Sei sulla pagina 1di 36

Aircraft Design with Active

Load Alleviation and Natural Laminar Flow


Jia Xu

and Ilan Kroo



Stanford University, Stanford, CA, 94305, U.S.A
We integrate maneuver load alleviation (MLA), gust load alleviation
(GLA) and natural laminar ow (NLF) into aircraft conceptual design. The
concurrent design of the aircraft and its active load alleviation systems can
yield signicant gains. The simultaneous application of MLA and GLA to a
short-haul aircraft leads to an 11% reduction in fuel burn and 7% reduction
in direct operating cost. These savings are diminished if MLA and GLA
are applied independently. We also explore the synergy between active
load control and natural laminar ow. It is possible to invest some of the
weight savings from load alleviation to achieve extensive laminar ow. The
combination of load alleviation and NLF increases the fuel and cost savings
to upwards of 18% and 11% respectively. We conclude that load alleviation
can shift the transonic transport optimum towards low-sweep, NLF wings.
Sensitivity studies nally demonstrate that contemporary commercial air-
craft should be able to meet the control requirements for eective gust load
alleviation.
Nomenclature
x
CG
Longitudinal center-of-gravity
C Global damping matrix
F Generalized forces
K Global stiness matrix
M Global mass matrix
u Generalized displacement

Ph.D. Department of Aeronautics & Astronautics. Member AIAA

Professor, Department of Aeronautics & Astronautics. Fellow AIAA


1 of 36
c Wing trailing edge extension ratio
dCp
dx
Airfoil rooftop pressure gradient
D
T
Drag-to-thrust ratio
AR Trapezoidal wing aspect ratio
AR
g
Gross wing aspect ratio
Re
x
Local Reynolds number
Re
e
9 Transition Reynolds number as predicted by e
9
ts

fuel
Fuel density
A
box
Section enclosed wing box area
b
w
Wing span plumbed for fuel
f Modal force
F
g
Unsteady gust load alleviation factor
H Boundary layer shape factor
h Gust gradient length
k
d
GLA derivative gain
k
p
GLA proportional gain
l
land
Landing eld length
l
to
Balanced takeo eld length
M
land
Landing Mach number
M
to
Takeo Mach number
R Range
S Stratford concavity parameter
S
h
Horizontal tail area
T
0
Sea-level static thrust
t
inv
Airfoil thickness distribution from the inverse solution
t
s
Wing box skin gauge thickness
t
w
Wing box web gauge thickness
U
ds
Design gust amplitude
V
cruise
Cruise speed
V
A
Design maneuver speed
V
r
Rotation speed
V
s
Stall speed
w
g
Gust vertical velocity
W
reserve
Reserve fuel weight
x
mg
Main landing gear position
x
root
Wing root leading edge position
x
te
Airfoil trailing edge
2 of 36
x
a
Section aft spar location
x
e
Section elastic axis location
x
f
Section forward spar location
x
mg
Longitudinal position of the main gears
x
ng
Longitudinal position of the nose gear
z
f
Final cruise altitude
z
i
Initial cruise altitude
C
l0
C
l
at 0 angle of attack
C
Lh
Horizontal tail lift coecient
C
Mac
Pitching moment about the aerodynamic center
C
p
LE
Leading edge C
p
L
e
4
Elastic wing roll moment due to aileron deection
L
r
4
Rigid wing roll moment due to aileron deection
M
r
Isobar-normal local Mach number at the start of recovery
C
lmax
Section maximum lift coecient
S
ref
Trapezoidal wing reference area
W
MTOW
Maximum takeo weight
W
MZFW
Maximum zero-fuel weight
Symbols

0
Zero-lift angle of attack
Trailing edge control surface deections

land
Landing ap

to
Takeo ap
Non-dimensional span = y/(b/2)
d
dt
Control surface deection rate
Second segment climb gradient

w
Usable fuel volume ratio for a enclosed wing box section
Wing taper ratio

1/4
Trapezoidal wing quarter-chord sweep

a
Allowable stresses

v
von Mises stress
Wing jig twist
Damping ratio
3 of 36
I. Introduction
Conceptual aircraft design studies often reduce the aircraft-level impacts of laminar ow
and load alleviation to empirical technology factors.
13
This approach greatly simplies
the design problem, but omits potentially important multidisciplinary interactions. For
example, aeroelastic and control power considerations can limit the eectiveness of active
load alleviation. Similarly, one should weigh the drag savings from transonic natural laminar
ow against potential aerostructural penalties. In this work, we develop a moderate delity,
physics-based framework to integrate active load alleviation and natural laminar ow into
the realm of conceptual aircraft design.
Active load control is not a new technology many modern commercial transport aircraft
incorporate some form of active load alleviation.
4,5
However, industry practice is to introduce
load alleviation after conceptual and preliminary design.
6
There are good reasons, however,
to include load alleviation in conceptual design, where the technology has the potential to
signicantly inuence the optimal wing and aircraft congurations.
7
Haghighat, Martins and
Liu
8,9
demonstrate that the concurrent optimization of a simple aircraft with its active load
control system can net a 20% improvement in endurance.
A. Natural Laminar Flow
Achieving extensive laminar ow over the wings can reduce skin friction and prole drag.
10,11
In active laminar ow control (LFC), which we illustrate in Figure 1b, the aerodynamic gains
from laminar ow must be weighed against the weight, maintenance and power burdens
associated with pneumatic boundary layer control.
-Cp
x
t
x
t
(a) Turbulent
-Cp
Suction
xt
xt
(b) Active laminar ow control
-Cp
x
t
x
t
(c) Natural laminar ow
Figure 1. Representative turbulent and laminar airfoil geometries and pressure distributions
(adapted from Joslin
10
).
An alternative approach to active LFC is to passively stabilize the boundary layer using
tailored pressure distributions. The potential fuel savings from this natural laminar ow is
variously quoted as between 5 to 12%.
1012
While NLF avoids the pneumatic and mechanical
complexities associated with active LFC, its design is nonetheless driven by a complex set of
aerostructural trade-os. This is particularly true in the case of transonic NLF wing design.
NLF airfoil like the one illustrated in Figure 1c can be designed with extensive regions of
ow acceleration to suppress streamwise Tollmien-Schlichting instabilities. A highly swept
4 of 36
0
5
10
15
20
25
30
35
40
0 5 10 15 20 25 30 35
S
w
e
e
p

(
D
e
g
)

