A Detection Method for Bridge Cables Based on Intelligent Image Recognition and Magnetic-Memory Technology

Abstract

In recent years, more and more disasters caused by the fracture of bridge cables have been reported. Brittle fracture of cables is mainly caused by corrosion fatigue, and thus it is crucial to find cable defects early. The present study developed a cable inspection method embedded with a lightweight deep-learning model and equipped with micromagnetic sensors, based on intelligent image recognition and magnetic memory technology. After four cable-stayed bridges were found with defects, five types of defects and features on the surface of cables were identified by the SqueezeNet network model with the image denoising algorithm and transfer-learning method, with accuracy of 97.18%. The corrosion along the cable was positioned with micromagnetic sensors. Four alerting levels were proposed and corresponding remedial measures were suggested to be implemented. The novelty of this work lies in the intelligent detection of bridge defects, as well as accurate evaluation of long-term performance of bridge cables.