1. 1 Effects of Inflow Forcing on Jet Noise Using Large Eddy Simulation P. Lew, A. Uzun, G. A. Blaisdell & A. S. Lyrintzis School of Aeronautics & Astronautics Purdue University, West Lafayette, IN. January 6, 2004 42 nd Aerospace Sciences Meeting and Exhibit Reno, NV AIAA 2004-0516

2. 2 Motivation Current CFD calculations which include the jet nozzle are mostly restricted to a Reynolds Average Navier-Stokes (RANS) approach –Prohibitive number of grid points to resolve the shear layer in LES. Forcing yields a way to replace a jet nozzle for LES –Pros: Computationally cheap and easy to implement –Cons: Results are sensitive to forcing parameters

3. 3 Motivation (cont’d) Flow development and far-field noise of the jet were affected when selected parameters were changed in the inflow forcing (Bogey & Bailly 2003, Bodony & Lele 2003) Parameter that had the greatest impact was the number of azimuthal forcing modes –Used 16 modes in total –Removing the first 4 modes resulted in a more quiet jet

4. 4 Objective Using our LES methodology, investigate and establish further trends on the effects of inflow forcing on: –Turbulent flow development –Far-field jet noise

5. 5 LES Methodology LES methodology developed by Uzun et al. (AIAA 2003-3322) –6 th -order compact scheme for interior nodes –4 th -order centered compact scheme for points next to the boundaries –3 rd -order one-sided compact scheme for boundary nodes –Sponge zone is attached downstream of the physical domain

6. 6 LES Methodology (Cont’d) –Tam & Dong’s 3-D radiation and outflow BCs on boundaries –6 th -order tri-diagonal compact spatial filter used as an implicit SGS model  Smagorinsky results sensitive to C sgs  Localized dynamic SGS model computationally expensive (50% increase in CPU)  Only looking for trends

7. 7 LES Methodology (Cont’d)

8. 8 Vortex Ring Forcing Proposed by Bogey et al. 2003 Simplified expressions Total number of modes = n modes + 1

9. 9 Setup Domain size: (x, y, z) = (25, ±15, ±15) r o Grid points: (Nx, Ny, Nz) = 287 x 128 x 128 Approx. 4.7 x 10 6 points ( Every other grid point is shown)

10. 10 Setup (Cont’d) Jet inflow conditions –Mach 0.9 and Re = 100,000 (Isothermal Jet) Runtime for one case: 4 days on 64 CPUs (IBM SP3) Original setup has 16 modes in total Test case name No. of modes removed Baseline None (16) rf4 First Four Modes (12) rf6 First Six Modes (10) rf8 First Eight Modes (08)

11. 11 Setup (Cont’d) Vortex Ring Forcing

12. 12 Results – Growth rates Under-prediction due to short domain length Need x > 45r o to get correct growth rates Test Case Growth Baseline 0.076 rf4 0.071 rf6 0.074 rf8 0.078

13. 13 Potential Core Lengths Test Case Location Baseline 11.54r o rf4 13.07r o rf6 13.43r o rf8 13.45r o Potential Core Jet develops slower as more modes are removed –Experiments: Transitional jet = 10 r o, (Raman, 1994) Initially turbulent jet = 14 r o (Arakeri, 2002) Current observation is in good agreement with Bogey and Bailly’s numerical experiments

14. 14 Turbulence intensities – axial (within shear layer, r = r o )

15. 15 Turbulence intensities – radial (within shear layer, r = r o )

16. 16 Turbulence intensities Shift in peak turbulence intensities due to a longer potential core –Trends so far agree well with Bogey However, radial peak intensities show unexpected increases for rf6 & rf8 –Uzun (2003) reported a similar observation for rf6 (M=0.9, Re = 400,000) –Further investigation is needed

17. 17 Far Field Aeroacoustics Methodology –Ffowcs Williams-Hawkings surface integral acoustic technique: Open and closed control surfaces are used –Acoustic data collected every five time steps over 25,000 time steps –Based on current spatial grid resolution we resolve a maximum Strouhal number of St = 1.1

18. 18 Far Field Aeroacoustics

19. 19 OASPL at 60r o (Open Control Surface)

20. 20 OASPL at 60r o (Closed Control Surface)

21. 21 SPL at r = 60r o @  = 60 o (Open CS)

22. 22 SPL at r = 60r o @  = 60 o (Closed CS)

23. 23 Conclusions The effect of removing modes for a vortex ring forcing was studied for an LES code (with a filter used as an SGS model) As more modes are removed –the potential core becomes longer –the peak radial turbulence intensities increase –OASPL increases slightly

24. 24 Recommendations Extend computational domain to about 60r o –Computationally costly Include part of nozzle geometry for LES to alleviate uncertainty of forcing

25. 25 Acknowledgements Indiana 21 st Research Century & Technology Fund National Computational Science Alliance under grant CT0100032N Simulations were run on SGI Origin 2000 and IBM SP4 at UIUC, Urbana-Champaign Also utilized Purdue University’s 320-node and Indiana University’s 600 node IBM SP3 supercomputers