You are on page 1of 19

p a s q u a l e @v t .

e d u

2011 May
Aerodynamic and Structural
Optimization of a General
Aviation Trainer Wing
AOE 5064 Final Project
Anthony Ricciardi
VirginiaTech
Abstract
A real world design evolution was used as inspiration for formulating a wing optimization
problem. Background information on the real life program is presented. An approach to
optimizing a general aviation trainer wing for minimum drag is formulated. A code that
joins a lifting line model, parasite drag model, and trim solver is used to calculate drag and
aerodynamic loads. This is coupled with empirical wing weight equations and a gradient
based optimizer to to solve the shape optimization problem. The internal structure of the
optimized shape is then sized according to stress constraints while subject to four ight
loading conditions. Results will show a signicant decrease in wing drag by evolving from
an untapered, untwisted rectangular wing, to a twisted and tapered wing with a slight
increase in aspect ratio. It is concluded that feedback from the structural sizing to the
aerodynamic shape optimization is not fully captured by the empirical weight equations.
Incorporating direct coupling of the analysis would be an obvious next step.
2
Contents
1 Introduction 4
1.1 Problem: Piper PA - 28 Wing Upgrade . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Aerodynamic and Loads Analysis . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Structural Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.3 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Design Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Shape Design Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Sizing Design Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Flight Loading Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 Analysis and Optimization Approach 9
2.1 Aerodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Structural Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Aeroelastic Inputs (Not Used) . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Stress Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Shear Stress Due to F
z
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.4 Shear Stress Due to F
y
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.5 Shear Stress Due to M
x
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.1 Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.2 Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Empirical Weight Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.1 Approximate Weight Method . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.2 Statistical Weight Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Results and Discussion 15
3.1 Using Approximate Weight Method . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Using Statistical Weight Method . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4 Conclusions 19
3
1 Introduction
1.1 Problem: Piper PA - 28 Wing Upgrade
In 1976 Piper Aircraft evolved the PA-28-160 Piper Cherokee to the PA-28-161 Piper Warrior
II. Signicant modications were an increase in gross weight, an additional row of windows in
the aft cabin, resizing of the horizontal tail, and modication of the outboard wing geometry and
wing structure. This work focuses on the wing redesign. The design variables were formulated
to match those that the Piper design team worked with. Figure 1 highlights the changes in the
planform.
46
45
PA-28-160 PA-28-161
176 15
53 36.2
Figure 1: Top view of the Piper PA-28-160 (right half) and the upgraded PA-28-161 (left half)
[2] [1].
The inboard wing geometry houses fuel tanks and other systems. Changes to the inboard
wing geometry would propagate into required changes of many other systems that are housed
in that section. Re-certifying systems and losing commonality with other models would signif-
icantly drive up the cost of the program. These cost factors are most likely why Piper decided
to limit the scope of the geometry changes to the outboard wing. The aircrafts performance is
very sensitive to changes in the geometry of the outboard wing. So large improvements could be
achieved even with the limited scope (especially when starting with an untwisted, untapered,
rectangular Hershey bar wing).
The taper, twist, and structure of the outboard wing were optimized. The inboard outer
mold line was xed, but the structural loading changed with the gross weight increase and the
geometry modications. Therefore the structure of the inboard wing was also optimized. The
exact objective and constraints that were used by the Piper design team are unknown. The
4
objective used in this study was to minimize drag at cruise. The design was constrained by
limiting the allowable stress in the material at several key loading conditions.
The shape and sizing portions of the optimization were decoupled. Ultimately only one way
coupling was used. The following sequence was followed for the optimization:
1. Optimize the shape variables for minimum drag using an empirical equation for the wing
structure and wing system weight.
2. Optimize the sizing variables for minimum weight based on loading conditions dependent
on the shape design.
The big assumption made: It was assumed that the structural component of the em-
pirical weight equation used in step one is accurate enough to assume that it is
equal to the optimized structural weight from step 2. Unfortunately, the structural and
system components are not given separately by the empirical equations so there is no way to
quantify this. It is also unfortunate that the structural weight calculation does not truly eect
the drag calculation. Based on the results it can be assumed that this is a mediocre assumption
at best.
Nevertheless, the assumption allows one way coupling of the optimization, and it allows
the minimum weight structural sizing optimization to be a subproblem of the minimum drag
shape optimization (i.e. the minimum drag structural design is the minimum weight design
that satises the stress constraints).
1.2 Background
The section provides a brief introduction to the methods used. More details are included in the
Analysis and Optimization Approach section.
1.2.1 Aerodynamic and Loads Analysis
Several key analysis components were needed to determine the trim lift and drag at cruise, and
the aerodynamic ight loads.
1. Three dimensional lift and lift-induced drag model
2. Parasite drag model
3. Trim calculator
These capabilities are integrated in a special version of the NATASHA aeroelastic solver. The
analysis components are based on the following models
1. Lifting line aerodynamic model
2. Skin friction coecient parasite drag model
3. Newton-Raphson aeroelastic system solver including trim calculation
The NATASHA code was used to calculate the drag objective as well as the aerodynamic, and
gravitational loads required for the constraints.
5
1.2.2 Structural Analysis
A method to convert loads to stresses was required. A wing box structural cross section was
assumed. A formulation was derived to go from cross-sectional forces and moments to stresses
at key points. A von-Mises stress constraint based on a percentage of the material yield stress
was enforced.
Other options for the structural analysis could include using VABS to recover the stresses,
or using a 3-D nite element structural model.
1.2.3 Optimization
Dierent gradient based optimization routines were considered and tested. The results in this
document were produced using MATLABs fmincon optimizer with the active-set algorithm.
Gradients were provided from a mixture of complex-step and analytical derivations.
1.3 Design Variables
1.3.1 Shape Design Variables
For the shape design problem the design variables are the outboard span (b
3
), tip chord (c
3
),
and the tip twist (
g
). There will be linear twist from zero twist at BL 93 to the tip.
c
2
=53
c
3
35%c

