Research on Feature Extraction for Flaw Ultrasonic Echo Based on Wavelet Pack Analysis

 Abstract—Non-destructive testing (NDT) is the technology of using the characteristics of sound, light and magnetic to detect all kinds of physical properties and organization of metal components without any damage to them in order to identify the defects of surface and inner of them. With the development of NDT technology, ultrasonic wave is used to detect the bonding defects of metal thin composite plate by more and more people because of its advantage of greater penetration and higher sensitivity. This paper applies wavelet analysis as the tool of processing signals to extracting features and proposes a new method of filtering based on the traditional wavelet soft and hard threshold filtering algorithm. Because of the limitation of the traditional wavelet theory, this paper applies wavelet pack technology to extracting energy torque of the echo of ultrasonic wave. Decomposed by wavelet pack, the original signal is divided into several frequency bands, whose number is relate to the number of decomposition level. Calculate the energy torque of each frequency band, and the results show that the energy torque can be viewed as the feature used to recognize the bonding defects of the metal thin composite plate.