You are on page 1of 15

American Institute of Aeronautics and Astronautics

1
Prediction of Laminar-Turbulent Transitional Flow over
Single and Two-Element Airfoils
Shirzad Hosseinverdi
1
and Masoud Boroomand
2
Nomenclature

AmirKabir University of Technology (Tehran Polytechnic), Tehran 15875-4413
The correlation-based transition model has been developed to predict precisely the
laminar-turbulent transitional flow over single and two-element airfoils. At relatively low
Reynolds number, laminar-turbulent transition has an important role on skin friction and
the onset of separation and eventually on aerodynamic characteristics. For this reason, the
accurate prediction of transition onset is fundamental to the modeling of such flows. This
model is based on coupling the turbulence model with empirical and semi empirical
correlations. The two-equation shear stress transport (SST) k- turbulence model of Menter
was employed to determine accurately turbulent flow. To predict the transition onset, two
different methods were used. The first method is the Cebeci & Smith correlation which is
based on Michel's method and Smith e
9
-correlation. The second method is based on the
linear stability theory, originally proposed by Smith and Van Ingen that is referred to the e
n

method. The laminar boundary layer properties required for calculation of the transition
onset and extent were calculated by solving a two-equation integral formulation with the
pressure distribution of the Navier-Stokes solution as the input. Whereas the extent of
transition calculated by using the intermittency function that is based on the work of
Dhawan and Narasimha. The developed transition model was used to predict the
incompressible transitional flow over the flat plate under zero pressure gradient, natural
laminar flow airfoil NLF(1)-0416, S809 wind turbine airfoil and two-element NLR 7301
airfoil with trailing edge flap. For the cases of flow over flat plate, good agreement of skin
friction between the transitional computed and the experimental data were observed. In the
case of airfoil flows, the significant improvements were obtained in drag prediction by using
the transitional computation in comparison with the fully turbulent simulation in the given
experimental data.
C = airfoil chord
C
f
= skin friction coefficient
Cd

= dissipation coefficient
C
D
= drag coefficient
C
L
= lift coefficient
C
P
= surface pressure coefficient
d = normal wall distance
H = shape factor
H* = kinetic energy shape factor
K = turbulent kinetic energy
Re
C
= Reynolds number based on chord length, U

C/
Re
X
= Reynolds number based on local distance, U
e
X/
Re

= Reynolds number based on momentum thickness, U


e
/
S = surface distance
Ti = free stream turbulence intensity
U

= free stream velocity


X
tr
= transition onset

1
Student, Department of Aerospace Engineering, hosseinverdi@aut.ac.ir, Student Member AIAA.
2
Associate Professor, Department of Aerospace Engineering, boromand@aut.ac.ir, Senior Member AIAA.
40th Fluid Dynamics Conference and Exhibit
28 June - 1 July 2010, Chicago, Illinois
AIAA 2010-4290
Copyright 2010 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

American Institute of Aeronautics and Astronautics


2
y
+
= non-dimensional wall-normal distance
= absolute value of vorticity
= boundary layer thickness
= momentum thickness
= molecular viscosity

t
= turbulent viscosity
= stream wise distance
= density

k
,

= turbulent model constants


= specific dissipation rate
= intermittency function
Subscripts
L = laminar
T = turbulent
Tr = transition
= free stream
0 = critical
I. Introduction
HE prediction of transitional flows which occur typically in aerospace applications has been objected of
research in fluid mechanics and aerodynamics for about century. However, until today, a complete
understanding of this phenomenon and what physically is happening in transitional region have not been fully
understood. Transitional boundary layer flows are important in many engineering applications, such as airfoil, wind
turbine and turbomachinery. In the design of aerodynamic vehicles, it is critical that the design group have an
accurate assessment of aerodynamic characteristics that are being considered. Errors in the aerodynamic coefficients
will result in errors in performance estimations and economic projection of flying vehicles. One of these
characteristics is drag. It is crucial that the drag value predicted from CFD simulations would be as close as possible
to the actual flight values. As an example of drag effects on flying objects, a reduction of one drag count (C
D
=10
-4
)
on a subsonic civil transport airplane means about 200 lb (one person) more in payload, Van Dam
26
.
Drag over a wing can be divided into three major components: viscous, induced and wave drag. Induced drag is
due to the induced downwash velocity by wing tip vortices; wave drag is result of the shock waves generated in
transonic flow and the viscous drag generated by skin friction and thus it is directly related to viscous flow behavior.
Induced and wave drag are generated by normal forces and they can be well predicted using Euler equations
simulations. On the other hand, viscous drag can only be predicted by solving the Navier-Stokes equations. This
study focuses on two-dimensional, incompressible transitional flow over flat plate and airfoils so the center of
attention is on the viscous drag component. In the drag evaluation, one of the key factors is laminar-turbulent
transition.
There are number of different transition mechanisms which depend on the factors such as: turbulence level of the
external flow, the pressure gradient along the laminar boundary layer, the geometrical details, the surface roughness
and the free stream Mach number.
In the aerodynamic flows, it is generally assumed that the most common transition mechanism is natural
transition which is the result of flow instability (Tollmien-Schlichting waves or cross-flow instability), where the
resulting exponential growth of two-dimensional waves eventually results in a non-linear break down to turbulence,
Schlichting.
Another transition mechanism that has also received heightened attention is called bypass transition. In
turbomachinery application, the main transition mechanism is bypass transition imposed on the boundary layer by
high levels of turbulence in free stream, Morkovin
27
. Free stream with high level of turbulence generated by
upstream blade rows is an example of this type of flows. Another example of by pass transition that can occur in
aircrafts flights is on the tail surfaces located in the wake of fuselage, or on flaps of a multi-element airfoil.
Moreover, the well known important transition mechanism is separation-induced transition, Mayle
13
, where a
laminar boundary layer separates under the influence of pressure gradient and transition develops within the
separated shear layer (which may or may not reattach).
Transition analysis is divided in two tasks; the first one is the prediction of transition onset while, the second one
is evaluation of the transition extent. In the first category, various methods have been used to predict transition that
can be classified in the following manner:
T