25%c

g
b
3
c
1
=73
b
1
=5
210
8
Elastic Axis
Aerodynamic
Center
b
2
=29
MLG Wheel and
Strut Attach Point
BL 0
BL 60 BL 93
Figure 2: Parameterized starboard wing of the aircraft. The red variables are the design
variables.
The aerodynamic center is at the quarter chord of the wing. The elastic axis is at the 35% chord
from BL 60 outboard. Inboard of BL 60 the elastic axis moves back relative to 35% chord and
remains strait. Notice the location of the main landing gear and strut attach point in Figure 3.
6
1.3.2 Sizing Design Variables
The wing box spar thickness (t
s
) and web thickness (t
w
) were designed at 21 cross-sections
across the semi-span. The thickness are assumed to be linearly varying in between sections.
The length dimensions of the cross-section are specied by the local chord length as shown in
Figure 3.
c

35% c


10% c

127
95
Figure 3: Parameterized wing cross section. The red variables are the design variables.
1.4 Objective
A coupled aerodynamic and structural optimization is a meaningful way to minimize the drag
of an aircraft. Consider Equation 1
C
D
= C
D
0
+
C
2
L
eAR
(1)
where
C
D
is the overall drag coecient
C
D
0
is the zero-lift drag coecient
C
L
is the aircraft lift coecient
e is the Oswald eciency number
AR is the wing aspect ratio
At its heart, this optimization is a problem of nding the best combination of reducing the
numerator and increasing the denominator of the second term in Equation 1, while reducing
the rst term (C
D
0
, which will be proportional to the total surface area).
A lighter design will reduce the cruise lift coecient. A close to elliptical lift distribution
provided by linear taper and twist of the outboard wing will drive the Oswald eciency number
close to one. Increasing the aspect ratio will increase the denominator of the second term, but
it will increase the structural weight. More structural weight will drive up C
L
.
Equation 1 is not evaluated directly in the analysis; it is referenced because it provides a
simple understanding for the drag trade-o problem.
1.5 Constraints
A two dimensional von-Mises stress constraint is checked at critical points.

