Network Models

Assignment Transportation Intro to Modeling/Excel How the Solver Works Sensitivity Analysis
15.057 Spring 03 Vande Vate 1

Objective
Mini Course on Networks
Introduction to modeling
In Excel and AMPL

Intuitive description of solution approach Intuitive description of sensitivity analysis

Intuitive and visual context for covering technical aspects

15.057 Spring 03 Vande Vate 2

Assignment Model
Autopower Europe
Manufactures UPS for major installations Four manufacturing plants
Leipzig, Germany Nancy, France Liege, Belgium Tilburg, The Netherlands

One VP to audit each plant

15.057 Spring 03 Vande Vate 3

Autopower, Europe
Netherlands

Tilburg Liege
*

Germany
*

Belgium

Leipzig

France
Miles
0 50 100

Nancy

15.057 Spring 03 Vande Vate

Assignment Problem
Whos to visit whom?
VPs expertise and plants needs Available time and travel requirements Language abilities ...

15.057 Spring 03 Vande Vate

The Challenge
Estimate costs (Done - Thoughts?) One VP to each plant One plant for each VP Minimize cost of assignments

15.057 Spring 03 Vande Vate

Autoppow e r Europe: Assignment Mode l Moore et al. pp224

01AssignmentModel.xls
Estimated Assignment Costs Leipzig Nancy Liege Tilburg 1 2 3 4 24 10 21 11 14 22 10 15 15 17 20 19 11 19 14 13 Leipzig 1 Assignme nts Nancy Liege 2 3 Tilburg 4

VP Finance (F) Marketing (M) Operations (O) Personnel (P)

VP Finance (F) Marketing (M) Operations (O) Personnel (P) VPs Assigned

Plants Assigned 0 0 0 0 0

0 Leipzig 1 0 0 0 0 0

0 Tilburg 4

VP Finance (F) Marketing (M) Operations (O) Personnel (P) Total Cost

Cost of Assignments Nancy Liege 2 3 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0

Total Cost 0 0 0 0 0

15.057 Spring 03 Vande Vate

A Challenge
Find best assignments

15.057 Spring 03 Vande Vate

Building a Network Model

In Excel Tools | Solver...

Set Target Cell: $F$28 By Changing Cells: $B$15:$E$18

15.057 Spring 03 Vande Vate 9

The Constraints
Each VP assigned to one plant
$F$15 $F$16 $F$17 $F$18 = = = = 1 1 1 1

Shortcut $F$15:$F$18 = 1

Each plant assigned one VP

$B$19 = 1 $C$19 = 1 $D$19 = 1 $E$19 = 1

Shortcut $B$19:$E$19 = 1
15.057 Spring 03 Vande Vate 10

Whats Missing

15.057 Spring 03 Vande Vate

11

Additional Constraints...
Non-negativity
The variables cannot be negative Handled separately

Integrality
The variables should have integral values We can ignore these because this is a network model!!!

15.057 Spring 03 Vande Vate

12

Model Components
Set Target Cell: Objective $F$28
The value we want to minimize/maximize

Equal to: Min

Min for Minimize or Max for Maximize

By Changing Cells: Variables or Adjustables $B$15:$E$18

The values we can change to find the answer

Subject to the Constraints .

$B$19:$B$18 = 1 $F$15:$F$18 = 1
15.057 Spring 03 Vande Vate 13

Excel Model

15.057 Spring 03 Vande Vate

14

Options

15

Limits
Max time: Limits the time allowed for the solution process in seconds Iterations:Limits the number of interim calculations. (More details to come)

15.057 Spring 03 Vande Vate

16

Precision
Controls the precision of solutions. Is 1/3 <= 0.3333? 0.333333?

15.057 Spring 03 Vande Vate

17

Quality of Solutions
Tolerance: For integer problems. Later Convergence: For non-linear problems. Later

15.057 Spring 03 Vande Vate

18

Review & Terminology

Objective: Target Cell Equal to: Max or Min Variables: By Changing Cells Constraints: Constraints
LHS: Reference Cell - a function of the variables RHS: Constraint - a constant (ideally)

Options:
Assume Linear Model Assume Non-negative

Solve
15.057 Spring 03 Vande Vate 19

What do you think?

