Deep learning-based corrosion inspection of long-span bridges with BIM integration

Abstract

Infrastructure operation and maintenance is essential for societal safety, particularly in Japan where the aging of infrastructures built during the period of high economic growth is advancing. However, there are issues such as a shortage of engineers and inefficiencies in work, requiring improvements in efficiency and automation for their resolution. Nevertheless, there are still many inefficiencies in the current procedures for bridge inspections. Usually, inspection engineers check for damage on bridges through close visual inspections at the site, then photograph the damaged parts, measure the size by touch, and create a report. A three-dimensional representation, considering the front and back of the structural elements, is needed for identifying damage, necessitating the creation of multi-directional three-dimensional drawings. However, this process is labor-intensive and prone to errors. Furthermore, due to the lack of uniformity in records, it is challenging to refer to past inspection histories. Especially for long bridges, without resolving such issues, the required labor and the number of mistakes could exceed acceptable limits, making proper management difficult. Therefore, in this study, we developed a method for automatically measuring the position and area of corroded parts by capturing images of the lower surface of the stiffening girder using a bridge inspection vehicle and utilizing image diagnosis technology. By integrating these results into a 3D model called BIM (Building Information Modeling), it becomes possible to manage the bridge more efficiently. We verified this method on actual long bridges and confirmed its effectiveness.