A Transshipment Problem:

Cars must be +100

shipped from
ports in Boston $30
Newark and $50 2 Newark –200
Jacksonville to 1
other locations
+60 Columbus
at minimum 3 $40
cost.
$40
The costs, $35 $30 Richmond +80
supplies, and Atlanta 4
demands are +170 5
shown on the $50
$25
network. $35 $45

+70 Mobile J'ville –300

$50 7
6

1
Defining the Decision Variables

For each arc in a network flow model

we define a decision variable as:
Xij = the amount being shipped (or flowing) from node i to node
j

For example…
X12 = the # of cars shipped from node 1 (Newark) to node 2 (Boston)
X56 = the # of cars shipped from node 5 (Atlanta) to node 6 (Mobile)

Note: The number of arcs determines

the number of variables!
Defining the Objective Function

Minimize total shipping costs.

MIN: 30X12 + 40X14 + 50X23 + 35X35

+40X53 + 30X54 + 35X56 + 25X65
+ 50X74 + 45X75 + 50X76
 Total Supply = 500 cars
Total Demand = 480 cars

If Total Supply > Total Demand then for every node,

Inflow - Outflow >= Supply or Demand

For example at Node 3,

Inflow = X23 + X53
Outflow = X35
=> X23 + X53 – X35 ≥ 60
Constraint for node 1:
–X12 – X14 >= – 200 (Note: there is no inflow for node 1!)

This is equivalent to:

+X12 + X14 <= 200
Defining the Constraints
Flow constraints
–X12 – X14 >= –200 } node 1
+X12 – X23 >= +100 } node 2
+X23 + X53 – X35 >= +60 } node 3
+ X14 + X54 + X74 >= +80 } node 4
+ X35 + X65 + X75 – X53 – X54 – X56 >= +170 } node 5
+ X56 + X76 – X65 >= +70} node 6
–X74 – X75 – X76 >= –300 } node 7

Nonnegativity conditions
Xij >= 0 for all ij
Constraints for Network Flow Problems:
The Balance-of-Flow Rules
ILP: there is a constraint that solutions should be integers

LP relaxation: ILP with integrality constraints removed (can accept decimals)

With integrity conditions: we can’t use Simplex but “Branch and bounds”

24.10.09
LP Model for Network Model

 The optimal solution of a Network Model can be found by solving an

associated Linear Programming model:

 xij = # of units of flow from node i to node j (for some pairs i-j)

 Minimize the total cost: ∑ cijxij

 Balance flow in each node i:

 If Total Supply > Total Demand: Inflow – Outflow ≥ bi
 If Total Supply < Total Demand: Inflow – Outflow ≤ bi
 If Total Supply = Total Demand: Inflow – Outflow = bi

 Limit each variable xij: lij ≤ xij ≤ uij

9
LP Model for Network Model

 Remember:
 For each arc there always is one decision variable
 For each node there always is one flow balancing constraint

 Special structure of the network flow constraints:

 Each variable appears in two constraints only, once with
coefficient equal to 1, and once with coefficient equal to -1.
 If Demand and Supply are integers, then the optimal solution
is always integer

10
Optimal Solution to the LP
+100

Boston $30
$50 2 Newark -200
120 1
20
Columbus 80
+60
3 $40
$40

40 Richmond
+80
Atlanta 4
+170 5

$45
210
70
+70 Mobile
6 J'ville -300
$50 7
LP Model in Excel

See file Fig5-2.xls and Fig5-2generic.xls

12
The Shortest Path Problem
 Many decision problems boil down to determining one
shortest (or cheapest) route or path through a network.
 Shortest way to drive a car from a city to another city
 Fastest way for an ambulance to get to an accident
 Easiest way to walk across the mountains
 The Shortest Path Problem is a special case of the
Network Flow Problem:
 There is one source node with 1 unit of supply (b i = –1)
 There is one destination node with 1 unit of demand (b i = 1)
 All other nodes do not have any supply or demand (b i = 0)

13
The American Car Association
What is the fastest way +0 L'burg 3.3 hrs
+1
to get from B’ham 9 Va Bch
to Virginia Beach? 5.0 hrs
11
2.0 hrs

4.7 hrs 2.7 hrs

+0 1.1 hrs
K'ville
5 2.0 hrs G'boro Raliegh
3.0 hrs
8 10 +0
+0
1.7 hrs
A'ville 1.5 hrs
6 +0 2.3 hrs

+0 Chatt. 2.8 hrs

3 2.0 hrs Charl.
7 +0
3.0 hrs 1.7 hrs

1.5 hrs
+0 G'ville
4
Atlanta +0
B'ham 2 2.5 hrs
1 2.5 hrs

-1
14
 +0
L'burg 5 pts +1
9 Va Bch
11
9 pts
4 pts

+0 9 pts
+0 4 pts

K'ville 3 pts
5 G'boro Raliegh
9 pts 4 pts 8 10
+0
5 pts
A'ville
6 +0 3 pts
+0 3 pts
Chatt.
3 7 pts
Charl.
8 pts 7
+0
4 pts 4 pts

