TUP50 　ポスター①　 8月29日 14号館1443教室 13:30-15:30

Linear optics correction of storage ring with closed orbit data based on machine learning

○Elham Salehi (UVSOR), Masahiro Katoh (UVSOR and HiSOR)

The LOCO algorithm based on orbit response matrix (ORM), which is the change in orbit at beam position monitors (BPMs) with changes in steering magnets, has been widely used for linear optics correction in storage rings. It computes the orbit response matrix and fits it to the experiment data linearly as making quadrupole strengths and others as free parameters. In order to solve the difficulty on convergence of ORM method when the change of focusing optics or the effect of nonlinear optics on the closed orbit is large, we propose a method based on machine learning. The closed orbit of the design lattice produced by each steering magnet at BPMs is calculated using ELEGANT simulation code which fully includes non-linear components and is fitted to the measured closed orbit data in the experiment. This method can nonlinearly compute the correction parameters of the lattice by minimizing the discrepancies between measured and predicted closed orbits. We have applied this method to the lattice of UVSOR_III. The latest results will be presented at the meeting.