Short Introduction

The prediction of aerodynamic noise generation is of ever-increasing importance for vehicle and airframe manufacturers. The reduction of noise is among the most important design criteria. For the numerical simulation of aeroacoustic sources and the propagation of noise into the far-field, various Computational Aeroacoustics (CAA) procedures have been developed. The most accurate numerical approach for aeroacoustic prediction is DNS with appropriate high order schemes since it resolves all turbulent eddies down to the Kolmogorov scale. However, for high Reynolds number flows, the required computational resources lie far beyond todays sufficiently available supercomputers. Therefore, cheaper methods that rely on simplified flow representations have been and still are developed. Moreover, the use of hybrid techniques allows to the restrict the expensive CFD calculation to the sources regions, whereas sound propagation is handled by specific numerical tools (acoustic analogies, simplified flow equations et cetera).

Among these cheaper CFD techniques, Large Eddy Simulation (LES) is closest to DNS. LES resolve only the largest scales that contain most of the flow energy and models the energy transfer to smaller scales. The LES method has demonstrated its capability in lots of investigations of turbulent low Reynolds number flows and recently for higher Reynolds number flows. At high and moderate Reynolds numbers, the LES method requires small grid sizing and corresponding small time steps to resolve the small turbulent scales of wall bounded layers in the flow field. The computational costs are a pitfall of the LES in Aeroacoustics of high Reynolds number flows. A new class of cheaper CFD techniques [Shur et al.] is based on a coupling between LES models in regions of highly unsteady turbulent flow and standard turbulence models (RANS) in steady low turbulence flow regions and in non-separated near wall flows. The advantage of this zonal approach is that time and space resolution can be coarser in the steady flow regions whereas an accurate down-to-grid LES-like resolution is only achieved in the regions where they are required. Although, the practical implementation and tuning of this Detached Eddy Simulation (DES) in various flow situations is still not fully understood, it has already demonstrated its capability to predict complex wall-bounded and separated flows [Schmidt & Thiele];[Strelets]. In the present paper, the feasibility of using DES to model broadband noise sources is investigated on a well-known test case that has been experimentally and numerically (using LES and RANS) studied [Jacob et al.]. Both mean flow and unsteady flow parameters from the experiment are compared to the present computations. The influence of the DES turbulence models onto the radiated sound and the flow field are investigated. The configuration is that of an airfoil located in the wake of a rod that contains both periodic and broadband fluctuations, in particular the broadening of noise of the main Strouhal peak. The rod wake combines a deterministic periodic vortex shedding with a random broadband turbulent wake. This test case is a simple model representing different types of industrially relevant noise sources as airfoil high-lift configurations (slat noise), trailing edge noise, landing gears and turbo machinery-noise of rotor-stator interaction.

For validation of the present study, the DES flow fields are compared to LES simulations from Boudet and Magagnato as well as measurements from Jacob consisting of hot-wire, wall pressure and PIV measurements. Some statistical results of the preliminary investigation of a 2D-URANS are shown. The acoustic far field is then computed from the flow field fluctuating values using the Ffowcs-Williams & Hawkings analogy for different integration surfaces. Spectra for some observer positions, as well as directivity plots are compared to experimental data.