Vous êtes sur la page 1sur 13

Multiobjective Optimization of Earth-Entry Vehicle Heat Shields

Joshua E. Johnson

and Mark J. Lewis

University of Maryland, College Park, Maryland 20742


and
Ryan P. Starkey

University of Colorado, Boulder, Colorado 80309


DOI: 10.2514/1.42565
A differential evolutionary algorithm has been executed to optimize the hypersonic aerodynamic and stagnation-
point heat transfer performance of Earth-entry heat shields for manned and unmanned missions. Objective
functions comprise maximizing cross range, minimizing heat ux, and minimizing heat load. Each considered heat-
shield geometry is composed of an axial prole tailored to t a base cross section. Axial proles consist of spherical
segments, spherically blunted cones, and power lawgeometries. Heat-shield cross sections include oblate and prolate
ellipses, rounded-edge parallelograms, and blendings of the two. Aerothermodynamic models are based on modied
Newtonian impact theory with semiempirical correlations for convection and radiation. Entry velocities of 11 and
15 km=s are used to simulate atmospheric entry conditions at lunar and Mars return conditions, respectively.
Results indicate that skip trajectories allow for vehicles with a low lift-to-drag ratio of 0.25 to achieve 1000 km cross
range, a factor of 4 increase in capability over direct entry. For 11 km=s entry with a 6g deceleration limit, the
spherical segment provides optimal performance. For 15 km=s entry with a 12g deceleration limit, the spherically
blunted cone produces an 8% lower heat ux when compared with spherical segments with similar aerodynamic
characteristics. This result is attributed to the blunted cones smaller shock standoff, which reduces the heat load
generated by thermal radiation, the dominant heat transfer mode.
Nomenclature
A = coefcient of power law
BC = ballistic coefcient, kg,m
2
l = exponent of power law
C = aerodynamic coefcient
D = drag, N
J = diameter, m
E = total emitted power density, J,(m
3
s)
e = eccentricity
g
n
= ratio of wall enthalpy to total enthalpy
g
1
, g
2
, g
3
= coefcients and exponents [Eq. (6)]
H = exponent [Eq. (7)]
h = length along y direction, m
h
i
= altitude, km
] = semimajor axis length of a blunt body, m
]
max
= maximum mesh points in x direction
k = semiminor axis length of a blunt body, m
k
max
= maximum mesh points in direction
L = lift, N
l = length along x direction, m
M
o
= freestream Mach number
m = mass, kg
m
1
= number of sides of the superellipse
^ n = normal unit vector, away from surface
n
crew
= crew number, person
n
max
= peak deceleration load, Earth g
n
1
, n
2
, n
3
= superelliptic parameters

xrs
= cross range, km
Q = heat load, kJ,cm
2
q = heat ux, W,cm
2
q
o
= freestream dynamic pressure, Pa
r = base radius, m
r
n
= nose radius of blunted cone, m
r
s
= radius of curvature of spherical segment, m
S = area of base cross section, m
2
i = time, s
i
J
= total mission duration, days
V
PR
= pressurized volume, m
3
:
1
, :
2
= superellipse parameters
V
o
= freestream velocity, m,s
x, y, z = coordinate values, m
o = angle of attack,

[ = sideslip angle,

; = trajectory ight-path angle; positive pointing away
from planet,

so
= shock-standoff distance, m
c = edge tangency angle,

j
:
= volumetric efciency
0
c
= half-cone angle,

0
s
= half-spherical segment angle,

, = density, kg,m
3
,
2
,,
1
= normal-shock density ratio
= sweep angle, rad

l
= bank angle,

Subscripts
A = axial force, N
l = base
cg = center of gravity
conv = convective
D = drag, N
EI = entry interface, km (for Earth, h
i
is 122 km)
eff = effective
EV = entry vehicle
] = nal
Received 4 December 2008; revision received 4 August 2011; accepted for
publication 4 August 2011. Copyright 2011 by the University of Maryland.
Published by the American Institute of Aeronautics and Astronautics, Inc.,
with permission. Copies of this paper may be made for personal or internal
use, on condition that the copier pay the $10.00 per-copy fee to the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include
the code 0022-4650/12 and $10.00 in correspondence with the CCC.

Graduate Research Assistant, Department of Aerospace Engineering,


3181 Glenn L. Martin Hall; currently Analyst, Johns Hopkins University,
Applied Physics Laboratory, 11100 Johns Hopkins Road, Laurel, Maryland
20723; Joshua.Johnson@jhuapl.edu. Member AIAA.

