Crack Angle and Depth Estimation Using Wavelet Preprocessed Neural Network

A crack characterization problem must involve analysis of a signal received by appropriate transducer to evaluate the shape and size of crack in test object. The presence of a crack in ferromagnetic tube or plate, tested by magnetic field leakage (MFL) inspection method, causes a redistribution of excited magnetic flux. A neural network (NN) approac (with analog output neurons) can be used for surface-breaking crack depth and angle oblique estimation using signal profile of leakage field