Speaker
Description
The beam position monitor (BPM) system of the High Energy Photon Source (HEPS) storage ring consists of 578 BPMs distributed over 48 cells. Since BPM faults can affect the fast orbit feedback and beam stability, real-time monitoring with precise localization is essential. This work proposes a channel-level anomaly detection framework that integrates a multi-window autoencoder with a physics-based diagnostic module. The autoencoder captures normal spatial correlations among neighboring BPMs, while the electrostatic constraint inherent to button BPM signals provides an independent cross-check. By analyzing the four electrode signals (A, B, C, D) of each BPM, the system identifies not only the faulty BPM but also the specific electrode channel. Implemented with a GUI for training, offline, and real-time detection, the method is validated on HEPS operational data, successfully detecting and localizing real anomalous signal channels.
| I have read and accept the Privacy Policy Statement | Yes |
|---|