The diagnostics of track superstructure, which involves geometric measurements, direct observation and railroad surveillance, provides the basis for making decisions regarding the commencement of repair works. Planning repairs and increasing the probability of making the right decision at the right time also requires knowledge of the basic performance specifications of a given railway line, especially the maximum train speed and the permissible traffic volume. The article discusses a way to plan the repairs of track superstructure using artificial neural networks. It features a description of the process of designing, building and training a neural network, based on which a way to predict the degree of urgency of repairs has been discussed. The conclusions point towards the potential advantages of neurocomputers in the process of track superstructure maintenance.