Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

2nd AIAA Flow Control Conference AIAA 2004-2697

28 June - 1 July 2004, Portland, Oregon

Numerical Simulation of Micro Vortex Generators

Erik Wik * and Scott T. Shaw†


School of Engineering, Cranfield University, Cranfield, Bedfordshire MK43 0AL, United Kingdom

A computational study of the aerodynamics of vane type micro-vortex generators has


been performed. Solutions of the incompressible Reynolds -Averaged Navier-Stokes
equations have been obtained for a rectangular vane type micro-vortex generator mounted
on a flat plate at a Reynolds number of 81,000 based upon vortex generator length. The
study demonstrates the importance of resolving the detailed flow around the vortex
generator. Simulations which ignore the finite thickness of the device or improperly resolve
the device boundary layer are shown to produce results that are in poor agreement with the
available experimental data. Comparisons of results obtained using conventional one- and
two-equation turbulent closures with full Reynolds stress models highlight inherent
weaknesses of conventional turbulence modeling for boundary layers containing discrete
vortical structures.

Nomenclature
c = chord
E = flux vector
h = device height
k = turbulent kinetic energy
n = unit normal
Q = conserved variables
R = core radius
s = surface area
t = time
x = axial displacement from device leading edge
y = vertical displacement from flat plate
z = horizontal displacement from device leading edge
G = circulation
d = boundary layer thickness
d* = boundary layer displacement thickness
? = boundary layer momentum thickness
O = volume
µt = turbulent viscosity
? = vorticity
? = turbulent dissipation

I. Introduction

F OR aircraft flying at high subsonic Mach numbers, local pockets of supersonic flow develop over the lifting
surfaces that are usually terminated by a shock wave. Associated with the appearance of this shock wave is an
increase in drag, resulting from the entropy losses due to the shock itself (wave drag) and from the corresponding
shock-boundary layer interaction (increased skin friction). For large transport aircraft, control of the shock can
reduce drag substantially at the cruise condition with obvious implications for fuel economy and range. In addition,
control of shock wave strength and buffet provides scope for noise reduction and flight envelope expansion for
rotorcraft and improved agility for military aircraft. The formation of shock waves, their interaction with the
boundary layer and their control have been the subject of extensive research, see for example Pearcey1 and Delery 2,3.

*
Post-graduate Student

Lecturer in Computational Aerodynamics, Department of Aerospace Sciences, Member AIAA.

1
American Institute of Aeronautics and Astronautics

Copyright © 2004 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
Early attempts to control the phenomena relied heavily on the use of passive devices to re -energize the boundary
layer ahead of the shock. Pearcey1 describes in detail the application of vortex generators to the problem of shock
control on transonic wings.
Most commonly vortex generators consist of small vanes fixed at right angles to the airframe surface and
inclined to the local stream direction. The vane produces discrete streamwise vortical structures that enhance mixing
between the retarded fluid of the boundary layer and the higher momentum fluid of the freestream. The aerodynamic
effects produced by vortex generators are strongly influenced by the size, shape, orientation and location of the
vortex generator device, Freestone4 provides an extensive review of the influence of these parameters.
The effectiveness, light weight and mechanical simplicity of vortex generators has meant that they have found
favor as a means of controlling boundary layer separation post-design to remedy unexpected problems and enhance
aerodynamic performance of growth wings. These benefits may be partly offset by adverse effects. For example,
during cruise where the vortex generators may not be needed for separation or shock control their presence may
increase aircraft drag.
To overcome these deficiencies Rao and Kariya5 have proposed the use of reduced profile sub-boundary layer
vortex generators (SBVG). SBVG devices have a height of h ~ 0.2d compared to conventional vane type vortex
generators which typically have heights h ~ 0.9d. While the basic control philosophy is unchanged, SBVG devices
are intended to redistribute momentum within the boundary layer rather than entrain higher momentum fluid from
the freestream. Fundamental studies of the fluid dynamics of sub-boundary layer vortex generators and their
application to separation control have been undertaken by Ashill and Fulker6,7 , Lin and his co-workers 8,9,10,11 and
Wendt12.
Bender13 and May 14 have proposed the use of empirical models, such as that of Bray15 to represent the influence
of vortex generators within computational analyses. This approach has the advantage that the detail of the vortex
generator does not need to be represented in the calculation. While the use of such models provides a practical tool
for airframe design care must be taken when using the model to predict flows that do not belong to the class of
problems from which the empirical model was constructed.
Numerical simulations of sub-boundary layer vortex generators have been reported by Allan16,17 and Wik18.
Allan obtained solutions of the Favre averaged Navier-Stokes equations for vortex generator devices mounted on
flat plates. Detailed comparisons of the computed data with experimental measurements show that in general the
trajectory of the vortex is correctly predicted but that the calculated vortex concentration diminishes rapidly
downstream of the device. Allan also found that the computed circulation was in poor agreement with the
experimental observations immediately downstream of the device. Wik solved the incompressible Reynolds
averaged Navier-Stokes equations for single sub-boundary layer vortex generators and arrays of co-rotating and
counter-rotating sub-boundary layer vortex generators.
This paper describes a preliminary study of the application of computational fluid dynamics to the simulation of
the flow induced by a single sub-boundary layer vortex generator. The main aim of the study was to understand and
quantify uncertainty in the modeling process. The study placed particular emphasis on understanding the geometric
fidelity of the test article, appropriate selection of a physical model (turbulence modeling) and the choice of grid.

