Speaker
Description
The SPIRAL2 superconducting LINAC at GANIL operates 26 quarter-wave resonator cavities whose online diagnostics currently rely on physics-based models limited to single operating points. This paper presents two complementary AI-based diagnostic tools: (i) neural-network heat-load virtual observers that estimate the cavity thermal dissipation — a proxy for the intrinsic quality factor Q0 — from cryogenic process signals, with prediction errors predominantly in [−2, +1] W@4.2 K for loads up to 20 W@4.2 K; and (ii) a machine-learning pipeline meant to detecting anomalies in LLRF data, predicting alarms before they fire, and classifying fault subtypes within the cavity-quench category ($F_1$ = 92%). This paper presents a state of progress on these two applications.
| In which format do you inted to submit your paper? | LaTeX |
|---|---|
| Preprint marking on your proceeding paper | I do not wish my paper to be marked as preprint. |