all Row (C Z) elements are not zero or negative.
X2 enters and, because the value for the second row is the smallest nonnegative, S2 leaves.
10. Perform row operations to complete the solution. (See Table 2.)
a. The pivot element is 3/2 because that element is at the intersection of the entering-variable
column and the leaving-variable row.
b. Divide the pivot row (second constraint row) of the second tableau through by the pivot
element (3/2). Enter this new row into the third tableau in the same second-row position:
Third Tableau
C 900 600 0 0 0
900 X1
600 X2 2,000 0 1 1/3 /
23 0
0 S3
Z
(C ⴚ Z)
c. Multiply this new row by a negative 1/2 to transform the first constraint row in the sec-
ond tableau:
1/2(2,000) 1/2(0) 1/2(1) 1/2(1/3) 1/2(2/3) 1/2(0)
1,000 0 1/2 16
/ 1/3 0
which are added to the first constraint row of the second tableau
2,000 1 12
/ 12
/ 0 0
to give the new first constraint row of the third tableau
1,000 1 0 23
/ 1/3 0
Multiply the first row entered in the third tableau by 1/2:
1/2(2,000) 1/2(0) 1/2(1) 1/2(1/3) 1/2(2/3) 1/2(0)
1,000 0 1/2 16
/ 1/3 0
which are added to the third constraint row of the second tableau
1,500 0 12
/ 1/2 0 1
to give the new third constraint row of the third tableau
500 0 0 1/3 1/3 1
d. These rows are now entered into the first- and third-row positions of the third tableau:
Third Tableau (Continued)
C 900 600 0 0 0
900 X1 1,000 1 0 / 23 /13 0
600 X2 2,000 0 1 1/3 2/3 0
0 S3 500 0 0 1/3 /3
1 1
Z 2,100,000 900 600 400 100 0
(C ⴚ Z) 0 0 400 100 0
e. Compute Rows Z and (C Z) to complete the third tableau. Row Z is computed as fol-
lows:
ZRHS (900)(1,000) (600)(2,000) (0)(500) 2,100,000
ZX1 (900)(1) (600)(0) (0)(1) 900
ZX2 (900)(0) (600)(1) (0)(0) 600

ZS1 (900)(2/3) (600)(1/3) (0)(1/3) 400
ZS2 (900)( / )
13 (600)( / ) (0)( / )
23 13 100
ZS3 (900)(0) (600)(0) (0)(1) 0
Row (C Z) is again computed by subtracting each element in Row Z from its coun-
terpart in Row C.
f. The solution to the third tableau is:
X1 1,000 circular saws to be manufactured per month
X2 2,000 table saws to be manufactured per month
S1 0 unused fabrication hours per month
S2 0 unused assembly hours per month
S3 500 unsatisfied market demand for saws per month
Z $2,100,000 profits per month
Does this solution look familiar? It should, because it is identical to Point B in Exam-
ple 8.4 of the graphical method.
11. Return to Step 5 and continue until optimality is achieved. Check solution for opti-
mality. If optimal, stop. If not, continue. The third tableau is optimal because all the ele-
ments in Row (C Z) are either negative or zero.
The graphical solution of maximization Problem LP-1, Example 8.4, identified three
solution points—A, B, and C. Each of these solutions was first identified as an inter-
section of the constraints, and then each was substituted into the objective function to
Table 2 Steps in Performing Row Operations
1. Identify the pivot element in the present tableau, which is found at the intersection of the column of
the entering variable and the row of the leaving variable. Circle this element. Row operations convert
the pivot column in this present tableau, element by element, to a new column in the new tableau. This
new column always has these features: The element that is the pivot element in the present tableau will
be a one in the new tableau, and all other elements in that column in the new tableau will be zero.
2. Convert the pivot element in the present tableau to a one in the new tableau by dividing the entire
pivot row in the present tableau by the pivot element (element by element). This new row with a
one in the pivot element is entered into the new tableau in the same row position as the pivot row
of the present tableau.
3. Next, convert all other elements in the pivot column of the present tableau to zero in the new
tableau. This is done by performing a separate operation on each remaining row of the present
tableau. This operation involves the development of a special transitional row that is added to each
remaining row in the present tableau. Each row in the present tableau for which we wish to convert
its pivot column element to zero requires its own unique transitional row. This transitional row is
developed by first determining the value of the element that we want to transform to zero. We then
take the negative of the value of this element and multiply it by the new row obtained in Step 2
above, element by element. When this transitional row is added to the row of the present tableau, a
new row results that has a zero in the pivot column. This row is entered in the new tableau in the
same row position as it occupied in the present tableau before it was transformed.
determine a value of Z. Finally, the optimal solution (maximum profit) was selected.
The simplex method solution to LP-1 follows this same general process with one ex-
ception. The first tableau begins with Z 0, and each subsequent tableau methodically
exhibits higher values of Z. You can be assured that each tableau will exhibit progres-
sively higher profits. This progression to better and better solutions is the only concep-
tual departure of the simplex method from the general process of the graphical method.
Simplex Minimization Solutions

The graphical solution of minimization Problem LP-2, Example 8.5, identified three
solution points—A, B, and C. The optimal solution to the problem was then determined
by investigating Points A, B, and C. The simplex solution to this problem begins the
first tableau with a very large value of Z. Subsequent tableaus exhibit progressively
lower values for Z until optimality is achieved. This progression from high to low val-
ues of Z is characteristic of the simplex solution of minimization LP problems.
There are only two basic differences between maximization and minimization LP
problem solutions with the simplex method:
1. Minimization LP problems are more likely to have and constraints, although
the procedures for treating them apply to both minimization and maximization LP
problems.
2. Minimization LP problems have different objective functions; Z is minimized.
In either minimization or maximization problems, and constraints are accommo-
dated by adding artificial variables to these constraints. For example, in the case of
constraints:
1. X1 2X2 500
2. X1 2X2 500 S1
3. X1 2X2 S1 500
4. X1 2X2 A1 S1 500
Note that in Step 2 above a slack variable (S1) is added to the right-hand side, which
must always be less than or equal to the left-hand side. The addition of S1 to the smaller
side of the expression allows us to convert the to . In Step 3 the S1 is moved to the
left-hand side by subtracting S1 from both sides. In Step 4 an artificial variable (A1) is
added to the left-hand side. Why do we do this? The only reason is to get a starting sim-
plex solution. Remember when we said earlier that a requirement for each tableau was
that the variable with a column which has a one in the first row and zeros for all other
elements in its column is the variable that must go in the SOL column in the first row?
What happens if no variable exists that has a column meeting this requirement? This is
exactly the situation we have with or constraints. A constraint has a 1 coeffi-
cient for S1, and this does not meet the requirement. Similarly, an constraint does not
have a slack variable (as we shall soon see); therefore we shall not be able to meet the
requirement here either. When such conditions exist, we must add an artificial variable
to or rows to meet the requirement and obtain a starting solution (complete the first
tableau). The artificial variables appear in the SOL column of the first tableau and are
then methodically driven from the solution in subsequent tableaus. The artificial vari-
ables have absolutely no meaning, and we shall not be concerned with them again.
When constraints occur, an artificial variable must also be added. For example:
3X1 2X2 1,000
3X1 2X2 A2 1,000
Again, the purpose of A2 is to achieve a starting simplex solution; A2 will have no sub-
sequent meaning.
The second basic difference is accommodated by converting objective functions
from the min to the max form. For example:
Min Z 5X1 3X2 becomes Max Z 5X1 3X2
This is achieved by multiplying each term of the Min Z objective function by a 1.
Minimization LP problems, after this conversion, are then solved as maximization prob-
lems. The exact same tableau procedures apply. After this conversion, slack variables
and artificial variables are added to the objective function as required. Slack variables
are assigned zero coefficients in the objective function as before. But what about the
artificial variables? What coefficients should be assigned to the artificial variables?
Would you believe a M, where the M is a very large number? Now, what the M ac-
tually represents is subject to speculation, but legend has it that Harvey Wagner first
used the big M method in LP studies at the Mercury Motors Division of the Ford Mo-
tor Company many years ago. Regardless of its origin, each artificial variable is always
assigned a M coefficient in the objective function in either max or min LP problems
when the objective function conversion described above is used.1 This M in Row C
of the simplex tableau avoids an artificial variable entering back into the solution, since
(C Z) will always be zero or negative because the C value is M . . . , a negative
very large number, say, minus infinity. Since whatever is subtracted from minus infin-
ity is either negative or zero, the artificial variable will never reenter the solution.
Example 2 solves Problem LP-2 from Chapter 8, a minimization problem, using
the simplex method.
1
If this conversion of the objective function is not used, the objective function is left in its original form and the
rule for the entering variable for minimization problems becomes: The entering variable is the one with the most
negative number in Row (C Z). The artificial variables must then be assigned M coefficients in the objective
function. All other procedures remain the same as in maximization problems. We prefer the conversion described
above because the procedures are all exactly the same for both maximization and minimization problems, and
we shall follow this convention throughout.
Example 2 Simplex Solution of LP-2, a Minimization Problem
The Gulf Coast Foundry is developing a long-range strategic plan for buying scrap metal for its
foundry operations. The foundry can buy scrap metal in unlimited quantities from two sources,
Atlanta (A) and Birmingham (B), and it receives the scrap daily in railroad cars. The scrap is
melted down, and lead and copper are extracted for use in the foundry processes. Each railroad
car of scrap from Source A yields 1 ton of copper and 1 ton of lead and costs $10,000. Each rail-
road car of scrap from Source B yields 1 ton of copper and 2 tons of lead and costs $15,000. If
the foundry needs at least 21/2 tons of copper and at least 4 tons of lead per day for the foresee-
able future, how many railroad cars of scrap should be purchased per day from Source A and
Source B to minimize the long-range scrap metal cost? LP-2 was formulated in Chapter 8 as:
Min Z 10,000X1 15,000X2
X1 2X2 4 (lead—tons)
X1 X2 2 / (copper—tons)
12
where:
X1 carloads of scrap purchased from Source A per day
X2 carloads of scrap purchased from Source B per day
1. Add slack variables to constraints to convert from 傶 to ⴝ:

Min Z 10,000X1 15,000X2
X1 2X2 S1 4
X1 X2 S2 21/2
2. Multiply objective function by ⴚ1 to convert to a maximization problem:
Max Z 10,000X1 15,000X2
X1 2X2 S1 4
X1 X2 S2 21/2
3. Add artificial variables to constraints to obtain a starting solution, and include all vari-
ables in all functions:
Max Z 10,000X1 15,000X2 MA1 MA2 0S1 0S2
X1 2X2 A1 0A2 S1 0S2 4
X1 X2 0A1 A2 0S1 S2 21/2
4. Place in first tableau and solve:
First Tableau
C 10,000 15,000 M M 0 0
SOL RHS X1 X2 A1 A2 S1 S2 ␾ Leaving
M A1 4 1 2 1 0 1 0 4/2 2 , Variable
M A2 21/2 1 1 0 1 0 1 21/2/1 21/2 (smallest
nonnegative)
Z ⴚ61/2M 2M 3M M M M M
(C ⴚ Z) 2M 3M
10,000 15,000 0 0 M M
+
Entering Variable
(largest nonnegative)
2 1/2 1 1/2 0 1/2 0
21/2 1 1 0 1 0 1
1/2 1/2 0 1/2 1 1/2 1
Second Tableau
C 10,000 15,000 M M 0 0
SOL RHS X1 X2 A1 A2 S1 S2 ␾
Leaving
15,000 X2 2 /
12 1 / 12 0 1/2 0 2/1/2 4 Variable
M A2 1/2 1/2 0 1/2 1 1/2 1 1/2/1/2 1 , (smallest
nonnegative)
ⴚ1/2M 1/2M /M 12 1/2M
Z ⴚ30,000 7,500 15,000 7,500 M 7,500 M
/M 12 /M 32 /M
12
(C ⴚ Z) 2,500 0 7,500 0 7,500 M

+
Entering Variable
(largest nonnegative)
1/2 1/2 0 12 / 1 1/2 1

2 1/2 1 12 / 0 1/2 0
1
1 /2 0 1 1 1 1 1
Third Tableau
C 10,000 15,000 M M 0 0
15,000 X2 1/12 0 1 1 1 1 1
10,000 X1 1 1 0 1 2 1 2
Z ⴚ32,500 10,000 15,000 5,000 5,000 5,000 5,000
M M
(C ⴚ Z) 0 0 5,000 5,000 5,000 5,000
The third tableau is optimal because all elements in Row (C Z) are zero or negative.
5. Interpret the solution:
The solution is deduced from the SOL and RHS columns of the last tableau. All variables
that do not appear in the SOL column are equal to zero:
X1 1 X2 11/2 S1 0 S2 0 Z 32,500
The Gulf Coast Foundry should purchase 1 carload of scrap per day from Source A and 11/2
carloads of scrap per day from Source B. The total daily scrap cost will be $32,500, and no
excess lead or copper above the minimum requirements will result. Note that the simplex
method does not guarantee whole number (integer) answers. This is ordinarily not a serious
difficulty. In this problem, for example, an average of 11/2 carloads of scrap from Source B
could be accommodated by either 3 carloads one day and none the next or two carloads one
day and one the next.
Pay particular attention to the conversion of constraints to the first tableau form
by adding artificial variables and subtracting slack variables. Note also that the sub-
scripts of the artificial and slack variables correspond to the order of the constraints.
A1 and S1 belong to the first constraint, and A2 and S2 belong to the second.
The key complication in minimization problems is the more frequent inclusion of
artificial variables. Negative M’s, negative very large numbers, appear in the C row and
column and consequently in Row Z and Row (C Z). If these M’s are treated as any
other very large number and are added, subtracted, and multiplied while the appropri-
ate signs are observed, the minimization problems are as straightforward to solve as
the maximization problems.
Interpreting Simplex Solutions

Example 1 used the simplex method to solve Problem LP-1. Let us now examine LP-1
and the last tableau from this example to determine what information is available to
POM decision makers. LP-1 was formulated as:
Max Z 900X1 600X2
2X1 X2 S1 4,000 (fabrication—hours)
where:
X1 number of circular saws to be manufactured per month
X2 number of table saws to be manufactured per month
S1 unused fabrication hours per month
S2 unused assembly hours per month
S3 unsatisfied market demand for saws per month
Z profits per month
The last tableau in the simplex solution to LP-1 in Example 1 was:
Third Tableau
C 900 600 0 0 0
900 X1 1,000 1 0 / 23 / 13 0
600 X2 2,000 0 1 1/3 2/3 0
0 S3 500 0 0 1/3 1/3 1
Z 2,100,000 900 600 400 100 0
(C ⴚ Z) 0 0 400 100 0
The solution to LP-1 is deduced as follows: X1, X2, S3 (all basic variables), and Z are
in the SOL column, and their values are shown in the RHS column. Because S1 and
S2 are not found in the SOL column (nonbasic variables), their values are zero. This
solution indicates that management should manufacture 1,000 circular saws and 2,000
table saws per month for a monthly profit of $2,100,000. All fabrication and assembly
production capacity would be used, and 500 additional saws could be sold to the mar-
ket. To check this, let’s examine the constraints of LP-1:
2X1 X2 S1 4,000 (fabrication—hours)
2(1,000) 2,000 S1 4,000
2,000 2,000 S1 4,000
S1 4,000 2,000 2,000
S1 0
The fabrication of circular saws (X1) and table saws (X2) has used up all the available
fabrication hours per month; therefore S1 0.
1,000 2(2,000) S2 5,000
1,000 4,000 S2 5,000
S2 5,000 1,000 4,000
S2 0
The manufacture of circular saws and table saws has used up all the available assem-
bly capacity per month; therefore S2 0.
1,000 2,000 S3 3,500
S3 3,500 1,000 2,000
S3 3,500
The sales of circular and table saws have fallen 500 saws short of completely satisfy-
ing the maximum monthly demand for saws; therefore S3 500.
This explains the solution to LP-1. There is, however, some additional informa-
tion in the last tableau that can be useful to operations managers. This information is
called shadow prices and is found in Row (C Z):
X1 X2 S1 S2 S3
(C ⴚ Z) 0 0 400 100 0
The zeros appear in the X1 and X2 columns because these variables are in the so-
lution column SOL in the last tableau. When nonzero shadow prices appear under X
variables in Row (C Z), these values indicate the change in Z as a result of forcing
one unit of an X variable into the solution. In maximization problems, the shadow
prices under X variables indicate how much Z would be reduced by the introduction
of one unit of the X variable into the solution. In minimization problems, the shadow
prices under X variables indicate how much Z would be increased by the introduction
of one unit of the X variable into the solution.
The value under an S variable in Row (C Z) represents the change in Z from a
one-unit change in a constraint’s RHS. S1 refers to the first constraint because its sub-
script is 1 and represents fabrication hours. Since S1 is not in the SOL column, S1
0, which means that all the monthly fabrication capacity is used. For example, assume
that management wanted to know the following information:
1. How much would monthly profits (Z) increase if we could find one more hour of
fabrication capacity per month (4,001 versus 4,000)?
2. How much would monthly profits (Z) decrease if we had one less hour of fabri-
cation capacity per month (3,999 versus 4,000)?
The answer to both of these questions is found in Row (C Z) and the S1 column:
$400. The new monthly profits would be $2,100,400 and $2,099,600, respectively.
The element in Row (C Z) and the S2 column indicates the change in Z if the
RHS of the second constraint changes by one unit:
1. How much would monthly profits (Z) change if we could find one more hour of
assembly capacity per month (5,001 versus 5,000)?
2. How much would monthly profits (Z) decrease if we had one less hour of assem-
bly capacity per month (4,999 versus 5,000)?
The answer to both of these questions is $100; the new monthly profits would be
$2,100,100 and $2,099,900, respectively.
How much would we be willing to pay to expand the market for our saw product
lines through advertising or promotion? The answer to this question—nothing—can be
found in Row (C Z) and the S3 column. This answer is also obvious from the solu-
tion to Problem LP-1. If we could sell 500 more saws in the market than we are presently
selling, we would not pay anything for more market demand.
An understanding of the shadow prices in Row (C Z) is valuable to manage-
ment. This information allows managers to evaluate whether resources (production ca-
pacity and market demand in this example) should be shifted from other products or
projects. If the cost of getting one unit of a resource is less than its shadow price, the
resource should be acquired.
Interpreting simplex solutions of LP minimization problems is essentially the same
as interpreting those of maximization problems. To demonstrate this similarity, let us ex-
amine Row (C Z) from the optimal tableau of Problem LP-2, a minimization problem:
X1 X2 A1 A2 S1 S2
(C ⴚ Z) 0 0 M M 5,000 5,000
5,000 5,000
The zeros appear under X1 and X2 because both X1 and X2 are in the solution. The
values under A1 and A2 have no meaning. The 5,000 under S1 means that if the lead
requirement is raised or lowered one unit (1 ton per day), Z will change by $5,000. If
the RHS of the first constraint were increased from 4 to 5, Z would increase from
$32,500 to $37,500. If the RHS were decreased from 4 to 3, Z would decrease from
$32,500 to $27,500. The meaning of the 5,000 under S2 is similarly the marginal im-
pact of 1 ton of copper on Z.
The minimization and maximization LP problems are interpreted exactly the same
way. It is usually helpful to have the original problem formulated and in front of you
when the last tableau is interpreted. The meanings of the variables in the solution and
the shadow prices are then easier to understand.
Now, to further develop your ability to interpret simplex solutions, let us move on
to a more realistic LP problem. Remember the Oklahoma Crude ingredient mix prob-
lem, Example 8.7, from Chapter 8. Example 3 presents this problem and its simplex
solution and fully interprets the meaning of the solution. Carefully reread the earlier
problem before you begin this example.
Example 3 Interpreting Simplex Solutions: The Oklahoma Crude Oil Company

Ingredient Mix Problem
The variable definitions of the Oklahoma Crude Oil Company problem are:
X1 thousands of gallons of Oklahoma crude to be purchased per month

X2 thousands of gallons of Texas crude to be purchased per month
X3 thousands of gallons of Kansas crude to be purchased per month
X4 thousands of gallons of New Mexico crude to be purchased per month
X5 thousands of gallons of Colorado crude to be purchased per month
S1 excess regular gasoline over minimum market requirement in thousands of gallons
S2 excess premium gasoline over minimum market requirement in thousands of gallons
S3 excess low-lead gasoline over minimum market requirement in thousands of gallons
S4 excess diesel fuel over minimum market requirement in thousands of gallons
S5 excess heating oil over minimum market requirement in thousands of gallons
S6 excess lubricating oil base over minimum market requirement in thousands of gallons
S7 unused Oklahoma crude supply in thousands of gallons
S8 unused Texas crude supply in thousands of gallons
S9 unused Kansas crude supply in thousands of gallons
S10 unused New Mexico crude supply in thousands of gallons
S11 unused Colorado crude supply in thousands of gallons
A1, A2, A3, A4, A5, and A6 no meaning
Min Z 200X1 140X2 150X3 180X4 120X5

0.4X1 0.3X2 0.3X3 0.2X4 0.3X5 5,000 (regular gasoline
market requirement)*
0.2X1 0.3X2 0.4X3 0.3X4 0.2X5 3,000 (premium gasoline
market requirement)
0.2X1 0.1X2 0.3X4 0.1X5 3,000 (low-lead gasoline
market requirement)
0.1X1 0.1X2 0.1X3 0.2X5 2,000 (diesel fuel
market requirement)
0.1X2 0.1X3 0.2X4 0.1X5 1,000 (heating oil
market requirement)
0.1X1 0.1X2 0.1X3 0.1X5 2,000 (lubricating oil base
market requirement)
X1 8,000 (Oklahoma crude supply)
X2 4,000 (Texas crude supply)
X3 5,000 (Kansas crude supply)
X4 3,000 (New Mexico crude supply)
X5 6,000 (Colorado crude supply)
*All requirements and supplies are in thousands of gallons.
A computer solution (last tableau) is presented in Table 3.

