Abstract: A neural network approach for solving an inverse problem of identification of crack width and depth is proposed. Radial Basis Function (RBF) neural networks (NN) perform the identification. It was trained using information from numerical simulated pulsed eddy current (PEC) nondestructive testing (NDT). The capability of the RBF NN was checked with information from numerical and physical experiment. The obtained results illustrate the efficiency of the approach.
Keywords: Nondestructive testing, inverse problem, neural networks.