Professional Documents
Culture Documents
A new correlation-based transition model has been developed, which is built strictly on
local variables. As a result, the transition model is compatible with modern CFD techniques
such as unstructured grids and massively parallel execution. The model is based on two
transport equations, one for intermittency and one for a transition onset criterion in terms
of momentum thickness Reynolds number. The proposed transport equations do not
attempt to model the physics of the transition process (unlike e.g. turbulence models), but
form a framework for the implementation of transition correlations into general-purpose
CFD methods. The transition model was initially developed for turbomachinery and
aeronautical flows. The main goal of the present paper is to validate the model for predicting
transition on wind turbines. In this paper, fully turbulent and transitional computations of
the 2D S809 airfoil along with a full 3D wind turbine rotor (that uses the S809 airfoil) have
been accomplished. The transitional results are in good agreement with the experimental
data and the transition model would appear to be well suited for the prediction of wind
turbine aerodynamics.
Nomenclature
*
CFD Development, Staudenfeldweg 12, 83624 Otterfing, Germany.
†
Intern, Staudenfeldweg 12, 83624 Otterfing, Germany.
‡
Head of CFD Development, Staudenfeldweg 12, 83624 Otterfing, Germany.
1
American Institute of Aeronautics and Astronautics
Copyright © 2006 by ANSYS Inc. . Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
U = local velocity
Uo = local freestream velocity
Uref = inlet reference velocity
u’ = local fluctuating streamwise velocity
x/C = axial distance over axial chord
y = distance to nearest wall
y+ = distance in wall coordinates, ρyµτ/µ
δ = boundary layer thickness
θ = momentum thickness
λθ = pressure gradient parameter, (ρθ2/µ)(dU/ds)
µ = molecular viscosity
µt = eddy viscosity
ρ = density
Ω = absolute value of vorticity, (2ΩijΩij)1/2
Ωij = vorticity tensor, 0.5(∂ui/∂xj - ∂uj/∂xi)
ω = specific turbulence dissipation rate
Subscripts
t = transition onset
s = streamline
I. Introduction
W ind energy is rapidly becoming an economically viable energy source. This is partially due to increases in the
cost of oil, but also due to significant improvements in the performance of modern wind turbines. Like many
aerodynamic devices such as turbine blades and wings, a better understanding of the flow field can result in design
changes that can significantly improve performance. As well, it is critical that the aerodynamic characteristics of the
wind turbine be known during the design phase in order to have accurate economic projections. Because of the
costs associated with performing wind tunnel experiments, there is a significant amount of interest in predicting the
aerodynamic characteristics of a wind turbine using computational fluid dynamics (CFD). However, a survey of the
available literature indicates that there are still some significant issues associated with accurately predicting wind
turbine performance using CFD. According to Wolfe and Ochs (1997) there are two main issues. The first is that
the standard k-ε turbulence model (which is the default in many commercial CFD codes) under predicts the amount
of separation on an airfoil at high angles of attack. This can lead to overly optimistic performance projections. The
second problem is that very few CFD codes attempt to model laminar-turbulent transition, which can also have a
significant effect on the predicted performance. There are a few CFD codes that have been modified to predict
transition using either an empirical correlation or the eN method. However, these approaches are based on non-local
operations such as integrating the boundary layer quantities or tracking the amplification of instabilities along a
streamline. As a result, they are only applicable to structured grids, can only predict transition in 2D flows and often
cannot be parallelized in order to run large cases (e.g. a full 3D wind turbine with 10 million grid nodes).
Recently, a new correlation-based transition model has been developed2, which is based strictly on local variables.
