Using large eddy simulation to explore sound-source mechanisms in jets

This paper presents an analysis of data generated by means of large eddy simulation for a single-stream, isothermal Mach 0.9 jet. The acoustic field is decomposed into Fourier modes in the azimuthal direction, and filtered by means of a continuous wavelet transform in the temporal direction. This allows the identification of temporally localised, high-amplitude events in the radiated sound field for each of the azimuthal modes. Once these events have been localised, the flow field is analysed so as to determine their cause. Results show high-amplitude, intermittent sound radiation for azimuthal modes 0 and 1. The mode-0 radiation, which dominates low-angle emission, is found to result from the temporal modulation of a basic axisymmetric wave-packet structure within the flow. Similar intermittent activity, observed, again within the flow, for azimuthal mode 1 suggests a link between the modes 0 and 1 dynamics. Both the amplitude and spatial extent of the axisymmetric wave-packet are modulated, and the strongest axisymmetric propagative disturbances are found to radiate from the downstream end of the wave-packet at moments when the wave envelope becomes truncated. The observed behaviour is modelled using a line-source wave-packet ansatz which includes parameters that account for the said modulation. Inclusion of these parameters, which allow the wave-packet to “jitter” in a manner similar to that observed, leads to good quantitative agreement (accurate to within 1.5 dB), at low emission angles, with the acoustic field of the LES. This result is in contrast with results obtained using a time-averaged wave-packet (one which does not jitter), for which a 12 dB error is observed. This result shows that the said modulations are the salient source feature for the low-angle sound emission of the jet considered. Analysis of a longer time series shows the occurrence of several similar high-amplitude bursts in the axisymmetric mode of the acoustic pressure, and a calculation of the radiated sound for this longer time-series, again using the wave-packet ansatz, once again leads to good agreement with the LES (now accurate to within 1 dB).