Você está na página 1de 33

OPTIMIZATION OF WING PLANFORM USING THREE

DIMENSIONAL PANEL METHODS


A Project Report
submitted by
VINAY KIRAN C K
in partial fullment of the requirements
for the award of the degree of
BACHELOR OF TECHNOLOGY
DEPARTMENT OF Aerospace Engineering
INDIAN INSTITUTE OF TECHNOLOGY MADRAS, CHENNAI
April 2010
CERTIFICATE
This is to certify that the project titled Optimisation of Wing Planform Using 3D Panel
Methods, submitted by Mr. Vinay Kiran C K (AE06B022), at the Department of Aerospace
Engineering, to the Indian Institute of Technology, Madras, for the award of the degree of
Bachelor of Technology, is a bonade record of the project work done by him under my
supervision. The contents of this report, in full or in parts, have not been submitted to any other
Institute or University for the award of any degree or diploma.
Prof. M. Ramakrishna
Professor
Dept. of Aerospace Engineering
IIT-Madras, 600 036
Prof. P. Sriram
Head of the Department
Dept. of Aerospace Engineering
IIT-Madras, 600 036
ACKNOWLEDGEMENTS
I wish to thank my guide Prof. M. Ramakrishna for giving me this opportunity to work under
him. I am grateful to him for guiding me all through, suggesting corrections and improvements
from time to time. Id like to thank my friends for making my stay at IIT a memorable one.
Special thanks are due to Nitesh Shah and Rajesh Gandham for their inputs and suggestions.
Finally, Id like to thank my parents and brother for their love and condence in my abilities.
i
ABSTRACT
KEYWORDS: Vortex Lattice Method, Unstructured Grid, Zimmerman
One of the problems in the design of Micro Air Vehicles is the choice of the shape of the
wing planform. The main objective of the project was arriving at the optimum shape by the
use of the Vortex Lattice Method. In this method, the wing is replaced by the mean surface,
and an unstructured grid is used to represent this mean surface. VLM is discussed in detail
in the book Low Speed Aerodynamics ((1)), which uses rectangular panels on structured grids.
In this project, the VLM is applied on an unstructured grid. Three planform congurations,
rectangular, zimmerman and the inverse zimmerman are investigated. Python was used to
design the code using object oriented principles. The solution to the problem is obtained as a
superimposition of more fundamental ows, like the vortex ow. The Vortex Lattice Method
demands less programming efforts and offers results that are good initial estimates.
ii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS i
ABSTRACT ii
LIST OF FIGURES 1
ABBREVIATIONS 2
1 INTRODUCTION 3
1.1 Elementary Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Source Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Vortex Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Vortex Filament . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Derivation for the Induced Velocity by a Straight Line Filament . . 6
1.3 Vortex Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Panel Methods 8
2.1 Geometry & Grid Generation . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Wing Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 The Vortex Lattice Method 12
3.1 Choice of Singularity Element . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Boundary Conditions & Inuence Coefcients . . . . . . . . . . . . . . . . 12
3.3 Modelling the Wake . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 Secondary Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4.1 Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4.2 Coefcient of Pressure . . . . . . . . . . . . . . . . . . . . . . . . 15
iii
3.4.3 Calculation of Lift . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Results 17
4.1 Rectangular Planform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 The Inverse Zimmerman Conguration . . . . . . . . . . . . . . . . . . . . 19
4.3 The Zimmerman Conguration . . . . . . . . . . . . . . . . . . . . . . . . 21
5 Concluding Remarks & Future Work 25
A Code Snippets 26
LIST OF FIGURES
1.1 Source Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 A Free Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 A Vortex Filament . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 A Straight Line Vortex Filament . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 A Triangular Vortex Ring Panel . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 A Gmsh example illustrating the effect of Characteristic Length. The gure on
the left uses a CL of 0.3. The one on the right has a CL of 0.6 . . . . . . . . 9
2.2 Camber Proles NACA 3-Digit Reex Airfoil Series . . . . . . . . . . . . 10
3.1 Horse Shoe Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Component of Circulation that Contributes to the Generation of Lift . . . . 16
4.1 C
p
distribution over the wing. Rectangular. . . . . . . . . . . . . . . . . . . 17
4.2 distribution over the wing. Rectangular. . . . . . . . . . . . . . . . . . . 18
4.3 Net Circulation per Panel Distribution over the wing. Rectangular. . . . . . 18
4.4 C
p
distribution over the wing. Inverse Zimmerman . . . . . . . . . . . . . 19
4.5 distribution over the wing. Inverse Zimmerman . . . . . . . . . . . . . . 20
4.6 Net Circulation per Panel Distribution over the wing. Inverse Zimmerman . 20
4.7 C
p
distribution over the wing. Zimmerman . . . . . . . . . . . . . . . . . . 21
4.8 distribution over the wing. Zimmerman . . . . . . . . . . . . . . . . . . 22
4.9 Net Circulation per Panel Distribution over the wing. Zimmerman . . . . . 22
4.10 C
l
vs curve. Rectangular. . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.11 C
l
vs curve. Zimmerman. . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.12 C
l
vs curve. Inverse Zimmerman. . . . . . . . . . . . . . . . . . . . . . 24
1
ABBREVIATIONS
AoA Angle of Attack
AR Aspect Ratio
BEM Boundary Element Method
CL Characteristic Length
C
l
Coefcient of lift
C
p
Coefcient of Pressure
MAV Micro Air Vehicle
SVD Singular Value Decomposition
VLM Vortex Lattice Method
2
CHAPTER 1
INTRODUCTION
One of the problems in the design of Micro Air Vehicles is the choice of the shape of the wing
planform. That is, given a numerical value for the magnitude of the area after wing loading
calculations, which shape is best suited for the planform. One could choose between several
options including Elliptical, Rectangular and Zimmerman kind of conguration. This project
aims at solving the problem of choosing the optimum planform shape using Three Dimensional
Vortex Panel Methods, which belong to the category of Boundary Element Methods.
For ows where the characteristic speeds are lesser than the speed of sound and where
viscous effects are negligible, i.e. in inviscid incompressible ows, the equation governing the
ow is simply the Laplacee Equation.

