Two parameter optimization methods for large aperture mirror
-
摘要: 大口径反射镜在自重作用下发生变形,变形过大会引起镜片通光表面的峰谷值较大,难以满足光机装置打靶精度的要求。建立了典型反射镜结构的有限元仿真模型,以计算镜片表面峰谷值;优化了安装柱的位置、尺寸及镜片厚度等参数,使镜片通光表面形变峰谷值最小。分别采用了基于有限元模型的直接优化和基于代理模型的优化,均获得了结构最优参数,使形变峰谷值比初始值降低了74%。相比而言,基于代理模型的优化还可提供目标量随设计变量的变化情况以及变量灵敏度分析,更有利于设计决策。当设计变量个数多、范围宽时,可先根据先验知识或低精度代理模型缩小设计范围,然后建立紧缩范围的高精度代理模型并进行优化,可有效避免大量计算。根据“三圆柱”支撑结构的参数优化结果,建议采用“四圆柱”支撑设计,以获得更高的面形精度。Abstract: Much deformation of the large aperture mirror by deadweight leads to large PV (peak-to-valley) value of its aperture surface, and it is hard to ensure the point accuracy on laser beams of optical facility. Finite element model is built up to calculate the PV value of the mirror surface in this paper. Moreover, parameters including position and size of erection column as well as the mirror thickness are optimized in order to minimize the PV value of the aperture surface. Both the direct optimization based on finite element model and the optimization based on surrogate model are adopted herein, and the PV value is reduced by 74% of its initial value as a result. In contrast to direct optimization, optimization based on surrogate model conveniently provides the objective variation with design variables and sensitivity analysis as well, which brings much benefit for structure design. In case of numerous and wide-range design variables, it is suggested to firstly decrease the design range according to prior knowledge or low-fidelity surrogate model and then optimize the parameters based on high-fidelity surrogate model within the shrunk range in order to avoid large amount of calculation. Furthermore, a scheme of four-cylinder support is recommended to replace the three-cylinder support for smaller PV value according to the optimization result.
-
Key words:
- peak-to-valley value /
- finite element model /
- optimization /
- surrogate model /
- sensitivity analysis
-
表 1 设计变量的范围和初值
Table 1. The range and initial value of the design variables
variables lower bound/mm upper bound/mm initial value/mm h1 60 240 150 h2 60 240 150 w 100 160 100 D 30 80 60 T 70 100 85 表 2 优化结果对比
Table 2. Comparison of the optimization results
method PV value/nm design variables/mm h1 h2 w D T initial value 254.8 150.0 150.0 100.0 60.0 85.0 direct optimization 66.3 240.0 68.3 116.2 80.0 100.0 optimization based on whole-range surrogate model 85.3 205.4 63.3 106.6 79.6 99.9 optimization based on shrunk-range surrogate model 67.1 239.7 69.3 117.7 78.7 99.9 -
[1] Trummer D J, Foley R J, Shaw G S. Stability of optical elements in the NIF target area building[R]. UCRL-JC-129682, 1998. [2] 徐元利, 吴文凯, 陈学前, 等. ICF装置靶场结构总体稳定性设计[J]. 强激光与粒子束, 2009, 21(10): 1485-1488. http://www.hplpb.com.cn/article/id/2362Xu Yuanli, Wu Wenkai, Chen Xueqian, et al. Structural systemic stability design of inertial confinement fusion facility target area. High Power Laser and Particle Beams, 2009, 21(10): 1485-1488 http://www.hplpb.com.cn/article/id/2362 [3] 曹庭分, 熊召, 徐旭, 等. 大口径KDP晶体夹持方式对面形的影响[J]. 激光与光电子学进展, 2011, 48(8): 138-141. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ201108021.htmCao Tingfen, Xiong Zhao, Xu Xu, et al. Mounting ways and surface figure of large-aperture KDP crystal. High Power Laser and Particle Beams, 2011, 48(8): 138-141 https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ201108021.htm [4] 陈晓娟, 傅学农, 吴文凯, 等. 基于Pro/M的大口径反射镜的变形分析及其支撑优化[J]. 机械设计, 2008, 25(2): 68-70. doi: 10.3969/j.issn.1672-1616.2008.02.038Chen Xiaojuan, Fu Xuenong, Wu Wenkai, et al. Deformation analysis on large caliber reflector and its support optimization based on Pro/M. Journal of Machine Design, 2008, 25(2): 68-70 doi: 10.3969/j.issn.1672-1616.2008.02.038 [5] 陈晓娟, 王美聪, 吴文凯, 等. 大口径反射镜波前畸变控制技术[J]. 强激光与粒子束, 2011, 23(12): 3325-3328. http://www.hplpb.com.cn/article/id/5704Chen Xiaojuan, Wang Meicong, Wu Wenkai, et al. Wavefront distribution control for large aperture mirror. High Power Laser and Particle Beams, 2011, 23(12): 3325-3328 http://www.hplpb.com.cn/article/id/5704 [6] Zhu J D, Wang Y J, Collette M. A multi-objective variable-fidelity optimization method for genetic algorithms[J]. Engineering Optimization, 2014, 46(4): 521-542. doi: 10.1080/0305215X.2013.786063 [7] Li M, Li G, Azarm S. A Kriging meta-model assisted multi-objective genetic algorithm for design optimization[J]. Journal of Mechanical Design, 2008, 130(3): 1-10. [8] Dey A, Sarkar D. A note on the construction of orthogonal Latin hypercube designs[J]. Journal of Combinatorial Designs, 2016, 24(3): 105-111. [9] Georgiou S D, Stylianou S. Block-circulant matrices for constructing optimal Latin hypercube designs[J]. Journal of Statistical Planning and Inference, 2011, 141(5): 1933-1943. [10] Krige D G. A statistical approach to some basic mine valuation problems on the Witwatersrand[J]. Journal of the Chemical, Metallurgical and Mining Engineering Society of South Africa, 1951, 52(6): 119-139. [11] Forrester A I J, Sóbester A, Keane A J. Engineering design via surrogate modelling: A practical guide[M]. Chichester: John Wiley & Sons Ltd., 2008. [12] Helton J C, Johnson J D, Sallaberry C J, et al. Survey of sampling-based methods for uncertainty and sensitivity analysis[J]. Reliability Engineering and System Safety, 2006, 91(10/11): 1175-1209.