Multiobjective optimization of the dynamic aperture using surrogate models based on artificial neural networks

Abstract

Modern synchrotron light source storage rings, such as the Swiss Light Source upgrade (SLS 2.0), use multi-bend achromats in their arc segments to achieve unprecedented brilliance. This performance comes at the cost of increased focusing requirements, which in turn require stronger sextupole and higher-order multipole fields for compensation and lead to a considerable decrease in the dynamic aperture and/or energy acceptance. In this paper, to increase these two quantities, a multi-objective genetic algorithm (MOGA) is combined with a modified version of the well-known tracking code tracy. As a first approach, a massively parallel implementation of a MOGA is used. Compared to a manually obtained solution this approach yields very good results. However, it requires a long computation time. As a second approach, a surrogate model based on artificial neural networks is used in the optimization. This improves the computation time, but the results quality deteriorates. As a third approach, the surrogate model is re-trained during the optimization. This ensures a solution quality comparable to the one obtained with the first approach while also providing an order of magnitude speedup. Finally, good candidate solutions for SLS 2.0 are shown and further analyzed.

X-ray FEL linear accelerator design via start-to-end global optimization

“…In this study, we used an improved multi-objective optimizer based on the differential evolution method in the accelerator global beam dynamics design optimization. In the future study, we would like to further improve the computational speed in the global design optimization by exploring methods to include surrogate models in the optimizer [34,35].…”

2022 Review of Data-Driven Plasma Science

“…Neural network-based surrogate models can be trained to quickly map between accelerator parameters and beam properties, providing faster estimates than possible with computationally expensive physics models. Surrogate models can also be used to generate data sets for ML training and for optimization studies [253][254][255][256][257][258][259].…”

Machine learning enabled fast evaluation of dynamic aperture for storage ring accelerators

“…To reduce the computational cost, machine learning (ML) techniques have recently inspired a surge of applications to DA evaluation, especially in the time-consuming DA optimization studies [29][30][31][32][33]. These proposals focus mainly on learning the map between the magnetic lattice settings of the storage ring (e.g., the strengths of magnets) and the corresponding DA size.…”

“…As a result, the ML model often predicts accurately for a small variation range in the variable space and/or for a small number of control variables. Although this limitation can be mitigated to some degree by continuously retraining the ML model with new data samples (will take extra computing time) [33], one needs to always concern about the DA prediction accuracy when applying the ML model to a new storage ring design, whose lattice settings are away from the distribution of the existing training data.…”