Professor, Department of Aerospace Engineering, 3181 Glenn L. Martin


Hall; lewis@umd.edu. Fellow AIAA.

Assistant Professor, Aerospace Engineering Sciences Department, 429


UCB; rstarkey@colorado.edu. Associate Fellow AIAA.
JOURNAL OF SPACECRAFT AND ROCKETS
Vol. 49, No. 1, JanuaryFebruary 2012
38
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

H = horizontal component relative to orbital plane
HS = heat shield
L = lift, N
l, m, n = rolling, pitching, yawing moments, N m
max = point of maximum heat transfer
N = normal force, N
= pressure
rad = radiative
s = stagnation point
sl = sea level
tot = total
V = vertical component relative to orbital plane
Y = side force, N
o = derivative with respect to o, ,rad
[ = derivative with respect to [, ,rad
I. Introduction
V
EHICLES returning to Earth from the moon or Mars must
survive an extreme hypersonic environment during atmos-
pheric entry. This involves entering Earths atmosphere at super-
orbital velocities ranging from 10 to 15 km,s with corresponding
Mach numbers from 30 to 50 [1] while withstanding 3000 K
temperatures. The forward heat shield, which protects the vehicle, is
the primary contributor to the vehicles aerothermodynamic perfor-
mance, i.e., the aerodynamic forces, moments, and heat transfer [2].
The rest of the vehicle experiences signicantly lower heat uxes and
regions of extremely low pressure due to ow separation.
Optimizing aerodynamic and heat transfer performance are
conicting objectives. To reduce surface heating and maximize
thermal energy transferred to the surrounding environment, vehicles
designed for hypervelocity entry have typically been blunt-body
designs that limit entry aerodynamic performance [3]. Conse-
quently, an aerothermodynamic balance must be achieved to satisfy
mission requirements without exceeding material technology
constraints.
In this work, aerothermodynamic performance is represented by
crossrange capability and stagnation-point heating. Crossrange
capability enables missions that require diverting from the vehicles
orbital ground track. It also allows an entry vehicle to execute course
corrections to counter offnominal atmospheric conditions and switch
landing sites [4]. Stagnation-point heat transfer, although it may not
be the point of maximum heating, is a good relative measure of the
expected high heating over the heat shield. From an overall
perspective, both are directly associated with mission requirements
and material constraints. This work seeks to nd optimal blunt-body
heat-shield designs from the standpoints of cross range and
stagnation-point heat transfer. Acomplete trajectory model with low-
order aerothermodynamic models is applied to balance the need for
delity with the desire to have practical computational times,
allowing the optimizer to consider a wide range of heat-shield
geometries.
Previous work has primarily emphasized heat shields based on
spherical and conic shapes: the ~25

spherical segment (SS) from


Project Apollo [58], and the 70

spherically blunted cone (SC) from


the Mars Viking [911] missions. Although these shapes have been
used over the past 40 years, it is unknown whether either provides
optimal aerothermodynamic performance for lunar and Mars return
missions. Properly broadening the design space would allow the
optimizer to determine which geometric features improve perform-
ance. Recent work completed by the authors [12] introduced non-
spherical designs consisting of elliptical and polygonal base cross
sections. Single design-condition optimization indicated improve-
ments in aerodynamic performance by generating oblate heat shields
with parallelogram cross sections, increasing the hypersonic lift-to-
drag ratio signicantly. In general, a larger lift-to-drag ratio increases
crossrange capability [4]. For both spherical and nonspherical heat-
shield designs, the present work extends the recent work by
incorporating entry trajectory analysis, thus allowing the optimizer to
account for peak heat ux, total heat load, cross range, and
deceleration loads.
To produce optimal tradeoff relationships between performance
parameters, this work uses a population-based multiobjective
optimization scheme, in which a differential evolutionary algorithm
is employed to optimize cross range and heat transfer simultaneously.
The optimized trajectories may be located on the bounds of the
feasible design space and likewise be sensitive to small deviations in
bank angle. Thus, the trajectories themselves may not be yable
during a real mission but are reported to recognize optimal perfor-
mance boundaries for each heat-shield/trajectory conguration.
Contributions of viscous shear forces and turbulence are not con-
sidered. The overall level of analysis presented in this work is
appropriate for assessment of the trade space at the conceptual design
level for heat-shield geometries.
II. Methodology
A. Mission Prole
To simulate Earth entry for lunar return, an initial entry velocity of
11 km,s is applied [13]. For Mars return, a fast 180 day return
requires entry velocities up to 14.7 km,s [14], and an initial entry
velocity of 15 km,s is applied. Although the hypersonic aero-
dynamics at these two velocities are similar for a given heat-shield
design, their heating environments are greatly different. While
convection typically dominates in heat load for a vehicle entering at
11 km,s, radiation is projected to be the primary heat transfer mode
for 15 km,s. The mission prole for the Orion crew exploration
vehicle (CEV) with an overall duration of 18 days, a crewof four, and
a pressurized volume of 5 m
3
,person is applied [15]. The Earth-
entry simulation begins at the atmospheric entry interface, at an
altitude of 122 km, and terminates after the freestreamMach number
becomes less than ve and accounts for the hypersonic aerodynamics
only. For blunt-bodied capsules (low L,D), whether the trajectory
ends at M
o
< 5 does not strongly affect the values of the three
optimization parameters in this analysis: cross range
xrs
, peak
stagnation-point heat ux q
s.max
, and total stagnation-point heat load
Q
s.tot
. This mission prole is used for both lunar and Mars return.
B. Heat-Shield Geometries
The heat-shield geometry is dened by two contours: the base
cross section of the heat shield and the axial prole that is swept about
the central axis and is tailored to t the base cross section. For
optimization, the ideal equation dening the geometry would
produce a wide, continuous range of cross sections. One such
equation is available using the superformula [16] of the superellipse.
It denes the cross-section radius for 0 2:
r() =

