In construction, where many organizations work on a project, document consistency is critical. Errors in calculations or drawings can lead to delays, additional costs and even threats to building safety. Projects that do not use building information modeling (BIM) - a technology that allows creating a digital model and automatically checking for clashes - are especially vulnerable. High cost, a shortage of specialists and technical barriers limit the use of BIM, so many projects still rely on traditional paper or electronic documents.
Researchers from Seoul National University of Science and Technology (South Korea), led by Associate Professor Jaeuk Chong, have developed an automated system based on artificial intelligence (AI) that detects human errors in design documentation. The work was published in the journal Engineering Applications of Artificial Intelligence.
The system works in several stages. First it collects and cleans data, then extracts information from calculation documents and drawings. For calculation documents, the DocLayNet-YOLO model is used to identify areas with the required information, and text is recognized using Surya-OCR. For drawings, where elements are listed in tables with different row and column widths, element codes are first found, and then, using OCR and image processing methods (Canny edge detector, Hough transform, morphological operations), the tables are located. The extracted data is standardized.
At the next stage, the system matches information from the two types of documents, using element codes and direction as key identifiers. It identifies inconsistencies and duplicates and classifies errors. A web application has been created for user convenience.
In a test residential building project in South Korea, information extraction accuracy was 95.56% for calculation documents and 90.47% for drawings. The check revealed 253 inconsistencies: 224 missing entries, 15 main rebar mismatches, three stirrup mismatches, eight dimensional discrepancies and three duplicates. These results show that the system can automate verification and reduce the burden on engineers.
"Our system can help engineers, designers, contractors and supervisory authorities identify discrepancies before they lead to construction errors, increasing the reliability of buildings and infrastructure," Chong notes. "It is especially valuable for small and medium-sized projects where implementing BIM is impossible."
The development offers an accessible way to improve the quality of design documentation without switching to expensive digital platforms. In the future, the authors hope to integrate the system into engineering workflows.