network optimization models - citadelmacs.citadel.edu/zhangli/courses-taught/f… · ppt file ·...

Post on 25-May-2018

220 Views

Category:

Documents

2 Downloads

Preview:

Click to see full reader

TRANSCRIPT

© 2015 McGraw-Hill Education. All rights reserved.

© 2015 McGraw-Hill Education. All rights reserved.

Frederick S. Hillier Gerald J. Lieberman

Chapter 10Network Optimization Models

© 2015 McGraw-Hill Education. All rights reserved.

Network Optimization Models

2

© 2015 McGraw-Hill Education. All rights reserved.

10.1 Prototype Example

• The road system for Seervada Park– Location O: park entrance– Location T: a scenic wonder– Trams transport sightseers from park

entrance to location T and back

3

© 2015 McGraw-Hill Education. All rights reserved.

10.1 Prototype Example

4

© 2015 McGraw-Hill Education. All rights reserved.

Prototype Example

• Summary: Park management faces three problems– Determine the route with the smallest total distance

• A shortest-path problem– Determine where telephone lines should be laid

• A minimum spanning tree problem– Determine how to route tram to maximize number of

trips during peak season• A maximum flow problem

5

© 2015 McGraw-Hill Education. All rights reserved.

6

• Network consists of a set of points and a set of lines connecting points

• Node: point (vertex) in the network• Lines: links, arcs, edges, or branches

– Labeled by naming the node at each end• From node precedes the to node

– Have a flow of some type through them• Directed arcs have unidirectional flow• Undirected arcs (links) allow bidirectional flow

10.2 The Terminology of Networks

© 2015 McGraw-Hill Education. All rights reserved.

The Terminology of Networks

7

© 2015 McGraw-Hill Education. All rights reserved.

8

The Terminology of Networks

• Directed network– Network has only directed arcs

• Undirected network– Network has only undirected arcs

• Path between two nodes– A sequence of distinct arcs connecting the

nodes

• Directed path from node i to node j– Sequence of connecting arcs toward node j

© 2015 McGraw-Hill Education. All rights reserved.

9

• Undirected path from node i to node j– Sequence of connecting arcs whose direction can be with

toward or away from node j (note a directed path also satisfies the definition of an undirected path, but not vice versa). Example: in Fig. 10.2, A->B->C->E is a directed path and B->C->A->D is an undirected path

• Cycle: A path begins and ends at the same node; DE-ED is a directed cycle; AB-BC-CA is an undirected cycle

• Connected network: Every pair of nodes in the network has at least one undirected path between them

• Tree (spanning tree): Connected network with no undirected cycles; every spanning tree has n – 1 arcs (the minimum number of arcs needed to have a connected network and the maximum number of arcs possible without having undirected cycles).

The Terminology of Networks

© 2015 McGraw-Hill Education. All rights reserved.

The Terminology of Networks

10

© 2015 McGraw-Hill Education. All rights reserved.

The Terminology of Networks

11

Thereare severalalternativechoicesof addingan arc ateach stageof the process.

© 2015 McGraw-Hill Education. All rights reserved.

12

• Arc capacity– Maximum amount of flow that can be carried

on a directed arc

• Supply node (or source node)– Flow out exceeds flow in

• Demand node (sink node)– Flow in exceeds flow out

• Transshipment node (intermediate node)– Flow in equals flow out

The Terminology of Networks

© 2015 McGraw-Hill Education. All rights reserved.

10.3 The Shortest-Path Problem

• Consider an undirected, connected network– Contains origin and destination nodes– Each link has a nonnegative distance

• The problem– Find the shortest path from origin to destination

13

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

This is a special type of linear programming problem (strictly speaking it is an integer programming problem, but due to the special structure of the model, the integer restriction on the decision variables can be relaxed).

Each node can be thought of as having a unit flow passing through if it is on the selected path, no flow otherwise. Thus, O is the source node, T is the demand node, and other nodes are transshipment nodes.

14

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

Minimize Z = 2xOA + 5xOB + 4XOC + 2XAB + 7xAD + xBC + 4xBD + 3xBE + xCB + 4xCE + xDE + 5xDT + xED + 7xET

s.t. (net flow constraints)

xOA + xOB + xOC = 1 (source node O)

-xDT – xET = -1 (sink node T)

-xOA + xAB + xAD = 0 (node A)

xBC + xBE + xBD – xOB – xAB – xCB = 0 (node B)

xCB + xCE – xOC – xBC = 0 (node C)

xDE + xDT – xAD – xBD – xED = 0 (node D)

xED + xET – xDE – xBE – xCE = 0 (node E)

xij ≥ 0

The optimal solution by Excel Solver (see course website) is xOA = xAB = xBE = xED = xDT = 1 (and other xij = 0) with Z = 13. 15

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

The Excel Solver uses the general simplex method. However, using the shortest path algorithm to solve the problem is much more efficient.

