Session 4a. Decision Models -- Prof. Juran2 Overview More Network Models –Assignment Model...
-
Upload
jayla-colt -
Category
Documents
-
view
223 -
download
3
Transcript of Session 4a. Decision Models -- Prof. Juran2 Overview More Network Models –Assignment Model...
Session 4a
Decision Models -- Prof. Juran
2
OverviewMore Network Models
– Assignment Model (Contract Bidding)• “Big Cost” trick
– Project Management (House Building)• More binary / integer tricks• Critical Path / Slack Time• Excel trick: Conditional Formatting• Cost Crashing
– Changing an objective to a constraint– Issues with Integers– Location Analysis (Hospital Location)
Decision Models -- Prof. Juran
3
Contract Bidding ExampleA company is taking bids on four construction jobs. Three contractors have placed bids on the jobs. Their bids (in thousands of dollars) are given in the table below. (A dash indicates that the contractor did not bid on the given job.) Contractor 1 can do only one job, but contractors 2 and 3 can each do up to two jobs.
Job 1 Job 2 Job 3 Job 4 Contractor 1 50 46 42 40 Contractor 2 51 48 44 — Contractor 3 — 47 45 45
Decision Models -- Prof. Juran
4
Formulation
Decision VariablesWhich contractor gets which job(s).ObjectiveMinimize the total cost of the four jobs.ConstraintsContractor 1 can do no more than one job.Contractors 2 and 3 can do no more than two jobs each.Contractor 2 can’t do job 4.Contractor 3 can’t do job 1.Every job needs one contractor.
Decision Models -- Prof. Juran
5
Formulation Decision VariablesDefine Xij to be a binary variable representing the assignment of contractor i to job j. If contractor i ends up doing job j, then Xij = 1. If contractor i does not end up with job j, then Xij = 0.
Define Cij to be the cost; i.e. the amount bid by contractor i for job j.
ObjectiveMinimize Z =
3
1
4
1i jijijCX
Decision Models -- Prof. Juran
6
Formulation Constraints for all j.
for i = 1.
for i = 2, 3.
13
1
i
ijX
14
1
j
ijX
24
1
j
ijX
01,34,2 XX
Decision Models -- Prof. Juran
7
Solution Methodology1234567
89
10
11121314
15
A B C D E F G HJob 1 Job 2 Job 3 Job 4
Contractor 1 50 46 42 40Contractor 2 51 48 44 10000Contractor 3 10000 47 45 45
Assignment of contractors to jobsJob 1 Job 2 Job 3 Job 4 Total Max
Contractor 1 0 0 0 0 0 <= 1Contractor 2 0 0 0 0 0 <= 2Contractor 3 0 0 0 0 0 <= 2
Total 0 0 0 0= = = =
Required 1 1 1 1
Total cost ($1000s) 0=SUMPRODUCT(B2:E4,B8:E10)
=SUM(E8:E10)
=SUM(B8:E8)
Decision Models -- Prof. Juran
8
Solution Methodology
Notice the very large values in cells B4 and E3. These specific values (10,000) aren’t important; the main thing is to assign these particular contractor-job combinations costs so large that they will never be in any optimal solution.
Decision Models -- Prof. Juran
9
Solution Methodology
Decision Models -- Prof. Juran
10
Optimal Solution
1234567
8910
11121314
15
A B C D E F G HJob 1 Job 2 Job 3 Job 4
Contractor 1 50 46 42 40Contractor 2 51 48 44 10000Contractor 3 10000 47 45 45
Assignment of contractors to jobsJob 1 Job 2 Job 3 Job 4 Total Max
Contractor 1 0 0 0 1 1 <= 1Contractor 2 1 0 1 0 2 <= 2Contractor 3 0 1 0 0 1 <= 2
Total 1 1 1 1= = = =
Required 1 1 1 1
Total cost ($1000s) 182
Decision Models -- Prof. Juran
11
Conclusions
The optimal solution is to award Job 4 to Contractor 1, Jobs 1 and 3 to Contractor 2, and Job 2 to Contractor 3. The total cost is $182,000.
