You are on page 1of 72

FACILITIES PLANNING & DESIGN

Alberto Garcia-Diaz
J. MacGregor Smith

CHAPTER 5
Location Models in Facilities Planning
Types of Models
 Single-Facility-Unit Location Model

 Location-allocation Model

 Quadratic Assignment Model


Distances
Rectilinear (Manhattan) distances
Single-Facility Location Model

1. Minisum Problem
2. Minimax Problem
Existing units/facilities
New unit/facility

|x-a|
b
|y-b| Distance= |x-a|+ |y-b|

y=?

x=? a
W = Number of trips from to

W
b
W
y

x a
Weighted Distance = W |x-a| + W |y-b|

W |x-a|
b
W |y-b|

x a
Minisum Location Problem with Rectilinear
Distances
m
Minimize f ( x , y )   ( Wi x  ai  Wi y  bi )
i 1
m m m m m m m m

Minimize f ( x, y )  Wi x  ai  Wi y  bi Minimize f ( X )  Wi x  ai  Wi y  bi Minmzef(X) Wixai  Wiy bi Minimize f ( X )  Wi x  ai  Wi y  bi
i1 i1 i1 i1 i1 i1 i1 i1

m m m m
Minimize f ( x )  Wi x  ai  Wi y  bi Minimize f ( X )  Wi x  ai  Wi y  bi
i1 i1 i1 i1 SEPARABLE
m m m m
MODELS
Minimize f ( y )  Wi x  ai  Wi y  bi Minimize f ( X )  Wi x  ai  Wi y  bi
i1 i1 i1 i1
Example
Machines 1, 2, 3, and 4 are located at the points (8, 5),
(4, 2), (11, 8) and (13, 2), respectively. There are 9, 6, 4,
and 12 trips per week, respectively, between the
machines and a new facility unit. Find the location of the
new facility unit.

y

(11, 8)
3
(8, 5)
1

(4, 2) (13, 2)
2 4
(0, 0) x
x-coordinate y-coordinate trips

Machine 1 a1 = 8 b1 = 5 W1 = 9
Machine 2 a2 = 4 b2 = 2 W2 = 6
Machine 3 a3 = 11 b3 = 8 W3 = 4
Machine 4 a4 = 13 b4 = 2 W4 = 12

Minimize f(x,y) = 9 | x-8 | + 6 | x-4 | + 4 | x-11 | + 12 | x-13| +


9 | y-5 | + 6 | y-2 | + 4 | y - 8 | + 12 | y-2 |
Reformulation as an LP Model

Let ri  ( x  ai )  and si  (ai  x) 


where
  y if y  0
( y)  
 0 otherwise
Note that : | x  ai | ri  si
and x  ai  ri  si
Exercise:
write LP models to find optimal location
Machines 1, 2, 3, and 4 are located at the points (8, 5),
(4, 2), (11, 8) and (13, 2), respectively. There are 9, 6, 4,
and 12 trips per week, respectively, between the
machines and a new facility unit. Find the location of the
new facility unit.

y

(11, 8)
3
(8, 5)
1

(4, 2) (13, 2)
2 4
(0, 0) x
Reformulation as an LP Model
Minimize Z1 = 9 (r1 + s1)+ 6 (r2 + s2) + 4 (r3 + s3) + 12 (r4 + s4)
x - r 1 + s1 = 8
x - r 2 + s2 = 4
x - r3 + s3 = 11
x - r4 – s4 = 13
ri, si > 0, i = 1,2,3,4
Minimize Z2 = 9 (r1 + s1)+ 6 (r2 + s2) + 4 (r3 + s3) + 12 (r4 + s4)
y - r 1 + s1 = 5
y - r 2 + s2 = 2
y - r 3 + s3 = 8
y - r 4 + s4 = 2
ri, si > 0, i = 1,2,3,4
LP OPTIMUM FOUND AT STEP 6
OBJECTIVE FUNCTION VALUE
1) 93.00
VARIABLE VALUE REDUCED COST
r1 3.000000 0.000000
s1 0.000000 18.000000
r2 7.000000 0.000000
s2 0.000000 12.000000
r3 0.000000 7.000000
s3 0.000000 1.000000
r4 0.000000 24.000000
s4 2.000000 0.000000
x 11.0000 0.000000
ROW SLACK OR SURPLUS DUAL PRICES
2) 0.000000
9.000000
3) 0.000000
6.000000
4) 0.000000 -
3.000000
5) 0.000000 -
12.000000
6) 3.000000
0.000000
7) 0.000000
0.000000
8) 7.000000
LP OPTIMUM FOUND AT STEP 4
OBJECTIVE FUNCTION VALUE
1) 51.00
VARIABLE VALUE REDUCED COST
r1 0.000000 18.000000
s1 3.000000 0.000000
r2 0.000000 5.000000
s2 0.000000 7.000000
r3 0.000000 8.000000
s3 6.000000 0.000000
r4 0.000000 0.000000
s4 0.000000 24.000000
Y 2.000000 0.000000
ROW SLACK OR SURPLUS DUAL PRICES
2) 0.000000 -9.000000
3) 0.000000 1.000000
4) 0.000000 -4.000000
5) 0.000000 12.000000
6) 0.000000 0.000000
7) 0.000000 0.000000
8) 0.000000 0.000000
9) 0.000000 0.000000
10) 3.000000 0.000000
11) 0.000000 0.000000
12) 6.000000 0.000000
13) 0.000000 0.000000
Minimize Σi wi(ri+si )
x - ri + si = ai, i = 1,2,...,m
ri, si > 0, i = 1,2,...,m

