Metal magnetic memory signal recognition by neural network for welding crack

Metal magnetic memory (MMM)is one of non-destructive testing method which inspection or evaluation ferromagnetic material used the inner magnetic information.It has been considered as a potential predominance for early diagnosis of crack.The wavelet analysis is employed to extract the MMM signal energy feature with or without welding crack for API 5L X70 pipeline steel at the condition of hydraulic pressure, and then the back propagation (BP)neural network is used to distinguish the weld with crack from free crack that energy feature is used as input eigenvector.The result shows that used the wavelet analysis and BP neural network can recognize the welding crack preferable.