Decision Models -- Prof. Juran
12
Sensitivity Analysis
1. What is the “cost” of restricting Contractor 1 to only one job?
2. How much more can Contractor 1 bid for Job 4 and still get the job?
Decision Models -- Prof. Juran
13
123456
78910111213141516171819202122
23242526272829
A B C D E F G HMicrosoft Excel 15.0 Sensitivity Report
Variable CellsFinal Reduced Objective Allowable Allowable
Cell Name Value Cost Coefficient Increase Decrease
$B$8 Contractor 1 Job 1 0 1 50 1E+30 1$C$8 Contractor 1 Job 2 0 1 46 1E+30 1$D$8 Contractor 1 Job 3 0 0 42 1 3$E$8 Contractor 1 Job 4 1 0 40 3 1E+30$B$9 Contractor 2 Job 1 1 0 51 1 1E+30$C$9 Contractor 2 Job 2 0 1 48 1E+30 1$D$9 Contractor 2 Job 3 1 0 44 1 1$E$9 Contractor 2 Job 4 0 9958 10000 1E+30 9958$B$10 Contractor 3 Job 1 0 9949 10000 1E+30 9949$C$10 Contractor 3 Job 2 1 0 47 1 1E+30$D$10 Contractor 3 Job 3 0 1 45 1E+30 1$E$10 Contractor 3 Job 4 0 3 45 1E+30 3
ConstraintsFinal Shadow Constraint Allowable Allowable
Cell Name Value Price R.H. Side Increase Decrease
$B$11 Total Job 1 1 51 1 0 1$C$11 Total Job 2 1 47 1 1 1$D$11 Total Job 3 1 44 1 0 1$E$11 Total Job 4 1 42 1 0 1$F$8 Contractor 1 Total 1 -2 1 1 0$F$9 Contractor 2 Total 2 0 2 1E+30 0$F$10 Contractor 3 Total 1 0 2 1E+30 1
Decision Models -- Prof. Juran
14
Conclusions
The sensitivity report indicates a shadow price of –2 (cell E29).
(Allowing Contractor 1 to perform one additional job would reduce the total cost by 2,000.)
The allowable increase in the bid for Job 4 by Contractor 1 is 3. (He could have bid any amount up to $43,000 and still have won that job.)
Decision Models -- Prof. Juran
15
Con. 3Con. 2Con. 1
Network Representation
Job 3Job 2 Job 4Job 1
Decision Models -- Prof. Juran
16
Optimal Solution
Con. 1
Job 3
Con. 2 Con. 3
Job 2 Job 4Job 1
4047
44
51
Decision Models -- Prof. Juran
17
House-Building Example
Activity Description Predecessors Duration (days) Activity A Build foundation — 5 Activity B Build walls and ceilings A 8 Activity C Build roof B 10 Activity D Do electrical wiring B 5 Activity E Put in windows B 4 Activity F Put on siding E 6 Activity G Paint house C, F 3
Decision Models -- Prof. Juran
18
Managerial Problem Definition
Find the critical path and the minimum number of days needed to build the house.
Decision Models -- Prof. Juran
19
Network Representation1
0Start
43
2
A5
B8
C10
E4
D5
G3
F6
5End
Decision Models -- Prof. Juran
20
Formulation
Decision VariablesWe are trying to decide when to begin and end each of the activities.ObjectiveMinimize the total time to complete the project.ConstraintsEach activity has a fixed duration.There are precedence relationships among the activities.We cannot go backwards in time.
Decision Models -- Prof. Juran
21
FormulationDecision VariablesDefine the nodes to be discrete events. In other words, they occur at one exact point in time. Our decision variables will be these points in time.Define ti to be the time at which node i occurs, and at which time all activities preceding node i have been completed.Define t0 to be zero.
ObjectiveMinimize t5.