Transition Reynolds Number (Million)
NLF Region
Figure 2. Experiment trends of laminar ow versus wing sweep. (Adapted from Rajnarayan
and Sturdza
13
)
NLF wing has to balance the stability of the streamwise against the crossow boundary
layer: favorable streamwise pressure gradients tend to be destabilizing for the crossow. The
ight test results consolidated in Figure 2 shows that laminar ow becomes progressively
more dicult to maintain over highly swept wings.
13
Flight test experiences suggest that a
NLF wing should be designed with minimum sweep. However, a low-sweep wing has to be
made thin to prevent drag divergence at transonic speeds. The thin wing would be heavier
and suer from degraded high-lift performance relative to a thicker, swept wing. Considered
in a multidisciplinary context, the structural and high-lift penalties associated with a thin,
unswept NLF wing can outweigh the drag savings from laminar ow.
B. Active Load Alleviation
The basic workings of maneuver load alleviation (MLA) is illustrated in Figure 3. Here
two aircraft are in symmetric maneuvers at the same load factor. The aircraft with MLA
uses coordinated control surface deections to concentrate lift inboard. So while both wings
carry the same lift, the wing with MLA experiences signicantly less bending stresses. It
can therefore be made lighter, or longer and thinner at the same weight. Studies show that
aircraft with MLA can be designed with 10-15% greater wing span at xed weight, which
can translate into 8-13% drag savings.
2,1417
Moreover, by reducing the structural penalties
associated with thin wings, MLA can help move the optimal transonic conguration toward
low-sweep, NLF wings.
As MLA reduces the stress associated with maneuver, dynamic loads such as those
produced by gusts and stiness requirements grow in relative importance.
18
The interplay
between static and dynamic loads in wing sizing suggests that a complimentary gust load
5 of 36
(a) Conventional (b) MLA
Figure 3. Maneuver lift distributions for aircraft with and without MLA.
alleviation (GLA) system is likely needed to help MLA realize its potential.
Like MLA, GLA also uses control surface deections to reduce the wing stresses. The
dierence is that GLA responds to unanticipated, dynamic loads. Sensor and actuator
bandwidth can therefore inuence GLA eectiveness. The stochastic nature of gusts and
the close coupling among aircraft conguration, structural dynamics and gust response
complicates the design of the GLA controller.
Previous studies have addressed the key components of GLA design problem: 1) nd the
worst-case gust for a given airplane,
1924
2) design the aircraft or wing structure to sustain a
given gust encounter
2528
and 3) design a GLA controller for a given airplane.
2931
Still others
have examined the integrated problem of concurrently designing an aircraft with its GLA
system.
79
The present work aims to further close the loop among aircraft design, active load
control and natural laminar ow.
II. Aircraft Design Framework
The design framework is based on the Program for Aircraft Synthesis Studies (PASS).
32
PASS uses fast, semi-empirical models to relate mission performance to aircraft conguration.
We extend PASS with physics-based aeroservoelastic and viscous design modules to integrate
load alleviation and NLF into conceptual design.
33,34
In this section, we discuss the aircraft and mission level design variables and constraints.
The aircraft-level design variables are as follows:
x
aircraft
=
_
S
h
S
ref
, x
root
, W
MTOW
, W
MZFW
, T
0
, x
mg
_
(1)
The horizontal tail is parameterized by an area ratio: S
h
/S
ref
. The longitudinal wing root
position x
root
important for trim, stability and gear integration is also a design variable.
The thrust and fuel consumption are computed using a rubberized engine.
32
The sea level
thrust to dry engine weight ratio and TSFC are anchored to published gures for the CFM-
56-7B27 turbofan.
35
The engine weight and dimensions scale with required thrust.
6 of 36
There are no weight iterations in the mission analysis. We employ the commonly used
technique of optimized-based decomposition (OBD)
3638
to pose the xed-point weight con-
vergence problem as an optimization constraint. Using OBD the weights and longitudinal
CG are converged with the optimized aircraft. The nose gear is also constrained to carry at
least 8% of the aircraft load for traction:
x
mg
=
max (x
CG
) 0.08x
ng
0.92
(2)
W
MTOW
= W
MZFW
+ W
fuel
+ W
reserve
(3)
Second Segment Climb (400 ft, 1.2V
s
)
Takeo Rotation
(S.L. V
r
)
Approach
(S.L. ,1.3V
s
)
1.3-g Maneuver (40,000 ft, V
c
)
Gust, (10,000 ft, V
c
)
Initial Cruise, Gust (z
i
, V
cruise
i
)
Final Cruise, Gust (z
f
, V
cruise
f
)
2.5-g Maneuver, (20,000 ft, V
c
)
Figure 4. The mission prole and design ight conditions.
The mission prole illustrated in Figure 4includes the typical takeo, climb, cruise,
approach and landing segments. Not shown is a standard diversion reserve. The mission
also includes representative gust encounters and limit load maneuvers. Important mission
parameters such as the initial and nal cruise altitudes are optimized with the aircraft:
x
mission
= [z
i
, z
f
, M
to
, M
land
,
to
,
land
] (4)
The aircraft and engines are sized to meet range, eld length, second segment climb gradient
and en-route climb constraints:
(D/T)
i
< 0.95, l
to
< l
tomax
, > 0.024,
(D/T)
f
< 0.95, l
land
< l
landmax
, R > R
req
(5)
The D/T constraint creates a thrust reserve to sustain en-route climb. The 0.95 limit is
derived from energy considerations at favorable climb speeds.
32
The trimmed aircraft is
subject to tail maximum lift and stability constraints at all ight conditions:
C
Lh
< C
Lhmax
(6)
7 of 36

dC
Mac
dC
L
> 0.05c (7)
We compute the fuselage and tail parasite drag using empirical shape factors.
32
The wing
drag and stability derivatives are solved using a combination of panel and inverse methods,
which we discuss in Section III.
III. Wing Design
The goal of the wing design module is to capture the aerostructural trade-os that domi-
nate transonic NLF wing design. The optimization trades laminar ow against compressibility
drag and wing structural eciency. We parameterize the wing using the following variables:
x
wing
=
_
S
ref
, AR, ,
1/4
, c, , t(x
e
), t(x
f
), t(x
a
), t
s
, t
w
_
(8)
The wing trailing edge extensions c and jig twists are dened at the six breakpoints
highlighted in Figure 5a. The wing box in Figure 5b extends from 20 to 65% chord. The
wing box is parameterized by the box height at the forward spar, aft spar and elastic axis.
The spar cap and web thicknesses t
s
and t
w
are design variables. The intermediate spanwise
geometries are linearly interpolated. The wing structural weight comes from the wing box
geometry and material density. The structure is sized against stress, minimum gauge and
stiness constraints. Aircraft structures are not sized by stress alone. For example, the wing
is stiened against local and global buckling. There are also non-structural weights from
rivets, fasteners and panels. To simplify the problem we use the semi-empirical relationship
of Gallman
39,40
to relate the wing structural to total weight. Gallman presents a validation
of this weight method against conventional aluminum transport designs.
40
Fuselage Width
=0.3
=0.5
=0.1
=0.7
=0
=1
(a) Wing Breakpoints
Fuel
x
e
x
f
x
a
t
w
t
s
(b) Wing Box (in structural axis)
Figure 5. The spanwise wing break locations and the wing-box denition.
Fuel volume plays an important role in the sizing of high-aspect ratio, thin wings. All
8 of 36
mission and reserve fuel are carried in the wings:
W
fuel
< 2
w

fuel
_
bw/2
0
A
box
(y)dy (9)
The inboard 70% of the wing box, including the fuselage carryover, is plumbed for fuel
(b
w
= 0.7). We assume that 90% of the enclosed wing box volume is available to carry usable
fuel (
w
= 0.9).
A. Aerodynamics
The wing lift distribution and stability derivatives are solved using the Weissinger lifting line
method
17,41
with Prandtl-Glauert corrections. Integration of the Trez plane normal-wash
yields the induced drag. We use 30 panels per semispan to balance speed against accuracy.
42
Trailing edge control surface deections are modeled as changes in the local C
l0
, similar
to the approach taken by Wakayama.
43
Over the apped span, an empirical C
l
()
44
is
added to the boundary condition of the Weissinger method as an equivalent twist

=
C
l
()
C
l
(10)
This method allows us to consider the 3-D inuence of control surface deections without
having to model chordwise pressure variations. We include maximum lift constraints using
the method of critical sections: the wing is assumed to stall if any station reaches its local
C
lmax
.
43
A C
l
margin of 0.2 is imposed over the outboard wing stations ( > 0.75) to preserve
control authority at high lift conditions:
C
l
() <
_

_
C
lmax
() < 0.75
C
lmax
() 0.2 0.75
(11)
Where C
lmax
is modeled as a function of airfoil properties and control surface deections:
43
C
lmax
() = f(t/c, Re
c
, M

,
1/4
, ,
s
) (12)
B. Inverse Design
We use a hybrid-inverse method to link the wing pressure distribution to laminar ow
and structural design. The inverse approach is motivated by the common shape of typical
transonic laminar ow pressure distributions: an extended region of ow acceleration, followed
by a moderate shock, and a turbulent recovery to the trailing edge.
9 of 36
Fuselage Width
=0.3
=0.5
=0.1
=0.7
=0
=1
(a) Inverse Design Sections
x
r
dC
p
/dx(u)
M
r
(u)
C
P
u
(x=0)
C
P
l
(x=0)
dC
p
/dx(l)
S
M
r
(l)
(b) Section Pressure Parameterization
Figure 6. The inverse design wing stations and pressure parameterization.
The wing root is always assumed to be turbulent to account for the eects of attachment
line instabilities. The assumption of 2-D ow also breaks down near the wing tip,
12,45,46
which
are also assumed to be turbulent. Figure 6b shows that the section pressure distributions are
parameterized as linear ramps followed by Stratford-type recoveries.
47
The pressure design
variables are dened as follows:
x
pressure
=
_
x
r
, C
p
LEl
, C
p
LEu
,
_
dC
p
dx
_
u
,
_
dC
p
dx
_
l
, S
u
_
(13)
We start by dening the compressible streamwise C
p
at the exposed stations in Figure 6a.
The wing stations highlighted in blue can be designed for laminar ow. The wing-fuselage
intersection and tip stations, highlighted in red, are assumed to be turbulent. The extent
of laminar ow and the peak Mach number are both strongly determined by the rooftop
pressure gradient
dCp
dx
. The Stratford concavity parameter S, also a design variable, controls
the speed of pressure recovery. A rapid pressure recovery extends the region of accelerating
ow, but also brings the ow closer to separation. The Stratford criteria for an arbitrary
S can be written in terms of the canonical pressure coecient C
p
and eective Reynolds
number Re
x
:
C
p
=
_

