Mécanique des fluides, transferts et rayonnement sonore

Transcription

1 Turbulence and Aeroacoustics Research team of the Centre Acoustique École Centrale de Lyon & LMFA UMR CNRS 5509 SMI - Turbulence en mécanique des fluides, Académie des Sciences, 14 juin 2011 Mécanique des fluides, transferts et rayonnement sonore Christophe Bailly & Christophe Bogey Université de Lyon, Ecole Centrale de Lyon & LMFA - UMR CNRS 5509 Institut universitaire de France 1

2 Acknowledgments Olivier Marsden (ECL, Assitant Prof.) Thomas Castelain (ECL, Assitant Prof.) Sébastien Barré (Dassault-Aviation) Julien Berland (EDF) Vincent Fleury (ONERA) Benoît André (ECL, Ph.D. Student) 2 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

3 Outline of the talk Motivations Example of aeronautics, commercial air transport & military applications Predicting jet mixing noise Scales and control parameters Large-eddy simulation (LES) and direct noise computation (DNC) Role of coherent structures Mixing and noise of turbulent subsonic jets Introduction to underexpanded supersonic screeching jets Concluding remarks 3 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

4 Motivations Physics-based predictions for real jets, i.e. dual, hot, with co-flow, shock-cells and noise reduction devices : shape optimization, variable geometry chevrons or fluidic actuators High-bypass-ratio nozzle (cfm56 type) chevrons on the fan and core nozzles (Loheac et al., SNECMA, 2004) QTD2 - Boeing - NASA AIAA Paper Castelain et al. AIAA Journal, 2008, 45(5) understanding of noise generation mechanisms providing reliable predictions giving insight for noise reduction Lobed exhaust ejector/mixer system CFM56-5C engine powering Airbus A340 4 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

5 Motivations Supersonic jets acoustic environment of space launchers at liftoff and protection of payloads military aircrafts (e.g. hearing protection of naval crew on aircraft carrier deck) Ariane V ECA - CNES flight st launch V broadband shock-associated noise in cruise conditions : cabin noise Pratt & Whitney FX631 jet engine (F-35 Joint Strike Fighter) Kleine & Settles, Shock Waves (2008) Take off from aircraft carrier (noise levels exceed 140 db) 5 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

6 Motivations Economic and societal impacts Noise levels (EPNdb, sideline), no normalization EPNdB (sideline) turbojet VC 10 Comet 4 1st generation B707 turbofan B720 Trident 1 2nd generation DC 8 Caravelle 10 DC 10 turbofan 3rd generation B turbofan B L 1011 A300 B A F28 B A B BAe 146 A B ATR42 A A A A380 A319 MD 80 F100 ERJ 145 Falcon 900 ATR72 G IV G V year of entry into service Advisory Council for Aeronautics Research in Europe (ACARE ) ; Traffic growth must be compensated for by quieter aircrafts Jet noise during take-off remains the major component of total aircraft community noise (between a third and half of the energy) Sound insulation tax, France (TNSA) 50 M /year 6 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

7 Subsonic turbulent jets Aeroacoustic scaling Acoustic Mach number M a M a = u j c noise M n a D u j Strouhal number St St = fd u j = f u j /D non-dimension frequency nozzle Reynolds number Re D Re D = u jd ν = D2 /ν D/u j viscous time convective time 7 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

8 Subsonic turbulent jets Reynolds number Re D = u j D/ν Prasad & Sreenivasan (1989) Re D 4000 Dimotakis et al. (1983) Re D 10 4 Kurima, Kasagi & Hirata (1983) Ayrault, Balint & Schon (1981) Mollo-Christensen (1963) Re D Re D Re D = Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

9 Subsonic turbulent jets Initial conditions at the nozzle exit Kolpin, J. Fluid Mech. (1964) Mollo-Christensen, Kolpin & Martucelli, J. Fluid Mech. (1964) Re D = Re D = Re D = Re D = Transition region moves from the mixing layer to the nozzle boundary layer as Re D ր 9 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

