Researchers model accelerator magnets’ history using machine learning approach
After a long day of work, you may feel tired or agitated. Either way, you are influenced by what happened to you in the past.
Accelerator magnets are no different. What they’ve been through — or what’s gone through it, like an electric current — affects how they’ll perform in the future.
Without understanding a magnet’s past, researchers may need to reset it completely before starting a new experiment, a process that could take 10 or 15 minutes. Some accelerators have hundreds of magnets and the process can quickly become time consuming and costly.
Now, a team of researchers from the Department of Energy’s SLAC National Accelerator Laboratory and other institutions has developed a powerful mathematical technique that uses concepts from machine learning to model a magnet’s past states and make predictions about future states. This new approach eliminates the need to readjust the magnets and immediately results in improvements in throttle performance.
“Our technique fundamentally changes how we predict magnetic fields in accelerators, which could improve the performance of accelerators around the world,” said SLAC associate scientist Ryan Roussel. “If the history of a magnet is not well known, it will be difficult to make future control decisions to create the specific beam you need for an experiment.”
The team’s model looks at an important property of magnets known as hysteresis, which can be thought of as residual (or leftover) magnetism. Hysteresis is like the leftover hot water in your shower pipes after you turn off the hot water. Your shower won’t go cold right away — the hot water left in the pipes has to flow out of the shower head before you’re left with only cold water.
“Hysteresis makes tuning magnets challenging,” said SLAC associate scientist Auralee Edelen. “The same settings in a magnet that resulted in one beam size yesterday may result in a different beam size today due to the effect of hysteresis.”
The team’s new model eliminates the need to reset magnets as often and could allow both machine operators and automated tuning algorithms to quickly see their current state, revealing what was once invisible, Edelen said.
A decade ago, many accelerators didn’t have to consider susceptibility to hysteresis errors, but with more precise facilities like SLAC’s LCLS-II coming online, predicting residual magnetism is more important than ever, Roussel said.
The hysteresis model could also help smaller accelerator facilities, which may not have as many researchers and engineers to reset magnets, to conduct higher-precision experiments. The team hopes to implement the method on a full set of magnets in an accelerator facility and demonstrate an improvement in predictive accuracy on an operational accelerator.
A new machine learning method streamlines the operation of particle accelerators
R. Roussel et al, Differentiable Preisach Modeling for Characterization and Optimization of Hysteresis Particle Accelerator Systems, Physical Assessment Letters (2022). DOI: 10.1103/PhysRevLett.128.204801
Quote: Researchers model the history of accelerator magnets using machine learning approach (2022, June 15) Retrieved June 15, 2022 from https://phys.org/news/2022-06-magnets-history-machine-approach.html
This document is copyrighted. Other than fair trade for personal study or research purposes, nothing may be reproduced without written permission. The content is provided for informational purposes only.