Decision Models -- Prof. Juran
22
FormulationConstraintsThere is really one basic type of constraint. For each activity x, let the time of its starting node be represented by tjx and the time of its ending node be
represented by tkx. Let the duration of activity x be
represented as dx.
For every activity x,
For every node i,
xjxkx dtt
0it
Decision Models -- Prof. Juran
23
Solution Methodology
123
456
789101112131415161718
A B C D E F G H I J K L M N O P
1
t0 t1 t2 t3 t4 t50 1 1 1 1 1
t0 t1 t2 t3 t4 t5A -1 1 0 0 0 0 1 >= 5B 0 -1 1 0 0 0 0 >= 8C 0 0 -1 0 1 0 0 >= 10D 0 0 -1 0 0 1 0 >= 5E 0 0 -1 1 0 0 0 >= 4F 0 0 0 -1 1 0 0 >= 6G 0 0 0 0 -1 1 0 >= 3
1
0Start
43
2
A5
B8
C10
E4
D5
G3
F6
5End
Decision Models -- Prof. Juran
24
Solution Methodology
The matrix of zeros, ones, and negative ones (B12:G18) is a means for setting up the constraints.
The sumproduct functions in H12:H18 calculate the elapsed time between relevant pairs of nodes, corresponding to the various activities.
The duration times of the activities are in J12:J18.
Decision Models -- Prof. Juran
25
Decision Models -- Prof. Juran
26
Optimal Solution123
456
789101112131415161718
A B C D E F G H I J K L M N O P
26
t0 t1 t2 t3 t4 t50 5 13 17 23 26
t0 t1 t2 t3 t4 t5A -1 1 0 0 0 0 5 >= 5B 0 -1 1 0 0 0 8 >= 8C 0 0 -1 0 1 0 10 >= 10D 0 0 -1 0 0 1 13 >= 5E 0 0 -1 1 0 0 4 >= 4F 0 0 0 -1 1 0 6 >= 6G 0 0 0 0 -1 1 3 >= 3
1
0Start
43
2
A5
B8
C10
E4
D5
G3
F6
5End
Decision Models -- Prof. Juran
27
Conclusions
The project will take 26 days to complete.
The only activity that is not critical is the electrical wiring.
Decision Models -- Prof. Juran
28
CPM Jargon
Any activity for which
is said to have slack time, the amount of time by which that activity could be delayed without affecting the overall completion time of the whole project. In this example, only activity D has any slack time (13 – 5 = 8 units of slack time).
xjxkx dtt
Decision Models -- Prof. Juran
29
CPM Jargon
Any activity x for which
is defined to be a “critical” activity, with zero slack time.
xjxkx dtt
Decision Models -- Prof. Juran
30
CPM JargonEvery network of this type has at least one critical path, consisting of a set of critical activities. In this example, there are two critical paths: A-B-C-G and A-B-E-F-G.
1
0Start
43
2
A5
B8
C10
E4
D5
G3
F6
5End
Decision Models -- Prof. Juran
31
Excel Tricks: Conditional Formatting1
2
3
45
6
789101112131415161718
A B C D E F G H I J K L M N O P
26
t0 t1 t2 t3 t4 t5
0 5 13 17 23 26
t0 t1 t2 t3 t4 t5A -1 1 0 0 0 0 5 >= 5B 0 -1 1 0 0 0 8 >= 8C 0 0 -1 0 1 0 10 >= 10D 0 0 -1 0 0 1 13 >= 5E 0 0 -1 1 0 0 4 >= 4F 0 0 0 -1 1 0 6 >= 6G 0 0 0 0 -1 1 3 >= 3
1
0Start
43
2
A5
B8
C10
E4
D5
G3
F6
5End
Decision Models -- Prof. Juran
32
Critical Activities: Using the Solver Answer Report
ConstraintsCell Name Cell Value Formula Status Slack
$H$12 A 5 $H$12>=$J$12 Binding 0$H$13 B 8 $H$13>=$J$13 Binding 0$H$14 C 10 $H$14>=$J$14 Binding 0$H$15 D 13 $H$15>=$J$15 Not Binding 8$H$16 E 4 $H$16>=$J$16 Binding 0$H$17 F 6 $H$17>=$J$17 Binding 0$H$18 G 3 $H$18>=$J$18 Binding 0
Decision Models -- Prof. Juran
33
House-Building Example, Continued
Suppose that by hiring additional workers, the duration of each activity can be reduced. Use LP to find the strategy that minimizes the cost of completing the project within 20 days.
