VLM 3dpanel

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

2020 3rd International Conference on Computing, Mathematics and Engineering Technologies (iCoMET)

Comparison of Flow-Solvers: Linear Vortex Lattice Method and Higher-


Order Panel Method with Analytical and Wind Tunnel Data
Tahura Shahid1, Faiza Sajjad1, Muneeb Ahsan2, Syed Farhan1, Shuaib Salamat1, Messam Abbas1
1
Aerodynamics, AvRID, Pakistan Aeronautical Complex, Kamra, Pakistan.
2
Department of Aerodynamics, Air University, Kamra, Pakistan.
Tahura671@gmail.com, faiza.sajjad97@gmail.com, muneebahsan@live.com, syedfarhannaveed@outlook.com,
ssalamat@aerospace.pk, messam.naqvi@gmail.com

Abstract— Two induced drag analysis techniques, Vortex moments.


Lattice Method (VLM) and panel method are renowned for Vortex Lattice Method (VLM) is another potential flow-
inviscid aerodynamic computations and are widely used in the solving technique that specifically uses vortices to compute
aerospace industry and academia. To demonstrate the
applicability of potential flow theory and to establish an aerodynamic forces over a geometry by dividing its two-
extensive correlation of linearized, attached potential flow- dimensional projection into a finite number of quadrilateral
solver codes for estimating lift and induced drag, a generic panels. This method uses a potential flow equation modified
rectangular wing planform is analyzed. Due to a wide range of with Prandtl-Glauert correction to solve flows in subsonic
applicability in conceptual design, the two solvers are compared and supersonic regimes. Both these methods are used to
for accuracy, computational time and input controllability to calculate the aerodynamic loading on the surface of an
find an optimum solver that can predict inviscid aerodynamics
accurately and efficiently but with the least amount of time. aircraft of a given configuration, as well as for optimizing an
VLM-based code is founded upon the Laplace equation. It aerodynamic surface for a given pressure distribution [1].
approximates a three-dimensional wing into a two-dimensional
A. Panel Method Code
planform, making it apposite for moderate aspect ratio and
thin-airfoil aircraft. A modified VLM is used that takes a suction In panel code [2], the planform is first divided into
parameter, calculated analytically, as an input to capture three- networks such as upper wing surface, lower wing surface, etc.
dimensional leading-edge thrust and vortex lift effects. On the and these networks are further sub-divided into quadrilateral
contrary, the higher-order panel method takes the complete panels. Over these panels, sources are distributed linearly and
wing surface and changes the wake orientation to model
modified flow to better predict the effects of downwash. These doublets are distributed quadratically. Each singularity is
codes, allow computation in both subsonic and supersonic placed at mid-chord of the panel and its influence is computed
regimes, as they include Prandtl-Glauert compressibility at a control point P for far-field calculations and at field point
correction. The rectangular wing is generated with an identical Q for near-field calculations. A single panel geometry is
number of panels and networks for coherent comparison. depicted in Fig. 1.
Distinguishable pre-processing techniques are utilized and In the figure, R is the geometric distance between the control
similar boundary conditions and flow conditions are maintained
over the surfaces that are then given to solvers. The induced point P and field point Q and is given by:
drag polar is plotted and compared with wind tunnel and R=#(xq -xp )2 +β2 [(yq -yp )2 +(zq -zp )2 ] (1)
analytical data.

Keywords— Analytical, NACA 0015, panel method, Vortex Where 𝛽2 = 1 − 𝑀2


Lattice Method (VLM), wind-tunnel This distance determines the velocity potential induced by
Point Q to Point P as shown below:
I. INTRODUCTION For point source:
−1
φ7 8 (x9⃗q) = (2)
Panel methods are numerical techniques for solving linear, R(x;7 , x;= )
inviscid, and irrotational flows about the aircraft in subsonic
as well as supersonic regimes using the Prandtl-Glauert For point doublet:
equation. These methods divide the aircraft, or some 9⃗(x;7 , x;= )
−R
(3)
φ7 > = n@.
approximation of it, into quadrilateral panels, and distribute RB
source, doublet or vorticity singularities over it. The
singularity distributions may be constant in strength- for
lower-order methods- or vary linearly or quadratically for
higher-order methods. To simulate the flow and maintain
flow-tangency, the boundary conditions are imposed on panel
surfaces. Multiple equations for each panel are solved
simultaneously to calculate aerodynamic forces and

