An Enhanced Artificial Immune System Approach For Assembly Line Balancing Problem Through Shifting Bottleneck Identification
Loading...
Date
2018-09
Authors
Mohd Nor Akmal Khalid
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
The manufacturing industry has evolved rapidly in the past few years, due to the
global competitive economy, high-quality market demands, and customized products
with the lowest possible costs. This is achieved by partitioning the workloads among
the available resource to obtain an equal amount of workloads in the assembly line
system, which defines the assembly line balancing (ALB) problem. The most
prominent ALB problem is the simple assembly line balancing (SALB) problem
which has been utilized for decades to provide a basis for testing different approaches.
Despite varieties of computational techniques have addressed the ALB problem,
which can be categorized as exact, heuristic, and meta-heuristic approaches, little
work had been done on SALB-E problem due to its difficulty of obtaining the optimal
solutions. Additionally, bottlenecks can still occur during the assembly operations
that affect the production quality and induce unnecessary costs. Identifying and
optimizing machines with the likelihood of the next operation bottleneck had been
rarely addressed in the assembly line especially when it shifts from one machine to
another (called shifting bottleneck). This study propose an effective computational
approach to address the SALB-E problem through the shifting bottleneck
identification. A bio-inspired approach had been frequently adopted for handling
complex and combinatorial optimization problem through a simple yet effective
manner. As such, a computational method, known as artificial immune system (AIS)
approach, had been proposed. Three variants of the AIS approaches were proposed, namely the contagious AIS (CAST), contagious AIS with discrete bottleneck
simulator (CASTOR) and contagious immune network with bottleneck indicator
matrix (COMET). These three approaches were tested on 24 real-world SALB-E data
sets with 242 instances. The experimental results showed that the proposed CAST,
CASTOR, and COMET approaches have solved up to 34.30%, 66.12%, and 100%
instances of the data sets, respectively. Additionally, comparative study against
approaches from the literature was conducted where statistically significant results up
to 99.5% confidence interval (p < 0:00001) were deduced. This study concludes that
problem representation, complexity reduction, and utilizing the shifting bottleneck
helps to guide solution improvement. By addressing the identified shifting bottleneck,
the computational complexity is mitigated by utilizing the problem-specific
information when the proposed approaches are faced with difficult SALB-E
problems.
Description
Keywords
Computer Science