The master’s thesis examines the use of AI agents in Building Information Modeling (BIM) processes, with a focus on the automated detection of changes between different versions of project documentation. The research investigates the extent to which an AI agent can automate the comparison of different types of project documentation and reliably alert the user to detected changes. For this purpose, we developed an automated workflow that integrates the Dalux Common Data Environment (CDE), the n8n platform, and a large language model (LLM), enabling the retrieval, preparation, comparison, and reporting of document changes. The system was tested on textual and tabular PDF files, PDF drawings, Excel files, DWG drawings, and IFC files, and its performance was evaluated using classification accuracy, precision, recall, and the F1 score. The results show that the AI agent performed best on structured, textual, and tabular data, while its performance was lower on graphically complex drawings due to a higher number of missed changes. For IFC files, despite the successful detection of structured changes, limitations were identified in the spatial interpretation of these changes. The findings indicate that an AI agent can take over part of the repetitive review of project documentation and facilitate the identification of changes for the user; however, due to the identified limitations, expert review of the results remains important. Further development should focus primarily on improving the analysis of graphically complex drawings, the spatial interpretation of changes in IFC files, and the preparation of input data.
|