American Institute of Aeronautics and Astronautics


3
A Direct Numerical Simulation (DNS) is performed by solving the full unsteady Navier-Stokes equations. Since
there is no Reynolds averaging there is no requirement for turbulence closure by a turbulence model. In order to
capture the small scales of turbulent a DNS computation requires extremely fine grid. Due to the extraordinary large
computational requirements, DNS is clearly not yet at a stage where it can serves as a practical tool for engineering
design applications.
Because of the significant computational costs associated with DNS, a number of researchers have applied the
concept of Large Eddy Simulation (LES) to transitional flows. One of the main problems with LES is that the
predicted transition location is very sensitive to the choice of Smagorinsky constant which is used to calibrate the
subgrid eddy viscosity. Germano et al.
28
proposed a dynamic subgrid-scale model which computes the Smagorinsky
constant locally. Nevertheless, the dynamic LES model is not a complete solution to the issues associated with
applying LES to predict transitional flows. Another approach is the application of low-Reynolds number turbulence
models. However, the ability of low-Reynolds turbulence models to predict transition seems to be coincidental. This
is based on reproducing the viscous sublayer behavior, not on predicting transition from laminar to turbulent flow. It
is now generally accepted that the use of turbulence models without any coupling to an intermittency equation
appears to be very delicate and often unreliable method of predicting transition. In addition, low-Reynolds
turbulence models can be applied to bypass transition and are therefore not suitable for aerodynamic flows.
One of the most common ways of predicting boundary layer transition onset location is employing empirical
correlations. The empirical correlations usually relate to the free stream turbulence intensity (Ti) and to the transition
Reynolds number based on the momentum thickness Reynolds number (Re
t
). A typical example is the Mayle
13

correlation which is based on fairly recent high-quality experimental data for transition onset. Another popular
correlation is that of Abu-Ghannam and Shaw
14
. It is based on the older data but it can account for the effect of
pressure gradient on the transition onset location. There are also other models such as the methods of Michel
30
,
Granville
31
, Smith and Gamberoni
8
, Van Ingen
9
, and Van Driest and Blumer
29
. Of those that are widely recognized
for their accuracy and ease of implementation in engineering applications are Michel's and Granville's methods.
Michel based his method on relating the transition momentum thickness Reynolds number Re
t
to the local distance
Reynolds number Re
x
through a single curve that matches fairly well with the experimental data for two-
dimensional airfoils in incompressible flows. Later Cebeci & Smith
1
in according with the Michel's method,
proposed a similar curve that is compatible with the e
9
method.
Finally, one of more widely used method for predicting transition is so-called e
n
method. This model is based on the
local linear stability theory and the parallel flow assumption in order to calculate the growth of disturbance
amplitudes from the boundary layer natural point to the transition location. Drela and Giles
3
develop the semi-
empirical model to predict the transition onset; this model is based on the Orr-Sommerfeld equation coupled with
empirical correlations originally developed for two-dimensional steady incompressible flow.
The other aspect of transition analysis process is that of the transition extent. That is the zone over which the
boundary layer undergoes a change from a fully laminar flow to a fully turbulent flow. Such a region starts at the
transition onset location and ends at the point where the flow is 100% turbulent. The region can be represented
quantitatively by using intermittency functions. Dhawan and Narasimha
4
have correlated the experimental data and
proposed a generalized distribution function across the transition region. Chen and Thyson
7
based their work on
Emmons spot theory for predicting the transition and they have suggested an intermittency function to achieve a
smooth transition from laminar into turbulent flow in the boundary layer. Later Cebeci
2
added some modifications
for the function, so that it will include cases of two-dimensional low-Reynolds-number flows where laminar
separation bubbles increase in size and in occurrences. Edward et al.
5
have developed a one- equation turbulence
transition model that is based on both the Spalart-Allmaras one-equation turbulence model and a transition model
that was developed by Warren and Hassan
6
. Intermittency function that used by Edward et al. consists of two
components. One is based on the work of Dhawan and Narasimha and labeled as the surface-distance-dependent
component. The other is a multidimensional component that is used to calculate the transition distribution normal to
the surface. Then the two components are combined together to describe the transition zone distribution in the flow.
It is to be concluded that by using the empirical and semi-empirical correlations developed to model both the
onset and extent of transition, major time and effort can be saved in comparison with the percent accuracy to be
gained by using more complex procedures.
This work is focused on the modeling of the transition of laminar-turbulent boundary layer. To predict the
transition onset location, the empirical correlation of Cebeci & Smith and semi-empirical correlation of e
n
method
are used. To capture the transition region the intermittency function of Dhawan and Narasimha with the surface-
distance-dependent component of Edward et al. is used. The accuracy of developed transition model is assessed for a
transitional flow over flat plate under zero pressure gradient, natural laminar flow airfoil NLF(1)-0416, S809 wind
turbine airfoil and two-element NLR 7301 airfoil with trailing edge flap.

American Institute of Aeronautics and Astronautics


4
II. Transition Modeling
In a typical flow simulation using RANS solver coupled with a turbulence model, the effective turbulence
viscosity is considered as
T L eff
+ = (1)
Where,
L
and
T
are the laminar and turbulent viscosity coefficients, respectively. One of the methods of
introducing transition in a fully turbulent boundary layer is by multiplying the turbulent viscosity by an
intermittency function. Then the modified effective viscosity is equal to
T L eff
+ = (2)
Thus for =0, the boundary layer is fully laminar and for =1, the boundary layer is fully turbulent. For any
value in between zero and one, the flow is in the transition region. Turbulent viscosity
T
is obtained from a
turbulence model and is calculated by using the onset prediction method and transition extent model.
A. Turbulence Model
Among the several variations of widely used two-equation turbulence models, the shear-stress transport (SST) k-
turbulence model of Menter
19
was adopted to properly resolve the flow properties in turbulent regime. This model
is a two-equation eddy-viscosity model which merges the k- model of Wilcox with a high Reynolds number k-
model (transformed into the k- formulation). The transport equations for the turbulent kinetic energy and the
specific dissipation rate of turbulent in Cartesian coordinate are as follows:
( )
( )

k P
x
k
ku
x t
k
k
j
t k j
j
*
=
(
(

(3)
( )
( ) ( )
j j
k
t j
t j
j
x x
k
f P
x
u
x t

+ =
(
(


1
1 2
2 , 1
2
(4)
Turbulent viscosity
t
and production of turbulent kinetic energy P
K
are computed thorough the following relations
( )
2 1
1
, max f a
k a
T

