You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264897040

A Numerical Study of the Aerodynamics of Cessna 172 Aircrafts in Echelon


formation

Conference Paper · January 2014


DOI: 10.2514/6.2014-1107

CITATIONS READS
0 1,235

2 authors:

Guna Sekar M Rinku Mukherjee


Indian Institute of Technology Madras Indian Institute of Technology Madras
5 PUBLICATIONS   2 CITATIONS    20 PUBLICATIONS   64 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PhD project View project

All content following this page was uploaded by Rinku Mukherjee on 22 September 2015.

The user has requested enhancement of the downloaded file.


A Numerical Study of the Aerodynamics of Cessna 172
aircrafts in Echelon formation
M.Gunasekaran∗ and Rinku Mukherjee†
Indian Institute of Technology Madras, India, 600036

A vortex-lattice numerical scheme that uses a novel decambering technique to predict


post-stall aerodynamic characteristics is used to study the aerodynamics of tandem Cessna
172 aircrafts flying in echelon formation. Results like CL − α from the current method
are compared with experiment. Additional results like section Cl distributions over wing
spans, CM − α and CLW − αtw (i.e. CL of the leading aircraft for different angles of attack
of the trailing aircraft) and analysis for (-)ve y-offset, which are available only from the
current numerical method are reported that supplement the experimental results. Detailed
post-stall numerical analysis and the effect of chord-wise, span-wise and vertical offsets on
the aerodynamics of the formation are also reported.

Nomenclature
cl Local lift coefficient Subscript
cm Pitching Moment coefficient visc viscous
α Angle of attack sec section
∆Cm Change in cm lw leading wing
δ1 , δ2 Decambering functions tw trailing wing
θ1 , θ2 Angular locations max maximum
x2 Location on the chord
αCL =0Zero-lift angle of attack Abbreviations
dx Chord-Wise shift V LM Vortex Lattice Method for 3D
dy Span-wise shift LLT Lifting line theory
dz Vertical shift J Jacobian

I. Introduction
Formation flying has been attracting the attention of scientists for sometime now as an aircraft flying in
formation has certain aerodynamic advantages. A significant saving in fuel can be obtained by flying in the
upwash field created by the leading aircraft. This advantage is not just limited to the trailing aircraft but
also to the leading aircraft, which is also affected by the bound and trailing vortex system of the trailing
aircraft.
Formation flying of aircrafts is motivated by observing birds in nature, which fly in close formation.
Several theories have been proposed for this behavior among birds, namely, it is mere social behavior or they
fly in formation to improve visual communication or as an act of defense against predators etc. However, the
explanation, which is of utmost interest to aerospace engineers, is that such close-formation flying provides
some aerodynamic advantages.1
A bird produces lift by flapping its wings that form a closed loop vortex with wavy vortex lines at
the aircraft tip and spanwise waves generated at every feather tip. Lift is produced by the aircraft during
downstroke and thrust is generated as the aircraft pivots forward on its axis. During upstroke the aircraft
rotates the humerus about its axis and curves in towards the body. It then rises and extends for the next
downstroke.2 The advantage of flying in formation comes from the existence of rolled-up tip vortices of the
∗ PhD Scholar, Dept. of Applied Mechanics, aerogunasekar@gmail.com
† Assistant Professor, Dept.of Applied Mechanics, Associate Member, AIAA, rinku@iitm.ac.in

1 of 12

American Institute of Aeronautics and Astronautics


leading bird that create upwash on the trailing bird. The leading bird in a formation therefore varies the
flow condition of the trailing bird, both temporally and spatially. In other words, a flapping wing in a bird
will result in discrete and/or periodic shedding of vortices. This is different from that of fixed aircrafts which
shed a continuous sheet of vortices.
Therefore, the numerical analysis in this paper reports results that are obtained by changing the conditions
in which the trailing aircraft is operating. The results are compared with experiment.3 The current numerical
method provides additional information like section Cl distributions for both aircrafts in formation, Cm − α
results for the trailing aircraft in the formation, Clw − αtw results (i.e. CL of the leading aircraft for different
α of the trailing aircraft) and post-stall results when one or more aircrafts in the formation has stalled, which
are not available from experiment but enhance the understanding of the aerodynamics of formation flight.

