Detection of Cracks According to Bayesian Wavelet Threshold on Metal Magnetic Memory Signals

Metal magnetic memory signals can reflect the stress concentration and defects on the surface of the ferromagnetic components, but the traditional criteria often can't detect the locations of cracks accurately. Using wavelet singularity detection principle, the locations of cracks can be detected. First, we analyzed the signal-to-noise ratios of 35 wavelet functions to the gradient values detected from the pipeline steel specimens with prefabricated crack defects. Then, we chose the wavelet function with maximum signal-to-noise ratio as the tool to recognize the crack defects. Compared with the four kinds of traditional threshold to detect the cracks by wavelet transformation, Bayesian hard and soft threshold wavelet transformation can find out the locations of the cracks more accurately, and hard threshold is more effective to detect the locations of cracks than soft threshold. Experiments show that in the range of +/- 1 mm, the recognition accuracy reaches 96%. The results show that the wavelet transformation with Bayesian threshold to the gradient values on metal magnetic memory signals can accurately recognize the location of cracks.