论文标题
部分可观测时空混沌系统的无模型预测
Enhanced Prediction of Three-dimensional Finite Iced Wing Separated Flow Near Stall
论文作者
论文摘要
储层计算是预测湍流的有力工具，其简单的架构具有处理大型系统的计算效率。然而，其实现通常需要完整的状态向量测量和系统非线性知识。我们使用非线性投影函数将系统测量扩展到高维空间，然后将其输入到储层中以获得预测。我们展示了这种储层计算网络在时空混沌系统上的应用，该系统模拟了湍流的若干特征。我们表明，使用径向基函数作为非线性投影器，即使只有部分观测并且不知道控制方程，也能稳健地捕捉复杂的系统非线性。最后，我们表明，当测量稀疏、不完整且带有噪声，甚至控制方程变得不准确时，我们的网络仍然可以产生相当准确的预测，从而为实际湍流系统的无模型预测铺平了道路。
Icing on three-dimensional wings causes severe flow separation near stall. Standard improved delayed detached eddy simulation (IDDES) is unable to correctly predict the separating reattaching flow due to its inability to accurately resolve the Kelvin-Helmholtz instability. In this study, a shear layer adapted subgrid length scale is applied to enhance the IDDES prediction of the flow around a finite NACA (National Advisory Committee for Aeronautics) 0012 wing with leading edge horn ice. It is found that applying the new length scale contributes to a more accurate prediction of the separated shear layer (SSL). The reattachment occurs earlier as one moves towards either end of the wing due to the downwash effect of the wing tip vortex or the influence of end-wall flow. Consequently, the computed surface pressure distributions agree well with the experimental measurements. In contrast, standard IDDES severely elongates surface pressure plateaus. For instantaneous flow, the new length scale helps correctly resolve the rollup and subsequent pairing of vortical structures due to its small values in the initial SSL. The computed Strouhal numbers of vortical motions are approximately 0.2 in the initial SSL based on the vorticity thickness and 0.1 around the reattachment based on the separation bubble length. Both frequencies increase when moving towards the wing tip due to the downwash effect of the tip vortex. In comparison, the excessive eddy viscosity levels from the standard IDDES severely delay the rollup of spanwise structures and give rise to "overcoherent" structures.
