An Anomaly Detection Approach for Maglev Trains’ Levitation System With Weighted Correlation Coefficient and Minimum Distance

The levitation system is a critical component of maglev trains, yet it is susceptible to abnormalities such as guideway settlement, fixture loosening, and component aging, which may lead to incipient failures. Timely and accurate detection of such abnormal states using sensor data is essential for implementing proactive maintenance and ensuring operational safety. To address this need, this study proposes a novel multivariate time series analysis (MTSA) method based on weighted correlation coefficients and minimum distance (WCC-MD-MTSA) for anomaly detection in levitation systems. The proposed approach leverages multi-dimensional sensor data to construct an anomaly detection index using weighted correlation coefficients, followed by a minimum-distance classifier for anomaly categorization. Furthermore, to overcome the challenge of determining sliding window sizes in MTSA, an adaptive data partitioning method based on autocorrelation length is introduced, providing a more principled and stable input configuration. A comprehensive anomaly processing framework is also established, integrating detection, classification, and maintenance guidance into a systematic workflow. The effectiveness and practicality of the proposed methods are demonstrated using real-world operational data from maglev train levitation systems.