Speaker
Li Zeng
(Institute of Advanced Science Facilities)
Description
Traditionally, the optimization of Free-Electron Laser (FEL) facilities has been performed manually by the FEL operators. This approach proves to be time-consuming due to the multitude number of parameters that require adjustments. The results of this manual optimization are highly contingent upon the operator’s experience. To address these challenges, the implementation of machine learning algorithms offers a rapid and adaptable alternative for achieving global optimization of FEL performance within limited timeframes. In this paper, a surrogate model has been constructed using neural networks to expedite simulations. Additionally, the simulation results of FEL pulse energy optimization utilizing reinforcement learning are presented.
Authors
Li Zeng
(Institute of Advanced Science Facilities)
Xiaofan Wang
(Institute of Advanced Science Facilities)
Yong Yu
(Institute of Advanced Science Facilities)
Jitao Sun
(Dalian Institute of Chemical Physics)
Weiqing Zhang
(Institute of Advanced Science Facilities)
