论文部分内容阅读
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。
In this paper, a design method of supercritical airfoil based on self-organizing feature map (SOM) neural network is adopted to study the design of supercritical airfoil. According to the geometric characteristics and aerodynamic characteristics of different sample airfoils, SOM neural network is used to group them to form a system of supercritical airfoil expert database. SOM neural network after training according to the design conditions, automatically select the most suitable set of airfoils as a reference airfoil. On this basis, the relationship between airfoil geometric parameters and aerodynamic parameters was established by confidence reasoning method. As the design basis, the best geometric parameters of airfoils were given by steepest descent method. The results show that SOM neural network can effectively differentiate a type of airfoil with the same characteristics and can be classified flexibly, which can provide direction guidance for the design work. Compared with the reference airfoil, the SOM neural network can effectively improve the performance of the airfoil Resistance ratio, with better overall aerodynamic performance.