Paper | Title | Page |
---|---|---|
MO2C3 |
Novel Approaches for Forecasting of Beam Interruptions in Particle Accelerator | |
|
||
The beam interruptions (i.e. interlocks) of particle accelerators, despite being necessary safety measures, lead to abrupt operational changes and a substantial loss of beam time. Novel data-driven time series classification approaches are applied in the High-Intensity Proton Accelerator complex of Paul Scherrer Institut, in order to forecast interlock events thus decrease beam time loss. The forecasting is performed through binary classification of single timestamps as well as windows of multivariate time series, with methods ranging from linear Lasso models based on statistical Two Sample Test, to deep learning model that generates Recurrence Plots followed by Convolutional Neural Network*. The "beam time saved" in any given time interval, a continuous evaluation metric, is established with preliminary experiments showing that interlocks could be circumvented by reducing the beam current. The models have been integrated with EPICS, and the best-performing interlock-to-stable classifier on real-time data potentially increases 5 min beam time per day for the users.
* Li S, Zacharias M, Snuverink J, et al. A novel approach for classification and forecasting of time series in particle accelerators[J]. Information, 2021, 12(3): 121. |
||
![]() |
|
|
please see instructions how to view/control embeded videos | ||
![]() |
Slides MO2C3 [5.032 MB] | |
Cite • | reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml) | |