DISCUSSION OF SOURCE TERMS IN COMPUTA- TIONAL AEROACOUSTICS OF ANISOTHERMAL FLOW USING A LOW MACH NUMBER APPROXIMATION. Cyril Nana, David Marx, Christian Prax and V´ eronique Fortun´ e Institute PPRIME, Departement of Fluid Flow, Heat Transfer and Combustion, Universit´ e de Poitiers, ENSMA, CNRS, T´ el´ eport 2 - Bd., Marie et Pierre Curie B.P. 30179, 86962 Futuro- scope Chasseneuil Cedex, France, e-mail: cyril.nana@univ-poitiers.fr A hybrid method is presented to compute the sound emitted by low Mach-number flows. The flow solver is a low Mach number approximation of the Navier-Stokes equations. It allows going beyond the incompressible approach by taking into account temperature and density inhomogeneities. The radiated sound computation is performed using linearized Euler equa- tions. It is well known that these equations support unstable vorticity modes that can spoil the acoustic result. One way to avoid the vortical mode development is to not excite it. To do this, it is proposed to use the pressure gradient as a source term. This source term can be used for both isothermal and anisothermal flows. The method has been applied to isothermal and anisothermal excited shear layers. The validity of the proposed method is assessed by com- parison to a direct noise computation. The isothermal case is presented here, the anisothermal case and temperature effects will be discussed at the congress. 1. Introduction Aerodynamically generated noise prediction has become a major issue in transport industry. Numerical aeroacoustic computations have established themselves as powerful tools to predict noise radiated by many types of flows. Two classes of methods are available. The first class of methods consists in performing Direct Noise Simulation (DNS). The compressible Navier-Stokes equations are calculated both in the aerodynamic source region and in the acoustic far field [1, 2, 3]. The connection between the dynamic flow and the sound produced by it is done naturally and requires no model for the sound source. This method requires large computational resources and is inefficient in the low Mach number range. This has motivated the second class of methods, known as hybrid methods [4]. For Mach number less than about 0.3, these methods can lead to a speed-up factor of up to 30 over the DNS [5]. They consist in splitting the full computation into a dynamic flow computation and a sound propagation computation, using a source model in between. The flow computation is typically incompressible [4, 5], but density and temperature inhomogeneities can also be taken into account [6, 7]. The noise computation can be done using some kind of perturbed equations, such as the linearized Euler equations (LEE) [2, 6, 8]. One well known problem with the LEE is that they can sustain unstable vortical modes that can spoil the noise computation. One strategy to avoid this mode is to modify the equations so that they do not support the mode anymore [2, 8, 9]. But a detrimental effect of this is to neglect some sound/flow interactions. Another strategy consists in not exciting the ICSV18, 10–14 July 2011, Rio de Janeiro, Brazil 1