II. Numerical Procedure


A vortex-lattice method algorithm based on the decambering approach4 is used for predicting formation
flight aerodynamics of aircrafts using known section data. Although the numerical code, VLM3D based on
this approach was originally developed with a view to predict post-stall aerodynamics of single wings or their
configurations, it has been found to be robust and powerful in the analysis of formation flight as well with
some modifications. The ability to change spatial offsets (chord-wise, span-wise and vertical), the ability
to change the angle of attack of the trailing aircraft for a particular angle of attack of the leading aircraft
and vice versa are incorporated into VLM3D to study formation flight aerodynamics. The unique feature
of formation flight is the interaction between the vortices and aircrafts (both leading and trailing), which
is captured well by the modified VLM3D. Post-stall results provide enhanced understanding of formation
flight aerodynamics.
The code VLM3D based on the decambering approach was developed wherein the chordwise camber
distribution at each section of the wing was reduced to account for the viscous effects at high angles of
attack. The approach uses either or both Cl and Cm section data and uses a two-variable function for
the decambering. In addition, unlike all earlier methods, the approach uses a multi-dimensional Newton
iteration that accounts for the cross-coupling effects between the sections in predicting the decambering
for each step in the iteration. The subsequent discussion illustrates briefly the decambering approach by
describing its application to model a two-dimensional flow past an example airfoil. The iterative approach
for three-dimensional geometries is discussed next.

A. Application to 2D Flow
The overall objective was to arrive at a scheme for incorporating the nonlinear section lift curves in wing
analysis methods such as LLT, discrete-vortex Weissingers method and vortex lattice methods. For this
purpose, it was assumed that the two-dimensional data (Cl and Cm as functions of α) for the sections
forming the wing were available from either experimental or computational results. The objective was that
for the final solution of the three-dimensional flow, the Γ distribution across the span would be consistent
with the distribution of the effective α for each section and the Cl and Cm for each section would be consistent
with the effective α for that section and the section Cl -α and Cm -α data.
This overall objective was achieved by finding the effective reduction in the camber distribution for each
section along the span. The typical flow past an airfoil at small angles of attack consists of a thin boundary
layer that remains attached to the surfaces of the airfoil. For these conditions, the Cl and Cm predicted
using potential flow analysis of the airfoil camberline agrees closely with the computational and experimental
results that account for viscosity as shown for α < 12o in Fig. 1.
With increasing angles of attack, the boundary layer thickens on the upper surface and finally separates,
as shown in Fig. 2 using both experiment and computation. It is this flow separation that causes the viscous
results for Cl and Cm to deviate from the predictions using potential flow theory for α > 12o as shown in
Fig. 1. The reason for the deviation can be related to the effective change in the airfoil camber distribution
due to the boundary-layer separation. If the decambering could be accounted for, then a potential-flow
prediction for the decambered airfoil would closely match the viscous Cl and Cm for the high-α flow past the
original airfoil shape. This decambering idea served as the basis for the formulation of the current approach
for the three-dimensional flow problem.
While the camber reduction due to the flow separation can be determined from computational flows,

2 of 12

American Institute of Aeronautics and Astronautics


−1
−5 0 5 10 15 20

angle of attack
Fig. 6 Cl curv
method with de
Fig. 4 Cl -α curves from potential and viscous
methods.
To verify the e
2.5 2.5 proach, the valu
0.04
viscous flow
potential flow viscous flow
potential flowflow
(decambered airfoil)
the viscous Cl -α
potential
2 2 viscous flow
potential flow (original camberline) the NACA-0012
0.03
∆C
l
ues were then ap
1.5
1.5
0.02 B C camberline for p

pitching moment coefficient


0012 airfoil using
1
A
6 shows for com
lift coefficient

1
0.01

lift coeficient
0.5
∆Cm Cl -α curve for th
0.5
0 result from XFO
0 parison for the C
0
−0.01 The agreement w
−0.5
fied that the two-
−0.5
−0.02 used to model non
−1
−5 0 5 10 15 20 curves for high an
−1 angle of attack
−5 0 5 10 15 20 −0.03
−5 0 5 10 15 20
angle of attack
Fig. 6 Cl curve obtainedangle using
of attacka potential-flow 0.04