978-1-7281-4970-7/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 09:52:05 UTC from IEEE Xplore. Restrictions apply.
Where VIC is velocity influence coefficient matrix and 𝑏𝑖 =
99999⃗
−𝑉 @ . Applying boundary condition on each point yields:
∞. 𝑛
[AIC]{λ} = {b} (11)

AIC is the Aerodynamic Influence Coefficient matrix and ʎ


is the unknown singularity parameter. 'b' is known from
boundary conditions. Solving for ʎ gives potential and
velocity distribution thus, the pressure distribution on each
panel surface S. Integrating the pressure values gives the
aerodynamic forces and moments.
B. VLM Code
In VLM code, the planform is described by a series of straight
Fig. 1. Panel geometry and control point [2] lines projected on X-Y Plane from the three-dimensional
geometry. This planform is divided by chord-wise and span-
wise lines along the surface to form quadrilateral panels..
Sum of sources and doublets yields another flow solution that
can be expressed in terms of Green’s Third Theorem as:
φ7 = C[σ (x9⃗q)K F Gx= , x7H + µ(x9⃗q)K J Gx= , x7H]dS= (4)

Where,
1 8
K F Gx= , x7 H = φ Gx; H (5)
k 7 =
1
K JGx= , x7 H = φGx;= H (6)
k
In the equations above, σ is the source strength, µ is the
doublet strength and, Kσ and Kμ are the amplitudes of
singularity strength distribution. Equation (4) shows velocity Fig. 2. 1/4th-3/4th chord rule on a panel geometry [3]
potential induced by both the source and doublet at point P
that is distributed over a surface S. Since ∇2∅ is zero as Kσ Potential flow theory in the form of Biot-Savart Law is used
is the function of point P. The above equation satisfies the to represent disturbances due to lift distribution on the
Prandtl-Glauert equation: planform [3]. Horseshoe vortices are placed on the quarter-
∂2 ∅ ∂2 ∅ ∂2 ∅ chord of each panel. The strengths of these vortices are
(1 − M 2 ) + + =0 (7)
∂x 2 ∂y 2 ∂z 2 computed on the control points placed at ¾ of each panel
chord by solving the following system of equations,
The velocity can be obtained as: stemming from tangent-flow boundary condition [3]:

999⃗
VP =∇P φP = ∬ [σ (x9⃗q)∇P K σGxq ,xp H+µ(x9⃗q)∇P K µ Gxq ,xp H]dSq (8) γq
nFp,q − Fr,q tan∅q u v y = 4π{α}} (12)
U
The equation for potential is utilized to produce the influence
coefficient matrix to relate the strength of source and doublet Here, Fw and Fv are influence coefficients for downwash and
to the potential over a surface S. The results for each panel side-wash respectively for nth panel, U is the free-stream
are then summed. The equations are solved for singularity velocity, and αk is local angle of attack at each panel chord.
strengths over every panel using boundary conditions and ∅q and γq are the potential and circulation value for nth panel.
then velocity potential is computed using the equations The Kutta-Joukowski theorem for lift per unit length of a
above. The total velocity is obtained as the sum of the induced vortex filament uses the vortex strength values to compute lift
velocity of each panel over the control point and the free for each panel in the following manner:
stream by relating it to the wall-tangency boundary condition l ̅ = ρVγ (13)
as:
Z The lift for each panel is summed to calculate the total lift
VW 9999⃗
VXY + 99999⃗ 999⃗X . n@
V^ _ . n@ = V (9) over the planform. Furthermore, leading edge thrust, leading
[\] edge suction and induced drag are calculated using near-field
potential in each panel [3].
For N singularities, above equation can be expressed as
Z

