REVISED

M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 158

C H A P T E R

Integer Programming, Goal Programming, Nonlinear Programming and the Branch and Bound Methods
TEACHING SUGGESTIONS
Teaching Suggestion 11.1: Topics in This Chapter. The overall purpose of this chapter is to provide a framework for the topics of integer programming, branch and bound, nonlinear programming and goal programming. These are fairly advanced topics in a mathematical sense, and the chapters intention is solely to introduce them through a series of simple graphical problems. Some of the topics are on the cutting edge of QA. For example, in integer and nonlinear programming, no one solution procedure exists to handle all problems. Teaching Suggestion 11.2: Using the Computer to Solve Mixed-Integer Programming Problems. Note that the Excel printout in Program 11.2 allows users to specify which variables are integers and which, by default, can be fractional. Teaching Suggestion 11.3: How the Branch and Bound Method Can Help. In this section we illustrate how branch and bound is used to solve small assignment and integer programming problems. But its real strength is in dealing with huge problems (for example, thousands of variables/constraints). Branch and bound allows us to divide a large problem into smaller parts, thereby eliminating one-half or two-thirds of the options and reducing the problem to a more manageable level. Teaching Suggestion 11.4: Multiple Goals. Ask studdents what other goals a company might have beyond maximizing prot. Socially conscious rms need to state as their mission a whole series of objectives. Encourage students to research an article showing a goal programming application. There is a wealth of research in journals. One interesting application is in the box later in this section that deals with budgeting for prisons. Teaching Suggestion 11.5: Deviational Variables Are the Key in Goal Programming. The concept of deviational variables requires careful explanation to the class. Students are accustomed to the decision variables of X1 and X2. Now they need to concentrate on goal achievement. The minus and plus signs on deviational variables need a thoughtful classroom discussion. Teaching Suggestion 11.6: Difculty of Graphical Goal Programming. Solving goal programming problems graphically can be a confusing concept relative to graphical LP. Students often have difculty with the direction of deviational variables. Teaching Suggestion 11.7: Using the Goal Programming Simplex Method. Point out the similarities and differences between the simplex method and the modied goal programming tableau. You can show that the structure is almost the same. The big change is the addition of two rows for each new goal. Surprisingly, the computation is not as difcult as it looks.

11

ALTERNATIVE EXAMPLES
Alternative Example 11.1: 01 Integer Programming. Indianas prison budget allows it to consider four new installations next year. They are X1 1 if maximum security prison in Ft. Wayne, 0 otherwise X2 1 if minimum security prison in Bloomington, 0 otherwise X3 1 if halfway house in Indianapolis, 0 otherwise X4 1 if expanded tricounty jail in South Bend, 0 otherwise The state wants to maximize the number of people that can be served, while only building one of the two prisons (X1 or X2) and observing cost and space limitations. Here is the formulation: maximize number served 3,000X1 900X2 4,000X3 1,500X4 subject to X1 X2 1 prison 4X1 2X2 7X3 3X4 12 acres available 3.5X1 1X2 2.5X3 9X4 12 million dollars budgeted Solution: Using software, we nd that X1 1, X2 0, X3 1, X4 0, number served 7,000. Alternative Example 11.2: The Quality University (QU) is a private noncredit training rm that specializes in total quality management (TQM) courses. QU wants to determine how many

158

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 159

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

159

of each of two programs to offer in order to maximize prot. Their integer program can be formulated as follows: maximize prot $8,500X1 $6,000X2 subject to X1 X2 courses max. of 10 $1,000X1 $700X2 instructors pay of $7,200 X1, X2 0 and are integers Using LP, the solution is: X1 X\c, X2 9Z\c, prot $61,667.

SOLUTION Initial upper bound (UB) $61,667 (X1 X\c, X2 9Z\c) Initial lower bound (LB) $54,000 (X1 0, X2 9) (See graph below for this example.) All nodes are either integer or infeasible, so the solution is seen to be X1 3, X2 6, prot $61,500.

X1 = 0 X2 = 10 P = $60,000

X1
X1 = 2/3 X2 = 9 1/3 P = $61,667

UB = $61,667 LB = $54,000 X 1 1

X
X1 = 1 X2 = 8 7/8 P = $61,643

9 2
8

Not Feasible X1 = 1 X2 = 8 P = $56,500 Not Feasible X1 = 2 X2 = 7 P = $59,000

X2

X1
X1 = 1 3/5 X2 = 8 P = $61,000

X1

2
X1 = 2 X2 = 7 3/7 P = $61,571

X2

Figure for Alternative Example 11.2

UB = $61,600 LB = $54,000 [X1 = 0, X2 = 9]

X2

7
X1 = 2 3/10 X2 = 7 P = $61,550

X1

X1

3
X1 = 3 X2 = 6 P = $61,500

Optimal Solution

Alternative Example 11.3: Minimize P1d1 P2d2 P3d3 P4d1 subject to 2x1 4x2 d1 d1 80 2x1 2.5x2 d2 d2 80 2x1 1.5x2 d3 d3 60 All variables 0 See the graph to the right:

