Speaker
Description
The design of multipole and other magnets for accelerators is typically an iterative process in which the magnet geometry is optimized for the required beam dynamics properties. Modelling the field for a given geometry can be computationally expensive, so exploring the parameter space can be a time-consuming procedure. The task is particularly challenging when complex field properties are needed (for example, in magnets with several multipole components or with longitudinal field variation). Combined function magnets with several multipole components are particularly useful in accelerators with tight spatial constraints such as an X-ray Free Electron Laser (XFEL). Surrogate models using neural networks can provide a way of rapidly generating possible magnet geometries for given field or beam dynamics requirements. In this contribution, we discuss how machine learning tools may be used to improve the efficiency of the design process for complex accelerator magnets, and present results from a case study based on a combined function magnet for the beam spreader in a future XFEL.
Funding Agency
This work is supported by Science and Technology Facilities Council, UK, through a grant awarded to The Cockcroft Institute
| In which format do you inted to submit your paper? | LaTeX |
|---|
