•  
  •  
 

CIB Conferences

Abstract

In the operational phase of buildings, Digital Twins (DT) is utilized to enhance maintenance quality and optimize energy efficiency. Building Information Modeling (BIM) serves not only as a visualization tool but as a semantic foundation within this framework, enabling dynamic IoT sensor data to be accurately mapped into physical elements. The process of integrating BIM models with IoT and DT for application development is lengthy. This governance dilemma arises from the gap between high-level organizational requirements and underlying technical implementation, or lack of clear, adaptable information requirement definitions and the inclusion of redundant data in BIM deliverables, which increases management costs. To address these weaknesses, this study proposes a conceptual framework based on the information management framework ISO 19650. The goal of this framework is to translate the " application purpose" into verifiable Organizational Information Requirements (OIR), Asset Information Requirements (AIR), and Level of Information Need (LOIN) by applying natural language processing technologies. Through a Model Context Protocol (MCP), a communication bridge is established between an AI Agent and BIM models. By linking Large Language Models (LLMs) to PyRevit scripts via MCP, the system automatically extracts the information from Revit models based on natural language instructions to support precise data mapping. The framework aims to automate the execution of minimum necessary information and integrate BIM with external data during DT construction. An AI agent interprets natural language instructions to determine the required information, reads semantic information from the building and systems model, and ultimately applies it in conjunction with data. This initiated theoretical framework aims to be adopted in a real case study from a residential community based in the United Kingdom, specifically by starting the tracking of indoor health and comfort as an example, to verify the practical feasibility of this framework.

Keywords

AI Agent, Model Context Protocol, Digital Twin, Information Requirements, BIM

Download

Share

COinS