Graph for Alternative Example 11.3

X2 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 X1 2X1 + 21/2X2 = 80 2X1 + 4X2 = 80 2X1 + 11/2 X2 = 60

Goal Constraints

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 160

160

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

First Priority Goal: Minimize d1

X2 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 X1 d 1 d 1+ 2X1 + 4X2 = 80

Third Priority Goal: Minimize d3

X2 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 X1 B d 3+ 2X1 + 11/2 X2 = 60 d 2+ d 2 d 1+ d 1 C

d 3 A

The area below the constraint line d1 is eliminated. Second Priority Goal: Minimize d2
X2 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30

The area above the constraint line 2X1 1Z\xX2 60 is eliminated. Fourth Priority Goal: Minimize d1
X2 45 40 B d 3+ 2X1 + 11/2X2 = 60 d 2+ d 2 d 1+ d 1

d 2+ d 2 d 1+ d 1 2X1 + 21/2X2 = 80

35 30 25 20 15 10 5 35 40 45 X1 0

d 3 A

X1 = 15 X2 = 20

10

15

20

25

D 30 35

40

45

X1

The area below the constraint line d2 is eliminated.

Cannot minimize d1 totally without violating rst two priority goals. SOLUTION X1 X2 d1

15 20 30

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 161

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

161

Alternative Example 11.4: Here is the simplex solution to the goal programming problem in Alternative Example 11.3. Initial Goal Programming Tableau
Cj l b Solution Mix d1 d2 d3 P4 P3 P2 P1

0 X1 2 2 2 0 0 2 2

0 X2 4 2 212 112
1

P1 d1 1 0 0 0 0 0 0 Pivot column

P2 d2 0 1 0 0 0 0 0

0 d3 0 0 1 0 0 0 0

P4 d1 1 0 0 1 0 0 1

0 d2 0 1 0 0 0 1 0

P3 d3 0 0 1 0 1 0 0 Quantity 80 80 60 0 0 80 80

Pivot row

The Second Goal Programming Tableau

Cj l b 0 P2 0 Cj Zj Solution Mix X2 d2 d3 P4 P3 P2 P1 0 X1 2 2 5 2
1 3

0 X2 1 0 0 0 0 0 0

Pivot row

The Third Goal Programming Tableau

Cj l b 0 P4 0 Cj Zj Solution Mix X2 d1 d3 P4 P3 P2 P1 0 X1 5 5 8 5
4 3

0 X2 1 0 0 0 0 0 0

P1 d1 0 1 0 1 0 0 1

P2 d2 5 5 35
1 2

0 d3 0 0 1 0 0 0 0

P4 d1 0 1 0 0 0 0 0

35 025 025 025 5 Pivot column

l

25 025 125 025

Pivot row

0 0 32 0

P1 P2 0 Cj Zj

012 012 212 412

P1 d1 2 52 32
1

P2 d2 0 1 0 0 0 0 0

0 d3 0 0 1 0 0 0 0

P4 d1 2 5 2 3 2
1

0 d2 0 1 0

P3 d3 0 0 1 0 1 0 0 Quantity 20 60 60 0 0 60 0

a
a

052 052 52 152

1 0 52 0 2

0 0 1 0 Pivot column

0 d2 5 25 3 5
1 2 5 025 025 025

P3 d3 0 0 1 0 1 0 0 Quantity 32 24 24 24 0 0 0

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 162

162

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

The Final Goal Programming Tableau

Cj l b 0 P4 0 Cj Zj Solution Mix X2 d1 X1 P4 P3 P2 P1 0 X1 0 0 1 0 0 0 0 0 X2 1 0 0 0 0 0 0 P1 d1 0 1 0 1 0 0 1 P2 d2 2 58 38
1

