© C.Hicks, University of Newcastle IGLS02/1 A Genetic Algorithm Tool for Designing Manufacturing Facilities in the Capital Goods Industry Dr Christian Hicks, University of Newcastle, England Email: [email protected]
Dec 19, 2015
© C.Hicks, University of Newcastle
IGLS02/1
A Genetic Algorithm Tool for Designing Manufacturing
Facilities in the Capital Goods Industry
Dr Christian Hicks,
University of Newcastle,
England
Email: [email protected]
© C.Hicks, University of Newcastle
IGLS02/2
Types of Facilities Design Problems
• Green field – designer free to select processes, machines, transport, layout, building and infrastructure
• Brown field – existing situation imposes many constraints
© C.Hicks, University of Newcastle
IGLS02/3
Facilities Layout Problem
Includes:• Job assignment – selection of
machines for each operation and definition of operation sequences
• Cell formation – assignment of machine tools and product families to cells
• Layout design – geometric design of manufacturing facilities and the location of resources
• Transportation system design
This paper considers cell formation and layout design
© C.Hicks, University of Newcastle
IGLS02/4
Cell Formation Methods
• “Eyeballing”
• Coding and classification• Product Flow Analysis• Machine-part incidence matrix
methods– Rank Order Clustering– Close Neighbour Algorithm
• Agglomerative clustering– Various similarity coefficients– Alternative clustering strategies
© C.Hicks, University of Newcastle
IGLS02/5
Rank Order Clustering Applied to data Obtained from a capital goods company
© C.Hicks, University of Newcastle
IGLS02/6
Similarity Coefficient
1
654
32
S ij = m ax(n ij/n i, n ij/n j)
S2,5 = m ax(2 /3 , 2 /2)
S2,5 = 1
© C.Hicks, University of Newcastle
IGLS02/7
Agglomerative clustering using the singlelinkage strategyEquation 1
© C.Hicks, University of Newcastle
IGLS02/9
Clustering applied to capital goods companies
Limitations• Few natural machine-part clusters• Long and complex routings mitigate
against self contained cells• Clustering only uses routing
information• Geometric information is not used.
© C.Hicks, University of Newcastle
IGLS02/10
Genetic Algorithm Design Tool
Based upon: • Manufacturing System Simulation
Model (Hicks 1998) • GA scheduling tool (Pongcharoen et
al. 2000)
© C.Hicks, University of Newcastle
IGLS02/11
Manufacturing Planing &Control System
Manufacturing Facility
Manufacturing System Simulation Model
Planned Schedule
Resourceinformation
CAPM modules used
System parameters
Product information
Operational factors
System dynamics Logic
Measures ofperformance
Flow measurementCluster AnalysisLayout generation methods
Tools
© C.Hicks, University of Newcastle
IGLS02/12
GA Procedure
• Use GAs to create sequences of machines
• Apply a placement algorithm to generate layout.
• Measure total direct or rectilinear distance to evaluate the layout.
© C.Hicks, University of Newcastle
IGLS02/13
Genetic Algorithm
Similar to Pongcharoen et al except, the repair process is different and it is implemented in Pascal
© C.Hicks, University of Newcastle
IGLS02/15
Case Study
• 52 Machine tools• 3408 complex components• 734 part types• Complex product structures• Total distance travelled
– Direct distance 232Km
– Rectilinear distance 642Km
© C.Hicks, University of Newcastle
IGLS02/17
Total Rectilinear Distance vs Generation
0
100000
200000
300000
400000
500000
600000
700000
800000
1 11 21 31 41 51 61 71 81 91 101
111
121
131
141
151
161
171
181
191Generation
Tota
l Rec
tilin
ear
Dis
tan
ce (
m)
Minimum
Average
Population size 200Generations 200Crossover 90%Mutation 18%
Total rectilinear distance travelled vs. generation (brown field)
© C.Hicks, University of Newcastle
IGLS02/19
Total rectilinear distance vs. generation (green field)
Total rectilinear distance travelled vs. generation(green field problem)
0
100000
200000
300000
400000
500000
600000
700000
800000
1
11 21
31
41
51
61
71
81
91
10
1
111
12
1
13
1
14
1
15
1
16
1
17
1
18
1
19
1
Generation
To
tal
rec
tili
ne
ar
dis
tan
ce
(m
)
Average
Minimum
Note the rapid convergence with lower totals than for the brown field problem
© C.Hicks, University of Newcastle
IGLS02/20
Resultant layout (green field)
Note that brown field constraints, such as wallsHave been ignored.
© C.Hicks, University of Newcastle
IGLS02/21
Conclusions
• Significant body of research relating to facilities layout, particularly for job and flow shops.
• Much research related to small problems.
• Capital goods companies very complex due to complex routings and subsequent assembly requirements.
• Clustering methods are generally inconclusive when applied to capital goods companies.
• GA tool shows an improvement of 70% in the green field case and 30% in the brown field case.
© C.Hicks, University of Newcastle
IGLS02/22
Future Work
• The GA layout generation tool is embedded within a large sophisticated simulation model.
• Dynamic layout evaluation criteria can be used.
• The integration with a GA scheduling tool provides a mechanism for simultaneously “optimising” layout and schedules.