基于代理模型与遗传算法的翼型优化设计方法研究  

Investigations on Airfoil Optimization Method Based on Surrogate Model and Genetic Algorithm

作　　者：王璐瑶 于佳鑫 王晓东[1] 陈江涛[2] 吴晓军[2] Lu-yao Wang;Jia-xin Yu;Xiao-dong Wang;Jiang-tao Chen;Xiao-jun Wu(North China Electric Power Unversity;China Aerodynamic Ressarch and Development Center)

机构地区：[1]华北电力大学电站能量传递转化与系统教育部重点实验室 [2]中国空气动力研究与发展中心

出　　处：《风机技术》2021年第6期69-75,I0002,共8页Chinese Journal of Turbomachinery

基　　金：国家数值风洞工程项目课题NNW2018-ZT7B14;国家自然科学基金(No.51876063)。

摘　　要：计算流体力学(CFD)广泛用于翼型的气动优化设计。由于CFD计算量大、计算时间长,常用响应面或人工神经网络等代理模型来代替CFD模拟进行气动性能评估。代理模型的预测精度关系着优化结果的可信度。本文研究基于代理模型与优化算法的翼型气动优化设计方法。采用CST函数建立了翼型的参数化方法。采用拉丁超立方实验设计方法,在设计空间内选择训练样本。基于开源CFD求解器OpenFOAM计算样本翼型的气动参数,建立基于径向基神经网络的代理模型,以减少计算量。以S809翼型为对象,升力最大为目标函数,最大厚度为约束条件,利用代理模型与遗传算法结合优化得到最优翼型,并采用了代理模型的由粗到精的外层迭代,以提高代理模型的精度和效率。结果显示:优化后的翼型较原S809翼型气动性能有了明显提升,升力系数提高,阻力系数降低;采用外层迭代后,代理模型的预测精度提高,保证了全局最优性,同时总计算量减少。Computational fluid dynamics(CFD)has been widely used in aerodynamic optimization of airfoils.Due to the large computational cost and long computational time of CFD,surrogate models are often used to predict the aerodynamic performance instead of CFD simulation.This paper investigates the aerodynamic optimization method of airfoil based on Radial Basis Function Neural Network(RBFNN)and genetic algorithm.The parameterization method of airfoil is established by using CST function.Using Latin hypercube design method,training samples are selected in the design space.Based on the CFD solver OpenFOAM,the aerodynamic parameters of the sample airfoils are calculated,and the surrogate model based on RBFNN is established to reduce the calculation cost.Taking S809 airfoil as the object,the maximum lift coefficient is the objective function,and the maximum thickness of the airfoil is the constraint condition.The surrogate model and genetic algorithm are combined to get the optimal airfoil.In order to improve the precision and efficiency of the surrogate model,the outer layer iteration of the surrogate model from coarse-to-fine is adopted.The results show that the aerodynamic performance of the optimized airfoil is significantly improved compared with the original S809 airfoil.The lift coefficient is increased,and the drag coefficient is reduced.By using the coarse-to-fine iteration,the prediction accuracy of the surrogate model is improved,the global optimum is guaranteed,and the total computation cost is reduced.

关 键 词：气动优化 代理模型 遗传算法 翼型 

分 类 号：TM614[电气工程—电力系统及自动化]