0 d3 2 38 58
1

P4 d1 0 1 0 0 0 0 0

0 d2 2 58 3 8
1 5

P3 d3 2 8 58
1 3

Quantity 20 15 15 15 0 0 0

58 0 1 0

38 0 0 0

8 0 0 0

38 1 0 0

SOLUTIONS TO DISCUSSION QUESTIONS AND PROBLEMS

11-1. a. Linear programming allows only one goal (for example, prot maximization) whereas goal programming permits multiple goals. b. LP always optimizes; goal programming sometimes only satises. c. In goal programming, we deal with deviational variables as well as real variables. 11-2. When a non-integer solution to an LP problem is found and branching is performed on one of the variables, an additional constraint is added to each of two subproblems. These subproblems have all of the previous constraints plus one new one. Therefore, the feasible region for the subproblem must be smaller than the feasible region for the original LP problem. No new points are added to the feasible region. Consequently, it is impossible for the subproblem to have a better objective function value than the previous LP problem. 11-3. a. Rounding off is the easiest way to solve an integer program, but it can give an infeasible or nonoptimal solution. b. Enumeration is simple in concept, but it can be very time consuming in large problems. c. The branch and bound method, which can be computerized, is especially useful when solving large problems where enumeration is impractical. It does not always reach an optimal solution in large problems, however. 11-4. The three types of integer programs are (1) pure integer programming, where all variables are integer; (2) mixed-integer programming, where some but not all variables are integer; and (3) zeroone integer programming, where all variables are either 0 or 1 in value. 11-5. The upper and lower bounds are limits set at each branch and bound stage on the highest and lowest possible costs of a possible assignment. The process is described in Section 11.2. The bounds help us decide which branches can be discarded. 11-6. Satiscing is a term used in goal programming because it is often not possible to optimize a multi-goal problem. We come as close as possible to reaching goals. 11-7. Deviational variables, similar to slack variables in LP, are the difference between set goals and the current solution. In LP

problems, only real variables are used, representing physical quantities. This is discussed in Section 11.3. 11-8. A college presidents goals might be to (1) increase enrollments by 1,000 students; (2) stay within budget; (3) keep class sizes down to an average of 25 students; (4) increase faculty salaries; (5) develop 10 new off-campus courses; (6) reduce average teaching loads to three courses per semester, and so on. There will be nancial, space, tenure, and many other constraints. 11-9. Ranking goals just means more weight can be placed on one goal over another. The higher-ranked goals must be achieved completely before goal programming moves on to meet lowerranked goals. 11-10. There are four differences between the LP and GP simplex methods. 1. GP has negative and positive deviational variables, each with a priority. 2. The negative deviational variables provide the initial basic feasible solutions and are analogous to slack variables in LP. 3. There is a separate Zj and Cj Zj row for each of the priority goals. 4. The highest-priority row and the most negative Cj Zj value determine the variable to enter the solution next. 11-11. a. b. c. d. e. 11-12. Linear Nonlinear because of 8X1X2 in objective Goal programming 2 Nonlinear because X1 in rst constraint Nonlinear and quadratic objective function

a. Let X number of prime time ads per week Y number of off-peak ads per week Maximize audience exposure 8200X 5100Y Subject to: 390X 240Y 1800 X2 Y6 X, Y 0 Solution: X 2; Y 4.25; audience 38,075 b. X 2, Y 4; audience 36,800 There are other good solutions. c. Optimal integer solution: X 4, Y 1; audience 37,900

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 163

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

163

11-13. Let Xi 1 if item i is selected and 0 otherwise, for i 1 to 8. Maximize 80X1 20X2 50X3 55X4 50X5 75X6 30X7 70X8 Subject to: 8X1 X2 7X3 6X4 3X5 12X6 5X7 14X8 35 Xi 0, 1 Solution using QM for Windows Mixed-Integer Programming Module: X1 X2 X4 X5 X6 X7 1 X3 X8 0 Objective function 310 11-14. X1 number of larger posters X2 number of smaller posters Maximize prot 3X1 2X2 subject to X1 3 X2 5 2X1 X2 10 X1, X2 0 See graph below. Figure for Problem 11-14
10 9 8 7 6 X2 5 4 3 2 1 0 0 1 2 X1 3 4 5 Optimal Solution c a b X2 5 2X1 + X2 10 X1 3

2 e X1 X1

1 e X2 0 (one-third 757s) X2 17 (planes) X1, X2 to be integers 0

800,000X1 500,000X2 $8,000,000 (maintenance)

This is a pure integer programming problem. The QM for Windows integer programming solution is: X1 5 X2 8 Passengers carried 1,273,000 11-16. a. b. c. d. e. X1 X2 X3 X4 X5 X6 3 X1 X4 1 X4 X6 2X5 X2 X3 X5 1 X2 + X3

11-17. Let Xi 1 if location i is selected and 0 otherwise, for i 1 to 6. Minimize X1 X2 X3 X4 X5 X6 Subject to: X1 X6 1 X1 X2 1 X2 X3 1 X1 X3 X6 1 X3 X4 X5 1 X3 X4 X6 1 X2 X5 1 X4 X5 1 Xi 0, 1 for i 1 to 6. Solution using QM for Windows Mixed-Integer Programming Module: X1 X2 X4 1; all other variables 0. Objective function value 3 This means only locations 1, 2, and 4 will be used. 11-18. a.
1 if location i is selected Let: X i = 0 if location i is not selected

Maximize prot $5,000X1 6,000X2 10,000X3 12,000X4 8,000X5 3,000X6 9,000X7 10,000X8 subject to $60,000X1 50,000X2 82,000X3 103,000X4 50,000X5 41,000X6 80,000X7 69,000X8 $300,000 b. X1 0, X2 1, X3 1, X4 0, X5 1, X6 1, X7 0, X8 1, Prot 37,000 11-19. a. Let X1 1 if apartment project is undertaken; 0 otherwise Let X2 1 if shopping center project is undertaken; 0 otherwise Let X3 1 if mini-warehouse project is undertaken; 0 otherwise Maximize NPV 18X1 15X2 14X3 Subject to: 40X1 30X2 20X3 80 301 20X2 20X3 50 X1, X2, X3 1 or 0