Realistic? Practical? Issues? Questions

15.057 Spring 03 Vande Vate

20

First Kind of Network Model

Sum across row = Const. Sum down column = Const. Each variable in two constraints: A row constraint A column constraint
VP Finance (F) Marketing (M) Operations (O) IT (I) Personnel (P)

Estimated Assignment Costs Leipzig Nancy Liege Tilburg 1 2 3 4 24 10 21 11 14 22 10 15 15 17 20 19 19 24 15 10 11 19 14 13 Leipzig 1 0 0 0 0 0 0 Leipzig 1 Assignments Nancy Liege 2 3 0 0 0 0 0 0 Tilburg 4 0 0 0 0 0 0 0 0 0 0 0 0

VP Finance (F) Marketing (M) Operations (O) IT (I) Personnel (P) VPs Assigned

Plants Assigned 0 0 0 0 0

VP Finance (F) Marketing (M) Operations (O) IT (I) Personnel (P) Total Cost

Cost of Assignments Nancy Liege Tilburg 2 3 4 Total Cost 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

15.057 Spring 03 Vande Vate

21

Influence of Optimization
Changes focus of negotiation about assignments from emotion and personal preferences to estimation of cost

15.057 Spring 03 Vande Vate

22

Motor Distribution
Netherlands 500

Amsterdam The Hague 800

* * *

Antwerp

Belgium

700

Tilburg 500

Germany
*

400

Liege

Leipzig

200

Nancy 900
23

Miles
0 50 100

Transportation Costs
From Origin Amsterdam Antwerp The Hague Leipzig 120 61 102.5 To Destination Nancy Liege 130 41 40 100 90 122 Tilburg 59.5 110 42

Unit transportation costs from harbors to plants

Minimize the transportation costs involved in moving the motors from the harbors to the plants
15.057 Spring 03 Vande Vate 24

A Transportation Model
Autopower Transportation Model
Unit Cost Leipzig From/To Amsterdam $ 120.0 Antwerp $ 61.0 The Hague $ 102.5 Shipments From/To Amsterdam Antwerp The Hague Total Required Total Cost From/To Amsterdam Antwerp The Hague Total Nancy $ 130.0 $ 40.0 $ 90.0 Liege $ 41.0 $ 100.0 $ 122.0 Tilburg $ 59.5 $ 110.0 $ 42.0

Leipzig 400

Nancy 900

Liege 200

Tilburg Total 500

Available 500 700 800

$ $ $ $

Leipzig -

$ $ $ $

Nancy -

$ $ $ $

Liege -

$ $ $ $

Tilburg -

Total $ $ $ $

Challenge
Find a best answer

15.057 Spring 03 Vande Vate

26

Building a Solver Model

Tools | Solver
Set Target Cell: The cell holding the value you want to minimize (cost) or maximize (revenue) Equal to: Choose Max to maximize or Min to minimize this By Changing Cells: The cells or variables the model is allowed to adjust
15.057 Spring 03 Vande Vate 27

Solver Model Contd

that limit the choices of the values of the adjustables. Click on Add
Cell Reference is a cell that holds a value calculated from the adjustables Constraint is a cell that holds a value that constraints the Cell Reference. <=, =, => is the sense of the constraint. Choose one. 15.057 Spring 03 Vande Vate

Subject to the Constraints: The constraints

28

What are the Constraints?

Supply Constraints
Amsterdam: G9 <= H9 Antwerp: G10 <= H10 The Hague: G11 <= H11 G9 is the total volume shipped from Amsterdam Short cut: G9:G11 <= H9:H11 C12 is the total volume shipped to Leipzig Short cut: C12:F12 => C13:F13
29

Demand Constraints
Leipzig: C12 => C13 Nancy: D12 => D13 Liege: E12 => E13 Tilburg: F12 => F13
15.057 Spring 03 Vande Vate

The Model

15.057 Spring 03 Vande Vate

30

Whats Missing?