W VICb[ λ[ = bb (10)
[\]

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 09:52:05 UTC from IEEE Xplore. Restrictions apply.
II. LITERATURE REVIEW due to spurious line-vortex terms at the panel edges. To tackle
developed in the 1960s to estimate aerodynamic this problem, Woodward modeled the fuselage with constant-
characteristics of aircraft numerically. With the advent of strength source singularities and the wing using vortex
modern computational aerodynamics, which permits singularities; whose strength varies linearly chord-wise but is
modeling of the entire aircraft surface including complex constant span-wise. This improved the modeling capability
geometries of weapons and engine intakes, these elementary but did not completely remove the instability problem [7].
methods were considered outdated. These methods, however, Morino modified the method by using constant-strength
are still in use today for the conceptual design phases of sources and doublets on hyperboloid panels on the mean
aircraft as they require significantly less time and surface but this too was unable to cater for the supersonic
computational resources as compared to the exhaustive CFD instability [8]. In addition to this, this method increased the
techniques. The quality of results is undoubtedly not refined chances of flow leakage as the end-points of panels were not
due to modeling limitations but is reasonable enough for co-planar. The instability problem was finally solved using a
initial estimates and design of an aerodynamic surface based continuous, quadratic strength-distribution of doublet
on a given loading. singularities over the aircraft surface [7]. A fortuitous side-
These codes, moreover, are not applicable in the transonic effect of this approach was the insensitivity of the results to
regime and viscous effects are not catered for. the size, shape, and arrangement of the panels. The higher-
Vortex Lattice Method was developed successively by a order distribution allows for actual-surface modeling of the
chain of aerodynamicists to cultivate a numerical technique aircraft in addition to eliminating the stability issues in
for computing aerodynamic forces on an aircraft. It is based supersonic flow. To prevent flow leakage, the panels were
upon the finite difference method to solve loading integral further divided into piece-wise flat sub-panels [2]. Panel
equations. Fluid models in CFD and VLM, are generally methods are based upon distributing singularities that can be
based upon partial differential equations that had no practical a source, vortex or a doublet. Since vorticity is a vector
significance until L.F. Richardson applied them to determine quantity and doublets are easier to work with, panel methods
stresses for masonry dam design in 1910 [4]. In 1918, Ludwig generally do not use vortices to save computational time and
Prandtl and his companions at Gottingen, Germany, avoid stability problems [9].
developed a method for airfoil analysis and extended it to the III. METHODOLOGY
3D wing. Collectively, they developed the vortex panel
method to solve for loads on the airfoil. They then formulated The comparison discussed in this paper is based upon two
the lifting line theory but failed to implement this theory potential flow solvers, a Vortex Lattice Method code that is
practically. With the advent of computers, their technique developed and modified by NASA and a panel method code
was excessively utilized to solve flow problems. V. Faulkner developed by Boeing and modified by Martin C. Hegedus. A
modeled the wing with a grid of discrete horseshoe vortices rectangular wing is chosen with the specified dimensions as
to predict wing surface loading and coined the name vortex shown in table 1 for analysis of both solvers with analytical
lattice method. However, this method required significant and wind tunnel data to establish a judgment upon accuracy,
computing time and power that strictly limited the number of computational time and input-controllability.
panels and the authenticity of results was questionable. The A. Wind Tunnel
solution to this problem had to wait until the development of
An experiment was performed by Tongsawang [10] to obtain
advanced computers [4]. Harloww and Fromm developed a
aerodynamic loading on a rectangular wing with the
method for numerical simulation in 1965 and served as a
dimensions mentioned in Table 1. The wind tunnel was a
basis for modern computational dynamics [4]. Vortex lattice
low-turbulence, open-loop, subsonic atmospheric wind
method was extended to improve the computation of
tunnel capable of a maximum of 25 m/s velocity. In the
aerodynamic forces over low aspect ratio thin wings and to
experiment, an airspeed of 6 m/s was used corresponding to
compute pressure distribution on thick wing surfaces by
a Reynolds number of 60000. Observations of aerodynamic
Charles W. Smith and Ishwar C. Bhateley [4]. Wake
forces were recorded for angles of attack of 2 to 22 degrees
modeling and its effect on the lifting surfaces and effect of
with an increment of 2 degrees. A 1.2mx1.2m test-section
flaps were studied by Brian Maskew using VLM [5]. Soon
was used to simulate the above-mentioned conditions.
after, the effect of blowing surfaces and jet interaction was
studied by C. Edward Lan [6]. The method was then B. Vortex Lattice Method
optimized by ambitious researchers to reduce cost and to In the OpenVSP geometry module, a wing is generated with
attain accurate results, with the minimum number of panels similar dimensions as that of [10]. To generate major panels,
and by increasing the convergence rate. the number of span-wise stations must be provided. This
Panel methods were originally developed as lower-order input is specified in the tessellation section of VSP and must
methods for incompressible and subsonic flows, however not exceed 50 [1]. The panels are further divided into sub-
supersonic panel methods were developed in the 1960s. The panels. These sub-panels are generated based on chord-wise
3D panel methods allowed for modeling of the entire aircraft stations which must not exceed 100 [1]. A paneled geometry
surface. For supersonic case, initially, mean-surface models conforming to these constraints is created. It is then exported
had to be used to avoid the numerical instabilities which arose

