Construction of a Knowledge-Driven Federated Database Model for Technical Knowledge Transfer in Port and Harbor Engineering

Abstract

In the port and harbor engineering field, many tasks depend heavily on the experience of skilled engineers, making technical knowledge transfer a critical challenge. This study proposes a Knowledge-Driven Federated Database (K-D FDB) type AI reference model for supporting technical knowledge transfer, built upon a Retrieval-Augmented Generation (RAG) framework that allows a Large Language Model (LLM) to reference structured external knowledge. In the proposed model, the RAG_DB (knowledge base) systematically organizes technical knowledge, design rules, and construction procedures. Furthermore, a Relational Database (RDB) and a knowledge graph are integrated through Excel, thereby linking drawing generation, quantity estimation, and construction scheduling. A case study targeting port revetment construction confirmed the operation of automatic cross-section CAD generation, automatic estimation of work quantities by work type, and schedule optimization. Through this model, the AI retains the tacit knowledge of skilled engineers as structured knowledge, and a mechanism is established in which junior engineers can acquire such knowledge via the AI, thereby enabling technical succession. The proposed framework is expected to serve as a foundational platform for advancing digital transformation (DX) across the entire port and harbor industry in the future.