邓晨, 陈功, 王文正, 等. 基于不确定度和气动模型的气动数据融合算法[J]. 空气动力学学报, 2022, 40(4): 117−123. doi: 10.7638/kqdlxxb-2020.0151
引用本文: 邓晨, 陈功, 王文正, 等. 基于不确定度和气动模型的气动数据融合算法[J]. 空气动力学学报, 2022, 40(4): 117−123. doi: 10.7638/kqdlxxb-2020.0151
DENG C, CHEN G, WANG W Z, et al. Aerodynamic data fusion algorithms based on aerodynamic model and uncertainly[J]. Acta Aerodynamica Sinica, 2022, 40(4): 117−123. doi: 10.7638/kqdlxxb-2020.0151
Citation: DENG C, CHEN G, WANG W Z, et al. Aerodynamic data fusion algorithms based on aerodynamic model and uncertainly[J]. Acta Aerodynamica Sinica, 2022, 40(4): 117−123. doi: 10.7638/kqdlxxb-2020.0151

基于不确定度和气动模型的气动数据融合算法

Aerodynamic data fusion algorithms based on aerodynamic model and uncertainly

  • 摘要: 飞行器气动数据的来源主要有风洞试验、数值模拟、飞行试验三种方式。受试验和模拟能力的限制,任意一种单一手段都难以准确地对飞行器全飞行包线进行覆盖。为弥补各种数据的“缺陷”,提出并实现了两种数据融合算法:一种是依据不确定度作为权值参考,进行加权融合的加权融合算法,利用高斯过程回归算法获得不同来源气动数据预测值的特征,并进行加权融合;另一种是基于模型的CoKriging融合算法,利用CoKriging算法直接建立融合模型。并以某型飞行器气动数据为例进行了对比分析。结果表明:使用单一精度数据建模时,在一定的范围内,样本数据越多,覆盖的设计变量空间越广,精度越高;与单独使用一种精度数据的建模算法相比,两种融合算法预测结果的精度都有较大的提高;相比于基于不确定度的融合算法,使用CoKriging算法建模得到的结果精度更高,提高了近一个数量级。融合数据对于提高数据精度和增强模型预测能力上有重要作用,在参数变量空间内,融合数据能够对高精度数据进行內填补充,同时在参数变量空间外的融合数据能对数据的变化趋势预测提供参考。

     

    Abstract: Wind-tunnel experiments, numerical simulations, and flight tests are the three major means for aircraft aerodynamic research. However, each method has its limitation, thus it is difficult to accurately predict aerodynamic characteristics of an aircraft in its full flight envelope by a single approach. Data fusion is important for improving accuracy and enhancing forecasting capabilities. In the parameter space, the fused data could be a supplement to high-fidelity data. Meanwhile, the fused data outside the parameter space can provide a reference for the variation trend of the data. Therefore, aerodynamic data obtained by different methods need to be fused. To this end, two data fusion algorithms are proposed. One is a weighted fusion algorithm based on uncertainty, which uses the Gaussian process regression to obtain the characteristics of aerodynamic data from different sources before performing a weighted fusion. The other is an agent fusion model which is established by using the CoKriging algorithm. Aerodynamic data of an aircraft are taken for comparative analyses. Results show that the prediction accuracy of the Kriging model using single-precision data can be improved by more samples. Both fusion algorithms have significant higher accuracy than the Kriging model using only single-precision data. But the accuracy of the CoKriging algorithm is roughly one order of magnitude higher than the fusion algorithm based on uncertainty.

     

/

返回文章
返回