_
_
C
p
m
3
+ 0.1893Re
x
1/5
ln
_
x
xm
_
S
2
_
1/3
C
p
< 4/7
1
ka

k
b
+
x
xm
C
p
4/7
(14)
We use constraints to ensure the compatibility of the pressure distributions with the rest
of the design. The pressure distribution has to integrate to at least the same section C
l
as
the Weissinger solution:
C
l
() <
_
c
0
_
C
p
l
(x) C
p
u
(x)
_
dx (15)
We match the sectional C
l
using inequality, rather than equality constraints to improve
problem scaling. The inequality constraints are sucient because any excess C
l
produced by
10 of 36
the inverse solution would necessarily increase super-velocity and incur drag penalties.
For a xed C
l
and airfoil thickness the steepness of the favorable pressure gradient is
limited by the isobar-normal Mach number at the recovery M
r
, which we take to be the
Mach number ahead of a weak shock. The local Mach number ahead of the shock is resolved
using the Lock transformation.
48
The optimal recovery Mach number is driven by a trade-o
between parasite and wave drag: a high Mach number at the recovery allows for a more
favorable pressure gradient, which can delay transition. But too high a recovery Mach number
can lead to drag divergence or worse, shock-induced ow separation. In this study, we limit
the maximum recovery Mach number to prevent drag divergence:
M
ru
< 1.1
M
rl
< 0.95
(16)
The limit in the upper surface recovery Mach number is consistent with the design intent and
aerodynamic characteristics of the NASA SC(2) series of supercritical airfoils.
49,50
We esti-
mate the compressibility drag coarsely using the quartic drag rise prole given by Obayashi:
51
C
dc
= C
dc
cos
3
(
xr
)
C
dc
=
_

_
0 M
r
< 1
0.02(M
r
2
1)
4.4
M
r
1
(17)
This simple, empirical drag model is only meant to render C
Dc
sensitive to the pressure
distribution; it cannot accurately estimate the wave drag for arbitrary airfoils. But by
enforcing the conservative Mach number constraint in Equation (16), we also make sure that
the compressibility drag is small.
With the pressure distribution dened, we use thin airfoil theory to solve for the airfoil
geometry. The sweep-taper theory of Lock
48
and the inverse Krmn-Tsien transformation
41
are used to convert the compressible, streamwise pressure distributions into their equivalent
isobar-normal, incompressible distributions. The following geometric constraints close the
loop between aerodynamic and structural design:
t
inv
(x
e
) > t
e
t
inv
(x
f
) > t
f
t
inv
(x
a
) > t
a
t
inv
(x
te
) = 0
(18)
The constraints require the airfoil geometry from the inverse solution to enclose the hexagonal
11 of 36
wing box directly dened by x
wing
. The airfoil must also close at the trailing edge.
C. Boundary Layer Solution
We employ the integral methods of Thwaites and Head to solve the laminar and turbulent
boundary layers.
52
For low sweep wings the incompressible approach is likely conservative
as compressibility tend to stabilize the streamwise boundary layer.
5355
To simplify surface
discretization and numerical integration we use the single-step H R
x
criteria of Wazzan to
predict ow transition:
56
log (Re
e
9) = 40.46 + 64.81H 26.75H
2
+ 3.38H
3
(19)
The predicted transition Reynolds number Re
e
9 is a function of the boundary layer shape
factor H. Transition is predicted to occur when Re
x
> Re
e
9. Rather than interpolating or
iteratively solving for x
t
, we specify x
t
as a design variable and enforce laminar ow before
transition:
[Re
x
(x) < Re
e
9(x)]|
(x<xt)
(20)
This application of optimizer-based decomposition eliminates the need for internal iterations
to solve for the transition point. It also reduces the laminar boundary layer evaluation to
just one point at x
t
. We nally integrate the turbulent boundary layer to the trailing edge
and compute the prole drag using the Squire-Young equation.
57
D. Structure
Loads experienced at dierent ight conditions, including limit maneuver and gust encounters,
size the wing structure. The structure model follows the work of Gallman
39
and is similar to
Drelas smeared thickness method.
58
Static and dynamic stress constraints combine to size
the wing:
{
v
()} <
a
(21)
The dynamic stress constraints are functions of both wing spanwise position and time. The
solution for the dynamic stresses are discussed in more detail in Section F.
{
v
(, t)} <
a
(22)
The skin carries the bulk of the bending loads. The web and skin carry the torsional and
shear stresses. The torsional stress is computed using Bredt-Batho theory. The sectional
stress constraints reference the von Mises yield criterion. Figure 7 shows that the y-stations
of the stress constraints are collocated with those of the aerodynamic control points. The
12 of 36
stress constraints are imposed individually for the web and skin at each wing station (and
also at each time step of the dynamic simulations). This formulation of the optimization
problem increases the computational cost of calculating derivatives. Fortunately, the active
set algorithm can eciently solve problems with large numbers of constraints.

v
()<
a

v
(,t)<
a
Figure 7. Aerodynamic control points and wing box stress constraints.
In addition to the stress constraints, we impose a minimum aileron eectiveness constraint
based on the reported operating parameters of the Boeing 747-100.
59
The aileron eectiveness
constraint is dened for a maneuver speed V
A
of 250 knots at 30,000 feet and requires the
aileron roll moment of the exible wing L
e
4
to be at least 53% of its rigid counterpart L
r
4
:
L
e
4
L
r
4

V
A
> 0.525 (23)
The aileron derivatives are solved using the Weissinger method. The elastic derivative includes
the eects of aeroelastic twist from the aileron deection.
The maneuver load calculations do not include the eects of static aeroelasticity from wing
bending. This is a non-trivial omission. Static deections in limit maneuvers can produce
load alleviation for a swept wing. Consequently, the present approach may over-predict the
relative gains from MLA for a highly swept wing.
E. Maneuver Load Alleviation
MLA use coordinated control surface deections to minimize wing stress in limit maneuver
conditions. In this study we aect load alleviation using the ailerons and ap arrangement
shown in Figure 8. The control surfaces are bounded by wing breakpoints and extend over
25% of the chord. The allocation of the control surfaces follow conventional patterns: the
inboard aps are assumed to have lower slew rates and are consequently used only for MLA.
The outboard ailerons have higher bandwidth and are used for both MLA and GLA.
Ideally, the aircraft and its MLA system should be designed for the entire Vn envelope.
To simplify the problem, we design the aircraft to the representative limit maneuvers detailed
13 of 36

2
MLA
MLA and GLA

4
Figure 8. Trailing edge control surface arrangement and allocation.
Load factor Altitude Speed
Maneuver 1 2.5-g 30,000 feet V
c
Maneuver 2 1.3-g 40,000 feet V
cruise
Table 1. Representative maneuver conditions.
in Table 1. V
c
denotes the structural design speed, which is 6% faster than the cruise speed..
The 2.5-g maneuver at 30,000 feet is a high stress condition. As the eectiveness MLA
becomes constrained by maximum lift at higher altitudes, the eective wing stresses can
increase. However, severe maneuvers at altitudes are both rare and constrained by available
thrust.
4,5
The more modest 1.3-g maneuver at high altitude is included primarily as a
constraint against buet.
The MLA design variables are the trailing edge control deections at each maneuver
condition.
x
MLA
= (24)
The MLA control deections are between 0 to 10

for the aps and -10 to 10

for the ailerons.


The limits reect the inability of typical slotted aps to reex.
0

<
i
< 10

, for i = [1, 2]
10

<
i
< 10

, for i = [3, 4]
(25)
Restricting the aps to positive (downward) travel is not likely to degrade MLA performance.
In MLA the aps should deect down to shift loads inboard to reduce the bending moments.
The positive control deection limits reect practical considerations of hinge moment and
ow separation. The 10

limit match the parameters of the load alleviation systems on the
Airbus A320 and A330.
4,5
Previous sensitivity studies
60,61
demonstrate that large MLA ap
and aileron travels do not necessarily improve load alleviation performance. Beyond about
10

, maximum lift and elastic constraints combine to limit the gains from load alleviation.
14 of 36
F. Gust Load Alleviation
Aircraft gust response is a complex function of gust shape, aircraft dynamics, structural
dynamics and their interactions with the GLA system. In this study, we concurrently design
the aircraft with a simple proportional-derivative (PD) GLA system:
(t) = k
p
(t) +k
d
(t) (26)
The gust-induced and drive the wing control surface deections (t). We optimize the
control gains concurrently with the aircraft. The commanded deections are subject to
magnitude and rate limits:
10

<
i
< 10

d
i
dt
< 30

/s
i = [3, 4]
(27)
As previously illustrated in Figure 8, only the two outboard control surfaces, which correspond
to ailerons, are allocated for GLA. This is because aps typically have slew rates of only
around 2

/s too slow for dynamic load control. The rate limit is consistent with the
slew rate of contemporary commercial aircraft ailerons at 35-40

