基于神经网络的飞行器总体优化设计

Conceptual Optimization Design of Aircraft Based on Neural Network

针对经验公式存在的精度低、适用范围受限的缺点,采用广义回归网络代替经验公式计算气动特性,并将其嵌入到优化设计流程之中,提出了一种基于神经网络的飞行器总体优化设计方法。神经网络的训练样本输入、输出数据通过试验设计和涡格法计算得到。以高空长航时无人机为例,选取航时与机翼结构重量系数作为目标函数,应用所提出的方法进行总体优化设计。计算结果表明,提出的方法是可行的,能够满足设计要求,同时降低优化过程的复杂程度,减少迭代次数。

For lower fidelity and poor flexibility of empirical formula, a method for conceptual optimization design of aircraft using neural network is proposed in this paper. In this method, a generalized regression network model instead of empirical formula for aerodynamic calculation is embedded in the processes of aircraft optimization design. The data set trained for neural network is obtained by design of experiments and vortex lattice method program. For application, endurance and coefficient of wing structural weight are selected as objective function, and then the optimization design of High Altitude Long Endurance （HALE） UAV is investigated. The results in this study illustrate that this method is feasible a.