留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

机器学习在储存环轨道校正中的应用研究

李瑞淳 张庆磊 米清茹 姜伯承 王坤 李昌亮 赵振堂

李瑞淳, 张庆磊, 米清茹, 等. 机器学习在储存环轨道校正中的应用研究[J]. 强激光与粒子束, 2021, 33: 034007. doi: 10.11884/HPLPB202133.200318
引用本文: 李瑞淳, 张庆磊, 米清茹, 等. 机器学习在储存环轨道校正中的应用研究[J]. 强激光与粒子束, 2021, 33: 034007. doi: 10.11884/HPLPB202133.200318
Li Ruichun, Zhang Qinglei, Mi Qingru, et al. Application of machine learning in orbital correction of storage ring[J]. High Power Laser and Particle Beams, 2021, 33: 034007. doi: 10.11884/HPLPB202133.200318
Citation: Li Ruichun, Zhang Qinglei, Mi Qingru, et al. Application of machine learning in orbital correction of storage ring[J]. High Power Laser and Particle Beams, 2021, 33: 034007. doi: 10.11884/HPLPB202133.200318

机器学习在储存环轨道校正中的应用研究

doi: 10.11884/HPLPB202133.200318
基金项目: 国家重点研发计划项目(2016YFA0402001);中国科学院青年创新促进会项目(2020287)
详细信息
    作者简介:

    李瑞淳(1996—),男,硕士研究生,从事储存环轨道校正研究;lirch@shanghaitech.edu.cn

    通讯作者:

    张庆磊(1984—),男,博士,副研究员,从事加速器束流动力学研究;zhangqinglei@zjlab.org.cn

  • 中图分类号: TL54+4

Application of machine learning in orbital correction of storage ring

  • 摘要: X射线同步辐射光源,是现代科学研究中最强大的工具之一。位于中国上海的上海光源,是一台能量为3.5 GeV的先进的第三代中能同步辐射光源。第三代同步辐射光源要提供高亮度、高稳定性的同步辐射来满足实验条件要求苛刻的前沿研究,因此对束流的轨道稳定性有很高的要求。为此,采用机器学习算法进行电子束轨道的控制和反馈。这种基于神经网络的轨道校正方法不依赖于具体的响应矩阵,建立非线性映射关系,并且还可以进行连续的在线再训练,对上海光源的轨道校正和提高束流轨道稳定性有重要意义。
  • 图  1  神经网络原理示意图(x代表输入信号,o代表输出信号)

    Figure  1.  Schematic of neural network (x represents the input signal,o represents the output signal)

    图  2  训练数据与测试数据的绝对平均误差

    Figure  2.  Mean absolute errors of training data and test data

    图  3  训练数据与测试数据的损失函数

    Figure  3.  Loss function of training data and test data

    图  4  机器学习程序的轨道反馈效果。(图中不同颜色曲线代表了不同的BPM数据,第一次扰动对应C01VCM2,第二次扰动对应C04VCM2)

    Figure  4.  Orbit feedback effect of machine learning program. (The different color curves in the figure represent different BPM data. The first disturbance corresponds to corrector C01VCM2 and the second disturbance corresponds to corrector C04VCM2)

    图  5  机器学习程序运行过程中校正铁的变化(图中不同颜色曲线代表了不同的校正铁电流数据,第一次扰动对应C01VCM2,第二次扰动对应C04VCM2)

    Figure  5.  Changes in correctors during the operation of machine learning program.(The different color curves in the figure represent different correctors data. The first disturbance corresponds to C01VCM2 and the second disturbance corresponds to C04VCM2)

    图  6  慢反馈程序的轨道反馈效果

    Figure  6.  Orbit feedback effect of slow feedback program

    图  7  慢反馈程序运行过程中校正铁的变化

    Figure  7.  Changes in correctors during the operation of slow feedback program

    图  8  机器学习程序的轨道反馈效果

    Figure  8.  Orbit feedback effect of machine learning program (corretor current loaded with random disturbance)

    图  9  机器学习程序运行过程中校正铁的变化

    Figure  9.  Changes in correctors during the operation of machine learning program (corretor current loaded with random disturbance)

    图  10  慢反馈程序的轨道反馈效果

    Figure  10.  Orbit feedback effect of slow feedback program (corretor current loaded with random disturbance)

    图  11  慢反馈程序运行过程中校正铁的变化

    Figure  11.  Changes in correctors during the operation of slow feedback program (corretor current loaded with random disturbance)