/s.
62
The control deection
limits are consistent with those of the MLA system at 10 degrees.
The simple control logic, combined with rate and deection saturation limits, capture the
essential impact of GLA. Future studies should examine the eectiveness of more sophisticated
model-predictive GLA controllers, such as those considered by Haghighat, Liu and Martins
9
and Gaulocher, Roos and Cumer.
63
Also important, but omitted from the present study, are
the eects of measurement uncertainty and sensor and controller delays.
h
U
ds
Figure 9. The 2D velocity eld dened by the 1Cosine gust.
The Federal Aviation Regulations (FAR) species both a continuous and discrete vertical
gust criterion.
22,64
The continuous criterion is important from the angle of structural fatigue
and passenger comfort. In this study, we focus on the discrete criterion used to repre-
sent signicant gust events. The discrete criterion requires aircraft structures to withstand
15 of 36
encounters with the so-called 1Cosine gust shown in Figure 9:
w
g
(t) =
U
ds
2
_
1 cos
_
V
ht
__
U
ds
= U
ref
F
g
_
h
350
_
1/6
h = [35, 350] feet
(28)
The design gust amplitude U
ds
is dened in terms of the reference velocity U
ref
and the
unsteady gust load alleviation factor F
g
. Both are in turn functions of the gust encounter
altitude. The gust gradient length h is one half of the gust wavelength. The power law
scaling of the gust amplitude is consistent with the theoretical scaling properties of severe
atmospheric turbulence.
We integrate the longitudinal pseudo-steady aircraft equations of motion to advance the
gust encounter simulations:
W
q

S
ref
cg
z = C
L

g
+ C
Lq

Wr
y
2
q

S
ref
cg

= C
M

g
+ C
Mq

g
= +
w
g
z
V

(29)
The stability derivatives and control inputs are updated at each time step. Although the
analysis is based on a pitch-plunge model, we typically restrict optimizations to use a plunge-
only model to avoid dynamically unstable designs.
The multi-point design methodology models gust encounters at a number of altitudes and
gust gradient lengths. We design the aircraft for 1Cosine gust encounters at 10,000 feet and
the initial and nal cruise altitudes. Low altitude gust encounters combine high dynamic
pressure and signicant gust amplitude to produce limit design conditions. The cruise stage
gust encounters capture the trade between span eciency and wing dynamic loads.
At each altitude we simulate gust encounters at 12 dierent gradient lengths ranging
from 35 to 800 feet. An inspection of Equation (28) would suggest that the highest amplitude
gusts are those with the longest wavelength. However, the aircraft would also have more time
to rise and fall with a longer, slower gust. This would reduce the eective angle of attack
and hence, the severity of the gust-induced stresses.
The aircraft structural response is also a strong function of the gust frequency. Short-
wave gusts contain little energy, but can rate-saturate the GLA system to produce limit
loads. However, gusts that are much faster than the wing natural frequency also do not
excite the wing structure. At Mach 0.78 and 30,000 feet for example, the FAR specied
16 of 36
gust gradients span the frequency range between 0.2 to 8 Hz, which brackets the natural
frequencies of typical transport wings. It is necessary therefore to characterize the wings
frequency response to gust.
We model the wing structure as a variable cross section beam in forced vibration. A simple
nite element (FEM) kernel is used to assemble the wing stiness and mass matrices.
65,66
The generalized displacement u can be written in terms of the mass, stiness and damping
matrices:
65,66
M u(t) +C u(t) +Ku(t) = F(t) (30)
Direct integration of the wing equations of motion can be expensive. We use the technique
of modal decomposition to isolate the energetic low-frequency modes. The decoupled ODEs
can then be integrated as piece-wise exact second-order impulse responses.
67
A problem
arises that modal decomposition is valid only for proportionally damped system. The wing
structural dynamics, on the other hand, is dominated by non-proportional aerodynamic
stiness and damping:

d

u
V

sin
e

s

2u
b
sin
e
(31)
We can approximate the aerodynamic stiness and damping at each time step as applied
forces. At each time-step, we compute the wing twist due to apping using the wing deection
and deection rates from the previous time-step. The panel method boundary conditions
are updated before the next step. The mass and stiness matrices remain constant. The
augmented equations of motion can be written as:
M u(t) +C u(t) +Ku(t) = F(t) +F
d
(t) +F
s
(t) (32)
Where F
d
and F
s
are the generalized damping and stiness pseudo-forces. We nally recover
the wing stresses from the time history of deections and twists. We impose spanwise stress
constraints at each time step:
{
v
(, t)} <
a
(33)
Since the simulations cover only a nite number of gust encounter altitudes and wave-
lengths, the optimization will not, in general, capture the eects of the worst-case gust. The
optimization of an aircraft against its worst-case gust is an adversarial optimization problem,
in which the aircraft and gust can evolve against one another. Moreover, the aircraft is not
designed by a single, worst-case gust. The current multi-point design method is a pragmatic
attempt to capture the impacts of gust loads on structural design.
17 of 36
IV. Optimization
We optimize the aircraft for minimum direct operating cost (DOC) computed using
an extended version of the Airline Transport Association (ATA) cost model.
68
. The cost
analysis is based on a fuel cost of $2.5/gal. The simple ATA model is useful for comparing
the relative performance of aircraft with comparable technology contents. The optimization
problem can be posed succinctly as:
minimize J = DOC
w.r.t. x
mission
= [z
i
, z
f
, M
to
, M
land
,
to
,
land
]
x
aircraft
=
_
S
h
S
ref
, x
root
, W
MTOW
, W
MZFW
, T
0
, x
mg
_
x
wing
=
_
S
ref
, AR, ,
1/4
, c, , t(x
e
), t(x
f
), t(x
a
), t
s
, t
w
_
x
pressure
=
_
x
r
, C
p
LEl
, C
p
LEu
,
_
dC
p
dx
_
u
,
_
dC
p
dx
_
l
, S
u
_
x
transition
= x
t
(for NLF designs)
x
MLA
= (for MLA designs)
x
GLA
= [k
p
, k
d
] (for GLA designs)
s.t. The constraints listed in Table 2
(34)
The wing geometry, pressure and transition variables are dened at the wing breakpoints.
The MLA control surface deections and GLA gains are dened for the appropriate trailing
edge control surfaces. There can be up to 100 design variables if MLA, GLA and NLF are
all included in the optimization.
The pressure, compressibility, wing box geometry and transition compatibility constraints
in Table 2 are imposed at the wing breakpoints in cruise. The transition constraints are
enforced for NLF surfaces. The maximum lift and static and dynamic stress constraints are
imposed at the same span stations as the Weissinger control points.
We use a monolithic, gradient-based optimizer (MATLAB fmincon). The gradients are
computed via nite dierence. Care is taken to maintain smooth numerics , and properly
scale the design variables and constraints to aid convergence. There are 146,000 constraints
in the optimization. The objective and constraints are converged to a tolerance of 1
6
. We
use multiple restarts to reduce the likelihood of converging to local minimums.
A typical optimization takes between 10 to 30 minutes of wall time to complete. Figure 10
shows a representative convergence history (Mach 0.78 turbulent design).
18 of 36
Constraints
Weight
W
MTOW
= W
MZFW
+ W
fuel
+ W
reserve
Landing Gear
x
mg
=
max (x
CG
) 0.08x
ng
0.92
Performance
(D/T)
i
< 0.95, l
to
< l
tomax
, > 0.024,
(D/T)
f
< 0.95, l
land
< l
landmax
, R > R
req
Trim
C
Lh
< C
Lhmax
Stability

dC
Mac
dC
L
> 0.05c
Maximum Lift C
l
() <
_
_
_
C
lmax
() < 0.75
C
lmax
() 0.2 0.75
Pressure
C
l
() <
_
c
0
_
C
p
l
(x) C
p
u
(x)
_
dx
Compressibility
M
ru
< 1.1
M
rl
< 0.95
Wing Box
t
inv
(x
e
) > t
e
t
inv
(x
f
) > t
f
t
inv
(x
a
) > t
a
t
inv
(x
te
) = 0
Fuel Volume
W
fuel
< 2
w

fuel
_
bw/2
0
A
box
(y)dy
Transition
[Re
x
(x) < Re
e
9(x)]|
(x<xt)
Static Stress
{
v
()} <
a
Dynamic Stress
{
v
(, t)} <
a
Aileron Reversal
L
e
4
L
r
4

V
A
> 0.525
Table 2. Optimization constraints.
19 of 36
0 10 20 30 40 50 60
0
0.5
1
1.5
O
b
j
e
c
t
i
v
e

f
u
n
c
t
i
o
n
0 10 20 30 40 50 60
0
2
4
Iteration
M
a
x

c
o
n
s
t
r
a
i
n
t

v
i
o
l
a
t
i
o
n
Figure 10. The convergence history of a turbulent, Mach 0.78 design.
V. Design Study
We now apply the design framework developed in the previous sections to a short-haul
transport design problem. The aircraft in this study are designed for a 3,000 nm cruise
mission. The takeo and landing eld requirements are 7,500 and 5,500 feet respectively.
These specications match the published performance gures of the Boeing 737-800.
The aircraft all have T-tails and fuselage-mounted engines to keep the wings clean for
laminar ow. The high lift system consists of simple, single-slotted aps and leading edge
Krueger aps. The Krueger aps keep the upper wing free of gaps that could lead to
transition. The aircraft are of conventional aluminum construction.
The 2-D boundary layer model discussed in Section III is only applicable to low sweep,
high aspect ratio wing, where transitions is dominated by Tollmien-Schlichting instabilities.
To avoid introducing crossow dominated designs, we limit the maximum NLF wing sweep
to no more than 10

. Turbulent wings, on the other hand, can be designed with high sweep.
This discontinuity in the design space leads us to dene the two optimization baselines
summarized in Table 3.
Transition (upper) Transition (lower) max(1
/4
)
Turbulent 5% 5% 40

NLF free 5% 10

Table 3. The turbulent and laminar baselines.


