Professional Documents
Culture Documents
DOI 10.1007/s00170-004-2161-0
ORIGINAL ARTICLE
Received: 28 November 2003 / Accepted: 1 March 2004 / Published online: 26 January 2005
Springer-Verlag London Limited 2005
Abstract In apparel manufacturing, experience and subjective
assessment of production planners are used quite often to plan
the production schedules in their fabric-cutting departments. The
quantities of cut-pieces produced by fabric-cutting departments
based on these non-systematic schedules cannot fulfil the cutpiece requirements of the downstream sewing lines and minimize the makespan. This paper proposes a genetic algorithms
(GAs) approach to optimize both the cut-piece requirements
and the makespan of the conventional fabric-cutting departments
using manual spreading and cutting methods. An optimization
model for the manual fabric cutting process based on GAs was
developed. Two sets of production data were collected to validate the performance of the model and the experimental results
were obtained. From the results, it can be found that both the
makespan and cut-piece fulfilment rates are improved in which
the latter is improved significantly.
Keywords Fabric-cutting Genetic algorithms
Production scheduling
Nomenclature
X
N
i
j
i , j
1 Introduction
In apparel manufacturing, fabric cutting is done before assembly.
The performance of the cutting department, which is generally
neglected by manufacturers, is a critical factor on the smoothness
of downstream operations in sewing lines and hence the overall
efficiency of the apparel manufacturing plant. Since the late 80s,
some apparel manufacturers have implemented the computerized
fabric-cutting systems in their apparel manufacturing process.
The demands on fabric-cutting departments for greater accuracy,
faster throughput, larger fabric and labour savings have driven
the adoption of computerized cutting systems.
However, many manufacturers still rely on the manual
method for the fabric-spreading and cutting operations in their
fabric-cutting department. Before daily spreading and cutting
operations start, the production planners of cutting departments
need to plan the production (spreading and cutting) schedule so
as to minimize the idle time of operatives and fulfil the fabric
cut-piece requirements from different sewing production lines.
The production planning is normally based on their experience
and subjective assessment which is not a systematic method
and an optimal schedule cannot be obtained. As a result, idle
times occur on the spreading and cutting operatives which in turn
increases the overall makespan of cutting departments. The cutpiece quantities produced cannot fulfil the different requirements
of each downstream sewing production line.
As most of the apparel manufacturers and researchers emphasize the importance of sewing process, research has been
done to improve the operation of sewing lines. However, the productivity of cutting departments, which plays a significant role
153
on the smooth flow of work in the sewing lines and thus the
whole manufacturing plant, is often neglected. Jones [1] investigated a table loading system to which a rough spread/cut ratio
was proposed by comparing the estimated total spreading time
to the estimated total cutting time so as to assist management in
handling balancing problems. The further the spread/cut ratio for
a specific spread was from the optimum ratio, the more dramatic
the effect of this spread would be on balancing the tables. However, the determination of operative assignments relied on the
experiences and subjective assessment of the production planners of cutting departments. Another limitation is that decisions
are made on an as-needed basis without planning.
Hui et al. [2] attempted to solve the problem of fabric
roll planning in spreading using genetic algorithms. The optimal combinations of the fabric roll sequences for each fabric
lay were derived in order to minimize fabric wastage during
spreading. Though the fabric roll sequence for each fabric lay
was studied, the spreading sequence for all fabric lays, which
has a great impact on the productivity of cutting departments
was not considered. A lack of proper planning and scheduling
causes much anguish among manufacturers when the supply of
cut pieces do not meet the requirements of production units.
Skyes and McGregor [3] proposed the use of object-oriented
technology to design a computer simulation model for the pinning and cutting processes in apparel manufacturing such that
the fabric-cutting manager of fabric-cutting departments could
estimate the effects of various resource allocation to meet production goals. Wong et al. [4] proposed a spreading and cutting sequencing model to minimize the idle time of a computerized fabric-cutting system. Studies of optimizing cut-piece
fulfilment, which refers to the quantities of cut-pieces required
by sewing lines, has been neglected. This is a critical factor
on the smoothness of downstream operations in sewing lines,
which affects the overall efficiency of the apparel plant and the
delivery date of the whole production order to the customers
substantially.
154
(1)
(2)
PO
CR
=1 =1
()
()
()
C /D w
(3)
where C() and D() are the respective completed and required number of garments of the th production order ( =
()
1, 2, . . ., PO) at the th checking run ( = 1, 2, . . ., CR), and w
is the corresponding cut-piece fulfilment weighting.
3.3 Initialization
The evolution procedure begins by randomly generating an initial population of integer strings (chromosome) in which each
such string represents a processing sequence, , of the job. Each
chromosome is processed according to the manual cutting job
placement mechanism described in Sect. 2, and thus obtains the
makespan time and cut-piece fulfilment rates which facilitate the
fitness evaluation using Eq. 1.
