Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
TEACHING SUGGESTIONS
Teaching Suggestion M7.1: Meaning of Slack Variables.
Slack variables have an important physical interpretation and represent a valuable commodity,
such as unused labor, machine time, money, space, and so forth.
Teaching Suggestion M7.2: Initial Solutions to LP Problems.
Explain that all initial solutions begin with X1 = 0, X2 = 0 (that is, the real variables set to zero),
and the slacks are the variables with nonzero values. Variables with values of zero are called
nonbasic and those with nonzero values are said to be basic.
Teaching Suggestion M7.3: Substitution Rates in a Simplex Tableau.
Perhaps the most confusing pieces of information to interpret in a simplex tableau are
“substitution rates.” These numbers should be explained very clearly for the first tableau because
they will have a clear physical meaning. Warn the students that in subsequent tableaus the
interpretation is the same but will not be as clear because we are dealing with marginal rates of
substitution.
Teaching Suggestion M7.4: Hand Calculations in a Simplex Tableau.
It is almost impossible to walk through even a small simplex problem (two variables, two
constraints) without making at least one arithmetic error. This can be maddening for students
who know what the correct solution should be but can’t reach it. We suggest two tips:
1. Encourage students to also solve the assigned problem by computer and to request
the detailed simplex output. They can now check their work at each iteration.
2. Stress the importance of interpreting the numbers in the tableau at each iteration.
The 0s and 1s in the columns of the variables in the solutions are arithmetic checks and
balances at each step.
Teaching Suggestion M7.5: Infeasibility Is a Major Problem in Large LP Problems.
As we noted in Teaching Suggestion 7.6, students should be aware that infeasibility commonly
arises in large, real-world-sized problems. This module deals with how to spot the problem (and
is very straightforward), but the real issue is how to correct the improper formulation. This is
often a management issue.
ALTERNATIVE EXAMPLES
Alternative Example M7.1: Simplex Solution to Alternative Example 7.1 (see Chapter 7 of
Solutions Manual for formulation and graphical solution).
1st Iteration
Cj® Solution 3 9 0 0
¯ Mix Quantity
X1 X2 S 1 S 2
X1 X2 S 1 S 2
0 S1 1 4 1 0 24
0 S2 1 2 0 1 16
Zj 0 0 0 0 0
Cj – Zj 3 9 0 0
2nd Iteration
Cj® Solution 3 9 0 0
¯ Mix Quantity
X1 X2 S 1 S 2
9 X2 1 0 6
0 S2 0 1 4
Zj 9 0 54
Cj – Zj 0 0
This is not an optimum solution since the X1 column contains a positive value. More profit
3 X1 1 0 -1 2 8
Zj 3 9 60
Cj – Zj 0 0
This is an optimum solution since there are no positive values in the Cj – Zj row. This says to
make 4 of item #2 and 8 of item #1 to get a profit of $60.
Alternative Example M7.2: Set up an initial simplex tableau, given the following two
constraints and objective function:
Minimize Z = 8X1 + 6X2
Subject to: 2X1 + 4X2 ³ 8
3X1 + 2X2 ³ 6
The constraints and objective function may be rewritten as:
Minimize = 8X1 + 6X2 + 0S1 + 0S2 + MA1 + MA2
2X1 + 4X2 – 1S1 + 0S2 + 1A1 + 0A2 = 8
3X1 + 2X2 + 0S1 – 1S2 + 0A1 + 1A2 = 6
The first tableau would be:
Cj® Solution
¯ Mix 8 6 0 0 M M Quantity
X1 X2 S1 S2 A1 A2
M A1 2 4 –1 0 1 0 8
M A2 3 2 0 –1 0 1 6
Zj 5M 6M –M –M M M 14M
M A2 3 2 0 –1 0 1 6
Zj 5M 6M –M –M M M 14M
Cj – Zj 8 – 5M 6 – 6M M M 0 0
The second tableau:
Cj® Solution 8 6 0 0 M M
¯ Mix Quantity
X1 X2 S1 S2 A1 A2
6 X2 1 0 0 2
M A2 2 0 –1 1 2
Zj 3 + 2M 6 –M M 12 + 2M
M M
Cj – Zj 5 – 2M 0 M 0
M M
The third and final tableau:
Cj® Solution 8 6 0 0 M M
¯ Mix Quantity
X1 X2 S 1 S2 A1 A2
6 X2 0 1
8 X1 1 0 1
Zj 8 6 17
Cj – Zj 0 0
M– M–
A minimal, optimum cost of 17 can be achieved by using 1 of a type #1 and of a type #2.
