花萼状涡流阵列传感器裂纹在线定量识别算法∗

摘要

Considering that the inverse algorithm of traditional eddy current testing can not meet the requirements of real-time, on-line, and simplification in structure health monitoring, an rosette eddy current sensor array was present,the output characteristics of the sensor while crack propagating were analyzed,and a crack quantitative i-dentification algorithm based on extraction of crack-characteristics was proposed. Then crack monitoring experiments of 304 stainless steel and TC-4 titanium alloy specimen were carried out to validate the effective of the algorithm. It was shown that the sliding window analysis mode could be employed in signal processing. The data stream within sliding window was processed by least square regression algorithm and recursive least square regression algorithm. Then the crack characteristics in data stream could be detected through regression parameters and threshold,and the crack length could be identified on-line quantitatively.%针对传统涡流检测逆向算法难以满足结构健康监测技术对实时性、在线以及算法复杂性低的要求，介绍了一种花萼状涡流阵列传感器，分析了传感器感应线圈输出信号在裂纹扩展过程中的变化特点和规律，提出了基于裂纹特征在线提取的裂纹在线定量识别算法，并通过304不锈钢和TC-4钛合金的疲劳裂纹在线监测试验对算法的有效性进行了验证。研究表明：采用滑动窗口分析方式，对窗口内数据流进行最小二乘回归和递推最小二乘回归，依据回归参数和阈值可以对裂纹特征进行有效识别，进而实现裂纹的在线定量监测。