Facilities Design

S.S. Heragu
Decision Sciences and Engineering
Systems Department
Rensselaer Polytechnic Institute
Troy NY 12180-3590

1
 Chapter
Chapter 13
13

Basic
Basic Models
Models
for
for the
the
Location
Location Problem
Problem

2
Outline
Outline
• 13.1 Introduction
• 13.2 Important Factors in Location
Decisions
• 13.3 Techniques for Discrete Space
Location Problems
- 13.3.1 Qualitative Analysis
- 13.3.2 Quantitative Analysis
- 13.3.3 Hybrid Analysis

3
Outline
Outline Cont...
Cont...
• 13.4 Techniques for Continuous Space
Location Problems
- 13.4.1 Median Method
- 13.4.2 Contour Line Method
- 13.4.3 Gravity Method
- 13.4.4 Weiszfeld Method
• 13.5 Facility Location Case Study
• 13.6 Summary
• 13.7 Review Questions and Exercises
• 13.8 References
4
McDonald’s
McDonald’s
• QSCV Philosophy
• 11,000 restaurants (7,000 in USA, remaining
in 50 countries)
• 700 seat McDonald’s in Pushkin Square,
Moscow
• $60 million food plant combining a bakery,
lettuce plant, meat plant, chicken plant, fish
plant and a distribution center, each owned
and operated independently at same location

5
McDonald’s
McDonald’s cont...
• Food taste must be the same at any
McDonald, yet food must be secured locally
• Strong logistical chain, with no weak links
between
• Close monitoring for logistical performance
• 300 in Australia
• Central distribution since 1974 with the help
of F.J. Walker Foods in Sydney
• Then distribution centers opened in several
cities
6
McDonald’s
McDonald’s cont...
• 2000 ingredients, from 48 food plants,
shipment of 200 finished products from
suppliers to DC’s, 6 million cases of food and
paper products plus 500 operating items to
restaurants across Australia
• Delivery of frozen, dry and chilled foods twice a
week to each of the 300 restaurants 98% of the
time within 15 minutes of promised delivery
time, 99.8% within 2 days of order placement
• No stockouts, but less inventory

7
Introduction
• Logistics management can be defined as the
management of transportation and
distribution of goods.

- facility location
- transportation
- goods handling and storage.

8
Introduction Cont...
Some of the objectives in facility location
decisions:
(1) It must first be close as possible to raw
material sources and customers;
(2) Skilled labor must be readily available in the
vicinity of a facility’s location;
(3) Taxes, property insurance, construction and
land prices must not be too “high;”
(4) Utilities must be readily available at a
“reasonable” price;

9
 Introduction Cont...
• (5) Local , state and other government
regulations must be conducive to business;
and
(6) Business climate must be favorable and the
community must have adequate support
services and facilities such as schools,
hospitals and libraries, which are important
to employees and their families.

10
Introduction Cont...
Logistics management problems can be
classified as:

(1) location problems;

(2) allocation problems; and

(3) location-allocation problems.

11
List of Factors Affecting
Affecting
Location Decisions
• Proximity to raw materials sources
• Cost and availability of energy/utilities
• Cost, availability, skill and productivity of labor
• Government regulations at the federal, state,
country and local levels
• Taxes at the federal, state, county and local
levels
• Insurance
• Construction costs, land price

12
List of Factors Affecting
Affecting
Location Decisions Cont...Cont...
• Government and political stability
• Exchange rate fluctuation
• Export, import regulations, duties, and tariffs
• Transportation system
• Technical expertise
• Environmental regulations at the federal,
state, county and local levels
• Support services

13
List of Factors Affecting
Affecting
Location Decisions Cont...
Cont...
• Community services, i.e. schools, hospitals,
recreation, etc.
• Weather
• Proximity to customers
• Business climate
• Competition-related factors

14
13.2
Important Factors in Location
Decisions
Decisions
• International
• National
• State-wide
• Community-wide

15
13.3.1
Qualitative
Qualitative Analysis
Analysis
Step 1: List all the factors that are important,
i.e. have an impact on the location decision.
Step 2: Assign appropriate weights (typically
between 0 and 1) to each factor based on the
relative importance of each.
Step 3: Assign a score (typically between 0 and
100) for each location with respect to each
factor identified in Step 1.