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 09:52:05 UTC from IEEE Xplore. Restrictions apply.
from VSP in the form of coordinates and given to the boundary conditions are given as input to the pre-processor.
preprocessor. The preprocessor converts this 3D wing to 2D The pre-processor converts all this data into a processor-
by mean airfoil approximation. A separate file is generated readable format [9]. This file is an input to the processor. Lift
by this pre-processor that is composed of major panel co- and drag coefficients, second-order pressure coefficient,
ordinates and mean camber line of the airfoil at each span- mass flux component, velocity components and potential are
wise cross-section as shown in Fig. 3. This file describes the obtained as an output [14].
entire geometry and is given as an input along with desired
flow conditions to the processor.
TABLE I. GEOMETRY SPECIFICATIONS [10]

Span 700mm
Aspect Ratio 3.5
Chord 200mm
Airfoil NACA 0015
Fig. 4. Gambit geometry generated for panel code

Pre-processor
Geometry in for generating Processor
OpenVSP
panels

Geometry & Pre-processor


panels in for input file Processor
Fig. 3. OpenVSP geometry generated for VLM code Gambit formatting

The processor generates an output file containing the lift and Fig. 5. Comparison of flow solvers’ methodology
induced drag coefficients corresponding to each Mach
number and angle of attack combination. The entire process can be easily understood with the flow
C. Panel Method chart depicted in Fig. 5.
For the panel method code, the wing geometry is generated D. XFOIL Corrected Data
in Solid Edge CAD software. The geometry is imported in The analytical estimation is used to predict the 3D wing
IGS format to Gambit v2.3.16 as seen in Fig. 4. Once in aerodynamics by using 2D airfoil drag polar from XFOIL.
Gambit, chord-wise and span-wise stations are defined on the XFOIL is a user-friendly program used for design and
wing's upper and lower surfaces. The number of span-wise analysis of subsonic, low Reynolds number airfoils. It is
stations and chord-wise stations is set such that the number of based upon higher-order panel method with both viscous and
panels is equal to the VLM case. Based on these points, inviscid analysis methods developed by Drela and Giles [15].
Gambit generates a grid consisting of quadrilateral panels The drag polar obtained is shown in Fig. 6.
over the upper and lower wing surfaces. From the numerous This drag polar departs from paraboloidal shape due to the
gridding options available in Gambit, the Quad Map viscous effects [16]. For estimating 3D effects, span
technique is used to create the panels; as only quadrilateral efficiency plays an important role. It represents the effect of
panels are supported by the program. A coordinate file, span-wise flow on drag and lift. As per the definition in [16]
similar to the one used for VLM code, is exported. These it is defined as:
coordinates are further processed using several techniques 1
e= (14)
and are arranged into networks. A network is a smooth part (1 + δ + 𝑎2 πAR)
of the configuration which is then divided into panels for where δ is a measure of deviation from elliptical loading and
distributing source and doublet singularities [11]. Each a2 is the measure of parabolic shape of the drag polar. For
network has a normal vector pointing in a particular direction. computing span efficiency, δ is neglected as its value is
It must be ensured that this vector is projecting outward from insignificant in comparison to a2πAR. k is computed by least
the network so that the flow remains tangent to the surface. square quadratic fits of Cd on Cl for low Reynolds numbers
At least two networks are required as input to the processor [16] and is shown in table 2.
[12]. Wake is attached to the trailing edge of the wing to
satisfy the Kutta condition [13]. The impermeable, thick
surface boundary condition is imposed upon each network.
These network coordinates and the file composed of flow and