2
= 0
Since the Laplaces Equation is a linear differential equation, solutions can be constructed
by linear superposition of other solutions. This property is exploited and one uses the funda-
mental ows, namely the Source Flow and the Vortex Flow to construct the solution to any
given problem. Since the source ow has zero circulation, it is not sufcient to model lifting
ows, in which case vortices are used to construct the solutions.
1.1 Elementary Flows
1.1.1 Source Flow
Source ows are the basic building blocks used in modelling non-lifting ows. The most basic
owis the point source, using which one can construct more complicated source sheets, which
can be used in 2D as well as 3D. The equation for the point source can be derived using the
assumption of radial ow. The following equations are quoted from Anderson(2).
V

= 0
V
r
=

2r
Figure 1.1: Source Flow
where is the source strength. Physically, is the rate of volume ow from the source,
per unit depth perpendicular to the page.

V (r, ) =

2r
r
1.1.2 Vortex Flow
The other fundamental ow is the Point Vortex. Due to its net non-zero circulation, it can be
used to model lifting ows. The following is quoted from Anderson(2).
V

=

2r
V
r
= 0

V (r, ) =

2r

4
where is the circulation or the strength of the vortex, and

is the unit vector in the tangential
direction.
Figure 1.2: A Free Vortex
1.2 Vortex Filament
The inuence of Vortex Filaments can be computed by treating it as a continuous distribution
of point vortices. And the summation is replaced by an integral. The formula is quoted from
(2)
(x, y) =
_
P
2
P
1
ds
2
ln r
where is the vortex strength per unit length.
Figure 1.3: A Vortex Filament
5
1.2.1 Derivation for the Induced Velocity by a Straight Line Filament
Consider the case where we have a vortex lament extending from point A to point B, and
we have to calculate the velocity induced by this segment at a point P in the ow domain.
By segment AB, we mean a segment that is a part of a closed vortex ring ( By Helmholtzs
Theorem a vortex lament cannot start or end in a ow eld). Then the velocity induced at P
due to AB, using the Biot-Savart Law is (from chap. 2 (1)),
Figure 1.4: A Straight Line Vortex Filament

V
P
=

4
r
1
r
2
r
1
r
2

2
r
0

_
r
1
r
1

r
2
r
2
_
where,
r
0
is the vector

AB
r
1
is the vector

AP
r
2
is the vector

BP
6
1.3 Vortex Ring
A vortex ring is a closed circuit vortex lament. The treatment is same as that for a vortex
lament, except that the integral is replaced by a closed loop integral. Hence,
=
_

