Turbulent RSI (rotor-stator interaction) mechanism is a major broadband source contribution of turbofan noise generation. Acoustic prediction tools used by Industry are based on flat-plate cascade response models with restrictive assumptions on flow and geometry. Due to huge CPU memory and time cost required, Large Eddy Simulations of the complete fan-OGV stage are still out of reach (apart from recent impressive results obtained using the Lattice Boltzmann Method). This paper presents an alternative approach based on the use of a 3-D CAA (Computational Aeroacoustics) code solving the linearized-Euler equations applied to the disturbances and coupled with a synthetic turbulence injection model. The inflow turbulence is synthetized by means of a sum of harmonic gusts with random phases. The Fourier-mode amplitudes are trimmed by a 2 or 3-wave number Von-Karman or Liepmann turbulence spectrum. Swirling convection of the synthetic turbulence is provided by a 3D RANS mean flow solution and interpolated at the nodes of the CAA grid. In this paper, our methodology is first validated on a benchmark case (fully annular duct with swirling flow and a prescribed turbulence) and then applied for the first time to an industrial turbofan in the framework of a European project, TurboNoiseBB. Previous implemented 2D formulation (2-wave number spectrum) for turbulence generation is extended here to 3D (axial, radial, and angular modes) in order to study the sensitivity on cascade effects.