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 09:52:05 UTC from IEEE Xplore. Restrictions apply.
on the versatility of these inputs, the quality of output is
Airfoil Drag Polar affected.
1.5
Analytical Data Induced-Drag Polar
1
1.2
Lift Coefficient

0.5 1

Lift Coefficient
0 0.8
0.6
-0.5
0.4
-1 Drag Coefficient
0.2
0 0.05 0.1 0.15 0.2
0
Induced Drag Coefficient
Fig. 6. Drag polar for airfoil obtained from XFOIL
0 0.05 0.1 0.15 0.2 0.25
Fig. 7. Analytical induced drag polar for wing

TABLE II. LEAST QUADRATIC SQUARE FITS OF CD AND CL


FOR LOW REYNOLDS NUMBER [16]
The common inputs to both programs are the geometric
parameters; such as wingspan, chord, and area, as well as the
flow conditions which include Mach number and angles of
attack. Further, the aircraft geometry is taken as an input in
the form of coordinates. The VLM code converts this
geometry into a two- dimensional form, while the panel code
retains the three-dimensionality. The VLM code requires full
geometry whereas, in the panel code, half of the geometry can
By plugging in the values of a2 and Aspect Ratio, the Oswald be input and the symmetry option utilized to get the final
efficiency was calculated to be 0.503. output, reducing computational time. Moreover, to control
Using the lift coefficient vs angle of attack curve, lift-curve the grid size, VLM code entails an explicit input of the
slope for the airfoil is obtained at Reynolds number 60000 by number of panels and sub-panels based on which it generates
XFOIL. For 3D effects, following formulation from [17] was a uniform grid over the surface. For the panel code, on the
used: other hand, the grid can be controlled in the CAD software.
a„
a = a (15) This lends more freedom over the grid bias as the denser grid
1+ „
πeAR can be generated over areas of interest such as wing leading
edge and areas with significant curvature, to correctly model
Using the straight-line equation for lift-curve slope [17], the
them. Over this grid, VLM code distributes vortices with
lift coefficient can be estimated at different angle of attack.
(16)
either constant or cosine variation, based on user input. The
C… = a(α − α…\† )
panel code, however, distributes the source singularities
linearly and the doublet singularities quadratically for every
Similarly, the induced drag coefficient can be obtained by
case. For both codes, these arrangements enable calculation
[18]:
in the supersonic regime. In contrast to the VLM code, the
C…2 panel code offers a variety of boundary conditions catering to
C‡ˆ = (17)
πeAR
varying requirements including mass flux, potential, velocity,
and wake boundary conditions. These boundary conditions
can be applied on each network independently which
IV. ANALYSIS facilitates realistic modeling of the problem. Panel code,
The results of the flow-solver codes are compared with wind unlike the VLM code, has a wake input option. The wake can
tunnel and are presented in Fig 11. These results are analyzed be applied automatically by giving the wake command, or by
based on control over inputs, processing time and validity or inputting the wake coordinates. The orientation of the wake
accuracy of results. Each of these analyses is discussed in can also be controlled to correctly model its direction with the
separate sections. angle of attack variation. Lastly, VLM requires leading-edge
suction as input. This enables the modeling of vortex-lift
A. Input Controllability effects which contribute to induced drag. However, vortices
In this section, the freedom of control over the output is are not formed at such low speeds therefore this factor is
analyzed by comparing the required inputs of the codes. The insignificant for the present study. Based on these
programs require several minimum inputs to run and based observations, it can be said that the panel code offers more