cos(
1
4
m
1
)
:
1

n
2

sin(
1
4
m
1
)
:
2

n
3

1,n
1
(1)
in which m
1
corresponds to the number of sides of a polygon
:
1
=:
2
=1, n
1
and n
2
are modiers, and n
3
=n
2
to render sharp or
rounded-edge polygons. Increasing n
2
transforms a polygon
(n
2
<2) into an ellipse (n
2
=2) and then into a rounded-edge
concave polygon (n
2
>2). Example cross sections are shown in
Fig. 1, and ranges of values for parameters m
1
, n
1
, and n
2
are listed in
Table 1. This work uses n
2
2.0 to avoid regions of high heat ux
generated by corner ow that may be present around concave
geometries. Note that an ellipse is generated by Eq. (1) with n
2
=2
independent of n
1
and the chosen type of polygon. For any n
2
, oblate
(e < 0) and prolate eccentricity (e > 0) are considered. Once the
cross section is chosen, the heat shields axial prole, which is the
shape that protrudes fromits base, is selected. Three axial proles are
considered: the SS, the SC, and the power law (PL). To generate a
heat-shield geometry, the axial prole is rendered and then swept
360

about the central body axis using Eq. (1), based on the selected
cross-section contour.
Figures 2 and 3 showthe SS and SCgeometries, respectively. The
PL formula y =Ax
l
offers axial proles with a wide range of
bluntness being controlled by coefcient Aand exponent l, as shown
JOHNSON, LEWIS, AND STARKEY 39
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