a. What should management do? In other words, to management decision makers, what is the
complete meaning of the values of the decision variables, slack variables, artificial variables,
and Z in the optimal solution?
b. What is the meaning of each element in Row (C Z)?
Solution
a. What should management do? What is the complete meaning of the values of the variables?
X1 8,000 (buy 8 million gallons of crude oil per month from Oklahoma)
X2 4,000 (buy 4 million gallons of crude oil per month from Texas)
X3 2,000 (buy 2 million gallons of crude oil per month from Kansas)
X4 1,3331/3 (buy 11/3 million gallons of crude oil per month from New Mexico)
X5 6,000 (buy 6 million gallons of crude oil per month from Colorado)
S1 2,0662/3 (2,066,667 gallons of excess regular gasoline will be supplied monthly)
S2 2,200 (2,200,000 gallons of excess premium gasoline will be supplied monthly)
S3 0 (no excess low-lead gasoline will be supplied monthly)
S4 600 (600,000 gallons of excess diesel fuel will be supplied monthly)
S5 4662/3 (466,667 gallons of excess heating oil will be supplied monthly)
18
Table 3 Optimal Tableau from Ingredient Mix Problem: Oklahoma Crude Oil Company
Appendix: Linear Programming Solution Methods

C 200 140 150 180 120 0 0 0 0 0 0 M M M M M M 0 0 0 0 0
X1 X2 X3 X4 X5 S1 S2 S3 S4 S5 S6 A1 A2 A3 A4 A5 A6 S7 S8 S9 S10 S11
120 X5 6,000 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
180 X4 1,333.33 0 0 0 1 0 0 0 3.33 0 0 0 0 0 3.33 0 0 0 0.67 0.33 0 0 0.33
0 S5 466.67 0 0 0 0 0 0 0 0.67 0 1 1 0 0 0.67 0 1 1 0.23 0.07 0 0 0.07
150 X3 2,000 0 0 1 0 0 0 0 0 0 0 10 0 0 0 0 0 10 1 1 0 0 1
0 S2 2,200 0 0 0 0 0 0 1 1 0 0 4 0 1 1 0 0 4 0.4 0.2 0 0 0.3
140 X2 4,000 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
200 X1 8,000 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0
0 S9 3,000 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 10 1 1 1 0 1
0 S4 600 0 0 0 0 0 0 0 0 1 0 1 0 0 0 1 0 1 0 0 0 0 0.1
0 S10 1,666.67 0 0 0 0 0 0 0 3.33 0 0 0 0 0 3.33 0 0 0 0.67 0.33 0 1 0.33
0 S1 2,066.67 0 0 0 0 0 1 0 0.67 0 0 3 1 0 0.67 0 0 3 0.03 0.07 0 0 0
Z ⴚ3,420,000 200 140 150 180 120 0 0 600 0 0 1,500 0 0 600 0 0 1,500 70 70 0 0 90
(C ⴚ Z) 0 0 0 0 0 0 0 600 0 0 1,500 M M M M M M 70 70 0 0 90
600 1,500
S6 0 (no excess lubricating oil base will be supplied monthly)

S7 0 (all Oklahoma crude oil available will be purchased monthly)
S8 0 (all Texas crude oil available will be purchased monthly)
S9 3,000 (3,000,000 gallons of Kansas crude oil will be available and not
purchased monthly)
S10 1,6662/3 (1,666,667 gallons of New Mexico crude oil will be available and not
purchased monthly)
S11 0 (all Colorado crude oil available will be purchased monthly)
A4, A5, and A6 no meaning
Z 3,420,000 ($3,420,000 crude oil cost per month will result)
b. Interpret Row (C Z):
1. The values in the A1, A2, A3, A4, A5, and A6 columns have no meaning.
2. The zeros in the X1, X2, X3, X4, and X5 columns because all of these variables are in
the solution.
3. The zeros in the S1, S2, S4, S5, S9, and S10 columns mean that a 1,000-gallon change in the
RHS of these constraints will not affect the monthly crude oil cost (Z), because each of
these slack variables is in the solution. For example, S1 2,0662/3 means that 2,066,667
gallons more than the 5,000,000 RHS of the regular gasoline minimum market requirement
is supplied. Therefore, raising or lowering the RHS a small amount will not affect Z.
Column Row (C ⴚ Z) Interpretation
S3 600 If the low-lead gasoline monthly market requirement were increased by
1,000 gallons, Z would increase $600.
S6 1,500 If the lubricating oil base monthly market requirement were increased
by 1,000 gallons, Z would increase $1,500.
S7 70 If the amount of Oklahoma crude available each month were increased
by 1,000 gallons, Z would decrease by $70.
S8 70 If the amount of Texas crude available each month were increased by
1,000 gallons, Z would decrease by $70.
S11 90 If the amount of Colorado crude available each month were increased
by 1,000 gallons, Z would decrease by $90.
Postoptimality Analysis
Postoptimality analysis, or sensitivity analysis as it is often called, manipulates the
elements of the last tableau of the simplex procedure to determine the sensitivity of
the solution to changes in the original problem. Here are some of the questions that
this analysis seeks to answer:
1. How will Z change if the RHS of any constraint changes? This question was ad-
dressed earlier in this section. The values in Row (C Z) under the slack vari-
ables provide this information.
2. If a decision variable is not in the optimal solution (equals zero), how will Z change
if one unit of the decision variable is forced into the solution (Xi 1)? This ques-
tion was also addressed earlier in this appendix. The values in Row (C Z) un-
der the decision variables provide this information.
3. Over what range can the RHS change and the shadow prices in Row (C Z) re-
main valid?
4. How will Z change if one of the coefficients of a decision variable in the objec-
tive function changes by one unit?
5. How will Z change if one of the coefficients of a decision variable in a constraint
changes by one unit?
These and other postoptimality questions are deduced from the optimal tableau, but the
methods of analysis required to answer Questions 3, 4, and 5 are beyond the scope of
this book. These questions assume that all other parts of the original problem remain
unchanged and that only the singular change under consideration occurs. Although this
is a popular topic with operations researchers, one rather obvious method for answer-
ing these and other postoptimality questions exists: Make the desired changes in the
original problem, input the new problem to the computer, and interpret the new results.
Unusual Features of Some LP Problems

Two linear programming situations deserve special attention: degeneracy and alternate
optimal solutions. Degeneracy is a condition in which there is a tie between two or
more leaving variables in any maximization or minimization simplex tableau. For ex-
ample, consider this problem:
Max Z 5X1 10X2
X1 3X2 6
2X1 2X2 4
First Tableau
C 5 10 0 0
SOL RHS X1 X2 S1 S2 ␾
0 S1 6 1 3 1 0 / 2
63
0 S2 4 2 2 0 1 / 2
43
Z 0 0 0 0 0
(C ⴚ Z) 5 10 0 0
+
Entering
Variable
In the first tableau, a tie for the leaving variable exists. Therefore, a condition of de-
generacy is present. Why is this a problem? It usually poses absolutely no problem at
all in arriving at an optimal solution. One of the variables is arbitrarily selected to leave,
and the simplex method is continued. In rare instances, however, looping can occur.
In other words, in the example above, if S2 is selected to leave the first tableau, S1 and
X2 are in the SOL column of the second tableau. S1 could leave and S2 could reenter
the third tableau, and S2 and X2 would be in the SOL column. S2 could leave and S1
reenter the fourth tableau, and S1 and X2 would be in the fourth tableau. This switch-
ing could conceivably continue endlessly, thus prohibiting an optimal solution.
Looping situations almost never happen in nontrivial real LP problems in POM.
When they do, a simple solution is to add or subtract an infinitesimally small amount
to either the RHS or coefficient in the ratio to break the tie. For example, the orig-
inal problem could be modified as follows:
Max Z 5X1 10X2
X1 3X2 6.0001
2X1 2X2 4
This slight modification breaks the tie of leaving variables and removes the degener-
acy condition.
Alternate optimal solutions exist when an element under a variable in Row (C Z)
is zero and that variable is not in the solution. For example, consider this problem:
Max Z X1 7X2
X1 7X2 14
7X1 X2 14
Second Tableau
C 1 7 0 0
7 X2 2 /
17 1 /
17 0 2/ / 14
17 Leaving
0 S2 12 /
48 7 0 1/7 1 12/48/7 13/4 , Variable
Z 14 1 7 1 0
(C ⴚ Z) 0 0 1 0
+
Entering
Variable
Note that the zero under X1 in Row (C Z) of this optimal tableau indicates that X1
can enter the solution (X2 2, X1 0, Z 14) with no change in Z. Let X1 enter to
check this out, X1 enters and S2 leaves:
1/7(13/4) 1/7(1) 1/7(0) 1/7(1/48) 1/7(7/48)

or 1/4 1/7 0 1/336 7/336
2 1/7 1 1/7 0
13/4 0 1 49/336 7/336
A new alternate optimal solution emerges in the third tableau where X1 13/4, X2
13/4, and Z 14. Z has not changed. Note that the zero in Row (C Z) under S2 in
the third tableau indicates also that an alternate optimal solution exists, the one in the
second tableau.
Third Tableau
C 1 7 0 0
7 X2 1/
34 0 1 /
49 336 / 7 336
1 X1 13/4 1 0 1/48 7/48
Z 14 1 7 1 0
(C ⴚ Z) 0 0 1 0
We should always examine Row (C Z) to inspect for alternate optimal solutions.

Why? Because these alternatives offer management the ultimate in flexibility—alter-
natives with the same profits or costs.
Now let us turn to a solution method designed to solve a special type of LP prob-
lem—the transportation problem.
TRANSPORTATION METHOD
In Chapter 8, we discussed selecting a set of shipments from sources to destinations

that minimized period transportation costs. These problems can be analyzed by either
the simplex method or the transportation method. We present the essential elements of
the transportation method here.
Characteristics of Transportation Problems

Transportation problems have these characteristics:
• A finite and homogeneous set of discrete units must be shipped from several sources
to several destinations in a particular time period.
• Each source has a precise number of units that must be shipped in the time pe-
riod.
• Each destination has a precise number of units that must be received in the time
period.
• Each discrete unit to be shipped has a specific transportation cost from each source
to each destination.
• The objective is to minimize the total transportation costs for the time period.
• The decision variables represent the number of units to be shipped from each source
to each destination during the time period.
Such problems were first formulated in the 1940s, and the solution procedures that will
be presented in this section were developed in the 1950s. Although the characteristics
listed above accurately depict the problems as they were originally formulated and
solved, later formulations and solutions allow a much broader range of problems. As
we shall see later in this section, problems with and constraints, maximization
objectives, demands greater than supply, and other characteristics are now routinely de-
scribed as transportation problems. Solution procedures that constitute the trans-
portation method also are used routinely to solve such expanded problems.
Solution Procedures of the Transportation Method

The procedures of the transportation method are exhibited in Table 4. To demonstrate
these procedures, Example 4 analyzes a problem of a manufacturing company that must
select a new location for a factory to produce computer peripheral units. Notice in Steps
2 and 3 of Table 4 that two alternative ways can be used to either test a solution for
optimality or obtain a starting solution. Example 4 uses only the northwest corner rule
and the stepping-stone method so that you can get the overall picture of how the trans-
portation method works. Later the same problem will be worked with the MODI and
VAM methods so that you can appreciate the relationships among the methods.
Table 4 Procedures of the Transportation Method
1. Formulate the problem in a transportation table.

