主成分分析在飞机多层结构层间腐蚀脉冲涡流检测中的应用

摘要

It’ s required to identify the defect signals from interfered signals caused by lift-off effect, and estimate the defect depth in aircraft multi-layered structure hidden corrosion detection with pulsed eddy current testing. A simulated sample of multi-layered riveted metal structure was made and different size and depth corrosion defects were tested. PCA was used and first three principal components were extracted. Results show that PCA method is able to distinguish pure lift-off signals against Signals with corrosion defects between the layers significantly. Pure lift-off signals were identified from pure corrosion signals and corrosion plus lift-off signals by applying K-means algorithm to the extracted principal components. But signals with both lift-off and corrosion components, can not be identified exactly from pure corrosions with PCA method.%在飞机多层铆接结构层间腐蚀缺陷的脉冲涡流检测中,需要识别提离效应造成的干扰信号和缺陷信号,同时也需要判断缺陷深度。制作了模拟飞机多层铆接金属结构的试样,对不同深度和大小的腐蚀缺陷进行了检测。采用主成分分析( Principal Component Analysis, PCA)方法对实验数据进行处理,并提取前3个主成分进行分析。结果表明：应用PCA方法,可以将纯提离信号与带层间腐蚀缺陷的信号显著区别开来,可以将不带提离时的纯腐蚀信号的深度识别出；将PCA提取的主成分应用K-means算法进行聚类,可以将纯提离信号与纯腐蚀信号和腐蚀提离混合信号区别开来。而对于带提离的腐蚀,试验发现其PCA分布与不同深度的纯腐蚀出现混淆,因而不能准确识别这两种信号。