Speaker
Description
Advanced Light Source (ALS) control systems operate with continuous streams of more than 200\,000 EPICS Process Variables~(PVs). While most alarms are handled by manually setting a threshold, we present a configurable service that scores live PV streams for anomalies in real time, providing a flexible fault discrimination tool for operators. The service applies a recurrent neural network to capture time-dependent correlations, trained on a given time range using unsupervised one-class loss that requires no labeled fault examples. Furthermore, we embed operationally-aware logic that suppresses false alerts during machine-off periods, ensuring detection quality without manual setpoint tuning per PV. We show the system in operation on live data from the ALS linac modulator capacitor voltage divider channels, and describe its architecture as a configurable, multi-instance service driven by a single JSON specification, enabling straightforward extension to additional PV groups as training data becomes available.
Funding Agency
This work was supported by the Director of the Office of Science of the U.S.Department of Energy under Contract No. DEAC02-05CH11231.
| In which format do you inted to submit your paper? | LaTeX |
|---|---|
| Preprint marking on your proceeding paper | I wish my paper to be marked as preprint. |
| I no longer wish to present this contribution, please withdraw it. | Keep my contribution |