Guided waves for stress corrosion crack detection in pipelines—feature selection and classification

Abstract

This paper describes a wavelet-based analysis of electromagnetic acoustic transducer (EMAT) signals for in-line inspection of flaws in natural gas pipeline. The main objective of the project has been to implement the use of EMATs for pipe flaw detection, specifically the ability to detect stress corrosion cracks (SCCs) that are undetectable by current techniques. In this approach, two EMATs are used; one is the transmitter, while the second one, located a few inches away from the first, is used to receive the induced signal. Using a four-level wavelet decomposition, the EMAT data are filtered based on frequency. The features used to classify are derived from the coefficients representing each level of the four-level decomposition of the signature. The objective of the project was to detect SCC with minimal false positive even if smaller SCCs (shallow) are not identified. Although many features could be used, selecting the right features that results in maximum separation between the classes (SCC flaw, other pipe artifacts, and no flaw) was a challenge. This paper describes the process of down-selecting the feature sets and separating the classes. The results using this approach have shown promise.