耦合粒子图像测速误差的连续伴随数据同化技术

Continuous-adjoint based data assimilation technique coupled with particle image velocimetry error

提出了一种耦合粒子图像测速(PIV)实验误差的连续伴随数据同化算法,通过优化目标损失函数,增强算法在不同误差场景下的鲁棒性。为了验证该算法的有效性,先对已知PIV流场植入合成误差进行同化对比测试,继而对PIV互相关算法不同参数设置所获得的流场进行同化研究。结果表明:相比于原连续伴随数据同化,耦合PIV实验误差的同化算法能够对实验观测数据去伪存真,抗误差干扰能力明显提升,鲁棒性更强,能够对高误差场景下的流动数据进行更好地同化,准确地预测流场的真实分布规律,还原流场细节。

A continuous-adjoint based data assimilation technique coupled with Particle Image Velocimetry(PIV) error was proposed to optimize the objective loss function,thereby enhancing the robustness of the technique in different error scenarios.For verification,a given PIV flow field implanted with synthetic errors was selected as a preliminary test,and a further data assimilation test was implemented in the flow fields obtained with different parameter settings of the PIV cross-correlation algorithm.The results indicated that the continuous-adjoint algorithm coupled with the PIV error can discard the false experimental observations and improve the anti-interference ability and robustness,compared with its original counterpart.The high-fidelity flow fields can be well obtained using this data assimilation technique even in large error scenarios.