in Figs. 4a and 4b, respectively. The xed-body coordinate system
for this work is included in Fig. 2. In the xed-body coordinate
system, it is convention to combine the three orthogonal force vectors
C
A
, C
N
, and C
Y
into the two aerodynamic force vectors C
L
and C
D
.
This results invertical and horizontal components of lift (with respect
to the z axis) represented by L
V
and L
H
, respectively, for the cases
with [ 0. In calculating rolling moment stability, the sign of L
V
determines the direction of the positive rolling moment shown in
Fig. 2. This allows C
l.cg.[
to be negative for all statically roll stable
congurations. The moment coefcients are dened as the ratio of
the moment to q
o
S], in which ] is the length of the semimajor axis of
the base cross section. Additional details of both the static stability
and the heat-shield shapes in this work are given in [17].
C. Entry Vehicle Mass Estimation
Mass estimation and scaling of the entry vehicle are based on the
mission prole, heat-shield geometry, and dimension requirements
for incorporating the crew or payload. Vehicle scaling is necessary
due to the wide range of heat shields in the design space. High 0
s
, for
example, can allow the heat shield to encompass part of or the entire
pressurized volume. If the entry vehicle geometry is assumed to have
the same base height h
l
(top to bottom) as the Orion with 5 m, then
heat shields with base cross sections of high eccentricity would have
over 15 times Orions volume. To scale the entry vehicle, a heat shield
is categorized into one of four cases based on l
HS
. This differentiates
the procedures applied for estimating the pressurized volume based
on the heat shields geometry. The scaled vehicle must also satisfy
required crew seating dimensions. It is assumed that the seat dimen-
sions required for a suited astronaut consist of a top-to-bottomheight
of 1.4 m, a width of 0.7 m, and a depth of 1.1 m.
Once the pressurized volume is closely matched and seat
dimensions are satised, the entry vehicle mass is estimated based on
the following empirical correlation [18]:
m
EV
=592(n
crew
i
J
V
PR
)
0.346
(2)
Based on the mission prole, Orions estimated mass of 7330 kg is
within 1%of the landing mass circa 2006 reported in the early design
phase [15]. The entry vehicle masses for this analysis are nearly
constant since heat-shield scaling is designed to produce a pres-
surized volume that meets the mission requirements as closely as
possible. Details on the method of calculation are documented in
[19]. This mass estimate is independent of heat load, which
determines the heat-shield materials thickness in a detailed design
analysis. Several new heat-shield designs are considered; thus, the
heat load for a given vehicle and ight path is unknown a priori. The
required iterative process, which would increase the computation
time by a few factors, has not been integrated into the optimization
setup. Some of the new heat-shield designs would require
unshrouded launch due to their size. The presence of a shroud is an
aerodynamic requirement on some launch vehicle designs, but there
have been reentry vehicles launched unshrouded on successful test
ights including the ve aerothermodynamic elastic structural
systems environmental tests from 19631965 [20]. In this program,
the aerothermodynamic structural test vehicle was attached to the top
of a Thor launch vehicle. For unshrouded launch, there could be areas
of the reentry vehicle that are protected by the heat shield during
reentry, but not during the launch phase, and may require additional
material and weight for thermal protection.
Uniform density has been assumed here to calculate the center of
gravity location of the heat shield, and the prescribed X
cg
has been
arbitrarily modied to equal 75% of the uniform density value.
Bringing the X
cg
forward increases the feasible design space by
allowing more slender blunt bodies with higher L,D to be
longitudinally statically stable. Using the uniform density assump-
tion is a limitation, as it does not account for the specic placement of
subsystems within the reentry vehicle geometry. In reentry vehicle
design, the placement of the center of gravity is important and
represents a signicant challenge. In some cases, ballasting or
redistributing mass may be needed to achieve the desired placement.
In reality, reentry vehicle conceptual design may generate a
signicantly different center of gravity than the uniform density
assumption used in this heat-shield shape analysis.
Fig. 1 Examples of base cross sections.
Table 1 Superformula parameters for
rounded-edge polygons
a
(n
3
n
2
)
m
1
n
1
4 1.00
5 1.75
6 2.30
7 3.20
8 4.00
9 5.50
10 7.00
a
1.3 n
2
2.0
Yawing
x
z

s
r
s
y
Rolling
C
L,V
> 0
C
L,V
< 0
Pitching
Fig. 2 SS,
s
60

, n
2
n
3
2.

r
n d

l

c
Fig. 3 SC axial prole, r
n
=d 0:25,
c
60

.
40 JOHNSON, LEWIS, AND STARKEY
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