x r1 s1 r2 s2 … ri si … rm sm RHS
1 -1 +1 0 0 … 0 0 … 0 0 a1

1 0 0 -1 +1 … 0 0 … 0 0 a2
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
1 0 0 0 0 … -1 +1 … 0 0 ai
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
1 0 0 0 0 … 0 0 … -1 +1 am
Rules to determine the optimal value of x
• Arrange the ai’s in increasing order
• Compute sum of weights (trips) for each value
• Optimal x is the value at which the cumulative
weight > ½ (sum of weights)
Rules to determine the optimal value of y
• Arrange the bi’s in increasing order
• Compute sum of weights (trips) for each value
• Optimal y is the value at which the cumulative
weight > ½ (sum of weights)
Calculation of Optimal x*

Machine Coordinate Trips


i ai wi Sum of wi
2 4 6 6
1 8 9 15
3 11 4 19
4 13 12 31

1 4 31
x 
*
Wi   15.5
2 i1 2
1 4 31
Calculation of optimal y* 
2 i1
Wi 
2
 15.5

Machine Coordinate Weight Sum of wi


i bi wi

2 2 6 6
4 2 12 18
1 5 9 27
3 8 4 31

y * Optimal Location: (x*, y*) = (11, 2)


y
(11, 8)

(8, 5)

(4, 2) (13, 2)

(0, 0) x

f(x*,y*) = 9 | 11-8 | + 6 | 11-4 | + 4 | 11-11 | + 12 | 11-13| +


9 | 2-5 | + 6 | 2-2 | + 4 | 2-8 | + 12 | 2-2 |
= 93 + 51
= 144
Exercise
Find the optimal location (x,y) for a new facility, using
the minisum model & rectilinear distances. There are
five existing machines located at P1 (1,1), P2 (5,2),
P3(2,8), P4 (4,4), P5 (8,6). Number of trips per day
between the new facility & the existing machines: 5, 6,
2, 4, 8.

• Arrange the ai’s in increasing order


• Compute sum of weights (trips) for each value
• Optimal x is the value at which the cumulative weight > ½ (sum
of weights)
4
1 25
Calculation of optimal x *

2 i 1
Wi 
2
 12.5

Machine Coordinate Weight Sum of wi


i ai wi

1 1 5 5
3 2 2 7
4 4 4 11
2 5 6 17
5 8 8 25

x*
Optimal Location: x* = 5
4
1 25
Calculation of optimal y *

2 i 1
Wi 
2
 12.5

Machine Coordinate Weight Sum of wi


i bi wi

1 1 5 5
2 2 6 11
4 4 4 15
5 6 8 23
3 8 2 25
y*

Optimal Location: (x*, y*) = (5, 4)


Graphical Approach
1. Cumulative Weight Diagram
2. Resultant Force Diagram
Cumulative Weight Diagram for X

0 4 8 11 13
X

4
6
9
12
11
31
6
25
9
16 1 4
Wi  15.5
2 i1
4
12

12

0
Cumulative Weight Diagram for Y

0 2 5 8
Y


6
4

12

Y=2
31

13

4
0
Exercise
There are five existing machines located at P1 (1,1), P2
(5,2), P3(2,8), P4 (4,4), P5 (8,6). Number of trips per
day between the new facility & the existing machines:
5, 6, 2, 4, 8. Construct the Cumulative Weight
Diagrams & find the optimal location for the new
facility.
Resultant Force Diagram
Key point: the number of trips is viewed as a
force. If we are finding x this force acts
horizontally. If we are finding y this force acts
vertically.
0 4 8 11 13
X

