• rectangular cylinder
• numerical simulations
• turbulent wakes
• dynamic stall model
• Wind Energy
• BARC benchmark
• upstream edge rounding

This thesis is aimed at contributing to the numerical modeling and simulation of large turbulent wakes originating from flow separation. Two classes of problems have been investigated, namely the flow around wind turbines and the flow around an elongated cylinder. As for the first problem, to reduce the computational cost of the wind turbine numerical simulations, actuator models are widely used. Nowadays, the most accurate model is the actuator line (ALM), whose accuracy is further investigated for the predictions of turbulent separated wakes. Since the ALM accuracy depends on some free parameters, an optimal setup for the ALM is obtained through a stochastic sensitivity analysis of the model to its parameters. For the vertical axis turbines applications, an extensive study has been done to model the so-called dynamic stall phenomenon under pitching conditions. In particular, the ``deep'' stall regime is considered. A model is proposed, which has a low implementation and computational complexity, but yet is able to deal with different types of dynamic stall conditions. The second problem is the flow around a rectangular cylinder, having a chord-to-depth ratio equal to 5, which is the object of the benchmark BAR. Some contributors to the benchmark highlighted a sort of paradox: high fidelity numerical simulations mismatch with the experiments, which are in good agreement with less accurate simulations. The main contribution to the benchmark by the present thesis is that this paradox can be explained by the fact that the upstream edges in the numerical simulations are perfectly sharp while they have a certain degree of roundness in experiments. This is shown through a sensitivity analysis of the LES results to the curvature radius.