You are on page 1of 6

RANS Predictions of Junction Flow with

Localized Suction
1
Shakeel Ahmed, 2Dr A. Malik, 3Dr Khalid Parvez
1,2,3
Department of Aeronautics and Astronautics
Institute of Space Technology
Islamabad, Pakistan

Abstract—Horseshoe vortex is a prominent feature in junction Work of Paciorri et al. [3] demonstrated that CFD can
flows which not only changes the local flow patterns but is also predict the junction flow with sufficient qualitative accuracy
responsible for aerodynamic loses in the region. In the past many and hence CFD may be used to understand the physics of the
investigations using wind tunnels and CFD have been made to junction vortex. Similar results were also obtained by
study the junction flow. These studies have shown that RANS
J.L.Coombs et al. [4] where they presented assessment of
prediction of the junction flow demonstrate an acceptable
qualitative match with the experimental results. The present various turbulence models for junction flows and showed that
investigations extend the CFD investigations to RANS predictions RANS based turbulence models are able to give acceptable
of the junction flow with boundary layer control using localized results in comparison to the experimental data. Similarly,
suction. Localized suction is an active technique that has proved A.M.Levchenya et al. [5] in their paper have simulated the
successful, in the past through wind tunnel measurements, in presence of horseshoe vortex in the junction region using two
elimination of the junction horseshoe vortex. The present versions of RANS based turbulence models (Mentor SST
investigations used these experimental results to evaluate the model). They found out that these models produced
RANS predictions from the current study. During the course of qualitatively correct results when compared the quantitative
this study, six cases of junction flows with and without localized
data with the experimental results. Despite so much research in
suction were numerically simulated and compared with the wind
tunnel measurements. For no suction case, CFD and experiment understanding the behavior of junction flows and horseshoe
have been found in total agreement where location and size of vortices, surprisingly attempts to understand horseshoe vortex
vortex core formation was observed to be similar. When suction behavior after application of localized suction using CFD is
was applied, CFD predicted its effect by showing reduction in still relatively a less explored area.
size of the vortex core similar to the results observed during the The present study investigated the CFD / RANS capability
wind tunnel experiment. RANS simulations have predicted fairly in prediction of junction flows with localized suction. Many
accurate results in almost all the cases in comparison with the researchers in the past have investigated various active and
experimental data. passive control techniques to eliminate or reduce the strength
Index Terms — Junction flow, Horseshoe vortex, RANS and size of the vortex. Among those attempts uses of localized
predictions, Localized suction, Flow control suction, by Phillips and Barberis, have been successful
attempts in removing or decreasing the strength of the
I. INTRODUCTION horseshoe vortex. In this paper wind tunnel experiments
Junction flows are observed whenever a boundary layer performed by Barberis et al. [6] have been numerically
growing on a flat plate like surface is encountered with an simulated to evaluate RANS capability in prediction of
obstruction in the surface. Therefore these types of flows are boundary layer control by localized suction just ahead of the
encountered in numerous engineering applications like aircraft model. The commercial CFD package Fluent [7] has been
wing-fuselage attachments, bridge support pillars, car side utilized to perform the simulations. The CFD predictions from
view mirrors, sail and hull junction on submarines etc. the current study will be bench marked against these studies. A
Horseshoe vortex is a prominent feature in these junction flows comparison of the results obtained by the turbulence model
which not only changes flow but is also responsible for with the wind tunnel experimental data (from Barberis) have
aerodynamic loses in the region [1]. Hence, the junction flow been made to determine the capability of this RANS based
has been the subject of many investigations. turbulence model for such cases. Based on the findings from
This kind of vortex starts as a two dimensional flow this investigation CFD may be used in the design of localized
separations that transforms in to three dimensional complex suction setups for various applications.
