A hybrid analysis methodology for improved accuracy in low cost jet noise modelling

摘要

The tools currently available for jet noise prediction range from high precision, high cost approaches such as DNS and LES, to lower cost methods which require the turbulence to be modelled in some way (e.g. statistical and SNGR type approaches). Where the latter are concerned the error incurred in modelling the turbulence can often be orders of magnitude greater than the acoustic energy generated. It is thus imperative that the models used in these approaches be extremely accurate. DNS or LES computations give the kind of precision necessary, but at a prohibitive computational cost. This work comprises a hybrid analysis for improved accuracy of low cost noise prediction schemes. The methodology consists in using data from a LES (this data having been validated by experiments) to validate and improve the models used in both statistical and SNGR jet noise prediction schemes. Analytical models for the spatiotemporal velocity corrlelation tensors which have been developed to deal with the inhomogeneous, anisotropic structure of a jet are compared with their numerically generated counterparts from the LES. The spatial evolution of their form and its relation to local quantities such as the integral scales and turbulence intensities are investigated using the LES, with a view to understanding how the analytical models can be adapted so as to capture this evolution.