Generating Synthetic Inhomogeneous Turbulence from RANS Statistics

摘要

In this contribution to the special session on Generation of Inflow Turbulence for Wall-Bounded LES, we consider a synthetic turbulence method designed to generate realistic eddies from inhomogeneous turbulence statistics. This method was originally devised as a set of source terms for hybrid RANS/LES, but is equally applicable as a stand-alone method for generating unsteady inlet turbulence from any given set of second moments and dissipation rates. This synthetic turbulence approach is based on Kraichnan's idea of generating energy distributions via a sum of Fourier modes. Tensor scaling allows an accurate reproduction of the target Reynolds stresses, while RANS-determined length- and time-scales are mimicked via the introduction of appropriate spatial and temporal wavelengths, along with windowing functions to avoid decorrelation at large distances and times when dealing with inhomogeneous turbulence statistics. The method allows LES content to be generated at inlets to LES domains using nothing but user-supplied (or previously-computed) RANS statistics. The method involves no additional transport equations, requires only modest computational and memory requirements and is applicable to arbitrary meshes, with filtering in the frequency domain ensuring that only resolvable-energy content is drained from the RANS kinetic energy. The method has been used at a range of Mach numbers, however, work is still underway to isolate the effects of dilatational and solenoidal turbulence, specifically for aeroacoustic applications, where the aim is to minimize the introduction of spurious noise sources arising from arbitrary divergence sources in the synthetic fluctuating velocity fields.