flows due to interaction between the approaching boundary II. SELECTION OF TURBULENCE MODEL
layers and the pressure field produced by the wing. This
horseshoe vortex originates from a stagnation point ahead of In this regard, RANS based eddy viscosity turbulence
the wing and wraps around the complete wing [2]. model i.e one equation Spalart Allmaras [8] have been
Understanding the physics of horseshoe vortex is of prime employed to study its utility. In all turbulent flows velocity
importance in order to improve the aerodynamic performance fields fluctuate rapidly both in space and time.
of junction flows. Computationally it is very expensive to directly simulate both
the small scale and high frequency fluctuations for any useful
engineering problem. There are two methods to eliminate the ̃
need to resolve these small scales and high frequencies; ( ̃) ( ̃ ) [ {( ̃) }
̃
namely, Reynolds Averaging and Filtering. The present ̃
investigations are only concerned with the Reynolds ( ) ] ̃ (3)
Averaging. In this method all flow variables are divided into a
mean component and a rapidly fluctuating component and then
Where, is the production of turbulent viscosity and is
all equations are time averaged to remove the rapidly
the destruction of turbulent viscosity that occurs in the near-
fluctuating components. For the continuity equation the new
wall region due to wall blocking and viscous damping. ̃ and
equation is identical to the original equation, except that the
are constants and ̃ is the molecular kinematic viscosity. ̃
transported variables now represent the mean flow quantities.
is a user-defined source term. More details about the turbulence
In the Navier-Stokes equation; however, new terms appear
model can be found in the FLUENT 6.3 User’s guide [10].
which involve mean values of products of rapidly varying
quantities. These new terms are known as the Reynolds IV. CFD MODEL DESCRIPTION, COMPUTATIONAL DOMAIN
Stresses, and solution of the Reynolds Averaged Navier-Stokes AND BOUNDARY CONDITIONS
(RANS) equation initially involves the construction of suitable
models to represent these Reynolds Stresses. These models Wind Tunnel Model Description
include various one equation and two equation models, like Barberis et al. [3] were able to significantly decrease the
Spalart Allmaras, K-omega, K-epsilon etc. magnitude and strength of the horseshoe vortex using an active
A low Reynolds number one equation Spalart-Allmaras control scheme. They employed localized suction just upstream
turbulence model has been selected for numerically simulating of the wing leading edge in the subsonic wind tunnel to
the junction flow with localized suction in order to study the eliminate the boundary layer which essentially removes the
RANS capability for prediction of such flows. This has been spanwise vorticity. They employed suction holes at different
selected due to the known ability of the turbulence model to locations in the symmetric plane in front of the leading edge
predict the flow within the boundary layer right down into the and concluded that applying suction in this manner
laminar sub-layer. This turbulence model is widely used for considerably reduces the horseshoe vortex strength. The wing
predictions of wall-bounded flows and it is believed to perform model was shaped like cylindrical obstacle of 360 mm
well for boundary layers subjected to adverse pressure diameter at the leading edge ending in a shaped trailing edge
gradients [9]. This aspect of the turbulence model is important [3]. The model had a total chord length of 1090 mm and height
in the planned three dimensional half wing applications, since of 940 mm. The aerofoil of the wing model as utilized in the
this study involves junction flow separation which originates reference paper [3] is shown in Fig. 1. Half wing geometry
due to the adverse pressure gradient effects upstream of the with the same dimensions as mentioned in the experiment has
wing leading edge. been modeled for numerical simulations as shown in Fig. 2.
III. NUMERICAL METHODOLOGY
Governing equations of the fluid flow are the full set of
Navier Stokes equations together with continuity equation.
These equations are subjected to the laws of conservation of
mass, momentum and energy. In Reynolds averaging, the exact
Navier Stokes equations are decomposed into mean and
fluctuating components. However, doing so produces
additional terms in the N-S equations to represent the effects of
turbulence. The Reynolds Averaged Navier Stokes (RANS)
equations are represented in Eq. 1 and Eq. 2.
Fig. 1. Definition of Obstacle (dimensions in mm) as appeared in [3]
( )   