II. Numerical Metho d

A. Governing Equations
The physical problem considered is that of viscous airflow involving shear layers (including boundary layers),
vortices and their interactions. The mathematical model is based upon the three-dimensional incompressible
Reynolds averaged Navier-Stokes equations written in an integral form,


QdΩ + ∫ E (Q ) ⋅ nds = 0
∂t ∫Ω
(1)
S

in which Q is the vector of primitive variables, E is a vector containing both convective and diffusive contributions
to the flux and H is a source term containing terms re lated to the generation and dissipation of turbulent viscosity. In
order to close this system of equations the flow is assumed to be of an ideal Newtonian gas and a turbulence model
is employed.

2
American Institute of Aeronautics and Astronautics
B. Turbulence Modeling
All of the flows computed in this study are assumed to have fully developed turbulent boundary layers. Three
turbulence closures were investigated; the one-equation model of Spalart and Allmaras, Menter’s two-equation k-?
SST model and a Reynolds stress model. The details of the models are described elsewhere 19,20,21,22 . The following
paragraphs describe briefly the salient features of each model for the present study.

1. Spalart Allmaras Model


The Spalart –Allmaras19 model requires the solution of a single transport equation for the turbulent viscosity that
has been derived from empirical considerations and dimensional analysis. The model employed in the current
investigation can be written in the form,

Dν~
Dt
1
σ
[ 2
]  υ~ 
= cb 1 S υ~ + ∇ ⋅ (υ + υ~ ) + cb 2 (∇ υ~ ) − cω1 f ω  
~
d
(2)

and the turbulent viscosity is calculated from,

υ t = υ f v1 (3)

In these equations cb1 , cb1 and σ are model constants, f v1 and f ω are the viscous and near wall damping functions
~
respectively and S is a scalar measure of the deformation tensor. In the original formulation of Spalart and Allmaras
~
S was simply equated to the magnitude of the local vorticity. More recently Dacles-Mariani et al20 recognizing the
limitations of the original formulation for flows containing vortical structures away from the wall have proposed
modifying the production terms in the original model in an effort to account for the influence of mean strain on the
production of turbulence. Both the original model (SA) and modified model (SA S-V) have been employed in the
present study.

2. Menter k -? SST Model


In Menter’s k-? SST model21 two equations for the transport of turbulent kinetic energy and turbulent dissipation
are solved,

Dk
= µT S 2 − ρβ ∗ kω
Dt
(4)
Dϖ ω 1 ∂k ∂ω
= α µT S 2 − ρβω 2 + 2(1 − F1 )ρσ ω 2
Dt k ω ∂ xi ∂ xi

and the turbulent viscosity is calculated from,

k
υt = F2 (5)
ω

β ∗ and σ ω 2 are model constants, F1 and F2 are blending functions controlling the
In these equations α , β ,
near wall behavior of the model and S is a measure of the mean rate of strain. In this work the mean rate of strain is
approximated by the magnitude of the vorticity.

