适用于弹性飞机飞行动力学仿真的气动力降阶方法

Aerodynamic order reduction method for elastic aircraft flight dynamics simulation

针对低速太阳能无人机在由阵风引起的大迎角非线性气动力情况下的高精度、高效率气动弹性及飞行动力学工程仿真问题,对气动力降阶方法及其在飞行动力学仿真中的应用进行深入探索。以二维翼型为例,分别基于自回归外部输入模型(ARX)方法、多小波Volterra级数方法、反向传播(BP)、径向基函数(RBF)神经网络方法和支持向量回归(SVR)方法建立了非定常气动力降阶模型,并比较了5种方法的收敛性及泛化能力。在此基础之上综合ARX和神经网络方法的优点提出一种计算精度高,收敛性、泛化能力较强的ARX-BP/RBF组合建模方法。将基于ARXRBF的组合气动力降阶方法和弹性飞机刚弹耦合动力学方程相结合建立基于气动力降阶的弹性飞机飞行动力学仿真模型。以一架太阳能无人机模型为例,开展数值验证,对算例对象在阵风及舵面激励下的响应进行仿真研究,并和计算流体力学(CFD)-计算结构动力学(CSD)结果及基于线性气动力的仿真结果进行对比。结果表明:所建立的弹性飞机仿真模型能有效反映气动力非线性特性,且仿真效率高于CFD-CSD分析方法,适用于工程实践。

We develop an aeroelasticity and flight dynamics simulation model with nonlinear aerodynamic reduced order model for low-speed solar unmanned aerial vehicles under high angle of attack caused by gust.Five aerodynamic reduced-order models are created in light of the aforementioned issues,and their use in simulating flight dynamics is further explored.The reduced order models include the model based on the autoregressive model with eXogenous inputs(ARX)method,multi-wavelet Volterra series method,back propagation(BP)neural network,radial basis function(RBF)neural network and support vector regression.Then the convergence and the generalization ability of the five models are compared,and a new model with high precision and strong generalization ability is proposed,which is a combination of ARX and neural network model.Taking a solar UAV model as an example,combining an aerodynamic reduced order model with a rigid-elastic coupling dynamic equation,the flight dynamics simulation model of elastic aircraft is established.Lastly,the computational fluid dynamics-computational structural dynamics(CFD-CSD)outputs and the outcomes of the simulation model based on linear aerodynamics are compared with the simulation results to determine how the UAV model responds to gusts.The results show that the performance of the elastic aircraft simulation model in aerodynamic prediction and gust response analysis is better than that of the model based on linear aerodynamics and is in good agreement with the CFD-CSD model,and the simulation efficiency is much higher than that of the CFD-CSD analysis.