(5)
j
i
ij ij k k i j j i t k
x
u
k u u u u P

|
.
|

\
|
+ =
3
2
3
2
. ,
(6)
Auxiliary functions (f
1
, f
2
) are given by
|
|
|
.
|

\
|
|
|
.
|

\
|
(
(

|
|
.
|

\
|
=
4
2
2
2 *
1
4
,
500
, max min tanh
y CD
k
y y
k
f
k


(7)

(
(

(
(

|
|
.
|

\
|
=
2
2 *
2
500
,
2
max tanh

y y
k
f (8)
|
|
.
|

\
|

=
20
2
10 ,
1
2 max
j j
k
x x
k
CD


(9)
In the above formulation, y is the normal-distance to the wall. The constants of model are: a
1
=0.31, *=0.09.

American Institute of Aeronautics and Astronautics


5
Finally, the coefficients of , ,
k
and

are obtained from the relation 10:


( )
2 1 1 1
1 f f + = (10)
Let
1
represents any constant in k- model (inner),
2
any constant in the transformed k- model (outer) and the
corresponding constant of k- (SST) model. Constants of inner and outer models are given in table 1.
Table 1. Coefficient of outer and inner models
Model


k


Inner (1) 5/9 3/40 0.85 0.5
Outer (2) 0.44 0.0828 1 0.856

B. Transition Onset Prediction Methods
To predict the transition onset, two different methods were used. First model is the correlation based on Michel's
method and Smith's e
9
-correlation which derived by Cebeci & Smith.
They suggest that the following correlation gives more accurate results than the empirical one-step model of Michel;
46 . 0
,
,
,
Re
Re
22400
1 174 . 1 Re
tr x
tr x
tr
(
(

+ =

(11)
The above expression is derived for attached and incompressible flow. According to this method, the boundary
layer development on the body is calculated for a laminar flow starting at the leading-edge of flow so that both Re


and Re
x
can be determined. The location where the (Re

, Re
x
) values interact this curve corresponds the onset of
transition location.
The second method is based on linear stability theory and is referred to the e
n
model that originally proposed by
Smith and Gamberoni and Van Ingen. The en assumes that transition occurs when the most unstable Tollmien-
Schlichting wave in the boundary layer has grown by some factor of e
n
, where n is defined as follows:
( ) ;
~
max x n n = ( )

=
x
x
i
dx x n
0
) ( ;
~
(12)
Where is the frequency, x
0
() is the onset location of instability, -
i
is the spatial growth rate of the TS wave and
n(x ;..) describe the amplitude growth of the disturbance along the surface. Given a velocity profile, the local
disturbance growth rate can be determined by solving the Orr-Sommerfeld eigenvalue equations. The amplification
factor is calculated by integrating the growth rate, usually the spatial growth rate, starting from the point of neutral
stability.
Drela and Giles, using the Falkner-Skan profile family, have solved the Orr-Sommerfeld equation for the various
shape factors and unstable frequencies. The logarithmic of the maximum amplification rate n is calculated by
integrating the local amplification rate downstream from the point of instability, simply assumes that the slope of the
amplification rate dn/dRe

is only a function of local calculated shape actor H and is given by the corresponding
self-similarity solution:

Re
Re
Re
Re
0 ,
d
d
dn
n

= (13)
Where, Re

is Reynolds number based on the boundary layer momentum thickness. The critical Reynolds number
(Re
,0
) and the slope dn/dRe

are expressed by the following formulas:



44 . 0
1
295 . 3
9 . 12
1
20
tanh 489 . 0
1
415 . 1
Re log
0 , 10
+

+
|
.
|

\
|

|
.
|

\
|

=
H H H

(14)
| | { }
5 . 0
2
25 . 0 ) 65 . 4 5 . 1 tanh( 5 . 2 7 . 3 4 . 2 01 . 0
Re
+ + = H H
d
dn

(15)

American Institute of Aeronautics and Astronautics


6

For non-similar flow, based on the properties of the Falkner-Skan profile family, the amplification factor with
respect to the spatial coordinate is determined as follow:

1
1
2
1
Re
Re
Re
2
e
e e e
e
u
d
du
u d
dn
d
d
d
dn
d
dn
|
|
.
|

\
|
+ = = (16)
An explicit expression for the integrated amplification factor then becomes:

d
d
dn
n

=
0
) ( (17)
Where
0
is the point at which Re

=Re
,0
. The n factor is empirically determined from several experimental data
and can vary from one flow situation to another. The free stream effect can be incorporated into the e
n
method by the
following correlation suggested by Mack
10
:
), 01 . 0 ln( 4 . 2 43 . 8 Ti n = 98 . 2 07 . 0 Ti (18)
C. Transition Extent Prediction
The extent of the transition region is obtained by using an intermittency function that will divide the flow into
three zones: the laminar zone with a corresponding intermittency function value of zero, the fully turbulent zone
with value of 1, and the transition zone that will have a value that varies between zero at the beginning of transition
and progressively increase in the transitional region until it reaches unity in the fully turbulent region.
The intermittency function used in the current work is composed of two parts, a surface-distance-dependant
component
S
based on the work of Dhawan and Narasimha and a multidimensional component
b
(x,y) of Edwards
et al. that serves to restrict the applicable range of the transition model to boundary layers. The two components are
blended together through the following equation:
| | 1 ) , ( 1 ) , ( + =
s b
y x y x (19)
The Dhawan and Narasimha expression
S
is defined along the surface of the geometry from the stagnation point:
) 412 . 0 exp( 1
2
=
S
(20)
, / ) max(
tr
s s =

75 . 0
,
Re 0 . 9 Re

=
tr s
(21)
The normal-distance-dependent function
b
is defined on the basis of the work of Edwards et al. as follows:
) tanh( ) , (
2
= y x
b
(22)
| |

+

=
t t
t t t
3
2 1
) , max( , 0 max
(23)