As a result, the transition model is compatible with modern CFD approaches such as unstructured grids, massive
parallel execution and can also predict transition in 3D flows. The model is based on two transport equations, one
for intermittency and one for the transition onset criteria in terms of momentum thickness Reynolds number. The
proposed transport equations do not attempt to model the physics of the transition process (unlike e.g. turbulence
models), but form a framework for the implementation of correlation-based models into general-purpose CFD
methods. One of the central variables in the formulation of model is the strain-rate (or sometimes vorticity3)
Reynolds number:
2
American Institute of Aeronautics and Astronautics
ρy 2 ∂u ρy 2
Rev = = S (1)
µ ∂y µ
where y is the wall distance, ρ is the density, µ is the molecular viscosity and S is the absolute value of the strain
rate. The importance of ReV lies in the relation of its maximum value inside the boundary layer to the momentum
thickness Reynolds number ReΘ, of Blasius (or more generally Falkner-Skan) profiles4:
where ~ y is the location where ReV has its maximum. The function Reν can be used on physical reasoning, by
arguing that the combination of y2S is responsible for the growth of disturbances inside the boundary layer, whereas
ν = µ / ρ is responsible for their damping. As y2S grows with the thickness of the boundary layer and ν stays
constant, transition will take place once a critical value of Reν is reached. The connection between the growth of
disturbances and the function Reν was shown by Van Driest and Blumer (1963) in comparison with experimental
data. The transition model was first detailed in Menter et al (2004) and has been extensively validated for predicting
transition in both turbomachinery6 and aeronautical flows2 under the influence of freestream turbulence, pressure
gradients and separation. It has been combined with the SST turbulence model7, which has been shown to be very
accurate for predicting turbulent separation on airfoils.
The first part of the paper will compute the 2D aerodynamic coefficients of the well-known S809 wind turbine
airfoil1 using the SST turbulence model along with the new transition model. The numerical results for this airfoil
will be compared to the experimental data obtained by Somers (1989,1997). The second part of the paper will detail
the numerical results obtained for the NREL Phase VI test case9, which was a 3D wind turbine that used the S809
airfoil for the blade profile. It will be shown that the transition model results in a significant improvement of the
predicted power output, particularly at the higher angles of attack. The differences between the 2D and 3D airfoil
characteristics will also be investigated, particularly at the higher angles of attack were there is a significant amount
of separated flow and the 3D effects are substantial.
The present transition model formulation is described very briefly for completeness, a detailed description of the
model can be found in Menter et al. (2004) along with some improvements for Natural and separation induced
transition which are detailed in Langtry and Menter (2005). The present transition model is summarized in the
following pages.
∂ (ργ ) ∂ (ρU j γ ) ∂ µ ∂γ
+ = Pγ 1 − Eγ 1 + Pγ 2 − Eγ 2 + µ + t (3)
∂t ∂x j ∂x j σ γ ∂x j
Pγ 1 = Flength c a1 ρ S [γFonset ] α ;
c
Eγ 1 = ce1 Pγ 1γ (4)
where S is the strain rate magnitude. Flength is an empirical correlation that controls the length of the transition
region. The destruction/relaminarization sources are defined as follows:
3
American Institute of Aeronautics and Astronautics
Pγ 2 = c a 2 ρ Ω γ Fturb ; E γ 2 = c e 2 Pγ 2 γ (5)
where Ω is the vorticity magnitude. The transition onset is controlled by the following functions:
ρy 2 S ρk
ReV = ; RT = (6)
µ µω
Fonset 1 =
Re v
2 . 193 ⋅ Re θ c
; ( (
Fonset 2 = min max Fonset 1 , Fonset 1 , 2 . 0
4
) ) (7)
4
R 3 R
− T
Fonset 3 = max 1 − T ,0 ; Fonset = max (Fonset 2 − Fonset 3 ,0 ); Fturb = e 4 (8)
2.5
Reθc is the critical Reynolds number where the intermittency first starts to increase in the boundary layer. This
~
occurs upstream of the transition Reynolds number, R eθt , and the difference between the two must be obtained
~
from an empirical correlation. Both the Flength and Reθc correlations are functions of R eθt .
The boundary condition for γ at a wall is zero normal flux while for an inlet γ is equal to 1.0.