2. Use the northwest corner rule or VAM to obtain a feasible starting solution.
3. Test the optimality of the solution by using the stepping-stone or MODI method. If the solution is
optimal, stop. If not, continue to the next step.
4. Develop a new transportation table that is an improved solution. Go back to Step 3.
Example 4 The Northwest Corner Rule and the Stepping-Stone Method
The Plain View Manufacturing Company presently has two factories at Amarillo and Waco, Texas,
and three warehouses at Dallas, San Antonio, and Houston, Texas. In recent months, Plain View
has been unable to produce and ship enough of its computer peripheral units to satisfy the mar-
ket demand at the warehouses. A new factory at Huntsville, Texas, is proposed to increase fac-
tory capacity. Bill Mayer, Plain View’s president, wants to determine what Plain View’s monthly
shipping costs will be with the new factory located at Huntsville.
The monthly capacities of the old and new factories, the monthly warehouse requirements,
and the transportation costs per unit from each factory to each warehouse are:
Monthly Transportation Costs

Monthly Warehouse Warehouse
Capacity Requirement San
Factory (units) Warehouse (units) Factory Antonio Dallas Houston
Amarillo 400 San Antonio 300 Amarillo $31 $21 $42
Waco 1,000 Dallas 900 Waco 20 21 30
Huntsville 600 Houston 800 Huntsville 23 20 15
Total 2,000 Total 2,000
a. Use the transportation method to determine the total monthly transportation costs if the new
factory is located at Huntsville. b. How many units per month should be shipped from each fac-
tory to each warehouse after the new factory is built? Follow the steps in Table 4 for the trans-
portation method. These steps are illustrated in Figure 1.
Solution
1. and 2. Formulate the problem in a transportation table, and use the northwest corner
rule to obtain a starting solution. Note that the monthly factory capacities are placed on
the right-hand side of the table opposite the appropriate factory row. Similarly, warehouse
requirements are placed along the bottom of the table under the appropriate warehouse col-
umn. The per-unit shipping cost is shown in a box within each factory–warehouse cell. Note
also that the total capacity for all factories equals the total warehouse requirements.
This starting solution shows how many units are shipped from each factory to each ware-
house. When a cell is empty, zero units are to be shipped. The initial solution is obtained by
beginning in the northwest cell (Amarillo–San Antonio) and allocating as many units as pos-
sible to this cell and proceeding likewise from left to right and downward. Only 300 units are
possible in the Amarillo–San Antonio cell because this amount satisfies the San Antonio ware-
house requirement. Moving to the right, we can allocate only 100 units to the Amarillo–Dallas
cell because this completes the Amarillo factory capacity of 400 units. Next, we move down-
ward and allocate 800 units to the Waco–Dallas cell, move right and allocate 200 units to the
Waco–Houston cell, and so on until all 2,000 units have been allocated.
3. Test the optimality of the solution by using the stepping-stone method. This step requires
systematically evaluating each of the empty cells in Transportation Table #1 to determine if
monthly transportation costs can be reduced by moving any units into the empty cells. The
stepping-stone evaluation method involves the procedures in Table 5. As shown in Figure 1,
Transportation Table #1 is not optimal because we have a negative circuit cost for the
Waco–San Antonio cell and we can reduce monthly transportation costs by moving some
units into this empty cell.
Figure 1 Transportation Tables of Example 4
Step 3a: First evaluate the Amarillo–Houston empty cell: Steps 1 and 2: Starting Solution Using the Northwest Corner Rule
Circuit cost 42 30 21 21 12. Place this cost in a circle. To Factory
San Antonio Dallas Houston
Factory From Totals
To San Antonio Dallas Houston
From Totals X 31 X 21 42
Amarillo 300 100 400
31 21 12 42
Amarillo 300 100 400 X 21 X 30
() () 20
Waco 800 200 1,000
20 21 30
Waco 800 200 1,000 X 15
() () 23 20
Huntsville 600 600
23 20 15
Huntsville 600 600
Warehouse 2,000
300 900 800
Warehouse 2,000 Totals 2,000
300 900 800
Totals 2,000 Step 4: Transportation Table #1
To Factory
From Totals
Step 3b: Next, evaluate the Waco–San Antonio empty cell: 31 21 12 42
Amarillo 300 100 400
Circuit cost 20 31 21 21 11. Place this cost in a circle.
Factory 11 20 21 30
To San Antonio Dallas Houston Waco 800 200 1,000
From Totals
31 21 42 7 23 14 20 15
Amarillo 300 100 400 Huntsville 600 600
() ()
20 21 30 Warehouse 2,000
Waco 11 800 200 1,000 300 900 800
() () Totals 2,000
23 20 15 The monthly costs of Transportation Table #1 are:
Huntsville 600 600
Units to Monthly
Warehouse 2,000 Be Shipped Transportation
300 900 800
Totals 2,000 Factory Warehouse per Month Costs
Amarillo San Antonio 300 $ 9,300
Amarillo Dallas 100 2,100
Step 3c: Next, evaluate the Huntsville–San Antonio empty cell: Waco Dallas 800 16,800
Circuit cost 23 31 21 21 30 15 7. Place this cost in a circle. Waco Houston 200 6,000
To Factory Huntsville Houston 600 9,000
From Totals Totals 2,000 $43,200
31 21 42
Amarillo 300 100 400 Step 5: Transportation Table #2
() ()
20 21 30 To Factory
Waco 800 200 1,000 San Antonio Dallas Houston
() () From Totals
23 20 15 11 31 21 12 42
Huntsville 7 600 600 Amarillo 400 400
() ()
Warehouse 2,000 20 21 30
300 900 800 Waco 300 500 200 1,000
Totals 2,000
23 20 15
Huntsville 18 14 600 600
Warehouse 2,000
Step 3d: Next, evaluate the Huntsville–Dallas empty cell: 300 900 800
Totals 2,000
Circuit cost 20 21 30 15 14. Place this cost in a circle.
The monthly costs of Transportation Table #2 are:
To Factory
San Antonio Dallas Houston Units to Monthly
From Totals
Be Shipped Transportation
31 21 42
Amarillo 300 100 400 Factory Warehouse per Month Costs
20 21 30 Amarillo Dallas 400 $ 8,400
Waco 800 200 1,000 Waco San Antonio 300 6,000
() ()
23 20 15 Waco Dallas 500 10,500
Huntsville 14 600 600
() () Waco Houston 200 6,000
Warehouse 2,000 Huntsville Houston 600 9,000
300 900 800
Totals 2,000 Totals 2,000 $39,900
Table 5 Stepping-Stone Method
Logic of the Stepping-Stones

In this procedure, we want to determine if the present transportation table is optimal and, if it is not, which
variable should enter the solution. Each unfilled cell (a variable not in the solution) is investigated to
determine if period costs would be reduced if units were moved from a filled cell (a variable in the
solution) to an unfilled cell. The stepping-stone circuit is simply a logical way of determining the per-unit
change in period costs if units are moved into an unfilled cell. This is analogous to examining the Row
(C Z) in the simplex method to determine the entering variable.
Steps of the Stepping-Stone Method
1. Select an empty cell from the transportation table.
2. Draw a closed circuit between the empty cell and other stones (occupied cells) by using only straight
vertical or horizontal lines. The circuit may skip over stones or other empty cells, but the corners of the
circuit may occur only at stones (occupied cells) and the empty cell that is being evaluated.
3. Beginning at the empty cell being evaluated, move clockwise and alternatively assign positive () and
negative () signs to the costs of the cells at the corners of the circuit from Step 2.
4. Total the per-unit costs of the cells at the corners of the circuit. The circuit total corresponds to the Row
(C Z) values of the simplex method, and these values mean the amount of change in total shipping
costs that can be realized by moving one unit to the empty cell under examination. Positive values mean
costs will rise; negative values mean costs will fall.
5. Return to Step 1 and continue until all empty cells have been evaluated. The new cell to enter the
solution is the cell whose circuit has the most negative circuit total cost.
6. If all the circuit totals are positive or zero, the solution is optimal. If negative circuit totals exist, develop
an improved solution.
Determining the New Solution
Once we determine that period costs would be reduced if some units were moved into an unfilled cell, how
many units should be moved? We should move as many as possible. The change in the solution is
determined by examining the stepping-stone circuit with the most negative total circuit cost. In this circuit,
among the filled cells with negative costs, locate the filled cell with the smallest number of units to be
shipped. Subtract the number of units from all filled cells with negative costs and add this number to all
filled cells with positive costs on that stepping-stone circuit. This is analogous to determining the value of
the entering variable in the simplex method.
4. Develop a new transportation table that is an improved solution. An improved solution

is obtained by moving as many units as possible into the empty cell of the last transporta-
tion table with the most negative circuit cost. But how many units can be moved into the
Waco–San Antonio cell, which had a negative circuit cost in Step 3? Let us again examine
the stepping-stone circuit for this cell.
The maximum number of units that can be moved into the Waco–San Antonio empty cell
is 300—the smallest number of units in a negative cell on the Waco–San Antonio stepping-
stone circuit. To complete the improved solution, subtract the smallest number of units in
negative cells, 300, from all negative cells and add this same number of units to the positive
cells of the circuit. All other cells not on this circuit remain unchanged. This new solution
is shown in Transportation Table #2 of Figure 1.
This solution is an improved one—$43,200 in Transportation Table #1 versus $39,900 in
Transportation Table #2—but is it optimal? Optimality can be determined once again by follow-
ing the stepping-stone procedures. The stepping-stone circuit costs of the empty cells are shown
in circles in Transportation Table #2 of Figure 1. Because all stepping-stone circuit costs are ei-
ther positive or zero, the solution in Transportation Table #2 is optimal.
Now let us answer the questions of Plain View’s location problem:
a. If the factory is located at Huntsville, Plain View’s total monthly transportation costs will be
$39,900.
b. Plain View should make these monthly shipments:
Number of
Factory Warehouse Units
Amarillo Dallas 400
Waco San Antonio 300
Waco Dallas 500
Waco Houston 200
Huntsville Houston 600
Total 2,000
Example 4 uses the well-known stepping-stone method to determine optimality and

