Defect classification using PEC respones based on power spectral density analysis combined with EMD and EEMD

• A new PEC feature (PSD of the IMFs) is investigated.
• The PSD distributions from EMD and EEMD transform are analyzed.
• The classifiers including PCA-LDA and PCA-Bayes are employed for defect classification.
• The multi-resolution IMFs provide an alternative to acquire more classifying features.

The defect classification is investigated by using features-based giant-magnetoresistive pulsed eddy current (GMR-PEC) sensor. The power spectrum density of the intrinsic mode functions (IMFs) is extracted as the classification feature, considering the disadvantage of selecting a wavelet base determined in previous work on spectral analysis combined with wavelet-decomposition. The IMFs are derived through empirical mode decomposition (EMD) and ensemble EMD. Principal component analysis, linear discriminant analysis, and Bayesian classifier are employed for defect classification algorithm. The proposed approach is validated by experiments, and results indicate that the cracks and cavities in the surface and subsurface can be classified satisfactorily.