Solving the Facility Layout Problem Using Genetic and Progressive Algorithms

In modern markets, manufacturers and service systems are faced with the need to constantly respond to changing product mixes, consumer’s tastes and demands. Facility planning is a significant strategy used to achieve this goal. A well-planned facility minimizes the materials and personnel flow distances, inventory build-up and consequently improves the production lead time and operation cost. Therefore, this thesis explores two variations of the facility layout problem: the Equal Areas dynamic facility layout problem (EA-DFLP) and the unequal area facility layout problem (UA-FLP). The EA-DFLP involves arranging and rearranging a number of equal departments in a facility such that the sum of material handling and rearrangement cost is minimized. The UA-FLP involves finding the best arrangement for a number of departments that minimizes the material handling cost without violating the shape constraints. In this thesis, a genetic algorithm is proposed to solve the EA-DFLP. The genetic algorithm is a direct implementation with the addition of an elitism criteria. This thesis also presents a new modeling approach called Progr essive Modeling (PM) and applies it to solve the Unequal Areas Facility Layout Problem (UA-FLP). A problem solution is represented by an object-oriented binary tree and an objective graph. PM introduces a component model to deploy the problem logic and its solution algorithm over several interacting components. Component models isolate the objective space from the search space in a black-box fashion. A novel solution algorithm is presented to demonstrate how the search process is managed and controlled while searching for optimal or near-optimal solutions. The solution process and the optimization algorithm are demonstrated using the developed software framework. The Genetic Algorithm and the Progressive Modeling were both tested using well-known problems from the literature. The Genetic Algorithm was applied to the EA-DFLP while the Progressive Modeling was used to solve the UA-FLP. The results using the genetic algorithm showed a better performance when compared to other solutions from literature. The genetic algorithm generated a new best solution (0.8% improvement) for one of the test problems and it obtained the best-found results for most of the rest. For small sized UA-FLP, all optimum or best-known solutions have been obtained. The PM generated the best solutions for the midsize problems (problems with 10, 12 and 14 departments). For eight values of aspect ratios tested for the large size problem, PM generated seven new best solution out of eight problems. The developed problem analysis, solution algorithm, and results demonstrate why the proposed modeling approach is promising and capable of handling more complex real-world problems. The comparison between progressive algorithms and genetic algorithms will be left for future research.

Name thesis:degree_name

Master of Applied Science (MASc)

Level thesis:degree_level

Master's

Discipline thesis:degree_discipline

Engineering - Industrial Systems

Grantor dc:publisher

Faculty of Graduate Studies and Research, University of Regina

Year dc:date.issued

2017

Author dc:creator

• Fowosere, Sodiq Olumide

Advisor dc:contributor.advisor

• Ismail, Mohamed

Committee members dc:contributor.committeemember

• Mehrandezh, Mehran
• Stilling, Denise

Language dc:language.iso

en

DOI

https://doi.org/10.82465/3885

OAI identifier oai:identifier

oai:uregina.scholaris.ca:10294/8463