Large-eddy Simulation of Variable Density Flows

Abstract

A fully compressible direct numerical simulation flow and combustion solver (S3D) is modified and turned into a Large-eddy simulation (LES) solver. In this study, Favre-averaged governing equations are formulated first, supplemented with the classical Smagorinsky model and the dynamic procedure. To simulate low-Mach number flows, the speed of sound is artificially reduced while preserving the zero-Mach number physics. This pseudo-compressibility method is called Acoustic Speed Reduction (ASR). With ASR, the code has the capability to compute low-Mach number flows in an efficient way.

The boundary conditions in S3DLES are based on a one-dimensional characteristic analysis. To stabilize the solution, a buffer layer treatment is introduced at outflow boundaries to reduce acoustic reflections. The resulting flow is stable and produces results that compare well with a reference study. The implementation of the Smagorinsky model and other sub-models are validated using published plane jet simulation results with well-defined flow and perturbation conditions.

A second test case is the simulation of a round thermal plume. The ASR method is adopted to increase the computational efficiency by a factor of at least 10, thus making the computation of a 3-D round plume feasible on a small-scale cluster.

A third configuration is the simulation of a saltwater plume that was studied experimentally at UMD and is analog to a gaseous thermal plume. A comparison methodology between saltwater and gaseous plumes is developed. It is found that the computational requirement of a configuration that includes both the near- and far-field remains large and grid-resolution in our simulations remains marginal.

The fourth and last simulation takes advantages of the compressible flow formulation and considers flow-acoustic-interactions as a part of a thermoacoustic study. Three streams of different densities and momentums are introduced into a wall-confined domain. The flow is acoustically excited by an acoustic driver. The amplitude and phase of the driver are controlled. The high frequency modal response of the chamber compares well with experimental results. A variety of numerical tests in 1D, 2D and 3D configurations reveal the mechanism of transverse resonance and the resulting flow-acoustic interactions. This suggests that S3DLES will be a good prediction tool for future combustion noise and combustion instability studies.

Overall, the series of tests presented in this work serve to document the strengths and weaknesses of the current version of S3DLES.