Feature analysis of metal magnetic memory signals for weld cracking - based on wavelet energy spectrum

Abstract

Feature extraction from the metal magnetic memory (MMM) signal is a very important step for MMM inspection. In order to know the relation between the MMM signal and the welding crack, a method based on continuous wavelet transform (CWT) energy spectrum is introduced in this paper. Using the scale-wavelet energy spectrum and the space-wavelet energy spectrum, research is focused on whether crack information exists in the MMM signal and on the locating of welding cracks. The conclusion can be drawn as follows: For a MMM signal with cracks, the distributing range of scale-wavelet energy spectrum is wide at a high-energy level and the energy spectrum will exponentially decrease after a peak; Space-wavelet energy spectrum indicates that a concentration is the position of the crack.