155
In roulette wheel selection, the chance of a parent being selected is directly proportional to its fitness. In the example shown
in Fig. 3, from a population of ten chromosomes with a set of fitness evaluations totalling 80, six individuals are selected by the
biased roulette wheel scheme, according to six random numbers
generated from the interval of 0 to 80.
3.6 Elitism
Since the biased roulette wheel selection processes are based on
the survival of the fittest and are in random nature, there is no
guarantee that some fit individuals will be selected. In order to
156
improve the selection mechanism, De Jong [8] therefore introduced elitism. Elitism is an addition to many selection methods
that forces the GAs to retain some of the best individuals in each
generation. This elitist strategy copies the best individuals of each
generation directly onto the succeeding generation. Such individuals might be lost if they are not selected to reproduce, or are
destroyed by crossover or mutation. The elitist strategy can increase the speed of the domination of populations using the best
individuals and provide an improvement of the GAs performance. Elitism is thus considered in this job sequencing problem.
Fig. 5. Layout of a fabric-cutting department consisting of four spreading
tables with examples of fabric lays being spread
3.7 Evolution
After the initialization, evolution is caused to occur in accordance with the standard genetic operations of crossover, mutation and selection. The evolutionary process is allowed to continue until no further increase is obtained in the finesse of the
fittest binary string or the pre-defined maximum number of generations is reached. Thus, the fittest string will result and the
optimal job sequencing can be determined. The operations of
GAs can be represented by a flow diagram as shown in Fig. 4.
4 Case studies
Two sets of real production data denoted as cases A and B, are
used to demonstrate this multiobjective job sequencing optimization problem. All the data listed in Table 1 was obtained from
the fabric-cutting department of a Hong Kong-based garment
manufacturing company in China. These two-day spreading production schedules were recorded in the fabric-cutting department
wT =
PO
2
w()
(4)
1
2
3
4
4
1
3
7
30 116 114 66
103 136 139 132
50 90 90 57
24 47 47 47
(A)-Job (X n )
Production order ()
Qty of garment ()
Marker length ()
Spreading time (s)
Cutting time (c)
25
9
2
68
17
24
(B)-Job (X n )
Production order ()
Qty of garment ()
Marker length ()
Spreading time (s)
Cutting time (c)
1
2
3
4
2
1
1
3
33 140 316 224
91 170 170 130
48 121 254 161
24 47 47 47
(B)-Job (X n )
Production order ()
Qty of garment ()
Marker length ()
Spreading time (s)
Cutting time (c)
26 27 28
6
5
7
33 300 94
91 158 132
48 233 77
24 47 47
25 26 27
5
3
1
14 78 58
137 105 170
23 59 57
47 47 47
5
6
7
8
9 10 11 12 13 14 15
6
8
8
8
8
7
5
9
2
8
6
15 224 224 224 300 118 300 10 200 300 98
89 130 130 130 130 172 158 106 175 130 169
30 161 161 161 209 104 233 20 170 209 87
24 47 47 47 47 47 47 47 47 47 47
16
2
14
87
29
24
29 30 31 32
2
7
6
3
33 14 140 146
91 137 170 140
48 23 121 113
24 47 47 47
33
9
8
81
23
24
34
3
2
73
16
47
35 36 37 38 39 40
9
1
8
4
3
6
3 78 224 53 228 316
72 105 130 171 148 170
18 59 161 96 172 254
24 47 47 24 47 47
5
2
3
72
18
24
9
2
14
87
29
24
10
2
5
81
20
24
11 12 13 14 15
3 10 11 10
3
21 94 118 94 116
73 132 172 132 136
34 77 104 77 90
24 47 47 47 47
6
7
2
2
8 316
81 170
23 254
24 47
28 29 30
2
6
6
42 300 300
91 130 130
60 209 209
24 47 47
31
1
42
91
60
24
8
2
4
85
19
24
32 33 34 35
2
4
3
1
2 228 224
6
68 148 130 101
17 172 161 21
24 47 47 24
36 37 38 39
1
3
1
2
14 300 104 10
87 158 171 106
29 233 174 20
24 47 24 47
17 18
9
6
4 200
85 175
19 170
24 47
23
4
13
93
28
24
24
8
21
73
34
24
41 42 43 44 45 46 47
6
6
4
2
2
3
4
14 58 104 98 316 94
6
87 170 171 169 170 132 101
29 57 174 87 254 77 21
24 47 24 47 47 47 24
48
9
5
81
20
24
16 17
1 10
13 146
93 140
28 113
24 47
19 20
2
6
42 52
91 171
60 53
24 47
21 22
6
2
42 140
91 170
60 121
24 47
18 19 20 21 22 23 24
1
1
3
1
2
3
3
33 30 300 200 140 224 224
91 103 158 175 170 130 130
48 50 233 170 121 161 161
24 24 47 47 47 47 47
40 41 42 43 44 45 46
1
2
4
1
1
5
2
15 98 114 98 53 66 200
89 169 139 169 171 132 175
30 87 90 87 96 57 170
24 47 47 47 24 47 47
47 48
4
2
2 52
73 171
16 53
47 47
157
Fig. 6. Job sequences adopted by the industrial practice for the case A
Fig. 8. Job sequences generated by the optimization model using GAs for
the case A
Fig. 7. Job sequences adopted by the industrial practice for the case B
Fig. 9. Job sequences generated by the optimization model using GAs for
the case B
and w()
= 1 for all production orders. The total production
orders are PO = 9 and PO = 6 for the cases A and B, respectively. The production targets at check-time 480 and 960 min are
listed in Table 2.