(a) ∆Cl method with decambering. potential flo


Fig. 4 Cl -α curves from potential and viscous Fig. 5 Cm -α curves (b) ∆Cm
from potential and viscous 0.03
viscous flo
potential flo
methods. methods.
To verify the effectiveness of the decambering ap-
Figure 1: Residuals in Coefficients of Lift and Pitching Moment 0.02
proach, the values of δ1 and δ2 were calculated for 13

pitching moment coefficient


0.04 approximation for a flat plate with a flap deflection.
viscous flow the viscous Cl -α and Cm -α data shown earlier for
potential flow These values of δ1 and δ2 in radians for given ∆Cl and 0.01

the NACA-0012 airfoil in Figs. 4 and 5. These val-


0.03 ∆Cm have been derived and are presented in Eqs. 1
ues were then applied as a correction to the flat-plate 0
and 2. In these equations, θ2 is the angular location
camberline for potential flow analysis of the NACA-
0.02 of the start point in radians for the function shown in
0012 airfoil using a lumped vortex method.13 Figure
pitching moment coefficient

−0.01
Fig. 3 and can be expressed as shown in Eq. 3 in terms
6 shows for comparison the predicted potential flow
0.01 of the x-location of this start point, x2 . In the current −0.02
∆Cm Cl -α curve for the decambered airfoil with the viscous
work, x2 was arbitrarily assumed to be 0.8, although
result from XFOIL analysis. Fig. 7 shows the com-
0
any value from 0.5 to 0.9 seemed to work well. −0.03
−5 0
parison for the Cm -α curve with the viscous result.
−0.01 The agreement was seen to be very good, which veri-
fied that the two-variable ∆Cm function can be Fig. 7 Cm curv
δ2 = 1 decambering 1 (1)
2 − as
sinas2θwell method with de
−0.02 used to model nonlinear 4lift 2 sin θ2
pitching moment
curves for high angles of attack.
−0.03 (a) Flow Visualisation with Experiment (b) Flow Visualisation with Numerical Analysis To compare th
−5 0 5 10 15 20
∆Cl − [2(π − θ2 ) + 2 sin θ2 ]δ2 two-variable func
angle of attack 0.04
δ1 = (2)
Figure 2: Flow separation pastpotential
an flowairfoil
(decambered airfoil) 2π from the viscous
viscous flow
Fig. 5 Cm -α curves from potential and viscous 0.03 potential flow (original camberline) made for α of 1
methods. 0012 example. Th
0.02
θ2 = cos−1 (1 − 2x2 ); x2 = 0.8 (3) from the bounda
pitching moment coefficient

13
no such detailed
approximation for information
a flat plate withis available from wind
a flap deflection. tunnel results that typically provide only the Cl -α and
These values
Cm -α curves. 1 of δ and δ 2 in radians for given ∆C l and 5 of 11 0.01

∆CIn m have been derived and are presented in Eqs. 1


the current method, the effective decambering for an airfoil was approximated using of
a Aeronautics
function ofand Astronautics P
American Institute 0
and variables
two 2. In theseδ equations, θ2 is the angular location
1 and δ2 . The two linear functions shown in Fig. 3 were superposed to obtain the final
of the start point in radians for the function shown in
decambering function. Two variables were used because the decambering was being backed out from two −0.01
Fig. 3 and can be expressed as shown in Eq. 3 in terms
pieces of information: the Cl and x
of the x-location of this start point,
Cm from the airfoil data for the α under consideration. This approximation
2 . In the current −0.02
will, of course, not match the actual viscous
work, x2 was arbitrarily assumed to be 0.8, although decambering, but the objective was to find an equivalent camber
reduction to match
any value from 0.5 tothe
0.9 viscous
seemed to Clwork
and well.
Cm for the α under consideration. −0.03
−5 0 5 10 15 20

angle of attack

∆Cm Fig. 7 Cm curve obtained using a potential-flow


δ2 = 1 (1)
4 sin 2θ2 − 12 sin θ2 method with decambering.

To compare the approximate decambering from the


∆Cl − [2(π − θ2 ) + 2 sin θ2 ]δ2 two-variable function with the actual decambering
δ1 = (2)
2π from the viscous solution, comparison plots have been
made for α of 10, 16 and 18 deg for the NACA-
0012 example. The actual decambering was computed
−1
θ2 = (a)
cosDecambering function
(1 − 2x2 ); x 2 = 0.8 δ1 (3) (b) Decambering
from the boundary-layer function thickness
displacement δ2 dis-

Figure5 3:
of Decambering
11

American Institute of Aeronautics and Astronautics Paper 2003–1097


Figure 3 shows the schematic diagram of modified camberline using decambering function((-)ve as shown).