16

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

• Algorithm (using a table or directly applying the algorithm on the graph)– Objective of nth iteration: find the nth nearest

node to the origin• Repeat for n = 1, 2… until destination is reached

– Input for nth iteration: n − 1 nearest nodes to the origin (these nodes are called solved nodes and other nodes are called unsolved nodes), including shortest path and distance from the origin

17

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

• Algorithm (cont’d.) – Candidates for nth nearest node: for each n – 1 solved

node, find an unsolved node (if any) with shortest connecting link to the solved node and make this unsolved node a candidate node (ties provide additional candidates)

– Calculation of nth nearest node• For each solved node and its candidate, add the

distance between them and the distance of the shortest path from the origin to this solved node

• Candidate with smallest total distance is the nth nearest node

18

© 2015 McGraw-Hill Education. All rights reserved.

19

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

• Shortest path for the Seervada park problem– Looking at last column in Table 10.2, two

potential shortest paths exist from the destination to the origin

• T→ D → E → B → A → O or T → D → B → A → O

• Total of 13 miles on either path

20

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

• Network simplex method– An alternate option for solving shortest-path problems

• Three categories of applications– Minimize total distance traveled– Minimize total cost of a sequence of activities– Minimize total time of a sequence of activities

• Other shortest path problems can be solved by slightly modify the algorithm

For example:

- shortest directed path (only directed arcs should be considered)

- shortest path from the origin to all other nodes (stop the algorithm until all nodes are solved nodes)

21

© 2015 McGraw-Hill Education. All rights reserved.

The Shortest-Path Problem

22

© 2015 McGraw-Hill Education. All rights reserved.

10.4 The Minimum Spanning Tree Problem

• Given: nodes of a network, potential links, and positive length of each link if it is inserted into the network– Design the network by inserting links– A path must exist between every pair of nodes

• Problem: minimize total length of links inserted into the network

• Network of n nodes requires only n−1 links– Choose the links to form a spanning tree

23

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Spanning Tree Problem

24

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Spanning Tree Problem

• Applications– Design of telecommunications networks– Design of a lightly-used transportation

network to minimize cost of providing links– Design network of power transmission lines– Electrical equipment wiring– Piping systems

25

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Spanning Tree Problem

• Algorithm– Select any node arbitrarily and then add a link

to connect it to its nearest node– Identify the unconnected node that is closest

to a connected node, and add a link between them

• Repeat until all nodes have been connected

– Ties may be broken arbitrarily• There may be multiple optimal solutions

26

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Spanning Tree Problem

• Example of graphical approach to implementing the algorithm– Problem: installing telephone lines in Seervada park

27

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Spanning Tree Problem

28

The optimal solution does notdepend on the initial node chosen.

Try the algorithmagain using adifferent initialnode.

© 2015 McGraw-Hill Education. All rights reserved.

10.5 The Maximum Flow Problem

• General problem description– All flow through a directed, connected network

originates at a source, and terminates at a sink

• Remaining nodes are transshipment nodes

– Flow through an arc is allowed in only one direction (indicated by the arrowhead)

• Maximum flow is given by arc capacity

– Objective: maximize total flow from source to sink

29

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

• Applications– Maximize flow through company’s distribution

network from factories to customers– Maximize flow through company’s supply

network from vendors to factories– Maximize oil flow through a system of

pipelines– Maximize water flow through aqueducts– Maximize flow of vehicles through a

transportation network30

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

• Algorithms– Simplex method can be used– Augmenting path algorithm is more efficient

• Residual network– Remaining arc capacities after some flows have been assigned

For example, if 5 units flow was sent from O to B with capacity 7, then the residual capacity of arc from O to B is 2. Note that the residual capacity of arc from B to O is 5 (for cancelling some previously assigned flow from O to B).

31

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

• Augmenting path– Directed path from source to sink in residual network such that

every arc on path has positive residual capacity. The minimum of these residual capacities is called the residual capacity of the augmenting path.