10 Subsonic turbulent jets Disparity of scales Isothermal jet, λ a /δ θ Re D /(M a St) (Re D = 10 6, M a = 0.9, r/d 10) p 2 a θ=90 o M7.5 a u a/u 10 4 p a/p 10 3 u a p a λ a δ θ u j D u p x c λ a /δ θ 10 3 u /u j 0.16 Mollo-Christensen (1963), Re D = laminar potential core length x c 10 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

11 Direct computation of aerodynamic noise High fidelity flow/noise simulation in a physically and numerically controlled environment fluctuating pressure field p outside of the flow non-reflecting boundary conditions & WEM an error of 1% on the aerodynamic pressure field yields an error of 100% on the acoustic field! artificial turbulent state at nozzle exit to mimic turbulent BL Re D, Re δθ, u e/u j vorticity ω in the flow Barré et al., Int. J. AeroAcous. (2006) Bogey & Bailly, J. Fluid Mech. (2007) 11 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

12 Subsonic turbulent jets First Direct Noise Computation (DNC) of a subsonic round jet Freund, J. Fluid Mech. (2001), Re D = 3600 & M = 0.9 see also Colonius, J. Fluid Mech. (1997) Kurima, Kasagi & Hirata (1983) Re D n x n r n θ = pts δ θ /r small random perturbation to seed the instabilities and turbulence 12 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

13 Subsonic turbulent jets Direct Numerical Simulation M = 0.9 & Re D = NEC SX-8 cluster at HLRS center in Stuttgart, Germany 212 GFlops, 250 GB of memory, 30,000 CPU hours 16 vorticity norm in the plane z = 0 full jet in top view 0 x 27.5r 0 (116 pts bottom figure) n x n y n z = pts k c η 1.5 k c grid cut-off wavenumber x = y = z = r 0 /68 simulation time T = 3000r 0 /u j, 295,000 iterations δ θ = 0.01 r 0 Bogey & Marsden in Advances in Parallel Computing, 19, 2009 see also Bogey & Bailly, J. Fluid Mech., 629, Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

14 Large Eddy Simulation of turbulent jets Energy budget across a self-preserving jet M = 0.9 & Re D = n x n y n z = = pts time steps Nec SX-5/ h CPU Expts of Panchapakesan & Lumley (1993) Self-similarity region for 120r 0 x 150r 0 All the terms are explicitly calculated including the filtering dissipation, to check the sum (normalized by ρ c u 3 c δ 0.5) vorticity norm in the plane z = y/δ 0.5 mean flow convection production dissipation turbulence diffusion pressure diffusion expt. data of P. & L. (1993) 14 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

15 Large Eddy Simulation of turbulent jets Energy budget across a self-preserving jet M = 0.9 & Re D = Interpretation of Panchapakesan & Lumley (J. Fluid Mech., 1993) versus Hussein, Capp & George (J. Fluid Mech., 1994) P. & L. H.C. & G y/δ 0.5 y/δ 0.5 in agreement with P. & L. who neglected presssure diffusion and not with H.C. & G. who used turbulence modelling for dissipation Bogey & Bailly, J. Fluid Mech. (2009) 15 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

16 Large eddy simulation Closure : subgrid-scale model e.g. Sagaut (2006), Lesieur (2007) Projection modeling by a spacial convolution filtering ū = G u t ū + (ūū) + ( p/ρ) + ν 2 ū = σ sgs structural approach, find a model for the sgs stress tensor functional approach, surrogate the mean action turbulent kinetic energy balance is more important turbulent kinetic energy cascade is dominant dissipation Two main classes of methods in physical space turbulent eddy viscosity models hyperviscosity ( σ sgs ν h 2n ) or high-order explicit filtering Pruett et al. Domaradzki, Adams et al. Visbal, Gaitonde, Rizzetta ADM... LES - Relaxation filtering σ sgs = χg ũ Berland et al., J. Comput. Phys. (2008) & JOT (2008) 16 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

17 Large eddy simulation Status of large eddy simulation full-scale for laboratory jets (typically D = 2 cm, M = 0.9, Re D = ) and nearly mature numerical tool (basic statistics, turbulent kinetic energy budget, two-point space-time correlations, Reynolds effects) advances in «alternative subgrid-scale models» e.g. removing energy at the smallest resolved scale by explicit filtering with N 10 8 points, DNS at Re D 10 4 and LES at Re D 10 5, St Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

