Optimizing dynamic facility layout problems: genetic algorithm with local search integration

• Published in: International Journal of Advanced Technology and Engineering Exploration Vol. 11; no. 115; p. 899
• Main Authors: Vineetha, G R, Shiyas, C R
• Format: Journal Article
• Language: English
• Published: Bhopal Accent Social and Welfare Society 30.06.2024
• Subjects: Combinatorial analysis; Convergence; Deviation; Effectiveness; Facilities management; Genetic algorithms; Layouts; Optimization; Optimization techniques; Performance evaluation; Searching; State-of-the-art reviews
• ISSN: 2394-5443; 2394-7454
• DOI: 10.19101/IJATEE.2023.10102424

The dynamic facility layout problem (DFLP) is one of the most complex combinatorial optimization challenges. Given that obtaining optimal solutions using exact methods requires substantial time and computational power, researchers often turn to nonconventional optimization techniques to achieve near-optimal solutions. This paper presents a genetic algorithm (GA) enhanced with a local search (LS) procedure for solving DFLPs. The algorithm employs roulette wheel selection (RWS), single-point crossover (SPC), and swap mutation (SM) as its genetic operators, with the 2opt neighborhood search serving as the LS operator. The termination criterion (TC) used in the proposed algorithm is the maximum number of generations. An extensive evaluation of the algorithm's performance was conducted in this research. It was tested on a diverse set of 48 problem instances, representing various problem sizes. To assess the effectiveness of the algorithm, the results produced were compared with those documented in existing literature and benchmarked against the best-known solutions previously reported. This rigorous comparison allows for an evaluation of the algorithm's performance relative to other established methods and state-of-the-art solutions available in the field. Through extensive experimentation on 48 test instances, the algorithm consistently delivers competitive results, achieving solutions within a margin of less than four percent deviation from the best-known solutions across all instances, with an average deviation ranging from 0% to 3.71%. Although the average runtime of the algorithm is provided, its comparison with existing literature is deemed irrelevant due to significant variations in machine configurations. This work introduces a hybrid genetic algorithm (hGA) specifically designed for solving DFLPs. By integrating fundamental genetic operations with a localized search approach, the proposed hGA demonstrates promising capabilities in tackling this complex optimization problem. The outcomes affirm the efficacy of the hGA in swiftly converging to near-optimal solutions for DFLPs, underscoring its potential for practical applications.