Ultrasonic Flaw Echo Signal Wavelet Threshold De-noising Algorithm Study

In recent years, ultrasonic testing technology with its penetrating capacity, equipment, portable and safe operation is widely used in various fields. The most commonly used ultrasonic testing technique is pulse-echo method, the main principle is through the pulse reflected back from the metal to determine the location of internal flaws of the metal, but these flaws are often mixed with various echo interference of waves. This makes the original weak echo flaw signal more susceptible to the noise, affecting us to accurately determine the location of flaws. Therefore if we want to get the clear and without distortion waveform, we must de-noise the detected signal.So the study of some effective de-noising methods have very important significance.The traditional method is to use Fourier transform to filter noise, but the ultrasonic flaw echo signal is a non-steady-state and time-varying pulse signal, it needs to analyze the frequency components contained each moment. The spectral characteristics of these signals extend along the entire timeline for these signals, the traditional Fourier transform is powerless. Wavelet analysis has good localization properties of time-frequency analysis, so it can be a good solution to this problem. Therefore, wavelet analysis method of ultrasonic flaw echo signal de-noising can achieve good results.In this issue, we start for the basic theory of wavelet. Then according to the characteristics of ultrasonic flaw echo signal, we use wavelet threshold de-noising method dealing with the mathematical model of ultrasonic flaw echo signal, focusing on the several parameters affects the de-noising effect such as wavelet, wavelet decomposition layers, threshold, threshold function. Among them, the selection of threshold function affects the reconstruction precision and continuity of the signal. If our choice is not appropriate, it will cause severe distortion of the signal and will affect de-noising effect. Now more commonly used method is Donoho’s soft and hard threshold, but both methods have their flaws. Hard threshold can keep the basic characteristics of the original signal.but the smoothness of the signal after de-noising is rather poor and it can cause shock easily. The smoothness of the de-noising signal with Soft threshold is better, but it is easy to lose the useful signal component after de-noising. In this paper, owing to the shortage of the two kinds of threshold function, we proposed a new threshold function basing on the original method. The simulation results show that after de-noising with the new threshold function, the SNR of signal improves and the root mean square error decreases. The de-noising effect is better than the traditional threshold function.At the same time we will use the improved algorithm for practical ultrasonic detection system. In the Lab VIEW software, it calls MATLAB node to de-noise the actual ultrasonic flaw detection data. The good de-noising effect confirms the value of the engineering application of the algorithm.