Decision Models -- Prof. Juran
34
Crashing Parameters
Activity Cost per Day to
Reduce Duration Maximum Possible Reduction (Days)
Build foundation $30 2 Walls and ceilings $15 3 Build roof $20 1 Electrical wiring $40 2 Put in windows $20 2 Put on siding $30 3 Paint house $40 1
Decision Models -- Prof. Juran
35
Managerial Problem Definition
Find a way to build the house in 20 days.
Decision Models -- Prof. Juran
36
Formulation
Decision VariablesNow the problem is not only when to schedule the activities, but also which activities to accelerate. (In CPM jargon, accelerating an activity at an additional cost is called “crashing”.)
ObjectiveMinimize the total cost of crashing.
Decision Models -- Prof. Juran
37
FormulationConstraintsThe project must be finished in 20 days.Each activity has a maximum amount of crash time.Each activity has a “basic” duration. (These durations were considered to have been fixed in Part a; now they can be reduced.)There are precedence relationships among the activities.We cannot go backwards in time.
Decision Models -- Prof. Juran
38
FormulationDecision VariablesDefine the number of days that activity x is crashed to be Rx.
For each activity there is a maximum number of crash days Rmax, x
Define the crash cost per day for activity x to be Cx
ObjectiveMinimize Z =
7
1xxxCR
Decision Models -- Prof. Juran
39
FormulationConstraints
For every activity x,
For every activity x,
For every node i,
xxjxkx Rdtt
xx RR max,
0it
Decision Models -- Prof. Juran
40
Solution Methodology12
3
45
6
7891011
12131415161718
A B C D E F G H I J K L M N O
30
t0 t1 t2 t3 t4 t5
0 1 1 1 1 1
20 <-- Max Completion Time
t0 t1 t2 t3 t4 t5 Improved Duration Basic Duration Crash Time Max Crash Cost/Time
A -1 1 0 0 0 0 1 >= 4 5 1 2 30$ B 0 -1 1 0 0 0 0 >= 8 8 0 3 15$ C 0 0 -1 0 1 0 0 >= 10 10 0 1 20$ D 0 0 -1 0 0 1 0 >= 5 5 0 2 40$ E 0 0 -1 1 0 0 0 >= 4 4 0 2 20$ F 0 0 0 -1 1 0 0 >= 6 6 0 3 30$ G 0 0 0 0 -1 1 0 >= 3 3 0 1 40$
1
0Start
43
2
A5
B8
C10
E4
D5
G3
F6
5End
Decision Models -- Prof. Juran
41
Solution Methodology
G3 now contains a formula to calculate the total crash cost. The new decision variables (how long to crash each activity x, represented by Rx) are in M12:M18.
G8 contains the required completion time, and we will constrain the value in G6 to be less than or equal to G8.
The range J12:J18 calculates the revised duration of each activity, taking into account how much time is saved by crashing.