3. Reynolds stress Model


The Reynolds stress model22 solves transport equations for the individual Reynolds stress terms. In addition to
the six equations describing the components of the Reynolds stress tensor a further equation modeling the transport
of the turbulent dissipation is also required. Full details of the Reynolds stress model employed in the current work
can be found in Reference 22.

3
American Institute of Aeronautics and Astronautics
By modeling the non-isotropic nature of the turbulence in a more rigorous manner than one-equation and two-
equation models employing the isotropic Boussinesq assumption, the Reynolds stress model offers the potential to
provide improved predictions for flows involving rotation and rapid changes in strain rate.

C. Discretization and Boundary Conditions


The governing equations were discretized using a pressure correction approach based upon a second-order
variant of the SIMPLE algorithm. Further details of this approach can be found in Reference 22.
The computations were performed for a single SBVG device using a structured multi-block grid. In order to
control the cost of he calculations the computational domain was limited in extent. In the streamwise direction the
domain extended 125mm forward of the device leading edge to 750mm aft of the device trailing edge. The
simulated domain was 300mm wide and 200mm high. Structured multi-block grids were generated around the
device. Grid points were distributed to provide adequate resolution of the solution gradie nts. In the wall normal
direction the initial spacing was chosen to ensure that the initial y+ was less than one. Clustering of grid points in
the normal direction was also used to provide appropriate resolution of the vortex developed by the device. In the
axial and lateral directions grid points were clustered to resolve the expected high gradients in the vicinity of the
device, a view of a typical surface grid is shown in Fig. 1. For computations in which an effort was made to resolve
the boundary layer developed on the SBVG device the local wall normal points distribution was chosen to ensure y+
was less than five.

Figure 1. Visualization of a typical surface grid

This approach to the grid generation resulted in grids containing between 300,00 points and 1,600,000 points for a
single device. This should be contrasted with the calculations reported by Allan16,17 who used 5,000,000 cells to
mesh the domain from the boundary layer trip (2250mm upstream of the device).
No slip boundary conditions were applied on the solid surfaces representing the flat plate and the SBVG device.
Upstream of the device the pressure, velocity and turbulence profiles were specified using previously computed
results for a fully developed two-dimensional boundary layer. On the remaining faces outflow conditions were
imposed. At these boundaries pressure was fixed and the remaining quantities extrapolated from the interior of the
domain.

III. Results and Discussion

A. Model Details
The Computations were performed for rectangular vane type vortex generators attached to a flat plate in a zero
pressure gradient with a freestream velocity of 34 m/s. The vanes were mounted 2.25m downstream of the plate
leading edge. The boundary layer had a thickness of 35mm at this location. The vanes had an aspect ratio of 5 and a
height equal to 20 % of the boundary layer thickness. The Reynolds Number was 81000 based on the length of the
vortex generator. The geometry and freestream conditions correspond with the experimental tests reported by Lin 11 .

4
American Institute of Aeronautics and Astronautics
B. Ge ometric Fidelity 0.50
Initially calculations were performed to quantify the
Lin
level of geometric detail required to perform successful 0.40

Circulation (m2/s)
simulations. Grids were generated around two 2D
configurations. The first configuration corresponds 0.30
exactly to the vortex generator tested in the wind tunnel 3D
experiments. It has a height of 7 mm (20% of the local 0.20
boundary layer height) an aspect ratio of 5 and a finite
thickness; the second configuration has the same 0.10
planform but has zero thickness.
The variation of circulation, Γ, with distance 0.00
downstream of the vortex generator is shown in Fig. 2. 0 20 40 60 80 100 120
Correlation between the two geometric models is x/h
initially good and agreement with the measured data is
acceptable. Further downstream, beyond about 40 Figure 2. Comparison of computed and measured
device heights, the computed data no longer correlate. circulation for infinitesimally thin (2D) and finite
The results obtained for the zero thickness SBVG (2D) thickness SBVG devices
appear to be in much better agreement with the
experimental data than those obtained for the finite
12000
thickness device (3D).
The maximum computed value of cross flow 10000
Peak Vorticity (1/s)