D. Trajectory Modeling and Validation
The University of Maryland Parallel Trajectory Optimization
Program(UPTOP) is applied for a three-degree-of-freedom(3-DOF)
entry trajectory analysis [21,22]. Its capability covers not only
trajectory optimization but also heat-shield shape optimization,
which is discussed in Sec. II.F. UPTOP uses a fourth-order Runge
Kutta routine to propagate the point-mass equations of motion for
rigid-body ight in a vertical plane [2123]. The time step is
normally set to 1 s. Arotating, ellipsoidal Earth model is applied with
a second harmonic gravity model J2 based on the WGS-84
Geocentric Equipotential Ellipsoid model [24]. The U.S. 1976
Standard Atmosphere [25] is applied for h
i
< 85 km, and the
NRLMSISEE-00 Atmosphere [26] is applied for h
i
85 km.
Trajectories generated with UPTOP are compared with those of
the benchmark Program to Optimize Simulated Trajectories (POST)
[27] inFig. 5. UPTOPis capable of optimizing multistage trajectories
where the vehicle may have multiple engines and fuel tanks. The
benchmark case [28] for the optimal space shuttle transport ascent
trajectory through space shuttle main engine cutoff is provided in
Fig. 5a to demonstrate UPTOPs comparable optimization capability.
The optimal pitch and altitude proles generated by UPTOP closely
match POSTs optimal proles. Additionally, the results from POST
as calculated by UPTOP match the POST proles. Validation for
Earth entry fromlunar returnat V
E
=11 km,s is givenin Fig. 5b. For
the given bank angle prole, which rotates the lift vector, the skip
trajectory generated in UPTOP matched POSTs and illustrates
UPTOPs suitability for high-velocity entry applications. The 3-DOF
entry analysis assumes that the reentry vehicles are dynamically
stable, and a higher-order analysis would be required to determine
where in the design space this is and is not the case.
E. Aerothermodynamics
The primary physical mechanisms that contribute to the hyper-
sonic aerothermodynamics, i.e., the aerodynamic forces, moments,
and heat transfer [2], in this work consist of 1) the surface pressure
distribution, 2) the velocity gradient along the heat shield, and 3) the
radiating shock layer. The local bow shock strength, imposed on the
vehicle by freestream conditions, strongly affects both the surface
pressure distribution and the resulting heat transfer along the heat
shield. The surface pressure distribution produces a velocity gradient
along the heat shield and sets the velocity at the edge of the boundary
layer. The resulting boundary-layer velocity gradient results in
convective heat transfer at the surface of the heat shield. The
thickness of the high-temperature shock layer inuences the thermal
radiative heat ux. Low-order methods have been applied to
determine the effects of these physical mechanisms on the aero-
thermodynamic performance.
1. Aerodynamics
Modied Newtonian ow theory is applied to produce the
hypersonic surface pressure distribution about a blunt-body heat-
shield design. This method assumes that the freestreamows normal
momentum is destroyed upon impact with the vehicle and that the
pressure is negligible on the portion of the vehicle not facing the
freestream[29]. The expression for the modied Newtonian pressure
coefcient is
C

=C
.max
(V
o
^ n,V
o
)
2
(3)
for V
o
^ n <0, and C

=0 for V
o
^ n 0, which represents the
region in aerodynamic shadow. The maximum pressure coefcient
Fig. 4 PL axial prole.
Fig. 5 Trajectory validation of UPTOP results with POST.
JOHNSON, LEWIS, AND STARKEY 41
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

occurs at the stagnation point, and it is calculated according to the
Rayleigh pitot tube formula [29]. Figure 6 presents Newtonian
results plotted against wind-tunnel data of the Apollo command
module [30]; note the excellent agreement. An in-depth validation of
modied Newtonian ow is provided in [17] and indicates that
results are generally within 10% for aerodynamic coefcients for
high Earth-entry velocities from both low Earth orbit and the moon.
2. Convective Heat Transfer
The properties of the high-temperature shock layer affect both
convection and radiation due to the presence of dissociated and
partially ionized air. Additionally, the heat-shield geometry directly
affects the surface pressure distribution, and thus the velocity
gradient along the heat shield at the edge of the boundary layer. A
smaller local radius of curvature increases the velocity gradient,
thereby increasing the local convective heat ux [3]. Low-order
correlations based on empirical data account for these upstream
effects with a local radius-of-curvature term. The stagnation-point
convective heat ux has been correlated [31] to nose radius and
freestream conditions:
q
s.conv
=(1.83 10
8
)r
0.5
n
(1 g
n
),
0.5
o
V
3
o
(4)
in which this work assumes g
n
1.
3. Radiative Heat Transfer
For a given set of freestream conditions and a shock layer with
emitted power density E, q
s.rad
will be greater for the heat shield with
the larger
so
[3]. Empirical results [32] indicate that the
so
for a
sphere is directly proportional to its radius as a function of the
normal-shock density ratio. Kaattaris semiempirical method [33,34]
is used to estimate
so
for zero and nonzero o. This solution applies
relationships for shock-standoff and shock surface inclinations near
the sonic point for elliptical bodies at specic heat ratios from 1.0 to
1.4 for a range of high normal-shock density ratios. For the present
work, it is assumed that the effective radius (r
eff
) for a given heat
shield is equal to the radius of that particular sphere that maintains an
equivalent shock-standoff distance at the stagnation point. Ried et al.
[32] offers an empirical curve t that renders an acceptable
approximation:

so
,r
eff
=

[(,
2
,,
1
) 1|
2
(,
2
,,
1
)

(2,
2
,,
1
) 1
p 1

1
(5)
In this way, the
so
is incorporated in the low-order method by
means of r
eff
. As a result, this effective radius term is different from
the radius-of-curvature term applied for convection. Two q
s.rad
correlations are applied over a range of freestream velocities. For
V
o
<9000 m,s, the correlation applies the following form:
q
s.rad
=r
eff
g
1
(3.28084 10
4
V
o
)
g
2

