Effects of Distance Metrics on Unequal-Area Facility Layouts under Predefined Aisle Constraints: An Experimental Study

บทคัดย่อ

          Objectives: Unequal-area facility layout problems under predefined aisle constraints are common in real manufacturing systems; however, most existing studies assume equal-area departments or unconstrained travel paths. This creates a research gap regarding how distance metrics affect layout quality when department sizes vary, and travel must follow actual aisles. Therefore, this study aimed (1) to examine the effects of search distance metrics (Rectilinear, Pathway, and Euclidean) and levels of department area heterogeneity (Space) on layout quality when performance is evaluated by total travel distance along actual aisles (Pathway distance), and (2) to compare the best layouts obtained from each metric and Space level through cross-evaluation to derive practical guidelines for selecting distance metrics in facility layout problems with predefined aisle constraints.

          Methods: Four benchmark quadratic assignment problems (Nug12, Nug16a, Nug20, and Nug24) were modified to represent unequal-area layouts using three Space levels (20–40, 20–60, and 20–80 m²). For each problem–Space combination, facility layouts were generated using a Genetic Algorithm under three search distance metrics, with 30 independent runs per setting. Layout performance was evaluated using the total material-handling distance calculated by the Pathway distance metric. The results were analyzed using a two-way analysis of variance with factors Search metric and Space, followed by post-hoc comparisons. The best layout from each search metric was further examined through cross-evaluation using all three distance metrics.

          Results: Space had the largest effect on Pathway distance, with very large effect sizes (η² ≈ 0.85–0.93), whereas the Search metric showed medium-sized effects, and the interaction was relatively small. When evaluated by Pathway distance, layouts obtained from Pathway search had the lowest mean total distances across all problem–space combinations, although statistically significant advantages over Rectilinear and Euclidean search were observed only in some instances. In many cases, the geometric search metrics yielded layouts whose Pathway performance was statistically indistinguishable from or only slightly worse than that of Pathway search. Overall, the results suggest using Pathway distance as the final evaluation metric, while employing Rectilinear or Euclidean distance in the search phase for simplicity and computational efficiency, with candidate layouts re-evaluated using Pathway distance before implementation.