vorticity is plotted against distance downstream of the


vortex generator device in Fig. 3. Substantial 8000
differences are observed between the computed and 6000
measured values immediately aft of the SBVG devices.
Further downstream correlation between the computed 4000
and measured data improves significantly. The finite
2000
thickness vortex generator produces a larger peak
vorticity than that of the zero thickness device, but 0
within a few device heights the two models produce 0 20 40 60 80 100 120
results that are in close agreement.
x/h
The interaction between the vortex and wall shear
layer results in a decay in vortex strength. This decay
Figure 3. Comparison of computed and measured
can be observed through the evolution of the vortex
peak streamwise vorticity for infinitesimally thin
decay parameter ln(Γ/Γ x=h), Fig. 4. The computed rate
(2D) and finite thickness SBVG devices
of decay is more rapid than observed in the experiment
suggesting tha t there are sources of diffusion in the
numerical computations in addition to those that have a 0.00
physical basis. This pathology can be observed in Fig. 5
in which computed and measured contours of cross flow
Vortex Decay

velocity are compared. -0.25


The computed and measured vortex trajectories are
shown in Fig. 6 (vertical and lateral components
respectively). The vortex trajectory is generally well
predicted, but the results from the finite thickness model -0.50
are generally in much better agreement with the
experimental measurements than those from the zero
thickness model. Generally both quantitative and -0.75
qualitative agreement between the current computations 0 5 10 15 20 25 30
and experiment is comparable with that obtained by
other authors; see for example Lin 11 . The qualitative x/h

Figure 4. Comparison of computed and measured


vortex decay for infinitesimally thin (2D) and finite
thickness SBVG devices

5
American Institute of Aeronautics and Astronautics
Fig. 5 Comparison of Computed and Measured Cross Flow Velocity Contours

6
American Institute of Aeronautics and Astronautics
behavior corresponds closely with that observed
experimentally, but there are gross quantitative errors, 4.00
particularly in the prediction of peak vorticity,
circulation and vortex decay.
The computed data suggest that the geometric detail 2.00
of the SBVG device is only important in the initial

z/h
stages of vortex development. This is clearly evident
from comparisons of the peak vorticity and circulation
(Fig. 2 and Fig. 3 respectively). As the vanes share the 0.00
same basic planform and toe angle there is no reason to
expect the bound circulation of the vanes to be
significantly different. However, the initial evolution of
the peak vorticity is sensitive to the detail of the vortex -2.00
generator because of the importance of geometry in 0 10 20 30 40 50
determining the in itial distribution of vorticity trailed x/h
behind the device. At later vortex ages, following roll up
of the initial filaments into a discrete vortex, this
sensitivity is lost and the vortices exhibit similar (a) Lateral Position
characteristics.
These observations suggest that computations 2.00
performed to gain a detailed understanding of the fluid
dynamic aspects of SBVG devices should include a 1.50
detailed geometric representation of the vane. For
design studies, in which only the interaction of the
y/h

