Supersonic jet noise prediction and noise source investigation for realistic baseline and chevron nozzles based on hybrid RANS/LES simulations

Open Access
Author:
Du, Yongle
Graduate Program:
Aerospace Engineering
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
May 06, 2011
Committee Members:
  • Philip John Morris, Dissertation Advisor
  • Philip John Morris, Committee Chair
  • Dennis K Mc Laughlin, Committee Member
  • Victor Ward Sparrow, Committee Member
  • Daniel Connell Haworth, Committee Member
Keywords:
  • Computational aeroacoustics
  • jet noise
  • noise sources
  • LES
Abstract:
Jet noise simulations have been performed for a military-style baseline nozzle and a chevron nozzle with design Mach numbers of Md=1.5 operating at several off-design conditions. The objective of the current numerical study is to provide insight into the noise generation mechanisms of shock-containing supersonic hot jets and the noise reduction mechanisms of chevron nozzles. A hybrid methodology combining advanced CFD technologies and the acoustic analogy is used for supersonic jet noise simulations. Unsteady Reynolds-averaged Navier-Stokes (URANS) equations are solved to predict the turbulent noise sources in the jet flows. A modified version of the Detached Eddy Simulation (DES) approach is used to avoid excessive damping of fine scale turbulent fluctuations. A multiblock structured mesh topology is used to represent complex nozzle geometries, including the faceted inner contours and finite nozzle thickness. A block interface condition is optimized for the complex multiblock mesh topology to avoid the centerline singularity. A fourth-order Dispersion-Relation-Preserving (DRP) scheme is used for spatial discretization. To enable efficient calculations, a dual time-stepping method is used in addition to parallel computation using MPI. Both multigrid and implicit residual smoothing are used to accelerate the convergence rate of sub-iterations in the fictitious time domain. Noise predictions are made with the permeable surface Ffowcs Williams and Hawkings (FWH) solution. All the numerical methods have been implemented in the jet flow simulation code "CHOPA" and the noise prediction code "PSJFWH". The computer codes have been validated with several benchmark cases. A preliminary study has been performed for an under-expanded baseline nozzle jet with Mj=1.56 to validate the accuracy of the jet noise simulations. The results show that grid refinement around the jet potential core and the use of a lower artificial dissipation improve the resolution of the predicted high frequency noise spectra. The results also show that the predicted low frequency noise spectra are sensitive to the axial extent of the acoustic data surface, and the high frequency noise spectra are affected by the radial size of the acoustic data surface. The baseline nozzle has been studied at several off-design conditions with Mj=1.36, 1.47 and 1.56. Although the noise levels at mid to high frequencies are over-predicted at several shallow polar angles, the predicted noise spectra in the peak noise radiation direction and upstream directions agree very well with the experimental measurements. More encouraging is that the frequencies and amplitudes of the broadband shock-associated noise (BBSAN) are captured accurately at all three operating conditions. Three techniques are used to examine the noise source characteristics. The two-point space-time correlation method is used to analyze the statistical characteristics of the turbulent eddies. The direct flow-acoustic correlation technique and the beamformed acoustic pressures are used to reveal the different noise generation mechanisms of the large-scale and fine-scale turbulent fluctuations. The chevron nozzle simulations have been performed at the same operating conditions to evaluate the noise reduction effects. Special treatments are proposed to address the numerical difficulties caused by the chevrons. The impact of chevrons on the near-field noise sources and far-field noise radiation is simulated using the immersed boundary method (IBM) to overcome the great difficulties in grid generation. A non-matching block interface condition is developed to allow the grids to be greatly refined around chevrons for a higher accuracy of simulations without increasing the mesh size significantly. The predicted noise spectra agree very well with the acoustic measurements of the baseline nozzle, considering the small noise reductions of the chevrons at the given operating conditions. No apparent over-prediction is observed. However, the noise reductions are over-predicted because of the over-prediction of the baseline nozzle noise level at some polar angles. Analysis shows that the chevrons generate strong streamwise vorticies and induce strong lateral secondary flows near the nozzle exit. The enhanced turbulent mixing increases the noise source intensity and efficiency near the nozzle exit, and creates a high frequency noise penalty. But it reduces the turbulence intensity in the main jet potential core, and decreases the low frequency noise level. Both the flow and noise results show that the effects of chevrons on the jet flow and noise reduction depend highly on the operating conditions.