This study develops a closed-loop optimization system that combines plasma flow control with a genetic algorithm to minimize the drag coefficient of a circular cylinder based on wind tunnel experiments. The system optimizes the plasma actuation parameters, namely the duty-cycle frequency f and the duty-cycle ratio \(\tau \) , by using experimental feedback data and significantly enhances the optimization efficiency at a Reynolds number of approximately 22,500. A pair of alternating surface dielectric barrier discharge plasma actuators is symmetrically positioned at azimuthal angles of \(\pm 90^\circ \) on both sides of the cylinder. The system communicates in real time with a signal generator and a pressure scanner through a LabVIEW program to adjust the actuator parameters. To reduce the influence of experimental errors during the optimization process, the drag coefficients corresponding to each parameter combination are cumulatively averaged and then incorporated into the genetic algorithm iteration. This averaging strategy increases the reliability of the optimization and shortens the required experimental time. The closed-loop system identifies and adapts to experimental asymmetries caused by slight errors, obtaining the true physical optimum rather than an idealized theoretical value. The optimization results show a drag reduction rate of 21.8%. Additionally, particle image velocimetry (PIV) confirms that optimized actuation suppresses vortex shedding, reduces wake velocity deficit, and stabilizes the wake structure, thereby lowering energy loss and improving aerodynamic performance. This study demonstrates the feasibility of machine learning-based closed-loop optimization in plasma flow control, providing a foundation for more robust, intelligent flow control strategies.