2r
ds
In the Vortex Lattice Method, one uses rectangular vortex rings, that is rings composed of
four straight line segments. In the current project, triangular vortex rings are used. This way,
one can benet from the time and effort saved in modelling complicated geormetries. Using an
unstructured grid helps in achieving uniform mesh characteristic length throughout the domain.
Figure 1.5: A Triangular Vortex Ring Panel
7
CHAPTER 2
Panel Methods
Panel Methods are a class of numerical techniques used to solve inviscid incompressible ows.
Since the equation being solved is the Laplaces Equation, it is also generally called potential
ow code. They were eventually replaced by Finite Volume Solvers. But the Panel methods
still provide good initial estimates for inviscid ows. The solution generally consists of three
main parts. The srt is the discretization of the geometry into panels of the desired size. For this
purpose, a software called Gmsh (5) was used. The second part is to obtain the vortex strengths
required to simulate the ow. This is done by setting up a system of linear equations, which are
obtained from the boundary conditions. The third and nal part is the post-processing which
involves performing the secondary computations viz. calculating the pressures, loads, C
l
, C
d
.
2.1 Geometry & Grid Generation
Gmsh(5) is a GUI based free open source mesh generator which also has pre and post-processing
facilities. Once the geometry is dened in it, a mesh is generated. The neness of the mesh can
be controlled by setting the Characteristic Length. The data can be outputted in a number of
formats. In this program, the vtk le format was chosen. The mid-points of the triangles are
chosen to be the control points, i.e. the points where the no normal condition will be applied.
Figure 2.1: A Gmsh example illustrating the effect of Characteristic Length. The gure on the
left uses a CL of 0.3. The one on the right has a CL of 0.6
2.2 Boundary Conditions
The Boundary Condition applied is Dirichlet Boundary Condition. Since the ow is inviscid,
only the normal component of velocity is set to zero. The no normal ow condition is applied
at each of the panel control points. Due to the thin wing approximation that has been made, the
boundary conditions are chosen at the mean camber rather than the actual surface. Since we
have N control points, we are provided with N equations. Also we have N unknowns i.e. the
vortex strengths. Hence, we have a system of equations that can be solved.
2.3 Wing Geometry
Flying wings generally use a reexed airfoil prole. This is done to compensate for the absence
of the restoring pitching moment which the tail wing would have provided in a conventional
aircraft. The reex towards the rear end of the airfoil can be thought of as a modied tail. For
our design course, we used the MH-60 airfoil (7), which has a history of having being used in
tail-less aircraft. For this program however, the NACA-251 was used. It is an airfoil from the
9
NACA 3-Digit Reex Airfoil family. The camber for this prole is given as,
y
c
=
k
1
6
__
x
c
r
_
3

k
2
k
1
(1 r)
3
x
c
r
3
x
c
+ r
3
_
, 0
x
c
r
y
c
=
k
1
6
_
k
2
k
1
_
x
c
r
_
3

k
2
k
1
(1 r)
3
x
c
r
3
x
c
+ r
3
_
, r <
x
c
1
where,
m is Chordwise Location for maximum ordinate of airfoil or camber line
r is chordwise location for zero value of second derivative of 3-digit or 3-digit-reex
camber-line equation
k
1
and k
2
are constants that determine the shape of the airfoil.
Figure 2.2: Camber Proles NACA 3-Digit Reex Airfoil Series
The values of the constants for airfoils are given below. Data is taken from NASA Technical
Memorandum 4741 (3)
The planform shapes used are the rectangular, Zimmerman and the Inverse Zimmerman
Congurations. (For shapes, see results.4 )
10
Table 2.1: NACA 3-Digit Reex Airfol Proles
Camber-line Designation m r k
1
k
2
/k
1
221 0.10 0.1300 51.990 0.000764
231 0.15 0.2170 15.793 0.00677
241 0.20 0.3180 6.520 0.0303
251 0.25 0.4410 3.191 0.1355
11
CHAPTER 3
The Vortex Lattice Method
The Vortex Lattice Method is an extension of the Vortex Panel Method to the three dimensional
case. In the vortex panel method, the airfoil is discretized as straight line panels with the
vortex strength varying in a certain way either constant strength, or linearly, or any other higher
order variation. ( The geometry can also be discretized as curved panels. The straight line
approximation is the simplest ). A point, mostly the mid-point, on each of these panels is
chosen as the control point where the no normal ow condition is applied. Finally, the Kutta
Condition requires that the ow leave the trailing edge smoothly. In the vortex panel methods,
this is done by forcing the two panels at the trailing edge to have the same vortex strength.
3.1 Choice of Singularity Element
In the Vortex Lattice Method, the wing, (the mean surface rather) is divided into triangular
panels. The approach follows the one outlined in (1), the difference being the method used in
the book uses rectangular panels. Once the mean surface is divided into triangles, each of the
triangles is now a Vortex Ring, and each of the arms of the ring is a vortex lament. The vortex
strength is constant over each panel, and at this stage is the unknown that has to be computed.
3.2 Boundary Conditions & Inuence Coefcients
The induced velocity at any point is the sum of the velocities induced by each of the vortex
panels and the freestream velocity V

. For the application of the boundary conditions, the


control points of the panels become points of interest.

