Machine Learning For Beamline Steering
This work addresses a domain-specific problem for operators of particle accelerator light sources by potentially automating a repetitive calibration task, though it appears incremental as it applies existing deep learning methods to a new dataset in this context.
The paper tackled the time-consuming and effort-intensive re-calibration of magnets in the LINAC To Undulator section of a particle accelerator beamline, which reduces scientific throughput, by using deep neural networks trained on archival data and validated on simulation data, achieving performance comparable to trained human operators.
Beam steering is the process involving the calibration of the angle and position at which a particle accelerator's electron beam is incident upon the x-ray target with respect to the rotation axis of the collimator. Beam Steering is an essential task for light sources. In the case under study, the LINAC To Undulator (LTU) section of the beamline is difficult to aim. Each use of the accelerator requires re-calibration of the magnets in this section. This involves a substantial amount of time and effort from human operators, while reducing scientific throughput of the light source. We investigate the use of deep neural networks to assist in this task. The deep learning models are trained on archival data and then validated on simulation data. The performance of the deep learning model is contrasted against that of trained human operators.