4 Note this…
6
9
12

Resultant Force = Total Force left – Total Force right


RF = 6 – 25 = -19

0 4 8 11 13
X
-31 -19 -1 7 31
4
6
9
12
0 4 8 11 13 0 4 8 11 13
X X
-31 -19 -1 7 31
4 4
6 6
9 9
12 12
11 11
31 -31

25 -19

16 1 4
Wi  15.5
2 i1
-1 0
12 7

0 31

Cumulative Weight Diagram Resultant Force Diagram


 

Y
Y
(11, 8)
8

(8, 5)
5

(4, 2) (13, 2)
2

(0, 0) 0
X

0 2 5 8
Y


6
4

12
0 2 5 8 0 2 5 8
Y Y


6 6
4 4

9 9

12 12

31 -31

13 5

4 23
0 31

Cumulative Weight Diagram Resultant Force Diagram


Exercise
There are five existing machines located at P1 (1,1), P2
(5,2), P3(2,8), P4 (4,4), P5 (8,6). Number of trips per
day between the new facility & the existing machines:
5, 6, 2, 4, 8. Construct the Resultant Force Diagrams &
find the optimal location for the new facility
Median Condition – Why does it work?
f(x) FU W
 90 x  750
1 3 4 10

90 x  750
2 4 6 20
 70 x  690 3 8 9 20
 30 x  530 4 12 8 40
10 x  210
f(x*)

x
𝑓 ( 𝑥 ) =10|𝑥 −3|+20| 𝑥 − 4|+20|𝑥 − 8|+40∨𝑥 − 12∨¿
0 x3
f ( x )  10(3  x )  20( 4  x )  20(8  x )  40(12  x )  90 x  750
3 x  4
f ( x )  10( x  3)  20( 4  x )  20(8  x )  40(12  x )  70 x  690
4 x8
f ( x )  10( x  3)  20( x  4)  20(8  x )  40(12  x )  30 x  530
8  x  12
f ( x )  10( x  3)  20( x  4)  20( x  8)  40(12  x )  10 x  210
12  x
f ( x )  10( x  3)  20( x  4)  20( x  8)  40( x  12)  90 x  750
Median Condition – Why does it work?

… wi-1 wi wi+1 …
 Consider interval ( wi-1 , wi ) with negative slope
w1    wi 1  wi    wn
Adding w1+…+wi-1 to both sides:
W
w1    wi 1 
2
Median Condition – Why does it work?

… wi-1 wi wi+1 …
 Consider interval ( wi , wi+1 ) with positive slope
w1    wi 1  wi  wi 1    wn
Adding w1+…+wi to both sides:
W
w1    wi 
2
Minimax Location Problem with Rectilinear
Distances
Existing facility
New facility

|x-a|

|y-b|
y=?

x=?
Minimax Location Problem with Rectilinear
Distances

Place new machine so that longest rectangular


distance will be as low as possible

x
Minimax Model
Minimize Maximum Distance

Min{z(x,y) = Max [|x–a1|+|y–b1|,|x–a2|+|y–b2|,…,|x–am |+|y–bm|]}

{
Min z(x,y) = Max [ |x – ai | + |y – bi | ]
i = 1,…,m
}
Diamond Covering

( 𝑎𝑖 − 𝑥 ) + ( 𝑦 − 𝑏𝑖 )= 𝑟 ( 𝑥 − 𝑎 𝑖 ) + ( 𝑦 − 𝑏𝑖 )= 𝑟

𝑟
(𝑎 𝑖 , 𝑏𝑖 )

( 𝑎𝑖 − 𝑥 ) + ( 𝑏𝑖 − 𝑦 )= 𝑟 ( 𝑥 − 𝑎 𝑖 ) + ( 𝑏𝑖 − 𝑦 )= 𝑟

x  ai  y  bi  r
c1 = min (ai+bi )
c2 = max (ai+bi )
c3 = min (- ai+bi )
c4 = max(- ai+bi )
c5 = max (c2-c1, c4-c3 )

Optimal Solution: line joining the following points

A(x*, y*) = ½ (c1 - c3, c1+c3+c5 )


B(x*, y*) = ½ ( c2 - c4, c2 + c4 - c5 )

Minimum Longest Distance, z* = c5 / 2


12 -x+y= 4,
gives c4=4
10
8 p5 p8

6 p2
y

x + y= 16,
p7
4 p4 gives c2=16
p6 x*
2 p 3 -x+y= -6,
p1 gives c3=-6
0 2 4 6 8 10 12
x

x + y= 0, c5 =16
gives c1=0
12

10
1. Construct a
smallest 8 P5 P8
rectangle whose
sides are at ±45°
with x and y 6 P2
axes
P7
2. Extend the 4 P6 P4
smaller edges of
this rectangle if
necessary to 2 P3
form a square
P1
3. Center of the 0 2 4 6 8 10 12
square formed
is an optimal
minimax
solution. Set of minimax
locations is the black
bold line
Exercise
Locate an emergency service unit to serve the following sub-
divisions of a facility:

Subdivision ai bi
A 4 3
B 5 11
C 13 13
D 10 6
E 4 6
F 10 10
G 14 2
Subdivision ai bi ai+bi -ai+bi
A 4 3 c1=7 -1 c1 = min (ai+bi )
B 5 11 16 c4= 6 c2 = max (ai+bi )
C 13 13 c2=26 0 c3 = min (- ai+bi )
D 10 6 16 -4 c4 = max (- ai+bi )
E 4 6 10 2 c5 = max (c2-c1, c4-c3 )
F 10 10 20 0
G 14 2 16 c3= -12

c5 = max {26-7, 6+12} = 19


z* = 19/2 = 9.5
A(x*, y*) = ½ (c1-c3, c1+c3+c5 )= (9.5,7)
B(x*, y*) = ½ ( c2-c4, c2+c4 - c5 )= (10, 6.5)
A (9.5,7)

B (10, 6.5)


Location-allocation Problem

Existing

New
Mathematical Model

n m
Minimize  =   zjiwjid(Xj,Pi) + g(n)
j 1 i1
n

Subject to: zji = 1, i = 1,2,...,m


j 1
Each existing facility is
n = 1,2, ..., m allocated to exactly
zji 
Cost (xjunit
per
{0, , yj) :distance
(a1}
location
Cost of
,i bi) : location
ofowning
oneof new facility
n
new facility j facilities
new
existing facility i
Solution Procedure

1. Find possible combinations


2. For each combination indicate allocation
3. Eliminate (if any) allocation combinations that yield
intersecting service regions
4. Use single-facility facility procedure to locate each new facility
5. For each group, compute the trip cost
6. Find the total cost for all groups
7. Select the global minimum
S(1, 3) = 1

1
Exercise S(2, 3) = 3
F1 F2

1 2,3

2 1,3

3 3 1,2

1
S(3, 3) = 1

1
Example of Location-Allocation Problem

Machine Location Frequency of


repair
1 (0, 0) 10

2 (1, 2) 5

3 (2, 1) 15

Cost per unit distance = $5


Cost of owning and operating each new facility = $300
2

1
Using the minisum 2
model
Solution 3
1
For n = 1, m = 3

Combination Allocations F1 Total cost


to F1 $
1 1,2,3 (1,1) 700

Cost = transport cost + owning cost


One-way transport cost = 5[10(1+1)+5(0+1)+15(1+0)]=200
Total Cost = 2(200) + 300 = $700
For n = 2, m = 3
Number of combinations = S(2, 3) = 3
The combinations are: F1 F2

1 2,3
Locate new facility F1 at 2
location of machine 2 2 1,3
(1, 2). Evaluate location 3
of F2. 3 1,2 1

Using the minisum model, we locate F2 at the location of


Machine 3: (2, 1)
2
F1

3 F2

1
2 F F1 F2
1

Combination 2 3
1 2,3
F2
2 1,3
1
3 1,2
Cost of any F = transport cost + owning cost
Total cost = F1 cost + F2 cost

F1 cost = 0 + 300 = $300

F2 cost= transport cost + owning cost


One-way transport cost = (5)(10)(2+1) = 150
Two-way transport cost = $300

Total cost = $300 + $300 + $300 = $900

Exercise: Do the same thing for combinations 1 & 3.


n=1
Combination Allocations to F1 Total
F1 Cost $
1 1,2,3 (1,1) 700
n=2
Combination Alloc to Alloc to F1 F2 Total
F1 F2 Cost $
1 1 2,3 (0,0) (2,1) 700
2 2 1,3 (1,2) (2,1) 900
3 3 1,2 (2,1) (0,0) 750
n=3
Combination Alloc Alloc Alloc F1 F2 F3 Total
to F1 to F2 to F3 Cost $
1 1 2 3 (0,0) (1,2) (2,1) 900
Claim: Assume that there are only two existing facilities at
locations . If the distance function is symmetric and observers the
triangular inequality, i.e., then always gives the unique optimum
solution for the single facility location problem with minisum
objective.

