基于统计特征聚类的飞行器表面测点位置快速确定方法

An efficient scheme for surface sensor placement on flight vehicles via clustered statistical features

  • 摘要: 在飞行器设计和使用过程中,为了通过表面稀疏测量值精确反演气动分布载荷,需要合理布置数量有限的传感器以提取更多有效信息。为了解决这一问题,本文提出一种基于统计特征聚类的飞行器表面测点位置快速确定方法。首先,通过均值、标准差、偏度、峰度四阶统计矩描述所有待选测点处的气动分布载荷统计特征。然后,采用K-means++算法对各测点处的统计特征进行聚类分析,并根据聚类结果,在每一簇中选择距离簇中心最近的测点,组合完成测点布置。最后,通过NACA 0012翼型和CRM飞机表面压力的测点布置算例,验证了本文方法所得测点的稀疏反演精度。结果表明:对于NACA 0012翼型,仅需在翼型表面布置2个测点,测试样本表面压力的平均绝对误差为10−3量级;对于CRM飞机,需在机身布置4个测点、机翼布置3个测点,测试样本机身和机翼的反演误差分别为10−4和10−3量级。

     

    Abstract: During the design and use of flight vehicles, to accurately infer the complete aerodynamic distributed loads from sparse surface measurements, it is essential to deploy the limited sensors in locations that encompass abundant information. To address this issue, we propose an efficient scheme based on clustered statistical features. Firstly, statistical features of the aerodynamic distributed loads at all candidate sensor locations are described through four statistical moments, i.e., mean, standard deviation, skewness and kurtosis. Secondly, the K-means++ clustering algorithm is employed to analyze the statistical features at each sensor location. Based on the clustered results, the sensor locations are determined by selecting the location closest to the center of each cluster. Finally, case studies on surface pressure sensor placement for the NACA 0012 airfoil and CRM aircraft are used to validate the sparse inversion accuracy of the measurement points obtained by the proposed method. The results show that for the NACA 0012 airfoil, only 2 sensors are required on the airfoil surface, and the mean absolute error of the inversed surface pressure for the test samples is on the order of 10−3. For the CRM aircraft, 4 sensors are placed on the fuselage and 3 on the wing, and the inversion error of the test samples for the fuselage and wing is on the orders of 10−4 and 10−3, respectively.

     

/

返回文章
返回