= =
+
= =

2
5
3
5 . 1
2
2
1
10 , ,
) (
,
500 U
t C t
d C
t
d
t
t
(24)

b
approaches to the one near the solid surface, and decays sharply to zero as the edge of boundary layer is
approached. By using the intermittency function, the RANS equations are coupled with the transition model in
following way:


=
t
t

0

t
t
x x
x x
>

(25)


American Institute of Aeronautics and Astronautics


7
D. Calculation of Laminar Boundary Layer Properties
According to the equations (11) to (21), the laminar boundary layer parameters, *, with the boundary layer
edge velocity u
e
, are required to calculate the transition onset and extent. The direct procedure to extract the
boundary layer parameters is using the definition of displace and momentum thickness in the following way:
, ) 1 (
0
*


d
u
u
e

=

d
u
u
u
u
e e
) 1 (
0

= (26)
Determination of boundary layer properties using the above equations and Navier-Stokes solution as the input are
in some difficulties and are not accurate because firstly the boundary layer thickness is not well defined. Secondly,
the turbulence starting from the transition point affects the integral parameters upstream. These results are differing
from their fully laminar value in the boundary layer. So an alternative procedure is required for calculating these
parameters. In the present work, the two-equation integral formulation based on the dissipation closure is chosen.
The first equation is the Von Karman integral relation given by:
2
) 2 (
f
e
e
C
d
du
u
H
d
d
= + +

(27)
Where is momentum thickness, H is shape factor, C
f
is skin friction coefficient, u
e
is the velocity at the boundary
layer edge and is he stream wise coordinate. The second equation is a combination of equation 27 and the kinetic
energy thickness equation, and is given by:
2
2 ) 1 (
* *
*
f
d
e
e
C
H C
d
du
u
H H
d
dH
= +

(28)
Where H
*
is the kinetic energy shape parameter and C
d
is the dissipation coefficient. For laminar flow, the two
ordinary first order differential equations can be solved with the following closure relationships for H*, C
f
and C
d

respectively:

+
=
H
H
H
H
H
2
2
*
) 4 (
040 . 0 515 . 1
) 4 (
076 . 0 515 . 1

4
4
>
<
H
H
(29)
) 1 (
) 4 . 7 (
01977 . 0 067 . 0
2
Re
2

+ =
H
H
C
f

4 . 7 < H (30)
| |

+
=
2
2
5 . 5
*
) 4 ( 02 . 0 1
) 4 (
003 . 0 207 . 0
) 4 ( 00205 . 0 207 . 0
2
Re
H
H
H
H
C
d

4
4
>
<
H
H
(31)
This model has successfully been used by Drela and Giles. To solve the two equations of 27 and 28 a third relation is
necessary. For flow over a curved object, the boundary layer edge velocity can be detected by using the Bernoulli
equation, because pressure can be assumed constant normal to the wall within the boundary layer (p
wall
=p
edge
). The
u
e
distribution is simply given by:
P e
C U u =

1 (32)
III. Numerical Method and Validation
The Reynolds-Averaged Navier-Stokes and turbulence model equations are solved by finite-volume method in a
non-uniform orthogonal body-fitted grid. Advection term is approximated by Quick scheme of Hayase et al.
21
. To

American Institute of Aeronautics and Astronautics


8
avoid checkerboard oscillation with the collocated storage arrangement Rhie and Chow
20
interpolation method
was used. The solution is affected by an iterative pressure-correction SIMPLE algorithm. The momentum integral
equation (Eq. 27) and kinetic energy thickness equation (Eq. 28) are solved using a fourth-order Runge Kutta
method. The coupling procedure used in the present method is simple. For a given surface pressure distribution
calculated from the code, a boundary layer edge velocity distribution is obtained by utilizing Bernoulli's equation.
Once the laminar boundary layer parameters are calculated along the body surface by solving equation 27 and 28,
those parameters are used to compute the transition onset.
A flow over NACA-2412, Abbot
22
at high Reynolds
number was used to validate a numerical procedure and
turbulence model. The simulation was performed at high
Reynolds number of 5.710
6
, because whatever the
turbulence region on the airfoil surface is greater the role
of laminar-turbulent transition is negligible. Figure 1
represents the variation of drag coefficient for the
current case at Reynolds number of 5.710
6
in wide
range of angles of attack in which the experimental data
are plotted as open symbols, and the solid line represent
the fully turbulent computations by k- SST. The
numerical results obtained by fully turbulent simulation
are in good agreement with the experimental data. A
slight discrepancy in results is due to the existence of
small laminar region on airfoil surface in low angle of
attacks. This problem is eliminated in higher angle of
attacks because; most part of drag at these angles is due
to the pressure drag.
IV. Results and Discussion
In order to verify the accuracy and effectiveness of developed transition model, four well-documented
experimental test cases were investigated: a transitional flow over flat plate under zero pressure gradient, a natural
laminar flow airfoil NLF(1)-0416, S809 wind-turbine airfoil, and two-element NLR 7301 airfoil with trailing edge
flap. The results obtained by using a developed transition model was compared with those obtained with fully
turbulent computation and were also compared with experimental data.
A. Transitional Flow over Flat Plate
The first case to be investigated is the simplest (though fundamentally important) transitional flow over flat plate
under zero pressure gradients. The developed transition model has been used to predict the natural transition of
Schabauer and Klebanoff
11
flat plate experiment, which was performed in a relatively quiet wind tunnel and had a
free stream turbulence intensity of only about 0.1
percent near the transition location.
The computational domain extends from 0.15m
upstream of leading edge to 2m downstream. The
domain height is 0.3m. The computational grid consists
of 500100 nodes, in which sufficient number of grid
lines clustered towards the leading edge with the first
grid point y
+
value below 1. Boundary conditions were
as follows: Symmetry boundary conditions were
imposed on the lower boundary faces located upstream
of the leading edge at x=0, and on the upper boundary of
domain. Along the flat plate the no-slip boundary
condition is used. A specified constant velocity inlet was
used together with zero stream-wise gradients at outlet.
The free stream turbulence quantities at the inlet are
specified as follow:
( )
2
5 . 1

= U Ti k ,

= l k /
2 / 3
(33)

