AUTOMATIC JUDGING METHOD FOR FLAW KIND OF FLAW DETECTION

摘要

PROBLEM TO BE SOLVED: To make it possible to judge the flow kind of flaw detection quickly and automatically without depending on the technique of a skilled inspecting person and to achieve paperless operation by processing the X/Y voltage output signals detected by an eddy-current flaw detector with a neural network. SOLUTION: The X/Y voltage output signals of the the detected flaw, which is detected by an eddy-current flaw detector along the longitudinal direction of a metal pipe, are inputted into a neural network, and the Lissajour's drawing is drawn. The total sum of the vectors or the X/Y voltage output signals inputted in the respective phase angle range of the Lissajous's figure is operated and made to be the respective intensity of the respective phase angle range. Then, the intensity distribution is inputted into a neuron input layer 1, and learning is performed so that what kind of flaw the intensity distribution can be outputted from an output layer 3 into the neural network. When the learning is performed in this way, the neural network can automatically output what kind of flaw the detected flaw is by the quick operation when the intensity distribution or the detected flaw of a certain kind is inputted into the input layer 1 from the output layer 3.