A Novel Wavelet Packet Denoising Method for Ultrasonic Testing in Austenitic Stainless Steel Welds

Abstract: The inhomogeneity and anisotropic characteristics of the grains in austenitic steels weld cause significant ultrasonic wave scattering. This leads to substantial noise contamination in the collected signals and makes it difficult to identify weld defects. To solve the problem, a novel ideal is put forward for ultrasonic signal denoising, in which ultrasonic dynamic scanning is considered, i.e., both the target signal and the signal collected at adjacent measurement points are taken into account. Based on this, a modified wavelet packet denoising method is proposed. In this method, firstly, the wavelet packet decomposition is performed on both the target signal and its adjacent signal. Secondly, according to the theory of signal cross-correlation, a set of optimized wavelet packet coefficients are obtained using the decomposed data. At last, threshold denoising and signal reconstruction are carried on the optimized wavelet packet coefficients. Artificial defects contained austenitic steels weld block are made and tested using ultrasonic pulse-echo technique. The collected ultrasonic signals are denoise processed using both the conventional and modified method and the results are compared and analyzed. The experimental findings demonstrated that the suggested approach can efficiently reduce noise in ultrasonic testing signals from austenitic stainless steel welds, so as to realize the identification and localization of defects. Moreover, the modified denoising method is more stable and better than the conventional one.