Design and Matching Study of LEBT Based on Surrogate Model

The low-energy beam transport section is an important part of particle accelerators whose design and optimization process usually requires substantial computation resources and time investment. We applies Latin hypercube sampling method to construct a Kriging surrogate model and verifies its accuracy. The results show that the root-mean-square (RMS) error of the overall prediction is below 0.01. The given output beam parameters are optimized by genetic algorithm, and the corresponding component parameters are determined reversely. The results indicate that 89% of the optimization deviations are below 0.001 while 100% of the optimization deviations are below 0.01. This method has high predictive accuracy and fast response time, significantly reducing the design and matching time of the low-energy transport section, and is expected to be applied to cyclotron injection, automated beam tuning, and other fields in the future.