Application of wavelet analysis to signal de-noising in ultrasonic testing of welding flaws

In ultrasonic testing of welding flaws, the reliability and quality of tests are considerably affected by noise and spurious signals. Thus, signal de-noising and increasing of the signal-noise ratio (SNR) are a key to successful application of ultrasonic NDT and NDE. At present, there are many methods for signal de-noising, such as median filtering, non-linear filtering, adaptive filtering, correlation technique, split spectrum processing, artificial neural network, etc. Although these methods are of a practical meaning for increasing SNR of ultrasonic testing, they have some limitations affecting the reliability of testing results. As a method of time-frequency analysis, wavelet analysis is one of the most effective and promising signal processing techniques with a lot of advantages. In this paper, the theory of wavelet analysis, including wavelet transform (WT), wavelet packet transform (WPT) and lifting wavelet transform (LWT), is introduced. And according to the characteristics of ultrasonic echo-signals of welding flaws, the theory and method of wavelet for signal de-noising are discussed. Finally, the experimental studies of signal de-noising by wavelet analysis are carried out for ultrasonic simulation signals and actual defect echo-signals. Experimental results show that noises can be well suppressed and SNR is improved obviously by wavelet analysis. Comparing LWT with WT, while SNR is similar, LWT is of flexible design and fast computation with a simple programme. In addition, for ultrasonic signal de-noising, WPT is more effective than WT, but its computation is complex. Thus, LWT has a good application prospect for signal de-noising, especially in real-time signal de-noising.