Detection of a Transverse Crack in Railheads with the Help of Wavelet Transforms and Neural Networks

We construct an artificial neural network, with the help of which and the coefficients of continuous wavelet transformation, we can automatically detect signals from transverse cracks in rails in the defectograms recorded by a flaw-detector car. In the case where the mother wavelet function of the continuous wavelet transform and a configuration of the artificial neural network are successfully chosen, we can detect signals from defects in the initial stages of their initiation, i.e., within the period of time when the signals are comparable with noise.