| Abstract | Within the engine room, a complex network of various equipment and pipes can be found. These components are installed and manufactured using a modular approach. Specifically, the equipment unit module comprises the equipment itself and the piping surrounding its inlet and outlet. When arranging these unit modules, engineers must consider the equipment's installation location and the interconnecting routes. Notably, a conventional piping pattern, based on expert knowledge and guidelines, is typically employed around the equipment's inlet and outlet. However, this design process heavily relies on an engineer's experience and needs more quantitative evaluation. As a result, non-experts find it challenging to understand the characteristics and design the arrangement of the equipment unit module. Moreover, due to the complexity of these patterns, the design review process requires substantial resources. To address these challenges, we propose a method for analyzing piping patterns and automatically generating optimal patterns for equipment unit modules. To identify the patterns, we preprocess the intricately interconnected piping group. Piping can be represented as connection lines since they serve the purpose of connecting equipment. By simplifying these connections, a topological approach based on equipment can be employed, enabling the analysis of overall patterns and the understanding of equipment relationships. In this study, we compared the results of each analysis using traditional techniques such as the Dijkstra and A* algorithms and the latest technique, a graph neural network (GNN) model based on deep learning. By applying the method that yielded the best results, we could develop a model capable of recommending optimal patterns. Finally, we applied the proposed method to analyze patterns and generate optimal patterns for major equipment unit modules within the engine room, verifying its effectiveness. |
|---|---|
| Publication Date | 2023-10-19 |