Researchers have developed an artificial intelligence system designed to improve structural damage assessments in historic buildings, combining specialized AI agents with sensor measurements, images and retrieved engineering knowledge.
Published in Scientific Reports, the study introduces a retrieval-augmented framework that coordinates language models and vision-language models to assess structural anomalies.
Tested on a historic masonry minaret, the system analyzes sensor relationships, examines visual evidence and combines findings to identify areas that may require closer inspection.
Its retrieval mechanism introduces relevant structural information into the assessment process, helping reduce potentially misleading interpretations from ambiguous images.
The research extends the growing use of AI in architectural design and structural understanding, while offering a different application for technologies also being explored in cultural heritage preservation.
The researchers emphasize that the framework is a proof of concept involving one structure. Its assessments indicate inspection priorities rather than independently verified damage, and wider field validation remains necessary.