Automated finite element modeling method for steel bridges integrating 3D point clouds and intelligent drawing recognition technology

Laser scanning is widely recognized for capturing bridge geometry, yet automation of information extraction and finite element model (FEM) generation remains limited by manual intervention. Therefore, an automated FEM framework for bridges is proposed by integrating point cloud with intelligent recognition techniques. This paper presents three key contributions: (1) A high-precision external dimension extraction algorithm is developed based on bridge-specific features and secondary segmentation, combining projection density, adaptive thresholding, and region-growing RANSAC; (2) An internal drawing extraction framework is established using deep learning-based search, optical character recognition (OCR), and large language models for automated retrieval of structural information; (3) A FEM generation process is implemented by aligning internal and external data through component naming conventions, using a three-step algorithm involving segmentation, element creation, boundary and load assignment. Validations on arch bridge model and pedestrian bridge are conducted. This paper provides an initial exploration toward automated digital modeling in bridge engineering.