Step 1. Optimal LP solution at a is (X1 212, X2 5, prot $17.50). Step 2. Integer solution at b is (X1 3, X2 4, prot $17). Integer solution at c is (X1 2, X2 5, prot $16). Hence the optimal integer solution is X1 3 large posters and X2 4 small posters (seen at point b). 11-15. X1 number of Boeing 757s purchased X2 number of Boeing 767s purchased Maximize passenger carrying capability 125,000X1 81,000X2 subject to 80X1 110X2 1,600 ($ million available) [X1 1 e (X1 X2)] or

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 164

164

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

b. The optimal solution is X1 1, X2 1, X3 0. NPV 33. This means that both the apartment project and the shopping center project will be undertaken. The amount of money spent in year 1 would be $70 (thousand) and in year 2 would be $50 (thousand). 11-20. a. X1 X2 This means that if the apartment is not built (X1 0), the shopping center cannot be built (X2 must equal 0). b. X1 X2 X3 2 11-21. a. Let Xij 1 if generator i is functioning during time period j, and 0 otherwise; where i 1, 2, 3 and j 1 for 62 time period; j 2 for 210 time period; j 3 for 610 time period. Let Yij megawatts produced by generator i in time period j, where i 1, 2, 3 and j 1 for 62 time period; j 2 for 210 time period. Minimize cost 6,000(X11 X12 X13) 5,000(X21 X22 X23) 4,000( X 3 1 X 3 2 X 3 3 ) 8( Y 1 1 Y 1 2 ) 9( Y 2 1 Y 2 2 ) 7(Y31 Y32) Subject to: Y11 Y21 Y31 3,200 megawatts requirements from 62 Y12 Y22 Y32 5,700 megawatts requirements from 210 Y11 2,400(X11 X13) Y12 2,400(X12 X13) Y21 2,100(X21 X23) Y22 2,100(X22 X23) Y31 3,300(X31 X33) Y32 3,300(X32 X33) X11 X12 X13 1 21 X22 X23 1 31 X32 X33 1 ij 0 or 1 for all i, j Yij 0 for all i, j b. The solution is: X12 1, X33 1, Y12 2,400, Y31 3,200, Y31 3,300, total cost $74,700. Thus, generator #1 will be utilized in the period 210 and will generate 2,400 megawatts of electricity. Generator #3 will be started at 6 and utilized for the entire 16 hours. It will generate 3,200 megawatts during the 62 time period, and 3,300 megawatts during the 210 time period. 11-22. Let T number of TV ads, R number of newspaper ads, B number of billboard ads, and N number of newspaper ads.
P2d2 P3d3 P4d4 P4d5 P4d6 P4d Minimize P1d1 7

(3) number of TV or radio ads

T R d 3 d3 6

(4) restriction on number of each individual type of ad

d T d4 4 10 R d 5 d5 10 B d 6 d6 10 N d 7 d7 10

All variables 0 b. T 0, R 0.73, B 0, N 88.86 c. Goal 1 (number of people reached) and goal 2 (budget) are met completely. The number of TV, radio, and billboard ads are each less than 10. The other goals are not met. 11-23. Maximize prot 2X1 3X2 subject to X1 3X2 9 3X1 X2 7 X1 X2 1 X1, X2 0 1. Solve graphically as an LP problem: X1 1.5 X2 2.5 prot $10.50 This provides an upper bound value. 2. Round down to X1 1, X2 2, prot $8.00 for a feasible solution. The lower bound is $8.00. 3. Branch on X2 to begin:
Subproblem A New constraint: X2 2 Optimal solution: X2 2 X1 1.6 prot $9.33 (new upper bound) Subproblem B New constraint: X2 3 Optimal solution: X2 3 X1 0 prot $9.00 (new lower bound)

maximum megawatts from #1 from 62 maximum megawatts from #1 from 210 maximum megawatts from #2 from 62 maximum megawatts from #2 from 210 maximum megawatts from #3 from 62 maximum megawatts from #3 from 210 generator #1 starts up at most once generator #2 starts up at most once generator #3 starts up at most once

4. Branch on X1 now from subproblem A:

Subproblem C New constraints: X2 2 X1 1 Optimal solution: X2 2 X1 1 prot $8.00 Subproblem D New constraints: X2 2 X1 2 Optimal solution: X2 1 X1 2 prot $7.00

Subject to: (1) number of people reached

40,000T 32,000R 34,000B 17,000N d 1 d1 1,500,000

(2) budget
d2 16,000 900T 500R 600B 180N d2 budget constraint

5. Both of these subproblems yield all-integer solutions. Comparing them to the lower bound of $9.00, we see they are both smaller (see the graph in the next column). The solution to the problem (see subproblem B) is X1 0, X2 3, prot $9.00.

