Experimental Validation of TNO Trailing Edge Noise Model and Application to Airfoil Optimization

Experimental Validation ofTNO Trailing Edge Noise Modeland Application to Airfoil Optimization

Franck Bertagnolio, Helge Aa. Madsen, and Christian Bak

Aero-Elastic Design, Wind Energy DivisionRisø DTU, National Laboratory for Sustainable EnergyRoskilde, Denmark

EWEC 2009, Marseille, 16-19 March 20092 Risø-DTU, Technical University of Denmark

Outline

• Trailing Edge Noise

• TNO Trailing Edge Noise Model

• Validation against Measurements

• Aeroacoustic Optimizationof a Wind Turbine Airfoil

EWEC 2009, Marseille, 16-19 March 20093 Risø-DTU, Technical University of Denmark

Trailing Edge Noise

2y3y

1y

1 2( )U y

( , )P k

1 2 3, ,u u u Far field sound ( )PS

EWEC 2009, Marseille, 16-19 March 20094 Risø-DTU, Technical University of Denmark

TNO Trailing Edge Noise Model

2

22 21 10 2 2 2 22 1 22 2 0

1 3 2

( , ) 4 ( ) ( , ) e dkym c

k UP L y u U k y

k k y

k k

Parchen (1998) combines a diffraction problem solution with knowledge of the turbulent fluctuations in the boundary layer

• Airfoil Surface Pressure Spectrum (Blake,1986) Lighthill analogy in spectral domain

Solution for the Mean shear-Turbulence interaction:

• Far Field Noise (Ffwocs Williams and Hall, 1970 ; Chandiramani, 1974; Chase, 1975; Howe, 1978; Brooks and Hodgson, 1981)

1 120 1

( ) ( , )d4span

P

LS P k k

R c k

EWEC 2009, Marseille, 16-19 March 20095 Risø-DTU, Technical University of Denmark

Model Practical Implementation

Input data originates either from XFOIL or EllipSys2D

Boundary layer quantities required as input are:

• Directly obtained from the codes (U1 (y2), Shear, kt (CFD), …)

• Classical turbulence theory results for 22 (Isotropic

turbulence spectrum, Von Karman), m (Gaussian), …

• Integral length scale: (Lutz et al, 2007)

3/ 2t

2 2( ) 0.387k

L y

EWEC 2009, Marseille, 16-19 March 20096 Risø-DTU, Technical University of Denmark

Validation: LM Glasfiber Wind Tunnel

Aerodynamic Test Facility

NACA0015 Airfoil Section

Surface Pressure Measurement Holes

EWEC 2009, Marseille, 16-19 March 20097 Risø-DTU, Technical University of Denmark

NACA0015: Surface Pressure Spectrum

Re=1.6x10^6 - No Turbulence Grid - x/C = 0.567=0o =4o

=8o =12o

EWEC 2009, Marseille, 16-19 March 20098 Risø-DTU, Technical University of Denmark

Validation: NACA0012 [Brooks and Hodgson]Anaechoic Wind Tunnel Facility at NASA Langley (1981)Re=1.6M, 2.9M – Aoa=0o

Surface Pressure

Far Field SPL

EWEC 2009, Marseille, 16-19 March 20099 Risø-DTU, Technical University of Denmark

Discussion on Validation

• Quantitative model results might be erroneous (Difference in conventions?? )

• but it may be that TNO model fails to accurately predict measurements(However, no fundamental differences between experiments NACA0015-LM / NACA0012-NASA…??)

• TNO model correctly captures tendencies observed

in measurements

• TNO model (using XFOIL) is a good candidate for aeroacoustic optimization

EWEC 2009, Marseille, 16-19 March 200910 Risø-DTU, Technical University of Denmark

Airfoil Optimization

• Goal: Reduce trailing edge noise

• AirfoilOpt: SIMPLEX algorithm (gradient based method) Cost function minimization subject to non-linear constraints Both cost function and constraints can involve:

1) Aerodynamic characteristics (XFOIL)2) Geometric characteristics

TNO model implemented in the code

• Noise Optimization Procedure: Maximum SPL from TNO model used as cost function Various constraints to preserve aerodynamic and geometrical

characteristics of original airfoil (PARAMETER STUDY, see paper)

EWEC 2009, Marseille, 16-19 March 200911 Risø-DTU, Technical University of Denmark

Relaxing Geometric Constraints

Initial airfoil: RISØ-B1-18Constraints relaxation: - Preserve all constraints - ymin and ymax along chord

- y,xx (Airfoil curvature)

EWEC 2009, Marseille, 16-19 March 200912 Risø-DTU, Technical University of Denmark

Relaxing Geometric Constraints

Far Field SPL A-Weighted SPL

EWEC 2009, Marseille, 16-19 March 200913 Risø-DTU, Technical University of Denmark

Mechanism Behind SPL Reduction Boundary Layer Development along Chord

Turbulent Kinetic Energy Profile

Near TrailingEdge

Original

Optimized

Original Optimized

EWEC 2009, Marseille, 16-19 March 200914 Risø-DTU, Technical University of Denmark

Conclusions

• TNO model validation gave mixed results (Convention problem / Model error ???)

• Airfoil noise optimization: SPL ~ -1 to 2dB

• Noise reduction reached through:reducing camber flattening of suction side

Reduction of TKE at trailing edge