Transcription

1 Turbulence and Aeroacoustics Research Team of the Centre Acoustique Laboratoire des Fluides et d Acoustique UMR CNRS 5509, Ecole Centrale de Lyon MUSAF II Colloquium Toulouse, September 2013 Ö Ø ÓÑÔÙØ Ø ÓÒ Ó Ø ÒÓ Ù Ò Ä Ö Ý Ë ÑÙÐ Ø ÓÒ Christophe Bogey LMFA, UMR CNRS 5509 Ecole Centrale de Lyon

2 ÇÙØÐ Ò Introduction : experimental results and open questions on jet noise Computations of jets using Large-Eddy Simulation numerical methods jet inflow specification Effects of jet initial conditions effects of nozzle-exit turbulence level effects of Reynolds number Influence of jet temperature Concluding remarks ¾

3 Motivations ÖÓÙÒ investigation of real jets (full scale, complex nozzle geometry, heated, two streams, chevrons, installation effects...) providing reliable predictions and better understanding giving insight for flow control and noise reduction Reynolds numbers Re D = u j D/ν 10 7 difficult to reach in experiments and simulations studies at lower Re D (DNS 10 3, LES/exp ) and most are for unheated jets High-bypass-ratio nozzle (CFM56 type) (Loheac et al., SNECMA, 2004) QTD2 - Boeing - NASA AIAA Paper Castelain et al. AIAA J., 2008, 45(5)

4 Shear-layer visualizations at Re D from Castelain et al., ECL Re D = 870,000 «Ø Ó Ò Ø Ð ÓÒ Ø ÓÒ Re D = 33,000 Re D = 120,000 as the Reynolds number increases coherent structures are weaker / fine-scale turbulence is stronger the laminar-turbulent transition moves upstream, from the mixing layer to the nozzle boundary layer the nozzle-exit flow parameters (boundary-layer momentum thickness δ θ /r 0 and shape factor H, Reynolds number Re θ = u j δ θ /ν, peak turbulence level u e /u j) vary

5 «Ø Ó Ò Ø Ð ÓÒ Ø ÓÒ Effects of initial conditions on flow and sound fields cf experiments by Hussain & Zedan (1978), Gutmark & Ho (1983), Zaman (1985), Raman Ø Ðº (1989), Bridges & Hussain (1987),... p M e.g. sound spectra at 90 deg. for untripped/tripped jets with laminar exit conditions turbulent exit conditions from Zaman, AIAA J. (1985) St D several exit parameters vary and may be unknown in experiments what should be prescribed in simulations? what is the influence of each parameter? there is an ongoing discussion on that issue (cf Viswanathan and Clark (2004), Zaman (2012), Karon & Ahuja (2012) and Bogey Ø Ðº ( ))

6 «Ø Ó Ø ÑÔ Ö ØÙÖ Effects of temperature on flow and sound fields cf experiments by Witze (1974), Lau (1981), Lepicovsky (1999), Fisher Ø Ðº (1974), Tanna (1977), Bridges(05),... overall noise increase for M = u j /c a < 0.7, and reduction for M > 0.7 noise reduction depending on the emission angle φ noise intensity Ú M sound levels Ú φ for M=0.9 Overall intensity (db) OASPL (db) far-field data cold/isot hot from Tanna(77) u /c j a φ (deg.) and Bridges(05)

7 «Ø Ó Ø ÑÔ Ö ØÙÖ Effects of temperature on noise components decrease of sound levels for all freq. in the sideline direction growth of low-freq. components in the downstream direction, attributed to entropy sources (cf Morfey Ø Ðº ( ))... questioned by Viswanathan (2004) (contamination? Re D effects?) is this low-freq. amplification obtained in simulations? 1/3 Octave intensity (db) at φ = /3 Octave center frequency (khz) PSD (db) at φ = St D far-field pressure spectra for M = 0.9 cold/isot hot from Tanna (77) and Bridges(05)

8 ÇÙØÐ Ò Introduction : experimental results and open questions on jet noise Computations of jets using Large-Eddy Simulation numerical methods jet inflow specification Effects of jet initial conditions effects of nozzle-exit turbulence level effects of Reynolds number Influence of jet temperature Concluding remarks

