Thick wall pipe ultrasonic signals processing based on wavelet packet and singular value decomposition

Abstract

The defect echo signals are often suffered from scan system and background noise when detecting internal and external flaws of thick wall pipe through ultrasonic testing method. Especially for detecting tiny flaws, the lower amplitude echo signals are easily submerged in noise. Wavelet packet threshold method can restrain the noise interference, however, the limitation and disadvantages of this method still exist. In response to this issue, a new approach of denoising based on wavelet packet and singular value decomposition (SVD) was advanced in this paper. The echo signals were decomposed with wavelet packets by utilizing the optimal wavelet functions. The high frequency noise was removed according to the energy distribution ratio of different frequency band and then the extracted energy distribution band was decomposed by singular value. Eventually an adaptive filter was realized after determining the effective echo signal area through singular entropy analysis. Experimentally analysis shows that the algorithm of denoising based on wavelet packet and singular value decomposition can not only remove the high frequency noise, but also effectively remove the low frequency noise, echo signal-to-noise ratio (SNR) was improved greatly.