Sie sind auf Seite 1von 65

Layout and Location

Problems

References:
1. Richard L. Francis, Leon F. McGinnis, Jr., and John A. White, Facility Layout
and Location: An Analytical Approach, 1992, Prentice Hall, NJ. (2nd edition)
2. Robert F. Love, James G. Morris, and George O. Wesolowsky, Facility
Location: Models & Methods, 1988, North-Holland,NY.
3. Andrew Kusiak, Computational Intelligence in Design and Manufacturing,
2000, John Wiley & Sons, Inc., NY.

Special case of multiple row

It can be considered as a case of


linear single row

Single-row problem It can be transformed into a linear program.

vk = [ 2 1 4 6 3 5 ]

How to decode chromosome vk = [ 2 1 4 6 3 5 ] ?

x2 = l2 = (2.0) = 1
x1 = x2 + (l2 + l1) + d12 = 0 + (2 +5) + 1 = 4.5

ok

x4 = x1 + (l1 + l4) + d14 + max { 0, d24 [x1 + (l1 + l4) + d14 x2 ] }

= 4.5 + (5 +6) + 2 + max { 0, 1 (4.5 + (5 +6) + 2 1) }


= 4.5 + 5.5 + 2 = 12
9
...

Ordinates yi are the same for all facility.

Roulette wheel selection + Elitist strategy

10

11

12

or add | yi y j | l0 d

( d : width required by AVG. d 0)

13

We may have different layout assumptions.


Row 4

3l0

Row 3

2l0

Row 2

l0

Row 1

Layout A
14

We may have different layout assumptions.


Row 4

3l0

Row 3

2l0

Row 2

l0

Row 1

Layout B

15

or add constraint | yi y j | l0 d
rectilinear
distance

( d : width required by AVG.)

1
(1 zik z jk ) M | xi x j | zik z jk (li l j ) d ij
2

?
add row-row
clearance
constraint
at least 2 rows

16

( zik = 1 if machine i is allocated to row k. )

17

[ {s1, s2, . . . , sk-1}, { mi1, mi2, . . . , min }, {i1, i2, . . . , in } ]


s1 < s2 < . . . < sk-1 , k < n (multiple row problems)

18

2- row problem

19

20

21

22

It may be profitable to have nonzero neat clearance.

Row-row clearance
constraint must be checked.

23

24

25

2l0
l0
L

26

27

28

29

Chromosome representation:
Real number encoding
Arithmetic Crossover: Randomly choose ( 0, 1)

v1 and v2

( parents )

v1 ' = v1 + (1 - ) v2

( offspring 1 )

v2 ' = v2 + (1 - ) v1

( offspring 2 )

30

31

crossover

s1, s2

s [ sL , sU ]

32

33

i 2 i
r
r

2i

ri
r

2ri
r

Select the best offspring.

34

35

36

between different rows

37

38

39

l0 = 8

L = 22

40

41

Local search may improve the performance of the algorithm.


42

( p median problem)

43

NP-hard problem

or qj zij

Euclidean distance

Each customer should be served by only one facility.

44

Two levels:
location
allocation

45

46

For chromosome,

solve a heuristic solution for the following problem by using


Lagrangian Multiplier Method.
m n

min cij zij


i
1 j 1

cij q j ( xik x j ) 2 ( yik y j ) 2

47

48

49

50

(+ ) selection: select best out of offsprings and old parents (+ ) as parents


of the next generation. (Bck and Hoffmeister)
51

52

Alternative location-allocation method

53

54

Obstacle

Obstacle

Obstacle
Obstacle

Figure 10.17 Obstacle location-allocation problem

55

56

57

58

59

60

61

62

63

64

65

Das könnte Ihnen auch gefallen