Optimization of rotor aerodynamic performance via machine learning-based plasma flow control

As the core lifting component of helicopters, the rotor operates in a complex three-dimensional unsteady flow environment, suffering from challenging flow phenomena such as spanwise flow and reverse flow regions that severely limit aerodynamic performance. While active flow control techniques—particularly plasma actuation—offer promising solutions, their application to rotors is limited by mechanical complexity and the inefficiency of conventional discrete-parameter optimization methods. To address these challenges, this study introduces a machine learning-based plasma actuation control framework for performance enhancement of a three-dimensional rotor. Based on Particle Swarm Optimization (PSO), the actuation parameters are globally optimized, resulting in a 14% relative increase in thrust coefficient in hover and 7.6% in forward flight. A Deep Neural Network (DNN) is further developed to map actuation parameters to performance outcomes, effectively expanding the experimental dataset. SHAP (Shapley Additive explanations)-based interpretability analysis identifies the duty cycle as the most critical control variable. The integrated approach demonstrates efficient parameter space exploration, data-driven modeling, and feature importance analysis, providing a viable strategy for rotor flow control under complex aerodynamic conditions.