基于BP神经网络的钢丝绳电涡流无损定量检测技术

钢丝绳在建筑、旅游、运输等行业中已得到了广泛应用,由其断丝、磨损等缺陷所引起的安全隐患备受人们关注。采用双探头低频透射式钢丝绳电涡流无损检测方案,选取感应信号相对于激励信号的峰-峰值差和相位差作为特征量,采用数字式峰-峰值算法和占空比原理计算信号特征量。应用BP神经网络对钢丝绳缺陷进行自动识别,以钢丝绳型号及其缺陷特征量为网络输入,以是否存在断丝及断丝数量为网络输出,通过离线训练方法获取神经网络辨识模型。对实验数据进行识别,结果表明BP神经网络能对断丝缺陷及其数量进行有效的定性及定量识别。

Eddy Current Nondestructive and Quantitative Detection of Wire Rope Based on Back Propagation Neural Network

Wire rope has been widely used in building,tourism,transportation and other related fields.Its safety problems caused by the defects that result from wire-breaking and wear attract so much in our community.The detection schematic of dual probe low frequency piercing eddy current was adopted to detect wire rope defects.The phase difference and peak-peak difference of response signal which related to excitation signal were selected to be characteristic quantities.Based on the digital algorithm about peak-to-peak and duty principle,the characteristic value of signal was calculated.For the eddy current non-destructive and quantitative detecting of wire rope,the BP neural network was used for automatically distinguish defects,the network structure was designed to get neural network models by offline training method based on the input and output of network.Measuring the experiment data,the identification results show the neural network can be used to effectively identify the condition and number of defects.A new idea for wire rope non-destructive and quantitative detecting is supported.