,
o
,
sl

g
3
(6)
in which g
1
=372.6, g
2
=8.5, and g
3
=1.6 from [35] for
V
o
<7620 m,s, and g
1
=25.34, g
2
=12.5, and g
3
=1.78 from
[36] for velocities 7620 to 9000 m,s. For velocities above
9000 m,s, Tauber and Sutton [37] apply
q
s.rad
=4.736 10
4
r
H
eff
,
1.22
o
](V
o
) (7)
in which
H =1.072 10
6
V
1.88
o
,
0.325
o
and
](V
o
)
=

3.9320679310
12
V
4
o
1.6137000810
7
V
3
o
2.4359860110
3
V
2
o
16.1078691V
o
39.494.8753 9000V
o
11.500m,s
1.0023310010
12
V
4
o
4.8977467010
8
V
3
o
8.4298251710
4
V
2
o
6.25525796V
o
17.168.3333 11.500<V
o
16.000m,s
Thermochemical equilibrium is assumed. The curve-t equation for
](V
o
) has a high number of signicant gures in order to have less
than 2% error with the published tabulated values [37].
The low-order heat transfer methods have been validated against
Apollo 4 and FIRE II ight data. In comparison with the ight
investigation of the reentry environment (FIRE) II calorimeter data
[38] shown in Fig. 7, the low-order code overpredicts the peak total
heat ux (radiative and convective) by 7%. Table 2 indicates
predictions of maximum total heat transfer within 9.2% of reported
values. Further descriptions of the method and validation are
provided in [12]. All of these heat transfer correlations are limited to
the stagnation area, and while this is appropriate for this level of
analysis, this work does not account for acreage heating, which could
Fig. 6 Modied Newtonian validation with Apollo wind-tunnel
data [30].
Fig. 7 FIRE II q
s;tot
with ight data [38].
Table 2 Apollo 4 comparison of total heat transfer [12]
Parameters Apollo 4, [8] Results, % error
q
max.tot
, W,cm
2
480 470 (2.1%)
Q
max.tot
, kJ,cm
2
42.6 38.7 (9.2%)
42 JOHNSON, LEWIS, AND STARKEY
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

have a signicant impact on thermal protection system (TPS) design
and requirements.
F. Optimization Setup
UPTOPs exible framework allows for an external code to
provide updated aerothermodynamics for a heat-shield design
throughout the trajectory calculation. The aerothermodynamic
models have been integrated into the UPTOP setup to perform heat-
shield shape optimization.
1. Optimization Method
Previous work [12] applied the modied method of feasible
directions gradient-based method to optimize over the geometric
design space (without trajectory analysis) for a single objective.
There were numerous local optima; over 200 runs were required to
locate the global optima for four objective functions. For the present
work, a more robust and global search algorithm is required to
account for both the additional complexity of multiobjective opti-
mization and the broader design space with trajectory analysis.
UPTOPapplies a differential evolutionary scheme (DES) [21,39] for
this optimization. As an evolutionary algorithm, DESbases its search
for an optimumon natures evolutionary principles [40]. It begins by
randomly selecting an initial population of designs, and 100s of
iterations are required to settle on an optimal solution. Each heat-
shield design, known as an individual in a population of designs, is
evolved throughout each iteration with other individuals based on
mutation intensity and crossover parameters. Details on DES and its
parameters are provided in [39,41].
2. Objective Functions
Three objective functions are applied in this work: minimizing
stagnation-point heat load Q
s.tot
, minimizing peak stagnation-point
heat ux q
s.max
, and maximizing cross range
xrs
. These objective
functions have been selected on the basis of 1) relevance to mission
requirements, 2) connection to low-performing or restrictive
capabilities of existing blunt-body designs, and 3) the availability of
accurate physical models suitable for optimization purposes.
The peak heat ux of the trajectory determines which materials are
capable of surviving the selected entry conditions. Minimizing heat
load reduces the heat shields thickness and mass indirectly.
Minimizing them together requires the capability to y hundreds of
entry trajectories and to calculate the heat ux along all of those
trajectories. As a result, low-order computational models of the
aerothermodynamics are implemented to balance the need for
delity with the desire to have practical computational times. Heat
transfer is tracked at the stagnation point. There is more condence in
well-validated stagnation-point correlations than in low-order
estimates of the maximum heat ux, especially when applied to a
wide range of geometries in extreme hypersonic conditions. They
also cost less computational time. In many cases, the stagnation-
point heat ux is not representative of the heat ux experienced over
the entire heat shield, which is commonly lower, but its value is
usually on the order of the maximum heat ux.
Crossrange capability can enable additional landing options.
Existing designs such as the ~25

