EI、Scopus 收录
中文核心期刊

多源多精度数据融合与气动特性智能外推

MULTI-SOURCE AND MULTI-FIDELITY DATA FUSION AND INTELLIGENT EXTRAPOLATION OF AERODYNAMIC CHARACTERISTICS

  • 摘要: 在虚拟飞行试验和飞行器优化设计中, 高维高精度的计算往往伴随着高昂的计算成本. 代理模型可以大幅度提高计算效率, 同时具备对多源多精度数据的融合能力, 可基于离散试验数据点进行气动特性校正. 基于此目的, 将控制方程向低维空间投影, 提出一种侵入式降阶的边界层修正算法, 在保证精度同时高效地实现采样数据库的扩充, 在此基础上构建的代理模型能够提升对飞行器阻力系数的预测精度. 其次, 通过在降阶模型中融合多源离散试验数据, 基于最小二乘的思想引入边界约束, 使得代理模型对物面压力分布特性的预测更加贴合给定的试验值. 随后构建Kriging桥函数, 将局部约束外推至整个物面, 实现物面任意位置压力和热流分布修正值的智能匹配, 为多源离散约束的融合提供一种新的算法. 上述方法在二维翼型和三维钝锥以及无舵飞行器中得到验证, 结果表明多精度数据融合算法构造的代理模型相比于单一精度源的代理模型预测流场残差更小, 阻力系数预测更为精准, 且当无黏采样的外场信息足够充足时, 多精度模型预测结果可与CFD计算结果基本无异. 进一步发展的多源数据融合算法和气动特性智能外推算法可充分融合离散点试验数据, 改善物面压力分布和热流分布的预测结果.

     

    Abstract: In virtual flight testing and aircraft design, high-dimensional and high-fidelity computations frequently entail significant computational expenses. Surrogate models are a useful technical approach that can significantly improve the computational efficiency, while possessing the capability to fuse multi-source and multi-fidelity data, and can be used for aerodynamic performances correction based on sparse experimental data points. To this end, the governing equations are projected into a lower-dimensional space, and an intrusive reduced-order boundary layer correction algorithm is proposed to efficiently expand the sampling database while ensuring accuracy. The surrogate model constructed on this basis can enhance the prediction accuracy of the aircraft drag coefficient. Furthermore, by fusing multiple sources of sparse experimental data into the reduced-order model and introducing boundary constraints based on the least squares method, the surrogate model’s prediction of surface pressure distribution performances is made more consistent with the given experimental data. Subsequently, a Kriging bridge function is constructed to extrapolate local constraints to entire surface, enabling intelligent matching of correction values for pressure and heat flux distribution at any position on the surface. This provides a new algorithm for the fusion of multiple sparse constraints. The method described above has been validated in two-dimensional airfoil, three-dimensional blunt cone, and tailless aircraft cases. The results indicate that the surrogate model constructed by the multi-fidelity data fusion algorithm predicts smaller residuals in the flow field and provides more accurate predictions of the drag coefficient compared with a surrogate model using single-fidelity sources. Moreover, when there is sufficient flow field information at the outer edge of the boundary layer, the predictions from the multi-fidelity surrogate model closely resemble the results obtained from computational fluid dynamics calculations. The further developed multi-source data fusion algorithm and the intelligent aerodynamic characteristics extrapolation algorithm can effectively fuse experimental data points, improving the prediction results of surface pressure and heat flux distribution.

     

/

返回文章
返回