Dynamic stall mitigation of a pitching aerofoil using a data-driven model

Abstract

Dynamic stall is an unsteady aerodynamic phenomenon which temporarily enhances lift and delays flow separation on a lifting surface, but is also associated with large load fluctuations that may compromise the structural integrity of the system. The present work, based on transient computational fluid dynamics (CFD) simulations, proposes a methodology to mitigate the undesired effects of dynamic stall on a pitching NACA 0018 aerofoil undergoing a large-amplitude sinusoidal oscillation. The study aims to alleviate the post-stall load fluctuations by introducing small modifications to the pitching kinematics of the aerofoil. The approach relies on the construction of a nonlinear data-driven model of the system, which is capable of predicting the time-varying lift, drag, and moment coefficients from a given angle-of-attack time series. This fast and accurate nonlinear model, based on neural networks, is coupled with a multi-objective genetic algorithm designed to optimise two competing objectives: the negative peak pitching moment coefficient and the mean lift coefficient. The optimised pitching parameters are identified by modifying the original sinusoidal motion through the superposition of two higher harmonics, with their amplitudes and phases being the design variables. The optimised aerofoil motion proposed by the genetic algorithm is subsequently evaluated through CFD analysis to verify the accuracy of the model predictions. Results show good agreement between the predicted and the actual transient aerodynamic coefficients, demonstrating that small adjustments to the pitching trajectory can lead to substantial reduction of the peak loads during deep dynamic stall. The obtained results further underscore the usefulness of nonlinear data-driven models, which are particularly well-suited for integration into optimisation and control frameworks that require both accuracy and a fast evaluation time.