Speaker
Description
The CERN Proton Synchrotron (PS) is equipped with several RF systems covering a wide range of revolution frequency harmonics. While the beam synchronous RF signal generation and cavity controllers are mostly digital, the global low-level RF beam loops still rely on analogue hardware. Upgrades to a fully digital system are underway.
Digitizing additional key signals from the present beam control recently enabled the exploration of automated monitoring methods. Subtle RF issues such as parameter drifts and intermittent anomalies often go undetected, delaying diagnosis until beam quality degrades or operation is disrupted. To address this, we investigate machine learning-based anomaly detection models that aggregate information from low-level RF signals, beam diagnostics, and contextual data (e.g. magnetic field). These models seek to automatically detect abnormal behaviour and connect beam effects with underlying causes linked to the RF systems, supporting both real-time alerts and root-cause analysis.
This proactive, data-driven approach aims to shorten the response time to performance degradation, improve reliability, and support preventative maintenance of the PS RF systems.
| Abstract Category | System and Operation |
|---|
