面向火箭构型论证的运载能力快速分析方法  

Rapid Analysis Method of Carrying Capability for the Demonstration of Launch Vehicle Configuration

作　　者：刁尹 王永海 卢宝刚[3] 韦常柱[1] DIAO Yin;WANG Yonghai;LU Baogang;WEI Changzhu(School of Astronautics,Harbin Institute of Technology,Harbin 150001,China;Beijing Institute of Nearspace Vehicle’s Systems Engineering,Beijing 100076,China;Beijing Institute of Space Long March Vehicle,Beijing 100076,China)

机构地区：[1]哈尔滨工业大学航天学院,哈尔滨150001 [2]北京临近空间飞行器系统工程研究所,北京100076 [3]北京航天长征飞行器研究所,北京100076

出　　处：《宇航学报》2022年第6期723-731,共9页Journal of Astronautics

基　　金：省部级项目(JCKY2019208C017);省部级项目(SAST2019-005)。

摘　　要：针对不同构型与任务条件下的运载能力快速计算问题,提出了基于高斯函数和组合神经网络的速度损失计算方法,并基于此对运载能力进行了快速分析。首先,基于状态量解析解计算分析,采用高斯函数对核心的重力速度损失项进行拟合计算;同时,为提高多构型与多任务样本的采样密度、简化数据建模过程并增强方法适应性,采用径向基网络(RBF)与深度神经网络(DNN)的组合形式进行状态量的提取与回归分析;然后将任务约束转化为需要速度增量,通过数值迭代得到运载能力。仿真结果表明,此运载能力分析方法精度偏差约为0.35%,计算耗时小于2 s,可为运载火箭总体参数快速论证与任务规划研究提供理论支撑。Aiming at the problem of fast calculation of the carrying capacity under different launch vehicle configurations and task conditions,a speed loss calculation method based on the Gaussian function and combined neural network is proposed.Firstly,based on the analytic solutions of state variables,the Gaussian function is used to fit the core velocity loss from gravity.Meanwhile,in order to improve the sampling density of multi-configuration and multi-task samples,simplify the data modeling process and enhance the adaptability of the method,a combination form of radial basis networks(RBF)and deep neural networks(DNN)is used for the extraction and regression analysis of state variables.Then the task constraints are transformed into the required velocity increments,and the carrying capacity is obtained by numerical iterations.The simulation results show that the accuracy deviation of the proposed calculation method is about 0.35%,and the calculation time is less than 2 seconds,which can provide theoretical support for the rapid demonstration of the overall parameters of the launch vehicle and the research of task planning.

关 键 词：运载火箭 运载能力 速度损失 高斯函数 神经网络 

分 类 号：V448.1[航空宇航科学与技术—飞行器设计]