Speaker
Description
Modern accelerator facilities often involve a large number of variables that could potentially influence the beam quality. While most variables are constrained by preset boundary conditions, long-period (from tens of minutes to one day) fluctuations within the boundaries can still significantly impact beam qualities. These variables are challenging for operators to identify and optimize due to their gradual nature and the difficulty of distinguishing meaningful trends from noise. This study explores the application of machine-learning algorithms to identify and analyze such slow-moving variables. By leveraging advanced techniques in time-series analysis and feature importance ranking, the algorithms uncover hidden dependencies between these variables and top beam quality indicator, that is, the ring loss for the Los Alamos Neutron Science Center (LANSCE). The results highlight the potential of ML to address longstanding beam quality issues that often trouble operation for weeks each time.
| Region represented | America |
|---|---|
| Paper preparation format | Word |
