Ultrasonic Defect Echoes Identification Based on Empirical Mode Decomposition and Stransform

Abstract:

In order to detect the minor defect echoes from noisy ultrasonic signals in nondestructive testing of metallic materials, a backscattering signal model of metallic materials was established. The unbiasedness of estimation of amplitudemodulated echo’s center frequency was discussed; and the defect echoes identification method based on empirical mode decomposition (EMD) and Stransform was proposed. In the first stage, the EMD was performed on the original signal and these IMFs with high frequency were removed, which realized a lowpass filter. Then a denoising method based on Stransform and OTSU was used for noise suppression of the reconstructed signal to eliminate the aliasing between useful signal and noise, yielding high SNR signal with relatively large amplitude attenuation. Finally, the Stransform was performed again on the resulting signal to mitigate the amplitudes attenuation caused by OTSUbased timefrequency spectrum denoising by means of multiplying a socalled amplitudegainfactor. With the above operations, the echoes became clear and their amplitudes were kept well. The processing result of simulation signals showed that the method can enhance signal significantly and highlight the defect echoes submerged by severe noise. And the processing results of experimental signals also showed the effectiveness of the method in noise suppression and defect identification.