18 lacements Subsonic turbulent jet flow Space-time velocity correlations by dual-piv (Fleury et al., AIAA Journal, 2008) Re D = , M = 0.9, D = 3.8 cm, δ θ /D init At x = 5D, L (1) D, Kolmogorov scale l η 10 4 D Space-time second-order correlation functions R 11 (x, ξ, τ) and R 22 (x, ξ, τ) measured at x = (6.5D, 0.5D) L (1) 11 2δ θ L (1) 22 δ θ ξ 2 D ξ 2 D τ = 0 µs τ = 50 µs τ = 150 µs τ = 250 µs ξ 1 /D ξ 1 /D ξ 2 D ξ 2 D ξ 1 /D ξ 1 /D ξ 2 D ξ 2 D ξ 1 /D ξ 1 /D ξ 2 D ξ 2 D ξ 1 /D ξ 1 /D 18 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

19 Subsonic turbulent jets Initial conditions at the nozzle exit (visualizations by T. Castelain, ECL) Re D Re D Re D fully laminar u e/u j < 1% nominally laminar u e/u j 1% Re D 10 5 (Re δθ 300) transitional jets nominally turbulent u e/u j 10% Re D fully turbulent Re D Re D = u j D/ν Re δθ = u j δ θ /ν σ ue = u e/u j 19 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

20 Influence of exit boundary-layer thickness Transition from initially laminar jets δ θ = 0.024r 0 M = 0.9 Re D = 10 5 σ ue 1% δ θ = 0.012r 0 δ θ = 0.006r 0 Smaller shear-layer thickness results in delayed jet development and longer potential core δ θ = 0.003r 0 All transitions are characterized by shear-layer rollingup and a first stage of strong vortex pairings Bogey & Bailly, J. Fluid Mech., 2010, Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

21 Transition from initially laminar jets Influence of exit boundary-layer thickness M = 0.9 Re D = 10 5 σ ue 1% 0.25 <u z u z > 1/2 /u j z/r 0 u z/u j along r = r 0 δ θ = 0.024r 0 δ θ = 0.012r 0 δ θ = 0.006r 0 δ θ = 0.003r 0 Fleury et al., AIAA Journal (2008), M = 0.9 & Re D = rms velocity profiles (dual-peak shape, high values) typical of a first stage of vortex pairings 21 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

22 Transition from initially laminar jets Influence of exit boundary-layer thickness M = 0.9 Re D = 10 5 σ ue 1% SPL (db/st) θ = 40 deg St SPL (db/st) Experimental data at Re D θ = 90 deg St Additional noise induced by vortex pairing at frequency f 0 /2, with St δθ = f 0 δ θ /u j 0.012, see Zaman, AIAA Journal (1985) 22 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

23 Consequences on noise prediction from turbulent jets Subsonic turbulent jets Need for considering initially turbulent jets, i.e. u e/u j 10%, to prevent any form of pairing noise like in jets at high Reynolds number : less noisy jets are indeed observed for a natural smooth development of their turbulent boundary layer. This is of course not a denial of the existence of coherent structures! Expts Zaman, 1985, AIAA Journal & J. Fluid Mech. Bridges & Hussain, 1987, J. Sound Vib. Raman, Zaman & Rice, 1989, Phys. Fluids A far field pressure spectra at θ = 90 o Re D = laminar jet (u e/u j 0.03%, δ θ /D ) tripped jet (u e/u j 0.09%, δ θ /D ) 4 db St 23 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

24 Subsonic turbulent jets Tripped subsonic round jets fully laminar u e/u j < 1% nominally laminar u e/u j 1% transitional jets nominally turbulent u e/u j 10% Re D 10 5 Re D nominally laminar nominally turbulent by tripping with a laminar mean velocity profile, u e/u j 10% and δ θ ր fully turbulent Re D Computing initially fully turbulent jets is still a challenge : only jets at Re D 10 5 can usually be considered using LES i.e. jets whose initial state should naturally be laminar 24 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

