摘要
激波抖振是一种由激波-边界层干扰引起的激波自激振荡现象,可能导致结构疲劳破坏,甚至引发飞行安全问题。激波抖振始发边界的准确预测对于运输类飞机的设计具有重要的工程意义。本文建立了一个融入定常流场中特征的神经网络(CFNN)模型,实现了对激波抖振始发迎角的准确预测。以NACA0012翼型为研究对象,利用卷积神经网络(CNN)模型提取抖振发生前后定常流场中的特征。随后,将提取的低维特征作为全连接神经网络(FNN)模型的隐含层,用以预测激波抖振的始发迎角。在较高马赫数的泛化预测中,CFNN模型预测的激波抖振始发迎角的平均相对误差,相较未融入特征的全连接神经网络(NN)模型,减小了约70%以上。研究结果表明,从定常流场中提取的低维特征能够辅助预测非定常激波抖振问题的始发迎角,提升神经网络模型的性能。
Shock buffet is a self-excited oscillation phenomenon caused by shock wave-boundary layer interference,which may lead to structural fatigue failure and even cause flight safety issues.The accurate prediction of shock buffet onset boundary is of great engineering significance for the design of transport aircraft.This paper establishes a Characteristics-integrated Fully connected Neural Network(CFNN)model that incorporates features from steady flow field,achieving accurate prediction of shock buffet onset angle of attack.Taking the NACA0012 airfoil as the research object,a Convolutional Neural Network(CNN)model extracts features from the steady flow field before and after the onset of the shock buffet.Subsequently,the extracted low dimensional features are used as hidden layers in the Fully connected Neural Network(FNN)model to predict the onset angle of attack of the shock buffet.In the generalization prediction of higher Mach numbers,the average relative error of the shock buffet onset angle of attack predicted by the CFNN model is reduced by more than 70%compared to the fully connected Neural Network(NN)model without incorporating features.The research results indicate that the low dimensional features extracted from the steady flow field can assist in predicting the onset angle of attack for unsteady shock buffet problems and improve the performance of neural network models.
作者
马启悦
高传强
孙健
Ma Qiyue;Gao Chuanqiang;Sun Jian(Northwestern Polytechnical University,Xi’an 710072,China;Beijing Institute of Astronautical Systems Engineering,Beijing 100076,China)
出处
《航空科学技术》
2024年第7期49-55,共7页
Aeronautical Science & Technology
基金
航空科学基金(2019ZH053003)。