9 Ø ÑÙÐ Ø ÓÒ ÈÖ ÒØ Numerical methods space derivatives : low-dispersion 11-point finite differences time integration : low-dispersion low-dissipation 6-stage Runge-Kutta 11-point selective filtering to relax subgrid energy radiation boundary conditions & outflow sponge zone Ref. : º ÓÑÔÙغ È Ý º, 2004, 194(1) - Ø Ù Ø, 2002, 88(4) - º ÓÑÔÙغ È Ý º, 2007, 224 È Ý º ÐÙ, 2006, (18)6 - º ÐÙ Å º, 2009, 627 LES based on relaxation filtering the flow variables are filtered explicitly after each time step to avoid pile-up of energy at smallest scales grid cut-off wave number (at 2 PPW) and filtering cut-off wave number (at 5 PPW) well separated energy drained by the filtering at smallest scales largest scales unaffected and damped by molecular viscosity

10 ÓÒ Ö Ð Ü Ø ÓÒ ÐØ Ö Ò Ä Ë In LES, largest scales must be Ú non-dimensional wave number well calculated, and mainly dissipated by viscosity (not by numeric/sgs models) to capture Reynolds number effects ÔÖ ÓÖ dissipation functions for viscosity : νk 2 = ν 2 (k )2 σ d for filtering : t D f (k ) with D f filter transfer fct π/32 π/16 π/8 π/4 π/2 π k viscosity filtering M = 0.9 and Re D = 1700, 2500 and 5000 vorticity and pressure Ref.: Bogey & Bailly, È Ý º ÐÙ¹, 2006, 18 - Ì, 2006, 20(1) ½¼

11 ÈÖ ÒØ Ø ÑÙÐ Ø ÓÒ LES parameters grid containing n r n θ n z = = 252 million points with r/r 0 = 0.35%, r θ/r 0 = 0.6%, z/r 0 = 0.7% at the nozzle lip 164,000 time steps the baseline LES of an isot jet at Re D = 10 5 shown to be accurate see Bogey Ø Ðº, ÈÓ, 2011, 23(3) the LES of the other jets with Re D 10 5 very likely reliable Far-field wave propagation to 60r 0 from the nozzle exit by solving the isentropic linearized Euler eq. from LES fields on a surface at r = 6.5r 0 n r n θ n z = = ½½

12 Â Ø Ò ÓÛ Ô Ø ÓÒ Boundary-layer tripping in a 2r 0 -long pipe Blasius laminar profile at the pipe inlet addition at z = r 0 of random vortical disturbances of magnitude chosen to provide the intended value of u e /u j laminar mean profile at exit (H 2.3) at z = 2r 0 at z = 0 <u z >/u j mean velocity Ú r/r 0 peak rms velocity Ú z/r 0 <u u > 1/2 /u z z j r/r z/r 0 ½¾

13 Â Ø Ò ÓÛ Ô Ø ÓÒ Spectra of velocity u z at r = r 0 and z = 0.4r 0 Ú axial and azimuthal wave numbers normalized by exit BL thickness δ Ú k z δ (using Taylor hyp.) Ú k θ δ/r PSD(u z ) PSD(u z ) k δ z k δ/r θ 0 baseline LES of an isot jet at Re D = 10 5 with u e /u j = 9% LES with twice the resolution DNS (turb. pipe flow, Ø Ðº Eggels (1994)) qualitative agreement with spectra in turbulent pipe flow and TBL (Tomkins & Adrian (PIV, 2005)) see Bogey Ø Ðº, ÈÓ, 2011, 23(9) ½

14 ÇÙØÐ Ò Introduction : experimental results and open questions on jet noise Computations of jets using Large-Eddy Simulation numerical methods jet inflow specification Effects of jet initial conditions effects of nozzle-exit turbulence level effects of Reynolds number Influence of jet temperature Concluding remarks ½

15 «Ø Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ Objective identify and distinguish between the effects of initial turbulence level and Reynolds number on laboratory-scale subsonic jets (Re D 10 5 ) LES of isothermal round jets at Mach M = 0.9 exiting from a pipe with similar Blasius BL velocity profiles of thickness δ 0 = 0.15r 0 (momentum thickness δ θ = 0.018r 0 ) and various peak turbulence intensities u e /u j 4 jets with Re D = 10 5 and u e /u j = 3%, 6%, 9% or 12% (δ θ = 0.018r 0 yielding Re θ 900) see in º ÐÙ Å º, 2012, jets with u e /u j = 9% and Re D = , , 10 5 or (and Re θ = 251, 486, 943 or 1856) see in ÈÓ, 2012, 24(10) ½

16 «Ø Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ Vorticity norm in the shear layers up to z = 3.75r 0 jets with Re D = 10 5 and u e /u j = 3% u e /u j = 6% u e /u j = 9% u e /u j = 12% jets with u e /u j = 9% and Re D = Re D = Re D = 10 5 Re D = large-scale structures observed at u e /u j = 3% and Re D = but stronger fine-scale turbulence with increasing u e /u j or Re D ½