15.057 Spring 03 Vande Vate

31

Options

15.057 Spring 03 Vande Vate

32

How the Solver Works

Not magic Quick and intuitive Not comprehensive Basic understanding of tool and terms

15.057 Spring 03 Vande Vate

33

How the Solver works

Netherlands 500

Amsterdam The Hague

800
*

* *

A Basic Feasible solution: This choice of variables admits one and only one answer and the answer is non-negative
Germany
*

Tilburg 500 Liege 200

*

Antwerp 700 Belgium

Leipzig

400

Nancy
900
34

Miles
0 50 100

A Basic Feasible Solution

Netherlands
500

Amsterdam The Hague 800 700

Antwerp
* *

500
*

Tilburg 500

Germany 400
*

Belgium

200
Liege
*

Leipzig

400

200

100

800
*

Nancy
900

Miles
0 50 100

15.057 Spring 03 Vande Vate

35

More Technical Detail

Netherlands
500

Amsterdam The Hague 800 700

Antwerp
* *

500
*

Tilburg 500

Germany 400
*

Belgium

200
Liege
*

Leipzig

400

200

100

800
*

Nancy
900

Miles
0 50 100

15.057 Spring 03 Vande Vate

36

Mathematically*
z are the basic variables y are the non-basic variables Write the constraints as Ax = Bz + Ny = b Fix the non-basic variables to y* The unique solution for the basic variables
x = B-1(b Ny*)

B must be invertible and so square Question: We have 7 constraints (3 ports, 4 plants) and only 6 basic variables. How so?
* For those who care to know
15.057 Spring 03 Vande Vate 37

More Technical Detail

Netherlands
500

Amsterdam The Hague 800 700

Antwerp
* *

500
*

Tilburg 500

Germany 400
*

Belgium

200
Liege
*

Leipzig

400

200

100

800
*

Nancy
900

Miles
0 50 100

15.057 Spring 03 Vande Vate

38

How Solver Works

Netherlands
500

Basic Feasible Solution because it is basic AND the basic variables satisfy the bounds ( 0)

Amsterdam The Hague 800 700

Antwerp
* *

500
*

Tilburg 500

Germany 400
*

Belgium

200
Liege
*

Leipzig

400

200

100

800
*

Nancy
900

Miles
0 50 100

15.057 Spring 03 Vande Vate

39

Simple Improvement
Netherlands

Consider sending product from The Hague to Liege

Amsterdam The Hague

Antwerp
* *

Tilburg Liege
*

Germany
*

Belgium

Leipzig

Nancy

Miles
0 50 100

15.057 Spring 03 Vande Vate

40

Conserving Flow

Amsterdam
*

500 800
*

The Hague

Tilburg
700

500

Antwerp

200 Liege

15.057 Spring 03 Vande Vate

Nancy

41

Conserving Flow
Costs $122 $ 40 $162 Saves $100 $ 90 $190 Net $28

Amsterdam
*

500 800
*

The Hague

Tilburg

500

Antwerp

$122
*

$100

200 Liege

$90 Reduced Cost $40

15.057 Spring 03 Vande Vate
*

Nancy

42

How Much Can We Save?

Costs $122 $ 40 $162
Current Flows
Amsterdam
*

Red flows increase Green flows decrease

*

The Hague

Tilburg
700 0
*

500

Saves Antwerp $100 200 $ 90 $190 How much can we send

*

200 Liege

800 100
*

Net $28

to Liege? How much do we save?

Spring 03 Vande Vate Whats the new 15.057 answer? 43

Nancy

New Answer
Costs $122 $ 40 $162 Saves $100 $ 90 $190 Net $28

Amsterdam
*

500 800
*

The Hague

Tilburg

500

Antwerp

200
*

200 Liege

600 300
15.057 Spring 03 Vande Vate
*

Nancy

44

The New Answer

Netherlands

Amsterdam The Hague

Antwerp
* *

Tilburg
*

Germany
*

Belgium

Leipzig

Liege

Nancy

Miles
0 50 100

15.057 Spring 03 Vande Vate

45

The New Answer

Netherlands

Amsterdam The Hague

*

500

800

500 Tilburg 500 200

Liege
*

What kind of solution is this?

