Application of 1D convolutional neural networks for anomaly detection in resistance measurements of railway wheelsets

The article presents the application of convolutional neural networks (CNN) for the classification of electrical resistance measurements of railway wheelsets. The aim of the study was to develop a model capable of automatically detecting incorrect measurement results based on data obtained from various measurement configurations. The training process used experimental data collected under real-world conditions. The developed model achieved high classification accuracy and was tested on variable-length data. The study demonstrates that CNN-based methods can be effectively applied in the diagnostics of measurement systems.