Unsupervised Anomaly Detection of Subway Vibration Signal Based on Ultra-weak FBG Sensing Array Combined with RCMDE and Local Density
Yan Yang, Qianru Yin, Lina Yue · Journal of Physics Conference Series · 2023
Abstract The subway vibration signal can reflect the health status of the passing train and tunnel structure. Identifying the abnormal vibration signal is a prerequisite for safe subway operation. In view of the difficulties in collecting anomalous samples in actual engineering and extracting effective features of high-dimensional time series, an unsupervised anomaly detection approach on the ground of RCMDE and local density has been put forward to identify vibration signals captured by the ultra-weak FBG sensing array cable. Two experiments demonstrated that compared with other methods, RCMDE could better evaluate the signal and extract features more precisely and that the abnormal vibration signals of passing trains consistent with the facts were accurately detected by the proposed method while other methods had errors.