Alternative Example M7.3: Referring back to Hal, in Alternative Example 7.1, we had a
formulation of:
Maximize Profit = $3X1 + $9X2
Subject to: 1X1 + 4X2 £ 24 clay
1X1 + 2X2 £ 16 glaze
where X1 = small vases made
X2 = large vases made
The optimal solution was X1 = 8, X2 = 4. Profit = $60.
Using software (see the printout), we can perform a variety of sensitivity analyses on this
solution.
Alternative Example M7.4: Levine Micros assembles both laptop and desktop personal
computers. Each laptop yields $160 in profit; each desktop $200.
The firm’s LP primal is:
Maximize profit = $160X1 + $200X2
subject to: 1X1 + 2X2 £ 20 labor hours
9X1 + 9X2 £ 108 RAM chips
12X1 + 6X2 £ $120 royalty fees
where X1 = no. laptops assembled daily
X2 = no. desktops assembled daily
Here is the primal optimal solution and final simplex tableau.
Cj® Solution $160 $200 0 0 0
¯ Mix Quantity
X1 X2 S1 S2 S3
200 X2 0 1 1 0 8
160 X1 1 0 –1 0 4
0 S3 0 0 6 –2 1 24
Zj 160 200 40 0 $2,240
Cj – Zj 0 0 –40 0
or X1 = 4, X2 = 8, S3 = $24 in slack royalty fees paid
Profit = $2,240/day
Here is the dual formulation:
Minimize Z = 20y1 + 108y2 + 120y3
subject to: 1y1 + 9y2 + 12y3 ³ 160
2y1 + 9y2 + 6y3 ³ 200
Here is the dual optimal solution and final tableau.
Cj® Solution 20 108 120 0 0
Here is the dual optimal solution and final tableau.
Cj® Solution 20 108 120 0 0
¯ Mix Quantity
y1 y2 y3 S1 S2
108 y2 0 1 2
20 y1 1 0 –6 1 –1 40
Zj 20 108 96 –4 –8 $2,240
Cj – Zj 0 0 +24 +4 +8
This means
y1 = marginal value of one more labor hour = $40
y2 = marginal value of one more RAM chip = $13.33
y3 = marginal value of one more $1 in royalty fees = $0
SOLUTIONS TO DISCUSSION QUESTIONS AND PROBLEMS
M7-1. The purpose of the simplex method is to find the optimal solution to LP problems in a
systematic and efficient manner. The procedures are described in detail in Section M7.3.
M7-2. Differences between graphical and simplex methods: (1) Graphical method can be used
only when two variables are in model; simplex can handle any dimensions. (2) Graphical method
must evaluate all corner points (if the corner point method is used); simplex checks a lesser
number of corners. (3) Simplex method can be automated and computerized. (4) Simplex method
involves use of surplus, slack, and artificial variables but provides useful economic data as a by-
product.
Similarities: (1) Both methods find the optimal solution at a corner point. (2) Both methods
require a feasible region and the same problem structure, that is, objective function and
constraints.
The graphical method is preferable when the problem has two variables and only two or three
constraints (and when no computer is available).
M7-3. Slack variables convert £ constraints into equalities for the simplex table. They represent
a quantity of unused resource and have a zero coefficient in the objective function.
Surplus variables convert ³ constraints into equalities and represent a resource usage above
the minimum required. They, too, have a zero coefficient in the objective function.
Artificial variables have no physical meaning but are used with the constraints that are = or ³.
They carry a high coefficient, so they are quickly removed from the initial solution.
M7-4. The number of basic variables (i.e., variables in the solution) is always equal to the
number of constraints. So in this case there will be eight basic variables. A nonbasic variable is
one that is not currently in the solution, that is, not listed in the solution mix column of the
tableau. It should be noted that while there will be eight basic variables, the values of some of
them may be zero.
M7-5. Pivot column: Select the variable column with the largest positive Cj – Zj value (in a
maximization problem) or smallest negative Cj – Zj value (in a minimization problem).
Pivot row: Select the row with the smallest quantity-to-column ratio that is a nonnegative
number.
Pivot number: Defined to be at the intersection of the pivot column and pivot row.
M7-6. Maximization and minimization problems are quite similar in the application of the
simplex method. Minimization problems usually include constraints necessitating artificial and
surplus variables. In terms of technique, the Cj – Zj row is the main difference. In maximization
problems, the greatest positive Cj – Zj indicates the new pivot column; in minimization problems,
surplus variables. In terms of technique, the Cj – Zj row is the main difference. In maximization
problems, the greatest positive Cj – Zj indicates the new pivot column; in minimization problems,
it’s the smallest negative Cj – Zj. The Zj entry in the “quantity” column stands for profit
contribution or cost, in maximization and minimization problems, respectively.
