ScholarWorks@성균관대학교

초록

Building digital twins (DTs) are essential for enhancing operational efficiency, optimizing energy consumption, and reducing costs in buildings. However, the inherent complexity of buildings, their long operational lifespans, and the specific nature of the construction industry pose significant challenges in creating digital twins for buildings. Intelligent digital twins (IDTs) address these challenges by integrating existing digital twin models with AI, enabling a comprehensive representation of the building lifecycle while incorporating expert input. This study proposes an AI agent-based IDT framework using an ontological approach, where AI agents are engineered by DT administrators with building operations and maintenance (O&M) data, information, and applications within an ontological DT environment. Data and information generated within this environment are expressed in the DT ontology, enabling AI agents to gain a holistic understanding of the target system. Applications are integrated as a tool, thereby enabling AI agents to expand their actions and gain additional information from results. To validate this framework, virtual in-situ modeling (VIM) and fault detection and diagnosis (FDD) algorithms were implemented as DT applications to demonstrate the operation of the IDT system. Four case studies were conducted to demonstrate IDT-enabled O&M services, and LangSmith was used to visualize the AI agents' reasoning process as part of the result validation. It shows that AI agents have capabilities of performing building O&M tasks with high-level reasoning. The significance of this study lies in demonstrating the feasibility of implementing IDT models in building O&M by enabling AI agents to provide comprehensive, domain-specific knowledge and perform operational tasks, thereby serving as an assistant for both users and operators. Finally, this study underscores the critical role of engineers in managing and maintaining ontology and applications within the DT environment. © 2025 Elsevier Ltd

키워드