基于支持向量拟合代理模型的卫星多学科设计优化  

Satellite Multidisciplinary Design Optimization Based on Support Vector Regression Surrogate Model

作　　者：史人赫 刘莉[1] 龙腾[1] 郭晓松[1] 彭磊[1] 

机构地区：[1]北京理工大学宇航学院,北京100081

出　　处：《南京航空航天大学学报》2014年第3期481-486,共6页Journal of Nanjing University of Aeronautics & Astronautics

基　　金：国家自然科学基金(51105040;11372036)资助项目

摘　　要：为了提高卫星的设计质量与效率,建立以卫星总质量最小为优化目标的多学科设计优化问题,主要考虑轨道、有效载荷、电源和结构4个分系统的设计并梳理其耦合关系,整理并建立了较为详细且贴近工程实际的学科分析模型。本文提出了基于支持向量拟合(Support vector regression,SVR)代理模型的优化策略,并将其应用于海洋卫星多学科设计优化中。本文算例卫星参考海洋一号卫星(HY-1),优化后整星质量相对于初始质量下降了约14.1%。优化结果表明了卫星多学科优化设计(Multidisciplnary design optimization,MDO)模型的合理性和该优化策略的有效性,为进一步探索代理模型技术在卫星MDO设计当中的应用奠定了基础。To improve the quality and efficiency in satellite design, the satellite multidisciplinary design optimization is presented with the object of minimizing total mass considering orbit, payload, power supply and structure. The paper combs the coupling tailed and practical disciplinary analysis model. The among the four subsystems and establishes the de optimization strategy is proposed based on support vector regresstion （SVR） method. The strategy is applied to multidisciplnary design optimization （MDO） of the marine satellite. The studied satellite takes Ocean 1 satellite （HY-1） as a reference, and the total mass of the satellite is reduced by 14.1 after optimization compared with the initial value. The optimization result proves the rationality of the satellite＇s MDO model and the effectiveness of the optimization strategy. It is of vital importance to the future work of the application of SVR metamodel in satellite MDO design.

关 键 词：海洋卫星 多学科优化设计 支持向量拟合 代理模型 

分 类 号：V19[航空宇航科学与技术—人机与环境工程]