摘要
为高效、高精度的获取高超声速飞行器表面热流数据,提出一种径向基优化的稀疏自编码器气动热预测模型。首先,利用稀疏自编码器模型对气动热数据进行表征学习,提取数据表示特征;然后将径向基函数网络模型与稀疏自编码器模型进行融合改造,最后利用改造后的模型实现对高超声速飞行器表面热流的预测。对双锥体高超声速武器表面热流预测结果表明,上述模型具有较高的精度和良好的外推性能,热流密度预测结果与多源融合数据偏差在10%以内。
To efficiently and accurately obtain heat flux data on the surface of hypersonic vehicles,a predictive model for aerodynamic heating based on a radial basis function(RBF)-optimized sparse autoencoder is proposed.Firstly,the sparse auto-encoder model is used to characterize and learn aerodynamic thermal data and extract data representation features.Then,the radial basis function network model is combined with the sparse auto-encoder model.Finally,the modified model is used to identify the surface heat flux of hypersonic vehicles.The results of identifying the surface heat flux of a dual cone hypersonic weapon show that this model has high accuracy and good extrapolation performance.The deviation between the recognition result of heat flux density and the multi-source fusion data is within 10%.
作者
司芳芳
汪文凯
秦诗牧
胡汉东
SI Fang-fang;WANG Wen-kai;QIN Shi-mu;HU Han-dong(Beijing Aerohydrodynamic Frontier Research Center,Beijing 100120,China)
出处
《计算机仿真》
2025年第3期50-54,共5页
Computer Simulation
关键词
高超声速飞行器
神经网络
数值模拟
气动热
模型
Hypersonic vehicle
Neural network
Numerical simulation
Aeroheating
Model
