Speaker
Description
Optimising SRF photoinjectors is a challenging task due to the high-dimensional, nonlinearly coupled parameters and competing objectives such as transverse emittance and bunch length. Conventional methods such as manual tuning or MOGA require thousands of evaluations and are impractical for routine operation or computationally expensive simulations. This work presents a multi-objective Bayesian optimisation (MOBO) approach that uses Gaussian-process surrogate models and tunable, uncertainty-aware acquisition functions to identify Pareto-optimal solutions in an order of magnitude fewer evaluations. When applied to the 1.4-cell SRF photoinjector at SEALab, and the 1.6-cell SRF gun and 20m injector beamline for EuXFEL, this optimisation outperforms MOGA in solution-efficiency and provides interpretable sensitivity information for injector tuning. These results demonstrate the potential of MOBO as an efficient, machine-ready strategy for SRF photoinjector optimisation.
| In which format do you inted to submit your paper? | LaTeX |
|---|