M7-7. The Zj values indicate the opportunity cost of bringing one unit of a variable into the
solution mix.
M7-8. The Cj – Zj value is the net change in the value of the objective function that would result
from bringing one unit of the corresponding variable into the solution.
M7-9. The minimum ratio criterion used to select the pivot row at each iteration is important
because it gives the maximum number of units of the new variable that can enter the solution. By
choosing the minimum ratio, we ensure feasibility at the next iteration. Without the rule, an
infeasible solution may occur.
M7-10. The variable with the largest objective function coefficient should enter as the first
decision variable into the second tableau for a maximization problem. Hence X3 (with a value of
$12) will enter first. In the minimization problem, the least-cost coefficient is X1, with a $2.5
objective coefficient. X1 will enter first.
M7-11. If an artificial variable is in the final solution, the problem is infeasible. The person
formulating the problem should look for the cause, usually conflicting constraints.
M7-12. An optimal solution will still be reached if any positive Cj – Zj value is chosen. This
procedure will result in a better (more profitable) solution at each iteration, but it may take more
iterations before the optimum is reached.
M7-13. A shadow price is the value of one additional unit of a scarce resource. The solutions to
the Ui dual variables are the primal’s shadow prices. In the primal, the negatives of the Cj – Zj
values in the slack variable columns are the shadow prices.
M7-14. The dual will have 8 constraints and 12 variables.
M7-15. The right-hand-side values in the primal become the dual’s objective function
coefficients.
The primal objective function coefficients become the right-hand-side values of dual
constraints.
The transpose of the primal constraint coefficients become the dual constraint coefficients,
with constraint inequality signs reversed.
M7-16. The student is to write his or her own LP primal problem of the form:
maximize profit = C1X1 + C2X2
subject to A11X1 + A12X2 £ B1
A21X1 + A22X2 £ B2
and for a dual of the nature:
minimize cost = B1U1 + B2U2
subject to A11U1 + A21U2 ³ C1
A12U1 + A22U2 ³ C2
M7-17. a.
b. The new optimal corner point is (0,60) and the profit is 7,200.
c. The shadow price = (increase in profit)/(increase in right-hand side value)
(7,200 – 2,400)/(240 – 80)
=
= 4,800/160
= 30
d. With the additional change, the optimal corner point in part B is still the optimal
corner point. Profit doesn’t change. Once the right-hand side went beyond 240, another
constraint prevented any additional profit, and there is now slack for the first constraint.
M7-18. a. See the table below.
Table for Problem M7-18
Cj® Solution $900 $1,500 $0 $0
¯ Mix Quantity
X1 X2 S1 S2
0 S1 14 4 1 0 3,360
0 S2 10 12 0 1 9,600
Zj 0 0 0 0 0
Cj – Zj 900 1,500 0 0
b. 14X1 + 4X2 £ 3,360
10X1 + 12X2 £ 9,600
X1, X2 ³ 0
c. Maximize profit = 900X1 + 1,500X2
d. Basis is S1 = 3,360, S2 = 9,600.
e. X2 should enter basis next.
f. S2 will leave next.
g. 800 units of X2 will be in the solution at the second tableau.
h. Profit will increase by (Cj – Zj)(units of variable entering the solution)
= (1,500)(800) = 1,200,000
M7-19. a. Maximize earnings = 0.8X1 + 0.4X2 + 1.2X3 – 0.1X4 + 0S1 + 0S2 – MA1 – MA2
subject to
X1 + 2X2 + X3 + 5X4 + S1 = 150
X2 – 4X3 + 8X4 + A1 = 70
6X1 + 7X2 + 2X3 – X4 – S2 + A2 = 120
b. See initial simplex tableau in Table M7-19b below.
Table for Problem M7-19b
Cj® Solution 0.8 0.4 1.2 -0.1 0 0 -M -M
¯ Mix Quantity
X1 X2 X3 X4 S1 S2 A1 A2
0 S1 1 2 1 5 1 0 0 0 150
–M A 0 1 –4 8 0 0 1 0 70
1 2 3 4 1 2 1 2
0 S1 1 2 1 5 1 0 0 0 150
–M A1 0 1 –4 8 0 0 1 0 70
–M A2 6 7 2 –1 0 –1 0 1 120
Zj –6M –8M 2M –7M 0 M –M –M –190M
Cj – Zj 0.8 + 6M 0.4 + 8M 1.2 – 2M –0.1 + 7M 0 –M 0 0
b.
