Automated Engineering Change Report Generation for Ships Based on GraphRAG

Jun Yin, Jingyu Sun

Abstract: With the deepening of the customization level of ship products, engineering change management has become increasingly important. However, existing digital technologies mainly focus on the optimization of the engineering change management process, and there is relatively little research on the subsequent generation of change reports. In particular, when faced with a large number of scattered change orders, it is becoming more and more difficult to write reports manually. Therefore, this study proposes an automated engineering change report generation framework based on Graph-based Retrieval-Augmented Generation (Graph RAG), which combines knowledge graphs with large language models. The construction of the knowledge graph is enhanced through GraphRAG to analyze the complex relationships in engineering changes accurately. Using the enhanced knowledge graph as a knowledge base, high-quality reports are generated with the help of large language models, and multi-dimensional evaluation is adopted to ensure the report quality. The experimental results based on the change orders of general cargo ships in Shipyard Z show that the reports generated by this framework are superior to traditional structured data methods in terms of integrity, information extraction, and semantic coherence.

External IDs:dblp:conf/whiceb/YinS25