2026/03/05 by Yating Hu, Yuting Dai, Yutian Cao +2
Engineering · Physics and Astronomy · #Aeroelasticity and Vibration Control #Biomimetic flight and propulsion mechanisms #Model Reduction and Neural Networks
paper · doi:10.2514/1.j066674
Particle image velocity of the gust-induced flowfield and lift for the elastic wing are measured simultaneously in the wind tunnel test. The experimental results indicate that the wing’s pronounced elastic motion alters the flow acceleration performance of the gust, which brings a challenge to control at varying gust frequencies. To address this, a deep reinforcement learning (DRL) control is presented for the first time in a gust alleviation wind tunnel test, which is not only trained but also tested using only experimental data. A GLA wind tunnel test is conducted with a seamless morphing trailing edge (TE). From the perspective of the phase offset between the gust and the morphing TE, the controller can adjust the TE morphing to its optimal phase across different gust frequencies, varying inflow velocities, and gust amplitudes. Hence, the aerodynamics induced by the morphing TE can cancel out the gust-induced lift, which can be suppressed significantly. Specifically, at [Formula: see text], [Formula: see text], and [Formula: see text], the gust-induced lift coefficient is alleviated by 76.0%. Furthermore, the presented DRL controller can also take effect under superimposed multifrequency sinusoidal gusts.