基于深度学习与漏磁探伤的桥梁缆索检测预警系统研究

Study of Bridge Cable Inspection Warning System Based on Deep Learning and Magnetic Flux Leakage Detection

为准确、全面地评估桥梁缆索的损伤,开发了基于深度学习和漏磁探伤的桥梁缆索检测预警系统。该系统主要由检测平台和预警平台两部分组成,利用检测平台中爬索机器人的高清摄像头和磁传感器列阵收集缆索表面的缺陷图像及漏磁信号数据,随后将缺陷图像输入到深度学习模型中对其进行自动分类与识别,利用小波分析处理漏磁信号数据以确定内部高强钢丝锈蚀缺陷位置,并根据检测到的数据提出了五级预警。为验证桥梁缆索检测预警系统的可靠性,利用该系统对4座在役斜拉桥的缆索进行检测。结果表明:该系统嵌入的深度学习模型和经过小波分析处理后的磁信号能够准确识别桥梁缆索表面的缺陷特征和内部钢丝锈蚀位置;该系统中预警平台可以将检测信息及时发送给管养部门,便于其采取相应的补救措施。

A bridge cable inspection warning system that is developed based on deep learning and the magnetic flux leakage detection method is proposed,aiming to provide solutions for the accurate and thoroughgoing inspection of bridge cables.The system mainly consists of two subsystems:the inspection system and the warning system.In the inspection system,the high-dimension cameras and magnetic sensor array can capture the images of surface deteriorations of cables and record the magnetic flux-leakage signals,afterwards automatically input the images of deterioration into deep learning models for categorization and identification,start wavelet analysis to process the magnetic flux-leakage signals to determine the corrosion defects locations of inner high-strength steel wires,and subsequently make a five-star warning.The bridge cable inspection warning system had been applied in four existing cable-stayed bridges to verify its reliability.The engineering practices prove that the deep-learning models embedded in the system and magnetic flux-leakage signals processed by wavelet analysis can accurately identify the traits of bridge cable surface defects and corrosion locations of inner steel wires.The warning subsystem can send the inspection data timely to the authorities for decision making.