16
13.3.1
Qualitative
Qualitative Analysis
Analysis
Step 4: Compute the weighted score for each
factor for each location by multiplying its
weight with the corresponding score (which
were assigned Steps 2 and 3, respectively)
Step 5: Compute the sum of the weighted
scores for each location and choose a
location based on these scores.

17
Example 1:
•A payroll processing company has recently
won several major contracts in the midwest
region of the U.S. and central Canada and wants
to open a new, large facility to serve these
areas. Since customer service is of utmost
importance, the company wants to be as near
it’s “customers” as possible. Preliminary
investigation has shown that Minneapolis,
Winnipeg, and Springfield, Ill., would be the
three most desirable locations and the payroll
company has to select one of these three.

18
Example 1: Cont...
Cont...
A subsequent thorough investigation of each
location with respect to eight important factors
has generated the raw scores and weights
listed in table 2. Using the location scoring
method, determine the best location for the new
payroll processing facility.

19
Solution:
Steps 1, 2, and 3 have already been completed
for us. We now need to compute the weighted
score for each location-factor pair (Step 4), and
these weighted scores and determine the
location based on these scores (Step 5).

20
Table
Table 2. Factors
Factors and
and Weights for
for
Three
Three Locations
Wt. Factors Location
Minn.Winn.Spring.
.25 Proximity to customers 95 90 65
.15 Land/construction prices 60 60 90
.15 Wage rates 70 45 60
.10 Property taxes 70 90 70
.10 Business taxes 80 90 85
.10 Commercial travel 80 65 75

21
Table
Table 2.
2. Cont...
Cont...

Wt. Factors Location

Minn. Winn. Spring.
.08 Insurance costs 70 95 60
.07 Office services 90 90 80

22
Solution: Cont...
From the analysis in Table 3, it is clear that
Minneapolis would be the best location based
on the subjective information.

23
Table
Table 3.
3. Weighted
Weighted Scores
Scores for
for the
the
Three
Three Locations
Locations
in
in Table
Table 22
Weighted Score Location
Minn. Winn. Spring.
Proximity to customers 23.75 22.5 16.25
Land/construction prices 9 9 13.5
Wage rates 10.5 6.75 9
Property taxes 7 9 8.5
Business taxes 8 9 8.5

24
Table
Table 3.
3. Cont...
Cont...
Weighted Score Location
Minn. Winn. Spring.
Commercial travel 8 6.5 7.5
Insurance costs 5.6 7.6 4.8
Office services 6.3 6.3 5.6

25
Solution: Cont...
Of course, as mentioned before, objective
measures must be brought into consideration
especially because the weighted scores for
Minneapolis and Winnipeg are close.

26
 13.3.2
13.3.2
Quantitative
Quantitative
Analysis
Analysis

27
General
General Transportation
Transportation Model
Model
Parameters
cij: cost of transporting one unit from
warehouse i to customer j
ai: supply capacity at warehouse i
bi: demand at customer j
Decision Variables
xij: number of units transported from
warehouse i to customer j

28
General
General Transportation Model
Minimize Total Transporta tion Cost
m n
Z   cij xij
i 1 j 1
Subject to
n

x
j 1
ij  ai , i  1,2,..., m (supply restriction at warehouse i)
m

x
i 1
ij  b j , j  1,2,..., n (demand requirement at market j)
xij  0, i, j  1,2,..., n (non - negativity restrictions)

29
Example
Example 2:
2:
Seers Inc. has two manufacturing plants at
Albany and Little Rock supplying Canmore
brand refrigerators to four distribution centers in
Boston, Philadelphia, Galveston and Raleigh.
Due to an increase in demand of this brand of
refrigerators that is expected to last for several
years into the future, Seers Inc., has decided to
build another plant in Atlanta or Pittsburgh. The
expected demand at the three distribution
centers and the maximum capacity at the Albany
and Little Rock plants are given in Table 4.