( ) ( ) [ (

)] ( ̅̅̅̅̅)

Where, the additional term ̅̅̅̅̅ is known as Reynolds


stresses and must be modeled for solving Eq. 2 [10]. The
simulations in this work have been carried out using ANSYS
Fluent CFD code, one equation Spalart Allmaras Turbulence
Model. Transport equation for Spalart Allmaras model is given
in Eq. 3. Fig. 2. Aerofoil of the Wing Model for RANS Simulation
Computational Domain note that for case with no suction, the face was selected as no
In order to reduce the computational cost, taking benefit of slip wall and for cases with suction, it was selected as velocity
the symmetric characteristics of the wing model, only half inlet and given the values with negative sign (to simulate
wing model has been used for numerical simulations. outflow).
Leading edge of the wing model was positioned at a
TABLE I. BOUNDARY CONDITION ON ALL THE FACES
distance of 3000 mm from the inlet on the center line of the
section to get a symmetric flow for the experiment. Face Name
Boundary
Value Remarks
Computational domain has been selected in a way to get the Condition
same flow conditions at the wing leading edge. Distance of the Vel Inlet Velocity Inlet 50 m/s For all cases
inlet has been kept 3000 mm from the leading edge of the wing “0” gauge
Pressure outlet Pressure Outlet For all cases
pressure
model; however, in order to avoid back flow at the outlet, it has
Outer Wall Wall No Slip For all cases
been extended 6000 mm downstream of the wing trailing edge.
Bottom Wall Wall No Slip For all cases
Height and width of the computational domain has been
selected same as that of the height and width of the test section Top Wall Wall No Slip For all cases
of the wind tunnel. Top view of the computational domain to Wing LE Wall No Slip For all cases
get clarity is shown in Fig. 3; whereas, three dimensional view Wing TE Wall No Slip For all cases
of computational domain is shown in Fig. 4. Wing Top Wall No Slip For all cases
Symmetry Symmetry Symmetry For all cases
For case-I
Wall No Slip
(no suction)
-5 m/s For case-II

Suction -10 m/s For case-III


Velocity Inlet
(z component -20 m/s For case-IV
of velocity) -40 m/s For case-V
-55 m/s For case-VI

V. GRID INDEPENDENCE
Fig. 3. Two Dimensional View of Computational Domain In order to carry out grid independence analysis, one of the
widely used methods i.e. the standard convergence analysis
The three dimensional view of the computational domain approach has been used. In this approach, a problem with lesser
along with the suction hole (50 x 82 mm2) at distance of 102 number of grid points is first solved and some parameters like
mm from the wing leading edge is shown in Fig. 4. velocity, pressure difference are recorded. Same parameters are
recorded with increasing the number of grid points until the
difference between results obtained in two consecutive grid
sizes become negligible. Finally, the 2nd last grid size is
selected for further study of the problem. For the purpose of
this study, three grids with different coarseness were generated
and results were simulated for the no suction case, only. The
grid with fine mesh settings was finally selected for further
analysis; the finally selected mesh has total 1485619 nodes.
Pictorial view of the selected 3D grid of the half wing model
within the computational domain is shown in Fig. 5.

Fig. 4. Computational Domain with Named Faces

Boundary Conditions
All the faces encompassing the computational domain were
given names for identification as can be clearly seen in Fig. 4.
A total of 6 different cases were numerically simulated for
comparison with the experimental data available in literature.
These 6 cases included the case with no suction, case with
suction rate of 5 m/s, 10 m/s, 20 m/s, 40 m/s and 55 m/s.
Boundary conditions on all the faces (except face named as
“suction”) were same for all the 6 cases. Summary of the Fig. 5. Computational Domain with Named Faces
boundary conditions are given in Table. 1. It is important to
VI. RESULTS AND DISCUSSIONS in size of the vortex core has been observed in the results
Results obtained for all the six cases have been compared obtained with RANS simulation as well. This phenomenon can
with the experimental results attained by Barberis et al. [3]. be seen in Fig. 9. In this case, the experimental results show
The parameters utilized for comparison purposes between that the vortex core has formed at a non-dimensional distance
experiment and numerical simulations are the streamlines of and height of about 0.14 and 0.025 respectively from wing
the flow and velocity vectors obtained on the symmetric face leading edge and bottom floor during wind tunnel experiment.
ahead of the wing leading edge. Symmetric plane is the origin Location of the vortex core with the numerical simulation is
of the junction horseshoe vortex and therefore streamlines and also same as that of experiment with negligible differences in
velocity vectors in this plane are a meaningfully parameters in streamline patterns. Hence, RANS has correctly predicted the
the junction flow investigations. Close observation of the behavior of flow in the symmetric plane with localized suction
streamlines and velocity vectors ahead of the wing leading at a rate of 5 m/s.
edge clearly shows the formation of the vortex which is in fact
the 2D representation of the horseshoe vortex responsible for
producing the aerodynamic drag. Results obtained by RANS
simulations would be compared with the experimental data for
all the six cases, one by one in coming paragraphs.
Case-I with No Suction
Fig. 6 and Fig. 7 show flow patterns in the symmetry plane
for the no suction case. The experiment shows development of Fig. 8. Mean velocity vector and streamlines on plane of symmetry:
vortex with high return velocity in the vicinity of wall [3] as suction rate of 5 m/s [3]
shown in Fig. 6. Results obtained with the RANS simulations
for this case is shown in Fig. 7. It can be seen by comparison of
Fig. 6 and Fig. 7 that for no suction case numerical simulation
have predicted fairly accurate results. In both cases, the vortex
core has formed at a non-dimensional distance of about 0.24
and 0.035 from wing leading edge and bottom floor,
respectively. A strong agreement has been observed between
experiment and numerical simulation for vortex formation.
Fig. 9. Mean velocity vector and streamlines on plane of symmetry:
suction rate of 5 m/s with RANS Simulation