Germany

Antwerp

700

300

200 600
*

400

Leipzig

400

Nancy

Miles
0 50 100

900
15.057 Spring 03 Vande Vate 46

An Optimal Basic Feasible Solution

Netherlands 500

500 The Hague 800 700

Antwerp

Amsterdam
* *

300
*

Germany

Tilburg 500
200

100
200
*

700

Liege 200

Leipzig

400

Nancy
900

Miles
0 50 100

15.057 Spring 03 Vande Vate

47

Summary
Solver
Finds a basic feasible solution
Satisfies all the constraints Using these variables there is just one answer

Computes reduced costs of non-basic variables one at a time

How would increasing the new variable affect cost?

Selects an entering variable

Increasing this non-basic variable saves money

Computes a leaving variable

What basic variable first reaches zero?

Repeats
15.057 Spring 03 Vande Vate 48

Sensitivity Analysis
How would the answer change if the data were a little different? Why is this important? Intuitive understanding

15.057 Spring 03 Vande Vate

49

Price Sensitivity
Amsterdam The Hague
*

Antwerp

Tilburg Liege

Leipzig

Nancy
15.057 Spring 03 Vande Vate

At what unit cost would Antwerp to Tilburg be attractive?

50

Price Sensitivity
Amsterdam The Hague $42
*

At what unit cost would Antwerp to Tilburg be attractive?

Costs: Leipzig $110 $ 90 $200 Savings: $42 $40 $82 Net $118
51

Antwerp

$110

Tilburg Liege

$40

$90

Nancy

The carrier would have to PAY us $8!

Try It!
Autopower Transportation Model
Unit Cost Leipzig Nancy Liege Tilburg From/To Amsterdam $ 120.0 $ 130.0 $ 41.0 $ 59.5 Antwerp $ 61.0 $ 40.0 $ 100.0 $ 110.0 The Hague $ 102.5 $ 90.0 $ 122.0 $ 42.0 Shipments From/To Amsterdam Antwerp The Hague Total Required Total Cost From/To Amsterdam Antwerp The Hague Total

Leipzig 400

Nancy 900

Liege 200

Tilburg Total 500

Available 500 700 800

$ $ $ $

Leipzig -

$ $ $ $

Nancy -

$ $ $ $

Liege -

$ $ $ $

Tilburg -

Total $ $ $ $

15.057 Spring 03 Vande Vate

52

Reduced Costs are...

The reduced cost of a variable is The rate of change in the objective if we are forced to use some of that variable The reduced costs of basic variables are 0

15.057 Spring 03 Vande Vate

53

Price Sensitivity: Basic Variables

Amsterdam The Hague
*

Antwerp

Tilburg Liege

Leipzig

If the unit cost from Amsterdam to Liege increases by $10 does the Nancy answer change?
54

15.057 Spring 03 Vande Vate

Checking Reduced Costs: Example

Amsterdam The Hague
*

$120 $102.5 Leipzig Tilburg

Antwerp

$100 $40

$51 = $41 + $10

Liege
$90

Example: Check the reduced cost of Antwerp to Liege

If the unit cost from Amsterdam to Liege increases by $10 does the Nancy answer change?
55

15.057 Spring 03 Vande Vate

Check All Reduced Costs

Amsterdam The Hague

Antwerp

Tilburg Liege

Leipzig
Why only these two?

Nancy
15.057 Spring 03 Vande Vate 56

Value of Price Sensitivity?

15.057 Spring 03 Vande Vate

57

Resource Sensitivity
How would the objective value change if we had more of a resource Tells us the marginal value of that resource If the optimal solution doesnt use all of the resource, then

15.057 Spring 03 Vande Vate

58

Resource Sensitivity
Amsterdam 500 The Hague 700

Antwerp 800

401 Tilburg 500 Leipzig 400

Liege 200

What would happen if we needed 1 more Nancy engine in Leipzig?