30
Example 2: Cont...
Cont...
Determine which of the two locations, Atlanta
or Pittsburgh, is suitable for the new plant.
Seers Inc., wishes to utilize all of the capacity
available at it’s Albany and Little Rock
Locations

31
Table 4. Costs, Demand and
Supply Information

Bost. Phil. Galv. Rale. Supply

Capacity
Albany 10 15 22 20 250
Little Rock 19 15 10 9 300
Atlanta 21 11 13 6 No limit
Pittsburgh 17 8 18 12 No limit
Demand 200 100 300 280

32
Table 5. Transportation Model
with Plant
Plant at
at Atlanta
Atlanta

Bost. Phil. Galv. Rale. Supply

Capacity
Albany 10 15 22 20 250
Little Rock 19 15 10 9 300
Atlanta 21 11 13 6 330
Demand 200 100 300 280 880

33
Table 6. Transportation Model
with Plant
Plant at
at Pittsburgh
Pittsburgh
Bost. Phil. Galv. Rale. Supply
Capacity
Albany 10 15 22 20 250
Little Rock 19 15 10 9 300
Pittsburgh 17 8 18 12 330
Demand 200 100 300 280 880

34
Min/Max Location Problem:

Location
d11 d12 d1n
d21 d22 d2n
Site

dm1 dm2 dmn

35
13.3.3
Hybrid Analysis
Analysis
• Critical
• Objective
• Subjective

36
Hybrid Analysis
Analysis Cont...
Cont...
CFij = 1 if location i satisfies critical factor j,
0 otherwise
OFij = cost of objective factor j at location i
SFij = numerical value assigned
(on scale of 0-1)
to subjective factor j for location i
wj = weight assigned to subjective factor
(0< w < 1)

37
Hybrid Analysis
Analysis Cont...
Cont...
p
CFM i  CFi1CFi 2    CFip   CFij ,
j 1

i  1,2,..., m
q  q
max i  OFij    OFij
OFM i   j 1  j 1 , i  1,2,..., m
 q
  q

max i  OFij   min i  OFij 
 j 1   j 1 
r
SFM i   w j SFij , i  1,2,..., m
j 1

38
Hybrid Analysis
Analysis Cont...
Cont...
The location measure LMi for each location is
then calculated as:

LMi = CFMi [  OFMi + (1- ) SFMi ]

Where  is the weight assigned to the

objective factor.

We then choose the location with the highest

location measure LMi

39
Example 3:
Mole-Sun Brewing company is evaluating six
candidate locations-Montreal, Plattsburgh, Ottawa,
Albany, Rochester and Kingston, for constructing a
new brewery. There are two critical, three objective
and four subjective factors that management
wishes to incorporate in its decision-making.
These factors are summarized in Table 7. The
weights of the subjective factors are also provided
in the table. Determine the best location if the
subjective factors are to be weighted 50 percent
more than the objective factors.

40
 Table
Table 7:
Critical,
Critical, Subjective
Subjective and
and Objective
Objective
Factor
Factor Ratings for
for six locations
locations for
for
Mole-Sun
Mole-Sun Brewing
Brewing Company,
Company, Inc.
Inc.

41
Table 7. Cont...
Location Factors
Critical
Water Tax
Supply Incentives
Albany 0 1
Kingston 1 1
Montreal 1 1
Ottawa 1 0
Plattsburgh 1 1
Rochester 1 1

42
Table 7. Cont...
Location Factors
Critical Objective
Revenue Labor Energy
Cost Cost
Albany 185 80 10

Kingston 150 100 15

Montreal 170 90 13
Ottawa 200 100 15
Plattsburgh 140 75 8
Rochester 150 75 11
43
Table 7. Cont...
Location Factors
Subjective
Community Ease of
Attitude Transportation
0.3 0.4
Albany 0.5 0.9
Kingston 0.6 0.7
Montreal 0.4 0.8
Ottawa 0.5 0.4
Plattsburgh 0.9 0.9
Rochester 0.7 0.65
44
Table 7. Cont...
Location Factors
Subjective
Labor Support
Unionization Services
0.25 0.05
Albany 0.6 0.7
Kingston 0.7 0.75
Montreal 0.2 0.8
Ottawa 0.4 0.8
Plattsburgh 0.9 0.55
Rochester 0.4 0.8
45
Table 8. Location Analysis of
Mole-Sun Brewing Company,
Company,
Inc., Using
Using Hybrid Method

