基于神经网络的跨介质飞行器水动力系数快速预示方法

Efficient Prediction Method for Hydrodynamic Coefficient of Trans-media Flight Vehicle Based on Neural Network

近年来,跨介质飞行器能够充分利用在不同介质中的优势,受到国内外的广泛关注。为充分提升跨介质飞行器水动系数的计算效率,提出了一种基于神经网络的跨介质飞行器水动力系数快速预示方法。首先,建立了跨介质飞行器的水动分析模型,并对模型进行校验。在此基础上,引入神经网络近似思想,建立了跨介质飞行器水动力系数神经网络预示模型;以飞行器在水中的攻角与速度为神经网络输入,升力系数与阻力系数为神经网络输出,通过少量的训练样本实现对跨介质飞行器在水中不同工况下的水动力系数快速预测。工程案例结果表明,文中提出的基于神经网络的跨介质飞行器水动力系数快速预示方法能够在保证水动力系数插值表精度的前提下,计算效率提升47%以上,从而验证了本文研究工作的有效性与工程实用性。

In recent years,trans-media flight vehicles have attracted much attention due to utilizing the advantages in different media.To improve the computational efficiency for calculating hydrodynamic coefficient of trans-media flight vehicle,a neural network based efficient prediction method is proposed in this paper.First,the hydrodynamic analysis model of trans-media flight vehicle is established and verified.Inspired by the idea of approximation using neural networks,a neural network based prediction model for the hydrodynamic coefficients of the trans-media flight vehicle are then constructed,where the input includes angle of attack and velocity and the output includes lift and drag coefficients.The prediction model can approximate the hydrodynamic coefficients efficiently and accurately,merely using few training samples.The engineering case demonstrates that the proposed method can improve the computational efficiency by 47% with a high approximation precision,which illustrates the effectiveness and practicality of the work.