Decision Models -- Prof. Juran
42
Solution Methodology
Decision Models -- Prof. Juran
43
Optimal Solution12
3
45
6
7891011
12131415161718
A B C D E F G H I J K L M N O
145
t0 t1 t2 t3 t4 t5
0 3 8 11 17 20
20 <-- Max Completion Time
t0 t1 t2 t3 t4 t5 Improved Duration Basic Duration Crash Time Max Crash Cost/Time
A -1 1 0 0 0 0 3 >= 3 5 2 2 30$ B 0 -1 1 0 0 0 5 >= 5 8 3 3 15$ C 0 0 -1 0 1 0 9 >= 9 10 1 1 20$ D 0 0 -1 0 0 1 12 >= 5 5 0 2 40$ E 0 0 -1 1 0 0 3 >= 3 4 1 2 20$ F 0 0 0 -1 1 0 6 >= 6 6 0 3 30$ G 0 0 0 0 -1 1 3 >= 3 3 0 1 40$
1
0Start
43
2
A5
B8
C10
E4
D5
G3
F6
5End
Decision Models -- Prof. Juran
44
ConclusionsIt is feasible to complete the project in 20 days, at a cost of $145.
1
0Start
43
2
A3
B5
C9
E3
D5
G3
F6
5End
Decision Models -- Prof. Juran
45
Conclusions
Basic Duration Crash Time Improved Duration Cost/Time CostA 5 2 3 30$ 60$ B 8 3 5 15$ 45$ C 10 1 9 20$ 20$ D 5 0 5 40$ -$ E 4 1 3 20$ 20$ F 6 0 6 30$ -$ G 3 0 3 40$ -$
145$
Decision Models -- Prof. Juran
46
An Alternative Solution
Basic Duration Crash Time Improved Duration Cost/Time CostA 5 2 3 30$ 60$ B 8 3 5 15$ 45$ C 10 0 10 20$ -$ D 5 0 5 40$ -$ E 4 0 4 20$ -$ F 6 0 6 30$ -$ G 3 1 2 40$ 40$
145$
Decision Models -- Prof. Juran
47
Excel Tricks: VLOOKUP
Looks for a specific value in the left column of a table and finds the row where that value appears, then returns the value corresponding to another specified column in that row.
Decision Models -- Prof. Juran
48
123456789
10111213141516171819202122232425262728293031
A B C D E F G H I JSchool CostNYU 31,746$
School Accept % Enroll % GMAT GPA Cost Minority pct Non-U.S. pct Female % SalaryStanford 8.8% 78.4% 718 3.58 31,002$ 24.6% 30.9% 38.7% 124,740$ Harvard 11.6% 87.3% 703 3.58 30,050$ 18.6% 37.2% 34.2% 121,979$ Penn 15.5% 70.2% 703 3.57 31,218$ 19.0% 43.1% 29.3% 121,200$ MIT 18.1% 69.9% 703 3.50 31,200$ 19.0% 35.5% 26.5% 118,381$ Northwestern 16.6% 61.8% 700 3.45 30,255$ 17.5% 30.8% 27.5% 114,664$ Duke 20.1% 52.8% 690 3.43 30,323$ 20.6% 32.7% 34.8% 115,444$ Chicago 28.4% 59.7% 695 3.33 30,596$ 18.2% 33.2% 23.6% 115,331$ Columbia 13.5% 68.7% 705 3.40 31,912$ 19.3% 26.5% 36.5% 117,989$ Dartmouth 18.8% 53.3% 690 3.40 30,250$ 17.2% 29.9% 27.6% 121,692$ Berkeley 14.5% 50.3% 688 3.43 21,242$ 19.4% 30.2% 30.4% 111,321$ Michigan 19.6% 58.5% 676 3.34 30,686$ 20.5% 31.2% 27.6% 117,498$ Virginia 16.6% 51.0% 681 3.40 27,283$ 15.1% 30.1% 29.1% 112,706$ NYU 22.3% 48.2% 689 3.40 31,746$ 19.0% 30.2% 34.1% 112,900$ Yale 20.0% 56.7% 686 3.50 29,055$ 17.3% 38.9% 27.5% 112,514$ UCLA 17.4% 48.5% 691 3.60 22,490$ 21.6% 24.0% 28.6% 103,364$ Cornell 25.8% 51.2% 669 3.32 30,455$ 22.6% 25.6% 27.1% 116,588$ North Carolina 23.8% 44.2% 674 3.30 25,525$ 16.2% 30.4% 30.6% 109,420$ Carnegie-Mellon 28.2% 60.1% 652 3.25 28,452$ 21.3% 38.3% 26.8% 111,211$ Texas 30.0% 52.4% 680 3.34 21,200$ 12.3% 26.7% 22.7% 106,859$ USC 29.0% 