Event-Triggered Adaptive Dynamic Programming for an Aeroservoelastic System: An Experimental Study

Abstract

The pursuit of higher efficiency has driven wind turbine blades and aircraft wings becoming increasingly slender and lighter. However, this results in structures that are more flexible and, consequently, more sensitive to atmospheric disturbances, leading to undesirable vibrations and even structural breakage. Efforts have been made to use active control algorithms to mitigate oscillations in these slender, air-operating structures, resulting in aeroservoelastic systems. However, the cutting-edge algorithms still fall short of adequately minimizing computational loads while fulfilling the robustness and adaptation needs of nonlinear, time-varying, uncertain, and underactuated aeroservoelastic systems. An event-triggered adaptive dynamic programming algorithm is proposed to overcome these obstacles. It also details the first real-world wind tunnel experiment of such an algorithm designed for an aeroservoelastic system. The experimental outcomes demonstrate that the proposed algorithm can learn from experiences online and autonomously adjust to various unforeseen situations. Additionally, it avoids the Zeno phenomenon both theoretically and practically. The designed triggering mechanism is capable of responding promptly to abrupt disturbances and decelerating the pace of control command updates once a gust has dissipated. Compared to its time-triggered counterpart, the experimental findings reveal that the proposed algorithm reduces the number of updates by 62.98% without compromising performance. This significantly reduces computational, communication, and actuation demands while enhancing structural longevity and safety.