17 Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ «Ø Momentum thickness u e /u j = 3% ր 12% Re D = ր δ θ /r ց u e ր δ θ /r ց Reր z/r z/r 0 for higher u e /u j or Re D, the shear layers develop more slowly with lower spreading rates ½

18 Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ «Ø Peak axial turbulence intensities u e /u j = 3% ր 12% Re D = ր <u z u z > 1/2 /u j u e ր <u z u z > 1/2 /u j Reր z/r z/r 0 lower rms velocities for higher u e /u j or Re D overshoot at u e /u j = 3% and Re D = but nearly monotonical trend at u e /u j = 12% and Re D = ½

19 Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ «Ø Fluctuating pressure u e /u j = 3% u e /u j = 6% u e /u j = 9% u e /u j = 12% Re D = Re D = Re D = 10 5 Re D = ½

20 Ó Ø Ø Ò Ø Ð ÓÒ Ø ÓÒ ÁÒ Ù Ò Sound levels at 60 radii from the nozzle exit u e /u j = 3% ր 12% Re D = ր OASPL (db) u e ր OASPL (db) Reր φ (deg.) φ (deg.), +, measurements at Re D as u e /u j or Re D increase, OASPL decrease and become closer to those for high-re D not very likely to generate vortex-pairing noise ¾¼

21 Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ «Ø Sound spectra at 60 deg. u e /u j = 3% ր 12% Re D = ր SPL (db/st) u e ր SPL (db/st) Reր St D measurements at Re D St D St θ = (half of ML dominant freq.) as u e /u j or Re D increase, the extra hump wrt high-re D measurements is weaker, as large-scale structures no longer dominate the shear layers ¾½

22 ÇÙØÐ Ò Introduction : experimental results and open questions on jet noise Computations of jets using Large-Eddy Simulation numerical methods jet inflow specification Effects of jet initial conditions effects of nozzle-exit turbulence level effects of Reynolds number Influence of jet temperature Concluding remarks ¾¾

23 «Ø Ó Ø Ø ÑÔ Ö ØÙÖ Objective to identify and distinguish btw temperature and Reynolds number effects on heated laboratory-scale subsonic jets see AIAA Q : is the downstream low-freq. noise amplification due to Re D? LES of round jets at Mach M = 0.9 exiting from a pipe with similar Blasius BL velocity profiles of thickness δ 0 = 0.15r 0 (momentum thickness δ θ = 0.018r 0 ) and peak turbulence intensities u e /u j one isot jet at Re D = 10 5 two hot jets at T j = 1.5T a and at T j = 2.25T a with the same diameter as the isot jet, yielding Re D = and Re D = due to the variations of viscosity with T j one hot jet at T j = 1.5T a with the same Re D = 10 5 as the isot jet ¾

24 Ó Ø Ø ÑÔ Ö ØÙÖ «Ø Shear-layer development momentum thickness δ θ /r 0 rms axial velocity [u z ] rms/u j δ θ /r <u z u z > 1/2 /u j z/r 0 z/r 0 isot, T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case with heating, the shear layers develop more rapidly with higher turbulence intensities for the jets at identical D, but shows much less change for a constant Re D... strong Reynolds nb effects ¾

25 «Ø Ó Ø Ø ÑÔ Ö ØÙÖ Vorticity norm in the jets and pressure outside with heating : emission of additionnal sound waves in the mixing layers for a constant D lower noise in the upstream direction at a fixed Re D ¾

26 Ó Ø Ø ÑÔ Ö ØÙÖ «Ø Sound levels at d = 60r 0 emission angle φ Ú OASPL (db) OASPL φ (deg.) OASPL (db) comp. with exp. data φ (deg.) isot, T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case exp. data for M = 0.9 jets with D = 5.1cm from Tanna and Bridges : isot/ T j = 2.3T a at d = 72D, cold/ T j = 1.43T a at d = 40D with heating, more noise for the same D, less noise for equal Re D... in the latter, resemblance to exp. data for high Re D ¾

27 «Ø Ó Ø Ø ÑÔ Ö ØÙÖ Difference in sound spectra wrt the isot case Ú St D = fd/u j 6 at φ = 30 4 SPL (db/st) St D T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case emergence of a low-freq. component independently of Re D and reduction of high-freq. noise, in agreement with exp. of Tanna (1977)... low-freq. amplification due to entropy noise sources? ¾

