Improving Noise Immunity of Audio Frequency Track Circuits Using Neural Networks and Data Classification

Abstract

Track circuits are key elements of railway automation systems, and train safety depends on their reliable operation. During the operation, they are exposed to numerous noises. The article considers a method to improve noise immunity of audio frequency track circuits (AFTC) so that the influence of noise can be reduced by opening the transmission path at the input of a track receiver in the intervals between signal current pulses. It allows to increase the signal-to-noise ratio at the input of a track receiver from 10% to 30%, depending on the noise parameters and the useful signal level. To eliminate noise in the intervals between pulses of the useful signal more effectively, a method of adaptive delay line control is proposed, which will allow to correct adaptively the interval of opening the transmission path, adjusting it to the parameters of the AFTC: the length of the rail line, the carrier frequency of the signal, the insulation resistance and the frequency of the modulating signal. A series of studies was conducted using simulation; according to the results of the studies, a database was created with tables of concordance of the values of the AFTC operation parameters with the signal transmission time. Once the data classification problem was solved, the optimal model structure based on neural networks was chosen, which implements the adaptive delay line control method.