Cj® Solution 10 8 0 0 Quantity
¯ Mix
X1 X2 S 1 S 2
0 S1 4 2 1 0 80
0 S2 1 2 0 1 50
Zj 0 0 0 0 0
Cj – Zj 10 8 0 0
This represents the corner point (0,0).
c. The pivot column is the X1 column. The entering variable is X1.
d. Ratios: Row 1: 80/4 = 20
Row 2: 50/1 = 50
These represent the points (20,0) and (50,0) on the graph.
e. The smallest ratio is 20, so 20 units of the entering variable (X1) will be brought into
the solution. If the largest ratio had been selected, the next tableau would represent an
infeasible solution since the point (50,0) is outside the feasible region.
f. The leaving variable is the solution mix variable in row with the smallest ratio.
Thus, S1 is the leaving variable. The value of this will be 0 in the next tableau.
g.
Second iteration
Cj® Solution 10 8 0 0 Quantity
¯ Mix
X1 X2 S1 S2
10 X1 1 0.5 0.25 0 20
0 S2 0 1.5 –0.25 1 30
Zj 10 5 2.5 0 200
Cj – Zj 0 3 –2.5 0
Third iteration
Cj® Solution 10 8 0 0 Quantity
¯ Mix
X1 X2 S1 S2
10 X1 1 0 0.3333 –0.3333 10
X1 X2 S1 S2
10 X1 1 0 0.3333 –0.3333 10
8 X2 0 1 –0.1667 0.6667 20
Zj 10 8 2 2 260
Cj – Zj 0 0 –2 –2
h. he second iteration represents the corner point (20,0). The third (and final) iteration
T
represents the point (10,20).
M7-22. Basis for first tableau: A1 = 80
A2 = 75
(X1 = 0, X2 = 0, S1 = 0, S2 = 0)
Second tableau: A1 = 55
X1 = 25
(X2 = 0, S1 = 0, S2 = 0, A2 = 0)
Graphical solution to Problem M7-22:
Zj 2 3 1 0 0 $6
Cj – Zj 0 0 –1 0 –M
b. The variable X2 has a Cj – Zj value of $0, indicating an alternative optimal solution
exists by inserting X2 into the basis.
c. The alternative optimal solution is found in the tableau in the next column to be X1 =
Cj – Zj 0 0 0 –M
0 S1 1 0 0 1 0 0 1,500
Zj M –M + 75 75 0 M 1,000M + 75,000
Cj – Zj –M + 50 M – 65 0 0 0
M–
Third iteration:
Cj® Solution 50 10 75 0 M M
¯ Mix Quantity
X1 X2 X3 S1 A1 A2
50 X1 1 –1 0 0 1 0 1,000
75 X3 0 1 1 0 0 1,000
75 X3 0 1 1 0 0 1,000
0 S1 0 1 0 1 –1 0 500
Zj 50 25 75 0 50 $125,000
Cj – Zj 0 –15 0 0 M – 50
M–
Fourth and final iteration:
Cj® Solution 50 10 75 0 M M
¯ Mix Quantity
X1 X2 X3 S1 A1 A2
50 X1 1 0 0 1 0 0 1,500
75 X3 0 0 1 –1 1 500
10 X2 0 1 0 1 –1 0 500
Zj 50 10 75 –15 65 $117,500
Cj – Zj 0 0 0 15 M – 65
M–
X1 = 1,500, X2 = 500, X3 = 500, Z = $117,500
M7-28. X1 = number of kilograms of brand A added to each batch
X2 = number of kilograms of brand B added to each batch
Minimize costs = 9X1 + 15X2 + 0S1 + 0S2 + MA1 + MA2
subject to X1 + 2X2 – S1 + A1 = 30
X1 + 4X2 – S2 + A2 = 80
Cj® Solution $9 $15 $0 $0 M M
¯ Mix Quantity
X1 X2 S1 S2 A1 A2