20 of 36
Table 3 shows that all lower wing surfaces are assumed to be turbulent. This conservative
assumption accounts for the gaps introduced by the leading edge Krueger aps, which retract
to the underside of the wings. Fortunately, achieving laminar ow on the suction surface,
where the supervelocity is greater, yields most of the drag savings.
11
A. Turbulent Designs
Figure 11 compares the optimized turbulent aircraft with dierent levels of load alleviation.
We summarize the important aircraft parameters in Table 4.
(a) Turbulent (b) Turbulent MLA (c) Turbulent GLA (d) Turbulent MLA+GLA
Figure 11. Optimized Mach 0.78 turbulent aircraft with dierent active load alleviation sys-
tems.
Turb. Turb. MLA Turb. GLA Turb. MLA+GLA
W
MTOW
(lb) 174,542 168,534 174,503 164,828
W
wing
(lb) 19,575 17,234 19,535 17,213
S
ref
(ft
2
) 1,610 1,376 1,610 1,378
b (ft) 137 148 137 159
AR 11.6 15.8 11.6 18.4

1/4
(Deg) 23.1 13.2 23.1 9.8
t/c 0.103 0.092 0.103 0.093
L/D 19.9 20.3 19.9 21.6
Table 4. Optimized Mach 0.78 turbulent aircraft parameters.
The optimized turbulent baseline (with no load alleviation) shown in Figure 11a is
representative of a contemporary short-haul transport. The 174,500 lb maximum takeo
weight agrees with that of the Boeing 737-800 at 174,200 lb.
69
The baseline designs relatively
large wing area is a consequence of the simplied high lift system: single slotted aps versus
the multi-slotted fowler arrangement on the 737.
The baseline aircrafts greater wing span may be attributable to the absence of gate
compatibility constraints in the optimization. Figure 12 shows the impact of a 120 feet gate
constraint on the optimized conguration. The optimized aspect ratio falls to 9.2.The weight
21 of 36
of the shorter wing falls by 20%. However, increased induced drag in cruise and climb lead
to a 3% increase in fuel weight and a 7% increase in required thrust. The net result is that
the W
MTOW
of the gate-constrained design falls only modestly, to 174,200.
(a) Turbulent (b) Turbulent (b < 120)
Figure 12. Optimized Mach 0.78 baseline and gate constrained congurations.
The MLA-equipped aircraft in Figure 11b has greater wing span, reduced wing sweep and
reduced wing thickness. Yet this thinner and longer wing weighs some 2,600 lb less than the
baseline wing. This result shows that maneuver load alleviation can signicantly increase the
eective structural eciency of the wing - allowing for longer wings at xed or even reduced
weight. MLAs measurable impact on the design suggests that maneuver loads are sizing a
substantial portion of the baseline wing.
The optimized GLA design in Figure 11c is unchanged from the baseline. The addition
of GLA nets no performance gain. The highly swept wing, with its high aeroelastic damping,
is not gust-critical.
There is only so much that a single load alleviation system can do to reduce wing structural
weight. As MLA reduces the severity of structural requirements in maneuver, gust loads
begin to drive structural weigh. The converse is true for a GLA-only design. More gains
should be possible with the simultaneous application of both MLA and GLA. The wing of
the MLA+GLA design pictured in Figure 11d is both 22 feet longer and 12.1% lighter than
the baseline wing. As MLA and GLA reduce the structural penalties associated with thin
wings, the optimizer is able to unsweep the wing to maximize span.
Turb. Turb. MLA Turb. GLA Turb. MLA+GLA
W
wing
/W
wing
baseline
1.000 0.880 0.998 0.879
W
fuel
/W
fuel
baseline
1.000 0.949 1.000 0.888
T
0
/T
0baseline
1.000 0.969 1.000 0.898
(L/D)/(L/D)
baseline
1.000 1.021 1.000 1.082
DOC/DOC
baseline
1.000 0.965 1.000 0.927
Table 5. Turbulent aircraft parameters normalized by the baseline.
22 of 36
The normalized aircraft parameters are shown in Table 5. To ease comparison we divide
the parameters by their corresponding values in the turbulent baseline. The results show
that the independent application of MLA leads to a 5% fuel and 3.5% cost reduction. The
gain is mainly structural: the L/D increases by less than 2% while the wing weight falls by
13%. The fuel and cost savings grow to 11.2% and 7.3% respectively when both MLA and
GLA are active. Here the savings are both aerodynamic and structural: an 8.2% gain in
L/D combined with a 12.1% reduction in wing weight relative to the baseline. The improved
weight and aerodynamics reduce the required sea-level static thrust by 10%.
0 0.2 0.4 0.6 0.8 1
0.5
0
0.5
1
1.5
2
Spanwise Position
C
l

c
/
c
r
e
f


Turb.
Turb. MLA
Turb. GLA
Turb. MLA+GLA
(a) 2.5-g at 30,000 feet
0 0.2 0.4 0.6 0.8 1
0.5
0
0.5
1
1.5
2
Spanwise Position
C
l

c
/
c
r
e
f


Turb.
Turb. MLA
Turb. GLA
Turb. MLA+GLA
(b) 1.3-g at 40,000 feet
Figure 13. Normalized turbulent maneuver lift distributions. (c
ref
= 10 feet)
Figure 13 overlays the normalized maneuver lift of the four turbulent designs. The MLA
system has clearly shifted the lift centroid inboard in the high stress, 2.5-g condition. The
MLA actions are less pronounced in the 1.3-g maneuver at high altitude. Here the variable
camber provides additional degrees of freedom for stall and buet protection. The aircraft are
not required to perform limit maneuvers at constant altitude; the inecient MLA maneuver
lift distributions do not increase the thrust requirements.
The wing stress plots in Figure 14 illustrate which load is sizing which portions of the
wing. The maximum gust stress is taken over simulated encounters with all gust wavelengths
and at all time-steps. Both the maneuver and gust-induced stresses are normalized by the
allowable stress
a
of the material. A section is stress-critical if the ratio of von Mises to
allowable stress,
v
/
a
, reaches 1 or -1. The allowable stress is related to the yield stress by
a safety factor of 1.5. The yield stress of aviation aluminum is conservatively assumed to be
54 KSI, which includes the detrimental eects of high-cycle fatigue.
The inboard wings are fully stressed. The outboard wings are sized by minimum gauge
23 of 36
0 0.2 0.4 0.6 0.8 1
1
0.5
0
0.5
1
v
o
n

M
i
s
e
s

S
t
r
e
s
s

v
/

a
Spanwise Position


2.5g (30,000 ft)
1.3g (40,000 ft)
gust (initial cruise)
gust (10,000 ft)
(a) Turbulent
0 0.2 0.4 0.6 0.8 1
1
0.5
0
0.5
1
v
o
n

M
i
s
e
s

S
t
r
e
s
s

v
/

a
Spanwise Position


2.5g (30,000 ft)
1.3g (40,000 ft)
gust (initial cruise)
gust (10,000 ft)
(b) MLA
0 0.2 0.4 0.6 0.8 1
1
0.5
0
0.5
1
v
o
n

M
i
s
e
s

S
t
r
e
s
s

v
/

a
Spanwise Position


2.5g (30,000 ft)
1.3g (40,000 ft)
gust (initial cruise)
gust (10,000 ft)
(c) GLA
0 0.2 0.4 0.6 0.8 1
1
0.5
0
0.5
1
v
o
n

M
i
s
e
s

S
t
r
e
s
s

v
/

a
Spanwise Position


2.5g (30,000 ft)
1.3g (40,000 ft)
gust (initial cruise)
gust (10,000 ft)
(d) MLA+GLA
Figure 14. Maneuver and gust-induced von Mises stress ratios (
v
/
a
) for turbulent designs.
24 of 36
and aileron reversal considerations. Maneuver loads dominate the baseline and GLA-only
wings to about 80% span. Low-altitude gust encounters size the MLA aircraft wing. As MLA
diminishes the importance of maneuver loads, gust loads begin to size the wing structure.
The addition of GLA at this point diminishes the eects of gust loads and reduces the
required structural mass. It follows then that the lighter MLA+GLA wing is designed by
a combination of maneuver and gust loads. We also note that the GLA system is more
successful at suppressing the gust loads at 10,000 feet, where there is no explicit trade
between the span loading and gust-induced stresses.
0.7 0.72 0.74 0.76 0.78 0.8 0.82
0.9
0.95
1
1.05
Cruise Mach Number
R
e
l
a
t
i
v
e

