Data-Driven Control of Shock Train Using Radial Basis Function Neural Network

The shock train is highly sensitive to variations in the incoming flow and backpressure, posing a potential safety risk if it is expelled from the inlet. To prevent unstart, detecting and controlling the shock train location are essential. The nonlinear dynamics of the shock train present a challenge for traditional control methods, which are typically linear and rely on precise mathematical models. The data-driven control approach employed in this study, which utilizes radial basis function neural network (RBF-NN), offers a solution to these problems. It directly derives control laws from data, enabling adaptive parameter adjustment. An experimental investigation was conducted to control the shock train in an inlet at Mach 4.2, using the shock train location as the feedback signal and the downstream flap angle as the control input. The standard deviation method was used to accurately evaluate the shock train location and mitigate errors caused by reflected shocks. Real-time control input required for determining shock train displacement was calculated using the RBF-NN. Considering disturbances and system lag, the favorable pressure gradient downstream of the throat significantly enhanced control performance. The development of data-driven control ensures stable and high-performance operation for high-speed, air-breathing aircraft across various flight conditions.