32

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

33

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

• Algorithm (each iteration follows these steps)– Identify an augmenting path

• If none exists, net flows already constitute an optimal flow pattern

– Identify the residual capacity, c* of this augmenting path

• It will equal the minimum residual capacity of the arcs on this path

– Increase the flow in this path by c* 34

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

• Algorithm (cont’d.)– Decrease by c* the residual capacity of each

arc on this augmenting path– Increase by c* the residual capacity of each

arc in the opposite direction on this augmenting path

– Return to the first step

• Example: Seervada park transportation problem

35

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

36

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

37

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

38

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

39

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

• Use Excel Solver to solve the problem (see the course website)

40

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

41

© 2015 McGraw-Hill Education. All rights reserved.

The Maximum Flow Problem

42

© 2015 McGraw-Hill Education. All rights reserved.

10.6 The Minimum Cost Flow Problem

• General description of the minimum cost flow problem– The network is directed and connected– At least one of the nodes is a supply node,

and one of the other nodes is a demand node• All remaining nodes are transshipment nodes

– Flow is only allowed in direction of the arrowhead

• Arc capacity gives maximum allowable flow

43

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• General description (cont’d.)– Network has enough arcs with sufficient

capacity to enable all flow generated at supply nodes to reach all demand nodes

– Cost of flow through each arc is proportional to the amount of flow

– Objective: minimize total cost of sending available supply through the network to meet the given demand

44

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

45

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• Linear programming problem formulation

46

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• Linear programming problem formulation

47

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• Feasible solutions property

• Integer solutions property– For minimum cost flow problems where every bi and

uij have integer values, all the basic variables in every basic feasible solution also have integer values

48

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• Example in Section 3.4 (Distribution Unlimited Co.)

49

One of the constraintsis redundant due tosummation of bi’s = 0.

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

50

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• Special cases that fit the minimum cost flow problem– The transportation problem– The assignment problem– The transshipment problem– The shortest-path problem– The maximum flow problem

• All the above problems can be modeled as the minimum cost flow problem by slightly adjusting the problems.

51

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• The transportation problem: there are no transshipment nodes and no capacity constraints on xij (i.e. uij = ∞)

• The assignment problem: same as the transportation problem and bi = 1 for each supply node and bi = -1 for each demand node.

• The shortest path problem: one supply node with supply 1 unit at origin and one demand node with demand 1 unit at the destination; replace each link by directed arcs except for the links with one end at the origin or destination; the distance between i and j becomes cij or cji and uij = ∞.

52

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

53

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• The maximum flow problem: source becomes supply node and sink becomes demand node; let cij = 0 for all existing arcs; select a quantity F which is a safe upper bound on the maximum feasible flow through the network and assign F to the supply node and –F to the demand node; add an arc going directly from the supply node to the demand node with cij = M (big number) and uij = ∞ (since cij = M, the minimum cost flow problem will send F from the supply node to the demand node through other paths as much as possible).

54

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

55

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

• Network simplex method: An alternative method to solving the special cases when the special-purpose algorithms are not available

56

© 2015 McGraw-Hill Education. All rights reserved.

The Minimum Cost Flow Problem

57

© 2015 McGraw-Hill Education. All rights reserved.

10.7 The Network Simplex Method• Streamlined version of the simplex method

– Same basic steps• Finding the entering basic variable• Determining the leaving basic variable• Solving for the new BF solution

• Incorporate the upper bound technique:– To deal with the arc capacity constraints like xij ≤ uij which are

handled like nonnegativity constraints. For example, for arc A->B with uAB = 10 and cAB = 2 and bA = 50 and bB = 40, suppose after some iterations xAB has become the leaving basic variable by reaching it upper bound 10. Then xAB = 10 is replaced by xAB = 10 – yAB, and yAB becomes the new nonbasic variable. Meanwhile, we replace arc A->B by arc B->A (reverse arc) with yAB as its flow quantity and assign this new arc a capacity of 10 and unit cost -2, i.e. uBA = 10, cBA = -2. 58

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

Also, to take xAB = 10 into account, decrease bA from 50 to 40 and increase bB from 40 to 50, i.e. bA = 40 and bB = 50.

