一种基于BP神经网络的自适应程序角制导方法  

An adaptive programmed guidance method based on BP neural network

作　　者：张佩俊 杨若眉 蔺睿 龚学兵 杨灿 ZHANG Peijun;YANG Ruomei;LIN Rui;GONG Xuebing;YANG Can(Xi'an Aerospace Propulsion Institute,Xi'an 710025,China)

机构地区：[1]西安航天动力技术研究所,陕西西安710025

出　　处：《飞行力学》2023年第6期44-50,共7页Flight Dynamics

摘　　要：针对固体火箭在大气层内耗尽关机后干扰因素导致落点散布大的问题,在BP神经网络的基础上,提出了一种自适应程序角制导方法。在末制导段,建立线性变化的程序角模型,利用牛顿迭代法和数值积分方法离线获得满足落点控制的程序角参数。之后,在弹道模型中加入多种干扰因素,通过数值计算获得多组飞行状态和程序角参数,利用BP神经网络建立了二者之间的映射关系。最后,将训练好的网络模型分别引入基于程序制导和闭路制导的六自由度仿真模型中进行数值验证。仿真结果表明,所提出的制导方法能够很好地抑制推力耗尽之后干扰因素的影响,使落点散布降低一个数量级,具有显著的落点控制效果。In order to solve the problem of large scattering of impact points caused by disturbance factors after solid rocket depletion shutdown in the atmosphere,an adaptive programmed guidance method based on BP neural network is proposed in this paper.In the terminal guidance phase,a linearly varying program angle model is established,and the program angle parameters satisfying the impact point are obtained offline by using the Newton iteration method and numerical integration method.After that,a variety of disturbance factors are added to the trajectory model and several groups of flight states,and program angle parameters are obtained through numerical calculation.The relationship between flight state and program angle parameters is established by using BP neural network.Finally,the trained network model is brought into the six degree of freedom simulation model based on programmed guidance and closed loop guidance respectively for numerical verification.The simulation results show that the guidance method proposed in this paper can well suppress the influence of disturbance factors after the thrust shutdown,and reduce the impact point dispersion by an order of magnitude.

关 键 词：耗尽关机 BP神经网络 自适应程序制导 

分 类 号：V448.2[航空宇航科学与技术—飞行器设计]