Optimizing the lattice design of a diffraction-limited storage ring with a rational combination of particle swarm and genetic algorithms

Abstract

Abstract:In the design of a diffraction-limited storage ring (DLSR) consisting of compact multi-bend achromats (MBAs), it is challenging to simultaneously achieve an ultralow emittance and a satisfactory nonlinear performance, due to extremely large nonlinearities and limited tuning ranges of the element parameters. Nevertheless, taking the High Energy Photon Source (HEPS) as an example, we demonstrate that the potential of a DLSR design can be explored with a successive and iterative implementation of the particle swarm optimization (PSO) and multi-objective genetic algorithm (MOGA). It turns out that with a hybrid MBA lattice, it is feasible for the HEPS to attain a natural emittance of about 50 pm.rad, and meanwhile, realize a sufficient ring acceptance for beam accumulation with an on-axis longitudinal injection scheme. Particularly, this study indicates that a rational combination of the PSO and MOGA is more effective than either of them alone in approaching the true global optima, for an explorative multi-objective problem with many optimizing variables and local optima.

Genetic algorithm enhanced by machine learning in dynamic aperture optimization

“…Population-based optimization techniques, such as evolutionary (genetic) [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16] and particle swarm [17][18][19] algorithms, have become popular in modern accelerator design. They are effective design tools for both linear and nonlinear system optimization.…”

Statistical analysis of the limitation of half integer resonances on the available momentum acceptance of the High Energy Photon Source

“…Apart from the presented HEPS design, other HEPS designs [26] with similar layout but with different emittance and nominal tunes were also tested. Basically the same criterion was obtained for safely crossing the HIRs.…”

Physics issues of the diffraction-limited storage ring light source

“…、模拟退火算法、蚁群算法等. MOPSO 能够在解空间中产生更多的多样性, 全局搜索能力强, 但收敛速度较慢; 而MOGA拥有更快的收敛速度, 却容 易在解多样性不足时陷入局部收敛 [43] . 模拟退火算法 局部搜索能力强, 但很容易陷入局部收敛.…”

“…(a) Optics parameters of complex bend Ⅰ; (b) the magnet layout of complex bend Ⅰ; (c) the magnet layout of complex bend Ⅱ http://engine.scichina.com/doi/10.1360/TB-2020-0165 现出更快的收敛速度. 研究表明, 将两种优化方法交替 使用, 其效果优于单独使用其中任何一种方法 [43] . [50,51] , 以及三高频频率的纵向注入方法 [52] .…”