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 165

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

165

Figure for Problem 11-23

Subproblem C X1 = 1 X2 = 2 P = 8.00

The best solution is X1 500, X2 400. The value of d3 100, meaning the two-drawer sales goal is underachieved by 100 cabinets. See the graph below. Figure for Problem 11-25

Subproblem A X1 = 1.6 X2 = 2 P = 9.33

2

X1

800
Infeasible, noninteger
X

X2

Upper Bound = $9.33 Lower Bound = $9.00

Subproblem D X1 = 1 X2 = 1 P = 7.00

733 700 650 600 500 X2 400

A Profit Target Three-Drawer Sales Limit B d4

+

X1 = 1.5 X2 = 2.5 P = 10.50

X

Upper Bound = $10.50 Lower Bound = $8.00

d 4 Two-Drawer Sales Limit d 3 200 400 d3

+

Subproblem B X1 = 0 X2 = 3 P = 9.00

300
Feasible, integer solution

Production Limit

200 100

Optimal Solution

d1 600 X1 800

d1

d2 d 2

11-24.

Let: X1 number of two-drawer cabinets produced each week X2 number of three-drawer cabinets produced each week d1 underachievement of prot goal d1 overachievement of prot goal d2 idle time in production capacity d3 underachievement of sales goal for twodrawer les d4 underachievement of sales goal for threedrawer les 11-26.

1,000 1,200 1,100 1,300

X1 number of 64MB chips produced X2 number of 256MB chips produced X3 number of 512MB chips produced d1 underlling customers orders of 64MB chips d2 underlling customers orders of 256MB chips d3 underachievement of sales quotas for 64MB chips d4 underachievement of sales quotas for 256MB chips d5 underachievement of sales quotas for 512MB chips d6 underutilization of plant capacity

Minimize deviations P1d1 P1d1 P2d2 P3d3 P3d4 subject to 10X1 15X2 d1 d1 $11,000 (prot target) 1X1 2X2 1X1 X2 All Xi, di variables 0 11-25. Because we want to achieve the prot goal as closely as possible (minimize both d1 and d1), the line ABC becomes the feasible region. When the P2 priority is included, the feasible region is reduced to the segment AB. P3 priority applies to both d3 and d4. The three-drawer goal (d4) is fully attained at point B and the two-drawer goal (d3) is almost reached. d2 1,300 hours (production limit) d3 600 (two-drawer sales limit) d4 400 (three-drawer sales limit)

Minimize deviations P1d1 P1d2 P2 d3 P2 d4 P2 d5 P3 d6 subject to X1 d1 d1 30 X2 d2 d2 35 X1 d3 d3 40 X2 d4 d4 50 X3 d5 d5 60 All variables 0

(64MB chips order) (256MB chips order) (64MB sales goal) (256MB sales goal) (512MB sales goal)

8X1 13X2 16X3 d6 1,200 (hours capacity)

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 166

166

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

11-27.
Cj l b 0 0 0 P4

Third tableau for Harrison Electric:

0 Solution Mix X1 X2 d3

0 X2 0 1 0 0 0 0 0 0 0 0 0 0

P1 d1
1 3 2 9 8 9 2 9 2 9 2 9

P2 d2
2 3 7 9 1 9 7 9 7 9 7 9

0 d3 0 0 1 0 0 0 0 0 0 0 0 0

P4 d4 0 0 0 1 1 0 0 0 0 0 0 0

0 d1
1 3 2 9 8 9 2 9 2 9 2 9

0 d2
2 3 7 9 1 9 7 9 7 9

P3 d3 0 0 1 0 0 0 0 1 0 0 0 0

0 d4 0 0 0 1 1 1 0 0 0 0 0 0 Pivot row 0 0 0 Quantity 2

8 3 14 3 13 3 13 3

X1 1 0 0 0 0 0 0 0 0 0 0 0

d4

P4

Zj C j Z j Zj C j Z j Zj C Zj j Zj C j Z j

7 9

0 0 0 0 0 1

0 0 0 1 0 0

0 0 0 0 0 0

0 0 0 0 0 0 a Pivot column

P3

P2

P1

Fourth tableau for Harrison Electric:

Cj l b 0 0 0 P4 Solution Mix d2

0 X1
3 2 7 6 1 6 7 6 7 6 7 6

0 X2 0 1 0 0 0 0 0 0 0 0 0 0

P1 d1
1 2 1 6 5 6 1 6

P2 d2 1 0 0 0 0 0 0 0 0 1 0 0

0 d3 0 0 1 0 0 0 0 0 0 0 0 0

P4 d4 0 0 0 1 1 0 0 0 0 0 0 0

0 d1
1 2 1 6 5 6 1 6 1 6

0 d2 1 0 0 0 0 0 0 0 0 0 0 0

P3 d3 0 0 1 0 0 0 0 1 0 0 0 0

0 d4 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 Quantity 3 5 5 2 2

X2 d3 d4

P4

Zj C Zj j

1 6 1 6

1 6

P3

Zj C j Z j Zj C Zj j Zj C j Z j

0 0 0 0 0 0

0 0 0 0 0 1

0 0 0 0 0 0

P2

P1

The third tableau corresponds to point C on the graph. The fourth tableau corresponds to point D on the graph.

