Feedback Neural Network Approach
Inverse problems in nondestructive evaluations involve the estimation of defect profiles in materials. Estimating defect profiles can be formulated as a function-approximation problem and the solution obtained using artificial neural networks. In order to retain the advantages of phenomenological and non-phenomenological solution techniques, and to overcome the disadvantages of phenomenological methods, a feedback neural network scheme was developed for solving the inverse problem. The feedback neural network approach is shown at right. Two neural networks are used in a feedback configuration. The forward network predicts the signal corresponding to a defect profile while the inverse network predicts a profile given an inspection signal. The forward network provides a reference for comparing the defect profile predicted by the inverse neural network. | |
Approach to Solving the Inverse Problem The overall approach to solving the inverse problem is shown here. The input signal, ƒ, from a defect of unknown profile is input to the characterization neural network (inverse neural network or INN) to obtain an estimate of the profile This estimate is then input into the forward neural network (FNN) to get the corresponding prediction of the MFL signal for that estimate of the profile. If the estimated defect profile is close to the true profile, the measured MFL signal and the predicted signal from the forward network will be similar to each other. This is the basis of the feedback neural network scheme. The following links provide details on the forward network, the inverse network, and the methodology used to optimize the network system:
|
