Sensitivity analysis of analytical models for the prediction of trailing edge noise.

Airfoil self-noise is defined as the acoustic emission by an airfoil immersed in uniform and steady fluid flow. As in most aeroacoustic noise generation situations, it is emitted by flow unsteadiness that interacts with the body surface and leads to broadband noise. There are mainly two broadband noise-generating mechanisms: the turbulence-interaction noise, involving the breakdown of oncoming vortexes on the leading edge of the airfoil, and the boundary-layer turbulence scattering at the trailing edge, known as trailing-edge (TE) noise. In future, the increasing air traffic and the proximity of airport to the cities will produce several noise problems for the inhabitants. For this reason, noise prediction models assume increasing importance. Noise emission has to be reduced in order to guarantee an high comfort for both passengers and citizens. Another interesting application field is related to the wind turbines. They produce typical broadband noise that has negative effects on health, mainly psychological, for who lives around the wind farms. Reducing broadband noise is one of the main requirement for wind turbine design. However, airfoil self-noise is only a contribution of the acoustic pollution. Keeping in mind the necessity of an accurate acoustic model during the design phase, this document provides a useful approach for noise prediction. The thesis consists firstly into a comparison of several analytical methods that allow to predict the sound pressure level for TE noise, in far field approximation, and generated by airfoils at different angle of attack, immersed into an high Reynolds flux. Several models are studied to find an optimum configuration that well approximates the experimental measurements. However, in experiments, the different sound sources, due to the turbulence phenomena around the body, are not completely discernible. The contribution of radiated TE noise is studied for two geometries: NACA0012, a classical example of symmetrical airfoil used for wind design, and DU96-W-180, asymmetrical profile applied in wind turbines. This allows to evaluate in which geometrical configuration the acoustic prediction is more accurate. The acoustics model is implemented in post processing data. The pressure, velocity and wall stress distribution around the profile are obtained using a numerical simulation based on Reynolds Averaged Navier-Stokes (RANS) equations with Spalart-Allmaras turbulent model. Amiet's theory, is the first model introduced in literature. As far as concern analytical aspects, the method is based on the Power Spectrum Density (PSD) and on the introduction of correlation length as function of the frequency, according to Corcos hypothesis. However, the condition of high shape ratio must be satisfied. The pressure spectrum is obtained using empirical models such developed by Goody, Rozenberg, Kamuruzzaman or Lee. A simple solution was proposed by Amiet in his principal work but omitted in following chapters. Finally, the solutions are compared with experimental data. A sensitivity analysis is carried out in order to understand how the output of each acoustic model is influenced by the boundary-layer quantities predicted by RANS simulations. Monte Carlo method is combined with Saltelli's approach which introduce global sensitivity factors useful to quantify the weight of each single input parameter and estimate the uncertainty propagation in different analytical models.