3 of 12

American Institute of Aeronautics and Astronautics


−1
−5 0 5 10 15 20

−1 angle of attack
−5 0 5 10 15 20

angle of attack
Fig. 6 Cl curve obtained using a potential-flow
method with decambering.
Fig. 4 Cl -α curves from potential and viscous
methods.
The procedure for calculating the values of δ1 and δ2 for To
obtaining theeffectiveness
decamberingof function for an airfoil
verify the the decambering ap-
at a given α can be summarized as follows: proach, the values of δ1 and δ2 were calculated for
0.04
viscous flow the viscous Cl -α and Cm -α data shown earlier for
1. Evaluate the viscous Cl and Cm for the α from experimental
potential flow
or computational
the NACA-0012 airfoil in Figs.data for5.theThese
4 and airfoil.
val-
0.03
ues were then applied as a correction to the flat-plate
2. Obtain the corresponding potential flow data using a lumped vortex model of the actual camberline
camberline for potential flow analysis of the NACA-
for the airfoil.
0.02
0012 airfoil using a lumped vortex method.13 Figure
pitching moment coefficient

3. 0.01Compute the difference between the viscous and the6 potential


shows forflow
comparison the predicted
results: ∆C potential flow
l = (Cl )visc - (Cl )potential
∆C Cl -α curve for the decambered airfoil with the viscous
and ∆Cm = (Cm )visc -(Cm )potential . m

0 result from XFOIL analysis. Fig. 7 shows the com-


4. These differences are shown schematically in Fig. 1parison
for an for the Cm -α airfoil
NACA-0012 curve analyzed
with the using
viscousXFOIL.
result.5
−0.01 The agreement was seen to be very good, which veri-
5. Use the difference between the viscous and the potential
fied thatflow results to calculate
the two-variable the values
decambering of can
function δ1 and
be
δ2 as explained below.
−0.02 used to model nonlinear lift as well as pitching moment
curves for high angles of attack.
−0.03
−5 0 5 10 15 20
0.04
2.5
angle of attack
potential flow (decambered airfoil)
viscous flow
Fig. 5 Cmpotential
-α flow curves
viscous flow
from potential and viscous
(decambered airfoil) 0.03 potential flow (original camberline)
2
methods. potential flow (original camberline)
∆C 0.02
l

pitching moment coefficient


1.5
13
approximation for a flat plate with a flap deflection.
B C

These 1values of δ1 and δ2 in radians for given ∆Cl and 0.01

A
∆Cm have been derived and are presented in Eqs. 1
lift coeficient

0
and 2.0.5 In these equations, θ2 is the angular location
of the start point in radians for the function shown in −0.01
Fig. 3 0and can be expressed as shown in Eq. 3 in terms
of the x-location of this start point, x2 . In the current −0.02

work,−0.5x2 was arbitrarily assumed to be 0.8, although


any value
−1
from 0.5 to 0.9 seemed to work well. −0.03
−5 0 5 10 15 20
−5 0 5 10 15 20 angle of attack
angle of attack
15 20

