Machine Learning to Boost Particle Accelerator Diagnostics
The U.S. Department of Energy's Thomas Jefferson National Accelerator has equipped operators of its primary facility, Continuous Electron Beam Accelerator Facility (CEBAF), with a new tool to help them rapidly address any issues that may arise.
In preliminary tests, the tool successfully used machine learning to identify glitchy accelerator components and the glitches they produced in near-real-time.
The results of the field test were published in the journal Physical Review Accelerators and Beams.
Machine learning tests at CEBAF
The CEBAF, a Department of Energy User Facility, features a unique particle accelerator to explore the fundamental structure of nuclear matter.
Powered by superconducting radiofrequency (SRF) cavities, CEBAF isn't immune from operational issues:
"The heart of the machine is these SRF cavities, and quite often, these will trip. When they trip, we'd like to know how to respond to those trips. The trick is understanding more about the trip: which cavity has tripped and what kind of fault it was," Chris Tennant, a Jefferson Lab staff scientist in the Center for Advanced Studies of Accelerators, explains in a press release.
In late 2019, Tennant and a team of CEBAF accelerator experts set out to build a machine learning system that could perform reviews in real-time. Such reviews would otherwise take operators a great deal of accumulative time.
Their custom data acquisition system pulls information on cavity performance from a digital low-level RF system that is installed on the newest sections of a particle accelerator in CEBAF.
The low-level RF system constantly measures the field in SRF cavities and tweaks the signal for each one to ensure optimal operation.
Efficient particle accelerator diagnostics
The new machine learning system was installed and tested during CEBAF operations over the course of a two-week testing period in early March 2020.
"For that two weeks, we had a few hundred faults that we were able to analyze, and we found that our machine learning models were accurate to 85% for which cavity faulted first and 78% in identifying the type of fault, so this is about as well as a single subject matter expert," Tennant explained.
The near-real-time feedback allowed CEBAF operators to make quick decisions on mitigating problems that arose in the machine during experimental runs.
"The idea is eventually, the subject matter experts won't need to spend all their time looking at the data themselves to identify faults," Tennant explained.
The next step for Tennant and his team is to analyze data from a second, longer test period. All going well, they then aim to begin new designs that would extend the machine learning system to include older SRF cavities in CEBAF.