vortex with flow structures at later wake ages is 1.00 Lin


important, significant reductions in computational 2D
expense may be realized with little reduction in 0.50 3D
accuracy by employing a simplified two-dimensional
Measured
geometric representation.
0.00
C. Grid Sensitivity 0 10 20 30 40 50
Great care was taken when constructing the x/h
computational grids used in the present study to ensure
that the solutions could be considered to be grid (b) Vertical Position
converged. The initial distribution of points normal to Figure 6. Comparison of computed and measured
the plate surface provided grid independent solutions for vortex trajectory
the corresponding two -dimensional boundary layer.
This distribution was modified though point addition to accommodate the SBVG device and the presence of the
vortex. In order to reduce computational cost initial calculations were performed with a relatively coarse distribution
of points in the direction normal to the vane surface.
The boundary layer developed on the surface of the SBVG was not captured using this meshing strategy. The
grid was refined in the direction normal to the vane surface until grid independent solutions were obtained. This
exercise resulted in the number of cells in the grid increasing from 1.1 million to 1.6 million. The sensitivity of the
results to the resolution of the vane boundary layer is presented in Fig. 7-10. Generally the differences between the
data are more significant than the differences between geometric models employing finite and zero thicknesses.
Fig. 7 and 8 show comparisons of the circulation and peak streamwise vorticity respectively. Comparing the
initial development of the circulation there appears to be little discernable difference between the two calculations,
although at later vortex ages the computation performed without resolving the vane boundary layer appears to be in
better agreement with the experimental data. The peak circulation immediately downstream of the SBVG device is
significantly lower in the calculation which fails to resolve the vane boundary layer. Comparisons of the vortex
decay parameter ln(G/Gx=h) show no significant difference when the results with and without a vane boundary layer
grid are compared, see Fig. 9. The path of the vortex core appears to be better predicted when the device boundary
layer is resolved, Fig. 10.
Based upon the present computations it appears that resolution of the SBVG boundary layer is important in
understanding the development of the generated vortex and its path. This requirement creates a considerable
additional computational requirement for the simulation of SBVG devices. It should be noted that in the present

7
American Institute of Aeronautics and Astronautics
4.00
0.50
Lin
Circulation (m2/s)

0.40 No Grid
With Grid 2.00
0.30 Measured

z/h
0.20
0.00
0.10

0.00
-2.00
0 20 40 60 80 100 120
0 10 20 30 40 50
x/h
x/h

Figure 7. Comparison of computed and measured


circulation with and without a boundary layer grid (a) Lateral Position
on the SBVG device
2.00
12000
10000 1.50
Peak Vorticity (1/s)

8000
y/h

1.00
6000
4000
0.50
2000
0 0.00
0 20 40 60 80 100 120 0 10 20 30 40 50
x/h x/h

Figure 8. Comparison of computed and measured (b) Vertical Position


peak streamwise vorticity with and without a
boundary layer grid on the SBVG device Figure 10. Comparison of computed and measured
vortex trajectory
0.00
-0.10
Vortex Decay

-0.20
-0.30
-0.40
-0.50
-0.60
-0.70
0 5 10 15 20 25 30
x/h

Figure 9. Comparison of computed and measured


vortex decay with an d without a boundary layer grid
on the SBVG device

8
American Institute of Aeronautics and Astronautics
computations the boundary layer on the SBVG device is
assumed to be fully turbulent. It is expected that this 0.50
simplification is a robust assumption because of the S-A
highly turbulent nature of the boundary layer in which 0.40

Circulation (m2/s)
the device operates. Based upon the current results it is S-A (S-
felt that it is unlikely that the nature of the boundary 0.30 V)
layer on the SBVG device will have a significant k-w
influence on the flow in the control interaction region. 0.20

D. Turbulence Modeling 0.10


The initial stages of the study have shown that in
order to produce reliable numerical simulations of 0.00
vortex generators it is necessary to both include the 0 20 40 60 80 100 120
thickness of the vortex generator in the geometric model x/h
and resolve the boundary layers developed on the
surfaces of the device itself. These requirements impose
Figure 11. Variation of circulation with distance
a severe computational burden; indeed the current grid,
downstream
which simulates a single device, contains 1.6 million
cells.
In the final part of the study the sensitivity of the 12000
computed data to the turbulence closure was 10000
investigated. Two popular one- and two-equation
Peak Vorticity