V
i
=

N
j=0

V
ij
+

V

where i denotes the i


th
control point and j denotes the j
th
vortex ring.
Now, since there is no mass ux across the mean surface, we must have

V n = 0
Now expanding the

V term, we get

N
j=0

V
ij
n =

V

V
i1
n
i
+

V
i2
n
i
+ ... +

V
iN
n
i
=

V

n
i
The application of the boundary condition at each of the N panels gives us N equations. In
matrix form, the above set of equations can be written as
_
_
_
_
_
_
_
_
_
_
_

V
11
n
1
V
12
n
1


V
1N
n
1

V
21
n
2
V
22
n
2


V
2N
n
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

V
N1
n
N
V
N2
n
N


V
NN
n
N
_
_
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_

n
1

n
2
.
.
.
.
.
.

n
N
_
_
_
_
_
_
_
_
_
_
_
(3.2.1)
Since the vortex strength is assumed to be constant over each ring, it factors out.
_
_
_
_
_
_
_
_
_
_
_
a
11
a
12
a
1N
a
21
a
22
a
2N
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
a
N1
a
N2
a
NN
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_

2
.
.
.
.
.
.

N
_
_
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_

n
1

n
2
.
.
.
.
.
.

n
N
_
_
_
_
_
_
_
_
_
_
_
(3.2.2)
In the above matrix equation, each of the terms is called an Inuence Coefcient. See the
appendix for the code (A) . An inuence coefcient a
ij
is physically the velocity induced by
unit strength j
th
vortex ring at the i
th
control point. So, nally we have a system of linear equa-
tions Ax = b, which can be solved by using a standard technique like Gaussian Elimination. In
13
the current case, the inuence coefcient matrix was often singular. A best t solution using the
method of Least Squares was obtained. The function pinv in the SciPy Linear Algebra package
computes the best t solution using the Singular Value Decomposition method.
Next, we are interested in calculating the tangential velocity distribution at the panel control
points.
3.3 Modelling the Wake
An important aspect here is that of modelling the wake. Following Katz(1), the wake is mod-
elled as a series of horse-shoe vortices. To satisfy the Kutta-Condition at the trailing edge, it
is required that the net circulation at the trailing edge vanish. The procedure followed is thus.
Take a trailing edge panel. Now add a horse-shoe vortex of the same strength to cancel the ef-
fect of the triangular panel. In practice, each arm of the horse-shoe vortex is treated as straight
line vortex lament. 1.2.1
Figure 3.1: Horse Shoe Vortex
14
3.4 Secondary Computations
3.4.1 Velocity
The velocity at a control point is computed by calling the voring (short for Vortex Ring) sub-
routine, this time with the real vortex strengths, to get the induced velocity due to each panel.
The contribution from each panel is summed up to get the total induced velocity at a control
point. The free-stream velocity is added to the give the total velocity.
3.4.2 Coefcient of Pressure
The Coefcient of Pressure is,
C
p
=
p p

1
2
V
2

The use of Bernoullis Equation in the above equation for an inviscid, incompressible, steady
ow leads us to the simpliation of the equation to
C
p
= 1
V
2
V
2

3.4.3 Calculation of Lift


Once the circulation (vortex strengths) over each panel have been determined, the lift is com-
puted by summing up the compnents of the circlation that cause lift. This is the component of
the circulation which is perpendicular to the direction of the free-stream velocity. According to
the Kutta-Joukowski theorem, the lift force per unit span is
L = V

where,
is the free stream density of air
V

is the free-stream velocity


15

is the net circulation around the wing


Figure 3.2: Component of Circulation that Contributes to the Generation of Lift
16
CHAPTER 4
Results
The code was run for different cases. For each different characteristic lengths were used. For
the same geometry, different characteristic lengths were used. All results were visualized in
Mayavi2 (4). The aspect ratio of the wings in all three cases is 1.73
4.1 Rectangular Planform
Below are plots for C
p
, , and net circulation per panel for a rectangular planform. The Char-
acteristic Length used was 0.1. The mesh contained 1100 triangles. The angle of attack is
0

.
Figure 4.1: C
p
distribution over the wing. Rectangular.
Figure 4.2: distribution over the wing. Rectangular.
Figure 4.3: Net Circulation per Panel Distribution over the wing. Rectangular.
18
4.2 The Inverse Zimmerman Conguration
Below are plots for C
p
, , and net circulation per panel for the inverse zimmerman congura-
tion. The Characteristic Length used was 0.1. The mesh contained 1100 triangles. The angle
of attack is 0