D
i
r
e
c
t

O
p
e
r
a
t
i
n
g

C
o
s
t


Turb.
Turb. MLA
Turb. GLA
Turb. MLA+GLA
Figure 15. The relative DOC of turbulent designs as functions of the Mach number.
We plot the DOC of aircraft optimized at dierent Mach numbers in Figure 15. Each
point corresponds to the cost performance of a dierent optimized aircraft. The DOC of each
aircraft is normalized by the DOC of the turbulent baseline at Mach 0.78. The gains from
load alleviations are consistent over a broad range of Mach numbers. The optimizer favors
speed over fuel burn to increase utilization and reduce cost. The direct operating costs fall
with increasing Mach number up to about Mach 0.8. The low sweep MLA+GLA design is
more sensitive to the eects of transonic drag rise its DOC starts to grow at about Mach
0.78.
B. Optimized Laminar Aircraft
The optimized NLF aircraft are compared in Figure 16 and Table 6. The transition lines are
shown in green on the starboard wing. All but the MLA+GLA design reach the maximum
allowed sweep of 10

. This is not surprising since the boundary layer model omits the eects
of crossow.
The transition lines in Table 6 show extensive laminar ow for only the MLA+GLA wing.
On the other aircraft, the laminar regions are restricted to the outboard wing. Here the
25 of 36
(a) NLF (b) NLF MLA (c) NLF GLA (d) NLF MLA+GLA
Figure 16. Optimized Mach 0.78 NLF aircraft with dierent active load alleviation schemes.
NLF NLF MLA NLF GLA NLF MLA+GLA
W
MTOW
(lb) 177,179 168,573 177,464 163,912
W
wing
(lb) 22,653 19,151 22,833 19,607
S
ref
(ft
2
) 1,500 1,394 1,509 1,464
b (ft) 145 159 145 172
AR 14.1 18.2 14.0 20.1

1/4
(Deg) 10.0 10.0 10.0 8.7
t/c 0.096 0.096 0.095 0.085
L/D 20.8 21.4 20.8 23.8
Table 6. Optimized Mach 0.78 NLF aircraft parameters.
26 of 36
optimizer balances structural eciency against laminar ow: the inboard wings experience
higher stresses and can therefore benet more from increased wing thickness. For a xed
section C
l
and maximum Mach number, a thick section cannot support a strongly favorable
pressure gradient without incurring strong shocks. The design framework captures this basic
transonic aero-structural trade.
NLF NLF MLA NLF GLA NLF MLA+GLA
W
wing
/W
wing
baseline
1.157 0.978 1.166 1.002
W
fuel
/W
fuel
baseline
0.977 0.909 0.978 0.817
T
0
/T
0baseline
0.994 0.930 0.995 0.838
(L/D)/(L/D)
baseline
1.043 1.072 1.044 1.196
DOC/DOC
baseline
0.995 0.944 0.995 0.891
Table 7. The NLF aircraft parameters normalized by the turbulent baseline.
Table 7 compares the laminar aircraft thrust, L/D, weight and cost, which are again
normalized by the corresponding turbulent baseline values. We see that the NLF aircraft
(without load alleviation) achieves a modest 2.3% fuel savings and virtually no cost savings
over the turbulent baseline. The aerostructural penalties associated with the thin, low sweep
wing oset much of the aerodynamic benets of laminar ow. The NLF wing achieves a 4.3%
higher L/D but also weighs 16% more. The increased weight has a even greater impact on
the DOC, which increases with not only fuel burn, but also aircraft weight.
The NLF MLA aircraft achieves tangible savings over the baseline: 9.1% in fuel and 5.6%
in cost. The laminar design achieves greater relative savings than the analogous turbulent
MLA aircraft the thinner, less structurally ecient wings of the laminar design benets
more from active load control.
As in the case of the turbulent designs, the introduction of just GLA does not change
the optimized laminar design. The most signicant gains again come from the simultaneous
application of MLA and GLA. The Laminar MLA+GLA aircraft achieves 18.4% fuel and
11% cost savings relative to the turbulent baseline. The aerostructural improvements leads
to a 20% increase in L/D at essentially the same wing weight.
C. Turbulent and Laminar Comparison
The turbulent and laminar MLA+GLA are compared in Table 8. The turbulent design gains
from improved structural eciency. The laminar design invests the weight savings towards
achieving extensive laminar ow.
The trends in Figure 18 demonstrate that active load alleviation increases the Mach
number at which NLF can be eciently exploited. NLF aircraft without load alleviation
have to cruise at lower Mach numbers to realize tangible cost savings. The NLF aircraft
27 of 36
0 0.2 0.4 0.6 0.8 1
0.5
0
0.5
1
1.5
2
Spanwise Position
C
l

c
/
c
r
e
f


NLF
NLF MLA
NLF GLA
NLF MLA+GLA
(a) 2.5-g at 30,000 feet
0 0.2 0.4 0.6 0.8 1
0.5
0
0.5
1
1.5
2
Spanwise Position
C
l

c
/
c
r
e
f


NLF
NLF MLA
NLF GLA
NLF MLA+GLA
(b) 1.3-g at 40,000 feet
Figure 17. Normalized laminar maneuver lift distributions. (c
ref
= 10 feet)
Turb. MLA+GLA NLF MLA+GLA
W
wing
/W
wing
baseline
0.879 1.002
W
fuel
/W
fuel
baseline
0.888 0.817
T
0
/T
0baseline
0.898 0.838
(L/D)/(L/D)
baseline
1.082 1.196
DOC/DOC
baseline
0.927 0.891
Table 8. The optimized MLA+GLA aircraft parameters normalized by the turbulent baseline.
0.7 0.72 0.74 0.76 0.78 0.8 0.82
0.9
0.95
1
1.05
Cruise Mach Number
R
e
l
a
t
i
v
e

D
i
r
e
c
t

O
p
e
r
a
t
i
n
g

C
o
s
t


Turb.
Turb. MLA+GLA
NLF
NLF MLA+GLA
Figure 18. Relative DOC of turbulent and laminar designs as functions of the cruise Mach
number.
28 of 36
(with no load alleviation) achieves the minimum DOC cruising near Mach 0.72. Even then,
the cost savings over the Mach 0.78 turbulent aircraft is only about 2%. The NLF designs
with active load control, on the other hand, perform signicantly better than the Turbulent
MLA+GLA aircraft up to Mach 0.78. Load alleviation shifts the optimum towards designs
with low-sweep, NLF wings.
A transonic NLF transport would likely carry more stringent maintenance and surface
nish requirements than its turbulent counterpart. The ATA model does capture the addi-
tional manufacturing and maintenance costs that may be associated with a NLF wing. The
estimated NLF savings are therefore likely to be non-conservative. Given that a design needs
to show signicant cost and environmental performance to justify an industrial program, the
results suggest two potential paths forward: 1) slow down and leverage NLF to reduce fuel
burn or 2) combine active load alleviation with NLF to achieve both cost and fuel savings at
high speed.
VI. Sensitivity Studies
The design studies in Section V are based on two important assumptions: 1) NLF wing
sweep are restricted to 10

or less and 2) control surface rates are limited to less than 30

/s.
In this section, we examine the sensitivity of the design to these assumptions.
A. NLF Wing Sweep
For completeness the low sweep NLF designs discussed in Section V should be measured
against comparable high-sweep NLF congurations, whose wing pressure distributions are
tailored to suppress both streamwise and crossow instabilities. More sophisticated ow
solvers are needed to analyze these crossow dominated boundary layers.
11,12
However, by
simply omitting the aerodynamic tradeos needed to stabilize the crossow boundary layer,
we can obtain an optimistic estimate of what 3-D boundary layer control can achieve.
The NLF-25 designs in Figure 19 are optimized with an allowable wing sweep of up
to 25

. In designs that do not have load alleviation, the increased allowable wing sweep
has a measurable impact on performance. In the MLA+GLA design the additional sweep
only becomes useful at high speed (Mach 0.8 and beyond). As load control minimizes the
structural penalties of a thin wing, there is less incentive to use wing sweep to manage
compressibility eects.
29 of 36
0.7 0.72 0.74 0.76 0.78 0.8 0.82
0.9
0.95
1
1.05
Cruise Mach Number
R
e
l
a
t
i
v
e

D
i
r
e
c
t

O
p
e
r
a
t
i
n
g

C
o
s
t


NLF
NLF25
NLF MLA+GLA
NLF25 MLA+GLA
Figure 19. Optimized aircraft cost as functions of Mach number.
B. GLA Bandwidth
Controller and actuator bandwidths can limit GLA eectiveness. Figure 20 shows the
variation of DOC as a function of control surface deection rates.
0 5 10 15 20 25 30 35
0.9
0.95
1
1.05
Load Control Deflection Rate (Deg/s)
R
e
l
a
t
i
v
e