46
Table 7. Cont...
Location Factors
Critical Objective Subjective LMi
Sum of SFMi
Obj. Factors
Albany -95 0.7 0
Kingston -35 0.67 0.4
Montreal -67 0.53 0.53
Ottawa -85 0.45 0
Plattsburgh -57 0.88 0.68
Rochester -64 0.61 0.56

47
 13.4
13.4
Techniques
Techniques For
For
Continuous Space Location Problems
Problems

48
13.4.1 Model
Model for
for Rectilinear
Rectilinear
Metric Problem
Consider the following notation:
fi = Traffic flow between new facility and
existing facility i
ci = Cost of transportation between new facility
and existing facility i per unit
xi, yi = Coordinate points of existing facility i

49
Model for Rectilinear Metric
Metric
Problem (Cont)
(Cont)
The median location model is then to minimize:

m
TC   c f [| x  x |  | y
i 1
i i i i  y |]

Where TC is the total distribution cost

50
Model for Rectilinear Metric
Metric
Problem (Cont)
(Cont)
Since the cifi product is known for each facility,
it can be thought of as a weight wi
corresponding to facility i.
m m
Minimize TC  w [ | x  x |] w [| y
i 1
i i
i 1
i i  y |]

51
Median Method:
Step 1: List the existing facilities in non-
decreasing order of the x coordinates.
Step 2: Find the jth x coordinate in the list at
which the cumulative weight equals or
exceeds half the total weight for the first
time, i.e.,
j 1 m j m
wi wi

i 1
wi  
i 1 2
and 
i 1
wi  
i 1 2

52
Median Method (Cont)
Step 3: List the existing facilities in non-
decreasing order of the y coordinates.
Step 4: Find the kth y coordinate in the list
(created in Step 3) at which the cumulative
weight equals or exceeds half the total
weight for the first time, i.e.,
k 1 m k m
wi wi

i 1
wi  
i 1 2
and 
i 1
wi  
i 1 2

53
Median Method (Cont)
Step 4: Cont... The optimal location of the new
facility is given by the jth x coordinate and the
kth y coordinate identified in Steps 2 and 4,
respectively.

54
Notes
1. It can be shown that any other x or y
coordinate will not be that of the optimal
location’s coordinates
2. The algorithm determines the x and y
coordinates of the facility’s optimal location
separately
3. These coordinates could coincide with the x
and y coordinates of two different existing
facilities or possibly one existing facility

55
Example 4:
Two high speed copiers are to be located in the
fifth floor of an office complex which houses
four departments of the Social Security
Administration. Coordinates of the centroid of
each department as well as the average number
of trips made per day between each department
and the copiers’ yet-to-be-determined location
are known and given in Table 9 below. Assume
that travel originates and ends at the centroid
of each department. Determine the optimal
location, i.e., x, y coordinates, for the copiers.
56
 Table 9. Centroid Coordinates
and Average
Average Number
Number of
of Trips
Trips to
to
Copiers
Copiers

57
Table 9.
Dept. Coordinates Average number of
# x y daily trips to copiers
1 10 2 6
2 10 10 10
3 8 6 8
4 12 5 4

58
 Solution:
Using the median method, we obtain the
following solution:
Step 1:

Dept. x coordinates in Weights Cumulative

# non-decreasing order Weights
3 8 8 8
1 10 6 14
2 10 10 24
4 12 4 28
59
Solution:
Step 2: Since the second x coordinate, namely
10, in the above list is where the cumulative
weight equals half the total weight of 28/2 =
14, the optimal x coordinate is 10.