-5 0 5 10 15
0.01
0.02
0.03
0.04
Experimental
Fully Turbulent Comp
CD

Figure 1. Variation of drag coefficient against the
angle of attack at Re
C
=5.710
6
for NACA-2412 airfoil

Re
1 2 3 4 5 6
0.002
0.004
0.006
0.008
0.010
Experimental: Schubauer and Klebanoff
Transitional Comp: e
Transitional Comp: Cebeci&Smith
Fully turbulent Comp
Laminar Theory
Turbulent Theory
x
n
x
Cf
10
-05

Figure 2. Distribution of skin friction coefficient for
natural transition test case




American Institute of Aeronautics and Astronautics


9
Figure 2 shows the skin friction coefficient, C
f
of Schabauer and Klebanoff, fully turbulent computation as well
as for transitional flow simulation using both e
n
and Cebeci & Smith models, the Blasius solution for laminar flow,
and turbulent correlation. The Blasius solution is given as:
2 / 1
Re 664 . 0

=
x f
C (34)
The turbulent correlation is given as:
) Re 06 . 0 ( ln / 455 . 0
2
x f
C = (35)
It is observed that using the Cebeci & Smith
criterion, transition was predicted upstream of the
location identified by experimental data. But e
n
data
indicates onset of transition at roughly the same location
as the experiment. In the transition region the computed
skin friction increase rapidly and then follows the
turbulent correlation. To consider the effect of turbulent
intensity on the transition location, the variation of Re
,t

against the turbulent intensity is displayed in Figure 3.
In this figure, some other experimental data at different
turbulent intensities, the results of two correlations of
Mayle and Abu-Ghannam & Shaw, e
n
, and Cebeci &
Smith are shown. As can be seen, for flow over flat plate
under zero pressure gradients, Cebeci & Smith
prediction is constant and does not varies with turbulent
intensity. The prediction of e
n
method is in good
agreement with the experimental data. According to the
Eq. (18) (Mack Correlation), n approaches to zero when
the level of turbulent intensity reaches to the 3 percent,
thus transition occurs at critical Reynolds number, Re
,0
.
The transition model developed in the previous
section has been also used to predict the ERCOFTAC
(European Research Community on Flow, Turbulence
and Combustion) T3A flat plate test case (Ti=3%,
U

=5.4m/s). These experiments were performed by


researchers at Rolls Royce in the early 1990s, Savill
17-
18
. Predicted skin friction distribution for T3A is given
in Figure 4. As seen from Figure 4, the location of
transition is close to the experiment but the growth of
skin friction is greater. In other words, the predicted
critical Reynolds number by e
n
method is in good
agreement with the experimental one in this test case.
B. Natural Laminar Flow Airfoil, NLF (1)-0416
The NLF(1)-0416 is a natural laminar flow airfoil
that was developed by D. Somers
23
at NASA Langley
Research Center to achieve a higher maximum lift
coefficients with a low cruise drag coefficient for
general aviation applications. The experimental data are
available in wide range of angle of attacks for Mach
number between 0.1 to 0.4 and Reynolds number
between 1 to 9 million. The numerical presented in this
section correspond to the following free stream
condition:
, 1 . 0 =

M
6
10 0 . 2 Re =
C
And 13 6 < < (36)
Tu (%)
0 1 2 3 4 5 6
300
600
900
1200
1500
1800
Experimental: ERCOFTAC
Experimental: Fashifar and Johnson
Experimental: Schubauer and Skramstad
Experimental: Sinclair and Wells
Correlation: Abu-Ghannam and Shaw
Correlation: Mayle
Cebeci&Smith
e
Re
t
n

Figure 3. Variation of Transition onset momentum
thickness Reynolds number with the turbulent
intensity




Re
0.005
0.010
0.015
Exprimental: T3A
Transitional Comp: e
Fully Turbulent Comp
Laminar Theory
Turbulent Theory
x
n
Cf
0.2 0.4 0.6 0.8 1.0 1.2
x10
-05

Figure 4. Distribution of skin friction coefficient
versus local Reynolds number for T3A test case





X/C
0 0.2 0.4 0.6 0.8
-0.2
-0.1
0.0
0.1
0.2
Y/C

Figure 5. Geometry of the NLF(1)-0416 airfoil







American Institute of Aeronautics and Astronautics


10
The C-topology computational mesh was used with
whole 35000 nodes. The near-wall mesh spacing is
0.710
-05
chord length, resulting in a peak value of y
+

vary from approximately 0.75 on pressure surface to a
peak value of 0.85 on the suction surface. This mesh is
fine enough to yield a mesh-independent result. The
domain extends from 12 chord lengths upstream to 35
chord lengths downstream. The airfoil geometry and
corresponding mesh is shown in Figures 5 and 6,
respectively.
Figure 7 shows the calculated and experimental surface
pressure coefficient distribution at =0. In general, the
numerical results agree well with the experimental data.
There is slight difference between fully turbulent and
transition simulations especially at x/c=0.4 on suction
surface and x/c=0.6 on pressure surface. As seen from
Figure 7, there is a strong adverse pressure gradient near
the 60% chord on the lower surface of the airfoil.
Experimental observation shows that the laminar
separation bubble forms in this region, Somers
23
.
Distribution of skin friction coefficient is displayed in
Figure 8. As can be seen, by using the fully turbulent
computation, the laminar separation bubble does not
occur, while an area of negative local skin friction is
captured by using the transitional simulation. Figure 9
compares the transition onsets predicted by the e
n
and
Cebeci & Smith with those obtained from experimental
data for different angle of attacks. To capture the
transition onset experimentally, the microphone was
used. Such a method is based on the concept that the
laminar flow is silent, whereas a turbulent flow is noisy.
Because the microphone is connected to individual
orifices on the model, the transition location can only be
determined as lying somewhere between two adjacent
orifices, Somers
23
. In Figure 9, the open symbols
represent orifice location at which the flow is laminar
and the solid symbols represent orifice locations at
which the flow is turbulent. The predicted transition
X/C
0.2 0.4 0.6 0.8 1
0
0.002
0.004
0.006
0.008
Transitional Comp: e
Fully Turbulent Comp
Cf
n

