Demonstration of Machine Learning-Based Model-Independent Stabilization of Source Properties in Synchrotron Light Sources

Abstract

Synchrotron light sources, arguably among the most powerful tools of modern scientific discov-9 ery, are presently undergoing a major transformation to provide orders of magnitude higher brightness and transverse coherence enabling the most demanding experiments. In these experiments, overall source stability will soon be limited by achievable levels of electron beam size stability, presently on the order of several microns, which is still 1-2 orders of magnitude larger than already demonstrated stability of source position and current. Until now source size stabilization has been achieved through corrections based on a combination of static predetermined physics models and lengthy calibration measurements, periodically repeated to counteract drift in the accelerator and instrumentation. We now demonstrate for the first time how application of machine learning allows for a physics-and model-independent stabilization of source size relying only on previously existing instrumentation. Such feed-forward correction based on a neural network that can be continuously online-retrained achieves source size stability as low as 0.2 µm (0.4%) rms which results in overall source stability approaching the sub-percent noise floor of the most sensitive experiments.

Comparative study of hard X-ray undulator beamline performance in the Korean 4GSR and the PLS-II

“…After innovative advances in deep supervised learning, machine learning (ML) has been perceived as one of the major disruptive technologies in the 21th century for recognizing patterns in big data [24]. The synchrotrons' accelerator parts have already approached the ML method to stabilize source size [25]. Given the fact that synchrotrons have encountered the issue of big data [26], the beamline parts also need to consider machine learning, especially in data collection, data reduction, and data analysis, for better optimization of the beamline or even discovering patterns hidden in the experimental data, which would be too much to handle otherwise.…”

Machine learning for orders of magnitude speedup in multiobjective optimization of particle accelerator systems

“…Then the model must be updated in a reliable fashion without manual intervention. Online retraining has been demonstrated for a feed-forward correction scheme to improve source size stability at the ALS [55]. However, for accelerators that frequently switch operating conditions and have very large operating ranges automatic retraining when a high prediction error is observed could result in a loss of valuable information about previously visited machine states.…”

Supervised learning-based reconstruction of magnet errors in circular accelerators

“…ML techniques have found their application in a wide range of accelerator control and optimization tasks [9][10][11][12], including early works on orbit corrections using artificial neural networks [13,14] and Bayesian approach for linear optics corrections [15]. More recent advances on building surrogate models using supervised learning are presented in [16,17]. In the following, we present the methodology of utilizing the supervised learning approach and linear regression for quadrupole field errors estimation based on measured optics perturbations.…”

“…Fig 16. Reconstruction of horizontal normalized dispersion deviations in beam 1 from noisy phase advance data.…”