43.4% 670 3.30 30,082$ 32.3% 23.8% 28.7% 95,213$ Indiana 24.7% 49.8% 651 3.35 20,696$ 14.3% 34.5% 20.7% 99,732$ Emory 31.5% 46.7% 651 3.30 28,012$ 10.7% 29.9% 23.4% 104,417$ Rochester 32.7% 42.2% 637 3.33 28,620$ 10.3% 54.2% 25.2% 103,466$ Georgetown 20.8% 46.2% 662 3.35 28,440$ 10.8% 38.7% 28.1% 99,528$ Michigan State 22.3% 55.0% 641 3.36 16,955$ 13.6% 37.9% 25.2% 88,746$ Ohio State 29.0% 41.4% 645 3.38 22,151$ 15.5% 29.2% 36.7% 99,598$ Minnesota 34.4% 40.6% 645 3.33 20,352$ 8.6% 29.1% 27.9% 91,140$
=VLOOKUP(A2,A5:J31,6,0)
Decision Models -- Prof. Juran
49
Integer Programs
• We have seen models where the decision variables had to be integers, but we didn’t have to impose a Solver constraint to make that happen (a special attribute of some transportation models)
• In general, you do need to impose a Solver constraint to force an integer solution
• No more Simplex algorithm!• Binary = Special case of integer
Decision Models -- Prof. Juran
50
Difficulties with Integer Programs
• Linear approximations are not necessarily feasible or optimal
• Integer optimal solutions require much more complicated algorithms
• Integer algorithms do not yield a sensitivity report
Decision Models -- Prof. Juran
51
Maximize z = YX 1121
Subject to: YX 47 13
X, Y 0 Linear Program
0
1
2
3
4
0 1 2 3 4
Optimal SolutionX1=1.857, X2=0
Decision Models -- Prof. Juran
52
Optimal linear-programming solution: X = 1.857, Y = 0Rounded to X = 2, Y = 0 is infeasibleRounded to X = 1, Y = 0 is not optimalOptimal integer-programming solution: X = 0, Y = 3
Integer Program
0
1
2
3
4
0 1 2 3 4
Optimal SolutionX1=0, X2=3
Decision Models -- Prof. Juran
53
Hospital Location Example
A county is going to build two hospitals.
There are nine cities in which the hospitals can be built.
The number of hospital visits per year made by people in each city and the x-y coordinates of each city are listed in the table below.
The county's goal is to minimize the total distance that patients must travel to hospitals. Where should it locate the hospitals?
Decision Models -- Prof. Juran
54
Hospital Location Example
City x y Hospital visits per Year 1. Kimsbury 0 0 3000 2. Jozefacki Junction 10 3 4000 3. Sapras Falls 12 15 5000 4. Lowthersburg 14 13 6000 5. Leesville 16 9 4000 6. Sanjay Beach 18 6 3000 7. Patel Point 8 12 2000 8. Rothsboro 6 10 4000 9. Nikolova City 4 8 1200
Decision Models -- Prof. Juran
55
Hospital Location ExampleCounty Map
0
2
4
6
8
10
12
14
16
0 2 4 6 8 10 12 14 16 18 20
X
Y
1. Kimsbury 2. Jozefacki Junction
9. Nikolova City
8. Rothsboro7. Patel Point
3. Sapras Falls
6. Sanjay Beach
5. Leesville
4. Lowthersburg
Decision Models -- Prof. Juran
56
Managerial Problem Definition
Decision Variables
We need to decide whether or not to build a hospital in each of nine cities.
We also need to decide how many visits there will be from each city to each hospital.