vm
=
_

2
n
+ 3
2
(2)
7
Where
n
is the normal stress in the beam axis, and is the shear stress. Details are given
in the Analysis and Optimization Approach section.
1.6 Flight Loading Conditions
A set of four ight loading conditions were considered. First and second: 3.8g symmetric
maneuver loads at maneuvering speed (V
A
) and never exceed speed (V
ne
). Then a quasi-static
gust load at maximum structural cruising speed (V
no
). The quasi-static gust load factor was
calculated according to Part-23 certication requirements.

g
=
2W
gS cC
L
(3)
K
g
=
0.88
g
5.3 +
g
(4)
n = K
g
U
0
V C
L
2W/S
(5)
Notice that the gust load factor will be dependent on the wing shape design. Only positive
maneuver loads are considered because the wing box structure being designed is symmetrical
about the horizontal centroidal axis. Figure 4 shows these rst three loading conditions in the
classical V n diagram.
0 50 100 150
-1
0
1
2
3
4
5
V-n Diagram, PA-28-160
Airspeed KCAS
L
o
a
d

F
a
c
t
o
r
V
A
V
no
V
ne
76.2
123
Figure 4: Classical V n diagram for aerodynamic loading conditions. Green circles indicate
symmetric maneuver load cases. The purple circle is the quasi-static gust load case (load factor
will change depending on design). The red dot is the cruise point; where design is optimized
for minimum drag.
A symmetric hard landing load was also considered. The stresses were checked with the main
landing gear loaded at 3.5g. The wing was xed at the root. See Figure 5 for an illustration.
8
3.5g
127
180
Figure 5: Symmetric hard landing load. Static case with the wing xed at the root.
2 Analysis and Optimization Approach
2.1 Aerodynamics
As discussed in Section 1.2, the aerodynamic model is based on the following three components:
1. Lifting line aerodynamic model
2. Skin friction coecient parasite drag model
3. Newton-Raphson aeroelastic system solver including trim calculation
These capabilities are integrated in a special version of the NATASHA aeroelastic solver. the
following aerodynamic coecients are assumed:
C
l
0
C
l
C
l

C
d
0
C
m
0
C
m

0.0 2 1 0.01 0.025 -0.25


The eects of the fuselage are ignored so incidence angle and C
l
0
are not signicant.
2.2 Structural Analysis
Linearly varying box beam structure is assumed as the wing spar cross-section. Figure 6 shows
the layout and cross-section variables.
9
64
65
Figure 6: Box beam cross-section.
The thickness and length are specied as 10% and 35% chord respectively. The spar and
web thickness (in red) are design variables. The forces and moments (in purple) are provided
by the loads analysis. The points AH (in green) are the points where the stress constraints
were enforced at 21 cross-sections across the semi-span.
2.2.1 Aeroelastic Inputs (Not Used)
NATASHA can solve for aeroelastic trim (and dynamic) deformations given beam properties
for the structure. The required inputs were formulated for the box beam case based on the
design variables. The following inputs would be required:
E Elastic Modulus
G Shear Modulus
J
x
Polar Moment Of Inertia About x-axis
I
y
Moment of Inertia About y-axis
I
z
Moment of Inertia About z-axis
Elastic and shear modulus are given by the material properties. The moments of inertia are
derived based on the thin wall assumption.
J
x
=
2
2

ts
+

tw
(6)
I
y
=
2t
w

3
12
+ 2t
s

2
_
2
(7)
I
z
=
2t
s

3
12
+ 2t
w

2
_
2
(8)
10
2.2.2 Stress Constraints
The stress constraint was that the von-Mises stress at any point in the material must be less
than 50% of the yield stress of 2024 aluminum (.5 50ksi = 25ksi =
allow
). The 50% buer
was added to help account for fatigue and buckling constraints which have been omitted. The
von-Mises stress was calculated using:

vm
=
_

2
n
+ 3
2
(9)
Where
n
is the normal stress in the beam axis, and is the shear stress.
106
148
Figure 7: Cross-section stresses used for stress constraint.
Figure 7 shows the sign convention used for the stresses. Three components contribute to
the shear stress: F
y
, F
z
and M
x
. A clockwise positive (about the beam axis) sign convention
is used for the shear stress.
Forces and moments are provided by the analysis. The stresses are determined using

n
=
M
y
z
I
y

M
z
y
I
z
+
F
x
A
(10)
= max
_
_

Fz
+
Mxz
_
,
_

Fy
+
Mxy
__
(11)
The area is
A = 2(t
s
+t
w
) (12)
Certain values of the considered points:
Point y z
A /2 /2
B 0 /2
C /2 /2
D /2 0
E /2 /2
F 0 /2
G /2 /2
H /2 0
11
2.2.3 Shear Stress Due to F
z
102
95




EA
y
z
V
z
s
A
H
B
C
D
E F G
Figure 8: Shear ow for shear stresses due to F
z
.
=
V
z
I
y
_
s
0
z(s)ds (13)

G
=
F
z
I
y
_

2
_

2
__
(14)

H
=
F
z
I
y
_

2
_

2
_


2
2
_

2
__
(15)

A
=
G
(16)

B
= 0 (17)

C
=
G
(18)

D
=
H
(19)

E
=
G
(20)

F
= 0 (21)
2.2.4 Shear Stress Due to F
y
131




EA
y
z
V
y
s
A
H
B
C
D
E F G
24
Figure 9: Shear ow for shear stresses due to F
y
.
12
=
V
y
Iz
_
s
0
y(s)ds (22)

C
=
F
y
I
z
_

2
_

2
_
_
(23)

B
=
F
y
I
z
_

2
_

2
_


2
2
_

2
_
_
(24)

A
=
C
(25)

D
= 0 (26)

E
=
C
(27)

F
=
B
(28)

G
=
C
(29)

H
= 0 (30)
(31)
2.2.5 Shear Stress Due to M
x
127
95




EA
y
z
s
M
x
A
H
B
C
D
E F G
Figure 10: Shear ow for shear stresses due to M
x
.
=
M
x
2 min(t
s
, t
w
)
(32)

Mxy
=
M
x
2t
s
(33)

Mxz
=
M
x
2t
w
(34)
2.3 Optimization
2.3.1 Process
The overall objective of the optimization process it to produce the minimum drag design. It is
a two stage process outlined in Figure 11.
13
Initial Design
Optimized Shape
Optimized Structure
Shape Optimization
Sizing Optimization
Emp. Weight Aerodynamics


Structure Loads
Min
Drag
Min
Weight
31
Step 1
Step2
One Way
Figure 11: Two stage optimization process with one way coupling.
First the shape variables where optimized for minimum drag using an empirical equation for the
wing weight. Then the structural sizing variables were optimized for minimum weight based on
loading conditions which are dependent on the shape design. The optimized structural weight
never feeds back to the shape optimization. It is assumed that the structural component of the
empirical wing weight is accurate.
The optimizer used was MATLABs fmincon with the active-set algorithm. Gradients
were provided.
2.3.2 Gradients
The analytical sensitivity of the cross-sectional area (objective function used the structural
sub-problem) with respect to the structural sizing variables is:
A
t
w
= .2c (35)
A
t
s
= .7c (36)
The remaining gradients are found using complex step. For cases that NATASHA was used,
the complex step sensitivities were integrated into the analysis. The complex step was applied
to the system linearized about the solution. This meant that the nonlinear system only needed
to be solved once to get the gradients.
14
2.4 Empirical Weight Equations
2.4.1 Approximate Weight Method
The Approximate Weight Method equation is taken from Table 15.2 of reference [3].
W
wing
= 2.5S (37)
Where S is the wing planform area. This equation is eected by the span and chord design
variables, but does not capture their eects independently. The equation will not be aected
by the twist design variable.
2.4.2 Statistical Weight Method
This method is also taken from reference [3]. On page 476 there is a list of statistically based
general aviation component weights. The wing weight equation is:
W
wing
= 0.036S
0.758
W
0.0035
fuel
AR
0.6
e
AR/30
q
0.006

