Speaker
Xinzhong Liu
(Shanghai Advanced Research Institute, Chinese Academy of Sciences)
Description
A deep-learning-based feedforward scheme has been developed to compensate insertion-device (ID) effects in the Shanghai Synchrotron Radiation Facility (SSRF). Neural networks predict orbit and betatron-coupling perturbations caused by ID gap and phase changes. The orbit model reduces residual closed-orbit distortion (COD) to below 2 um and shortens preparation time by about a factor of 50 compared with conventional feedforward-table measurements. A coupling model trained with turn-by-turn (TBT) beam-position-monitor (BPM) data reaches an R^2 value above 0.95. These results show that deep learning can support fast and reproducible ID compensation during light-source operation.
| Paper status | Resubmitted proceeding files received and assigned to an editor. Accepted. |
|---|
Author
Xinzhong Liu
(Shanghai Advanced Research Institute, Chinese Academy of Sciences)
Co-authors
LINGLONG MAO
(Shanghai Institute of Applied Physics, Chinese Academy of Sciences, University of Chinese Academy of Sciences)
Liyuan Tan
(Shanghai Advanced Research Institute, Chinese Academy of Sciences)
Yihao Gong
(Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai Institute of Applied Physics, Chinese Academy of Sciences, University of Chinese Academy of Sciences)
Shunqiang Tian
(Shanghai Advanced Research Institute, Chinese Academy of Sciences, University of Chinese Academy of Sciences)