Orbit correction based on machine learning
-
摘要: 轨道校正是加速器束流调节最基本的步骤之一，也是目前各加速器实验室共同面对的问题之一。在传统方法中，线性代数工具被应用于各种类型的响应矩阵，以解决响应矩阵的奇异性等问题。提出一种基于机器学习的加速器轨道校正方法，可以避免处理响应矩阵的问题通过直接读取BPM数据和校正磁铁强度值实时构建机器学习模型快速地对轨道进行修正。对机器学习的轨道校正方法进行了介绍，并从数学公式、算法模型、在模拟和真实数据上的测试等方面对该方法进行了讨论。结果表明，在误差范围内该方法能有效的对加速器束流轨道进行校正。Abstract: Orbit correction is one of the most fundamental processes used for beam control in accelerators. Algorithms have been developed at various laboratories to meet specific demands. Typically, linear algebraic tools are applied to various response matrices to solve related problems. However, there are still many problems faced by orbit correction algorithms such as lengthy measurement and computation time. A new approach based on machine learning to develop an orbit correction program is introduced. In this method a machine learning program is trained with correctors data and BPMs data for applying to orbit correction. Mathematical formulation, algorithms prototyped and tested on simulated and real data, and future possibilities are discussed.
-
Key words:
- orbit correction /
- machine learning /
- operation data /
- magnet error /
- data cleaning
-
图 1 测试数据集中模型输出校正磁铁值和真实值的比对
Figure 1. Difference of output value and real value in test data set
图 5 SVD方法校正后轨道和轨道偏离统计
Figure 5. Orbit and statistics of orbit deviation with SVD correction
图 6 机器学习方法和SVD方法校正磁铁强度比较
Figure 6. Correctors’ value of machine learning (ML) correction and SVD correction
图 9 校正磁铁真实电流值的波动及与设定值的差异
Figure 9. Variation of correctors’ real value and difference between real value and setting value
表 1 磁铁误差设置范围
Table 1. Magnet error parameters
∆x/mm ∆y/mm ∆z/mm ∆θx/mrad ∆θy/mrad ∆θz/mrad dipole 0.10 0.10 0.10 0.10 0.10 0.05 quadrupole 0.10 0.10 0.25 0.25 0.25 0.10 
下载: 导出CSV
表 2 数据集参数
Table 2. Data set parameters
features numbers train data test data total data BPM 66 350 150 500 corrector 33 350 150 500
下载: 导出CSV
-
[1] Hou Jie, Liu Guiming. Transport line orbit distortion correction based on response matrix and SVD algorithm[J]. High Energy Physics and Nuclear Physics-Chinese Edition, 2006, 30: 37-39. [2] Chung Y, Decker G, Evans K. Closed orbit correction using singular value decomposition of the response matrix[C]//Proceedings of 1993 IEEE Particle Accelerator Conference. 1993: 2263-2265. [3] Grote H, Iselin F C, Keil E, et al. The MAD program[C]//Particle Accelerator Conference. 1989. [4] Chu Zhongming, Xiao Dengjie, Qiao Yusi, et al. Machine learning applications for particle accelerators[J]. Frontiers of Data and Computing, 2019, 1(2): 110-120. [5] 魏源源. BEPCⅡ储存环基于响应矩阵的束流参数校正[D]. 北京: 中国科学院研究生院, 2007: 30-36.Wei Yuanyuan. Optics correction using response matrix at BEPCⅡ storage rings[D]. Beijing: Institute of High Energy Physics, Chinese Academy of Sciences, 2007: 30-36 [6] 刘渭滨. BEPCⅡ直线加速器束流光学轨道校正程序设计[D]. 北京: 中国科学院研究生院, 2004: 13-17.Liu Weibin. BEPCⅡ Linac optics calculations and orbit correction program design[D]. Beijing: Institute of High Energy Physics, Chinese Academy of Sciences, 2004: 13-17 [7] Xiao Dengjie, Chu Zhongming, Qiao Yusi. Orbit correction with machine learning[C]//Proceedings of the 10th International Particle Accelerator Conference. 2019: 2608-2610. [8] Bai Sha, Gao Jie, Geng Huiping. Orbit correction in CEPC[C]//Proceedings of the 6th International Particle Accelerator Conference. 2015: 2022-2024. [9] 乔予思. 大型加速器数据存储查询和处理的软件技术研究[D]. 北京: 中国科学院大学, 2019: 89-109.Qiao Yusi. Study of large particle accelerator’s data archiving, retrieving and processing software technology[D]. Beijing: University of Chinese Academy of Sciences, 2019: 89-109 [10] Mcdonald G C. Ridge regression[J]. Wiley Interdisciplinary Reviews: Computational Statistics, 2010, 1: 93-100. [11] 杨楠. 岭回归分析在解决多重共线性问题中的独特作用[J]. 统计与决策, 2004(3):14-15. (Yang Nan. Ridge regression for nonorthogonal problems[J]. Statistics and Decision, 2004(3): 14-15 doi: 10.3969/j.issn.1002-6487.2004.03.007 [12] Hoerl A E, Kennard R W. Ridge regression: biased estimation for nonorthogonal problems[J]. Technometrics, 2012(12): 55-67. -
点击查看大图
计量
- 文章访问数: 1762
- HTML全文浏览量: 400
- PDF下载量: 136
- 被引次数: 0