Figure 8. Distribution of skin friction coefficient at
=0 Re
C
=2.010
6
for NLF(1)-0416 airfoil


Figure 6. Grid used for NLF(1)-0416 airfoil with close
up view of leading and trailing edges' mesh





X/C
0 0.2 0.4 0.6 0.8 1
-1.0
-0.5
0.0
0.5
1.0
Experimental
Transitional Comp: e
Fully Turbulent Comp
Cp
n

Figure 7. Distribution of surface pressure coefficient
at =0 and Re
C
=2.010
6
for NLF(1)-0416 airfoil









Xtr/C
0 0.2 0.4 0.6 0.8
-5
0
5
10
15 Experimental: Upper
Experimental: Lower
e : Upper
e : Lower
Cebeci&Smith: Upper
Cebeci&Smith: Lower

n
n

Figure 9. Calculated and measured transition onset
location for the NLF(1)-0416 airfoil at Re
C
=2.010
6







American Institute of Aeronautics and Astronautics


11
locations by two methods are in the vicinity of the ranges given by the experimental data for a large range of angle
of attacks. Figure 10 displays the variation of drag coefficient with angle of attack for four sets of data: one set is
given experimentally, Somers
23
, another one is obtained by using the fully turbulent simulation and other two sets
are obtained numerically by developed transitional model. Significant improvement achieved in prediction of drag
coefficient by switching from a fully turbulent modeling to the developed transition model one. At zero angle of
attack, the experimental drag coefficient is 0.0069
whereas the predicted value by fully turbulent
computation is 0.011 with about 60 percent error. The
corresponding drag polar in Figure 11 illustrates the
importance of free transition prediction capability. The
value obtained by e
n
method is slightly better than the
Cebeci & Smith one.
C. S809 Wind Turbine Airfoil
The next case considered in this work is S809 airfoil
whose aerodynamic characteristics are representative of
horizontal axis wind turbine (HAWT) airfoils. The S809
is a 21% thick, laminar-flow airfoil designed specifically
for HAWT applications, Somers
24
. Geometry of the
airfoil is shown in Figure 12. A 600mm chord model of
the S809 was tested in the 1.8 m 1.25 m, low-
turbulence wind tunnel at the Delft University of
Technology. The results of these tests are reported by
Somers
24
and were used in this work for comparison
with the numerical results. This case was identified as a
good candidate to test the developed transition model.
The grid used to perform the numerical calculation is
C-topology similar to the one used for the previous test
case. The mesh composed of 37000 nodes that the
distance of first cell from the wall is 10
-05
chord length.
Surface pressure distribution at zero angle of attack
and Re
C
=210
6
is shown in Figure 13. As it can be seen
a good agreement between experimental result and
numerical simulation obtained by fully turbulent and
transition computation. Experimental observations show
a laminar separation bubble at about 45% x/c on both
upper and lower surfaces for this configuration.
The calculated surface pressure distribution follows
closely the experimental data, except in these separated
regions as shown in Figure 13. The grid resolution is not
fine enough to properly resolve the flow features in the
bubble region, hence the discrepancy in pressure data.
The laminar bubbles regions are captured by using
developed transition model as evident from the skin
friction distribution in Figure 14, whereas the fully
turbulent simulation does not predict any separation.
During the wind tunnel tests of this airfoil, boundary
layer transition was measured using the oil flow
visualization technique as well as a probe containing a
microphone, Somers
24
.
Figure 15 compares the experimental and computed
transition points over an angle-of-attack range. In this
case, the e
n
method used only to predict the transition
onset. As seen from Figure 15, the value obtained by e
n

is in excellent agreement with the experimental results.

-5 0 5 10 15
0.01
0.02
0.03
Experimental
Transitional Comp: e
Transitional Comp: Cebeci&Smith
Fully Turbulent Comp

CD
n

Figure 10. Variation of drag coefficient against the
angle of attack for NLF(1)-0416 airfoil at
Re
C
=2.010
6












C
0.01 0.02 0.03
0.0
0.5
1.0
1.5
2.0
Experimental
Transitional Comp: e
Transitional Comp:Cebeci&Smith
Fully Turbulent Comp
CL
D
n

Figure 11. Drag polar for the NLF(1)-0416 airfoil at
Re
C
=210
6












X/C
0 0.2 0.4 0.6 0.8 1
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
Y/C

Figure 12. S809 airfoil geometry


American Institute of Aeronautics and Astronautics


12
The variation of drag coefficients with the angle of
attack and drag polar obtained from the fully turbulent
computations and developed transition model
simulations were compiled together with the
experimental data and are given in Figures 16 and 17,
respectively. These figures again illustrate the
significant improvement in the drag values predicted
using the free transition model as opposed to those
obtained using the fully turbulent model. In this test
case, the difference between the experimental results
and fully turbulent simulation is more considerable than
the previous test case (NLF (1)-0416).
The error values between the fully turbulent and
transition modeling with experimental data are given in
table 2.




Xtr/C
0 0.2 0.4 0.6 0.8
-5
0
5
10
15
Experimental: Upper
Experimental: Lower
e :Upper
e :Lower
n
n


Figure 15. Experimental and calculated transition onset
location for S809 airfoil at Re
C
=2.010
6


X/C
0 0.2 0.4 0.6 0.8 1
0
0.002
0.004
0.006
0.008
Transitional Comp: e
Fully Turbulent Comp
Cf
n

Figure 14. Distribution of skin friction coefficient for
S809 airfoil at =0 Re
C
=2.010
6

-5 0 5 10
0.004
0.008
0.012
0.016
Experimental
Transitional Comp: e
Fully Turbulent Comp
C
n
D

Figure 16. Comparison of drag between calculated
and measured for S809 airfoil at Re
C
=2.010
6


C
0.005 0.01 0.015 0.02
-0.5
0.0
0.5
1.0
Experimental
Transitional Comp: e
Fully Turbulent Comp
CL
n
D

Figure 17. Calculated and experimental results of
drag polar for S809 airfoil at Re
C
=2.010
6


X/C
0 0.2 0.4 0.6 0.8 1
-1.0
-0.5
0.0
0.5
1.0
Experimental
Transitional Comp: e
Fully Turbulent Comp
Cp
n

