DATA CLASSIFICATION METHOD AND NEURAL NETWORKS APPLICATION FOR INCREASING NOISE IMMUNITY OF AUDIO FRUQENCY TRACK CURCUIT

Abstract

The article proposes the optimization of the existing device for improving the noise immunity of audio frequency track circuits. Due to the fact that the signal current in the audio frequency track circuit is a tone frequency signal modulated by pulses, by opening the transmission path of the audio frequency track circuit at the input of the track receiver during the intervals between signal current pulses, the track receiver can be protected from interference. To take into account the delay time for the signal to pass from the output of the track generator to the input of the track receiver, a delay line is provided in the device for increasing noise immunity. But the time of passage of the signal current in the path of the audio frequency track circuit may vary depending on its operating parameters. To take into account these fluctuations and to increase the efficiency of eliminating interference in the intervals between useful signal pulses, a method of adaptive control of the delay line is proposed, which allows adapting the delay time parameter depending on the length of the rail line, the carrier frequency of the signal, the insulation resistance and the frequency of the modulating signal. By solving the problem of classifying data containing information about the influence of the operation parameters of the audio frequency track circuit on the signal transit time, the optimal structure of the model based on neural networks was chosen. This model implements the method of adaptive control of the delay line.