摘要
针对NACA 0012翼型,在马赫数为0.176的来流条件下,首先利用数值模拟研究了翼型前缘下弯角度、前缘偏转位置、后缘下弯角度和后缘偏转位置等因素对翼型气动性能的影响规律;其次,以升阻比为目标,上述4个因素为设计变量,利用神经网络建立升阻比与4个设计变量间的预测模型;然后,充分考虑优化精度和神经网络训练数据库的计算量,构造了一种翼型优化过程与神经网络预测耦合的迭代优化策略,基于该优化策略得到最优变弯度翼型构型。对比优化翼型和原始翼型,升阻比提高约22%,较大程度改善了翼型的气动特性;并且通过远场噪声分析,发现优化翼型表现出了较好的声学性能,在1000 Hz附近单音噪声最大可降低12 dB。
Taking the NACA 0012 airfoil as research object,the flight condition selects 0.176 Mach number of incoming flow.Firstly,the influences of the different factors on the aerodynamic performance of the airfoil have been investigated by the numerical simulation,such as the downward bending angle and deflection position of the leading and trailing edgefor the airfoil and so on.Secondly,the prediction model between the object variable lift-drag ratio and the above four design variables is established by the neural network.Then under consideration of the optimization precision and the computation cost of neural network training database,an iterative optimization strategy is present,which is coupled with the neural network and optimization process.Furthermore,the optimal variable camber airfoil configuration is obtained based on the optimization strategy.Compared with the original airfoil,the optimized one increases the lift-drag ratioby about the 22%,and the aerodynamic characteristics of airfoil have been greatly improved.Moreover,by far field noise analysis,it is found that the optimized airfoil has better acoustic performance,and the tonal noise of around 1000 Hz can be reduced by up to 12 dB.
作者
保女子
彭叶辉
冯和英
杨成浩
BAO Nyuzi;PENG Yehui;FENG Heying;YANG Chenghao(School of Mathematics and Computational Science,Hunan University of Science and Technology,Xiangtan 411201,Hu'nan,China;Hunan Provincial Key Laboratory of Health Maintenance for Mechanical Equipment,Hunan University of Science and Technology,Xiangtan 411201,Hu'nan,China)
出处
《机械科学与技术》
CSCD
北大核心
2023年第2期309-320,共12页
Mechanical Science and Technology for Aerospace Engineering
基金
国家自然科学基金项目(51875194)
湖南省自然科学基金项目(2020JJ4306)。
关键词
变弯度
偏转位置
升阻比
神经网络
气动性能
variable camber
deflection position
lift-drag ratio
neural network
aerodynamic performance