Latent Bayesian optimization for the autonomous alignment of synchrotron beamlines

Morris, Thomas W.; Du, Yonghua; Fedurin, Mikhail; Giles, Abigail C.; Moeller, Paul; Nash, Boaz; Rakitin, Max  .; Romasky, Brianna; Walter, Andrew L.; Wilson, Noah; Wojdyla, Antoine

doi:10.1117/12.2677895

Cited by 3 publications

(2 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Typically, these methods involve running the optimiser on real-time measurements acquired during beamline operation, iteratively refining the beam position until the optimiser converges to an optimum. Further work in this regard has been undertaken demonstrating the use of Bayesian optimization for the alignment of beamlines and demonstrated using a digital twin (Morris et al, 2023).…”

Section: Introductionmentioning

confidence: 99%

Automated spectrometer alignment via machine learning

Feuer-Forson,

Hartmann,

Mitzner

et al. 2024

J Synchrotron Radiat

View full text Add to dashboard Cite

During beam time at a research facility, alignment and optimization of instrumentation, such as spectrometers, is a time-intensive task and often needs to be performed multiple times throughout the operation of an experiment. Despite the motorization of individual components, automated alignment solutions are not always available. In this study, a novel approach that combines optimisers with neural network surrogate models to significantly reduce the alignment overhead for a mobile soft X-ray spectrometer is proposed. Neural networks were trained exclusively using simulated ray-tracing data, and the disparity between experiment and simulation was obtained through parameter optimization. Real-time validation of this process was performed using experimental data collected at the beamline. The results demonstrate the ability to reduce alignment time from one hour to approximately five minutes. This method can also be generalized beyond spectrometers, for example, towards the alignment of optical elements at beamlines, making it applicable to a broad spectrum of research facilities.

show abstract

Section: Introductionmentioning

confidence: 99%

Automated spectrometer alignment via machine learning

Feuer-Forson,

Hartmann,

Mitzner

et al. 2024

J Synchrotron Radiat

View full text Add to dashboard Cite

show abstract

“…An ML framework well-suited for expensive-to-sample functions is Bayesian optimization, which performs well with no prior information on optimization problems that are expensive-to-sample, high-dimensional, and potentially very noisy. Bayesian optimization has been applied in such a wide variety of contexts as synchrotron light sources [16,17], free-electron lasers [18], particle colliders [19], and laser-plasma-based ion sources [20]. These implementations, however, are typically applicable to single experiments; indeed, much of the difficulty in implementing machine learning solutions to any problem is the trade-off of specificity and generality where an algorithm that is specific enough to be effective in some context is too specific to be applied generally.…”

Section: Introductionmentioning

confidence: 99%

On-the-fly optimization of synchrotron beamlines using machine learning

Morris

Rakitin

Giles

et al. 2022

Optical System Alignment, Tolerancing, and Verification XIV

View full text Add to dashboard Cite

Autonomous methods to align beamlines can decrease the amount of time spent on diagnostics, and also uncover better global optima leading to better beam quality. The alignment of these beamlines is a high-dimensional, expensive-to-sample optimization problem involving the simultaneous treatment of many optical elements with correlated and nonlinear dynamics. Bayesian optimization is a strategy of efficient global optimization that has proved successful in similar regimes in a wide variety of beamline alignment applications, though it has typically been implemented for particular beamlines and optimization tasks. In this paper, we present a basic formulation of Bayesian inference and Gaussian process models as they relate to multiobjective Bayesian optimization, as well as the practical challenges presented by beamline alignment. We show that the same general implementation of Bayesian optimization with special consideration for beamline alignment can quickly learn the dynamics of particular beamlines in an online fashion through hyperparameter fitting with no prior information. We present the implementation of a concise software framework for beamline alignment and test it on four different optimization problems for experiments at x-ray beamlines of the National Synchrotron Light Source II and the Advanced Light Source and an electron beam at the Accelerator Test Facility, along with benchmarking on a simulated digital twin. We discuss new applications of the framework, and the potential for a unified approach to beamline alignment at synchrotron facilities.

show abstract