Proof: Assume the claim is wrong and there is a point which


minimizes the minisum objective. Then we have the optimal
solution value .
Now consider the objective function value for locating the new
facility at which i + .
But then we have which contradicts being the minisum solution
value. Hence, the result follows.
A Greedy Heuristic
Customer Warehouse Sites
Location A B C D E
1 100 500 1,800 1,300 1,700
2 1,500 200 2,600 1,400 1,800
3 2,500 1,200 1,700 300 1,900
4 2,800 1,800 700 800 800
5 10,000 12,000 800 8,000 900
Fixed Cost 3,000 2,000 2,000 3,000 4,000

TCA=19,900; TCB=17,700; TCC=9,600;


TCD=14,800; TCE=11,100
Add C(A)=3000+(100-1800)+(1500-2600)=200
2nd location C(B)=-2200, C(D)=-100, C(E)=3900
A Greedy Heuristic
 The new total annual cost is
9600-2200 = 7400
 After warehouses B and C, if we want to add a
new one, the new C(j) from the third WH j will
be:
– C(A)= 3000-400= 2600
– C(D)= 3000-900 = 2100
– C(E)= 4000-0 = 4000
 No additional WH are justified.
 Built WHs at sites B and C at a cost of 7400.
A Greedy Heuristic
Customer Warehouse Sites
Location A B C D E
1 100 500 1,800 1,300 1,700
2 1,500 200 2,600 1,400 1,800
3 2,500 1,200 1,700 300 1,900
4 2,800 1,800 700 800 800
5 10,000 12,000 800 8,000 900
Fixed Cost 3,000 2,000 2,000 3,000 4,000

Add C(A)=3000-400=2600
3rd location C(D)=3000-900=2100
C(E)=4000
No additional WH are justified.
Built WHs at sites B and C at a cost of 7400.
Exercise
Exercise (solution)

Add MC(A)=6000-300=5700
2nd location MC(B)=5000-2300-300-500=1900,
MC(C)=5400-1500-4000=-100,
MC(D)=7000
Solution: Open E and C with a total cost of :
6000+5400+200+2800+1900+1000+5000= 22300
Quadratic Assignment Problem (QAP)
rAC Flow from facility A to C

Facilities A,B,C A B C

rCA
xA1 xC3 Binary decision variable,
=1 if facility C is assigned to
d13 location 3
=0 otherwise

Locations 1,2,3 1 2 3
Distance from location 3 to 1
d31

( rAC d13 + rCA d31 ) xA1 xC3


Terms in objective function:
(rABd12 + rBAd21)xA1xB2 + (rABd13 + rBAd31)xA1xB3 +
(rABd23 + rBAd32)xA2xB3 +
(rACd12 + rCAd21)xA1xC2 + (rACd13 + rCAd31)xA1xC3 +
(rACd23 + rCAd32)xA2xC3 +
(rBCd12 + rCBd21)xB1xC2 + (rBCd13 + rCBd31)xB1xC3 +
(rBCd23 + rCBd32)xB2xC3
(rABd12 + rBAd21)xA2xB1 + (rABd13 + rBAd31)xA3xB1 +
(rABd23 + rBAd32)xA3xB2 +
(rACd12 + rCAd21)xA2xC1 + (rACd13 + rCAd31)xA3xC1 +
(rACd23 + rCAd32)xA3xC2 +
(rBCd12 + rCBd21)xB2xC1 + (rBCd13 + rCBd31)xB3xC1 +
(rBCd23 + rCBd32)xB3xC2
Exercise
Consider the data given in the tables, and write the three
of the objective function terms for the quadratic
assignment problem

Flows (trips) Distances

From A B C From 1 2 3
/To /To
A 0 3 8 1 0 7 10

B 2 0 4 2 8 0 4

C 8 4 0 3 10 5 0
Exercise
Flows (trips) Distances
From A B C From 1 2 3
/To /To
A 0 3 8 1 0 7 10

B 2 0 4 2 8 0 4

C 8 4 0 3 10 5 0

Objective Function
(37) xA1xB2 + (50)xA1xB3 + (22)xA2xB3 …
Constraints
Subject to:

xA1 + xA2 + xA3 = 1, for facility A


xB1 + xB2 + xB3 = 1, for facility B
xC1 + xC2 + xC3 = 1, for facility C
xA1 + xB1 + xC1 = 1, for location 1
xA2 + xB2 + xC2 = 1, for location 2
xA3 + xB3 + xC3 = 1, for location 3
xij = 0, 1, for i = A,B,C & j = 1,2,3
Exact Solution
1 B
2 C
3 A
Objective Function Z = 182
QAP Model
n n n n
Min z   rij d rs xir x js
i 1 j 1 r 1 s 1
Subject to:
n

x
r 1
ir  1, i  1,  , n
n

x
i 1
ir  1, r  1,  , n

xir  {0,1}, i  1,  , n; r  1,  , n

You might also like