A Computational Study of the Aerodynamics and Aeroacoustics of a Flatback Airfoil Using Hybrid RANS-LES

1. A Computational Study of the Aerodynamics andAeroacoustics of a Flatback Airfoil Using HybridRANS-LES Christopher Stone Computational Science and Engineering Matthew Barone Sandia National Laboratories C. Eric Lynch and Marilyn J. Smith Georgia Institute of Technology ASME Wind Energy Symposium Orlando, FL 5 January 2009

2. Outline • Motivation • Flatback airfoils • Goals of the present CFD analysis of flatback airfoils • Method • CFD codes and models • Aeroacoustic prediction methods • Results • Aerodynamic performance • Vortex-shedding noise • Wake visualization and interpretation • Conclusions 2009 ASME Wind Energy Symposium

3. Flatback Airfoils: Background • Flatback airfoil shapes have been proposed for the inboard region of wind turbine blades. • Thickness is added about a given camber line – different from “truncated” airfoil. • Benefits • Structural benefit of larger sectional area and larger moment of inertia for a given maximum thickness. • Aerodynamic benefits of larger sectional Clmax , larger lift curve slope, and reduced aerodynamic sensitivity to leading edge soiling. • Drawbacks • Increased drag due to separated base flow. • Introduction of an aerodynamic noise source due to trailing edge vortex shedding. References: C.P. van Dam et al., SAND 2008-2008, SAND 2008-1782, J. Solar Energy Eng., 128:422, 2006 . 2009 ASME Wind Energy Symposium

4. Overview of Available CFD Methods • The “industry-standard” method for high-Reynolds number CFD is Reynolds-Averaged Navier-Stokes (RANS) • Turbulent fluid flow is modeled using a turbulence model • Typically a steady-state solution is found • Necessary for affordable calculation of thin turbulent boundary layers • Drawbacks of RANS • Inaccurate for massively separated flow (flatback wake) • No prediction of unsteady turbulent fluctuations (flatback noise sources) • Large Eddy Simulation (LES) • Directly computes the large scale features of a turbulent flow • Models the small-scale, or “sub-grid” features of turbulence • Hybrid RANS/LES • Combines RANS in the near-wall turbulent boundary layer region with LES in regions of separated flow 2009 ASME Wind Energy Symposium

5. Goals and Approach • Goal: Assess the ability of hybrid RAN/LES methods to: • Predict flatback airfoil lift and drag • Simulate vortex-shedding in the wake for aeroacoustic predictions. • Approach • Use multiple CFD codes with hybrid RANS/LES capability • Use measured flatback airfoil shape from Virginia Tech wind tunnel experiment • Run simulations independently on different meshes and using different RANS/LES models (not a rigorous code-to-code comparison) • Compare two different methods for aeroacoustic predictions 2009 ASME Wind Energy Symposium

6. Flatback Airfoil Definition • TU-Delft DU97-W-300, 30% thick, 1.5% chord trailing edge thickness • DU97-flatback: 10% chord base thickness • DU97-flatback with splitter plate 2009 ASME Wind Energy Symposium

7. CFD Codes • Multi-block, structured grid finite volume code • Stable low-dissipation finite volume scheme • 2nd order implicit time advancement • Spalart-Allmaras RANS model • Detached Eddy Simulation hybrid model OVERFLOW SACCARA FUN3D • Overset grid finite difference code • 4th order finite difference scheme • 2nd order implicit time advancement • Spalart-Allmaras, SST RANS models • GT-HRLES hybrid model • Unstructured grid finite volume code • Node-centered finite volume scheme • 2nd order implicit time advancement • SST RANS model • GT-HRLES hybrid model 2009 ASME Wind Energy Symposium

8. Computational Meshes OVERFLOW • 3D Domain: Span-wise extent/number of grid points • OVERFLOW: 0.5c with 33 grid points • SACCARA: 0.2c, 0.4c, 0.8c with 64 grid cells • FUN3D: two-dimensional only • Periodic boundary conditions in spanwise direction SACCARA FUN3D 2009 ASME Wind Energy Symposium