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 09:52:05 UTC from IEEE Xplore. Restrictions apply.
options to model the problem as close to reality as possible. Computational Time vs Accuracy
0.05 25
TABLE III. COMPARISON OF INPUTS FOR PANEL AND VLM
CODES
0.04 20
Input Panel Method VLM 15
0.03
Mach/AoA Information ü ü 10

Time
CDi
Geometric Parameters ü ü 0.02
5
Number of Panels û ü
0.01 0
Manual Meshing ü û
Symmetry ü û 0 -5
Boundary Conditions ü û 0 20 40 60 80 100
Wake ü û Number of Panels
Leading-Edge Suction û ü Cdi
CDi Vs
vs Num.
No. of of Panels
Panels Computational Time
Fig. 8. Grid Independence analysis for computational time and accuracy
B. Time
Computational time is important to scrutinize which flow
code to use for analysis. It also determines the operational
cost and effective computer memory. To find an optimum
computational time with minimum number of panels and not
compromising upon the accuracy, multiple iterations for each
code are performed. For Panel Method code, the number of
panels are controlled by span-wise and chord-wise divisions.
It was found that higher the number of panels, more the
computational time required. Computational time and cost
are directly proportional therefore it can be deduced that it
will have the same trend as time. A plot of time, accuracy and
number of panels can be seen in Fig. 8. It can be seen that as
the number of panels increases, computational time also Fig. 9. Grid independence analysis for VLM
increases linearly. On the contrary, accuracy is achieved at an
optimum number of panels. Computational time can also be There are multiple files generated with the output file for
significantly reduced by incorporating symmetry effects. these codes. The files vary in sizes and have different
Calculation can be performed on half configuration and information related to the studies. Total space used for panel
results can be obtained for full configuration. method, keeping constant the number of cases, is 97.8 MBs.
Therefore, the number of panels was fixed to twenty for panel VLM, however, utilized 1.716 MBs and includes fewer files.
method code as it requires two seconds and results are fairly
converged. C. Accuracy
The comparison for panel and VLM programs along with
For VLM code, the major panels cannot exceed 20 and wind-tunnel and XFOIL corrected results is depicted in Fig.
number of camber points cannot be more than 50, as stated in 10. At higher lift values, the graphs significantly digress from
the code limitations. The computational time required for 20 the experimental data. It is because these linear, inviscid
panels is the same as required for 11 panels, i.e. less than five codes do not cater for aerodynamic stall. The figure also
seconds, hence 20 panels are used to ensure accuracy. The
shows that the induced drag polar for the Panel Method code
grid independence is performed by varying sub-panels and
closely follows the experimental results. The higher-order
major panels both. It can be seen from Fig. 9 that the desired
panel method accurately predicts the pressure difference
accuracy is achieved for small number of major and sub-
panels. between the upper and lower surfaces of the wing. The panel
code results also exhibit concurrence with analytical results
as XFOIL is also based on a panel method. VLM curve, on
the other hand, deviates from both analytical and
experimental data as the angle of attack increases. That is
because the VLM code is based upon thin airfoil
approximation and therefore under-predicts the lift. In thick
airfoils, such as NACA 0015, the pressure difference between
upper and lower surface is significantly greater at higher
angles of attack as compared with thin airfoils. Since this

