Semiautomatic IFC-Based BIM Reconstruction for As-Designed Bridges from 2D Plans Leveraging Semantic Segmentation and Enrichment

Building information modeling (BIM) is an innovative approach that enables efficient life-cycle management of construction projects. The increasing application of BIM in the bridge sector can be attributed to its ability to streamline communication, enhance collaboration, and mitigate risks. Consequently, BIM contributes to the enhanced efficiency and effectiveness of bridge project workflows. However, the process of manually constructing a BIM instance model for bridges based on two-dimensional (2D) drawings can be labor-intensive, time-consuming, and error-prone. In this paper, the authors proposed a new framework to semi-automate the reconstruction of bridge BIM based on 2D portable document format (PDF) plans. The proposed framework builds on the state-of-the-art PDF2BIM, a 3D BIM reconstruction tool, and further enriches the model through semantic segmentation, information extraction, and semantic enrichment techniques. The integration of these steps helps address the gap of the output from the PDF2BIM technology in the lack of (1) segmented components; and (2) semantic information. It produces semantically segmented and enriched 3D Industry Foundation Classes (IFC)-based bridge BIM instance models. Semantic segmentation was utilized to break the bridge model into different components, such as piers and decks, leveraging as-designed point cloud data. To extract essential semantic information from the bridge plans, information extraction algorithms were developed iteratively. The developed algorithms leverage optical character recognition (OCR) and natural language processing (NLP), with the aim for extracting Specifications for the National Bridge Inventory (SNBI) items as per federal requirements. The proposed framework was tested on six bridges in the state of Indiana, United States. It achieved 97.7% precision and 94.4% recall in automated information extraction. Additionally, it reduced the overall time consumption on creating a bridge BIM instance model by 94.9% compared to the manual approach. The semantically segmented and enriched bridge IFC models can enhance data interoperability and integrity. The models can also improve stakeholder collaboration and support the life-cycle management of bridges.