Study On Denoise Of Ultrasonic Testing Signal Based On Wavelet Analysis

Ultrasonic detection technology is an important tool to ensure the operation of equipment safety. However, when carrying out ultrasonic testing especially in the coarse-grained material, a mass of structural noise makes the useful information lost in the noise. At the present, the wavelet thresholding method is the most widely in singnal denoise processing. When processing the ultrasonic signal by the wavelet thresholding method, the selection of decomposition level and the mother wavelet and the determination of threshold value and threshold processing method are the two key issues. Based on the theory of signal processing and the characteristics of ultrasonic signal, the paper analyzed that the characteristics of mother wavelets for ultrasonic signal analysis, the result shows that Daubechies wavelet family, Symlets wavelet family, Coiflets wavelet family and Biorthogonal wavelet family are suitable for ultrasonic signal analysis according to the properties of existing mother wavelets. Subsequently, the constructed noisy ultrasonic signals was processed by combining that four kinds of wavelet family with decomposition levels(1～10). The noises of ultrasonic signal is unknown were considered in practical application,"sln"and"mln"had been adopted when the paper selected threshold rescaling methods, the constructed noisy ultrasonic signal was processed by four kinds of threshold selection methods. Moreover, considered disadvantages of soft and hard thresholding function and analyzed advantages of three kinds of improved thresholding methods, the paper processed the constructed noisy ultrasonic signal by soft and hard thresholding methods and three kinds of improved thresholding methods. The result indicates that for the ultrasonic signals, the five wavelets, such as db12, sym8, sym10, coif2 and biors6.8, have better denoise effect, and the optimal decomposition level is 4; When"sln"and"mln"was selected, the optimal threshold selection method is"minimaxi", whose SNR increase by 0.1～0.7 dB compared with the other three threshold selection methods; In addition, the denoise effect of three kinds of improved thresholding methods is better than hard thresholding method, but there is little difference between soft thresholding method and improved compromise method of soft-hard threshold, and the value of the difference between two methods basically keeps 0.01～0.03 dB. When the noise grade is 1～12(stronger noise), the denoise effect of improved compromise method of soft-hard threshold is better than the other two improved methods, and SNR of improved compromise method of soft-hard threshold increase by 0.2～1.5 dB compared with the other two improved methods. When the noise grade is 13～24(weaker noise), the denoise effect of exponential type thresholding method is the same to the effect of hyperbolic type thresholding method, and SNR of the two methods increase by 0.2～1.8 dB compared with improved compromise method of soft-hard threshold,but the denosie visual effect of exponential type thresholding method is better than the effect of hyperbolic type thresholding method. Therefore, when the noise is weaker, the exponential type thresholding method is choosed. Finally, the paper removed noises from two practical collected ultrasonic signals of austenitic stainless steel welds, and from the denoise effect the optimal parameters of the wavelet thresholding method suited to the ultrasonic echo signals are proved to be the reasonable ones.