turb ulence models were investigated; the Spalart- 8000


Allmaras (SA) and k-ω SST (SST) models together
6000
with a seven equation Reynolds stress model RSM). In
addition to the basic formulation of the Spalart-Allmaras 4000
model a strain-vorticity (SA-SV) formulation was also
studied. 2000
Computed variation of circulation down -stream of 0
the device shows little sensitivity to the turbulence
0 20 40 60 80 100 120
model, Fig. 11. Although all of the models produced
initial circulation values in good agreement with the x/h
experimental data the subsequent development was
monotonic. The models therefore fail to predict the Fig. 12: Variation of peak streamwise vorticity with
initial behaviour observed in the measured data which distance downstream
clearly shows a rapid rise to the peak circulation
followed by the computed monotonic behaviour.
The Reynolds stress model provides predictions of peak streamwise vorticity, Fig. 12, that correspond more
closely with the experimental data than the predictions made using t he one- and two- equation models. The original
Spalart-Allmaras model performs worst of the models considered. The use of the strain-vorticity formulation of the
model improves the prediction of the Spalart-Allmaras model producing results that are comparable with those
obtained using the k-? SST model.
In Fig. 13 the horizontal and vertical trajectory of the vortex core are plotted against distance downstream of the
vortex generator. All of the models investigated compare favourably with the experimental data close to the SBVG
device. Further downstream the discrepancies between the models become larger.
The lateral trajectory of the vortices predicted using the original and strain-vorticity formulations of the Spalart-
Allmaras model remain in reasonable agreement over the whole computational domain, but significant differences
are observed in the predicted vertical trajectory. With the original formulation the vortex remains at almost constant
height, behaviour that can be observed in inviscid computations, and is in poor agreement with the measured data.
The k-ω SST model remains in reasonable agreement with the experimental data over the whole domain, while
correlation between the Reynolds stress model and the measured data shows significant improvements comp ared
with the other models.
A measure of the growth of the vortex core radius is shown in Fig. 14. Computations performed using the one-
and two-equation turbulence models are in poor agreement with the experimental measurements. The core of the
vortex predicted by the original Spalart-Allmaras model grows rapidly in size leading to very poor comparison with
the experimental data. The use of the strain -vorticity modification produces results that compare favourably with
9
American Institute of Aeronautics and Astronautics
those obtained using the k-ω SST model, but the core
radius remains much larger than that measured in the 4.00
experiment. Of the models tested only the Reynolds
stress model provides reliable predictions of the core
radius, although at later vortex ages the core is much 2.00
smaller than that observed in the experiment.

z/h
In order to understand this behaviour it is necessary
to consider the vorticity field in more detail. Computed 0.00
contours of streamwise vorticity obtained using the k-ω 0 10 20 30 40 50
SST and Reynolds stress model are compared with
experimental meas urements at several stations -2.00
downstream of the vortex generator device in Fig. 15. x/h
Initially both models appear to provide good qualitative
and quantitative agreement with the measured data, but
(a) Lateral Position
at later wake ages (beyond x/h = 3) there is evidence of
rapid diffusion in the k-ω SST computation, this
behaviour is also evident in both the simulation results 2.00
obtained using the Spalart-Allmaras model.
The principle reason for the rapid diffusion of the 1.50
vortex core lies in the failure of the simpler turbulence y/h
models to provide reliable estimates of the turbulent 1.00
viscosity (and consequently the Reynolds stresses). This
is clear from Fig. 15 and Fig. 16. The differences 0.50
between the computed turbulent viscosity obtained using
the original Spalart-Allmaras model and the k-ω SST 0.00
model are shown in Fig. 16 at several stations
downstream of the vortex generator device. The data 0 10 20 30 40 50
clearly explain the discrepancy in core radius between x/h
the two models. The turbulent viscosity predicted by the
Spalart-Allmaras model is much larger than that (b) Vertical Position
obtained from the k-ω SST computation leading to the
observed rapid diffusion in the vortex core. Figure 13. Comparison of computed and
Fig. 17 shows computed contours of vorticity and measured vortex trajectory
turbulent viscosity 50 device heights downstream of the
vortex generator device. The data presented in this
2.00
figure provide some additional insight into the
mechanisms responsible for the rapid dissipation of the
vortex core. For the original formulation of the one- 1.50
equation model excessive turbulent viscosity is
generated at the vortex core. This arises because the
R/h

model employs an isotropic Boussinesq assumption and 1.00


is therefore not suited for flows involving complex
strain. The reliance of the model on vorticity as an 0.50
indicator when determining the significance of the near
wall damping is also problematic. The relationship
between near wall damping and vorticity was 0.00
determined empirically using simple data from simple 0 10 20 30 40 50 60
wall bounded flows and is therefore unsuitable for flows x/h
involving discrete vortical structures. Fig. 14 Variation of vortex core radius with distance
The inclusion of the strain -vorticity modification
downstream
helps to alleviate this problem resulting in less
turbulence production allowing the vortex structure to be
maintained to later vortex ages.
Compared to the Reynolds stress model the k-ω SST
under predicts turbulent viscosity in the vortex

