The research of defect recognition for radiographic weld image based on fuzzy neural network

This paper presents a method of automatic recognition of weld defects based on fuzzy neural network (FNN). The weld image is preprocessed to extract defect features, according to which 8 characteristic parameters are selected, and the FNN model used for defects recognition is set up. Here inputted samples of the FNN model are fuzzified using π function. And weld defects of different types are processed using a three-layer neural network and BP learning algorithm. Using forty-two training samples and seven testing samples for examination, the results show that this model can recognize weld defects with better effects. This research indicates that FNN has excellent performance for the defect recognition in weld image.