Final work : Low speed dynamic stall modelling for 2D airfoils

Abstract

Research on low Reynolds unsteady aerodynamics is of great importance on the design of small unmanned aerial vehicles. Dynamic stall is the periodic separation and re-attachment of the boundary layer around oscillating wings. The physics of the phenomenon are very complex and have previously been described for different applications, such as helicopter rotor blades and horizontal axis wind turbines.

Dynamic stall can be modelled by using semi-empirical models. Two examples are the Leishman-Beddoes and ONERA models. An adaptation of the Leishman-Beddoes model was previously developed in the department, accounting for different behaviour in low Reynolds flow conditions.

In this thesis, CFD simulation results of the flow around a NACA0012 profile are compared to the corresponding experimental load coefficients in low Reynolds (~2*10^4), static conditions. There are some differences between both, especially for the chordwise force. Static stall onset is compared to the experimental results. It is shown that Kirchhoff's relation between the separation point and the normal force coefficient does not hold for the CFD results. It overestimates the attached portion of the airfoil. The CFD stall onset mechanism is also different from the one observed from the experiments, due to the lack of a laminar separation bubble.

Other important result is the exploration time evolution of dynamic stall in several different cases, using 2D Unsteady RANS simulations. In general, the load coefficients obtained compare well to the experimental results, though their agreement is not perfect. CFD predicts a large dip in the normal force and pitching moment not present in the experiments.

Vortex onset and evolution is described by using several pressure coefficient maps, overlaid by the instant velocity vectors. The simulated mechanism is explained in further detail. It is shown that several vortices form in each cycle, both over the suction surface and at the trailing edge.

From the time evolution of the flow field it is possible to obtain several parameters required for the semi-empirical model, which are then compared to the experimentally determined ones. While stall onset is delayed in the CFD simulations, the model parameters show good agreement with the experimentally determined ones.