飞行器总体设计中主要参数的神经网络估算

Neural Network for Estimating the Main Parameters in Flight Vehicle Preliminary Design

运用近年来迅速发展的神经网络技术，成功构造了飞行器总体设计过程中具有重要作用的总体质量估算网，升力系数估算网，以及阻力系数估算网，结果表明训练后的参数估算网比通常使用的近似估计公式具有更高的精度．同时，根据所涉及问题的数据特点，为了提高对网络的训练精度，对现有ＭＢＰ 算法作了进一步改进，仿真结果证明改进的ＭＢＰ 算法具有更高的训练效率．这一思路和方法可适用于机械及航空航天其它产品的总体概念设计过程．

With neural network technology developed rapidly in recent years,three parameter estimating networks for total mass,lift coefficient and drag coefficient are respectively structured successfully that play an important role in flight vehicle preliminary design.The result shows that trained parameter estimating networks can give more accurate value than usual approximating functions.And,according to the characteristics of the data involved here,a modification has been made for promoting the accuracy of MBP algorithm.The simulating result proved that the modified MBP algorithm can work more effectively.This method can be used in other product's preliminary design of machinery,aeronautics and astronautics.