Speaker
Description
The University of Hawai‘I at Mānoa (UHM) linac delivers up to 45 MeV electron beams to a Free-Electron Laser(FEL) oscillator.
As the linac is being recommissioned for renewed FEL operation, we are developing simulation and optimization tools to recover operational settings and to explore the landscape of beam-manipulation techniques for future experimental apparatus.
This paper benchmarks classical deterministic methods and Bayesian optimization (BO) algorithms on three representative beam-optics tuning scenarios using a beam dynamics simulation model developed inhouse.
For problems with only a few free parameters, classical methods converge reliably, while finite-difference derivative information improves the performance of constrained
gradient-based solvers.
For the higher-dimensional case, BO with a Gaussian Process (GP) surrogate and SOBOL initialization provides a more robust path toward convergence.
The emphasis is on the number of optimization iterations required for each scenario, in order to anticipate the computational cost of applying the same workflow to higher-fidelity models.
Funding Agency
The U.S. Department of Energy, Office of the Science, under Contract No. DE-SC0025583
| In which format do you inted to submit your paper? | LaTeX |
|---|---|
| Preprint marking on your proceeding paper | I do not wish my paper to be marked as preprint. |
| I no longer wish to present this contribution, please withdraw it. | Keep my contribution |