The makespan and cut-piece fulfilment rates are compared
in Table 3 for job sequences adopted by industrial practice and
those optimized by GAs. It can be found that with the use
of GAs, the makespan was improved slightly from 1209 to
1200 min for case A and from 1209 to 1203 min for case B. In
addition to the makespan improvement, the cut-piece fulfilment
1
97
2
422
3
292
4
103
5
300
6
484
7
146
8
759
9
16
194
843
584
206
600
968
292 1517
32
(B) PO
D(1)
1
561
2
463
3
856
4
339
5
99
6
300
927 1711
678
198
600
(2)
D(2)
1122
Table 3. Comparison on makespan time and cut-piece fulfilment rates of job sequences adopted by industrial practice and those generated by GAs
(A) PO
(1)
(1)
C /D Ind.
C(2) /D(2) Ind.
C(1) /D(1) GA
7
100%
8
0%
9
100%
Avg.
64%
100%
100%
100%
49%
100%
82%
1
0%
2
100%
3
67%
100%
71%
50%
67%
makespan
100%
47%
100%
100%
100%
55%
100%
69%
100%
86%
C(2) /D(1) GA
100%
89%
84%
74%
100%
70%
100%
65%
84%
85%
(B) PO
C(1) /D(1) Ind.
1
100%
2
100%
3
0%
6
0%
Avg.
61%
72%
74%
55%
100%
100%
50%
75%
C /D GA
97%
100%
35%
100%
100%
100%
89%
GA
59%
92%
73%
100%
100%
100%
87%
(1)
(1)
C(2) /D(1)
1209 min
1200 min
makespan
1209 min
1203 min
158
5 Conclusions
This paper proposes a genetic algorithms approach to solve the
optimization problem of the manual fabric-cutting process for
apparel manufacturing. Experiments were conducted to validate
the performance of the proposed method. The experimental results have indicated that the production (spreading and cutting)
schedules generated by GAs can improve both the cut-piece fulfilment rate and the makespan using the same number of operatives. The results also indicated that the shop-floor management
in the industry is capable of generating a production schedule
with a short makespan since slight improvement can only be obtained by using GAs approach. However, their schedule with a
short makespan cannot guarantee that the cut-pieces provided
by the cutting department can fulfil the requirements of sewing
lines. The proposed GAs approach has been proven as an effective method to increase the cut-piece fulfilment rates significantly
which directly affects the smoothness of downstream apparel assembly processes and thus ultimate delivery time of apparel to
References
1. Jones C (1993) Table planning takes a turnabout. Bobbin 35(7):4648
2. Hui PCL, Ng FSF, Chan KCC (2000) A study of the roll planning of
fabric spreading using genetic algorithm. Int J Clothing Sci Technol
12(1):5062
3. Sykes DA, McGregor JD (1991) Object-oriented modelling in the apparel industry. Proc Fifth International Conference of Technology of
Object-oriented Languages and System Tools, pp 219229
4. Wong WK, Chan CK, Ip WH (2000) Optimization of spreading and
cutting sequencing model in the garment manufacturing. Comput Ind
43:110
5. Holland JH (1975) Adaptation in natural and artificial systems. MIT
Press, Cambridge, MA
6. Bck T, Fogel DB, Michalewicz Z (1997) Handbook of evolutionary
Computation. Institute of Physics Publishing and Oxford University
Press, Oxford
7. Goldberg DE (1989) Genetic algorithms in search, optimisation, and machine Learning. Addison-Wesley, Boston, MA
8. De Jong KA (1975) An analysis of the behavior of a class of genetic
adaptive systems. Dissertation, University of Michigan, Ann Arbor, MI