D
i
r
e
c
t

O
p
e
r
a
t
i
n
g

C
o
s
t


Turbulent MLA+GLA
NLF MLA+GLA
Figure 20. MLA+GLA aircraft costs as functions of ap rate.
For turbulent designs the DOC trend attens near a slew rate of 25

/s. The more gust-


sensitive low-sweep laminar designs see performance gains up to 30-35

/s. High-bandwidth
actuators can clearly reduce the impact of high frequency gusts. However, gusts that are
fast enough to saturate a high-speed controller can also be too fast to excite any structural
response. Commercial aircraft ailerons, which have slew rates of 35-40

/s,
62
should support
eective GLA.
30 of 36
VII. Conclusions
We develop a method to concurrently design the aircraft with its load alleviation system.
The design framework is used to study the synergy between active load alleviation and
natural laminar ow. The gust encounter simulations incorporate both aircraft and wing
structural dynamics to capture the frequency dependence of aircraft gust response. By
directly operating on the wing pressure distribution, we also capture the trade-o among
structure eciency, compressibility drag and natural laminar ow.
We observe that the independent applications of MLA can net modest gains. The savings
increase signicantly when we simultaneously apply both MLA and GLA. A minimum cost
turbulent design with MLA and GLA achieves a 11% reduction in fuel burn and 7% reduction
in cost.
Active load alleviation reduces the structural penalties associated with thin wings, and
makes low sweep, NLF wings practical at high speed. The fuel and cost savings grow to 18%
and 11% respectively as the optimizer invests the weight savings from load alleviation to
achieve extensive laminar ow,
Sensitivity studies show that active load alleviation can shift the optimal transonic con-
guration towards high-span, low-sweep laminar ow wings. The control requirements for
eective GLA are also likely within the parameters of todays commercial aircraft actuators.
Future research can investigate the impact of more sophisticated GLA control policies at
the level of aircraft design. The control analysis should account for measurement uncertainties
and time delays. The eects of static aeroelasticity from wing bending should also be added
to improve the accuracy of the results, particularly for highly swept wings. Finally, utter
and its suppression are important considerations for high span, exible wing, and should be
included in future studies.
VIII. Acknowledgment
This research is supported by the NASA Subsonic Fixed Wing Project.
IX. Appendix A: Validation
The multidisciplinary and coupled nature of the design framework complicates system-
level validation. The inverse wing design methodology, in particular, makes validation
against known congurations dicult. We take therefore a component-based approach to
the validation problem.
We validate our implementation of the Weissinger panel method against known solutions
computed using LinAir
42
a commercial vortex lattice code. The validation cases form an
31 of 36
integral part of the unit test suite. The C
l
distribution, wing lift curve slope and induced
drag computed using the Weissinger method module exactly match LinAir solutions for
representative swept wings.
The inverse design, boundary layer solver and transition prediction tools are validated
using XFoil.
70
We use XFoil to solve for the C
p
distribution and boundary layer properties of
airfoils designed using the inverse solver. The C
p
outputs from XFoil show good agreement
with the C
p
inputs of the inverse solver. Unlike XFoil, the solver discussed in Section III
is not interactive. Some dierences in the results are therefore expected. Agreement also
breaks down near the leading edge, where the small disturbance assumptions of thin airfoil
theory, which forms the basis of our inverse solver, do not hold.
The FEM kernel and beam element formulations are validated against analytic solutions
of simple beam deections. Figure 21 compares the static deection and moment distribution
of a uniform cantilever under uniform distributed load against analytic results.
0 0.5 1
0
0.05
0.1
0.15
Displacement
x
z
0 0.5 1
0
0.2
0.4
0.6
Moment
x


FEM
Exact
Figure 21. FEM uniform beam deections under uniform distributed load.
The results of the modal decomposition is compared against the analytic solution of
a cantilever beam in free vibration. The natural frequency in Table 9 and mode shapes
in Figure 22 agree with the analytic solutions.
Analytic Solution FEM Solution (50 Elements)
Mode 1 3.516 3.516
Mode 2 22.035 22.034
Mode 3 61.698 61.701
Table 9. Analytic and FEM natural frequencies of a uniform cantilever beam in free vibration.
To validate the time-integration of the modal equations of motion, we compare numerically
integrated responses of second order linear systems (with varying damping and natural
frequencies) against known analytic solutions. The shapes, overshoots and peak times of the
numerical solution match their analytic counterparts.
32 of 36
0 0.2 0.4 0.6 0.8 1
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
First 3 Normalized Modes for a Cantilever Beam
x
z