25 Tripped subsonic round jets Influence of the initial turbulence levels M = 0.9 Re D = 10 5 δ θ /r 0 = 1.8% σ ue = 0%, 3%, 6%, 9%, 12% n r n θ = n z = = 252 million pts as the exit turbulence level increases, coherent structures (and consequently vortex rolling-ups and pairings) gradually disappear higher initial turbulence levels lead to longer potential cores from L c = 9.3r 0 for σ ue = 0% to 17r 0 for σ ue = 12% 25 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

26 Tripped subsonic round jets Influence of the initial turbulence levels M = 0.9 Re D = 10 5 δ θ /r 0 = 1.8% 0.25 <u z u z > 1/2 /u j z/r 0 u z/u j along r = r 0 σ ue = 0% σ ue = 3% σ ue = 6% σ ue = 9% σ ue = 12% as the initial turbulence level increases, the shear layers develop more slowly with lower rms velocity peaks (from 22.6% to 14.5% of u j )... in particular for σ ue = 12%, flat profile of rms velocities 26 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

27 Tripped subsonic round jets Computing initially fully turbulent jets is still a challenge M = 0.9 Re D = 10 5 δ θ /r 0 = 1.8% Re δθ = 900 σ ue = 9% snapshots of vorticity norm ω and ω z component at x = r 0 Large scales, i.e. integral length scales L (θ) u i u i, must be well discretized mesh grid should be nearly isotropic near the nozzle exit r, r 0 θ and z < δ θ /2 seems recommended n r n θ n z = Bogey et al., Phys. Fluids (2010) 27 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

28 Underexpanded supersonic screeching jets Schlieren pictures NPR = 3.68 M j = 1.50 free jet boundary incident shock triple point Mach diamond p e > p e c Mach disk reflected shock slip line André et al., AIAA Journal (2011) 28 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

29 Underexpanded supersonic screeching jets Acoustic spectra NPR = 3.68 M j = 1.50 r = 53.2D p 10 db θ = 30 θ = 50 harmonics of screech tone DSP (db/st) θ = 70 θ = 90 θ = 110 St e = Tam s model for BBSAN u c (D e /u e ) L s (1 M c cos θ) 90 θ = θ = St 29 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

30 Underexpanded supersonic screeching jets Frequency of screech tones mode A 1 (symmetric) mode A 2 (symmetric) mode C (helical) St s 0.4 mode b (sinuous) mode B (sinuous) M j André et al., AIAA Paper Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

31 Underexpanded supersonic screeching jets Flight effects : spark Schlieren pictures of the jet plume M j = 1.5 M f = 0. M j = 1.5 M f = Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

32 DNC of a screeching plane jet Underexpanded supersonic screeching jets Computation of the generation of screech tones in a underexpanded plane jet p R /p = 2.48, D = 5.76 cm p e /p = 2.48, M j = 1.67 Westley & Wooley, Prog. Astro. Aero., 43, 1976 M j = 1.55 & Re h = p e /p = 2.09 Berland, Bogey & Bailly, Phys. Fluids, 19, Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

33 Concluding remarks Turbulent shear flows at high Reynolds number LES should display the same initial conditions as experiments Slower development in the laminar case but stronger turbulent transition ; jets with initially laminar BL are noisier : not the same physics To evaluate control efficiency, it seems interesting to find the quietest configuration by changing exit flow conditions in laboratory for instance, even by using unrealistic devices for an aircraft, but by considering relevant noise generation mechanisms.... Zaman (1985) «Turbulence and noise are suppressed at the most to the asymptotic levels which occur for the high-speed jets» Further efforts to understand the mixing in initially turbulent shear flows at high Reynolds number and consequences on acoustics are still required, even if the subject is undoubtedly difficult and requires high-fidelity large eddy simulations. 33 Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey

34 Concluding remarks Some open questions Research tools should not always induce research topics, but research topics should drive to the development and/or selection of suitable tools. Challenge : how to transpose high-order numerical algorithms in industrial contexts? LES is not so often. Evolution from RANS to LES is usually done for high Reynolds number flows - currently difficult. Complexity in the considered physics, and not only in pratical geometries Interaction of turbulence with shock-waves (BBSAN), time dependent inflow conditions, impedance in time domain with TBL, TBL noise, Académie des Sciences, 14 juin 2011 Christophe Bailly & Christophe Bogey