Figure 13. Distribution of surface pressure coefficient
for S809 airfoil at =0, Re
C
=2.010
6




American Institute of Aeronautics and Astronautics


13
Table 2. Error comparisons between fully turbulent and transition computation

Deg
Experimental
Data
Fully
Turbulent Comp
Developed
Transition Comp
Error %
(Fully Turbulent)
Error %
(Transition Comp)

-4 0.0071 0.0112 0.00664 57.74647887 -6.478873239
-2 0.0070 0.0111 0.00663 58.57142857 -5.285714286
0 0.0070 0.0109 0.00681 55.71428571 -2.714285714
2 0.0072 0.0112 0.00700 55.55555556 -2.777777778
4 0.0069 0.0123 0.00720 78.26086957 4.347826087
5 0.0070 0.0130 0.00753 85.71428571 7.571428571
6 0.0096 0.0139 0.00970 44.79166667 1.041666667
7.5 0.0134 0.0156 0.01337 16.41791045 -0.223880597
D. Two-Element NLR 7301 Airfoil with Trailing Edge Flap
Final investigated test case is well-documented two-element NLR 7301 airfoil with trailing edge flap. The
geometry and experimental data used for comparison are given in the AGARD AR-303, Van Den Berg
25
. The
experimental data are available for following free stream condition:
, 185 . 0 =

M , 10 51 . 2 Re
6
=
C

0 0 0
1 . 13 , 1 . 10 , 6 = (37)
The flap was positioned with a deflection angle of 20 degree, an overlap of 5.3%C and a gap of 2.6%C in the
experiment .A C-grid was used for the basic airfoil and for the flap with the wall spacing of 10
-05
chords as seen in
Figure 18. The outer boundary extended downstream to about 30 times of the chord length. Computational surface
pressure coefficients were compared with the experimental data at an angle of attack of 6.0 and 13.1 degree in
Figures 19 and 20, respectively. These plots show a good agreement between experimental results and the results
obtained by developed transition model.
In Figure 21, the transition onset predicted using the e
n
model is plotted alongside the experimental values for the
three available angles of attack, namely 6, 10.1 and 13.1 deg. Transition onset location is well predicted as can be
seen from Figure 21. The drag coefficient obtained from the simulations is compiled together with the experimental
data and are given in Figure 22. Figure 22 shows a significant improvement in the drag values predicted using the
developed transition model as opposed to those obtained using the fully turbulent model. For an angle of attack 6
deg, the experimental drag coefficient is equal to 0.0229. The difference between the numerical value computed by
the fully turbulent model and the experimental one is
equal to 0.007726. This difference drops to 5 drag
counts with the developed transition model. The results
obtained for the other two angles of attack (10.1 and
13.1 deg) show the same behavior. model was used to
simulate several test cases; flow over flat plate under
zero pressure gradient, natural laminar flow airfoil
NLF(1)-0416, S809 wind turbine airfoil and two-
element NLR 7301 airfoil with trailing edge flap.
turbulence model of Menter was employed to determine
accurately turbulent flow. The developed transition
Predicted transition points are compared with
experimental data and generally, agreement is good. In
the single and two-element airfoil flow cases, the drag
values predicted using the developed transition model
are closer to the experimental values when compared
with those computed using the fully turbulent model.
Nonetheless, it can be concluded that the developed
transition model, presented in this work, resulted in a
significant improvement in drag prediction for airfoils
and skin friction coefficient of flat plate in transitional
flow.



Figure 18. Grid for NLR 7301 airfoil with trailing
edge flap


American Institute of Aeronautics and Astronautics


14










V. Conclusion
A simple but robust and accurate correlation-based transition model has been developed for 2-D Navier-Stokes
flow analysis. This developed transition model predicts both the onset and the extent of transition. The onset point of
transition was obtained by using the e
n
method and Cebeci & Smith correlation and the extent of transition was
obtained by using the intermittency function that is based on the work of Dhawn and Narasimha. The laminar
properties are calculated by solving a two-equation integral formulation with the pressure distribution of the Navier-
Stokes solution as the input. The two-equation shear stress transport (SST) k- turbulence model of Menter was
employed to determine accurately turbulent flow. The developed transition model was used to simulate several test
cases; flow over flat plate under zero pressure gradient, natural laminar flow airfoil NLF(1)-0416, S809 wind turbine
airfoil and two-element NLR 7301 airfoil with trailing edge flap. Predicted transition points were compared with
experimental data and generally, agreement is good. In the single and two-element airfoil flow cases, the drag values
predicted using the developed transition model are closer to the experimental values when compared with those
computed using the fully turbulent model. Nonetheless, it can be concluded that the developed transition model,
presented in this work, resulted in a significant improvement in drag prediction for airfoils and skin friction
coefficient of flat plate in transitional flow.
X/C
0.1 0.3 0.5 0.7 0.9 1.1
-6
-4
-2
0
Transition Comp: Main Wing
Transition Comp: Flap
Experimental: Main Wing
Experimental: Flap
Cp

Figure 19. Surface pressure distribution for NLR
7301 with trailing edge flap at =6, Re
C
=2.5110
6


Xtr/C
0 0.2 0.4 0.6 0.8 1 1.2 1.4
0
5
10
15
Experimental: Upper Surface
Experimental: Lower Surface
Experimental: Flap
e :Upper Surface
e :Lower Surface
e :Flap
n
n

n

Figure 21. Calculated and measured transition onset
location for NLR 7301 with trailing edge flap at
Re
C
=2.5110
6


0 5 10 15
0.02
0.04
0.06
0.08
Experimental
Transitional Comp: e
Fully Turbulent Comp

CD
n

Figure 22. Drag comparison between experimental
and numerical results for NLR 7301 with trailing
edge flap at Re
C
=2.5110
6


X/C
0.1 0.3 0.5 0.7 0.9 1.1
-12
-9
-6
-3
0
Transition Comp: Main Wing
Transition Comp: Flap
Experimental: Main Wing
Experimental: Flap
Cp

Figure 20. Surface pressure distribution for NLR
7301 with trailing edge flap at =13.1, Re
C
=2.5110
6



American Institute of Aeronautics and Astronautics


15
References
1
Cebeci, T., and Smith, A. M. O., CS Method for Turbulent Boundary Layers, Analysis of Turbulent Boundary Layers,
1st ed., Academic Press, New York, 1974, pp. 329384.