Fig. 6 (a) ∆C
δ2 = obtained
Cl curve
m
Cl using (1)
a potential-flow
Fig. 7 Cm curve obtained
(b) Cm using a potential-flow
1 1 method with decambering.
4 sin 2θ2 −
method with decambering. 2 sin θ2
l and viscous Figure 4: Effectiveness of the Decambering Method
To compare the approximate decambering from the
To verify the effectiveness of the decambering ap-
∆Cl − [2(π − θ2 ) + 2 sin θ2 ]δ2 two-variable function with the actual decambering
proach, the values
δ1 =
The effects of δ1 of
andδ1 δ2and
on δ2 were calculated for(2)
2π the change in Cl and Cm for from a the
given α cansolution,
viscous be computed reasonably
comparison well been
plots have using
the airfoil
thin Cl -α and
viscoustheory and aCthree-term
m -α data shown
Fourierearlier for
series approximation for a flat plate with a flap deflection. 6
These
made for α of 10, 16 and 18 deg for the NACA-
the NACA-0012
values of δ1 and δairfoil in Figs. 4 and 5. These val-
2 in radians for given ∆Cl and ∆Cm are 0012presented
example.inThe Eqs. 1 anddecambering
actual 2. In thesewas
equations,
computedθ2
isues
thewere then location
angular applied−1asofathe
correction
start
θ2 = cos (1 − 2x2 ); x2 = 0.8
to the
point in flat-plate
radians
(3)for the function δ shown in Fig. 3 and can be
from the boundary-layer displacement thickness dis-
2 expressed
ascamberline
shown in forEq.potential
3 in termsflowofanalysis of the NACA-
the x-location13 of this start point, x2 . In the current work, x2 was arbitrarily
0012 airfoil using a lumped vortex method. Figure of 11
assumed to be 0.8, although any value from 0.5 to 50.9 seemed to work well.
6 shows for comparison the predicted potential flow
American Institute of Aeronautics and Astronautics Paper 2003–1097
∆Cm Cl -α curve for the decambered airfoil with the viscous ∆Cm
result from XFOIL analysis. Fig. 7 shows δthe 2 = com-
1 1 (1)
parison for the Cm -α curve with the viscous result. 4 sin2θ2 − 2 sinθ2
The agreement was seen to be very good, which ∆l −veri-
[2(π − θ2 ) + 2sinθ2 ]δ2
δ2 =
fied that the two-variable decambering function can be 2δ (2)
2
used to model nonlinear lift as well as pitching moment
curves for high angles of attack. θ = cos−1 (1 − 2x ); x = 0.8 (3) 2 2 2
15 20

To verify the effectiveness of the decambering approach, the values of δ1 and δ2 were calculated for the
0.04

viscous Cl −α and Cm −α data and were then applied as a correction to the flat-plate camberline for potential
potential flow (decambered airfoil)
viscous flow
l and viscous 0.03
flow analysis of the NACA-0012 airfoil using a lumped vortex method.4 Figure 4 shows the comparison of
potential flow (original camberline)

the predicted potential flow Cl − α for the decambered airfoil with the viscous result from XFOIL analysis.
0.02
pitching moment coefficient

ap deflection.13 The agreement is seen to be very good, which verified that the two-variable decambering function can be
given ∆Cl and used to model nonlinear lift as well as pitching moment curves for high angles of attack.
0.01

nted in Eqs. 1
0
ngular location
ction shown in −0.01
Eq. 3 in terms
In the current −0.02
4 of 12
e 0.8, although
rk well. −0.03
−5 0 5 American
10 Institute
15 of
20 Aeronautics and Astronautics

angle of attack
B. Application to a Finite 3D Wing
The objective was to incorporate the two-variable decambering function into a three-dimensional analysis
method such as a vortex lattice method (VLM) in an iterative fashion. In a typical VLM, the wing is divided
into several spanwise and chordwise panels. Associated with each of these panels is a horseshoe vortex. In
the current approach, each spanwise section j (composed of a row of chordwise panels) had two variables, δ1j
and δ2j for defining the local decambering geometry. Unlike the two-dimensional case, where the δ1 and δ2
were selected to match the difference between the potential-flow and the viscous-flow results, in the three-
dimensional case, changing a δ on one section is likely to affect the neighboring sections and the sections on
the downstream lifting surfaces. To account for these effects, a 2N-dimensional Newton iteration was used
to predict the δ1 and δ2 at each of the N sections of the wing so that the ∆Cl and ∆Cm at these sections
approached zero with an increasing number of iterations. A 2N X 2N matrix equation has to be solved for
each step of the Newton iteration,7 as shown in Eq. 4.

J.δx = −F (4)
where F is a 2N-dimensional vector containing the residuals of the functions fi to be zeroed, δx is the
2N-dimensional vector containing the corrections required to the 2N variables xi to bring the vector F closer
to zero, and J is the 2N X 2N Jacobian of the system containing the gradient information. For each step of
the iteration, F and J are determined, and δx is computed using Eq. 4. The corrections are then applied to
the variables to bring the residuals closer to zero. In the current scheme, the residual functions fi were the
values of the ∆Cl and ∆Cm for each of the wing sections, and the variables xi were the values of δ1 and δ2
for each of the sections. The Jacobian can be partitioned into four submatrices as shown in Eq. 5. Equations
6- 9 show the elements of the four submatrices.
!
Jl1 Jl2
J= (5)
Jm1 Jm2
∂∆Cli
(Jl1 )i,j = (6)
∂δ1,j
∂∆Cmi
(Jm1 )i,j = (7)
∂δ1,j
∂∆Cli
(Jl2 )i,j = (8)
∂δ2,j
∂∆Cmi
(Jm2 )i,j = (9)
∂δ2,j
The iteration procedure can be summarized as follows:

1. Assume starting values of δ1 and δ2 for each section of the wing.

2. Compute the wing aerodynamic characteristics using the VLM code.


3. Compute the local section effective angles of attack αsec using the local section coefficient of lift, Clsec
as shown in Eq. 10. It is to be noted here that in Eq. 10, the variables δ1 , δ2 and θ2 are defined for each
section of the wing and are equivalent to those used earlier for the two-dimensional case in Eqs. 1- 3.

4. Compute the residuals ∆Cl = (Cl )visc -(Cl )sec and ∆Cm = (Cm )visc -(Cm )sec . The (Cl )visc and
(Cm )visc are obtained from the known section data for the angle of attack corresponding to αsec .
5. Calculate the Jacobian matrix for the Newton iteration.
6. Solve matrix Eq. 4 to obtain the perturbations to δ1 and δ2 at each section and update values of δ1
and δ2 .
7. Repeat steps 2-6 until ∆Cl and ∆Cm are close to zero within a specified tolerance.

5 of 12

American Institute of Aeronautics and Astronautics


(Cl )sec θ2 sinθ2
αsec = − δ1 − δ2 [1 − + ] (10)
2π π π
It must be mentioned that for cases where the experimental/computational viscous data for the airfoil
section does not have Cm or for cases where the decambering approach is applied to an analysis method
that cannot compute the section pitching moments (e.g. LLT or a discrete-vortex Weissinger method),
the decambering is modeled as a function of a single variable δ1 ; δ2 is assumed to be set to zero. In this
case, the viscous decambering function becomes similar to the α-reduction approach used by Tseng8 and
VanDam.9 However, in the current approach, the cross coupling between the sections was still accounted
for in predicting the δ1 values for the next step. In the earlier approaches, the sections are assumed to be
decoupled, and the δ1 values for each section are predicted using just the local values of the ∆Cl . For this
reason, it is believed that the current method will be more effective in handling situations where the section
flows are closely coupled.

III. Results
Results are now presented using VLM3D for similar configurations of tandem Cessna 172 aircrafts used
by Bangash3 to conduct experiments.

Figure 5: Echelon configuration used in VLM3D

VLM3D requires the 2D airfoil data, i.e. the Cl − α and Cm − α data to predict the results for a 3D wing
but no experimental 2D airfoil data is reported by Bangash.3 Hence, three different 2D data are used in the
analysis as described here: (a) simulated 2D nonlinear data of Selig11 (b) simulated 2D data of Bangash and
(c) wind tunnel data of NACA2412 airfoil.
To simulate the 2D data of Selig11 in (a) and that of Bangash in (b), the 3D baseline data for a single
wing is used as input into VLM3D and it is operated for a wing of infinite span. The output from VLM3D is
considered to simulate the 2D data and is further used for the configuration studies. The wind tunnel data
of NACA2412 airfoil in (c) is taken from Abbott and Von Doenhoff10 at Re ≈ 3.1X106 since the Cessna 172
uses a modified NACA2412 airfoil.
The wind-tunnel model has 3000 span and 4.500 chord and operated at Re ≈ 2.4X105 . To simulate the
same dimensions in VLM3D, tandem rectangular wings, each of AS≈7 are used. Figure 5 shows the geometry
definition used in VLM3D for the echelon formation.

6 of 12

American Institute of Aeronautics and Astronautics


A. Different 2D Cl -α input data on 3D CL -α curve
The 3D CL -α curves for the trailing aircraft from the current and experimental methods are shown in Fig. 6
for the longitudinal offset dy = 0, vertical offset dz = -0.11, angle of attack of the leading aircraft αlw = 10o
is unchanged and the angle of attack of the trailing aircraft undergoes a sequence αtw = −5o to 29o . Various
2D input data are used for VLM3D, which are also plotted along with the 3D results in Fig. 6.
The following observations are made:

1. The 2D result of Bangash3 shows a stall angle, αstall ≈13o and an alpha zero-lift, αCL=0 ≈−3o . The
3D result on the other hand shows αstall ≈14o and αCL=0 ≈−6.5o .
2. The 2D result of Selig shows an αstall ≈17o and the CLmax is less than the 2D result of Bangash.
3. The NACA2412 result from Abbott10 shows αstall ≈15o and αCL=0 ≈−2.5o .