M A1 1 2 –1 0 1 0 30
M A2 1 4 0 –1 0 1 80
Zj 2M 6M –M –M M M 110M
Cj – Zj –2M + 9 –6M + 15 M M 0 0
First iteration:
Cj® Solution $9 $15 $0 $0 M M
¯ Mix Quantity
X1 X2 S1 S2 A1 A2
15 X2 1 0 0 15
M A2 –1 0 2 –1 –2 1 20
Zj 15 –M M 225 + 20M
–M + 2M – 2M
Cj – Zj 0 M 0
+M – 2M 3M –
Second iteration:
Cj® Solution $9 $15 $0 $0 M M
¯ Mix Quantity
X1 X2 S 1 S 2 A1 A2
15 X2 1 0 0 20
0 S 0 1 –1 10
15 X2 1 0 0 20
0 S1 0 1 –1 10
Zj 15 4 0 $300
Cj – Zj 0 0 M
M–
Third and final iteration:
X1 = 0 kg, X2 = 20 kg, cost = $300
M7-29. X1 = number of mattresses
X2 = number of box springs
Minimize cost = 20X1 + 24X2
subject to X1 + X2 ³ 30
X1 + 2X3 ³ 40
X1, X2 ³ 0
Initial tableau:
Cj® Solution $20 $24 $0 $0 M M
¯ Mix Quantity
X1 X2 S1 S2 A1 A2
M A1 1 1 –1 0 1 0 30
M A2 1 2 0 –1 0 1 40
Zj 2M 3M –M –M M M 70M
Cj – Zj –2M + 20 –3M + 24 M M 0 0
Second tableau:
Cj® Solution $20 $24 $0 $0 M M
¯ Mix Quantity
X1 X2 S 1 S2 A1 A2
M A1 0 –1 1 10
$24 X2 1 0 0 20
Zj 24 –M 0 10M + 480
M + 12 M – 12 M + 12
Cj – Zj 0 M 0
M + 12 M + 12 M – 12
Final tableau:
Cj® Solution $20 $24 $0 $0 M M
¯ Mix Quantity
X1 X2 S 1 S 2 A1 A2
$20 X1 1 0 –2 1 2 –1 20
$24 X2 0 1 1 –1 –1 1 10
Zj 20 24 –16 –4 16 4 $640
Cj – Zj 0 0 16 4 M – 16 M – 4
X1 = 20, X2 = 10, cost = $640
subject to X1 + X2 £ 10
X1 + 2X2 £ 12
X1, X2 ³ 0
Initial tableau:
Cj® Solution $9 $12 $0 $0
¯ Mix Quantity
X1 X2 S 1 S 2
$0 S1 1 1 1 0 10
$0 S2 1 2 0 1 12
Zj 0 0 0 0 $0
Cj – Zj 9 12 0 0
Second tableau:
Cj® Solution $9 $12 $0 $0
¯ Mix Quantity
X1 X2 S 1 S 2
$0 S1 0 1 4
$12 X2 1 0 6
Zj 6 12 0 6 $72
Cj – Zj 3 0 0 –6
Final tableau:
Cj® Solution $9 $12 $0 $0
¯ Mix Quantity
X1 X2 S 1 S 2
$4 X1 1 0 2 –1 8
$12 X2 0 1 –1 1 2
Zj 9 12 6 3 $96
Cj – Zj 0 0 –6 –3
X1 = 8, X2 = 2, profit = $96
M7-31. Maximize profit = 8X1 + 6X2 + 14X3
subject to 2X1 + X2 + 3X3 £ 120
2X1 + 6X2 + 4X3 = 240
X1, X2 ³ 0
Initial tableau:
Cj® Solution $8 $6 $14 0 -M
¯ Mix Quantity
X1 X2 X3 S1 A1
0 S1 2 1 3 1 0 120
–M A2 2 6 4 0 1 240
Zj –2M –6M –4M 0 –M –240M
Cj – Zj 8 + 2M 6 + 6M 14 + 4M 0 0
Second tableau:
Cj – Zj 8 + 2M 6 + 6M 14 + 4M 0 0
Second tableau:
Cj® Solution $8 $6 $14 0 -M
¯ Mix Quantity
X1 X2 X3 S 1 A1
$0 S1 0 1 80
$6 X2 1 0 40
Zj 2 6 4 0 1 $240
Cj – Zj 6 0 10 0 –M – 1
Final tableau:
Cj® Solution $8 $6 $14 0 -M
¯ Mix Quantity
X1 X2 X3 S1 A1
$14 X3 0 1
$6 X2 1 0
Zj 6 14
Cj – Zj –1.1 0 0
–M –
The last two constraints can be rewritten as:
–10 – 4D £ 0 or D³
From the S1 column, we require that
–20 – 2D £ 0 or D ³ –10
1
–20 – 2D £ 0 or D ³ –10
$ £ Cj (for X1) ¥
b. The range of insignificance is
M7-3