60
 Solution:
Step 3:

Dept. x coordinates in Weights Cumulative

# non-decreasing order Weights
1 2 6 6
4 5 4 10
3 6 8 18
2 10 10 28

61
Solution:
Step 4: Since the third y coordinates in the
above list is where the cumulative weight
exceeds half the total weight of 28/2 = 14, the
optimal coordinate is 6. Thus, the optimal
coordinates of the new facility are (10, 6).

62
Equivalent
Equivalent Linear
Linear Model
Model for
for the
the
Rectilinear
Rectilinear Distance,
Distance, Single-
Single-
Facility Location Problem
Parameters
fi = Traffic flow between new facility and existing
facility i
ci = Unit transportation cost between new facility
and existing facility i
xi, yi = Coordinate points of existing facility i
Decision Variables
x, y= Optimal coordinates of the new facility
TC = Total distribution cost
63
Equivalent
Equivalent Linear
Linear Model
Model for
for the
the
Rectilinear
Rectilinear Distance,
Distance, Single-
Single-
Facility Location Problem
The median location model is then to

m m
Minimize TC  w [ | x  x |] w [| y
i 1
i i
i 1
i i  y |]

64
Equivalent
Equivalent Linear
Linear Model
Model for
for the
the
Rectilinear
Rectilinear Distance,
Distance, Single-
Single-
Facility Location Problem
Since the cifi product is known for each facility,
it can be thought of as a weight wi
corresponding to facility i. The previous
equation can now be rewritten as follows
m m
Minimize TC  w [ | x  x |] w [| y
i 1
i i
i 1
i i  y |]

65
Equivalent Linear Model for the
Rectilinear Distance, Single-
Facility Location Problem
Define
 ( xi  x ) if xi  x  0
x i 
0 otherwise
 ( x  xi ) if xi  x  0
x i 
0 otherwise
We can observe that, whether ( xi  x )  0 or  0,
xi  x  x  i  x  i and
( xi  x )  x  i  x  i 66
Equivalent
Equivalent Linear
Linear Model
Model for
for the
the
Rectilinear
Rectilinear Distance,
Distance, Single-
Single-
Facility Location Problem
A similar definition of y  i , y  i yields
yi  y  y  i  y  i and
( yi  y )  y  i  y  i

67
Equivalent
Equivalent Linear
Linear Model
Model for
for the
the
Rectilinear
Rectilinear Distance,
Distance, Single-
Single-
Facility Location Problem
Transformed Linear Model
n
Minimize  i
w
i 1
( x 
i  x 
i  y 
i  y 
i)

Subject to
( xi  x )  x  i - x  i , i  1,2,..., n
( yi  y )  y  i - y  i , i  1,2,..., n
x  i , x  i , y  i , y  i  0, i  1,2,..., n
x , y , unrestricted in sign
68
 13.4.2
13.4.2
Contour
Contour Line
Line Method
Method

69
Algorithm
Algorithm for
for Drawing Contour
Contour
Lines:
Step 1: Draw a vertical line through the x
coordinate and a horizontal line through the y
coordinate of each facility
Step 2: Label each vertical line Vi, i=1, 2, ..., p
and horizontal line Hj, j=1, 2, ..., q where Vi=
the sum of weights of facilities whose x
coordinates fall on vertical line i and where
Hj= sum of weights of facilities whose y
coordinates fall on horizontal line j

