2026/01/01 by Susanne Reetz, Thorsten Neumann, Douwe Buursma +1 · 1 voice
Engineering · #Machine Fault Diagnosis Techniques #Railway Engineering and Dynamics #Railway Systems and Energy Efficiency
paper · doi:10.1093/iti/liag005
openalex publication_date 2026/01/01 · openalex created_date 2026/04/25 · openalex updated_date 2026/06/18
Abstract Digitization and artificial intelligence for monitoring the health conditions of switches and point machines in the railway infrastructure has become increasingly important. By using a large unlabeled dataset of motor current curve measurements collected over two years from 58 electro-mechanical point machines in the Netherlands—containing over 1.7 million samples—this paper introduces an unsupervised health indicator. In contrast to many existing approaches, the indicator is comparable across point machines and operating temperatures. It also features transparent training and prediction pipelines, and its outputs are easily interpretable by maintenance personnel. The resulting ranking can be used to quickly gain an overview and prioritize planned maintenance between machines. Evaluation of the health indicator on the large dataset demonstrates behaviors that highlight the importance of condition monitoring systems that go beyond the analysis of individual current curves. Instead, short-, mid-, and long-term observations should be synthesized into a coherent overall health assessment.