~
The transport equation for the transition momentum thickness Reynolds number, R eθt , reads:
~
~
(
∂ ρ R e θt ) (
∂ ρ U j R e θt ∂ ) ~
∂ R e θt
+ = Pθ t + σ θ t (µ + µ t ) (11)
∂t ∂x j ∂ x j ∂ x j
ρ ~ 500 µ
Pθ t = c θ t
t
(Re θt )
− R e θ t (1 . 0 − F θ t ); t =
ρU 2
(12)
y
4
γ − 1 / ce 2
2
−
Fθt = min max Fwake ⋅ e δ
,1.0 − ,1.0 (13)
1.0 − 1 / ce 2
~
R eθ t µ 15 50 Ω y
θ BL = ; δ BL = θ BL ; δ = ⋅ δ BL (14)
ρU 2 U
4
American Institute of Aeronautics and Astronautics
2
Re
ρωy 2 − ω
Reω = ; Fwake = e 1E +5 (15)
µ
~
The model constants for the R eθt equation are:
~ ~
The boundary condition for R eθt at a wall is zero flux. The boundary condition for R eθt at an inlet should be
calculated from the empirical correlation based on the inlet turbulence intensity.
The model contains three empirical correlations. ReΘt is the transition onset as observed in experiments and it is used
in Eq.12. Flength is the length of the transition zone and goes into Eq. 4. ReΘc is the point where the model is activated
in order to match both, ReΘt and Flength, and it goes into Eq. 7. At present these empirical correlations are
proprietary and are not given in the paper.
~ ~
( )
ReΘt = f (Tu, λ ); Flength = f ReΘt ; ReΘc = f ReΘt ( ) (17)
The first empirical correlation is a function of the local turbulence intensity, Tu, and the Thwaites’ pressure gradient
coefficient λθ defined as:
λθ = (θ2/ν)dU/ds (18)
The transition model interacts with the SST turbulence model7, as follows:
∂ ∂ ∂
( ρk ) +
~ ~
( ρu j k ) = Pk − Dk + (µ + σ k µ t ) ∂k (19)
∂t ∂x j ∂x j ∂x j
~ ~
Pk = γ eff Pk ; Dk = min(max(γ eff ,0.1),1.0)Dk (20)
8
Ry
−
ρy k 120
Ry = ; F3 = e ; F1 = max (F1orig , F3 ) (21)
µ
where Pk and Dk are the original production and destruction terms for the SST model and F1orig is the original SST
blending function. Note that the production term in the ω-equation is not modified.
In order to capture the laminar and transitional boundary layers correctly, the grid must have a y+ of approximately
one. If the y+ is too large (i.e. > 5) than the transition onset location moves upstream with increasing y+. All
simulations have been performed with the CFX CFD code using a bounded 2nd order upwind biased discretisation
for the mean flow, turbulence and transition equations. For the case where system rotation is present (i.e. a 3D wind
turbine rotor) a moving frame approach is used to solve the equations. The equations where solved in steady-state
mode and the forces (i.e. lift and drag) where monitored in order to judge convergence. A relatively large time step
of 1/5 the airfoil characteristic time (i.e. chord/relative freestream velocity/5) was specified and this was possible
due to the implicit nature of the CFX code. This time step was found to be small enough to give relatively good
convergence for the equation residuals while being large enough to damp out any unsteadiness, which would have
been significant in the high angle of attack cases where a large amount of separated flow was present
5
American Institute of Aeronautics and Astronautics
III. Results and Discussion
The results obtained for the lift and drag polars are Figure 5 Pressure Distribution (Cp) for AoA = 20°
shown in Figures 7 and 8. Between 0° and 9° the lift
coefficients (Cl) predicted by the transitional CFD
results are in very good agreement with the
experiment while both the fully turbulent CFD and X-
Foil results appear to under-predict the lift curve by
approximately 0.1. As well, between 0° and 9° the
drag coefficient (Cd) predicted by the transitional
CFD and X-Foil results are in very good agreement
with the experiment while the fully turbulent CFD
over-predicts the drag. For angles of attack greater
than 9° all of the numerical results significantly over-
predict the lift coefficient and under-predict the drag
coefficient. One potential explanation for the poor
numerical results could be the presence of a
significant amount of 3D flow in the separated regions
of the wind tunnel model, which the 2D numerical
results obviously cannot resolve. As well, it is well
known that the RANS turbulence models typically fail
in massively separated regions and it is possible that a
hybrid RANS-LES model such as DES or SAS may Figure 6 Transition location (xt/c) vs angle of attack
be needed to capture the separated flow physics for the S809 airfoil
correctly11.