70
Algorithm
Algorithm for
for Drawing Contour
Contour
Lines (Cont)
(Cont) m
Step 3: Set i = j = 1; N0 = D0 = w i

i=1
Step 4: Set Ni = Ni-1 + 2Vi and Dj = Dj-1 + 2Hj.
Increment i = i + 1 and j = j + 1
Step 5: If i < p or j < q, go to Step 4. Otherwise,
set i = j = 0 and determine Sij, the slope of
contour lines through the region bounded by
vertical lines i and i + 1 and horizontal line j
and j + 1 using the equation Sij = -Ni/Dj.
Increment i = i + 1 and j = j + 1
71
Algorithm
Algorithm for
for Drawing Contour
Contour
Lines:
Step 6: If i < p or j < q, go to Step 5. Otherwise
select any point (x, y) and draw a contour line
with slope Sij in the region [i, j] in which (x, y)
appears so that the line touches the boundary
of this line. From one of the end points of this
line, draw another contour line through the
adjacent region with the corresponding slope
Step 7: Repeat this until you get a contour line
ending at point (x, y). We now have a region
bounded by contour lines with (x, y) on the
boundary of the region
72
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines
1. The number of vertical and horizontal lines
need not be equal
2. The Ni and Dj as computed in Steps 3 and 4
correspond to the numerator and
denominator, respectively of the slope
equation of any contour line through the
region bounded by the vertical lines i and i +
1 and horizontal lines j and j + 1

73
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
Consider the objective function when the new facility
is located at some point (x, y), i.e., x  x, y  y
m m
TC   wi xi  x   wi yi  y
i 1 i 1

74
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
By noting that the Vi’s and Hj’s calculated in
Step 2 of the algorithm correspond to the sum
of the weights of facilities whose x, y
coordinates are equal to the x, y coordinates,
respectively of the ith, jth distinct lines and that
we have p, q such coordinates or lines (p < m, q
< m), the previous equation can be written as
follows
p q
TC   Vi xi  x   H i yi  y
i 1 i 1

75
 Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
Suppose that x is between the sth and s+1th
(distinct) x coordinates or vertical lines (since
we have drawn vertical lines through these
coordinates in Step 1). Similarly, let y be
between the tth and t+1th vertical lines. Then
s p t q
TC   Vi (x  xi )   V ( x  x)   H ( y  y )   H ( y
i i i i i i  y)
i 1 i  s 1 i 1 i t 1

76
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
Rearranging the variable and constant terms in
the above equation, we get
 s p
  t q

TC   Vi   Vi  x   H i   H i  y
 i 1 i  s 1   i 1 i t 1 
s p t q
  Vi xi  V x   H y   H y
i i i i i i
i 1 i  s 1 i 1 i t 1

77
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
The last four terms in the previous equation can
be substituted by another constant term c and
the coefficients of x can be rewritten as follows

s p s s
TC   Vi  V  V  V
i i i
i 1 i  s 1 i 1 i 1

Notice that we have only added and

subtracted this term s

Vi i 1

78
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
s m
Since it is clear from Step 2 that V   w ,
i 1
i
i 1
i

the coefficient of x can be rewritten as

s
 s p
 s p
2 Vi   Vi   Vi   2 Vi   Vi
i 1  i 1 i  s 1  i 1 i 1
s m
 2 Vi   wi
i 1 i 1
t m
Similarly, the coefficient of y is
2 H i   wi
i 1 i 1
79
Notes on Algorithm for Drawing
Contour Lines (Cont)
 s m
  t m

Thus, TC  2 Vi   wi  x  2 H i   wi  y  c
 i 1 i 1   i 1 i 1 
• The Ni computation in Step 4 is in fact calculation
of the coefficient of x as shown above. Note that
Ni=Ni-1+2Vi. Making the substitution for Ni-1, we get
Ni=Ni-2+2Vi-1+2Vi
• Repeating the same procedure of making
substitutions for Ni-2, Ni-3, ..., we get
• Ni=N0+2V1+2V2+...+2Vi-1+2V1= m i
  wi  2 Vk
i 1 k 1

80
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
m i

Similarly, it can be verified that Di   wi  2 H k

i 1 k 1

 s m
  t m

Thus, TC  2 Vi   wi  x  2 H i   wi  y  c
 i 1 i 1   i 1 i 1 
 N s x  Dt y  c
which can be rewritten as
Ns
y x  (TC  c)
Dt

81
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
The above expression for the total cost function
at x, y or in fact, any other point in the region
[s, t] has the form y= mx + c, where the slope
m = -Ns/Dt. This is exactly how the slopes are
computed in Step 5 of the algorithm