2
Cebeci, T., Essential Ingredients of a Method for Low Reynolds-Number Airfoils, AIAA Journal, Vol. 27, No. 12, Dec.
1989, pp. 16801688.

3
Drela, M., and Giles, M. B., Viscous-Inviscid Analysis of Transonic and Low Reynolds Number Airfoils, AIAA Journal,
Vol. 25, No. 10, Oct. 1987, pp. 13471355.

4
Dhawan, S., and Narasimha, R., Some Properties of Boundary Layer Flow During the Transition from Laminar to
Turbulent Motion, Journal of Fluid Mechanics, Vol. 3, J an. 1958, pp. 418436.

5
Edwards, J . R., Roy, C. J., Blottner, F. G., and Hassan, H. A., Development of a One-Equation Transition/Turbulence
Model, AIAA Journal, Vol. 39, No. 9, Sept. 2001, pp. 16911698.

6
Warren, E. S., and Hassan, H. A., Transition Closure Model for Predicting Transition Onset, Journal of Aircraft, Vol.
35, No. 5, Sept.Oct. 1998, pp. 769775.

7
Chen, K., and Thyson, N., Extension of Emmons Spot Theory to Flows on Blunt Bodies, AIAA Journal, Vol. 9, No. 5,
May 1971, pp. 821825.

8
Smith, A.M.O. and Gamberoni, N., Transition, Pressure Gradient and Stability Theory, Douglas Aircraft Company,
Long Beach, Calif. Rep. ES 26388, 1956.

9
Van Ingen, J .L., A suggested Semi-Empirical Method for the Calculation of the Boundary Layer Transition Region,
Univ. of Delft, Dept. Aerospace Engineering, Delft, The Netherlands, Rep. VTH-74, 1956.
10
Mack, L. M., Transition Prediction and Linear Stability Theory, AGARD Conference Proceedings, CP-224, AGARD,
London, 1977, pp. 1-1-1-22.

11
Schubauer, G.B. and Klebanoff, P.S., Contribution on the Mechanics of Boundary Layer Transition, NACA TN 3489,
1955.

12
Schubauer, G.B. and Skramstad, H.K., Laminar-boundary-layer oscillations and transition on a flat plate, NACA Rept.
909, 1948.

13
Mayle, R. E., The Role of LaminarTurbulent Transition in Gas Turbine Engines, Journal of Turbomachinery, Vol.
113, No. 4, 1991, pp. 509537.

14
Abu-Ghannam, B. J., and Shaw, R., Natural Transition of Boundary Layersthe Effects of Turbulence, Pressure
Gradient, and Flow History, Journal of Mechanical Engineering Science, Vol. 22, No. 5, 1980, pp. 213228.

15
Fashifar, A. and J ohnson, M. W., An Improved Boundary Layer Transition Correlation, ASME Paper, ASME-92-GT-
245, 1992.

16
Sinclair, C. and Wells, J r., Effects of Freestream Turbulence on Boundary-Layer Transition, AIAA Journal, Vol. 5, No.
1, 1967, pp. 172-174.

17
Savill, A. M., Some Recent Progress in the Turbulence Modeling of By-Pass Transition, Near-Wall Turbulent Flows,
edited by R. M. C. So, C. G. Speziale, and B. E. Launder, Elsevier, Amsterdam, 1993, pp. 829848.

18
Savill, A. M., Further Progress in the Turbulence Modeling of By-Pass Transition, Engineering Turbulence Modeling
and Experiments 2,edited by W. Rodi and F. Martelli, Elsevier, Amsterdam, 1993, pp. 583592.

19
Menter F. R., Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications, AIAA Journal, 1994,
32, pp. 269-289.

20
Rhie, C.M. and Chow, W.L., Numerical study of the turbulent flow past an airfoil with trailing edge separation, AIAA
Journal, Vol. 21, 1983, pp. 1525-1532.

21
Hayase, T., Humphery, J . A. C., and Grief, R., A consistently formulated QUICK scheme for fast and suitable
convergence using finite volume iterative calculation procedure, J Comput Phys, Vol. 98, 1992, pp. 108-118.

22
Abbot I. H., and Von Doenhoff A. E., Theory of Wing Sections, McGraw-Hill Book Company, New York, 1949

23
Somers D. M., Design and experimental results for a natural laminar flow airfoil for general aviation applications,
NASA TP-1861, 1981.

24
Somers D. M., Design and experimental results for the S809 airfoil, NREL SR-440-6918, 1997.

25
Van Den Berg, B., and Gooden, J . H. M., Low-Speed Surface Pressure and Boundary Layer Measurement Data for the
NLR-7301 Airfoil Section with Trailing Edge Flap, Selection of Experimental Test Cases for the Validation of CFD Codes,
AGARD AR-303, Aug. 1994.

26
Van Dam, C. P., Aircraft Design and the Importance of Drag Prediction, VKI Lecture Series 2003-2: CFD-Based
Aircraft Drag Prediction and Reduction, National Inst. of Aerospace, Hampton, VA, Nov. 2003.

27
Morkovin, M.V., On the Many Faces of Transition, Viscous Drag Reduction, C.S. Wells, ed., PlenumPress, New York,
1969, pp 1-31.

28
Germano, M., Piomelli, U., Moin, P., Cabot, W.H., A Dynamic Subgrid-Scale Eddy Viscosity Model, Physics of Fluids
A 3, Number 7, 1991, pp. 1760-1765.
29
Van Driest, E.R. and Blumer, C.B., Boundary Layer Transition: Freestream Turbulence and Pressure Gradient Effects,
AIAA Journal, Vol. 1, No. 6, J une 1963, pp. 1303-1306.
30
Michel, R., Etude de la transition sur les profils d'aile establissment d'un point de transition et culcul de la trainee de
profil en incompressible, ONERA Report No. 1/1578A, 1952.
31
Granville, P. S., The calculation of the viscous drag of bodies of revolution, David Taylor Model basin report
849,1953.

You might also like