Pitch Perfect - tuning low-order aerodynamic models for non-periodic flows

无人驾驶飞行器(UAV)在旅途中可能会遇到外部干扰，并绕过障碍物。由于飞行器偏离了稳定的水平飞行，其周围的流动在统计上不再是静止的。如果没有定义好的平均(时间平均)流动，开发能够捕捉车辆周围连贯流动结构的降阶模型就更具挑战性。无人机上的大多数传感器只测量固体边界附近的流量，这一事实加剧了这一困难。如果无人机了解周围的速度场，它在激烈的动作中或面对阵风时就会有更好的控制权。因此，该项目的目标是利用机器学习和降阶建模技术，将有限的表面传感器测量结果转换为对车辆周围速度场的准确估计。这可以解释为非定常流动的超分辨问题。预计该项目的成果可以进一步扩展到更复杂的机动动作，例如俯仰和涌动，这是大规模流动应用所固有的，如直升机叶片和经历动态失速的风力涡轮机。

Unpiloted aerial vehicles (UAVs) can be expected to encounter external disturbances and to maneuver around obstacles during its journey. As a result of the vehicle deviating from steady level flight, the flow around it is no longer statistically stationary. Without a well-defined mean (time-averaged) flow, it is more challenging to develop reduced-order models that capture coherent flow structures around the vehicle. This difficulty is compounded by the fact that most sensors onboard a UAV only measure flow quantities near a solid boundary. If the UAV had knowledge of the surrounding velocity field, it would have better control authority during violent maneuvers or in the face of gusts. The objective of this project, therefore, is to exploit machine learning and reduced-order modeling techniques that convert limited surface sensor measurements into an accurate estimation of the velocity field around the vehicle. This can be interpreted as a super resolution problem for non-stationary flows. It is expected that the results from this project could be extended further to more complex maneuvers, e.g. pitching and surging, that are inherent to large-scale flow applications such as helicopter blades and wind turbines undergoing dynamic stall.