Surrogate-based aerodynamic shape optimization for delaying airfoil dynamic stall using Kriging regression and infill criteria
Surrogate-based aerodynamic shape optimization for delaying airfoil dynamic stall using Kriging regression and infill criteria
复制标题
DOI：
10.1016/j.ast.2021.106555
复制
发表时间：
2021-02
影响因子：
5.6
通讯作者：
Vishal Raul;Leifur Þ. Leifsson
中科院分区：
文献类型：
--
作者：
Vishal Raul;Leifur Þ. Leifsson
动态失速现象的特征在于前缘涡流的形成，这是造成不利的空气动力和力矩的原因，不利地影响系统的结构强度和寿命。气动外形优化（阿索）提供了一种经济有效的方法来延迟或减轻动态失速特性。不幸的是，阿索需要对精确但耗时的计算流体动力学（CFD）模拟进行多次评估，以产生最佳设计，使得优化过程在计算上昂贵。本文提出了一种基于替代物的优化技术，以减轻阿索的计算负担，从而延迟和减轻翼型的深度动态失速特性。特别是，克里格回归代理模型用于近似的目标和约束函数。机翼几何形状使用六个PARSEC参数进行参数化。目标函数和约束函数的计算采用非定常雷诺平均Navier-Stokes方程和Menter剪切应力输运湍流模型。该方法在雷诺数为135，000、马赫数为0.1的垂直轴风力涡轮机翼型上进行了演示，该翼型经历了频率为0.05的正弦振荡。代理模型是用60个初始样本构建的，并使用预期的改进用20个填充样本进一步细化。基于20个测试数据样本，用归一化均方根误差对生成的代理模型进行了验证。改进的代理模型用于使用多起点基于梯度的搜索来寻找最优设计。与基线相比，最佳翼型具有更大的厚度、更大的前缘半径和后拱。这些几何形状的变化使动态失速角延迟超过3 °，并降低了俯仰力矩系数波动的严重程度。最后，利用Sobol指数对优化设计进行了全局灵敏度分析，揭示了影响翼型动态失速特性的最重要的形状变量及其相互作用效应。
The dynamic stall phenomenon is characterized by the formation of a leading-edge vortex, which is responsible for adverse aerodynamic forces and moments adversely impacting the structural strength and life of a system. Aerodynamic shape optimization (ASO) provides a cost-effective approach to delay or mitigate the dynamic stall characteristics. Unfortunately, ASO requires multiple evaluations of accurate but time-consuming computational fluid dynamics (CFD) simulations to produce optimum designs rendering the optimization process computationally costly. The current work proposes a surrogate-based optimization (SBO) technique to alleviate the computational burden of ASO to delay and mitigate the deep dynamic stall characteristics of airfoils. In particular, the Kriging regression surrogate model is used for approximating the objective and constraint functions. The airfoil geometry is parametrized using six PARSEC parameters. The objective and constraint functions are evaluated with the unsteady Reynolds-averaged Navier-Stokes equations with a C-grid mesh topology and Menter's shear stress transport turbulence model. The approach is demonstrated on a vertical axis wind turbine airfoil at a Reynolds number of 135,000 and a Mach number of 0.1 undergoing a sinusoidal oscillation with a reduced frequency of 0.05. The surrogate model is constructed with 60 initial samples and further refined with 20 infill samples using expected improvement. The generated surrogate model is validated with the normalized root mean square error based on 20 test data samples. The refined surrogate model is utilized for finding the optimal design using multi-start gradient-based search. The optimal airfoil has a higher thickness, larger leading-edge radius, and an aft camber compared to the baseline. These geometric shape changes delay the dynamic stall angle by over 3∘ and reduces the severity of the pitching moment coefficient fluctuation. Finally, global sensitivity analysis is conducted on the optimal design using Sobol'indices revealing the most influential shape variables and their interaction effects impacting the airfoil dynamic stall characteristics.