A wavelet network approach for predicting surface cracks shapes

A wavelet network based technique is proposed for predicting crack depth profile using the output signal of a surface magnetic field measurement (SMFM) probe. The technique utilizes a wavelet network with Gaussian-derivative activation function. The main feature of this technique is that it requires only the sensor output signals along the crack edge. The learning process is done using the estimated probe output signals from a simulator. The application of the proposed technique to several surface cracks with various depth profiles demonstrates its ability to accurately predict the crack shape, in addition, the results indicate that the proposed technique is superior to a conventional multiplayer perceptron neural network.