7
American Institute of Aeronautics and Astronautics
Figure 7 Lift Coefficient (Cl) Polar (top) and enlarged Figure 8 Drag Coefficient (Cd) Polar (top) and
view from 0° to 10° (bottom) enlarged view from 0° to 10° (bottom)
8
American Institute of Aeronautics and Astronautics
(i.e. increasing angle of attack) and each run was
computed overnight on a 16 CPU Linux cluster.
IV. Conclusion
The model development and validation at ANSYS CFX was funded by GE Aircraft Engines and GE Global
Research. Prof. G. Huang and Dr. B. Suzen from the University of Kentucky have supported the original model
development with their extensive know-how and their in-house codes.
References
1
Wolfe, W.P. and Ochs, S.S. “CFD Calculations of S809 Aerodynamic Characteristics”, AIAA Paper AIAA-97-0972, 1997.
2
Langtry, R.B., and Menter, F.R., “Transition Modeling for General CFD Applications in Aeronautics”, AIAA Paper 2005-
522, 2005.
3
Van Driest, E.R. and Blumer, C.B., “Boundary Layer Transition: Freestream Turbulence and Pressure Gradient Effects,”
AIAA Journal, Vol. 1, No. 6, June 1963, pp. 1303-1306.
4
Menter, F. R., Esch, T. and Kubacki, S., “Transition Modelling Based on Local Variables”, 5th International Symposium on
Turbulence Modeling and Measurements, Spain, 2002.
5
Menter, F.R., Langtry, R.B., Likki, S.R., Suzen, Y.B., Huang, P.G., and Völker, S., "A Correlation based Transition Model
using Local Variables Part 1- Model Formulation”, ASME-GT2004-53452, ASME TURBO EXPO 2004, Vienna, Austria.
6
Langtry, R.B., Menter, F.R., Likki, S.R., Suzen, Y.B., Huang, P.G., and Völker, S., "A Correlation based Transition Model
using Local Variables Part 2 – Test Cases and Industrial Applications”, ASME-GT2004-53454, ASME TURBO EXPO 2004,
Vienna, Austria.
7
Menter, F.R., Two-equation eddy-viscosity turbulence models for engineering applications, AIAA Journal, Vol. 32, No. 8,
1994, pp. 1598-1605.
8a
Somers, D.M., “Design and Experimental Results for the S809 Airfoil”, Airfoils, Inc., State College, PA, 1989.
8b
Somers, D.M., “Design and Experimental Results for the S809 Airfoil”, NREL/SR-440-6918, January 1997.
9
Simms, D., Schreck, S., Hand, M, and Fingersh, L.J. , “NREL Unsteady Aerodynamics Experiment in the NASA-
Ames Wind Tunnel: A Comparison of Predictions to Measurements.”, NREL Technical report, NREL/TP-500-29494, 2001.
10
Drela, M., and Giles, M. B., “Viscous-Inviscid Analysis of Transonic and Low Reynolds Number Airfoils”, AIAA Journal,
Vol. 25, No. 10, October 1987, pp. 1347 – 1355.
11
Menter, F.R., and Egorov, Y., “A Scale-Adaptive Simulation Model using Two-Equation Models” AIAA Paper 2005-1095,
2005.
12
Duque, E.P.N., Burklund, M.D., Johnson, W., “Navier-Stokes and Comprehensive Analysis Performance Predictions of the
NREL Phase VI Experiment”, AIAA Paper 2003-0355, 2003.
11
American Institute of Aeronautics and Astronautics