Getting smarter overnight: how automated experiments help unwind unsteady vortex-dominated flows*

Typical unsteady vortex-dominated flows like those involved in bio-inspired propulsion, unsteady airfoil separation, and vortex-induced vibrations can be prohibitively expensive to simulate and impossible to measure comprehensively. They are inherently non-linear, often involve moving boundaries, high-dimensional parameter spaces, and multiscale flow structures. The classical way to get around these challenges has been to reduce the experimental complexity by using canonical motions (e.g. ramp-up or sinusoidally pitching motions) or simplified unsteady inflow conditions (e.g. one-minus-cosine or trapezoidal gust shapes). In our lab, we design automated experiments that can run continuously and autonomously such that we can explore and exploit higher-dimensional parameter spaces that cover more realistic and technically relevant unsteady conditions compared to what is traditionally feasible when conducting supervised canonical motion experiments. This approach give us the ability to derive more robust and generalizable models and control solutions while still discovering rare and extreme events. Our recent experiments allowed us to uncover flapping wing kinematics that maximize lift and efficiency, to optimize blade pitching kinematics that improve the power production of vertical axis wind turbines, and to gain insight into the influence of morphology on the forces of thin flexible objects. Once the experiments are started, the experimentalists' input is minimized while the potential for scientific discovery is maximized.