28 «Ø Ó Ø Ø ÑÔ Ö ØÙÖ Difference in sound spectra wrt the isot case Ú St D = fd/u j 6 at φ = 60 4 SPL (db/st) St D T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case 1/8, 1/4 and 1/2 of the freq. initially dominating in the mixing layers extra noise components as Re D decreases... generation of vortex-pairing noise? ¾

29 «Ø Ó Ø Ø ÑÔ Ö ØÙÖ Difference in sound spectra wrt the isot case Ú St D = fd/u j 6 at φ = 90 4 SPL (db/st) St D T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case exp. data for M = 0.9 jets with D = 5.1cm: Ç btw cold and T j = 1.43T a (Bridges) for the jet at Re D = 10 5, noise reduction for nearly all freq. in line with exp. of Tanna (1977) and Bridges (2005) ¾

30 Ö Ñ Ö ÓÒÐÙ Ò Large-Eddy Simulations of jets even still expensive, they can now allow us to carefully investigate problems encountered for laboratory-scale jets... and to complement and clarify experimental results e.g. regarding effects difficult to distinguish, which can mutually amplify or oppose one another effects of Reynolds number (Re D /Re θ ) and exit turbulence levels effects of temperature and Reynolds number They are mature enough to provide a better understanding of noise generation mechanisms be applied to more complex configurations ¼

31 ËÙ ÓÒ ØÙÖ ÙÐ ÒØ Ø Shear-layer visualizations at different Re D Re D = 870,000 Re D = 33,000 Re D = 120,000 from Castelain et al., ECL the nozzle-exit parameters vary with Re D, including the boundary-layer momentum thickness δ θ /r 0 ( 0.1 1%) and its corresponding Reynolds number Re θ = u j δ θ /ν the peak turbulence level u e /u j ( 0 10%) the shape factor H = δ /δ θ of the mean velocity profile (H 2.5 : Blasius laminar profile - H 1.4 turbulent profile) ½

32 ÓÒ Ø ÓÒ Ò Ù ÓÒ Ø ÆÓÞÞÐ ¹ Ü Ø Initial flow state at the nozzle exit peak exit rms velocity u e /u j for jets with Re D between 50,000 and 300, Re D 10 3 from Zaman, AIAA J. (1985) for Re D 100,000 : u e /u j < 1% and H 2.5 º the jets are initially fully laminar for 100,000 º Re D º 500,000 : 1% u e /u j 10% the jets are initially transitional for Re D 500,000 : u e /u j 10% and H 1.4 ² the jets are initially fully turbulent ¾

33 «Ø Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ Nozzle-exit flow profiles for the jets with Re D = 10 5 and u e /u j ր <u z >/u j mean axial velocity <u z >/u j <u z u z > 1/2 /u j rms axial velocity [u z ] rms/u j r/r r/r 0 Jet0% Jet3% Jet6% Jet9% Jet12% Zaman s data for a Re D = 10 5 tripped jet as desired, laminar velocity profiles (H 2.3), momentum thickness δ θ 0.018r 0, and rms velocity peaks u e /u j = 3,6,9,12% δ θ yielding momentum Reynolds numbers Re θ 900

34 «Ø Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ Nozzle-exit flow profiles for the jets with u e /u j = 9% and Re D ր mean axial velocity <u z >/u j rms axial velocity [u z ] rms/u j 0.12 <u z >/u j <u z u z > 1/2 /u j r/r r/r 0 JetRe25e3 JetRe50e3 JetRe100e3 JetRe200e3 Zaman s data for a Re D = 10 5 tripped jet as desired, laminar velocity profiles (H 2.3), momentum thickness δ θ 0.018r 0, and rms velocity peaks u e /u j 9% δ θ providing momentum Reynolds numbers Re θ = 251, 477, 925 and 1830

35 Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ «Ø Centerline mean axial velocity u e /u j = 3% ր 12% Re D = ր u c /u j ր u e ր u c /u j ր Reր z/r z/r 0 o,, measurements for Mach 0.9 jets at Re D with rising u e /u j or Re D, the jet spreads farther downstream... fair agreement with high Re D data

36 Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ «Ø Centerline axial turbulence intensities u e /u j = 3% ր 12% Re D = ր <u z u z > 1/2 /u j ց u e ր <u z u z > 1/2 /u j ց Reր z/r z/r 0 o,, measurements for Mach 0.9 jets at Re D as u e /u j or Re D increase, rms velocity peaks are reached farther downstream but do not differ much

37 «Ø Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ pressure fields for the jets with u e /u j = 9% and Re D ր strong waves emitted between z = 2r 0 and z = 5r 0 at lower Re D, attenuated at higher Re D