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 09:52:05 UTC from IEEE Xplore. Restrictions apply.
pressure difference is not being correctly modeled, the [2] M. A. M. A. E. Epton, "PANAIR: A Computer Program for
expected deviation between VLM and experimental data is Predicting Subsonic or Supersonic Linear Potential Flows about
Arbitrary Configurations using a Higher-Order Panel Method".
evident from Fig. 10. It can, therefore, be deduced that the
[3] J. E. L. Richard J. Margason, "Vortex Lattice FORTRAN Program
panel method yields more realistic results as it caters for wing for Estimating Subsonic Aerodynamic Characteristics of Complex
thickness in calculations. Planforms," Washington, D.C., 1971.
[4] J. DeYoung, "Historical Evolution of Vortex-Lattice Methods," in
Lift Coefficient vs Induced Drag Coefficent Vortex Lattice Utilization, Langley Research Center, Virginia, 1976.
1.2 [5] B. Maskew, "A Quadrilateral Vortex Method Applied to
Cofigurations with High Circulation," in Vortex-Lattice Utilization,
Langley Research Center, Virginia, 1976.
1
[6] C. E. Lan, "Upper-Surface Blowing Jet-Wing Interaction," in Vortex-
Lift Coefficient

0.8 Lattice Utilization, Langley Research Center, Virginia, 1976.


[7] L. L. Erikson, "Panel Methods- An Introduction," 1990.
0.6 [8] J. L. Hess, "Panel Methods in Computational Fluid Dynamics".
[9] G. A. Saaris, E. N. Tinoco, P. E. Rubbert and J. L. Lee, A502I User's
0.4 Manual- PANAIR Technology Program for Solving Problems of
Potential Flow about Arbitrary Configurations, Boeing, 1992 .
0.2
Induced Drag Coefficient [10] K. Tongsawang, "Stall Control of a NACA0015 Aerofoil at Low
Reynolds Numbers," Sheffield, England, 2015.
0
0 0.1 0.2 0.3 [11] T. Dirbyshire and K. W. Sidewell, " PANAIR Summary Document,"
Ames Research Center; Langley Research Center .
VLM Wind Tunnel
[12] M. Tarkian and F. J. Tessier, "Aircraft Parametric 3D Modelling and
PM XFOIL Corrected
Panel Code Analysis for Conceptual Design," Linkoping, Sweden,
Fig. 10. Comparison of flow-solvers with analytical and wind tunnel induced-drag 2007.
polar [13] A. M. Jaffri, "Inviscid Analysis of Aerodynamic Characteristics of
High-Performance Fighter Aircraft," Rawalpindi , 2005.
[14] F. T. Johnson, "A General Panel Method for the Analysis and Design
V. CONCLUSION of Arbitrary Configurations in Incompressible Flows," 1980.
[15] M. Drela and H. Youngren, "Web.mit.edu," 30 November 2001.
The panel method and VLM are used for estimating [Online]. Available:
aerodynamic forces on the aircraft surface. These methods Web.mit.edu/drela/public/web/xfoil/xfoil_doc.txt.
are based upon linear equations rather than complex Navier- [16] G. R. Spedding and M. J., "Span Efficiencies of Wings at Low
Stokes equation used in CFD; therefore, requiring less time Reynolds Numbers," Journal of Aircraft, vol. 47, no. 1, pp. 120-128,
and computational resources. In this paper, a comparison of 2010.
two potential flow-solvers based upon accuracy, [17] J. D. Anderson, Fundamentals of Aerodynamics.
computational time and input controllability is carried out. [18] J. D. Anderson, Aircraft Performance and Design, McGraw Hill ,
This research aims to determine an optimum solver that can 1999.
predict inviscid aerodynamics accurately and efficiently with
the least amount of time. It is observed that the panel method
code provides additional input controllability which enables
more realistic modeling of the flow as compared to VLM
code. On the contrary, it requires more computational time
and memory than VLM code. While VLM results correspond
to experimental and analytical data at low lift coefficients, the
departure is significant for high-lift conditions because the
thickness is not catered for. As this limitation does not exist
for panel code, its results conform to the experimental data as
well as to the panel-method-based XFOIL corrected results.
It can be concluded that the panel method is better than VLM
upon accuracy and input controllability but it requires
significantly more computational time and memory as
compared to VLM.

VI. REFERENCES

[1] L. R. Miranda, "A Generalized Vortex Lattice Method for Subsonic


and Supersonic Flow Applications".

Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 09:52:05 UTC from IEEE Xplore. Restrictions apply.

You might also like