Enhancing quench detection in SRF cavities at the EuXFEL: Towards machine learning approaches and practical challenges

Detecting anomalies in superconducting cavities at the EuXFEL is essential for reliable operation. We began with a model-based anomaly detection approach focused on residual analysis. To improve fault discrimination, particularly for quench events, we augmented the detection with a machine learning-based classification. Key challenges are posed by the transition to real-time operation, requiring computational and integration adjustments. For the online application, we deployed two servers at one of the 25 stations to detect and log anomalies with a software implementation. In parallel, we pushed the development of a firmware solution that will counteract critical faults in real-time. At the current stage only the anomaly detection is in online operation, which is planned to be augmented with the online fault classification in the future. The resulting detection system delivers reports across various timescales, supporting both immediate responses and long-term maintenance.