38 Ó Ø Ø ÑÔ Ö ØÙÖ «Ø Nozzle-exit profiles mean axial velocity <u z >/u j rms axial velocity [u z ] rms/u j <u z >/u j <u z u z > 1/2 /u j r/r 0 r/r 0 isot, T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case exp. data for a Re D = 10 5 tripped jet as desired, (laminar) Blasius velocity profiles (shape factor H 2.3), momentum thickness δ θ 0.019r 0, and rms velocity peak u e /u j 9%... δ θ providing Re θ = 943, 485, 254 and 941

39 «Ø Ó Ø Ø ÑÔ Ö ØÙÖ Vorticity norm in the shear layers up to z = 3r 0 stronger large-scale structures in the hot jets with equal diameter and decreasing Re D similar vorticity fields for a constant Re D

40 Ó Ø Ø ÑÔ Ö ØÙÖ «Ø Centerline mean axial velocity u c /u j variations with z/r 0 with (z z c )/r u c /u j u c /u j z/r 0 (z z c )/r 0 isot, T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case exp. data at M = 0.9 and Re D : Ç,, T j T a, T j = 1.76T a in all cases, the potential core shortens and the velocity decay is faster with heating, in good agreement with exp. data... temperature effects are predominant ¼

41 Ó Ø Ø ÑÔ Ö ØÙÖ «Ø Centerline turbulence intensities rms axial velocity [u z ] rms/u j 0.16 <u z u z > 1/2 /u j z/r 0 isot, T j = 1.5T a, T j = 2.25T a for the same D (and Re D ց) T j = 1.5T a for the same Re D as the isot case exp. data at M = 0.9 and Re D :, T j T a with rising temperature, the peak turbulence intensities increase, in agreement with exp. findings of Bridges (2006) ½

42 Ó Ø ÒÓ ÖÖ ÓÙØ Ò ¾¼¼¼ Ä Ë Simulations without nozzle of subsonic jets at Mach M = 0.9 : feasibility (JEAN) Ì, 2003, 16(4)- ², 2006, 35(10) influence of forcing Á Â, 2005, 43(5) influence of subgrid-scale model (Smagorinsky vs filtering) Á Â, 2005, 43(2) - ÈÓ, 2006, 18(6) - ÁÂÀÀ, 2006, 27 analysis of two noise components for Reynolds Re D 10,000 Simulations with pipe nozzle preliminary simulations : Ì, 2006, 20(10) - Â Å, 2007, 583 single-stream subsonic jet ÁÂ, 2008, 7(1) supersonic screeching jet ÈÓ, 2007, 19(7) dual-stream jet (CoJeN) ÈÓ, 2009, 21(3) ¾

43 Ó Ø ÒÓ ÖÖ ÓÙØ Ò ¾¼¼¼ Ä Ë recent simulations with pipe nozzle : jet at M = 3.3 and Re D = 100,000 Á Â, 2011, 49(10) initially laminar jets at M = 0.9 and Re D = 100,000 initially turbulent jets at M = 0.9 and Re D 100,000 grid convergence and investigation of exit turbulence  Å, 2011, 23(3) ÈÓ, 2011, 23(3) - ÈÓ, 2011, 23(9) effets of initial turbulence  Å, 2012, 701 effets of Reynolds number ÈÓ, 2012, 24(10) effets of exit-boundary-layer thickness ÈÓ, 2013, 25(05) ongoing work : heated jets at M = 0.9 and Re D 100,000 AIAA Ô Ô Ö Ú Ð Ð Ø ØØÔ»» ÓÙ Ø ÕÙ º ¹ÐÝÓÒº Ö» ÔÙ Ð ÖºÔ Ô

44 ÓÙØ Ö Ö ÓÐÙØ ÓÒ Ò Ø Ä Ë Highest resolutions in recent LES of subsonic single jets jet conditions Re D Re θ δ BL /r 0 BL trip u e /u j initial state Bogey & Bailly (JFM, 2010) no < 1% fully laminar Bogey Ø Ðº (PoF, 2011) yes 9% nominally turbulent Shur Ø Ðº (JSV, 2011) DES na na grid resolutions wrt δ BL n r n θ n z δ/ r min δ/(r 0 θ) δ/ z e remarks Bogey similar results (2010) when r & z/2 Bogey very close mixing layer solutions (2011) when δ/ r, δ/(r 0 θ) & δ/ z 2 Shur (2011) in Bogey (2011) : integral length scales shown to be well discretized and affected by molecular viscosity rather than by the sgs model