Mode 1
Mode 2
Mode 3
Figure 22. FEM (circle) and analytic (solid) mode shapes of a uniform cantilever beam in free
vibration.
It is acknowledged that more validations may be needed to fully anchor the wing weight
and drag buildup.
References
1
Marty, B. and Christopher, D., Subsonic Ultra Green Aircraft Research: Phase I Final Report, NASA
Technical Report, Vol. NASA/CR2011-216847, 2011.
2
White, R., Improving the Airplane Eciency by Use of Wing Maneuver Load Alleviation, Journal
of Aircraft, Vol. 8, 1971, pp. 769.
3
Arcara Jr, P., Bartlett, D., and McCullers, L., Analysis for the Application of Hybrid Laminar Flow
control to a Long-Range Subsonic Transport Aircraft, National Aeronautics and Space Administration,
Hampton, VA., 1991.
4
Airbus, A319/A320/A321 Flightdeck and Systems Brieng for Pilots, 1998.
5
Airbus, A330 and A340 Flight Crew Training Manual , 2004.
6
Flaig, A., Solutions to the Aerodynamic Challenges of Designing the Worlds Largest Passenger
Aircraft, Royal Aeronautical Society Hamburg Branch Lecture Series, 2008.
7
Livne, E., Integrated Aeroservoelastic Optimization: Status and Direction, Journal of Aircraft,
Vol. 36, No. 1, 1999, pp. 122145.
8
Haghighat, S., Martins, J. R. R. A., and T. Liu, H. H., Aeroservoelastic Design Optimization of a
Flexible Wing, Journal of Aircraft, Vol. 49, No. 2, 2012, pp. 432443.
9
Haghighat, S., Liu, H. H. T., and Martins, J. R. R. A., A Model Predictive Gust Load Alleviation
Controller for a Highly Flexible Aircraft, Journal of Guidance, Control and Dynamics, Vol. 36, 2012,
pp. 17511766.
10
Joslin, R., Aircraft Laminar Flow Control, Annual Review of Fluid Mechanics, Vol. 30, No. 1, 1998,
pp. 129.
11
Green, J., Laminar Flow Control: Back to the Future? 38th AIAA Fluid Dynamics Conference and
Exhibit, 2008.
12
Allison, E. and Kroo, I., Aircraft Conceptual Design with Laminar Flow, 27
th
International Congress
of the Aeronautical Sciences, 2010.
33 of 36
13
Rajnarayan, D. and Sturdza, P., Extensible Rapid Transition Prediction for Aircraft Conceptual
Design, 29
th
AIAA Applied Aerodynamics Conference, 2011.
14
Szodruch, J. and Hilbig, R., Variable Wing Camber for Transport Aircraft, Progress in Aerospace
Sciences, Vol. 25, No. 3, 1988, pp. 297328.
15
Carter, D., Osborn, R., Hetrick, J., and Kota, S., The Quest for Ecient Transonic Cruise, 7
th
AIAA
ATIO Conference, Belfast, Northern Ireland, 2007.
16
Hilbig, R. and Szodruch, J., The Intelligent Wing Aerodynamic Developments for Future Transport
Aircraft, 27th AIAA Aerospace Sciences Meeting, January 9-12, 1989/Reno,Nevada, 1989.
17
Ning, S. and Kroo, I., Multidisciplinary Considerations in the Design of Wings and Wing Tip Devices,
Journal of Aircraft, Vol. 47, No. 2, 2010.
18
Lehner, S., Lurati, L., Smith, S., Bower, G., Cramer, E., Crossley, W., Engelson, F., Kroo, I., Smith,
S., and Willcox, K., Advanced Multidisciplinary Optimization Techniques for Ecient Subsonic Aircraft
Design, AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition,
2010.
19
Karr, C., Zeiler, T., and Mehrotra, R., Determining Worst-Case Gust Loads on Aircraft Structures
Using an Evolutionary Algorithm, Applied Intelligence, Vol. 20, No. 2, 2004, pp. 135145.
20
Scott, R., Pototzky, A., and Perry, B., A Computer Program to Obtain Time-Correlated Gust
Loads for Nonlinear Aircraft Using the Matched Filter Based Method, National Aeronautics and Space
Administration, 1994.
21
Pototzky, A. and Zeiler, T., Maximum Dynamic Responses Using Matched Filter Theory and Random
Process Theory, National Aeronautics and Space Administration, 1988.
22
Jones, J., Statistical Discrete Gust Method for Predicting Aircraft Loads and Dynamic Response,
Journal of Aircraft, Vol. 26, 1989, pp. 382392.
23
Watson, G., Gilholm, K., and Jones, J., A Wavelet-based Method for Finding Inputs of Given Energy
which Maximize the Outputs of Nonlinear Systems, International Journal of Systems Science, Vol. 30,
No. 12, 1999, pp. 12971307.
24
Fidkowski, K., Engelsen, F., Willcox, K., and Kroo, I., Stochastic Gust Analysis Techniques for
Aircraft Conceptual Design, AIAA Paper Number 2008-5848, 2008., 2008.
25
Rao, S. S., Optimization of Airplane Wing Structures under Gust loads, Computers & Structures,
Vol. 21, No. 4, 1985, pp. 741749.
26
Suzuki, S. and Yonezawa, S., Simultaneous Structure/Control Design Optimization of a Wing Struc-
ture with a Gust Load Alleviation System, Journal of Aircraft, Vol. 30, No. 2, 1993, pp. 268274.
27
Yang, J., Nikolaidis, E., and Haftka, R., Design of Aircraft Wings Subjected to Gust Loads: A System
Reliability Approach, Computers & Structures, Vol. 36, No. 6, 1990, pp. 10571066.
28
Petersson, O., Optimization of Aircraft Wings Including Dynamic Aeroelasticity and Manufacturing
Aspects, 5th AIAA Multidisciplinary Design Optimization, 2009.
29
Aouf, N., Boulet, B., and Botez, R., Robust Gust Load Alleviation for a Flexible Aircraft, Canadian
Aeronautics and Space Journal , Vol. 46, No. 3, 2000, pp. 131139.
30
Lucas, A., Valasek, J., and Strganac, T., Gust Load Alleviation of an Aeroelastic System using
Nonlinear Control, 50
th
AIAA Structures, Structural Dynamics, and Materials Conference, 2009.
31
Moulin, B. and Karpel, M., Gust Loads Alleviation Using Special Control Surfaces, Journal of
Aircraft, Vol. 44, No. 1, 2007, pp. 17.
34 of 36
32
Kroo, I., An Interactive System for Aircraft Design and Optimization, Journal of Aircraft, Vol. 92-
1190, 1992.
33
Xu, J. and Kroo, I., Aircraft Design with Maneuver and Gust Load Alleviation, 29
th
AIAA Applied
Aerodynamics Conference, Honolulu, Hawaii, 2011.
34
Xu, J. and Kroo, I., Aircraft Design with Maneuver Load Alleviation and Natural Laminar Flow,
11
th
AIAA Aviation Technology, Integration, and Operations Conference, Virginia Beach, Virginia, 2011.
35
European Aviation Safety Agency, CFM International CFM56-7B Series Engine Type Certication
Data Sheet, 2012.
36
Cramer, E. J., Dennis, J. E., Frank, P. D., Lewis, R. M., and Shubin., G. R., Problem Formulation
for Multidisciplinary Optimization, SIAM Journal on Optimization, Vol. 4, No. 4, 1994, pp. 754776.
37
Kroo, I., MDO for Large-Scale Design, Multidisciplinary Design Optimization: State of the Art.
SIAM, Philadelphia, 1997, pp. 2244.
38
Martins, J. R. R. A. and Lambe, A. B., Multidisciplinary Design Optimization: A Survey of Archi-
tectures, AIAA Journal , 2013.
39
Gallman, J. and Kroo, I., Structural Optimization for Joined Wing Synthesis, Journal of Aircraft,
Vol. 33, No. 1, 1996, pp. 214223.
40
Gallman, J. W., Structural and Aerodynamic Optimization of Joined Wing Aircraft, Ph.D. thesis,
Stanford University, 1992.
41
Bertin, J. and Smith, M., Aerodynamics for Engineers, Prentice Hall, 1998.
42
Kroo, I., LinAir 4: A Nonplanar, Multiple Lifting Surface Aerodynamics Program, Desktop Aeronautics,
2011.
43
Wakayama, S. and Kroo, I., Subsonic Wing Planform Design Using Multidisciplinary Optimization,
Journal of Aircraft, Vol. 32, No. 4, 1995, pp. 746753.
44
Abbott, I. and Von Doenho, A., Theory of Wing Sections, Dover, 1959.
45
Lee, J. and Jameson, A., Natural Laminar Flow Airfoil and Wing Design by Adjoint Method and
Automatic Tansition Prediction, 47th AIAA Aerospace Sciences Meeting, 2009.
46
Cella, U., Quagliarella, D., Donelli, R., and Imperatore, B., Design and Test of the UW-5006 Transonic
Natural Laminar Flow Wing, Journal of Aircraft, Vol. 47, No. 3, 2010, pp. 783795.
47
Stratford, B., The Prediction of Separation of the Turbulent Boundary Layer, Journal of Fluid
Mechanics, Vol. 5, No. 01, 1959, pp. 116.
48
Lock, R., An Equivalence Law Relating Three and Two-Dimensional Pressure Distributions, Aero-
nautical Research Council , 1964.
49
Harris, C. D., Aerodynamic Characteristics of a 14 Percent Thick NASA Supercritical Airfoil Designed
for a Normal Force Coecient of 0.7, National Aeronautics and Space Administration, Hampton, VA. Langley
Research Center, 1975.
50
Harris, C. D., NASA Supercritical Airfoils: a Matrix of Family-related Airfoils, NASA Technical
Paper 2969, 1990.
51
Obayashi, S. and Takanashi, S., Genetic Optimization of Target Pressure Distributions for Inverse
Design Methods, AIAA Journal , Vol. 34, 1996, pp. 881886.
52
Cebeci, T. and Cousteix, J., Modeling and Computation of Boundary Layer Flows: Solutions Manual
and Computer Programs, Springer, 2001.
35 of 36
53
Somers, D., Design and Experimental Results for a Natural Laminar Flow Airfoil for General Aviation
Application, National Aeronautics and Space Administration, Scientic and Technical Information Branch,
1981.
54
Vijgen, P., Dodbele, S., Holmes, B., and Van Dam, C., Eects of Compressibility on Design of
Subsonic Natural Laminar Flow Fuselages, Journal of Aircraft, Vol. 25, No. 9, 1986, pp. 776782.
55
Arnal, D. and Vermeersch, O., Compressibility Eects on Laminar-turbulent Boundary Layer Transi-
tion, International Journal of Engineering Systems Modeling and Simulation, Vol. 3, No. 1, 2011, pp. 2635.
56
Wazzan, A., Gazley Jr, C., and Smith, A., HR-x Method for Predicting Transition, AIAA Journal ,
Vol. 19, 1981, pp. 810812.
57
Smith, A. and Stokes, T., Optimum Tail Shapes for Bodies of Revolution, Journal of Hydronautics,
Vol. 15, 1981, pp. 6773.
58
Drela, M., Simultaneous Optimization of the Airframe, Powerplant, and Operation of Transport
Aircraft, Hamilton Place, London, 2010.
59
Bindolino, G., Ghiringhelli, G., Ricci, S., and Terraneo, M., Multilevel Structural Optimization for
Preliminary Wing Box Weight Estimation, Journal of Aircraft, Vol. 47, No. 2, 2010, pp. 475489.
60
Xu, J., Aircraft Design with Active Load Al leviation and Natural Laminar Flow, Ph.D. thesis, Stanford
University, 2012.
61
Xu, J. and Kroo, I., Aircraft Design with Active Load Alleviation and Natural Laminar Flow, 53
rd
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2012.
62
Aviation Safety and Pilot Control: Understanding and Preventing Unfavorable Pilot-Vehicle Interac-
tions, National Research Council, 1997.
63
Gaulocher, S. L., Roos, C., and Cumer, C., Aircraft Load Alleviation During Maneuvers Using
Optimal Control Surface Combinations, Journal of Guidance, Control, and Dynamics, Vol. 30, No. 2, 2007,
pp. 591600.
64
Fuller, J., Evolution and Future Development of Airplane Gust Loads Design Requirements, SAE
International, 1997.
65
Rao, S., The Finite Element Method in Engineering, Butterworth-heinemann, 2005.
66
Yang, T., Finite Element Structural Analysis, Vol. 450, Prentice-Hall Englewood Clis, NJ, 1986.
67
Ibrahimbegovic, A. and Wilson, E., Simple Numerical Algorithms for the Mode Superposition Analysis
of Linear Structural Systems with Non-proportional Damping, Computers & Structures, Vol. 33, No. 2,
1989, pp. 523531.
68
Thomas, E., ATA Direct Operating Cost Formula for Transport Aircraft, SAE Technical Papers,
1966.
69
Boeing, 737 Airplane Characteristics for Airport Planning, 09/2013.
70
Drela, M., XFOIL: An Analysis and Design System for Low Reynolds Number Aerodynamics, Con-
ference on Low Reynolds Number Aerodynamics, 1989.
36 of 36

Potrebbero piacerti anche