Validation of the Fast Random Particle-Mesh Method for Broadband CAA of a Forward-Facing Step and its prediction sensitivity for geometrical modifications

Numerical studies regarding broadband noise of a Forward-Facing Step at Re_h=8000 are performed and compared to already existing experimental data obtained in the Aeroacoustic Windtunnel of University of Erlangen-Nuremberg. Results are transferred to an experimentally-decoupled setup at Re_h=30000 with a more relevant Reynolds number for industrial applications. The aeroacoustic near- and farfield is computed by a Hydrodynamic-Acoustic Splitting approach by means of a perturbed convective wave equation. Aeroacoustic source terms are computed either by incompressible LES or by the FRPM method, a stochastic noise source method modelling velocity fluctuations in time-domain based on time-averaged turbulence statistics. The synthetized velocity fluctuations reproduce target two-point space-time correlations and are transformed with a Poisson equation to hydrodynamic pressure fluctuations obtained by combining the incompressible momentum and continuity equation. The wall-resolved LES is in good agreement with experimental data and serves as an extended validation database for FRPM at the lower Reynolds number. Time-averaged field quantities are extracted from LES and two RANS models to validate the stochastic method. Results of FRPM-CAA based on time-averaged LES and RANS data are within good agreement to the LES in both their farfield radiation levels as well as spanwise and streamwise nearfield correlations. For the higher Reynolds number, a good agreement can be achieved by reducing the empirical correlation length of the synthetized vortices in FRPM. Modelling the acoustic radiation with frozen turbulence properties is sufficient. Geometrical modifications like different step heights, chamfered and rounded steps are used to investigate the Delta-prediction sensitivity. The step heights and chamfered steps are in good agreement whereas the rounded geometries deviate slightly to the LES results. However, the trend of noise reduction is reproduced. CAA excited by FRPM based on RANS data only needs 12% of the computational effort compared to the conventional hybrid LES-CAA approach and therefore reveals its enormeous potential to achieve aeroacoustic broadband noise predictions for optimization purposes.