to develop improved solutions. A newer and more frequently used method is the MODI
method (modified distribution method). This procedure is similar to the stepping-stone
method, but it is more efficient in computing the improvement costs (circuit costs in the
stepping-stone method) for the empty cells.
Example 5 uses the MODI method to test Transportation Table #1 in Figure 1 for
optimality. Before you begin this example, however, perhaps it would be helpful if you
would do two things. First, review the procedures of the transportation method from
Table 4 to get the overall view of the procedure again. Notice that the only way that
the MODI method affects the procedures of the transportation method is in testing each
transportation table for optimality. Everything else in the procedure stays the same—
formulating the transportation tables, using either the northwest corner rule or VAM to
obtain a starting solution, and developing new transportation tables that are improved
solutions. Next, review Step 3 in Figure 1. This step uses the stepping-stone method
to check the optimality.
Example 5 MODI Method of Testing Transportation Tables for Optimality
Transportation Table #1
Kj
K1 31 K2 21 K3 30
Ri
To San Factory
From Antonio Dallas Houston Totals
31 21 42
R1 0 Amarillo 300 100 12 400
20 21 30
R2 0 Waco 11 800 200 1,000
23 20 15
R3 15 Huntsville 7 14 600 600
Warehouse 2,000
Totals 300 900 800 2,000
Calculating Ri and Kj Calculating Circuit Costs

for Filled Cells for Unfilled Cells
Ri Kj Cij, where Cij represents the transportation Cij Ri Kj
cost of the ij cell Amarillo–Houston
R1 0 C13 R1 K3
R1 K1 C11 42 0 30 12
0 K1 31, K1 31 Waco–San Antonio
R1 K2 C12 C21 R2 K1
0 K2 21, K2 21 20 0 31 11
R2 K2 C22 Huntsville–San Antonio
R2 21 21, R2 0 C31 R3 K1
R2 K3 C23 23 (15) 31 7
0 K3 30, K3 30 Huntsville–Dallas
R3 K3 C33 C32 R3 K2
R3 30 15, R3 15 20 (15) 21 14
In Example 5, we first compute the Ri and Kj for the table. R1 is always set equal
to zero; this allows us to compute all other values of Ri and Kj for filled cells. After
we know these values, we can directly calculate the circuit costs for the empty cells of
the transportation table, but without the necessity of drawing the stepping-stone cir-
cuits for all of these cells as in the stepping-stone method. Next, if there are any neg-
ative circuit costs, we would draw the stepping-stone circuit for the empty cell that has
the most negative circuit cost, just as we did in Figure 1, and develop a new trans-
portation table with an improved solution as before. Although the Ri and Kj must be
recomputed for each transportation table, the MODI method is more efficient than the
stepping-stone method and tends to be used more frequently in practice.
The northwest corner rule provides a starting solution to transportation problems,
but one that is arbitrary. In most problems, this results in too many transportation ta-
bles. This source of inefficiency may not seem too important to you now after having
worked through Example 4 in just two transportation tables, but in problems of more
realistic proportions, say 25 sources and 40 destinations, many tables would be required
if we used the northwest corner rule to obtain a starting solution. Vogel’s approxi-
mation method (VAM) was developed to obtain a more efficient starting solution. In
fact, in many problems the starting solution is optimal. Although the VAM method is
more complicated than the northwest corner rule, in realistic problems, VAM is a more
practical way to obtain starting solutions.
Example 6 develops a starting solution to the transportation problem of our pre-
vious example. In working through the procedures of the VAM method, refer to Table
6, which explains the steps of the method that are applied in Example 6. In Step 1 from
Table 6, in the example the D1 row and column are first completed. These values rep-
resent the difference between the lowest unit cost and next lowest unit cost for each
row and column. In Step 2, the Houston column has the largest difference on the first
iteration and it is selected. In Step 3, the Huntsville–Houston cell has the lowest unit
cost within the Houston column, and we therefore allocate 600 units to that cell, the
most possible. Because all of the Huntsville row has been allocated, you may draw a
line all the way through this row to eliminate it from further consideration. This com-
pletes the first iteration.
Example 6 Starting Solutions with Vogel’s Approximation Method (VAM)
To San Factory
From Antonio Dallas Houston Totals D1 D2 D3 D4
31 (4) 21 42
Amarillo 400 400 10 10 10
(3) 20 (4) 21 (2) 30
Waco 300 500 200 1,000 1 1 1
23 20 (1) 15
Huntsville 600 600 5
2,000
Warehouse
Totals 300 900 800 2,000
D1 3 1 15
D2 11 0 12
D3 11 0
D4 0
Di Row Units Allocated

Di Column Calculations Calculations at Iteration (i)
31 21 10, (1) 600/Huntsville–Houston
D1 23 20 3, 21 20 1, 21 20 1,
30 15 15 20 15 5
D2 31 20 11, 21 21 0, 31 21 10, (2) 200/Waco–Houston
42 30 12 21 20 1
D3 31 20 11 , 31 21 10, (3) 300/Waco–San Antonio
21 21 0 21 20 1
D4 21 21 0 (4) 500/Waco–Dallas
400/Amarillo–Dallas
Next, we begin the second iteration in the D2 row and column. The Houston col-
umn has the largest difference and is selected, and 200 units are allocated to the
Waco–Houston cell within the Houston column because this cell has the lowest re-
maining unit cost. This allocation exhausts the Houston column and a line may be
drawn through the column. This completes the second iteration. The third and fourth
iterations are similarly completed.
The starting solution obtained in this example is the same optimal solution obtained
in Figure 1 of Example 4. This is not always the case and although the VAM method
does yield an efficient starting solution when compared to the northwest corner method,
it still must be considered only a starting solution and all the steps of the transporta-
tion method listed in Table 4 must be followed. In other words, in Example 6, after the
last iteration we would need to use either the stepping-stone or the MODI method to
test the solution for optimality and proceed with the entire transportation method.
Table 6 Steps of the VAM Method
1. For each row and column of the transportation table, compute the difference between the lowest
unit cost and the next lowest unit cost and record this difference. Place the row differences in a
column to the right of the table under a heading of Di and the column differences in a row across
the bottom of the table with a heading of Di, where i represents the number of times you have done
this step.
2. Select either the row or column with the largest difference. If ties occur, arbitrarily select among
tying elements.
3. Allocate as many units as possible to the cell with the lowest cost in the row or column selected in
Step 2.
4. If the units in a row or column have been exhausted in Step 3, that row or column may be
eliminated from further consideration in subsequent calculations by drawing a line through it.
5. When differences cannot be calculated in Step 1 because only one row or one column remains, this
is not an unusual occurrence as we near the end of the process. Calculate the differences that are
possible and continue.
6. Return to Step 1 and continue until the units in all the rows and columns have been allocated.
The reason that the VAM method yields better starting solutions than the north-
west corner rule is that the northwest corner rule does not consider any cost infor-
mation when the starting solution is determined—units are arbitrarily allocated on
a northwest diagonal regardless of the costs. In the VAM method, an opportunity
cost principle is applied. At each iteration, the difference between the lowest unit
cost and the next lowest unit cost is the opportunity cost of not allocating units to a
row or a column. By selecting the largest difference, the largest opportunity cost is
avoided. By taking into account the costs of alternative allocations, the VAM method
yields very good starting solutions that are sometimes optimal, particularly in sim-
ple problems.
Unbalanced Problems
Example 4 involved a problem where the total number of units to be shipped from
sources exactly equaled the number of units required at destinations. This is called a
balanced transportation problem. It is not unusual to have an unbalanced trans-
portation problem, where the number of units that can be shipped from sources ex-
ceeds the number required at destinations, or vice versa. Figure 2 is the optimal trans-
portation table of such an unbalanced transportation problem.
A dummy destination column is entered into the table to account for the differ-
ence between destination requirements and source shipments. Note that the shipping
cost from any source to the dummy destination is zero. The only function of the ficti-
tious destination is to balance the problem. The interpretation of the 2,000 units in the
Source C–dummy destination cell is that 2,000 units of the capacity at Source C will
not be shipped. The dummy destination column therefore serves the same purpose as
slack variables in the simplex method. All other solution procedures (the northwest cor-
ner rule, VAM, the stepping-stone method, or the MODI method) discussed earlier are
followed to solve unbalanced transportation problems. The interpretation of the final
solution is exactly the same in either balanced or unbalanced problems, with the ex-
ception of the dummy row or column interpretation.
Figure 2 An Unbalanced Transportation Problem
To Dummy Source
From 1 2 3 Destination Totals
A 3.5 1,000 2.0 4.0 0 1,000
B 4.0 2,000 2.5 3,000 1.5 0 5,000
C 1,000 2.0 1,000 3.0 3.0 2,000 0 4,000
Destination 10,000
Totals 1,000 4,000 3,000 2,000 10,000
Degeneracy
Degeneracy is another complication that can be encountered in transportation prob-
lems. The number of occupied cells in a transportation table must equal the number of
sources plus the number of destinations minus 1. Thus, in Example 4 all transporta-
tion tables always had 5 occupied cells (3 sources 3 destinations 1 5). De-
generacy is present when less than this minimum number of occupied cells occurs. De-
generacy, if present, interferes with the drawing of stepping-stone circuits when the
stepping-stone method is used, or it makes it impossible to compute the Ri and Kj if
the MODI method is used to check for optimality of transportation tables. Therefore,
special procedures are necessary to solve these problems successfully.
If degeneracy occurs in the initial transportation table solution after employing
either the VAM method or the northwest corner rule, a zero is assigned to one of the
empty cells. The empty cell selected is usually one that creates an unbroken chain of
occupied cells from NW to SE across the transportation table. The zero cell is treated
as an occupied cell with zero units occupying the cell when stepping-stone circuits
are drawn or Ri or Kj is calculated. This is analogous to having a variable in the so-
lution, but with a zero value as in the simplex method. This manipulation allows us
to complete our check for optimality without changing the nature of the transporta-
tion problem.
When degeneracy occurs in transportation tables beyond the initial solution, a
slightly different procedure is used. Figure 3 shows Transportation Tables #1 and #2 of
a transportation problem where degeneracy exists. Notice that the initial solution to this
transportation problem has 6 occupied cells, exactly the required minimum (3 sources
4 destinations 1 6). The stepping-stone circuit of the San Diego–Seattle cell
(superimposed on Transportation Table #1) has the most negative circuit cost of any
empty cell in this table, and an improved solution must be developed from this circuit.
Remember that we first identify the smallest number of units in the negative cells (5,000
units) and then subtract this number from all negative cells and add it to all positive
cells on the circuit.
Ordinarily, this procedure makes one occupied cell go to zero units, but in this cir-
cuit both the Miami–Seattle and Chicago–Denver cells go to zero. That is what causes
Figure 3 A Degenerate Transportation Table
To Factory
From Seattle Denver New Orleans New York Totals
1.2 0.7 0.5 0.6
Miami 5,000 5,000 10,000
() ()
0.7 0.5 0.5 0.6