Case-III with Suction Rate of 10 m/s


In this case, suction was applied ahead of the wing through
the suction hole at a rate of 10 m/s. With the increase in suction
rate, the vortex size started to reduce and also the vortex core
started to move closer towards the leading edge. The vortex
Fig. 6. Mean velocity vector and streamlines on plane of symmetry for No core has formed at a non-dimensional distance of about 0.12
Suction case [3] and 0.020 from wing leading edge and bottom floor
respectively during wind tunnel experiment. Fig. 10 shows
flow patterns as seen in the wind tunnel by Barberis et al. [3];
whereas, result obtained by RANS simulations has been
depicted in Fig. 11. The comparison show that RANS
simulation has also predicted the phenomenon of reduction in
size of vortex with increase in suction rate as the location of
vortex formation is same as that of experiment with
insignificant differences in streamline and velocity vector
patterns.
Fig. 7. Mean velocity vector and streamlines on plane of symmetry for No
suction case with RANS Simulation

Case-II with Suction Rate of 5 m/s


When suction was applied ahead of the wing through the
suction hole at a rate of 5 m/s, the vortex formed at the junction
was observed to reduce in size as shown in Fig. 8. It is also
noted that the vortex core has moved closer to the wing leading
Fig. 10. Mean velocity vector and streamlines on plane of symmetry:
edge and floor of the test section. Similar behavior of reduction suction rate of 10 m/s [3]
Fig. 11. Mean velocity vector and streamlines on plane of symmetry: Fig. 14. Mean velocity vector and streamlines on plane of symmetry:
suction rate of 10 m/s with RANS Simulation suction rate of 40 m/s [3]

Case-IV with Suction Rate of 20 m/s


In this case, suction rate has further increased to 20 m/s.
The streamlines and velocity vectors showing the formation of
horseshoe vortex during the wind tunnel experiment is shown
in Fig. 12. Result obtained by RANS simulations with suction
rate of 20 m/s is shown in Fig. 13. As described earlier, with
increasing the suction rate the vortex core size is reducing and
it is moving closer towards the wing leading edge. Similar to
the first three cases, results obtained by experiment and Fig. 15. Mean velocity vector and streamlines on plane of symmetry:
simulations are the same in terms of vortex location and size suction rate of 40 m/s with RANS Simulation
for this case as well. The vortex core has formed at a non-
Case-VI with Suction Rate of 55 m/s
dimensional distance of about 0.095 and 0.018 from wing
leading edge and bottom floor respectively in both the In this last case, suction rate through the suction hole was
approaches. increased to 55 m/s. The vortex core in this case has formed at
a non-dimensional distance of about 0.05 and 0.010 from wing
leading edge and bottom floor respectively during experiment
as well as in the RANS simulation. Results from experiment is
shown in Fig. 16; whereas, results from numerical simulation is
shown in Fig. 17. The strong agreement between the wind
tunnel experimental and numerical simulation results is a
testimony of the capability of RANS based turbulence model in
prediction of junction flow behaviors with localized suction.
Fig. 12. Mean velocity vector and streamlines on plane of symmetry: As predicted earlier, the size of the vortex core which is in fact
suction rate of 20 m/s [3] two dimensional representation of the three dimensional
horseshoe vortex has reduced with each increase in the suction
rate.