82
Notes on Algorithm for Drawing
Drawing
Contour
Contour Lines
Lines (Cont)
(Cont)
3. The lines V0, Vp+1 and H0, Hq+1 are required for
defining the “exterior” regions [0, j], [p, j], j = 1,
2, ..., p, respectively)
4. Once we have determined the slopes of all
regions, the user may choose any point (x, y)
other than a point which minimizes the
objective function and draw a series of
contour lines in order to get a region which
contains points, i.e. facility locations, yielding
as good or better objective function values
than (x, y)
83
Example 5:
Consider Example 4. Suppose that the weight
of facility 2 is not 10, but 20. Applying the
median method, it can be verified that the
optimal location is (10, 10) - the centroid of
department 2, where immovable structures
exist. It is now desired to find a feasible and
“near-optimal” location using the contour line
method.

84
Solution:
The contour line method is illustrated using
Figure 1
Step 1: The vertical and horizontal lines V 1, V2, V2
and H1, H2, H2, H4 are drawn as shown. In addition
to these lines, we also draw line V0, V4 and H0, H5
so that the “exterior regions can be identified
Step 2: The weights V1, V2, V2, H1, H2, H2, H4 are
calculated by adding the weights of the points
that fall on the respective lines. Note that for this
example, p=3, and q=4

85
Solution:
4

Step 3: Since w
i 1
i  38

set N0 = D0 = -38

Step 4: Set
N1 = -38 + 2(8) = -22; D1 = -38 + 2(6) = -26;
N2 = -22 + 2(26) = 30; D2 = -26 + 2(4) = -18;
N3 = 30 + 2(4) = 38; D3 = -18 + 2(8) = -2;
D4 = -2 + 2(20) = 38;
(These values are entered at the bottom of each
column and left of each row in figure 1)
86
Solution:
Step 5: Compute the slope of each region.
S00 = -(-38/-38) = -1; S14 = -(-22/38) = 0.58;
S01 = -(-38/-26) = -1.46; S20 = -(30/-38) = 0.79;
S02 = -(-38/-18) = -2.11; S21 = -(30/-26) = 1.15;
S03 = -(-38/-2) = -19; S22 = -(30/-18) = 1.67;
S04 = -(-38/38) = 1; S23 = -(30/-2) = 15;
S10 = -(-22/-38) = -0.58; S24 = -(30/38) = -0.79;
S11 = -(-22/-26) = -0.85; S30 = -(38/-38) = 1;
S12 = -(-22/-18) = -1.22; S31 = -(38/-26) = 1.46;
S13 = -(-22/-2) = -11; S32 = -(38/-18) = 2.11;

87
Solution:
Step 5: Compute the slope of each region.

S33 = -(38/-2) = 19;

S34 = -(38/38) = -1;

(The above slope values are shown inside each

region.)

88
Solution:
Step 6: When we draw contour lines
through point (9, 10), we get the
region shown in figure 1.

Since the copiers cannot be placed at the

(10, 10) location, we drew contour lines
through another nearby point (9, 10).
Locating anywhere possible within this
region give us a feasible, near-optimal
solution.

89
 13.4.3
13.4.3
Single-facility Location Problem with
Squared Euclidean
Euclidean Distances
Distances

90
La Quinta Motor Inns
Moderately priced, oriented towards business
travelers
Headquartered in San Antonio Texas
Site selection - an important decision
Regression Model based on location
characteristics classified as:
- Competitive, Demand Generators,
Demographic, Market Awareness, and
Physical

91
La Quinta Motor Inns (Cont)
Major Profitability Factors - Market awareness,
hotel space, local population, low
unemployment, accessibility to downtown office
space, traffic count, college students, presence
of military base, median income, competitive
rates

92
 Gravity
Gravity Method:
Method:
The cost function is

 
m
Minimize TC  ii i
c f
i 1
( x  x ) 2
 ( y i  y ) 2

As before, we substitute wi = fi ci, i = 1, 2, ..., m

and rewrite the objective function as
m m
Minimize TC   i i
w (
i 1
x  x ) 2
  i i
w ( y
i 1
 y ) 2