Figure 6: CL − α of the trailing wing in Echelon formation using different 2D Cl − α input data
for VLM3D.

As noted above, the three 2D results are significantly different from each other. The 3D results from
VLM3D using the above 2D results also corroborate that. Overall, the trailing wing CLmax is less than that
of the airfoil Clmax and the trailing wing stall is delayed compared to the airfoil.

B. Effect of using Cm − α along with Cl − α as input into VLM3D


Using the pitching moment, Cm − α data along with the Cl − α data of NACA2412 from Abbott10 as an
input into VLM3D results in some interesting results as listed below:

1. Wing CL − α curve
The Cl − α curves for the trailing aircraft from current method are shown in Fig. 7 by varying the span-wise
offset for dz = −0.11, constant angle of attack of the leading aircraft, αlw = 10o while the trailing aircraft
undergoes a sequence, αtw = −5o to 29o . It is seen that the CL − α curves do not vary significantly with a
change in the span-wise offset.
Overall, the trailing wing CLmax is less than that of the airfoil Clmax and the trailing wing stall is delayed
compared to the airfoil.

7 of 12

American Institute of Aeronautics and Astronautics


Figure 7: CL -α of the trailing wing in Echelon formation for different y-offsets using 2D Cl -α
and Cm -α as input data for VLM3D

2. Effect of varying chord-wise offset dx


The effect of varying the chord-wise offset, dx is shown in Fig. 8. It is seen that varying the chord-wise offset
does not affect the performance of the trailing wing B at all.

Figure 8: CL − α and section Cl due to varying dx

3. Effect of varying span-wise offset dy


The section Cl distribution over both wings is shown in Fig. 9 for dx = 0, dz = −0.11, αlw = 10o and the
angle of attack of the trailing wing undergoing a sequence. Results are shown for selected angles of attack
of the trailing wing, αtw = −3o , 10o , 18o and 21o , which throw some light on the pre-stall, near-stall and
post-stall regions of the flight.
The very first inference that is made from these plots is that unlike the wing CL -α results shown in section
(i), the section Cl distribution varies significantly with change in conditions. There is a marked asymmetry,
which decreases with increase in the y-offset. This is due to the asymmetric distribution of upwash caused
by the leading wing A on the trailing wing B, which is maximum on the left tip and gradually reduces to a

8 of 12

American Institute of Aeronautics and Astronautics


minimum on the right tip of wing B. Since the upwash is inversely proportional to distance, the upwash and
hence the asymmetry decreases with increase in the y-offset.

Figure 9: Section Cl distribution for varying (+)ve dy

Figure 10: Section Cl distribution for varying (-)ve dy

The trailing wing stalls at dy= 0, 0.44 and αtw = 21o , which is asymmetric unlike the conventional
root-stall of a rectangular wing. However, the same trailing wing does not stall at dy = 0.88 and αtw = 21o .
The leading wing remains unstalled for all cases.
The section Cl distribution over both wings for overlapping y-offsets or (-)ve dy is shown in Fig. 10. Here
the asymmetry is more pronounced with very sharp and sudden changes, which can result in strong rolling
moments and are a cause for potential dangerous situations. Here the trailing vortices of the leading wing
A cause downwash on the overlapping part of wing B and upwash on the remaining part. The trailing wing
stalls with asymmetric section Cl distribution on all the three cases shown.
Figure 11 shows the section Cl distribution over the wing-span of both wings, where both of them
undergo a sequence in the angle of attack. Selected results are presented here. Asymmetry in the section Cl
distribution is dominant for the trailing wing. The trailing wing stalls at αtw = 21o for all the cases but the

9 of 12

American Institute of Aeronautics and Astronautics


Figure 11: Section Cl distribution for varying αlw

nature of the asymmetric stall varies. At αlw = 22o , the leading wing also stalls and has symmetric root-stall.
This also induces an additional downwash on the trailing wing. This can be seen from more sections of the
trailing wing stalling at αtw = 21o .