Chicago 5,000 10,000 5,000 20,000
() ()
0.5 0.7 0.8 1.2

San Diego 15,000 15,000
() ()
Warehouse 45,000
Totals 5,000 10,000 10,000 20,000 45,000
To Factory
From Seattle Denver New Orleans New York Totals
1.2 0.7 0.5 0.6
Miami 0 10,000 10,000
0.7 0.5 0.5 0.6
Chicago 10,000 10,000 20,000
0.5 0.7 0.8 1.2
San Diego 5,000 10,000 15,000
Warehouse 45,000
Totals 5,000 10,000 10,000 20,000 45,000
the solution in Transportation Table #2 to be degenerate, a condition comparable to a

tie for leaving variables in the simplex method. We handle this degenerate situation in
the transportation method by assigning a zero number of units to either one of the cells
reduced to zero units. In Figure 3, the zero is assigned to either the Miami–Seattle cell
or the Chicago–Denver cell. The zero is then treated as an occupied cell, but with zero
units, when applying either the stepping-stone or the MODI method of checking sub-
sequent transportation tables for optimality. All other procedures of the transportation
method are followed as before.
When more than two occupied cells are eliminated, more than one zero must be
assigned to these cells to overcome the condition of degeneracy. Add enough zero cells
so that the number of occupied cells equals the number of sources plus the number of
destinations minus 1. These problems are otherwise solved as before, with standard
transportation method procedures.
ASSIGNMENT METHOD
The assignment method is another method for solving LP problems. Like the trans-
portation method, the assignment method is easier to work than the simplex method,
but it can be used only on LP problems with special characteristics. These character-
istics are even more restrictive than those of LP problems using the transportation
method:
1. n objects must be assigned to n destinations.
2. Each object must be assigned to some destination.
3. Each destination must be assigned an object.
4. The objective is to minimize the total cost of the assignments.
These problems can be solved by the transportation method where the units to be
“shipped” is 1 for all rows and columns and the number of rows equals the number of
columns. In this formulation, the “sources” are the objects to be assigned. Although
this formulation is straightforward, its solution with the transportation method can be-
come difficult owing to extreme cases of degeneracy. Because of this complication, the
assignment method offers computational advantages for solving assignment problems.
Example 7 demonstrates the procedures of the assignment method. In this exam-
ple, the Mercury Electric Motor Company needs to assign five motor overhaul jobs to
five motor rewinding centers. The assignment method is a simple and efficient method
for solving such problems because it guarantees assignments that result in minimum
processing cost, maximum profit, or minimum processing time for all the jobs.
Example 7 demonstrates the minimization algorithm that is appropriate for mini-
mizing costs or processing times. If we wished to maximize profits, we would multi-
ply all the profits in the first table by minus 1 (1) and then follow the same proce-
dures as in the minimization case. Regardless of whether the problem is of the
maximization or minimization type, the optimal job assignment—or job assignments,
since multiple optimal solutions are possible—results from the use of the assignment
method.
Example 7 Using the Assignment Method to Assign n Jobs to n Work Centers
The Mercury Electric Motor Company overhauls very large electric motors used in industrial
plants in its region. Bill Tobey has just received five electric motor overhaul jobs and is trying to
decide to which rewinding work centers the jobs should be assigned. Because some of the work
centers specialize in certain types of jobs, the cost for processing each job varies from work cen-
ter to work center. Bill has estimated the processing costs for the five jobs at five rewinding work
centers, and he uses the assignment method to make a minimum cost assignment of the jobs to
work centers by following this procedure:
1. Place the cost information for assigning the jobs to work centers in an assignment table for-
mat. This is shown in Step 1 in Figure 4.
2. Subtract the smallest cost in each row from all other costs in that row. The resulting table is
shown in Step 2 in Figure 4.
3. Subtract the smallest cost in each column from all other costs in that column. See Step 3 in
Figure 4.
4. Draw the least number of vertical or horizontal straight lines to cover the zero cells. See Step
4 in Figure 4. Note that several different schemes may be possible, but the same number of
minimum lines should result. If n lines (five in this example) were the minimum number of
lines to cover the zero cells, the optimal solution would have been reached. If that were the
case, the optimal job assignments would be found at the zero cells. Because less than n or
only four lines were required, we must perform a modification to the table in Step 4 in Fig-
ure 4.
5. Select the smallest cost not covered by lines in the table in Step 4 in Figure 4 (the smallest
cost is 50). Subtract this cost from all uncovered costs and add this cost to cells at the in-
tersections of lines in Step 4. Then transfer these new costs and the costs that are unchanged
in Step 4 to a new table in Step 5. Redraw the lines to cover the zero cells. Because five
lines are required in Step 5, the solution is optimal. If fewer than five lines had been required,
Step 5 would have to be repeated.
Figure 4 Tables of the Assignment Method in Example 7
Step 1 Step 2
Work Centers Work Centers
1 2 3 4 5 1 2 3 4 5
A $150 $300 $225 $350 $250 A 0 150 75 200 100
Jobs
Jobs
B 300 200 400 300 250 B 100 0 200 100 50
C 150 100 100 200 150 C 50 0 0 100 50
D 300 100 200 250 200 D 200 0 100 150 100
E 150 350 230 375 260 E 0 200 80 225 110
Step 3 Step 4
Work Centers Work Centers
1 2 3 4 5 1 2 3 4 5
A 0 150 75 100 50 A 0 150 75 100 50
Jobs
Jobs
B 100 0 200 0 0 B 100 0 200 0 0

C 50 0 0 0 0 C 50 0 0 0 0
D 200 0 100 50 50 D 200 0 100 50 50
E 0 200 80 125 60 E 0 200 80 125 60
Step 5
Work Centers
1 2 3 4 5
A 0 100 25 50 0
Jobs
B 150 0 200 0 0
C 100 0 0 0 0
D 250 0 100 50 50
E 0 150 30 75 10
Once the optimal table has been obtained, as in Step 5, you are ready to read the
optimal solution from this table. To do so, select any row or column with only one zero
in the optimal table. In this example, you could choose either Column 3 or Row D or
E. Suppose you choose Row D. Make the assignment at the cell with a zero in Row
D: D–2. Eliminate Row D and Column 2 from further consideration by drawing a line
through this row and column. Now only consider the 4 4 matrix with Rows A, B,
C, and E and Columns 1, 3, 4, and 5. Select any row or column with just one zero in
this new matrix. Make that assignment, eliminate that row and column from further
consideration, and repeat this process until all assignments have been made. The opti-
mal solution in this example is A–5, B–4, C–3, D–2, and E–1, giving a total cost of
$250 300 100 100 150 $900.
REVIEW AND DISCUSSION QUESTIONS
1. Where in the optimal simplex tableau is the LP so- 11. What methods may be used to test a transportation
lution found? table for optimality?
2. Where are the shadow prices in the optimal simplex 12. What advantage does the MODI method have over
tableau? the stepping-stone method? Explain.
3. What are shadow prices? What information do they 13. What advantage does the VAM method have over the
provide managers? northwest corner rule? Explain.
4. Why must row operations be performed in the sim- 14. What is an unbalanced transportation problem? What
plex method? modifications are necessary in our ordinary trans-
5. What purpose do artificial variables serve? portation method when working unbalanced prob-
6. What are slack variables? What purpose do they serve? lems?
7. What determines the subscripts for artificial and 15. Why is degeneracy a difficulty in the transportation
slack variables? method? What specific difficulties are encountered
8. What are the characteristics of transportation prob- in the northwest corner rule, stepping-stone method,
lems? and the MODI method?
9. What are the procedures of the transportation method? 16. What are the characteristics of assignment problems?
10. What are the procedures of the northwest corner 17. Describe the procedures of the assignment method.
rule? Why is it used?
PROBLEMS
Simplex Method
1. Perform row operations on this second tableau, and complete the third tableau. Is
the third tableau optimal? If not, which variables will enter and leave the third
tableau?
Second Tableau
C 40 75 0 0
75 X2 75 /
12 1 /
14 0
0 S2 50 4 0 1/2 1
Z 5,625 /
75 2 75 75 4/ 0
(C ⴚ Z) /
52 0 75/4 0
2. Perform row operations on this second tableau, and complete the third tableau. Is
the third tableau optimal? If not, which variables will enter and leave the third
tableau?
Second Tableau
C 50 60 70 0 0 0
SOL RHS X1 X2 X3 S1 S2 S3 ␾
70 X3 100 0 /
14 1 / 14 0 0
0 S2 0 2 1/2 0 1/2 1 0
0 S3 400 4 1/2 0 1/2 0 1
Z 7,000 0 /
35 2 70 /
35 2 0 0
(C ⴚ Z) 50 /
85 2 0 35/2 0 0
3. Perform row operations on this second tableau. Complete the third tableau. Is the
third tableau optimal? If not, which variables will enter and leave the third tableau?
Second Tableau
C 30 40 M M 0 0 0
SOL RHS X1 X2 A1 A2 S1 S2 S3 ␾
40 X2 30 /
12 1 /
12 0 1/2 0 0
M A2 50 3/2 0 5/2 1 5/2 1 0
0 S3 20 3/2 0 1/2 0 1/2 0 1
Z ⴚ50M 3/2M 5/2M 5/2M
40 M M 0
ⴚ1,200 20 20 20
/M
32 7/2M /M
52
(C ⴚ Z) 0 0 M 0
10 20 20
(LP-A, B, C, D, E, F, G, H, I, and J problems are found at the end of Chapter 8.)