900
59

15.057 Spring 03 Vande Vate

Infeasible
Supply
Amsterdam 500 Antwerp 700 The Hague 800 Total 2,000

Demand
Leipzig Nancy Liege Tilburg Total 400+1 900 200 500 2,000+1
15.057 Spring 03 Vande Vate 60

Resource Sensitivity
Amsterdam 500 The Hague 700

Antwerp 800

Tilburg 500 Leipzig 400

15.057 Spring 03 Vande Vate

What would Liege 200 happen to cost if we moved supply for 1 engine from Nancy Amsterdam to 900 Antwerp?

61

Resource Sensitivity
Amsterdam -1 The Hague

$120
Tilburg

$102.5 Leipzig

Antwerp

+1 $40

Whats the value of Liege supply in Antwerp $90 relative to supply in Amsterdam?
Nancy
15.057 Spring 03 Vande Vate 62

Value of Resource Sensitivity

15.057 Spring 03 Vande Vate

63

A Special Feature
We can eliminate any one of the constraints in this problem without changing the answers! Why?

15.057 Spring 03 Vande Vate

64

Redundant Constraint
Supply
Amsterdam 500 Antwerp 700 The Hague 800 Total 2,000

Know shipments from

Amsterdam Antwerp

Demand
Leipzig Nancy Liege Tilburg Total 400 900 200 500 2,000

And they provide...

Leipzig Nancy Liege Tilburg 200 600 0 400
65

15.057 Spring 03 Vande Vate

That means...
We can arbitrarily set the (relative) value of one constraint to 0. (the one we throw away) Set the shadow price or marginal value of supply in Amsterdam to 0, then the shadow price of supply in Antwerp is -$67.5. Why negative? If we had extra supply, where would we want it? Amsterdam or Antwerp?

15.057 Spring 03 Vande Vate

66

Internally Consistent
Given the Shadow Prices
Constraint Shadow Price Amsterdam Antwerp -67.5 The Hague -17.5 Leipzig 120 Nancy 107.5 Liege 41 Tilburg 59.5

Example: Antwerp-Tilburg Reduced Cost = Cost - Origin - Dest. 118 = 110 - (-67.5) - 59.5 = 110 + 8
Edge Amsterdam Leipzig Amsterdam Nancy Amsterdam Liege Amsterdam Tilburg Antwerp Leipzig Antwerp Nancy Antwerp Liege Antwerp Tilburg The Hague Leipzig The Hague Nancy The Hague Liege Value Reduced Cost 300 22.5 200 8.5 700 126.5 118.0 100 200 98.5 Cost 120.0 130.0 41.0 59.5 61.0 40.0 100.0 110.0 102.5 90.0 122.0

We can calculate the Reduced Costs

15.057 Spring 03 Vande Vate

67

Summary
Solver can tell us at what price a non-basic (inactive) variable will be attractive through the Reduced Cost. Solver can tell us how changes in the price of a basic variable affect the solution Solver can tell us the value of a resource via the Shadow Price or Marginal Value

15.057 Spring 03 Vande Vate

68

Sensitivity Info From Solver

15.057 Spring 03 Vande Vate

69

Sensitivity Report: Price

Adjustable Cells Cell $C$9 $D$9 $E$9 $F$9 $C$10 $D$10 $E$10 $F$10 $C$11 $D$11 $E$11 $F$11 Name Amsterdam Leipzig Amsterdam Nancy Amsterdam Liege Amsterdam Tilburg Antwerp Leipzig Antwerp Nancy Antwerp Liege Antwerp Tilburg The Hague Leipzig The Hague Nancy The Hague Liege The Hague Tilburg Final Reduced Objective Value Cost Coefficient 300 120 22.5 130 200 41 59.5 8.5 60.99999999 700 40 126.5 99.99999998 118.0 110 100 102.5 200 90 98.5 122 500 42
15.057 Spring 03 Vande Vate

Allowable Increase 0 1E+30 98.49999999 1E+30 1E+30 8.499999988 1E+30 1E+30 8.499999988 22.49999998 1E+30 0