59

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

• Network representation of BF solutions– Basic arcs: arcs corresponding to basic variables

• Key property: they never form undirected cycles– Nonbasic arcs: arcs corresponding to nonbasic

variables

60

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

61

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

• BF solutions can be obtained by solving spanning trees– For arcs not in the spanning tree, set the

corresponding variables (xij or yij) equal to zero

– For arcs in the spanning tree, solve for the corresponding variables (xij or yij) in the system of linear equations provided by the node constraints

62

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

• Feasible spanning tree– Spanning tree whose solution from the node

constraints also satisfies all the other constraints

• Fundamental theorem for the network simplex method– Basic solutions are spanning tree solutions

(and conversely)– BF solutions are solutions for feasible

spanning trees (and conversely)63

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

64

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

65

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

Iterations: Iteration I (1) selecting entering basic: candidate arcs are BA, ED and AC.Consider arc AC: add AC to the initial feasible spanning tree with flow Ө. This will create an undirected cycle. Within the cycle, arcs having the same direction as arc AC increase their net flow by Ө, and arcs having opposite direction as arc AC decrease their flow by Ө. Arcs not in the cycle their net flow is unchanged. As a result, we compute the change in Z: ΔZ = cAB*θ + cCE*θ + cDE*(-θ) + cAD*(-θ) = 4θ + θ -3θ - 9θ = -7θ

66

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

67

ΔZ= -2θ+9θ+3θ-θ-3θ = 6θ

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

68

ΔZ = 2θ + 3θ = 5θ

Since it’s a minimizing problem, adding arc AC decreases Z value. Thus,xAC is the entering basic variable.

(2) Selecting the leaving basic variable: By adding arc AC, consider the upper bound of the arcs which have the same direction as AC (sincethe net flow increased by θ) and the lower bound of the arcs which havethe opposite direction as arc AC (the net flow decreased by θ).

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

69

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

70

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

71

• Iteration II

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

72

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

• Iteration III:

73

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

74

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

Iteration IV:

75

No improvement in Z, thus the current BF solution is optimal.

© 2015 McGraw-Hill Education. All rights reserved.

The Network Simplex Method

76

© 2015 McGraw-Hill Education. All rights reserved.

10.8 A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Network based OR techniques developed in the 1950s– PERT (Program Evaluation Review

Technique)– CPM (Critical Path Method)– Both are used in project management

• Concepts have merged into PERT/CPM

• CPM method for time-cost tradeoff– Addresses a project with a specific deadline

77

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• CPM method for time-cost trade-off (cont’d.)– Problem: find optimal plan for expediting

activities to minimize the total cost of completing the project within the deadline

• General approach– Use a network to display the various activities

• And the order in which they need to be performed

– Form optimization model• Solve using linear programming

78

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Prototype example– The Reliable Construction Co. won the contract to

construct a new plant within a time period of 40 weeks– See Table 10.7

• Project network (a network used to represent a project) options– Activity-on-arc (AOA)

• Each activity is represented by an arc• Nodes separate activities from predecessors• Used by original versions of PERT and CPM

79

© 2015 McGraw-Hill Education. All rights reserved.

80

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Project network options (cont’d.)– Activity-on-node (AON)

• Each activity is represented by a node• Arcs show precedence relationships between

activities• Has several advantages over AOA• May become the standard format for project

networks

81

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

82

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Path: One of the routes following the arcs from start to finish

• The critical path (the length of the longest path through the project network)– Equals the estimated project duration– Ties provide more critical paths

83

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Crashing an activity– Taking special costly measures to reduce an activity’s duration

(by special costly measures such as using overtime, hiring additional temporary help and etc.)

– Crashing the project involves crashing a number of activities (normal point: no crashing activity; crash point: activities are fully crashed)

84

© 2015 McGraw-Hill Education. All rights reserved.

85

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Example problem: determine least expensive way to crash activities to reduce overall duration to 40 weeks and the company will be paid $5.4 million (normal point: 44 weeks and cost $4.55 million and crash point: 28 weeks and cost $6.15 million)

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Solution methods– Marginal cost analysis

In each iteration, crash an activity (which is allowed to be crashed) on the critical path with the smallest cost to reduce its duration by a week. If there is a tie, crash the activity on all critical paths that costs the least.

86

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

87

Total crashing cost is $140,000. Figure 10.30 shows the resulting project network, where the darker arrows show the criticalpaths.

If we need go further (for example, the deadline is 39 weeks),the next step would require looking at the activities on allthree critical paths to find the least expensive way of shortening all three critical paths by a week, and that’d be B).

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

88

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Solution methods cont’d– Linear programming

89

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• For example, consider activity J

90

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

91

© 2015 McGraw-Hill Education. All rights reserved.

A Network Model for Optimizing a Project’s Time-Cost Trade-off

• Excel Solver solution (also see on the course website)

92

© 2015 McGraw-Hill Education. All rights reserved.

10.9 Conclusions

• Problems addressed with network models– Optimizing an existing network– Designing a new network

• Minimum spanning tree problem

• CPM method of time-cost trade-offs– Powerful way of applying network optimization

to project management

93

top related