Issue34

Z. Jijun et alii, Frattura ed Integrità Strutturale, 34 (2015) 590-598; DOI: 10.3221/IGF-ESIS.34.65 596 Figure 10: Results of angle adaptive method. Figure 11: Horizontal integral projection. Identify the defect of wire rope with neural network Identify the superficial defect of the wire rope with BP neural network, Sigmoid function as excitation function, and error signal back propagation algorithm [16-20] as training algorithm, parameters are as below: number of nodes at input layer is 4, and the input values are respectively quantity of effective pulse, width of pulse, sum of height of pulse and radius of wire rope under detection. We can identify whether the different kinds of wire ropes have defect. Besides, the number of nodes at implicit strata is 6 and at output layer is 1. Assume the transmission function and output of the j nerve cell at L-1 layer of the k training sample are respectively 1 ( ) L j u k  and 1 ( ) L j z k  , then: 1 1 1 1 1 ( ) 2 ( ) 2 ( ) ( ) 2 ( ) ( ) L K i ij L L k L i ij K L L L i im m L k ij K L L i j k u k E E u k k z k k z k                                        In the formula, E is the overall error; ij L  is weight; ( ) i k  is delta error. 1 1 1 1 1 1 ' 1 1 ( ) ( ) ( ) ( ) ( ) ( ( )) ( ) ( ( )) ( ) L M L m i L L L m i m i J K L L L m mj j L k j i M L L L j m mi m u k E E k u u k u k k f u k u k f u k k                                        The renewal process of weight is as below: 1 1 ( 1) ( ) ( ) ( ) ( ) ( 1) ( ) K L L L L ij ij i j k L L L ij ij ij t t k z k t t t                       