The favorable localization features of discrete wavelet provide a new method for detecting the mutational
points of electric spark signal. In this paper, by means of discrete wavelet function called db5, using the way of 6 scales
wavelet, analyzing the gathered electric spark signal and by extracting the modulus maxima of the 6 layers detailed signal
coefficient, the signal’s mutational points were located exactly and successfully. In addition, via the modulus maxima to
calculate Lipschitz index, measuring signal’s singularity, the signal’s mutational time was confirmed. The result of the
simulation shows that this method can detect not only the time and size effectively if the ring fire happens but also the
failure of the locomotive traction dc motor, timely and precisely. In this way, the operation safety of the train is ensured.
Optimal function approximation with ReLU neural networks