0.04
(100 t/c)
0.3
(N
z
W
dg
)
0.49
(38)
Where W
fuel
is the fuel weight, q is the design dynamic pressure, is the taper ratio, t/c is the
thickness to chord ratio, N
z
is the ultimate load factor, and W
dg
is the design gross weight.
The span and tip chord deign variables eect this equation more directly though the aspect
ratio and taper ratio terms. However, the tip twist still does not eect the weight.
3 Results and Discussion
3.1 Using Approximate Weight Method
Figure 12 shows a table of results from the described optimization scheme using the empirical
wing weight equation from the Approximate Weight Method. For the Piper designed PA-28-
160 and PA-28-161, results are shown for calculated drag, empirical wing weight, and optimized
structural wing weight (The Piper designed shapes were used as an input to the sizing opti-
mization). The tip twists for each planform are also shown; the tip twist of the PA-28-161 was
not known, so it was optimized for minimum drag (not likely the actual PA-28-161 tip twist,
this best case scenario drag wise for this analysis). The power requirement to overcome wing
drag is on the order of 20% less for the PA-28-161 than for the PA-28-160.
It was expected that for each individual wing the empirical wing weight would be higher
than the optimized structural wing weight because the empirical weight includes the structural
weight plus the weight of everything else in the wing. Strangely, the empirical weight calculation
shows the PA-28-161 wing being heavier than the PA-28-160 wing, but the structural sizing
optimization found a lower structural weight for the PA-28-161. This seems unlikely. It is
suspected that the discrepancy is due to the fact that the empirical weight equation does not
correctly capture the eect that changes in the design variables will have on the structural
weight. The untwisted untapered wing will have more load near the wing tips. Changes in
load near the tip can have large eect on the bending moments at the wing root. The inboard
cross-sectional forces and moments end up being larger for the PA-28-160 than for the other
two wings. Evidence of this is provided in Figure 13; the optimizer added more area to the
cross sections of the PA-28-160 inboard wing than the other two. Of course the addition of
more loading conditions could change the results.
15
Original Method
PA-28-160
PA-28-161
Shape
Optimized
Model Planform
53
36
29
15
176
188
123.5 lbs
41.7 hp
Emp. Weight
Str. Weight
Drag
Power Req.
155.7 lbs
52.5 hp
119.2 lbs
40.2 hp
424 lbs
192 lbs
452 lbs
114 lbs
460 lbs
144 lbs
AR = 5.4
AR = 6.9
AR = 7.6
Tip Twist
0
o

-2.6
o

-0.6
o

34
18
Figure 12: Results using the empirical weight equation from the Approximate Weight Method
for the optimization. Emp. Weight is the weight calculated using the empirical equation. Str.
Weight is the structural weight calculated from the sizing optimization.
Note the results of the shape optimization. The result is a higher aspect ratio, more tapered,
and less twisted wing than the PA-28-161. Although the design variables are noticeably dier-
ent, the power required advantage is less than two horsepower. The sensitivity of the objective
in the the design space between the PA-28-161 and the shape optimized wing is much less that
in the design space between the PA-28-160 and the PA-28-161 wing. This would indicate that
correctly capturing the structural weight (and everything else eecting the objective function)
is even more important in this region. The empirical weight of the shape optimized wing is
only slightly higher than that of the PA-28-161, but the optimized structural weight is roughly
20% higher. A better weight model for the shape optimization is warranted.
16
Original Method
83
0 2 4 6 8 10 12 14 16 18 20
0
2
4
6
8
10
12
14
Cross-sectional Area
Span Location, ft
A
r
e
a
,

