Noise Prediction from Rectangular Jet Flow Using LES

Abstract

We investigate the acoustic radiation from a 1:5 aspect ratio transitional rectangular jet predicted by Large-Eddy Simulation (LES) employing the approximate deconvolution model (ADM). Previous comparison of results from Direct Numerical Simulation (DNS) for acoustic far-fleld prediction based on Lighthill’s acoustic analogy with LES prediction has shown good agreement for the low frequency range, however, higher frequencies in the LES prediction have been contaminated by spurious noise. In the present study the origin of this non-physical noise was analyzed. As flrst reason, the acoustic far-fleld analysis method applied was found to be a limiting factor. A cutofi frequency estimate has been obtained beyond which LES prediction is not reliable but up to which frequency LES prediction was good. Second, the in∞uence of the subgrid-scale model was assessed by analyzing (i) the efiect of reduced deconvolution order and (ii) an LES with no subgridscale modelling but spatial flltering after each timestep instead. Additionally, a higher resolution LES was performed with similar fllter width as in the original LES in order to reduce numerical errors close to the fllter cutofi. Both, reduced deconvolution order and spatial flltering, resulted in a more rapid drop-ofi of the far-fleld spectra up to the cutofi frequency, which is attributed to the higher dissipation in the ∞ow simulation. This suggests that the type of employed subgrid-scale model is not of importance for aeroacoustic prediction given that enough dissipation is provided to dampen spurious waves but as little as possible to reproduce the spectra at higher frequencies.