Journals of Accelerator Conferences Website (JACoW)
JACoW
is a publisher in Geneva, Switzerland that publishes the
proceedings of accelator conferences held around the world
by an international collaboration of editors.
| URL | https://doi.org/10.18429/JACoW-IBIC2024-WEP47 |
|---|---|
| Title | Deep learning framework for fault detection in accelerators |
| Authors |
|
| Abstract | The main goal of NSRC SOLARIS is to provide the scientific community with high-quality synchrotron light. To achieve this, it is necessary to constantly monitor many subsystems responsible for beam stability and to analyze data about the beam itself from various diagnostic beamlines. This work presents an in-depth analysis of multi-modal, deep learning-based frameworks for fault detection within big research infrastructures, with a specific focus on accelerator facilities. The study explores diverse approaches and architectures for identifying anomalies indicating potential faults in operation. At the present stage, a binary classification is performed: stable beam operation or unstable beam operation / no beam with the accuracy of 90%. The models and the results obtained so far are discussed, along with plans for future development. |
| Paper | download: WEP47.pdf |
| Cite | BibTeX, LaTeX, Text/Word, RIS, EndNote |
| Conference | Proc. 13th International Beam Instrumentation Conference |
| Series | |
| Location | Beijing |
| Date | 09-13 Sep 2024 |
| Publisher | JACoW Publishing, Geneva, Switzerland |
| Editorial Board | Jianshe Cao (IHEP), Junhui Yue (IHEP), Lin Bian (IHEP), Ivan Andrian (Elettra-Sincrotrone Trieste S.C.p.A.), Volker RW Schaa (GSI), Zhichu Chen (SARI) |
| Online ISBN | 978-3-95450-249-3 |
| Online ISSN | 2673-5350 |
| Received | 03 September 2024 |
| Revised | 07 September 2024 |
| Accepted | 07 September 2024 |
| Issued | 11 December 2024 |
| DOI | 10.18429/JACoW-IBIC2024-WEP47 |
| Pages | 382-385 |
| Copyright | Published by JACoW Publishing under the terms of the Creative Commons Attribution 4.0 license. Any further distribution of this work must maintain attribution to the author(s), the published article's title, publisher, and DOI. |