i
n
2


PA-28-160
PA-28-161
PA-28-SO
0 2 4 6 8 10 12 14 16 18 20
0
0.05
0.1
0.15
0.2
0.25
Spar Thickness
Span Location, ft
t
s
,

i
n


PA-28-160
PA-28-161
PA-28-SO
Figure 13: Spanwise cross sectional areas and spar thicknesses.
The sized wing box is shown in Figure 13. A plot of the web thicknesses is omitted because
they all went to the minimum gage value of 0.01 inches. The PA-28-161 and the shape optimized
wing show very similar results. The PA-28-160 wing ends up being heavier overall due to the
additional cross-sectional area that the optimizer added to the inboard sections. The lack of
taper and twist at the tip led to the larger bending moments in the inboard sections.
3.2 Using Statistical Weight Method
The empirical wing weight equation from the Statistical Weight Method oers a more explicit
relationship to the span and tip chord design variables. However, the tip twist design variable
(structural weight critical) is still not accounted for. The results using the Statistical Weight
Method are disappointingly similar to the results using the Approximate Weight Method. It
is important to restrict comparisons in the sizing optimization and drag with the previous results
to qualitative thinking. The dierent empirical wing weight equation is introducing a change
in weight that is aecting the aircraft weight (lift induced drag) and the load calculations.
17
With Stat Method
Planform
53
36
211
15
176
180.4
121.2 lbs
40.9 hp
Emp. Weight
Str. Weight
Drag
Power Req.
143.1 lbs
48.3 hp

119.1 lbs
40.1 hp
291 lbs
147 lbs
366 lbs
117 lbs
382 lbs
128 lbs
AR = 5.4
AR = 6.9
AR = 7.4
Tip Twist
0
o

-2.6
o

-0.8
o

PA-28-160
PA-28-161
Shape
Optimized
Model
34
18
Figure 14: Results using the empirical weight equation from the Statistical Weight Method
for the optimization. Emp. Weight is the weight calculated using the empirical equation. Str.
Weight is the structural weight calculated from the sizing optimization.
The results are similar to the previous case. That shape optimized wing has a slightly smaller
span and larger chord. The comparisons in weight and drag are qualitatively the same.
With Stat Method
83
0 2 4 6 8 10 12 14 16 18
0
1
2
3
4
5
6
7
8
9
10
Cross-sectional Area
Span Location, ft
A
r
e
a
,

i
n
2


PA-28-160
PA-28-161
PA-28-SO
0 2 4 6 8 10 12 14 16 18
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Spar Thickness
Span Location, ft
t
s
,

i
n


PA-28-160
PA-28-161
PA-28-SO
Figure 15: Spanwise cross sectional areas and spar thicknesses.
The sizing results are qualitatively the same as the previous case.
18
4 Conclusions
An approach to optimizing a general aviation trainer wing has been presented. The objective
of minimizing drag was used for the aerodynamic shape optimization. The wing weight used
in the shape optimization was based on one of two empirical equations. The structural sizing
was done was done subject to a number of ight loading conditions. The feedback from the
structural sizing to the aerodynamic shape optimization is not totally captured by the empirical
weight equations. Incorporating direct coupling of the analysis would be an obvious next step.
Signicant reductions in wing drag were observed when moving from the at rectangular wing
to a tapered and twisted wing.
References
[1] Piper Aircraft Corporation. Piper PA-28-160 Cherokee Pilots Operating Handbook. 1972.
[2] Piper Aircraft Corporation. Piper PA-28-161 Warrior II Pilots Operating Handbook. 1982.
[3] D.P. Raymer. Aircraft design: a conceptual approach - 3rd ed. American Institute of
Aeronautics and Astronautics, 1999.
19

You might also like