Data-Driven Model Predictive Control of Airfoil Flow Separation

This paper numerically investigates the performance of active flow control based on adaptive model predictive control without prior analytical modeling of any kind. Recursive least squares with variable-rate forgetting is used for online closed-loop system identification. The identified model is then used for receding-horizon optimization based on quadratic programming. This technique is applied to an airfoil model simulated using computational fluid dynamics. Two cases are considered. In the first case, the tangential component of the flow velocity is commanded at the location of a single flow-velocity sensor. In the second case, an array of flow-velocity sensors is used to estimate the location of the flow-separation point, which is used to command the flow-separation point.