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 167

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

167

11-28.
Cj l b P1 P2 0

a.
0 Solution Mix d1 d2

0 X2 4 10 6 0 0 0 0 10 10 4 4 a Pivot column

P1 d1 1 0 0 0 0 0 0 0 0 1 0

P2 d2 0 1 0 0 0 0 0 1 0 0 0

0 d3 0 0 1 0 0 0 0 0 0 0 0

P4 d1 1 0 0 0 1 0 0 0 0 1 1

0 d2 0 1 0 0 0 0 0 1 1 0 0

P3 d3 0 0 1 0 0 0 1 0 0 0 0 80 320 0 Quantity 80 320 240 0 k Pivot row

X1 2 8 8 0 0 0 0 8 8 2 2

d3

P4

Zj C j Z j Zj C j Z j Zj C j Z j Zj C j Z j

P3

P2

P1

11-28.
Cj l b 0 P4 0

b.
0 Solution Mix X2 d1 X1 X1 0 0 1 0 0 0 0 0 0 0 0 0 X2 1 0 0 0 0 0 0 0 0 0 0 P1 d1 0 1 0 1 1 0 0 0 0 0 1 P2 d2
1 4 5 8 3 16

0 d3
1 4 3 8 5 16

P4 d1 0 1 0 1 0 0 0 0 0 0 0

0 d2
1 4 5 8 3 16

P3 d3
1 4 3 8 5 16

Quantity 20 30 15 30

P4

Zj C j Z j

5 8

3 8 3 8

5 8 5 8

3 8

5 8

3 8

P3

Zj C j Z j
Zj C j Z j
Zj C j Z j

0 0 0 1 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 1 0 0 0 0

P2

P1

The best solution is X1 15 X2 20 d1 30

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 168

168

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

11-29.

a. d1 underachievement of class and study goal d1 overachievement of class and study goal d2 overachievement of sleeping goal d3 underachievement of social time goal

subject to 100X1 130X2 5,000 hours X1, X2 0 11-32. Let X1 no. of XJ6s and X2 no. of XJ8s
2 2 5X2 .2X2 a. Maximize Z 4X1 .1X1 subject to X1 2X2 40

Major Blighs objective function becomes minimize d1 d1 d2 d3 subject to constraints (per week) 1X1 1X2 1X3 1X4 168 1X3 d1 d1 30 1X1 d2 49 1X4 d3 20 All variables 0 Since the goals have priority, they can be rewritten in this order, yielding to the absolute completion of each goal before attempting to achieve the next goal. The objective function would become minimize P1d1 P1d1 P2 d2 P3d3 where P1 meet class and study goal P2 meet sleeping goal P3 meet socializing goal b. X1 49 X2 69 X3 30 X4 20 All goals are fully met. 11-30. a. Let S dollars invested in stocks; B dollars invested in bonds;

X1, X2 0 b. X1 18.3; X2 10.8; revenue $70,420 11-33. The optimal solution found using Solver in Excel is X 62.73, Y 8.64, Prot 720.41. 11-34. The optimal solution found using Solver in Excel is X 0.333, Y 0.667, with a variance of 0.102 and a return of 0.09. 11-35. a. Total prot (P16)X1 (P28)X2 b. The optimal solution found using Solver in Excel is X1 260, X2 140, P1 20, P2 17.33, prot $4,946.67. 11-36. a. Z $665,000
Variable X1 X2 X3 X4 X5 (South Orlando) X6 X7 X8 (Apopka) X9 (Lake Mary) X10 (Cocoa Beach) Value 0 0 0 0 1 0 0 1 1 1

R dollars invested in real estate Minimize d1 d2 d3 Subject to 0.13S + 0.08B + 0.10R d1 d1 25,000 B d2 d2 = 75,000 R d3 d3 = 0.50(S + B) S B R 250,000 S 150,000 150,000 150,000 S, , 0 b. S $50,000 invested in stocks B 75,000 invested in bonds; R $125,000 invested in real estate The total return is $25,000 (10%). The amount invested in real estate is not less than half than the amount invested in stocks and bonds. This is the only goal that is not met. 11-31. Maximize prot X1(1,800 50X1) X2(2,400 70X2) 2 1,800X1 50X1 2 2,400X2 70X2 X1, X2 0 Return is at least 10% Amount in bonds is at least 30% Real estate is less than half of stocks and bonds

b. The expected return drops to $625,000. Osceola opens and Cocoa Beach closes. c. As seen below, with Apopka corrected, the new solution has a return of $635,000 but the same locations as part a. Solution: Z $635,000
Variable X1 X2 X3 X4 X5 Value 0 0 0 0 1 Variable X6 X7 X8 X9 X10 Value 0 0 1 1 1

