Investigating The Performance Of Metaheuristics To Optimize The Revenue Of Semiconductor Supply Chain
Loading...
Date
2022-07-23
Authors
Roslee, Muhammad Sharifuddin
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Investigating the performance of metaheuristics to optimise the revenue of the
semiconductor supply chain is the final year project that focuses on to identify the
supply chain shortage problem related to semiconductors and to compare among the
20 metaheuristics algorithm optimization programmes capable of optimising the
semiconductor supply chain. Thorough planning is required at many levels, from the
initial layout design through the building of the facility itself. It is imperative that the
correct planning decisions are made from the beginning and that the operating policies
in existing and proposed factories maximise the product output without sacrificing
product quality or factory reliability if the increasing consumer demands for greater
product complexity and diversity at lower cost are to be met profitably.
On a typical wafer, there are between 300 and 500 manufacturing steps, and the
product cycle time is frequently more than a month. In addition, machines are divided
into serial and batch types based on the quantity of items being processed at the same
time. Wafer lots would have to wait longer for batch operations due to batch size
transformations. It is difficult to analyse the production planning with all of these
aspects in consideration. As the rivalry expands, semiconductor manufacturers must
be nimble and flexible in order to stay in business. Changes in the product mix make
the system even more complicated. An enormous number of distinct products must be
processed by machines in a semiconductor manufacturing facility (fab), leading to
resource sharing issues and the potential for a lengthy line. Due to their simplicity, flexibility, vast number of accessible methods, and ability to avoid local optimums throughout the computing process, metaheuristics are becoming increasingly popular for addressing
optimization issues. The algorithms used to tackle optimization issues in several scientific and technological sectors do not require major modifications to be successful. Optimisation is possible for every objective function, regardless of whether or not it is continuous or differentiable. Due to the random nature of metaheuristic approaches, there is a high probability that global optimal
solutions will be discovered. A predetermined optimization goal can be achieved by quantifying the quality
measures, also known as fitness functions, objective functions, or goodness levels.
These measurements aim to maximise desired qualities while decreasing undesirable
ones. Small fitness values are often indicative of close-to-optimal control-variable
values, as shown in this case. Below is the comparison of the fitness value among 20
metaheuristics algorithms run through Python programming.