Automated Checking of Highway Bridge BIM Models Based on Large Language Models

Building Information Modeling (BIM) is widely used in highway bridge engineering, making compliance with complex design specifications crucial. Existing checks rely heavily on manual review, which is time-consuming and inefficient. This study proposes an automated framework using large language models (LLMs) to parse unstructured design specifications and extract structured rules with 79.5% accuracy, stored in a knowledge graph. IFC-formatted BIM component attributes are then compared with these rules to check structural completeness and compliance, achieving 84.4% precision. The results indicate that the framework offers an effective solution for automated rule extraction and has the potential to improve compliance-checking efficiency and accuracy in engineering practice.