SOLUTIONS TO INTERNET HOMEWORK PROBLEMS

11-37. Maximize return 50X1 100X2 30X3 45X4 65X5 20X6 90X7 35X8 subject to 500X1 1,000X2 350X3 490X4 700X5 270X6 800X7 400X8 3,000 X1 X2 X3 X4 X5 X6 X7 X8 5 X1 X2 X3 X4 X5 All Xi 0 or 1 1 2 X6 X7 X8 2

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 169

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

169

A2 B1 $106 C3 D4 Feasible Not Feasible

A2

A1 B2 $130 C3 D4
A1

A3 B2 $150 C1 D4
A3

Feasible Not Feasible

Feasible

B2

A4 B2 $55 C3 D4 Lower Bound = $55 Upper Bound = $105

C1

A4 B2 $100 C1 D4

A4 B1 $46 C3 D4 Lower Bound = $46

Lower Bound = $100 Upper Bound = $105

D 3

C2
D

A4 B1 $131 Not Feasible C2 D4

Upper Bound = A4 B1 $105 C3 D2 Feasible

A4 B2 $105 C3 D1 Feasible

A4 B2 $150 C1 D3 Feasible

Optimal Solutions

Branch and Bound Solution for Problem 11-38.

11-38. Lower bound set on rows with assignment A4 ($10), B1 ($6), C3 ($5), D4 ($25): Total cost $46. Two optimal solutions (see the gure above) with a total cost of $105:
Assignment A4 B1 C3 D2 Cost $ 10 6 5 $184 $105 Assignment A4 B2 C3 D1 Cost $ 10 15 5 $175 $105

11-40. The rst two priorities, P1 and P2, are fully satised by the region ABC. But the P3 priority requires that we select a solution above the exposure constraint line (minimize d4). Point A comes closest to reaching the P3 goal. The best solution is X1 10 TV spots X2 35 newspaper ads Total exposure here is 8,250,000 people, so d4 750,000 people. 3 In other words, the exposure goal was underachieved by 4 million people. Notice that in this problem d2 and d3 are of equal (P2) priority and hence are equally important. See the graph below. Figure for Problem 11-40
Ads 70 60 50 X2 40 30 20 10 0 5 10 15 X1 d 2 d 2+ TV Spots Constraint

11-39. X1 number of TV spots X2 number of newspaper ads d1 deviation above budget funds of $120,000 d2 number of TV spots below 10 d3 number of newspaper ads below 20 d4 deviation below exposure of 9 million persons desired Minimize deviations P1d1 P2d2 P2d3 P3d4 subject to 5,000X1 2,000X2 d1 $120,000 (budget constraint) X1 d2 10 (TV spots) X2 d3 20 (newspaper ads) 300,000X1 150,000X2 d4 9,000,000 (exposures) All variables 0

Budget Constraint A C B

Newspaper Ads Constraint d3 d 1+ d 1 20 d 4 25 d 4+

+

d 3

Exposure Constraint

30 TV Spots

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 170

170

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

11-41. week

Let S number of Standard blenders produced each

These constraints assume no more than one assignment per manager. For project managers: X11 X12 X13 X14 d5 d5 1 X21 X22 X23 X24 d6 d6 1 X31 X32 X33 X34 d7 d7 1

D number of Deluxe blenders produced each week C number of Chefs Delight blenders produced each week Minimize
d1

Gardener Ruth Hardgraves

d1

d 2

d 3

d 4

d 5

Subject to: (1) use 240 hours per week

1.5S 2D 2.5C d 1 d1 240

(2) produce 60 of the Chefs Delight blenders C d2 d2 60 (3) produce 60 of the Deluxe blenders D d3 d3 60 (4) produce 60 of the Standard blenders S d4 d4 60 (5) generate prot of at least $3,500 28S 32D 35C All variables 0 11-42. The constraints are the same as in Problem 11-41. The objective function changes to:
Minimize d 1 d1 0.5d2 0.5d3 0.5d4 0.333d5

These constraints permit assigning three managers to four clients while minimizing positive and negative deviational variables (d5, d6, d7). Gardener to CBT restriction: X14 d8 0 This constraint looks at the deviation of d8 from 0. In other words, the closer d8 is to 0 (not assigning Gardener to CBT), the closer it comes to meeting the restriction. Ruth earns $3,000 or more: 2,700X21 3,200X22 3,000X23

d 5

d5

3,500

3,100X24 d9 d9 $3,000 Here d9 represents underachievement of the goal, while d9 is overachievement. The coefcients are the costs per assignment. Total costs:

(C ij X ij ) d10+ = 0
j =1 i =1

SOLUTION TO SCHANK MARKETING RESEARCH CASE

1. The rst part of this case is an assignment problem that can be formulated with LP. A dummy project manager can be added to create a balanced 4 4 cost matrix. Minimize