Figure 12: CL − α and section Cl due to varying dz

4. Effect of varying vertical offset dz


The effect of varying the vertical offset, dz is shown in Fig. 12. As noticed in earlier case studies, the wing
CL does not show any marked difference but the section Cl shows asymmetry and is highly sensitive to a
small change in the vertical offset.

C. Effect of using different input 2D data in VLM3D


The effect of using different 2D input data, including with and without 2D Cm -α data is shown in Fig. 13.
It is seen that the 3D results are highly dependent on the 2D data. At αtw = 21o , only the 2D input data
with Cm -α shows expected stall.

10 of 12

American Institute of Aeronautics and Astronautics


Figure 13: CL -α and lift distribution curve of different 2D input in VLM3D for Echelon formation

IV. Conclusion
A thorough analysis of tandem Cessna wings in echelon formation is conducted to get insights into the
behaviour.
The dominant observation in the analysis of the echelon formation using VLM3D is the existence of
asymmetry in the section Cl distribution of the trailing aircraft. This asymmetry is pronounced when the
span-wise offset is nearly zero and decreases when the span-wise offset is large. Such asymmetry can result
in large rolling moments, which is undesirable. When negative horizontal offset is used sharp changes in
section Cl is seen. This too is undesirable. Increase in negative vertical offset results in increase in CL of
the trailing aircraft. However, this is accompanied by asymmetry in the section Cl distribution.
The results presented in this paper give aerodynamicists a host of data which can be used to design a
formation where the enhancements in the aerodynamics of an aircraft in echelon formation can be utilized
keeping in mind the trade-offs in terms of asymmetry in section Cl distributions resulting in rolling moments.
Post-stall analysis provided in this paper is further useful in determining the geometry of the formation for
best results.
In conclusion, VLM3D is a rapid tool that can be used both for analysis and preliminary design stages
of formation flight.

References
1 Seiler,P., Pant, A. and Hedrick, K., Analysis of bird formations, Proceedings of 41st IEEE Conference on Decision and
Control, Las Vegas, Nevada, USA, December, 2002.
2 Poore, S. O., Sanchez-Heiman, A. and Goslow, G. E. Jr. Aircraft upstroke and evolution of flapping flight, Nature, Vol

387, pp 799-802, June 1997.


3 Bangash, Z. A., Sanchez, R. P. and Ahmed, A. Aerodynamics of Formation Flight, Journal of Aircraft, Vol 43, No. 4, pp

907-912, July-August 2006.


4 Mukherjee, R., and Gopalarathnam, A., Post-Stall Prediction of Multiple-Lifting-Surface Configurations Using a Decam-

bering Approach, Journal of Aircraft, Vol. 43, No. 3, pp. 660-668, May-June 2006.
5 Drela, M., XFOIL: An Analysis and Design System for Low Reynolds Number Airfoils, Low Reynolds Number Aerody-

namics, edited by T. J. Mueller, Vol. 54 of Lecture Notes in Engineering, Springer-Verlag, New York, June 1989, pp. 1-12.
6 Katz, J. and Plotkin, A., Low-Speed Aerodynamics From Wing Theory to Panel Methods, McGraw-Hill, Inc., 1991.
7 Press, W. H., Teukolsky, S. A., Vetterling, W. T., and Flannery, B. P., Numerical Recipes in Fortran-The Art of Scientific

Computing, Cambridge University Press, New York, 2nd ed., 1992.


8 Tseng, J. B. and Lan, C. E., Calculation of Aerodynamic Characteristics of Airplane Configurations at High Angles of

Attack, NASA CR 4182, 1988.


9 van Dam, C. P., Kam, J. C. V., and Paris, J. K., Design-Oriented High-Lift Methodology for General Aviation and Civil

Transport Aircraft, Journal of Aircraft, Vol. 38, No. 6, Nov-Dec 2001, pp. 10761084.

11 of 12

American Institute of Aeronautics and Astronautics


10 Abbott, Ira H., and Von Doenhoff Albert E. (1959). Theory of Aircraft Sections, Dover, Mineola, NY.
11 Roskam, J., Airplane Flight Dynamics and Automatic Flight Controls, Part I, DARC Corp.,Lawrence, KS, 1995.

12 of 12

American Institute of Aeronautics and Astronautics

View publication stats

You might also like