Application of time-frequency analysis for automatic hidden corrosion detection in a multilayer aluminum structure using pulsed eddy current
- 作者：Saleh Hosseini ; Aouni A. Lakis ; aouni.lakis@polymtl.ca
- 关键词：Multi-layer aluminum structures ; Pulsed eddy current ; Time&ndash ; frequency analysis ; Feature extraction ; K-means clustering ; Expectation&ndash ; maximization algorithm
- 刊名：NDT and E International
- 出版年：2012
- 期刊代码：38_09638695
- 类别：ph
- 出版时间：April, 2012
- 卷：47
- 期：Complete
- 页码：70-79
- 文件大小：2154 K
摘要
Pulsed eddy current (PEC) is a non-destructive testing method used to detect corrosion and cracks in multilayer aluminum structures which are typically found in aircraft applications. Corrosion and metal loss in thin multi-layer structures are complex and variable phenomena that diminish the reliability of pulsed eddy current measurements. In this article, pulsed eddy current signals are processed to improve the accuracy and reliably of these measurements. PEC's results (time domain data) are converted by time-frequency analysis (Rihaczek distribution) to represent data in three dimensions. The time-frequency approach generates a large amount of data. Principal component analysis is applied as feature extraction to reduce redundant data to provide new features for classifiers. K-means clustering and expectation-maximization are applied to classify data and automatically determine corrosion distribution in each layer.