基于 Kriging与多点取样算法的压气机特性建模

针对航空燃气轮机流量系数插值建模过程中,获取样本数据时的实验成本较高以及插值模型精度偏低这 2个问题,根据样本点在设计空间中的分布与模型精度的关系,在 Kriging插值模型的基础上,采用把垂距作为设计变量取值标准的取样算法。首先,将样本点与相邻 2点之间连线的垂距与垂距阈值进行比较,筛选出符合条件的数据点,形成候选点集；其次,选用高斯函数判定基础点集和候选点集的相关性；最后,使用该算法筛选后的样本点构建 Kriging插值模型。实验结果表明,该方法使用的样本点为原来数量的 70%左右,并且预测精度得到了提高,该方法是有效可行的。

Compressor Characteristic Modeling Based on Kriging and Multipoint Sampling Algorithm

In order to solve the two problems of high cost of obtaining sample data and low accuracy of interpolation model inthe process of aviation gas turbine flow coefficient interpolation modeling, according to the relationship between the distribu.tion of sample points in design space and model accuracy, based on Kriging interpolation model, a multipoint sampling algo.rithm that takes the vertical distance as the value criterion for the design variable is adopted. Firstly, the vertical distance be.tween the sample point and the adjacent two points is compared with the vertical distance threshold to filter out the eligibledata points and form a candidate point set. Secondly, the Gaussian function is used to determine the correlation between thebasic point set and the candidate point set. Finally, the Kriging interpolation model is constructed using the sample points fil.tered by the algorithm. The experimental results show that the sample points used in this method are about 70% of the origi.nal ones, and the prediction accuracy is improved. The method is effective and feasible.