This attempts to minimize total cost, bringing it as close to zero as possible; d10 is the deviation from the goal. Objective function: minimize Z P1d2 P2d8 P3(2d1 d3) P4d9 P5d10

Cij Xij
j =1 i=1

where

Xij

1 0

if project leader i is assigned to client j if otherwise

SOLUTION TO THE OAKTON RIVER BRIDGE CASE

For a given set of requirements, the smallest number of toll collectors that will meet them can be obtained from the following integer linear programming problem: minimize Z X1 X2 X3 X4 X5 X6 X7 subject to X1 X2 X3 X4 X5 X6 X7 R5 X1 X2 X3 X4 X5 X2 X3 X4 X5 X6 X1 X1 X2 X1 X2 X3 X1 X2 X3 X4 All variables 0 where Xj is the number of collectors starting on day j ( j 1 is Sunday) and Rj is the number required on day j. 1. The following table summarizes the requirements for shifts A, B, and C for each of the three days of the week along with the allocations that yield the minimum numbers of collectors starting each: 18 for shift A, 16 for shift B, and 18 for shift C. R5 R6

i 1, 2, 3, 4 for Gardener, Ruth, Hardgraves, Dummy j 1, 2, 3, 4 for Hines, NASA, General, CBT 2. This part is a goal programming formulation with ve goals, ranked from P1 (highest) to P5 (lowest): P1: assign a manager to the NASA account. P2: do not assign Gardener to CBT Television account. P3: meet demands of Hines; they are twice as important as those of General Foundry. P4: place Ruth on a project that will earn him $3,000 or more. P5: minimize the total cost of all assignments. Constraints For clients demand: X11 X21 X31 d1 1 X12 X22 X32 d2 1 X13 X23 X33 d3 1 X14 X24 X34 d4 1 Hines NASA General CBT

X3 X4 X5 X6 X7 R7 X3 X4 X5 X6 X7 R1 X5 X6 X7 R2 X6 X7 R3 X7 R4

REVISED
M11_REND6289_10_IM_C11.QXD 5/12/08 12:04 PM Page 171

CHAPTER 11

INTEGER PROGRAMMING, GOAL PROGRAMMING, NONLINEAR PROGRAMMING

171

Toll Collector Requirements for Oakton River Case

SHIFT DAY Sun. Mon. Tue. Wed. Thu. Fri. Sat. Total A Req. Start 8 13 12 12 13 13 15 0 3 5 0 5 1 14 18 B Req. Start 10 10 10 10 10 13 15 0 1 5 1 5 1 13 16 C Req. Start 15 13 13 12 12 13 8 5 2 1 4 1 5 10 18 Mix Req. Start 33 36 35 34 35 39 38 3 9 8 6 9 7 18 50

the annual rent constraint 4.4X 1 6.1X2 8.3X3 24.0X4 19.5X5 20.7X6 7.7X7 19.4X8 11.7X9 15.2X10 3.9X11 3.2X12 11.3X13 16.0X14 9.6X15 130 the construction cost constraint 24.6 X1 32.0X2 41.4X3 124.4X4 64.8X5 79.8X6 38.6X7 66.8X8 45.1X9 54.3X10 15.0X11 13.4X12 42.0X13 63.7X14 40.0X15 700 at least one clothing store X1 X2 X3 1 at least one hard goods store X8 X9 X10 1 at least one miscellaneous-type store X11 X12 X13 X14 X15 1 at least two restaurants X4 X5 X6 X7 2 no more than two clothing stores X1 X2 X3 2 miscellaneous types cannot exceed total of clothing and hard goods X1 X2 X3 X8 X9 X10 X11 X12 X13 X14 X15 0 The optimum solution is to include stores 1, 4, 5, 6, 8, 10, 12, 14, and 15. The present value is $647,400, all 16,000 square feet of space will be used, the annual rent is $132,000, and the construction cost is $531,800.

Note: Alternative optimal solutions for each shift may be possible. 2. If mixing of shifts is allowed, the daily requirements become the sum of the shift requirements, as shown in the second part of the table. The minimum number of collectors starting each day is shown in the last day. The total 50 is a reduction of two from the total required without allowing for the mixing of shifts.

SOLUTION TO PUYALLUP MALL CASE

The problem can be expressed as the following integer linear programming problem with Xi being a 01 variable, 1 if store i is to be included and 0 if not: Maximize 28.1X1 34.6X2 50.0X3 162.0X4 77.8X5 100.4X6 45.2X7 80.2X8 51.4X9 62.5X10 18.0X11 11.6X12 50.4X13 73.6X14 51.2X15 subject to the space constraint 1.0X1 1.6X2 2.0X3 3.2X4 1.8X5 2.1X6 1.2X7 2.4X8 1.6X9 2.0X10 0.6X11 0.5X12 1.4X13 2.0X14 1.0X15 16