9. Flatback Wake Visualizations 3D GT-HRLES 2D DES 2009 ASME Wind Energy Symposium

10. Aerodynamic Results • Chord Reynolds number = 3 million • Angle of attack = 4 and 10 degrees • Boundary layer transition at x/c = 0.05/0.10 on upper/lower surfaces 2009 ASME Wind Energy Symposium

11. Time-averaged DU97-flatback Lift • AOA = 4 deg., All simulations: 0.84 < CL < 0.91 • Experimental CL = 0.81 ± 0.04 • AOA = 10 deg., All simulations: 1.56 < CL < 1.66 • Experimental CL = 1.57 ± 0.07 • Decent agreement between simulation and experiment • Time-averaged lift relatively insensitive to code, grid, spanwise domain size, turbulence model 2009 ASME Wind Energy Symposium

12. Lift Fluctuation Spectrum • Experimental acoustic Strouhal number, : St = 0.24 ± 0.01 • All simulation results within the range 0.18 < St < 0.23 • 3D hybrid RANS/LES results within the range 0.18 < St < 0.20 • Overflow splitter plate simulation: Strouhal number shifted from 0.20 to 0.28 • Experimental acoustic Strouhal number shifted from 0.24 to 0.30 Overflow, AOA = 4 deg. 2009 ASME Wind Energy Symposium

13. Time-averaged DU97-flatback Drag • Experimental free transition CD = 0.06 ± 0.005 • Steady RANS: 0.034 < CD < 0.047 • Unsteady 2D RANS: 0.052 < CD < 0.144 • 3D Hybrid RANS/LES: 0.056 < CD < 0.110 • Results for drag are sensitive to: code, turbulence model, grid, spanwise domain size • Best agreement: FUN3D unsteady RANS (SST model) and Overflow GT-HRLES, CD = 0.055 – 0.056 2009 ASME Wind Energy Symposium

14. Drag from 3D Hybrid RANS/LES 0.4c/64 0.2c/64 0.4c/64 0.8c/64 0.5c/33 0.5c/33 2009 ASME Wind Energy Symposium

15. Effect of Splitter Plate (Overflow results) • Splitter plate lengthens wake recirculation zone and decreases intensity of the turbulent Reynolds stress • Drag is reduced by 18-27 %, compared to 45-50% reduction observed in experiment 2009 ASME Wind Energy Symposium

16. Aeroacoustic Results 2009 ASME Wind Energy Symposium

17. Aeroacoustic Prediction Methods 1. Approximate vortex-shedding theory assuming a compact line source imperfectly correlated along the span. • Computes far-field noise using integration of airfoil surface pressures Sectional lift coefficient spanwise correlation: Correlation length Centroid of correlation area L L 2. PSU WOP-WOP Aeroacoustic Prediction Tool (courtesy of James Erwin, Penn State) 2009 ASME Wind Energy Symposium

18. Peak Vortex-Shedding Noise Simulation results were for tripped b.l. 2D 2D 0.2c/64, 0.4c/64 0.4c/64 0.5c/33 0.8c/64 2009 ASME Wind Energy Symposium

19. Wake VisualizationSACCARA, AOA=4 deg. Z=0.4c, Nz=64 CD = 0.11, SPL = 98.6 Z=0.8c, Nz=64 CD = 0.084, SPL = 89.9 Z=0.8c, Nz=128 2009 ASME Wind Energy Symposium

20. Wake VisualizationOVERFLOW, AOA=10 deg. 2009 ASME Wind Energy Symposium

21. Conclusions • Time-averaged lift insensitive to code, grid, turbulence model, 2D/3D and agreement with experiment is decent • Strouhal number is relatively insensitive to the above parameters and is consistently lower than experimental values • Overflow simulations predicted qualitative effects of splitter plate: reduction in drag, reduction in noise, increase in shedding frequency • Time-averaged drag and noise is sensitive to the above parameters • 3D simulations exhibit 2 different vortex-shedding behaviors depending on spanwise domain size and grid resolution • well-defined 2D rollers with streamwise braid vortices : higher drag and louder tone • distorted 2D rollers breaking down into randomly oriented vortices : lower drag and quieter tone • More domain size and resolution studies are needed 2009 ASME Wind Energy Symposium