Allowable Decrease 17.5 22.49999998 41 0 8.499999988 1E+30 126.5 118 0 8.499999988 98.49999999 59.5
70

Sensitivity Report: Resource

Constraints Cell $G$9 $G$10 $G$11 $C$12 $D$12 $E$12 $F$12 Name Amsterdam Total Antwerp Total The Hague Total Total Leipzig Total Nancy Total Liege Total Tilburg Final Shadow Constraint Allowable Allowable Value Price R.H. Side Increase Decrease 500 500 1E+30 0 700 (67.5) 700 200 0 800 (17.5) 800 300 0 400 120.0 400 0 300 900 107.5 900 0 200 200 41.0 200 0 200 500 59.5 500 0 300

15.057 Spring 03 Vande Vate

71

Value
If our proposal comes up non-basic, reduced cost tells us how much harder we have to work to make it attractive. If we are unsure of prices, price sensitivity can tell us whether it is worth refining our estimates of the values Marginal values can help us target investments in capacity
15.057 Spring 03 Vande Vate 72

Caveats
Sensitivity Analysis is pretty nerdy stuff Technical difficulties Only meaningful for changes to a single value Only meaningful for small changes Doesnt work for Integer Programming Can always just change the values and re-solve, but...
15.057 Spring 03 Vande Vate 73

Bad Example
Autopower Transportation Model
Unit Cost Leipzig From/To Amsterdam $ 120.0 Antwerp $ The Hague $ 120.0 Shipments From/To Amsterdam Antwerp The Hague Total Required Nancy Liege $ 120.0 $ 120.0 $ $ $ $ 120.0 Tilburg $ $ 120.0 $ 120.0

Leipzig 400 400 400

Nancy 100 800 900 900

Liege 200 200 200

Tilburg Total 500 500 500

Available 500 500 700 700 800 800 -

Total Cost Leipzig From/To Amsterdam $ Antwerp $ The Hague $ Total $ -

$ $ $ $

Nancy -

Liege $ $ $ $ -

Tilburg $ $ $ $ -

Total $ $ $ $

15.057 Spring 03 Vande Vate

74

Sensitivity
Moving one unit from Liege to Tilburg should save $120
Constraints Cell $C$12 $D$12 $E$12 $F$12 $G$9 $G$10 $G$11 Name Total Leipzig Total Nancy Total Liege Total Tilburg Amsterdam Total Antwerp Total The Hague Total Final Shadow Constraint Allowable Allowable Value Price R.H. Side Increase Decrease 400 120 400 0 0 900 120 900 0 0 200 120 200 0 0 500 500 0 500 500 500 1E+30 0 700 (120) 700 0 0 800 (120) 800 0 0

Solver says the price is not good for 1 unit

15.057 Spring 03 Vande Vate 75

Try It!
Autopower Transportation Model
Unit Cost Leipzig From/To Amsterdam $ 120.0 Antwerp $ The Hague $ 120.0 Shipments From/To Amsterdam Antwerp The Hague Total Required Nancy Liege $ 120.0 $ 120.0 $ $ $ $ 120.0 Tilburg $ $ 120.0 $ 120.0

Leipzig 400 400 400

Nancy 100 800 900 900

Liege 200 200 200

Tilburg Total 500 500 500

Available 500 500 700 700 800 800 -

Total Cost Leipzig From/To Amsterdam $ $ Antwerp The Hague $ $ Total

$ $ $ $

Nancy -

Liege $ $ $ $ -

Tilburg $ $ $ $ -

Total $ $ $ $

15.057 Spring 03 Vande Vate

76

Thursday
Spreadsheet Models
03ShortestPathModel.xls 04TransshipmentModel.xls 05SingaporeElectricGeneratorModel.xls

Download the free student version of the AMPL/CPLEX 8.0 System from www.ampl.com
http://www.ampl.com/cm/cs/what/ampl/DOWNLOADS/cplex71.html#new

AMPL Example Model

TransportationModel.mod

Access Database
TransportationData.mdb
15.057 Spring 03 Vande Vate 77