Objective
We want to minimize the total distance traveled to the hospital by all county residents.
Decision Models -- Prof. Juran
57
Managerial Problem Definition
Constraints
The cities can’t be moved.
Exactly two hospitals will be built.
All of the planned hospital visits must be accounted for and included in the total distance calculation.
No hospital visits are allowed to a city that has no hospital.
Decision Models -- Prof. Juran
58
Network Representation
9
8
7
6
5
4
3
2
1
Decision Models -- Prof. Juran
59
Network Representation
9
8
7
6
5
4
3
2
1 One Hospital
Decision Models -- Prof. Juran
60
FormulationDecision Variables
Define Vij to be the number of visits from people in city i to the hospital in city j.
Define Xj to be a binary variable. If a hospital is built in city j, then Xj = 1; otherwise, Xj = 0.
These Vij and Xj are the decision variables.
There are 81 + 9 = 90 decision variables here.
Objective
Define Dij to be the distance from city i to city j.
Minimize Z = iji j
ijDV
9
1
9
1
Decision Models -- Prof. Juran
61
Formulation
Decision Models -- Prof. Juran
62
Solution Methodology123
456
789
101112131415
16171819202122232425262728
A B C D E F G H I J K L M N O P Q R S T U V WXj
1 2 3 4 5 6 7 8 9 Sum Required Total miles (x 1000)1 1 1 1 1 1 1 1 1 9 = 2 0.00
Vij to City coordinates1 2 3 4 5 6 7 8 9 Sum Ri x y
1 3.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3 = 3 1 0 02 0.0 4.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 4 = 4 2 10 33 0.0 0.0 5.0 0.0 0.0 0.0 0.0 0.0 0.0 5 = 5 3 12 154 0.0 0.0 0.0 6.0 0.0 0.0 0.0 0.0 0.0 6 = 6 4 14 13
from 5 0.0 0.0 0.0 0.0 4.0 0.0 0.0 0.0 0.0 4 = 4 5 16 96 0.0 0.0 0.0 0.0 0.0 3.0 0.0 0.0 0.0 3 = 3 6 18 67 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0 0.0 2 = 2 7 8 128 0.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0 4 = 4 8 6 109 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.2 1.2 = 1.2 9 4 8
3.0 4.0 5.0 6.0 4.0 3.0 2.0 4.0 1.2Logical upper bounds <= <= <= <= <= <= <= <= <= Distances 1 2 3 4 5 6 7 8 9
32.2 32.2 32.2 32.2 32.2 32.2 32.2 32.2 32.2 1 0.0 10.4 19.2 19.1 18.4 19.0 14.4 11.7 8.92 10.4 0.0 12.2 10.8 8.5 8.5 9.2 8.1 7.83 19.2 12.2 0.0 2.8 7.2 10.8 5.0 7.8 10.64 19.1 10.8 2.8 0.0 4.5 8.1 6.1 8.5 11.25 18.4 8.5 7.2 4.5 0.0 3.6 8.5 10.0 12.06 19.0 8.5 10.8 8.1 3.6 0.0 11.7 12.6 14.17 14.4 9.2 5.0 6.1 8.5 11.7 0.0 2.8 5.78 11.7 8.1 7.8 8.5 10.0 12.6 2.8 0.0 2.89 8.9 7.8 10.6 11.2 12.0 14.1 5.7 2.8 0.0
=SUM($N$7:$N$15)*C3
=SUMPRODUCT(C7:K15,N18:V26)
=SUM(C10:K10)
=SUM(K7:K15)
=SQRT((VLOOKUP($M22,$P$6:$R$15,2)-VLOOKUP(O$17,$P$6:$R$15,2))^2+(VLOOKUP($M22,$P$6:$R$15,3)-VLOOKUP(O$17,$P$6:$R$15,3))^2)
Decision Models -- Prof. Juran
63
Solution Methodology
The 9 Xj decision variables are in the range C3:K3.