93
Gravity
Gravity Method
Method (Cont)
(Cont)
Since the objective function can be shown to be
convex, partially differentiating TC with respect
to x and y, setting the resulting two equations
to 0 and solving for x, y provides the optimal
location of the new facility
TC m m
 2 wi x  2 wi xi  0
x i 1 i 1
m m
 x   wi xi w i
i 1 i 1
94
Gravity
Gravity Method
Method (Cont)
(Cont)
Similarly,
TC m m
 2 wi y  2 wi yi  0
y i 1 i 1
m m
 y   wi yi w i
i 1 i 1

Thus, the optimal locations x and y are simply

the weighted averages of the x and y coordinates
of the existing facilities
95
Example 6:
Consider Example 4. Suppose the distance
metric to be used is squared Euclidean.
Determine the optimal location of the new
facility using the gravity method.

96
Solution - Table 10
Department i xi yi wi wixi wiyi
1 10 2 6 60 12
2 10 10 10 100 100
3 8 6 8 64 48
4 12 5 4 48 20
Total 28 272 180

From table 10, we conclude that

x  272 28  9.7 and y  180 28  6.4
97
Example 6. Cont...
Cont...
If this location is not feasible, we only need to
find another point which has the nearest
Euclidean distance to (9.7, 6.4) and is a feasible
location for the new facility and locate the
copiers there

98
 13.4.4
13.4.4
Weiszfeld
Method
Method

99
Weiszfeld Method:
The objective function for the single facility
location problem with Euclidean distance can
be written as:
m
Minimize TC  c i f i (x i  x) 2  (y i  y) 2
i 1

As before, substituting wi=cifi and taking the

derivative of TC with respect to x and y yields

100
Weiszfeld Method:
TC 1 m w i  2(x i  x)
 
x 2 i 1 (x i  x) 2  (yi  y) 2
m
wixi
  
i 1 (x i  x) 2  (yi  y) 2
m
wix
 0
i 1 (x i  x) 2  (yi  y) 2

101
Weiszfeld Method:
m
wixi

i 1 (x i  x) 2  (y i  y) 2
x  m
wi

i 1 (x i  x) 2  (y i  y) 2

102
Weiszfeld Method:
TC 1 m w i  2(yi  y)
 
y 2 i 1 (x i  x) 2  (yi  y) 2
m
w i yi
  
i 1 (x i  x) 2  (yi  y) 2
m
wiy
 0
i 1 (x i  x) 2  (yi  y) 2

103
Weiszfeld Method:
m
w i yi

i 1 (x i  x) 2  (y i  y) 2
y  m
wi

i 1 (x i  x) 2  (y i  y) 2

104
Weiszfeld Method:
Step 0: Set iteration counter k = 1;
m m

w x i i w y i i
xk  i 1
m
; yk  i 1
m

w
i 1
i w
i 1
i

105
Weiszfeld Method:
Step 1: Set
m
wixi

i 1 (x i  x k ) 2  (y i  y k ) 2
k 1
x  m
wi

i 1 (x i  x k ) 2  (y i  y k ) 2

106
Weiszfeld Method:
m
wixi
i 1 (x i  x )  (y i  y ) 2
k 2 k
 x k 1  m
wi
i 1 (x i  x )  (y i  y ) 2
k 2 k

• Step 2: If xk+1 = xk and yk+1 = yk, Stop. Otherwise, set k =

k + 1 and go to Step 1
107
Example 7:
Consider Example 5. Assuming the distance
metric to be used is Euclidean, determine the
optimal location of the new facility using the
Weiszfeld method. Data for this problem is
shown in Table 11.

108
 Table
Table 11.
11.
Coordinates
Coordinates and
and weights
weights for
for
44 departments
departments

109
Table 11:
Departments # xi yi wi
1 10 2 6
2 10 10 20
3 8 6 8
4 12 5 4

110
Solution:
Using the gravity method, the initial seed can
be shown to be (9.8, 7.4). With this as the
starting solution, we can apply Step 1 of the
Weiszfeld method repeatedly until we find that
two consecutive x, y values are equal.

111