4. Formulate Problem LP-A. Develop the initial simplex tableau.
5. Solve LP-B using the simplex method. What is the optimal solution?
6. Solve LP-C using the simplex method. What is the optimal solution?
7. Solve LP-D using the simplex method. What is the optimal solution?
8. Solve LP-E using the simplex method. What is the optimal solution? (Hint: Con-
vert the objective function to cents.)
9. Solve LP-G using the simplex method. What is the optimal solution?
10. Solve LP-H using the simplex method. What is the optimal solution?
11. Solve LP-I using the simplex method. What is the optimal solution? Fully inter-
pret the meaning of the solution.
12. Below are the variable definitions and the last (optimal) tableau for Problem
LP-A.
a. What should management do? In other words, to management decision mak-
ers, what is the complete meaning of the values of X1, X2, S1, S2, S3, S4, A3,
A4, and Z in the optimal tableau?
X1 number of chairs to produce per week

X2 number of bookshelf units to produce per week
S1 excess labor hours unused per week
S2 excess wood unused per week (board-feet)
S3 number of chairs above minimum per week
S4 number of bookshelf units above minimum per week
Z dollars of profit per week
Optimal Tableau
C 15 21 0 0 0 0 M M
SOL RHS X1 X2 S1 S2 S3 S4 A3 A4 ␾
0 S1 20 0 0 1 /12 0 3 0 3
0 S3 50 0 0 0 1/8 1 3/2 1 3/2
15 X1 150 1 0 0 1/8 0 3/2 0 3/2
21 X2 100 0 1 0 0 0 1 0 1
Z 4,350 15 21 0 15 8/ 0 /
32 0 3/2
M
(C ⴚ Z) 0 0 0 15/8 0 3/2 M
3/2
13. Below are the variable definitions and the last (optimal) tableau for Problem LP-B.
ers, what is the complete meaning of the values of X1, X2, S1, S2, and Z in
the optimal solution?
X1 hundreds of gallons of premium ice cream to produce per week

X2 hundreds of gallons of light ice cream to produce per week
S1 unused mixing machine hours
S2 unused gallons of milk
Z dollars of profit per year
Optimal Tableau
C 100 100 0 0
100 X2 175 0 1 /30 8 / 1 80
100 X1 175 1 0 35/12 1/48
Z 35,000 100 100 83.33 0.833

(C ⴚ Z) 0 0 83.33 0.833
14. Below are the variable definitions and the last (optimal) tableau for Problem LP-C.
ers, what is the complete meaning of the values of X1, X2, S1, S2, S3, and Z
in the optimal solution?
X1 cases of relays per quarter

X2 cases of capacitors per quarter
S1 unused labor-hours
S2 unused stamping machine hours
S3 unused testing machine hours
Z dollars of profit per year
Optimal Tableau
C 250 200 0 0 0
0 S1 4285.7 0 0 1 /
275 7 / 100 7
250 X1 142.86 1 0 0 5/14 1/7

200 X2 314.286 0 1 0 3/14 2/7
Z 98,572 250 200 0 46.429 21.429

(C ⴚ Z) 0 0 0 46.429 21.429
15. Below are the variable definitions and the last (optimal) tableau for Problem LP-D.
ers, what is the complete meaning of the values of X1, X2, S2, S3, and Z in
the optimal solution?
X1 tons of sand per day
X2 tons of gravel per day
S2 below maximum sand limit
S3 above minimum gravel limit
Z $ cost of sand and gravel per day
Optimal Tableau
C 6 8 M M 0 0
0 S3 100 0 0 1 1 1 1
6 X1 800 1 0 0 0 1 0
8 X2 700 0 1 1 0 1 0
Z ⴚ10,400 6 8 8 0 2 0
(C ⴚ Z) 0 0 M8 M 2 0
16. Below are the variable definitions and the last (optimal) tableau for Problem LP-E:
in the optimal tableau?
X1 pounds of oats fed to each hog each day
X2 pounds of corn fed to each hog each day
S1 excess mineral units per hog per day
S2 excess calories per hog per day
S3 excess vitamin units per hog per day
Z daily cents of feed cost per day
Fourth Tableau
C 12 7 0 0 0 M M M
SOL RHS X1 X2 S1 S2 S3 A1 A2 A3 ␾
12 X1 /
25 2 1 0 /3 400 /
1 400 0 /
3 400 / 1 400 0
0 S3 250 0 0 1/4 3/4 1 1/4 3/4 1
7 X2 5 0 1 1/200 1/200 0 1/200 1/200 0
Z ⴚ185 12 7 /
11 200 /
1 200 0 11/200 1/200 0
M M
(C ⴚ Z) 0 0 11/200 1/200 0 M
11/200 1/200
17. The computer solution to Problem LP-F from the POM Software Library follows.
Min Z 7.9 X1 4.7 X2 3.2 X3 8.3 X4
6.1 X5 8.7 X6 4.5 X7 7.6 X8
5.4 X9 6.6 X10 7.1 X11 5 X12
ST
(1) 1 X1 1 X2 1 X3 1 X4 32000
(2) 1 X5 1 X6 1 X7 1 X8 40000
(3) 1 X9 1 X10 1 X11 1 X12 25000
(4) 1 X1 1 X5 1 X9 30000
(5) 1 X2 1 X6 1 X10 18000
(6) 1 X3 1 X7 1 X11 22000
(7) 1 X4 1 X8 1 X12 27000

SOLUTION:

VARIABLE MIX SOLUTION
X3 14000.000
X5 30000.000
X7 8000.000
X2 18000.000
Slack 2 0.000
X12 25000.000
X8 2000.000
Z 488600.000

SENSITIVITY ANALYSIS:

CONSTRAINTS:
——————————————————————————————————
RANGE OF RHS
CONSTRAINT TYPE OF SHADOW FOR WHICH SHADOW
NUMBER CONSTRAINT PRICE PRICE IS VALID
1 1.300 32000. –– 40000.
2 0.000 40000. –– INF
3 2.600 25000. –– 27000.
4 6.100 0. –– 30000.
5 6.000 10000. –– 18000.
6 4.500 14000. –– 22000.
7 7.600 25000. –– 27000.
NOTE: THE SHADOW PRICE REPRESENTS THE AMOUNT Z WOULD CHANGE IF

A CONSTRAINT’S RHS CHANGED ONE UNIT.

DECISION VARIABLES:

NONBASIC AMOUNT Z IS REDUCED (MAX) OR INCREASED (MIN)
VARIABLE FOR ONE UNIT OF X IN THE SOLUTION
X1 3.1
X4 2.
X6 2.7
X9 1.9
X10 3.2
X11 5.2
——————————————————————————————————

ers, what is the complete meaning of the values of the decision variables, slack
variables, and Z in the optimal solution?
b. What is the meaning of each value in the sensitivity analysis section of the
solution?
18. Below are the variable definitions and the last (optimal) tableau for Problem LP-G:
in the optimal tableau?
X1 number of cases of Product A per week

X2 number of cases of Product B per week
S1 unused Cutting Department hours per week
S2 unused Trimming Department hours per week
S3 unused Coating Department hours per week
Z dollars of profit per week
Third Tableau
C 64 87 0 0 0
0 S1 21 2 / 0 0 1 5/8 39/160
64 X1 15 1 0 0 1/4 1/16
87 X2 10 0 1 0 0 1/8
Z 1,830 64 87 0 16 /
55 8
(C ⴚ Z) 0 0 0 16 /
55 8
Transportation Method
19. a. Use the northwest corner rule and the stepping-stone method to solve the trans-
portation problem below.
To Source
From 1 2 3 Totals
$0.80 $0.40 $0.70
A 300
0.90 0.50 0.80

B 300
0.60 0.70 0.60
C 400
Destination 200 500 300 1,000

Totals 1,000
b. What is the optimal solution?

c. Fully interpret the meaning of the solution.
20. The Apex Company produces electric transformers for electric distributors at two
factories located in Atlanta and San Jose. Apex has four regional warehouses lo-
cated in Dallas, Seattle, Philadelphia, and Chicago. The monthly warehouse re-
quirements, monthly factor capacities, and per-unit transportation costs are:
Monthly
Factory Transportation
Capacity Cost
Factory (units) Source Destination per Unit
Atlanta 2,000 Atlanta Dallas $16
San Jose 3,000 Seattle 40
Philadelphia 20
Chicago 18
San Jose Dallas 30
Seattle 15
Philadelphia 45
Chicago 33
Monthly
Warehouse
Requirement
Warehouse (units)
Dallas 1,300
Seattle 1,200
Philadelphia 1,400
Chicago 1,100
a. Use the northwest corner rule and the stepping-stone method to solve the trans-
portation problem above.
b. What is the optimal solution?
c. Fully interpret the meaning of the solution.
21. a. Solve Problem LP-F using the northwest corner rule and the stepping-stone
method. What is the optimal solution? Fully interpret the meaning of the so-
lution.
b. Solve Problem LP-F using the VAM and MODI methods. What is the opti-
mal solution? Fully interpret the meaning of the solution.
22. a. Solve Problem LP-J using the northwest corner rule and the stepping-stone
method.
b. What is the optimal solution? Fully interpret the meaning of the solution.
23. a. Solve Problem LP-J using the VAM and MODI methods.
b. What is the optimal solution? Fully interpret the meaning of the solution.
24. Three fabrication departments—A, B, and C—produce three slightly different
products that are processed further in four assembly departments—1, 2, 3, and 4.
Each fabrication and assembly department has a unique monthly capacity, and it
is desirable that each department operate at capacity. Although any of the three
products coming from the three fabrication departments can be processed in any
of the four assembly departments, the assembly cost per unit differs among the
products. Additionally, the materials-handling costs per unit differ among all the
fabrication and assembly departments. The monthly fabrication department ca-
pacities, monthly assembly department capacities, and per-unit materials-handling
and assembly processing costs are:
Total
Materials-Handling
Monthly Materials-Handling Assembly and Assembly
Fabrication Capacity Fabrication Assembly Cost per Cost per Cost per
Department (units) Department Department Unit Unit Unit
A 12,000 A 1 $0.20 $0.40 $0.60
B 8,000 2 0.25 0.65 0.90
C 14,000 3 0.10 0.60 0.70
4 0.15 0.35 0.50
B 1 $0.30 $0.70 $1.00
Monthly 2 0.15 0.95 1.10
Assembly Capacity 3 0.15 0.65 0.80
Department (units) 4 0.25 0.65 0.90
1 7,000 C 1 $0.10 $0.70 $0.80
2 11,000 2 0.25 0.35 0.60
3 6,000 3 0.30 0.90 1.20
4 10,000 4 0.15 0.55 0.70
Develop a monthly production schedule for the three fabrication and four assem-
bly departments. In other words, how many units should be processed in each fab-
rication department and moved to and processed in each assembly department so
that monthly assembly and materials-handling costs are minimized?
a. Use the northwest corner rule and the stepping-stone method to solve the prob-
lem. What is the optimal solution? Fully interpret the meaning of the solution.
b. Use the VAM and MODI methods to solve the problem. What is the optimal
solution? Fully interpret the meaning of the solution.
25. Three recycling plants—A, B, and C—receive and process scrap paper and ship
paper stock to three regional warehouses—1, 2, and 3. The monthly recycling plant
capacities, monthly regional warehouse requirements, and the shipping cost per
100 pounds are:
Monthly Monthly Shipping

Recycling Capacity Regional Requirement Recycling Regional Cost per
Plant (100 pounds) Warehouse (100 pounds) Plant Warehouse 100 Pounds
A 800 1 700 A 1 $1
B 600 2 900 2 4
C 1,100 3 600 3 2
B 1 $2
2 4
3 1
C 1 $1
2 3
3 2
How many pounds of paper stock should be shipped from each recycling plant to
each regional warehouse per month to minimize monthly shipping costs?
a. Use the northwest corner rule and the stepping-stone method to solve this prob-
lem. What is the optimal solution? Fully interpret the meaning of the solution.
b. Use the VAM and MODI methods to solve this problem. What is the optimal
solution? Fully interpret the meaning of the solution.
Assignment Method
26. Four jobs must be assigned to four work centers. Only one job can be assigned to
each work center, and all jobs must be processed. The cost of processing each job
through each work center is shown below:
Work Centers
1 2 3 4
A $50 $40 $70 $35
B 25 25 55 40
Jobs
C 30 35 55 60
D 20 30 75 45
a. Use the assignment method to determine which jobs should be assigned to

which work centers to minimize total processing costs.
b. What is the cost of your assignments in Part a?
27. Four employees must be assigned to four projects. Only one employee can be as-
signed to a project, and all projects must be performed. The costs of the employ-
ees performing the projects are:
Projects
1 2 3 4
Ann $400 $300 $250 $300
Employees
Ben 350 400 275 250

