Penerapan Multi-Objective Adaptive Large Neighborhood Search (MOALNS) untuk Optimasi Double Row Layout Problem dengan Mempertimbangkan Safety Distance dan Geometrical Constraint
Belinda Tirza Amalia, Ir. Achmad Pratama Rifai, S.T, M.Eng, Ph.D.,IPM
2026 | Skripsi | TEKNIK INDUSTRI
Production activities on manufacturing lines with a double row configuration face the challenge of determining the optimal layout while simultaneously considering various operational aspects. The Double Row Layout Problem (DRLP) in this study also takes into account the safety distance between machines and geometrical constraint in the form of differences in row length due to the presence of fixed-position facilities, in order to better represent real-world conditions in the industrial sector. Two conflicting objective functions necessitate the use of a multi-objective approach to generate a set of non-dominated solutions with varying levels of trade-off between the two objectives that cannot be resolved through a single-objective approach.
The Multi-Objective Adaptive Large Neighborhood Search (MOALNS) algorithm was designed as a metaheuristic solution for this problem. The algorithm consists of seven destroy operators and six repair operators selected via a roulette wheel mechanism, with a solution acceptance mechanism adapted from AMOSA. The MOALNS algorithm was run on 27 benchmark instances with ten repetitions per instance. Four evaluation metrics were used to measure the quality of the generated solutions: Hypervolume (HV), Euclidean Distance (ED), Sum of Objectives (SO), and Spacing.
The results of MOALNS were compared against four algorithms from the reference paper (AMOSA, NSGA-II, MOVNS, and IMOVNS). MOALNS outperforms the comparison algorithms in 26 out of 27 instances for the HV, SO, and Spacing metrics. For the ED metric, MOALNS outperforms the comparison algorithms in 23 out of 27 instances, as its performance is lower than the comparison algorithms in small-sized instances. The resulting Pareto front offers several layout alternatives with varying levels of compromise between cost efficiency and safety distance compliance.
Production activities on manufacturing lines with a double row configuration face the challenge of determining the optimal layout while simultaneously considering various operational aspects. The Double Row Layout Problem (DRLP) in this study also takes into account the safety distance between machines and geometrical constraint in the form of differences in row length due to the presence of fixed-position facilities, in order to better represent real-world conditions in the industrial sector. Two conflicting objective functions necessitate the use of a multi-objective approach to generate a set of non-dominated solutions with varying levels of trade-off between the two objectives that cannot be resolved through a single-objective approach.
The Multi-Objective Adaptive Large Neighborhood Search (MOALNS) algorithm was designed as a metaheuristic solution for this problem. The algorithm consists of seven destroy operators and six repair operators selected via a roulette wheel mechanism, with a solution acceptance mechanism adapted from AMOSA. The MOALNS algorithm was run on 27 benchmark instances with ten repetitions per instance. Four evaluation metrics were used to measure the quality of the generated solutions: Hypervolume (HV), Euclidean Distance (ED), Sum of Objectives (SO), and Spacing.
The results of MOALNS were compared against four algorithms from the reference paper (AMOSA, NSGA-II, MOVNS, and IMOVNS). MOALNS outperforms the comparison algorithms in 26 out of 27 instances for the HV, SO, and Spacing metrics. For the ED metric, MOALNS outperforms the comparison algorithms in 23 out of 27 instances, as its performance is lower than the comparison algorithms in small-sized instances. The resulting Pareto front offers several layout alternatives with varying levels of compromise between cost efficiency and safety distance compliance.
Kata Kunci : double row layout problem, multi-objective optimization, adaptive large neighborhood search, safety distance, geometrical constraint
