The ship engine room is a complex design space requiring the simultaneous satisfaction of various constraints, including pipe connectivity, maintenance accessibility, and spatial utilization. Despite its complexity, engine room equipment arrangement still heavily relies on expert experience and manual operations, often leading to inconsistent design quality. This study proposes a systematic optimal arrangement design framework that integrates Case-Based Reasoning (CBR) with a multi-objective optimization approach. To quantify and reuse the implicit knowledge embedded in existing designs, we represent engine room configurations as system-wise graphs and employ Graph Edit Distance (GED) to calculate structural similarity between target and past designs. This similarity is utilized as a key objective function within the NSGA-II optimization algorithm, alongside objectives for minimizing pipe length and maximizing spatial efficiency. Furthermore, we introduce an automated passageway generation method using Generalized Voronoi Diagrams (GVD) and skeletonization to ensure design feasibility and accessibility. The framework was validated through three case studies, where the results demonstrated that the proposed method effectively preserves realistic arrangement patterns while significantly improving spatial utilization and accessibility compared to conventional optimization techniques. This approach provides a robust decision-making tool for designers, reducing dependency on individual expertise and enhancing overall design consistency. • A new engine room arrangement framework integrates case-based reasoning with multi-objective optimization. • Graph modeling and graph edit distance quantify structural similarity between target and past designs. • A similarity-driven objective function balances realistic arrangement patterns with spatial efficiency. • Generalized Voronoi diagrams and skeletonization automatically generate feasible passageway candidates. • Case studies show enhanced consistency, spatial utilization, and accessibility compared to conventional methods.
Kim et al. (Wed,) studied this question.