.
Figure 4.4: C
p
distribution over the wing. Inverse Zimmerman
19
Figure 4.5: distribution over the wing. Inverse Zimmerman
Figure 4.6: Net Circulation per Panel Distribution over the wing. Inverse Zimmerman
20
4.3 The Zimmerman Conguration
Below are plots for C
p
, , and net circulation per panel for the zimmerman conguration. The
Characteristic Length used was 0.1. The mesh contained 1110 triangles. The angle of attack is
0

.
Figure 4.7: C
p
distribution over the wing. Zimmerman
21
Figure 4.8: distribution over the wing. Zimmerman
Figure 4.9: Net Circulation per Panel Distribution over the wing. Zimmerman
22
Presented below are the C
l
vs curves for the three respective planforms. The code was
run for angles 0, 2, 4, 6, 8, 10 degrees. The magnitudes are off; this part requiers more work.
Figure 4.10: C
l
vs curve. Rectangular.
Figure 4.11: C
l
vs curve. Zimmerman.
23
Figure 4.12: C
l
vs curve. Inverse Zimmerman.
24
CHAPTER 5
Concluding Remarks & Future Work
The Vortex Lattice Method was implemented using an unstructured mesh on rectangular, Zim-
merman and Inverse Zimmerman conguration planforms and an attempt was made to optimize
the planform shape based on the C
l
values. A code was written in Python using NumPy and
Scipy, and C
p
, C
l
, and velocity distribution over the wing was obtained. VLM is much
faster than the more involved Finite Volume Method, and can be used to obtain quick initial
estimates.(1)
Improvements include using higher order approximations in representing the mean sur-
face and nally, directly modelling the actual surface using higher order elements. For two-
dimensional ows, Ramachandran et. al (6) showed that using cubic panels for representing
the geometry, and using linear vortex strength variation on a panel, improves the accuracy of
the solution and accurate results can be obtained using fewer panels.
APPENDIX A
Code Snippets
Listing A.1: Inuence Coefcient Calculations
for i in range(N):
panel_i = TriPanel(i, 1.0, points_list, triangle_list)
ctrl_pt[i] = panel_i.ctrl_point()
S[i] = panel_i.area()
area_of_panel[i] = norm(S[i])
n_cap[i] = S[i]/area_of_panel[i]
rhs[i] = np.dot(-v_inf, n_cap[i])
for j in range(N):
if is_a_te_triangle(j, te_triangles)==1:
wake_vel = wake_influence(j, 1.0, ctrl_pt[i])
panel_j = TriPanel(j, 1.0, points_list, triangle_list)
vel = panel_j.voring(1.0, ctrl_pt[i])
total_vel = vel + wake_vel
coeff[i][j] = np.dot(total_vel, n_cap[i])
Here, in the nested for loop, the inuence coefcients are computed and the rhs matrix is
set up. The outer loop cycles through all the control points, whereas the inner loop computes
the inuence coefcients. TriPanel is a class that contains method voring (vortex ring). Every
call to TriPanel sets up a triangle based on the input vtk data. voring calls function vortex line
three times, once for each arm of the (triangular) vortex ring. The inuence coefcient is the
dot product of the induced velocity due to the j
th
vortex ring and the normal of the i
th
panel.
REFERENCES
[1] Joseph Katz and Allen Plotkin, Low-Speed Aerodynamics, pp. 378 412, (1991).
[2] John D. Anderson, Fundamentals of Aerodynamics, pp. 115-173, (1991).
[3] Charles L. Ladson and Darrell W. Sproles, Computer Program To Obtain Ordinates for
NACA Airfoils, NASA Technical Memorandum 4741, 91-1596, (1996).
[4] Prabhu Ramachandran and Ga el Varoquaux, The Mayavi data visualizer, http://
code.enthought.com/projects/mayavi, (2005).
[5] Christophe Geuzaine and Jean-Francois Remacle, Gmsh: a three-dimensional nite ele-
ment mesh generator with built-in pre and post-processing facilities, International Journal
for Numerical Methods in Engineering, Vol. 79, Issue 11, pp 1309-1331, (2009).
[6] Prabhu Ramachandran, S. C. Rajan and M. Ramakrishna, A Fast Multipole Method For
Higher Order Vortex Panels In Two-Dimensions, SIAM Journal on Scientifc Computing,
Vol. 26, no. 5, 1620-1642, (May 2005)
[7] Martin Hepperle, http://www.mh-aerotools.de/airfoils/mh60koo.htm
27

Você também pode gostar