Defect classification by pulsed eddy current technique in con-casting slabs based on spectrum analysis and wavelet decomposition

• A new PEC feature (spectrum analysis after wavelet decomposition) is investigated.
• The spectrum of the detail decomposited exceeding 1.7 kHz is more effective.
• The high-frequency components are dominant among the surface and sub-surface defects.
• The PCA-LDA and Bayesian classifier can correctly classify all kinds of defects.

By comparison with traditional eddy current inspection, abundant response in wide frequency range and more utilized information both in the time-domain and frequency-domain can be realized by pulsed eddy current (PEC). However, the lift-off effect, measurement noise, and the surface oxidation are observed to negatively effects time-domain feature during PEC classification of the con-casting slabs (CCS). In this work, the spectrum analysis (SA) and SA after wavelet decomposition (SA-WD) are proposed for feature extraction. The principal component analysis (PCA) plus to PCA linear discriminant analysis (PCA-LDA) are aimed at counteracting the dimensional methods which are used for supplying the lower dimensional features. The Bayesian classifier is also applied to defect classification. The experimental results demonstrate that the cracks and cavities in surface and sub-surface can be classified satisfactorily by the proposed methods, which have the potential for gauging automatic in situ classification for CCS.