The Potential of Applying Artificial Intelligence Technologies to Track Section Condition Analysis Based on Input Impedance

Abstract

This paper demonstrates the potential of applying artificial intelligence technologies to track section condition analysis. Using Rython programming tools, an artificial neural network (ANN) with a multilayer perceptron structure was developed and trained. This network allows for determining the location of rolling stock based on the components of the complex value of the track line's input impedance, taking into account the matching equipment and its length. The trained ANN is capable of determining the direction of the rolling stock relative to the measurement point, as well as the distance to its boundary (head or tail). The results obtained during the study allow us to conclude that, under changing external factors, the error in determining the position of a rolling stock does not exceed 100 m.