为探究kriging插值算法应用于五孔探针试验数据插值的有效性,通过标准风洞试验,完成了燃气轮机流场测试专有设计五孔探针标定,获得了角度校准系数、总压及静压校准曲面,提出采用基于Kmeans聚类算法的kriging插值算法对校准数据有序化处理,并与基于最小二乘法的传统多项式插值结果比较,结果表明：在不同来流马赫数下、在整个气动特性试验角度范围内,探针角度校准曲线、静压及总压特性曲面均具有较高的几何相似性、对称性,且角度校准曲线网格方正,证明了探针加工质量较好,标定样本数据准确可靠；对于变马赫数特性下的校准数据三维插值,kriging插值精度优于传统多项式插值,但二者精度均较高,两种插值模型均可靠有效；对于定马赫数特性下的校准数据二维插值,传统多项式插值由于龙格现象在边缘点处误差较大、在内部点误差较小,kriging插值可以规避多项式插值的龙格现象,其在所有点处插值误差均匀且均接近0,精度优于多项式插值,验证了kriging插值算法应用于五孔探针校准数据插值的有效性。

In order to explore the effectiveness of kriging interpolation algorithm applied to five-hole probe test data interpolation, the calibration of five-hole probe specially designed for gas turbine flow field test was completed through standard wind tunnel test, and the Angle calibration coefficient, total pressure and static pressure calibration surface were obtained. kriging interpolation algorithm based on Kmeans clustering algorithm is proposed to process the calibration data in order, and compared with the traditional polynomial interpolation results based on least square method, the results show that: Under different incoming Mach number and in the whole range of aerodynamic characteristics, the probe Angle calibration curve, static pressure and total pressure characteristic surface have high geometric similarity and symmetry, and the Angle calibration curve grid is square, which proves that the probe machining quality is good and the calibration data is accurate and reliable. For the 3D interpolation of calibration data under variable Mach number characteristics, the accuracy of kriging interpolation is better than that of traditional polynomial interpolation, but both accuracy are higher, and both interpolation models are reliable and effective. For the two-dimensional interpolation of calibration data under constant Mach number characteristics, the traditional polynomial interpolation has larger error at the edge points and smaller error at the internal points due to the Runge phenomenon, while the kriging interpolation can avoid the Runge phenomenon of polynomial interpolation. The interpolation errors are uniform and close to 0 at the points, and the accuracy is better than that of polynomial interpolation. The validity of kriging interpolation algorithm applied to five-hole probe calibration data interpolation is verified.