Investigations on classifying pulsed eddy current signals with a neural network

This paper investigates the classification of pulsed eddy current signals with neural networks. In the experiments, a pulsed eddy current inspection was carried out to assess the condition of a multi-layered structure, which contains metal loss at various locations. The test specimen is an engineered specimen designed to assist in the development of inspection techniques for corrosion damage detection in aircraft fuselage splices. Separate neural network classifiers were developed using features extracted in the time domain, the coefficients of a transfer function based on a system identification approach, and the coefficients from a wavelet transform. The experimental results show the potential of applying neural networks to classify pulsed eddy current signals and to quantify the material loss evaluation. The robustness of this approach is also discussed.