Search results

The quality of a track circuit monitoring system is defined to the great extent by its ability to automatically analyse the acquired data, i. e. identify the state of a track circuit. Machine learning techniques can be applied to implement this functionality. Designing a machine learning algorithm requires a learning data set from the subject domain. In this article we discuss a basic principle of track circuit mathematical model design that could be used to generate such data set. We also apply this principle and a combination of some existing methods to design and demonstrate a mathematical model of the 25 Hz AC track circuit.