Direct Aeroacoustic Simulation of Feedback Mechanisms

Transcription

1 Direct Aeroacoustic Simulation of Feedback Mechanisms Hannes Frank, M. Werner, D. Flad and C.-D. Munz Institute for Aerodynamics and Gasdynamics, U.Stuttgart, Germany DLES10 Limassol Cyprus, May 28th 2015

2 Numerical analysis of tonal noise sources Tonal noise is perceived as particularly annoying Object of investigation: side view mirror Provide deeper understanding of source mechanisms Find possible means for tonal noise reduction/avoidance Experimental and numerical project at IAG with AUDI AG 2

3 Outline Motivation High order approach for LES and direct noise computation Tonal noise through acoustic feedback mechanisms Side view mirror Conclusions 3

4 Discontinuous Galerkin Spectral Element Method (DGSEM) Direct acoustics using the compressible Navier-Stokes equations Hybrid of inner-cell finite element and cross-cell finite volume approach Local high order polynomial representation gives arbitrary high order Favorable approximation properties: few points per wavelength required, low numerical dispersion and dissipation Unstructured meshes of hexahedra Low to medium Reynolds numbers ( ): high quality no-model LES 1 Efficient parallelization: communication only with direct neighbor cells 1 A.D. Beck et. al: High order discontinous Galerkin spectral element methods for transitional and turbulent flows (2014 I. J. Numerical methods in fluids) 4

5 Side view mirror: acoustic measurements Side view mirror NACA 0012 airfoil Ladder-type structures in frequency spectra Frequencies scale with Similar to acoustic feedback mechanism known from airfoils from: Plogmann et al. (2013) 2 2 B. Plogmann et al: Experimental investigations of a trailing edge noise feedback mechanism on a NACA 0012 airfoil (2013) 5

6 Acoustic feedback loop (5) a u c (1) (2) (3) d u inst (4) (1) Laminar flow separation near the TE (2) Amplification of TS-modes in wide frequency range within shear layer (3) The most amplified mode rolls up into vortices (4) Interaction of vortices with the TE leads to acoustic radiation at same frequency (5) Acoustic waves travel upstream and feed back into the boundary layer by receptivity (6) Initiation of TS-waves of same frequency Closed loop requires zero phase difference over one loop cycle 2 Discrete resonance frequencies 2 B. Plogmann et al.: Experimental investigations of a trailing edge noise feedback mechanism on a NACA 0012 airfoil (2013) 6

7 Large eddy simulation of mirror geometry Isolated mirror on wind tunnel floor, scheme, grid points ( per element) Nonconforming refinement (P4EST library, C. Burstedde 4 ) No-model LES using polynomial dealiasing Wall clock time on 3288 cores 4 C. Burstedde et al: p4est: Scalable algorithms for parallel adaptive mesh refinement on forests of octrees (2013 Siam J. Sci. Comput.) 7

8 Results Mean flow statistics averaged in y=-120 z=110 z=91 z=71 Wall friction Pressure coefficient 8

9 Comparison with PIV measurements Velocity magnitude at z=110mm Exp. Sim. Velocity magnitude RMS of velocity fluctuations 9

10 Instantaneous flow field Isosurfaces of, colored with velocity magitude 10

11 Hydrodynamic fluctuations Power spectra of pressure at the trailing edge, interval U S 15 5 top side side surface 11

12 Acoustic results Radiation pattern of spectral peaks: 2.9, 3.5, 4.4kHz U S U 12

13 Global stability analysis Disturbance calculation with localized impulse of linearly low amplitude 3 Time-averaged baseflow Initial disturbance Isosurfaces of, contours of 3 L.E Jones, R.D. Sandberg: Numerical analysis of tonal airfoil self-noise and acoustic feedback-loops (2011 J. Sound and Vibration) 13

14 Global stability analysis Wall pressure signal on the side surface Reoccuring wavepacket reveals presence of the acoustic feedback loop so F loop =566Hz 14

15 Global linear dynamic modes Linear global dynamic modes are approximated using dynamic mode decomposition (DMD) Snapshots with Flow field (disturbance calculation) is decomposed into linear modes Amplification spectrum 2834Hz 3679Hz 4471Hz 15

16 Global linear dynamic modes Frequencies and phase condition F [Hz] F loop [Hz] F/F loop Phase condition is fulfilled in good approximation for the selected modes. Spatial structure (isosurfaces of streamwise velocity perturbation) 2238 Hz 2834 Hz 3679 Hz 4471 Hz 16

17 Conclusions High order LES with DNC: Means to analyse aeroacoustic source mechanisms in detail Very good agreement with experiment in hydrodynamic data Tonal noise sources are localized in the same regions Feedback is recovered in the eigenmodes of the linear compressible system 17

18 Thank you for your attention! Thanks to my co-workers: M. Werner, D. Flad, C.-D. Munz and many more and the colleagues at the AUDI aerodynamics/aeroacoustics department Numerics research group DAGA 2015, Nürnberg, March 16 18

19 Direct noise computation with high order methods Solve the compressible Navier-Stokes equations for both the hydrodynamic and acoustic field High order numerical methods are mandatory for wave propagation problems favorable for time- and scale resolving CFD simulations (DNS and LES) Taylor Green vortex, Dr. Andrea Beck DNS 23

20 Acoustic feedback on the NACA0012 airfoil Validation case: Jones and Sandberg scheme, grid points (2D simulation) Acoustic Spectrum Main tone frequency, Jones & Sandberg f= L.E Jones, R.D. Sandberg: Numerical analysis of tonal airfoil self-noise and acoustic feedback-loops (2011 J. Sound and Vibration) 26

21 Boundary layer profiles Boundary layer profiles (top) and RMS fluctuations (bottom) at z=110 29

22 Acoustic results Wall-bounded acoustic array on wind tunnel floor, interval Tonal peaks from side surface: Sim: khz, Exp: 3.6 khz Ch. 32 side surface upper surface 30

23 Global linear dynamic modes Linear global dynamic modes are approximated using dynamic mode decomposition (DMD) Snapshots with, Nyquist frequency Flow field is decomposed into linear modes Tonal frequencies are identifed as unstable global modes! Amplification spectrum Amplitude spectrum 2.84kHz 3.67kHz 4.48kHz 32