Reconstruction of Storage Ring 's Linear Optics with Bayesian Inference (1902.11157v3)

Abstract: A novel approach of accurately reconstructing storage ring's linear optics from turn-by-turn (TbT) data containing measurement error is introduced. This approach adopts a Bayesian inference based on the Markov Chain Monte-Carlo (MCMC) algorithm, which is widely used in data-driven discoveries. By assuming a preset accelerator model with unknown parameters, the inference process yields the their posterior distribution. This approach is demonstrated by inferring the linear optics Twiss parameters and their measurement uncertainties using a set of data measured at the National Synchrotron Light Source-II (NSLS-II) storage ring. Some critical effects, such as radiation damping rate, decoherence due to nonlinearity and chromaticity can also be included in the model and inferred. These effects are usually ignored in existing approaches. One advantage of the MCMC based Bayesian inference is that it doesn't require a large data pool, thus a complete optics reconstruction can be accomplished from a limited number of turns in a single data snapshot, before a significant machine drift can happen. The precise reconstruction of the parameter in accelerator model with the uncertainties is crucial prior information for applying the them to improve machine performance.