14th European Conference on Turbomachinery Fluid dynamics & Thermodynamics
In the present paper, large eddy simulations (LES) are carried out in order to study two mechanisms of Fan/OGV broadband noise : turbulence interaction noise and airfoil self noise. First, the interaction noise from a thin flat plate impinged by an incoming synthetic turbulent flow is studied. Far-field acoustic spectra are compared to Amiet’s leading edge noise analytical model showing very good agreement and validating the numerical setup. Except for the effect of the far-field mesh refinement on the cut-off frequency, no significant influence of other numerical parameters is observed. The second case deals with the prediction of airfoil self noise from a plate with a sharp trailing edge and a chord-based Reynolds number of about 10e6. To avoid noise sources associated with Tollmien-Schlichting instabilty waves that may appear at such Reynolds number, the boundary layer has been tripped either by including a rectangular step in the numerical grid or by adding a source term in the stream-wise momentum equation. A parametric study is carried out on the sub-grid scale model, near-wall modelling, the grid topology and refinement in the near-wall and wake regions, the spanwise domain extent and the tripping method. The results are analyzed in terms of boundary layer’s and wake’s mean and RMS velocity profiles, far-field acoustic spectra as well as spanwise correlation length and wall pressure spectra (WPS) in the vicinity of the trailing edge. These last two characteristics are major inputs for trailing edge noise models. Empirical and analytical models as well as existing DNS data are used for validation purposes. Finally, an Ultra High Bypass Ratio fan geometry is studied at a subsonic regime over a limited radial slice of the rotor-stator stage. The original 16 blades and 31 vanes configuration has been transformed to 32 vanes to allow for a 2π/16 angular periodicity by adjusting the vanes’ chord length to maintain the same solidity. RANS computations comparing the performance of the modified and original configurations show no significant differences in terms of flow topology. Based on the RANS results, the radial slice is taken at 80% of the span to minimize 3D-effects. Aerodynamic and aeroacoustic predictions from LES are analyzed and compared to the RANS results and to analytical models.