Large eddy simulation of compressible turbulent jets

Increasing noise regulation at urban airports force jet engine manufactures to develop and build quieter engines. Over the last 10-15 years, a significant reduction in fan and mechanical noise has been achieved. However, the reduction in jet noise over the same time period is fairly small and a major reduction in acoustic emissions of jet engines has to come from a reduction in jet noise. Traditional jet mixing-noise predictions are based on a statistical description of the jet turbulence; space-time correlations of the Lighthill quadruples (Lighthill 1952, Lighthill 1954) are specified in the jet. Further refinements are, however, needed to obtain a reasonable prediction of the directivity and spectral distribution of the radiated noise (Goldstein 1976). Significant improvements are achieved by modeling the effects of source-convection (Lighthill 1954, Ffowcs Williams & Hawkins 1969) and mean-flow refraction (Mani 1976). Reynolds-averaged mean flow calculations can be used to provide the source-strength distribution and the length and time-scale estimates needed in the source models (Bechara et al. 1994). The empirical input needed in such an approach places limits on the range of applicability of this method. It is, therefore, desirable to develop methods which obtain the unsteady flow data with much less empirical input. Due to the inherently high Reynolds numbers of the jet formed by a jet engine, LES seems to be the only feasible candidate to obtain the necessary unsteady data for the jet. Some attempts for round jets have been made by Choi et al. (1999) and Boersma & Lele (1999). For simpler geometries like plane mixing layers, various simulations have also been performed by Vreman et al. (1996). In this paper we will describe an LES method which uses the standard compressible LES filtered variables (no Favre averaging). The LES equation are discretized with accurate numerical schemes with very little artificial diffusion. So the LES models have to supply all the damping of the non-linear terms. In LES models which use low order numerics, damping can come from the LES subgrid model or from the numerics, and it is often very difficult to separate those two effects. This makes the physical interpretation of the results, especially those related to subgrid motions, difficult.