10
American Institute of Aeronautics and Astronautics
Reynolds Stress
k-? SST Model Experiment

Fig. 15 Variation of streamwise vorticity with distance downstream of the SBVG device
Spalart-Allmaras k-? SST
Turbulent Viscosity Streamwise Vorticity
3h
SA

5h

SA-SV
10h

17h
k-? SST

50h

RSM
109h

Fig. 16 Computed contours of turbulent viscosity Fig. 17 Calculated contours of turbulent viscosity
downstream of the device and streamwise vorticity (x=50h)

core leading to improved vortex capturing properties. The improvements obtained in the Spalart -Allmaras and the
k-ω SST calculations are due to the way in which both models effectively limit the influence of strain or vorticity in
the generation of turbulence. In the Spalart-Allmaras model this is done by blending vorticity with strain to reduce
the influence of the near wall damping terms, comparing with the data obtained using the Reynolds stress model this

11
American Institute of Aeronautics and Astronautics
appears to be a good strategy. In the case of the k-ω SST model the influence of strain is controlled in both the
turbulent production and in the formulation relating turbulent viscosity to the turbulence variables. It appears that
this strategy results in an under prediction of viscosity, resulting in a stronger vortex than observed in the Reynolds
stress model. Of the models examined only the Reynolds stress model appears to be capable of resolving the
complex interaction between the generated vortex and the boundary la yer of the flat plate. The improvements
obtained with the Reynolds stress model are attributed to improved representation of complex strain associated with
vortex. While the simpler models are attractive in terms of computational expense their use of an isotropic
Boussinesq assumption renders them incapable of resolving the flow associated with discrete vortical structures
adequately.

IV. Conclusion
A computational study of the aerodynamics of vane type micro-vortex generators has been performed. Solutions
of the incompressible Reynolds-Averaged Navier-Stokes equations have been obtained for a rectangular vane type
micro -vortex generator mounted on a flat plate at a Reynolds number of 81,000 based upon vortex generator length.
The study examined uncertainty in the computational simulation of SBVG devices related to geometric fidelity, grid
generation and turbulence modeling.
It was shown that the geometric fidelity of the model used in the simulation was important in understanding the
detailed flow induced by the SBVG device at early wake ages. At later wake ages results obtained for a full model
and a simplified model of the device were in good agreement suggesting that for flow control studies the simpler
model could be employed without loss of accuracy.
Studies of the grid generation requirements for SBVG devices reveal that it is necessary to adequately resolve
the device boundary layer. Computations performed with grids that did not resolve this feature of the flow generally
compared poorly to measured data.
The choice of turbulence model was shown to be important in resolving the observed behavior of the vortex
trajectory and vortex diffusion. Simple models performed poorly. This was attributed to two main deficiencies;
firstly the use of an isotropic Bousinesq assumption to relate the Reynolds stresses to a turbulent viscosity and the
use of functions of vorticity in the production of turbulence. Improved predictions were obtained using a strain-
vorticity formulation of the Spalart -Allmaras model and the k-ω SST models both of which attempt to correct the
second of these deficiencies. Further improvements were obtained using the Reynolds stress model which accounts
for complex strain in a more rigorous fashion.