G'ville 2 pts What is the most scenic

4
Atlanta way to get from B’ham to Va
B'ham +0
2 3 pts Bch?
1
3 pts +0
-1
Solving the Problem
There are two possible objectives for this problem

 Finding the quickest route (minimizing travel time)

 Finding the most scenic route (maximizing the scenic rating

points)

See file Fig5-7.xls

The Equipment Replacement Problem

 The problem of determining when to replace

equipment is another common business problem.

 It can also be modeled as a shortest path problem…

 The Compu-Train Company
 Compu-Train provides hands-on software training.

 Computers must be replaced at least every two years.

 A lease contract is being considered:

Initially the equipment is bought at a cost of $62,000

Prices increase 6% per year

60% trade-in for 1 year old equipment
15% trade-in for 2 year old equipment
 Cost of trading after 1 year =
1.06*62,000 - 0.6*62,000 = $28,520

 Cost of trading after 2 years =

1.062*62,000 - 0.15*62,000 = $60,363

24.10.09
Network for Contract
+0 +0
$63,985
2 4
$28,520 $30,231 $33,968
$32,045

-1 1 3 5 +1
$60,363 $67,824
+0

Cost of trading after 1 year: 1.06*$62,000 - 0.6*$62,000 = $28,520

Cost of trading after 2 years: 1.062*$62,000 - 0.15*$62,000 = $60,363
etc, etc….
24.10.09
Compu-Train – model
xij = 1 if replacement associated with arc i-j is chosen
= 0 otherwise
Min 28,520x12 + 60,363x13 + 30,231x23 + 63,985x24 +
32,045x34 + 67,824x35 + 33,968x45 (total cost)
s.t. – x12 – x13 = –1 (1)
x12 – x23 – x24 = 0 (2)
x13 + x23 – x34 – x35 = 0 (3)
x24 + x34 – x45 = 0 (4)
x35 + x45 = 1 (5)
All xij >= 0


22
See file Fig5-12.xls – one model for both contracts.
Worker Assignment Problem
 Three workers, Ann, Bob, and Cindy, should be assigned
jobs such that the total work time is minimized.
 Three jobs are available and the work time (in hours) for
each worker on each job is given in the table below.
Jobs
Worker 1 2 3
Ann 21 50 40
Bob 35 30 22
Cindy 55 20 25
 Each worker should be assigned exactly one job.

23
 Assignment Model
Xij = 1 if worker i is assigned to job j, Employees Work tim e Jobs

=0 otherwise i=A..C, j=1..3 Ann

21
-1 1 +1
A 50

min 21XA1 + 50XA2 + 40XA3 + 40

35XB1 + 30XB2 + 22XB3 +

35
55XC1 + 20XC2 + 25XC3 Bob 30
2
-1 +1
B
(total work time) 22

XA1 + XA2 + XA3 = 1 (Ann) 55

XB1 + XB2 + XB3 = 1 (Bob) -1

Cindy 20
3 +1
C
XC1 + XC2 + XC3 = 1 (Cindy) 25

XA1 + XB1 + XC1 = 1 (job 1)

XA2 + XB2 + XC2 = 1 (job 2)
What if another worker was available?
XA3 + XB3 + XC3 = 1 (job 3) What if job 3 required 2 workers?
All X24
ij >= 0
Some network flow problems don’t have
trans-shipment nodes; only supply and
demand nodes

Processing
Supply Groves Distances (in miles) Plants
Capacity

Mt. Dora
21
Ocala
These problems are
200,000
275,000
1 50 4 implemented more
40
effectively using the
35 technique described
Eustis 30 Orlando 600,000
400,000
2
22
5 Previously.

55

Clermont 20 Leesburg
300,000 225,000
3 25 6

24.10.09
Transportation Problem: Tropicsun
 Transportation problem is a special case of the network flow
Xij = # of bushels shipped from node i
Processing
to node j, i=1..3, j=4..6 Supply Groves Distances (in miles) Plants
Capacity
min 21X14 + 50X15 + 40X16 + Mt. Dora
21
275,000 Ocala 200,000
35X24 + 30X25 + 22X26 + 1 50 4
40
55X34 + 20X35 + 25X36
(total bushel-miles) 35
Eustis 30 Orlando 600,000
X14 + X15 + X16 = 275,000 (1) 400,000
2 5
22
X24 + X25 + X26 = 400,000 (2)
X34 + X35 + X36 = 300,000 (3) 55
X14 + X24 + X34 <= 200,000 (4) Clermont 20 Leesburg
300,000 225,000
X15 + X25 + X35 <= 600,000 (5) 3 25 6