SS and the 70

SC have low
crossrange performance due to trajectory design and lowL,D. For a
vehicle ying a direct entry trajectory, lunar return with a hypersonic
L,D=0.30, the maximum cross range is limited to ~200 km
assuming a 5g constraint [1]. To increase crossrange capability, both
skip trajectories, which have been shown to increase cross range
signicantly [1], and higher L,D designs are considered feasible.
3. Multiobjective Optimization
In single-objective optimization, the one optimal or nondominated
solution is better than all other solutions. In multiobjective optimi-
zation, two or more objective functions are optimized simultaneously
to produce a set of optimal or nondominated solutions known as the
Pareto frontier. When two objective functions are optimized
simultaneously, a Pareto frontier has the form of a curve that
represents the optimal tradeoff between the two objectives. Shown in
Fig. 8, the results of minimizing Q
s.tot
and q
s.max
are provided
together. This Pareto frontier is composed of those solutions in the
feasible population that are not dominated with respect to both
objective functions; each point on the frontier represents an optimal
solution. In general, the Pareto frontier is a set of nondominated
solutions, in which one solution is better than another with respect to
at least one objective, but not all objectives [40].
Multiobjective optimization is used to optimize conicting
objectives. Since an increase in cross range produces a larger heat
load, maximizing cross range and minimizing heat load are
conicting objectives. Nonoptimal results may produce higher heat
loads than necessary for a desired cross range. Minimizing heat load
and minimizing peak heat ux are also conicting objectives. For this
work, optimal solutions are provided in the form of Pareto frontiers
between two objectives to highlight performance tradeoffs and
provide comparisons among axial proles. For entry velocities of 11
and 15 km,s, optimization is performed using two objective
function sets: 1) maximizing
xrs
and minimizing Q
s.tot
and
2) minimizing q
s.max
and Q
s.tot
.
The authors conducted a parametric study of the effects of
population size, crossover probability, and mutation intensity on the
Pareto frontier for maximizing
xrs
and minimizing Q
s.tot
in [42]. It
was found that the most comprehensive Pareto frontier for this design
space is produced with a crossover probability of 0.8 and a mutation
intensity that is randomly varied for each generation between 0 and 1.
An initial population of 390 individuals is applied since diminishing
returns were observed with increasing population. Twelve AMD
2.2 GHz Opteron 248 processors were used in this analysis, and the
time required per run was approximately 10 h.
4. Design Variables
For each axial prole, the design variables along with their side
constraints are listed in Table 3. For 0
s
, the lower limit of 5

provides
a blunt body that has a large but nite radius of curvature. For both 0
s
and 0
c
, the upper limit 89

removes numerical issues present if the


upper limit is set to 90

. For the blunted cone, the upper limit of r


n
,J
is chosen to provide overlap and continuity between the blunted cone
and SS design spaces. For the PL, Newtonian impact theory may
have an accuracy issue given the quick slope changes shown in
Fig. 4b for l =0.1 or smaller; a lower limit value of 0.2 for l has
been chosen arbitrarily. The upper limit of l excludes the PL proles
with sharper nose regions. The arbitrary upper limit for A has been
chosen based on its similarity to the SS near this value. In generating
the base cross section, the lower limit of n
2
produces slightly
rounded-edge polygons. A maximum eccentricity of 0.968 was
chosen to limit the axes ratio ],k to four. Additional reasons for the
chosen side constraints that are geometry-related have been detailed
in [12]. A mesh convergence study was completed in [19] to reduce
Fig. 8 Multiobjective function population with Pareto frontier, SS,
V
EI
15 km=s.
JOHNSON, LEWIS, AND STARKEY 43
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

computational time; for optimization, the geometrys mesh is
]
max
=45, k
max
=101.
Awide range of ;
E
including the Apollo [8] missions ;
E
~6.5

is allowed. To modify the vehicles ight path through rotation of the


lift vector, a bank angle prole with ve control points is available.
The optimizer can modify the ve bank angles
l.0
through
l.4
, as
well as the three intermediate times at which the bank angles are
initiated. Connecting the control points (i.
l
) produces the
l
prole. Alimit of 0 to 180

lowers the size of the design space; angles


181 through 359

are not necessary since longitude and latitude


constraints are not considered.
5. Design Constraints
Boundaries for the feasible design space are provided in Table 4.
These constraints account for trajectory design limits, theory
limitations, and static longitudinal, directional, and roll stabilities.
Aerodynamic and geometric constraints for this optimization are
detailed in [12]. The requirement for roll static stability changes sign
when C
L.V
changes sign, as explainedin [17], thus requiring a change
in the constraint. Note that, for heat shields with c <15

