An improved ant system algorithm for unequal area facility layout problems

To date, a formal Ant Colony Optimization (ACO) based metaheuristic has not been applied for solving Unequal Area Facility Layout Problems (UA-FLPs). This study proposes an Ant System (AS) algorithm for solving UA-FLPs using the Flexible Bay Structure (FBS) representation. In addition, this study pr...

Full description

Saved in:
Bibliographic Details
Main Author: Komarudin, Komarudin
Format: Thesis
Language:English
Published: 2009
Subjects:
Online Access:http://eprints.utm.my/id/eprint/11342/6/KomarudinMFKM2009.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To date, a formal Ant Colony Optimization (ACO) based metaheuristic has not been applied for solving Unequal Area Facility Layout Problems (UA-FLPs). This study proposes an Ant System (AS) algorithm for solving UA-FLPs using the Flexible Bay Structure (FBS) representation. In addition, this study proposes an improvement to the FBS representation when solving problems which have empty spaces. The proposed algorithm uses several types of local search to improve its search performance. It was extensively tested using 20 well-known problem instances taken from the literature. The proposed algorithm is effective and can produce all of the best FBS solutions (or even better) except for 2 problem instances. In addition, it can improve the best-known solution for 7 problem instances. The improvement gained by the proposed algorithm is up to 21.36% compared to previous research. Evidently, the proposed algorithm is also proven to be effective when solving large problem sets with 20, 25, and 30 departments. Furthermore, this study has implemented a Fuzzy Logic Controller (FLC) to automate the tuning of the AS algorithm. The experiments involved tuning four parameters individually, i.e. number of ants, pheromone information parameter, heuristic information parameter, and evaporation rate, as well as tuning all of them at once. The results showed that FLC could be used to replace manual parameter tuning which is time consuming. The results also showed that instead of using static parameter values, FLC has the potential to help the AS algorithm to achieve better objective function values.