X16 + X26 + X36 <= 225,000 (6)

All Xij >= 0

26
Note: can state the constraints as inflow-outflow<=0
LP Model for Network Model

 The optimal solution of a Network Model can be found by solving an

associated Linear Programming model:

 xij = # of units of flow from node i to node j (for some pairs i-j)

 Minimize the total cost: ∑ cijxij

 Balance flow in each node i:

 If Total Supply > Total Demand: Inflow – Outflow ≥ bi
 If Total Supply < Total Demand: Inflow – Outflow ≤ bi
 If Total Supply = Total Demand: Inflow – Outflow = bi

 Limit each variable xij: lij ≤ xij ≤ uij

27
Relationship of Network Flow Problems
Generalization Integer solutions
guaranteed if all
bi, lij, uij are integer

Special Case

 Some problems that do not look like network flow

problems may have a network flow problem structure.
 Example: equipment replacement problem

28
Generalized Network Flow Problems
 In some problems, a gain or loss occurs in flows.
 Example of gain
 Interest or dividends on investments
 Examples of loss
 Oil or gas shipped through a leaky pipeline
 Imperfections in raw materials entering a production process
 Spoilage of food items during transit
 Theft during transit

 Generalized Network Flow Problem can be used:

 Arc i-j has yield aij (gain if >1, loss if <1)
 Outflow from i is xij cij aij
 Inflow to j is aijxij i j
[lij, uij]
29
Coal Bank Hollow Recycling Problem
 Coal Bank Hollow Recycling specializes in collecting and recycling paper.
Collected used paper goes through recycling process 1 or 2, and then it is used to
produce paper pulp for newsprint, packaging paper, and print stock. Supply,
demand, costs, and yields (changes of mass) are shown below.
 Taking the Raw Material trough Process 1 and 2:
 To Process 1 To Process 2
 Material Cost Yield Cost Yield Supply
 Newspaper $13 90% $12 85% 70 tons
 Mixed Paper $11 80% $13 85% 50 tons
 White Office Paper $9 95% $10 90% 30 tons
 Cardboard $13 75% $14 85% 40 tons
 Using the paper material from process 1 and 2 to produce final products
 Newsprint Packaging Paper Print Stock
 Pulp Source Cost Yield Cost Yield Cost
Yield
 From Process 1 $5 95% $6 90% $8 90%
 From Process 2 $6 90% $8 95% $7 95%
 Demand 60 tons 40 tons 50 tons

30
Network for Recycling Problem
-70 Newspaper $13
1
$12 Newsprint
95%
90% +0 pulp +60
$5
7
$11 80%
Mixed Recycling 90%
-50 paper 95%
Process 1 $6
2 5
$13 $8
75% 90% Packing
paper
pulp
+40
$9 85% 95% 8
White
85%
-30 office
paper
$6
3 $10 Recycling $8 90%
90% Process 2
$7 Print
6
$13 85% 95%
stock
pulp
+50
+0 9

-40 Cardboard
$14
4

31
Model part 1
 Xij = # of tons of paper material flowing out of node i
on arc i-j (note: flow into node j = yield*Xij)
 Minimize total cost
 Min 13X15 + 12X16 + 11X25 + 13X26
+ 9X35 + 10X36 + 13X45 + 14X46
+ 5X57 + 6X58 + 8X59 + 6X67 + 8X68 + 7X69
 Raw materials constraints
 – X15 – X16 >= –70 (node 1)
 – X25 – X26 >= –50 (node 2)
 – X35 – X36 >= –30 (node 3)
 – X45 – X46 >= –40 (node 4)
32
Model part 2
 Recycling processes constraints
 0.9X15+0.8X25+0.95X35+0.75X45-X57- X58-X59 >= 0 (node 5)
 0.85X16+0.85X26+0.9X36+0.85X46-X67-X68-X69 >= 0 (node 6)

 Paper pulp
 +0.95X57 + 0.90X67 >= 60 (node 7)
 +0.90X57 + 0.95X67 >= 40 (node 8)
 +0.90X57 + 0.95X67 >= 50 (node 9)

 Nonnegativity
 All Xij >= 0

 See file Fig5-17.xls

33
Important Modeling Point
 In generalized network flow problems, gains and/or losses associated
with flows across each arc effectively increase and/or decrease the
available supply.
 This can make it difficult to tell if the total supply is adequate to
meet the total demand. When in doubt, do the following:
 Assume sufficient supply (inflow – outflow >= b i) and try to solve the
problem in Excel.
 If infeasible, assume insufficient supply (inflow – outflow <= b i) and
solve the problem again.
 Unlike in Network Flow Problems, in Generalized Network Flow
Problems integer solutions are not guaranteed because of fractional
gains or losses.

34