Low-Order Modeling and Sensor-Based Prediction of Stalled Airfoils at Moderate Reynolds Number
Low-Order Modeling and Sensor-Based Prediction of Stalled Airfoils at Moderate Reynolds Number
复制标题
DOI：
10.2514/1.j062475
复制
发表时间：
2023-05
期刊：
影响因子：
2.5
通讯作者：
D. Carter;B. Ganapathisubramani
中科院分区：
文献类型：
--
作者：
D. Carter;B. Ganapathisubramani
我们在NACA 0012和NACA 65-410翼型的沿流表面法向面对平面时间分辨粒子图像测速场进行了分析[公式：见文本]，重点分析了中高迎角下的失速现象[公式：见文本]。从安装在翼片上的称重传感器获得的升力谱中确定的主要流动频率，突出了崖体脱落的存在[[公式：见文]的[公式：见文]，其中[公式：见文]是频率，[公式：见文]是翼型弦，[公式：见文]是自由流速度]在所有情况下和突出的低频[[公式：见文]的[公式：见文]]在瞬态失速（TS）的情况下。通过适当的正交分解（POD）数据驱动的建模框架表明，低频与流动分离和再附着有关，这些分离和再附着是由垂直于吸力面的底层膨胀和收缩驱动的。此外，利用线性随机估计（LSE）对TS和深失速（DS）情况下的前、中弦和尾缘（伪）压力探头预测低阶特征的能力进行量化。该框架确定逆流区域的质心，DS和TS的误差分别为5%和20%。值得注意的是，发现控制低阶POD相关性的LSE系数不强烈依赖于翼型几何形状。使用NACA 0012案例的探针训练LSE来预测NACA 65-410速度场的比较性能证明了这一点，反之亦然。这项工作证明了POD对湍流停滞翼型的低阶特征以及跨几何形状预测流动特征的相似性所提供的洞察力，无需重新训练LSE库。
We present analysis from planar time-resolved particle image velocimetry fields in the streamwise surface-normal plane of turbulent flow surrounding NACA 0012 and NACA 65-410 airfoils at [Formula: see text] focusing on stall phenomena at intermediate to high angles of attack [Formula: see text]. Dominant flow frequencies, identified from the lift spectra obtained from a load cell mounted to the foils, highlight the presence of bluff-body shedding [[Formula: see text] of [Formula: see text], where [Formula: see text] is the frequency, [Formula: see text] the airfoil chord, and [Formula: see text] the freestream velocity] for all cases and prominent low frequencies [[Formula: see text] of [Formula: see text]] for cases in transient stall (TS). A data-driven modeling framework via the proper orthogonal decomposition (POD) reveals that the low frequencies are associated to flow separation and reattachment driven by underlying expansion and contraction normal to the suction surface. Further, the ability to predict the low-order features from (pseudo) pressure probes at the leading, midchord, and trailing edges for both TS and deep stall (DS) cases is quantified using linear stochastic estimation (LSE). The framework pinpoints the centroid of the region of reverse flow with error on the order of 5 and 20% for DS and TS regimes, respectively. Notably, it is found that LSE coefficients governing the low-order POD correlations do not strongly depend on the airfoil geometry. This is demonstrated by the comparative performance of training the LSE using the probes of the NACA 0012 cases to predict the NACA 65-410 velocity fields and vice versa. This work demonstrates the insight afforded by the POD on the low-order features of turbulent stalled airfoils as well as the similarity of such features across geometries toward predicting flow features for potentially any airfoil geometry without the need to retrain an LSE library.