The 81 Vij decision variables are in the range C7:K15.
The objective function is in cell P3.
The matrix in the range N18:V26 calculates the distance between each pair of cities using a long and ugly Excel function based on the famous Pythagorean theorem.
Cell N3 is used to keep track of constraint (1).
Cells L7:L15 and N7:N15 are used to keep track of constraint (2).
Cells C16:K16 and C18:K18 are used to keep track of constraint (3).
Decision Models -- Prof. Juran
64
Decision Models -- Prof. Juran
65
Optimal Solution123
456
789
101112131415
1617181920212223242526
A B C D E F G H I J K L M N O P Q RXj
1 2 3 4 5 6 7 8 9 Sum Required Total miles (x 1000)0 0 0 1 0 0 0 1 0 2 = 2 132.50
Vij to City coordinates1 2 3 4 5 6 7 8 9 Sum Ri x y
1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3.0 0.0 3 = 3 1 0 02 0.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0 4 = 4 2 10 33 0.0 0.0 0.0 5.0 0.0 0.0 0.0 0.0 0.0 5 = 5 3 12 154 0.0 0.0 0.0 6.0 0.0 0.0 0.0 0.0 0.0 6 = 6 4 14 13
from 5 0.0 0.0 0.0 4.0 0.0 0.0 0.0 0.0 0.0 4 = 4 5 16 96 0.0 0.0 0.0 3.0 0.0 0.0 0.0 0.0 0.0 3 = 3 6 18 67 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0 2 = 2 7 8 128 0.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0 4 = 4 8 6 109 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.2 0.0 1.2 = 1.2 9 4 8
0.0 0.0 0.0 18.0 0.0 0.0 0.0 14.2 0.0Logical upper bounds <= <= <= <= <= <= <= <= <= Distances 1 2 3 4 5
0.0 0.0 0.0 32.2 0.0 0.0 0.0 32.2 0.0 1 0.0 10.4 19.2 19.1 18.42 10.4 0.0 12.2 10.8 8.53 19.2 12.2 0.0 2.8 7.24 19.1 10.8 2.8 0.0 4.55 18.4 8.5 7.2 4.5 0.06 19.0 8.5 10.8 8.1 3.67 14.4 9.2 5.0 6.1 8.58 11.7 8.1 7.8 8.5 10.09 8.9 7.8 10.6 11.2 12.0
Decision Models -- Prof. Juran
66
Optimal Solution
If the county wants to build two hospitals, then the optimal locations are in City 4 (Lowthersburg) and City 8 (Rothsboro). The total miles traveled would be 132,500.
Decision Models -- Prof. Juran
67
Network Representation
9
8
7
6
5
4
3
2
1 Two Hospitals
Decision Models -- Prof. Juran
68
Network Representation
9
8
7
6
5
4
3
2
1 Three Hospitals
Decision Models -- Prof. Juran
69
Network Representation
9
8
7
6
5
4
3
2
1 Four Hospitals
Decision Models -- Prof. Juran
70
Network Representation
9
8
7
6
5
4
3
2
1 Five Hospitals
Decision Models -- Prof. Juran
71
Network Representation
9
8
7
6
5
4
3
2
1 Six Hospitals
Decision Models -- Prof. Juran
72
Network Representation
9
8
7
6
5
4
3
2
1 Seven Hospitals
Decision Models -- Prof. Juran
73
Network Representation
9
8
7
6
5
4
3
2
1 Eight Hospitals
Decision Models -- Prof. Juran
74
Network Representation
9
8
7
6
5
4
3
2
1 Nine Hospitals
Decision Models -- Prof. Juran
75
SummaryMore Network Models
– Assignment Model (Contract Bidding)• “Big Cost” trick
– Project Management (House Building)• More binary / integer tricks• Critical Path / Slack Time• Excel trick: Conditional Formatting• Cost Crashing
– Changing an objective to a constraint– Issues with Integers– Location Analysis (Hospital Location)