SIGNAL+DRAHT | Issue 12/2021
Connecting measurement data and knowledge-based engineering for heavy rail switch fault diagnosis
The application of Artificial Intelligence (AI) in prognostics and health management, especially fault diagnosis, for railway infrastructure is complicated by the largely unlabelled databases and the necessity for traceability due to strict safety regulations. Promising approaches include feature engineering, unsupervised learning and knowledge-based systems. This article discusses how to treat the current curve measurements of railway point machines against this background and connect them with a human-interpretable Bayesian network model for diagnostic purposes...