Cal 500 350 200 300
Dee 600 475 375 425
a. Use the assignment method to assign these employees to these projects.

b. What is the total cost of your assignments in Part a?
28. Five customers must be assigned to five stockbrokers in a brokerage house. The
estimated profits for the brokerage house for all possible assignments are shown
below:
Brokers
1 2 3 4 5
A $400 $400 $375 $450 $400
Customers
B 325 525 400 525 425
C 450 575 425 550 400
D 225 450 500 600 450
E 500 550 475 525 500
a. Use the assignment method to assign the five customers to the five different
brokers to maximize profits for the brokerage house.
b. What are the profits from your assignment in Part a?
COMPUTER SOLUTIONS
Several of the problems at the end of Chapter 8 and some of the problems in this ap-
pendix are appropriate for solution with a computer. The three computer problems/cases
at the end of Chapter 8—Quality Pixels Inc., Integrated Products Corporation (IPC),
and Jane Deere Company—are especially intended for solution with the POM Soft-
ware Library, which accompanies this book. In solving these problems, take care to
define the decision variables, determine the solution of the problems using a computer,
and fully interpret the meaning of the solution. This means not only determining the
value of the decision variables (X’s), slack variables (S’s), and the objective function
(Z) but also fully interpreting the meaning of the solution to management. In other
words, what would you recommend that management do as a result of your analysis?
This interpretation should also include an explanation of the meaning of the shadow
prices. Also, you should be aware of unusual features in the solution; for example, you
should determine if alternate optimal solutions are present.
SELECTED BIBLIOGRAPHY
Anderson, David R., Dennis J. Sweeney, and Thomas A. Pannell, David J. Introduction to Practical Linear Program-
Williams. An Introduction to Management Science: Quan- ming. New York: John Wiley & Sons, 1996.
titative Approaches to Decision Making, 9th ed. Cincin-
Perry, C., and K. C. Crellin. “The Precise Management Mean-
nati, OH: South-Western College Publishing, 2000.
ing of a Shadow Price.” Interfaces 12, no. 2 (April 1982):
Camm, Jeffrey D., and James R. Evans. Management Science 61–63.
and Decision Technology. Cincinnati, OH: South-Western
Walker, Russell C. Introduction to Mathematical Programming.
College Publishing, 2000.
Upper Saddle River, NJ: Prentice Hall, 1999.
Carter, Michael W., and Camille C. Price. Operations Research:
Winston, Wayne L., and S. Christian Albright. Practical Man-
A Practical Introduction. Boca Raton, FL: CRC Press, 2001.
agement Science, 2nd ed. Pacific Grove, CA: Brooks/Cole,
Hillier, Frederick S., and Gerald J. Lieberman. Introduction to 2001.
Operations Research, 7th ed. Boston: McGraw-Hill, 2001.
ANSWERS TO ODD-NUMBERED PROBLEMS
(1) Third tableau is optimal; X2 68.75, X1 12.5, Z 5,656.25. • (3) Third tableau
is not optimal; X1 will enter and S3 will leave. • (5) X1 175 hundred gallons of pre-
mium ice cream, X2 175 hundred gallons of light ice cream, S1 0 unused mixing
capacity, S2 0 unused high-grade milk, Z $35,000 profit next week. • (7) X1
800 tons of sand, X2 700 tons of gravel, S2 0 so exactly 40% of the concrete is
sand, S3 100 so amount of gravel in the concrete exceeded the 30% requirement by
100 tons, Z $10,400 total cost of sand and gravel each day. • (9) X1 15 cases of
Product A per week, X2 10 cases of Product B per week, S1 10.5 hours of un-
used Cutting Department capacity per week, S2 0 hours of unused Trimming De-
partment capacity per week, S3 0 hours of unused Coating Department capacity per
week, Z $1,830 total profit per week. • (11) X1 500,000 pounds of carrier, X2
100,000 pounds of filler/color, X3 100,000 pounds of adhesive, Z $370,000, S1,
S2, and S3 0. • (13) a. Make 175 hundred gallons of premium ice cream and 175
hundred gallons of light ice cream per week for a profit of $35,000 per week. All mix-
ing machine capacity and milk available will be used; b. X1 (0), X2 (0), because these
variables are in the optimal solution; S1 (83.33) means weekly profit would increase
$83.33 for each additional mixing machine hour that could be found; S2 (0.833)
means weekly profit would increase $0.833 for each additional gallon of milk that could
be found. • (15) a. Use 800 tons of sand and 700 tons of gravel per day for a cost of
$10,400 per week. Sand will be at the maximum limit, and gravel will be above the
minimum limit by 100 tons; b. X1 (0), X2 (0), because these variables are in the opti-
mal solution; S2 (2) means daily cost would decrease $2 for each additional ton in-
crease in the maximum limit of sand; S3 (0) means daily cost would not change if the
minimum limit of gravel were lowered. • (17) a. From Mill 1 ship 18,000 units to Ware-
house B and 14,000 units to Warehouse C; from Mill 2 ship 30,000 units to Warehouse
A, 8,000 units to Warehouse C, and 2,000 units to Warehouse D; from Mill 3 ship
25,000 units to Warehouse D; b. The values in the constraints section represent the
amount of change in Z for a one-unit change in each constraint’s RHS. For example,
the 2.6 in the third constraint means that if Mill 3’s capacity were increased from 25,000
to 25,001, Z would decrease from $488,600 to $488,597.40. The values in the deci-
sion variable section represent the amount of change in Z if one unit of a nonbasic
variable were forced into the solution. For example, if we were forced to ship one unit
from Mill 1 to Warehouse A, Z would be increased from $488,600 to $488,603.10.
• (19) From Source A ship 200 units to Destination 2 and 100 units to Destination 3;
from Source B ship 300 units to Destination 2; from Source C ship 200 units to Des-
tination 1 and 200 units to Destination 3; total shipping cost will be $540. • (21) See
the answer to Problem 17. • (23) Atlanta plant should ship 400 systems to Chicago and
1,200 to New York; El Paso plant should ship 100 systems to Chicago, 1,000 to Dal-
las, 800 to Denver, and 500 to San Jose; monthly shipping cost will be $191,000. • (25)
Alternative optimal solutions are available for this problem. One solution: From Plant
A ship 70,000 pounds to Warehouse 1; from Plant B ship 60,000 pounds to Warehouse
B; from Plant C ship 90,000 pounds to Warehouse 2; monthly shipping cost will be
$4,000. • (27) Assign Ann to Project 2, Ben to Project 1, Cal to Project 3, and Dee to
Project 4; cost of the assignments will be $1,275.

Appendix 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

Appendix Linear Programming

Uploaded by

Copyright:

Available Formats

Appendix 1

Appendix: Linear Programming

Introduction Unbalanced Problems

Simplex Maximization Solutions

Table 1 Steps in the Simplex Solution Method

1. Formulate the objective and constraint functions.

Example 1 Simplex Solution of LP-1, a Maximization Problem

S1 4,000 unused fabrication hours per month

First Tableau (Continued)

Second Tableau (Continued)

ZX2 (900)(0) (600)(1) (0)(0) 600

Table 2 Steps in Performing Row Operations

Simplex Minimization Solutions

Example 2 Simplex Solution of LP-2, a Minimization Problem

1. Add slack variables to constraints to convert from 傶 to ⴝ:

(C ⴚ Z) 2,500 0 7,500 0 7,500 M

1/2 1/2 0 12 / 1 1/2 1

Interpreting Simplex Solutions

Example 3 Interpreting Simplex Solutions: The Oklahoma Crude Oil Company

X1 thousands of gallons of Oklahoma crude to be purchased per month

Min Z 200X1 140X2 150X3 180X4 120X5

A computer solution (last tableau) is presented in Table 3.

Appendix: Linear Programming Solution Methods

S6 0 (no excess lubricating oil base will be supplied monthly)

Unusual Features of Some LP Problems

1/7(13/4) 1/7(1) 1/7(0) 1/7(1/48) 1/7(7/48)

1 X1 13/4 1 0 1/48 7/48

We should always examine Row (C  Z) to inspect for alternate optimal solutions.

In Chapter 8, we discussed selecting a set of shipments from sources to destinations

Characteristics of Transportation Problems

Solution Procedures of the Transportation Method

Table 4 Procedures of the Transportation Method

1. Formulate the problem in a transportation table.

Example 4 The Northwest Corner Rule and the Stepping-Stone Method

Monthly Transportation Costs

Figure 1 Transportation Tables of Example 4

Table 5 Stepping-Stone Method

Logic of the Stepping-Stones

4. Develop a new transportation table that is an improved solution. An improved solution

Example 4 uses the well-known stepping-stone method to determine optimality and

Example 5 MODI Method of Testing Transportation Tables for Optimality

Calculating Ri and Kj Calculating Circuit Costs

Example 6 Starting Solutions with Vogel’s Approximation Method (VAM)

Di Row Units Allocated

Table 6 Steps of the VAM Method

Figure 2 An Unbalanced Transportation Problem

A 3.5 1,000 2.0 4.0 0 1,000

B 4.0 2,000 2.5 3,000 1.5 0 5,000

C 1,000 2.0 1,000 3.0 3.0 2,000 0 4,000

Figure 3 A Degenerate Transportation Table

0.7 0.5 0.5 0.6

0.5 0.7 0.8 1.2

the solution in Transportation Table #2 to be degenerate, a condition comparable to a

Example 7 Using the Assignment Method to Assign n Jobs to n Work Centers

Figure 4 Tables of the Assignment Method in Example 7

B 100 0 200 0 0 B 100 0 200 0 0

REVIEW AND DISCUSSION QUESTIONS

(LP-A, B, C, D, E, F, G, H, I, and J problems are found at the end of Chapter 8.)

X1 number of chairs to produce per week

X1 hundreds of gallons of premium ice cream to produce per week

100 X1 175 1 0 35/12 1/48

Z 35,000 100 100 83.33 0.833

X1 cases of relays per quarter

(C ⴚ Z) 2,500 0 7,500 0 7,500 M

1/2 1/2 0 12 / 1 1/2 1

1/7(13/4) 1/7(1) 1/7(0) 1/7(1/48) 1/7(7/48)

1 X1 13/4 1 0 1/48 7/48

We should always examine Row (C Z) to inspect for alternate optimal solutions.

100 X1 175 1 0 35/12 1/48

250 X1 142.86 1 0 0 5/14 1/7