Noise Prediction for Increasingly Complex Jets. Part I: Methods and Tests

Abstract

This Part I presents a detailed description of a numerical system built and tested with the final goal of reaching an accuracy of 2–3 dB over a meaningful range of frequencies for airliner engine noise, while having low empiricism and a general-geometry capability. The turbulence is treated by Large-Eddy Simulation with grids of around 1 million points, slightly upwind-biased high-order differencing, and implicit time integration. The code can incorporate boundaries and multi-block grids (thus avoiding the centerline singularity), and capture shocks. The sub-grid scale model is de-activated, because on present grids it strongly interferes with transition in the mixing layer. Without unsteady inflow forcing, the shear-layer roll-up and three-dimensionalization are realistic and reasonably insensitive to the grid. The far-field noise is computed using the permeable Ffowcs-Williams/Hawkings (FWH) formulation without external quadrupoles. The treatment of the disk that closes the FWH surface near the outflow must be approximate, since the turbulent region is unbounded, and is crucial; it benefits from a change of variable from density to pressure, and other mitigating steps. Tests are presented in support of the key elements of the strategy. In a simple isothermal jet, the system is close to the 2–3 dB target both in terms of directivity and of spectrum, up to a Strouhal number of about 1.5. In Part II, the following effects are explored with overall success: jet Mach numbers from 0.3 to slightly supersonic with under-expansion (generating shock cells), jet heating, co-flow, and “synthetic chevrons.”