摘要
本文采用了基于自组织特征映射(SOM)神经网络的超临界翼型设计方法,研究了超临界翼型设计问题。根据不同样本翼型的几何特征和气动特征,利用SOM神经网络对其进行分组,形成系统的超临界翼型专家数据库。训练后的SOM神经网络能够根据设计条件,自动挑选出最合适的一组翼型作为参考翼型。在此基础上,采用置信度推理法建立了翼型几何参数与气动参数之间的关系,作为设计基准,采用最速梯度下降法给出翼型的较佳几何参数。研究结果表明:SOM神经网络能够有效地区分有相同特征的一类翼型,分类灵活,可以为设计工作提供方向性指导;最终得到的设计翼型与基准翼型相比,有效地提高了升阻比,具有较优的综合气动性能。
A supercritical airfoil design method was brought forwards based on self-organizing feature map（SOM） neural network.The SOM network was used to classify the different airfoils according to their geometry characteristics and aerodynamic characteristics,and then a supercritical airfoil expert database was built.A reference group of airfoils was automatically selected from the expert data base by the well-trained network according to the design requirement.A certainty factor inference method was used to build the relationship between the geometry characteristics and aerodynamic characteristics which gave the basic principles to design the airfoil geometry using gradient descent optimization method.The design results indicate that the SOM network can efficiently classify the airfoils by the same characteristics which can provide the professional guidance to the design work,and the designed airfoil has higher lift-to-drag ratio than the referenced airfoils,so that the design results have better integrated aerodynamic performance.
出处
《力学季刊》
CSCD
北大核心
2011年第3期411-417,共7页
Chinese Quarterly of Mechanics
基金
上海市经济信息化委支持项目"产-36-专项-1"
关键词
SOM神经网络
专家数据库
超临界翼型设计
置信度推理
SOM neural network
expert database
supercritical airfoil design
certainty factor inference
