首页 | 本学科首页   官方微博 | 高级检索  
     

基于深度学习的混合翼型前缘压力分布预测
引用本文:廖鹏,姚磊江,白国栋,赵航. 基于深度学习的混合翼型前缘压力分布预测[J]. 航空动力学报, 2019, 34(8): 1751-1758. DOI: 10.13224/j.cnki.jasp.2019.08.014
作者姓名:廖鹏  姚磊江  白国栋  赵航
作者单位:西北工业大学无人机特种技术重点实验室,西安710072;西北工业大学航空学院,西安710072;西北工业大学无人机特种技术重点实验室,西安,710072
摘    要:提出了一种基于深度学习的混合翼型前缘压力分布预测方法,通过对翼型几何特征提取、压力分布曲线的参数化,建立了卷积神经网络模型(CNN),并利用计算流体力学(CFD)的计算结果作为其训练样本,实现对混合翼型前缘压力分布的预测。结果表明:两种方法计算结果的拟合优度大于0.98,基于深度学习的计算方法耗时1.7 s,CFD方法耗时大于50 s,计算时间大大缩短。该方法能够在满足计算精度的条件下提高计算效率并可应用于其他的翼型设计过程。 

关 键 词:混合翼型  深度学习  卷积神经网络(CNN)  参数化方法  压力分布
收稿时间:2019-01-21

Prediction of hybrid airfoil leading edge pressure distribution based on deep learning
Abstract:A prediction model on the leading edge pressure distribution of the hybrid airfoil based on deep learning was proposed. A convolutional neural network model (CNN) was established on the basis of the geometric feature extraction of the hybrid airfoil and the parameterization of the pressure distribution curve. A group of hybrid airfoils with different trailing edges were analyzed by a verified CFD method. The CFD results were used as the training set of the CNN. Results show that the goodness of fit of the calculation results of the two methods exceeds 0.98. The proposed prediction method based on deep learning takes 1.7 s and CFD method takes more than 50 s, computation time is greatly reduced. The proposed method can improve the computational efficiency with satisfying the calculation accuracy and it can be applied to the design processes of other airfoils.
Keywords:hybrid airfoil  deep learning  convolutional neural network(CNN)  parameterization method  pressure distribution
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《航空动力学报》浏览原始摘要信息
点击此处可从《航空动力学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号