涡流检测信号的独立分量分析预处理

提出将一种求解盲源分离问题的独立分量分析(Independent Component Analysis,ICA)算法应用于自然裂纹涡流检测(Eddy Current Testing,ECT)信号的预处理中。利用一种基于负熵极大的FastICA算法,分别对实验产生的疲劳裂纹和应力腐蚀裂纹ECT信号进行了处理,实现了ECT信号中缺陷分量与探头提离信号、部分噪声信号的有效分离。为了验证算法的有效性,同时采用小波分析算法对相同ECT信号进行了去噪处理。结果表明ICA算法在ECT信号处理中具有独特优势。

ICA and Its Application in Signals Preprocessing of

This paper presents an approach to preprocess eddy current testing (ECT) signals of natural crack by means of a fast algorithm of independent component analysis (ICA), an effective algorithm for blind source separation (BSS) problem. A FastICA algorithm based on maximum negentropy is used in preprocessing of fatigue crack and stress corrosion crack ECT signals. The signals of defect, lift-off of probe, surrounding noises are successfully separated. To prove the validity of ICA algorithm, the wavelet denoise algorithm is also used in processing the same ECT signals, the results show the unique advantages of ICA application in ECT signal preprocessing.