  • [1] Jiang Bocheng, Liu Guimin, Zhao Zhentang. Simulation of a transverse feedback system for the SSRF storage ring[J]. High Energy Physics and Nuclear Physics, 2007, 31(10): 956-961.
    [2] Jiang Bocheng, Lin Guoqiang, Wang Baoliang, et al. Multi-bunch injection for SSRF storage ring[J]. Nuclear Science and Techniques, 2015, 26: 050101.
    [3] Zhang Q, Jiang B C, Tian S Q, et al. Study on beam dynamics of a Knot-APPLE undulator proposed for SSRF[C]//Proceedings of the 6th International Particle Accelerator Conference. 2015: 1669-1671.
    [4] Jiang Bocheng, Zhao Zhentang, Liu Guimin. Study of Touschek lifetime in SSRF storage ring[J]. High Energy Physics and Nuclear Physics, 2006, 30(7): 693-698.
    [5] Jiang Bocheng, Xia Guoxing, Han Lifeng, et al. Investigation of fast ion instability in SSRF[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2010, 614(3): 331-334.
    [6] 卜令山, 赵振堂, 殷立新, 等. 第三代同步辐射光源储存环支撑组件振动控制研究[J]. 中国物理C, 2008, 32(s1):37-39. (Bu Lingshan, Zhao Zhentang, Yin Lixin, et al. Vibration control research for the 3rd generation synchrotron light source storage ring mechanical components[J]. Chinese Physics C, 2008, 32(s1): 37-39
    [7] Zhao Z T, Xu H J, Ding H. Commissioning of the Shanghai Light Source[J]. Energy, 2015, 3: 3-51.
    [8] Nagaoka R, Bocchetta C J, Iazzourene F, et al. Orbit correction in ELETTRA[C]//Proc 4th EPAC. 1994: 1009.
    [9] Tsai HJ, Chang H P, Chou P J, et al. Closed orbit correction of TPS storage ring[C]//Proceedings of EPAC 2006.2006: 2029-2031.
    [10] Li Jingyi, Liu Gongfa, Li Weimin, et al. Closed orbit correction of HLS storage ring[C]//Proceedings of the 2001 Particle Accelerator Conference. Chicago: IEEE, 2001: 1255-1257.
    [11] Wang Faya, Song Minghao, Edelen A, et al. Machine learning for design optimization of storage ring nonlinear dynamics[DB/OL]. arXiv preprint arXiv: 1910.14220, 2019.
    [12] Leemann S C, Liu S, Hexemer A, et al. Demonstration of machine learning-based model-independent stabilization of source properties in synchrotron light sources[J]. Physical Review Letters, 2019, 123: 194801. doi: 10.1103/PhysRevLett.123.194801
    [13] 刘祖平. 同步辐射光源物理引论[M]. 合肥: 中国科学技术大学出版社, 2009: 161-196.

    Liu Zhuping. Introduction to physics of synchrotron radiation source[M]. Hefei: University of Science and Technology of China Press, 2009: 161-196).
    [14] Chung Y, Decker G, Evans K, et al. Global DC closed orbit correction experiments on the NSLS X-ray ring and SPEAR[C]//Proceedings of International Conference on Particle Accelerators. Washington: IEEE, 1993: 2275-2277.
    [15] LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444. doi: 10.1038/nature14539
    [16] 陈亚秋, 陈德钊, 胡上序, 等. 多层前传神经网的广义误差反传训练与模式分类[J]. 模式识别与人工智能, 1996, 9(2):161-165. (Chen Yaqiu, Chen Dezhao, Hu Shangxu, et al. Generalized error back-propagation training for multi-layered feedforward neural nets[J]. Pattern Recognition and Artificial Intelligence, 1996, 9(2): 161-165
    [17] LeCun Y, Boser B, Denker J S, et al. Backpropagation applied to handwritten zip code recognition[J]. Neural Computation, 1989, 1(4): 541-551. doi: 10.1162/neco.1989.1.4.541
    [18] Ruder S. An overview of gradient descent optimization algorithms[J]. arXiv preprint arXiv: 1609.04747, 2016.
    [19] Chollet F, Others. Keras: the python deep learning library[J]. Astrophysics Source Code Library, 2018, 1806: 1022.
    [20] Abadi M, Barham P, Chen Jianmin, et al. TensorFlow: a system for large-scale machine learning[C]//Proceedings of the 12th USENIX Conference on Operating Systems Design and Implementation. 2016: 265-283.
    [21] Dean J, Corrado G S, Monga R, et al. Large scale distributed deep networks[C]//Proceedings of the 25th International Conference on Neural Information Processing Systems. 2012: 1223-1231.
    [22] Zhang Chiyuan, Liao Qianli, Rakhlin A, et al. Theory of deep learning IIb: optimization properties of SGD[DB/OL]. arXiv preprint arXiv: 1801.02254, 2018.
    [23] Bottou L, Curtis F E, Nocedal J. Optimization methods for large-scale machine learning[J]. SIAM Review, 2016, 60(2): 223-311.
    [24] Nair V, Hinton G E. Rectified linear units improve restricted Boltzmann machines[C]//Proceedings of the 27th International Conference on Machine Learning. 2010: 807-814.
    [25] Glorot X, Bordes A, Bengio Y. Deep sparse rectifier neural networks[J]. Journal of Machine Learning Research, 2011, 15: 315-323.
    [26] Nowlan S J, Hinton G E. Simplifying neural networks by soft weight-sharing[J]. Neural Computation, 1992, 4(4): 473-493. doi: 10.1162/neco.1992.4.4.473
  • 加载中
图(11)
计量
  • 文章访问数:  1444
  • HTML全文浏览量:  310
  • PDF下载量:  144
  • 被引次数: 0
出版历程
  • 收稿日期:  2020-11-23
  • 修回日期:  2021-01-19
  • 网络出版日期:  2021-03-30
  • 刊出日期:  2021-03-05

目录

    /

    返回文章
    返回