Fig. 13. Mean velocity vector and streamlines on plane of symmetry:


suction rate of 20 m/s with RANS Simulation Fig. 16. Mean velocity vector and streamlines on plane of symmetry:
suction rate of 55 m/s [3]
Case-V with Suction Rate of 40 m/s
The suction rate in this case has been increased up to 40
m/s. The size of the vortex is observed to be reducing but now
at a slower pace as compared to the increase in suction rate.
Formation of vortex core at a non-dimensional distance of
about 0.07 and 0.015 from wing leading edge and bottom floor
respectively during experiment is shown in Fig. 14. Result
obtained by RANS simulations has the same vortex core
location and is shown in Fig. 15. Similar to all the previous
Fig. 17. Mean velocity vector and streamlines on plane of symmetry:
cases, results obtained by experiment and simulations are the suction rate of 55 m/s with RANS Simulation
same in terms of vortex location and size for this case as well.
VII. CONCLUSION wake flow” Experiments in Fluids 14, Springer and Verlag,
1993, pp. 366-378.
The primary objective of this study was to examine the
[2] D. B. Philips, J. M. Cimbala, A. L. Treaster, “Suppression of
capabilities of an eddy viscosity RANS based turbulence
Wing-Body Junction Vortex by Body Surface Suction” Journal
model i.e one equation Spalart Allmaras in predicting the of Aircraft, Vol. 29, No. 1, Jan-Feb 1992, pp. 118-122.
turbulent junction flows with and without localized suction in
[3] R. Paciorri, A. Bonfiglioli, A. Di Mascio, B. Favini, “RANS
the region ahead of the wing leading edge. It has been clearly Simulation of a Junction Flow” International Journal of
established through comparison of the simulation results with Computational Fluid Dynamics, Vol. 19, No. 2, February 2005,
the wind tunnel experimental data available in literature from pp. 179-189.
Barberis et al. [3], that Spalart Allmaras turbulent model can [4] J.L.Coombs et al, “Assessment of Turbulence Models for a
provide reasonable estimates of complex behavior of junction Wing-in-Junction Flow” 18th Australian Fluid Mechanics
flows with localized suction applied to reduce the horseshoe Conference, Launceston, Australia, 3-7 December, 2012.
vortex, due to its ability to capture the small vortices appearing [5] A.M.Levchenya, E.M.Smirnov, V.D.Goryachev, “RANS-based
in the boundary sub layers. It can also be concluded that CFD numerical simulation and visualization of horseshoe vortex
can be utilized to design experimental set up with localized system in the leading edge end wall region of a symmetric
suction in junction flow when wing is fixed at zero degree to body” International Journal of Heat and Fluid Flow, Volume 31,
the incoming flow. No. 6, December 2010, pp. 1107-1112.
The study can further be extended in future to large eddy [6] D. Barberis, P. Molton, T. Malaterra, “Control of 3D Turbulent
simulation (LES) of the turbulent junction flows with localized Boundary Layer Separation Caused by a Wing-Body Junction”
suction. Results obtained by LES can be used to compare with Experimental Thermal and Fluid Science, Vol. 16, 1998, pp. 54-
the results obtained with RANS simulation performed during 63.
this study and the experimental results available in literature. [7] Fluent-Commercial CFD software package based on finite
This will help to evaluate the optimized scheme for numerical Volume Method. Fluent Inc., Centerra Resource Park, 10
Cavendish court, Lebanon, NH 03766, USA.
simulations of complex flows encountering in real life
engineering problems. [8] Spalart P.R. & Allmaras, (1992), “A one equation turbulence
model for aerodynamic flows”, Reno, NV, 30th Aerospace
ACKNOWLEDGMENT Sciences Meeting and Exhibit, January 6-9, 1992, AIAA Reston
No. 92-0439.
Authors would like to acknowledge the all-out support
[9] Fluent Inc., (September 2006), “Tutorial 3-Modeling external
extended by Department of Aeronautics and Astronautics, compressible flow”, Fluent Inc., Centerra Resource Park, 10
Institute of Space Technology, Islamabad for carrying out the Cavendish Court, Lebanon, NH 03766, USA.
numerical simulations in its Modeling and Simulations Lab. [10] Fluent Inc., (September 2006), “FLUENT 6.3 User’s Guide”,
Fluent Inc., Centerra Resource Park, 10 Cavendish Court,
REFERENCES
Lebanon, NH 03766, USA.
[1] J.L. Fleming, R.L. Simpson, J.E. Cowling and W.J. Devenport, .
“An experimental study of a turbulent wing-body junction and

You might also like