References
1
Pearcey, H. H., “Shock-induced separation and its prevention by design and boundary layer control,” Boundary layer and
flow control, Vol. 2, Pergamon Press, 1961, pp. 1170-1344
2
Delery, J.M., ''Shock wave/turbulent boundary layer interaction and its control,” Progress in Aerospace Sciences, Vol. 22,
No.4, 1985, pp209-280.
3
Delery, J.M., “Shock phenomena in high speed aerodynamics: still a source of major concern”, Aeronautical J., Vol. 103,
No.1019, 1999, pp19-34.
4
Freestone, M., “Vortex generators for control of shock-induced separation,” Engineering Sciences Data Unit Technical
Memorandum 93-024, 1993.
5
Rao, D.M. and Kariya, T.T., “Boundary layer submerged vortex generators for separation control – an exploratory study ,”,
AIAA 88-3546CP, 1988
6
Ashill, P.R., Fulker, J.L. and Hackett, K.C., “Research at DERA on sub-boundary layer vortex generators (SBVGs),” AIAA
39th Aerospace Sciences Meeting and exhibit, AIAA 2001-0887, AIAA, Reno, Nevada, 2001
7
Ashill, P.R., Fulker, J.L. and Hackett, K.C., “Studies of flows induced by sub-boundary layer vortex generators (SBVGs),”
AIAA 40th Aerospace Sciences Meeting and exhibit, AIAA 2002-0968, AIAA, Reno, Nevada, 2002
8
Lin, J.C., Howard, F.G. and Selby G.V., “Small submerged vortex generators for turbulent separation control,” J. Spacecraft
and Rockest, Vol.27 (5), 1990, pp. 503-507
9
Lin, J.C., Robinson, S.K., McGhee, R.J. and Valarego, W.O., “Separation control on high-lift airfoils via micro-vortex
generators,” J. Aircraft, Vol. 31 (6), 1994, pp. 1317-1323
10
Lin, J.C., “Control of turbulent boundary layer separation using micro-vortex generators,” AIAA 99-3404, 1999
11
Yao, C-S and Lin, J.C., “Flow-field measurement of device induced embedded streamwise vortex on flat plate,” AIAA 1st
Flow Control Conference, AIAA 2002-3162, AIAA, St. Louis, Missouri, 2002
12
Wendt B.J., “Measurements and Modelling of Flow Structure in the Wake of a Low Profile “Wishbone” Vortex
Generator,” AIAA 34 th Aerospace Sciences Meeting and exhibit, AIAA 96-0807, AIAA, Reno, Nevada, 1996
13
Bender E.E., Anderson, B.H. and Yagle, P.J., “Vortex generator modeling fpr Navier-Stokes codes”, ASME Paper
FEDSM88-6919, 1999
14
May N.E., “A New Vortex Generator Model For Use in Complex Configuration CFD Solvers,” AIAA 2001-2434, 2001

12
American Institute of Aeronautics and Astronautics
15
Bray, B.G., “Numerical Simulations of Vortex Generator Vanes and Jets on a Flat Plate,” AIAA 1st Flow Control
Conference, AIAA 2002-3160, AIAA, St. Louis, Missouri, 2002
16
Allan, B.G., “Numerical Simulations of Vortex Generator Vanes and Jets on a Flat Plate,” AIAA 1st Flow Control
Conference, AIAA 2002-3160, AIAA, St. Louis, Missouri, 2002
17
Allan, B.G., “Numerical Simulations of Vortex Generator Vanes and Jets on a Flat Plate,” AIAA 1st Flow Control
Conference, AIAA 2002-3160, AIAA, St. Louis, Missouri, 2002
18
Wik, B.G., “Numerical Simulations of Vortex Generator Vanes and Jets on a Flat Plate,” AIAA 1st Flow Control
Conference, AIAA 2002-3160, AIAA, St. Louis, Missouri, 2002
19
Spalart, P. and Allmaras, S., “A one-equation turbulence model for aerodynamic flows,” AIAA 30th Aerospace Sciences
Meeting and exhibit, AIAA 92-0439, AIAA, Reno, Nevada, 1992
20
Dacles-Mariani, J. Zilliac, G.G., Chow, J.S. and Bradshaw, P., “Numerical/Experimental Study of a Wingtip Vortex in the
Near Field,” AIAA Journal, Vol. 33 (9), 1995, pp. 1561-1568
21
Menter, F.R., Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications.
AIAA Journal, 32(8):1598-1605, August 1994
22
Anonymous, “ Fluent User Guide,” Fluent, 2003.

13
American Institute of Aeronautics and Astronautics

You might also like