EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Four-dimensional beam matching and optimisation in a free-electron laser via multi-objective optimisation

Runfeng Luo, Michael P Backhouse, Zulfikar Najmudin

Machine Learning: Science and Technology · 01.08.2026 · 7 (4) · 045064

DOI: 10.1088/2632-2153/ae97b4 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Machine Learning: Science and Technology, online erschienen am 21.08.2026. Verfasst von Runfeng Luo und 2 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Science and Technology Facilities Council.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Transverse beam matching and optimisation are standard procedures in accelerator design, aimed at minimising beam losses and maximising beam brightness. Studies in accelerator physics often rely on single-objective formulations with predefined weighting factors, which can introduce bias and limit control over competing design requirements. In this work, we investigate the impact of objective formulation and optimiser choice by comparing single-objective and multi-objective genetic-algorithm-based optimisation with a gradient-based single-objective method for four-dimensional transverse beam matching. The SPARC free-electron laser beamline is used as a realistic case study. Within the genetic algorithm framework, the multi-objective formulation exhibits faster convergence than its single-objective counterpart and provides direct access to Pareto-optimal solutions, enabling transparent exploration of trade-offs between conflicting beam parameters. A knee-point selection strategy is employed to connect the Pareto front to practical operational requirements. Benchmarking against a gradient-based optimiser using a range of weighting factors reveals systematic stagnation, step-like penalty evolution, and frequent constraint violations, highlighting structural limitations of weight-based scalarisation. These results demonstrate that framing transverse beam matching as a multi-objective optimisation problem, addressed using genetic algorithms, improves robustness, convergence behaviour, and interpretability for free-electron laser and accelerator optimisation.

Förderung

  • Science and Technology Facilities Council

Zitieren

APA

Luo, R., Backhouse, M. P., & Najmudin, Z. (2026). Four-dimensional beam matching and optimisation in a free-electron laser via multi-objective optimisation. Machine Learning: Science and Technology, 7(4), 045064. https://doi.org/10.1088/2632-2153/ae97b4

Vancouver

Luo R, Backhouse MP, Najmudin Z. Four-dimensional beam matching and optimisation in a free-electron laser via multi-objective optimisation. Machine Learning: Science and Technology. 2026;7(4):045064. doi:10.1088/2632-2153/ae97b4

BibTeX
@article{Luo2026,
  author = {Luo, Runfeng and Backhouse, Michael P and Najmudin, Zulfikar},
  title = {Four-dimensional beam matching and optimisation in a free-electron laser via multi-objective optimisation},
  journal = {Machine Learning: Science and Technology},
  year = {2026},
  volume = {7},
  number = {4},
  pages = {045064},
  doi = {10.1088/2632-2153/ae97b4}
}
Kostenlos starten →