面向变形飞行器的时变气动参数在线辨识方法  

Online identification method for morphing vehicles with time-varying aerodynamic parameters

作　　者：卢昕玥 张鹏宇 霍文霞 张严雪 王剑颖 LU Xinyue;ZHANG Pengyu;HUO Wenxia;ZHANG Yanxue;WANG Jianying(School of Aeronautics and Astronautics,Sun Yat-Sen University,Guangzhou 510275,China;Science and Technology on Space Physics Laboratory,Beijing 100190,China)

机构地区：[1]中山大学航空航天学院,广东广州510275 [2]空间物理重点实验室,北京100190

出　　处：《哈尔滨工程大学学报》2024年第8期1520-1526,共7页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(62103452).

摘　　要：针对变形飞行器快时变气动参数的在线高精度获取问题,本文提出一种基于BP神经网络模型的气动参数在线辨识方法。基于变形飞行器气动模型的非线性输入/输出映射关系,建立能够在一定精度范围内逼近变形飞行器气动模型的BP神经网络模型。根据在线实测动力学参数观测数据,采用扩展卡尔曼滤波方法在线训练神经网络,实时校正并获取神经网络模型参数,基于神经网络模型快速计算并预测气动参数,从而跟踪快时变、非线性气动模型的变化。通过对变形飞行器连续变形/构型突变的气动参数辨识进行数学仿真验证。结果表明:提出的方法收敛速度快、在线辨识精度较高,可以实现对变形飞行器气动参数的有效辨识。Owing to environmental variations and shape changes during actual flight,the complex aerodynamic characteristics of morphing vehicles are time-varying and highly nonlinear.This paper proposes an online identification method based on a BP neural network model to obtain the time-varying aerodynamic parameters of morphing vehicles with high precision.First,a BP neural network model was established to approximate the aerodynamic model within a certain precision range based on the nonlinear relationship between input and output.Then,the neural network was trained online using the extended Kalman filter method with observed data from actual aerodynamic parameter tests.The BP neural network model could quickly calculate and predict the aerodynamic parameters after real-time correction and obtaining the neural network parameters.This enabled the tracking of changes in the rapidly time-varying and nonlinear aerodynamic model.Finally,a mathematical simulation was conducted to identify the aerodynamic parameters of a morphing air vehicle during successive deformation/structure mutation.The results verified that the proposed method has a fast convergence speed and high accuracy,demonstrating its effectiveness in identifying the aerodynamic parameters of morphing vehicles.

关 键 词：变形飞行器 快时变气动参数 非线性气动模型 气动参数辨识 在线辨识 智能辨识 神经网络 卡尔曼滤波 

分 类 号：V448.2[航空宇航科学与技术—飞行器设计]