Journals of Accelerator Conferences Website (JACoW)
JACoW
is a publisher in Geneva, Switzerland that publishes the
proceedings of accelerator conferences held around the world
by an international collaboration of editors.
| URL | https://doi.org/10.18429/JACoW-IPAC2026-MOP6310 |
|---|---|
| Title | Machine-Learning–Assisted Bayesian Uncertainty Quantification for Accelerator Digital Twin Modeling and Control |
| Authors |
|
| Abstract | Digital twins of particle accelerators are increasingly used for experiment planning, machine studies, and model‑based control. Achieving high‑fidelity predictions requires knowledge of machine properties that are difficult to measure directly, such as magnet alignments, transfer function variations, nonlinearities, and stray fields. In this work, we introduce parameterizations to capture these effects and employ Bayesian inference to estimate their values and uncertainties by calibrating a digital twin to orbit response measurements from the AGS Booster at Brookhaven National Laboratory. A machine‑learning emulator trained on a perturbed ensemble of Bmad simulations enables computationally efficient sampling of the high‑dimensional posterior. The resulting joint parameter distribution incorporates BPM uncertainties and provides data‑constrained variations that, when inserted back into the digital twin, significantly improve agreement with measured beam orbits while yielding uncertainty estimates on both parameters and predictions. |
| Paper | download: MOP6310.pdf |
| Cite | BibTeX, LaTeX, Text/Word, RIS, EndNote |
| Conference | 17th International Particle Accelerator Conference |
| Series | |
| Location | Deauville, France |
| Date | 17-22 May 2026 |
| Publisher | JACoW Publishing, Geneva, Switzerland |
| Editorial Board | Editorial Board |
| Online ISBN | 978-3-95450-252-3 |
| Online ISSN | 2673-5350 |
| Received | 11 May 2026 |
| Revised | 15 May 2026 |
| Accepted | |
| Issued | 20 July 2026 |
| DOI | 10.18429/JACoW-IPAC2026-MOP6310 |
| Pages | 403-406 |
| 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. |