Self-organzied critical control for the European XFEL using black box parameter identification for the quench detection system

Abstract

Abstract-The European Free Electron Laser (XFEL) consists of a large and complex plant, with many cost intensive and technological high-end components. It is therefore important that the XFEL can be operated reliably and safely using exception handling and fault detection systems. A crucial part of the system are the superconducting cavities for which especially quenches, i.e. the break down of the superconductivity have to be avoided. The paper shows the interaction of the fault detection system with the Low Level RF (LLRF) control system to maximize the field gradients. This is an example for process supervision, which can neither be classified as fault-tolerant, nor is it a reconfiguration system, but uses the result of the fault detection to operate the system at its fault critical limit. This scheme simulates a system which behaves like a selforganized critical system, and drives the process at its critical performance limit. It is therefore called Self-organized Critical Control (SOCC). The paper shows the basic set-up and quench detection methods of the European XFEL and gives an example for an application of SOCC.

Opportunities in Machine Learning for Particle Accelerators

“…Anomaly and breakout detection algorithms may provide added capabilities beyond those of existing machine protection systems by detecting subtle behaviors of key variables prior to negative events. This approach has been used for detection of quench precursors in both superconducting magnets [49] and superconducting RF cavities [50].…”

Intelliquench: An Adaptive Machine Learning System for Detection of Superconducting Magnet Quenches

“…This motivates Deep Learning (DL) is a sub-field of machine learning that focuses on learning features from data through multiple layers of abstraction [4]. Several DL techniques have been explored for anomaly prediction & detection of superconducting magnet quenches [5]- [7]. However, the techniques are still in their early development phase, and there is yet to be a practical setup and a well-defined procedure to be used for real-time prediction of magnet quenches.…”

Anomaly Detection in Particle Accelerators using Autoencoders

“…For example, understanding and predicting faulty behavior in superconducting radio frequency (RF) cavities and magnets is of interest due to the potential catastrophic nature of a failure of these devices. ML tools have been applied to detect anomalies in superconducting magnets at CERN [9] and RF cavities at DESY [10]- [12]. Additionally, machine learning has been used to identify and remove malfunctioning beam position monitors in the Large Hadron Collider (LHC), prior to application of standard optics correction algorithms [13].…”