Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKBP: Arbeitskreis Beschleunigerphysik
AKBP 6: Synchrotron Radiation Sources (SR and FEL)
AKBP 6.1: Vortrag
Dienstag, 20. März 2018, 16:30–16:45, NW-Bau - HS5
Optimization of Synchrotron Light Sources using Machine Learning — •Tobias Boltz1, Edmund Blomley2, Erik Bründermann2, Patrik Schönfeldt2, Marcel Schuh2, Minjie Yan2, and Anke-Susanne Müller1,2 — 1LAS, KIT, Karlsruhe, Germany — 2IBPT, KIT, Karlsruhe, Germany
The operation of particle accelerators often requires manual fine-tuning to achieve optimal conditions. For synchrotron light sources in particular, the machine settings have to be additionally tailored to the particular needs of different users, i.e. specific applications and experiments. Typical requirements concentrate e.g. on the intensity and temporal resolution of the generated light pulses. The optimization of these parameters yields challenging demands on beam characteristics and dynamics. As these are controlled by a multitude of different knobs (e.g. magnetic field strengths), optimization is a highly non-trivial problem. However, with recent developments in computer science similar problems have been solved in various fields by applying machine learning techniques. These are enabled by increasing amounts of data being collected as well as steadily rising computing power. In this contribution, we present recent efforts to employ machine learning techniques to optimize accelerator operation at the storage ring KARA (KArlsruhe Research Accelerator) as well as the linear accelerator and test facility FLUTE at KIT.