, an aftbody
with an angle greater than 15

is assumed to allow heat shields with


low 0
s
(for example, 0
s
=10

) to be feasible for [o[ 15

. Since
guidance laws are not considered in the trajectory analysis, optimal
trajectories with similar entry interface characteristics but with less
complex
l
proles tend to generate longer duration trajectories. A
compromise was chosen to be 2 h, i
]
7200 s, which is greater than
twice the upper limit to the estimated i
]
for the Orion CEV.
An arbitrary maximumaltitude for skip trajectories has been set to
3000 km. Anal altitude h
i.]
no greater than 75 km has been chosen
arbitrarily to ensure that the vehicles trajectory ends within the
atmosphere. This allows for higher L,D vehicle designs to be
feasible since their optimal trajectories may result in deceleration at
higher altitudes in the atmosphere. For V
EI
=11 km,s, a peak g load
of 6g was chosen since it is the maximum allowable acceleration
level for a deconditioned astronaut in a reclined position [43]. It is
also lower than the 7g that Apollo 10 experienced [44]. Preliminary
analysis indicated that, for a 15 km,s entry, this optimization setup
would be overconstrained with a 6g upper limit. This limit was
increased to 12g based on previous work that indicates a pilot can
sustain 12g for up to 60 s and still continue to perform the assigned
tasks [45]. Although this is not expected to conform with future
standards for manned Mars return, the results provide a sense of the
heating environment when entering at high hyperbolic velocities.
Results for V
EI
=15 km,s can be applied at least toward most
unmanned missions.
III. Results
Optimization has been performed using three types of heat-shield
axial proles: the SS, the SC, and the PL. A heat shield with a
parallelogram-form base cross section (m
1
=4) provides the largest
hypersonic L,D [12]. Since greater L,D increases crossrange
capability, this analysis focuses optimization on base cross sections
of parallelogram and elliptical forms, and blendings of the two. For
initial entry velocities of 11 and 15 km,s, Pareto frontiers are
provided for two multiobjective function sets: 1) minimizing heat
load Q
s.tot
and maximizing cross range
xrs
, and 2) minimizing heat
load Q
s.tot
and heat ux q
s.max
. Design variable distributions are
provided for selected Pareto frontiers, and specic designs listed in
Table 5 are discussed. Axial proles and base cross sections for the
specic designs selected from the Pareto frontiers are shown in
Figs. 9 and 10, respectively. Note that the base cross section for
design C is circular.
A. Optimal Congurations for V
EI
11 Kilometers per Second
1. Minimizing Q
s.tot
and Maximizing
xrs
The lowest possible heat load is expected to increase with cross
range
xrs
provided that down range is relatively constant or
increasing. A Pareto frontier is given for each type of axial prole in
Fig. 11a for cross ranges up to 1500 km and heat loads from 11 to
33 kJ,cm
2
. For optimal designs with
xrs
of 500, 1000, and
1500 km, the values of Q
s.tot
are 14.7, 22.6, and 29.4 kJ,cm
2
,
respectively. The optimizer produced similar Pareto frontiers for all
three axial proles. The PL Pareto frontier is expected to be the least
accurate since an articial effective nose radius is applied. All three
frontiers closely match for
xrs
> 750 km; close inspection indicates
that, for this region, the SC and PL proles are disguised SSs.
Figure 12 shows the design variable distribution for the SS Pareto
frontier shown in Fig. 11a. The transformation variable n
2
is nearly
constant at 2.0, indicating an elliptical cross section rather than a
parallelogram form is optimal for this set of
xrs
. These results
indicate that for lowL,Ddesigns, an elliptical cross section is better
due to its larger drag area (C
D
S =D,q
o
), resulting in a lower BCfor
a given m
EV
. For
xrs
700 km, nondominated or optimal designs
have highly oblate e =0.968, which is the lowest allowed value.
With an increased heat-shield radius of curvature, this design allows
for less convective heat transfer. Higher e also increases the drag area,
thus decreasing BC. The nondominated heat-shield geometry is held
constant by the optimizer until
xrs
=700 km, at which point there is
a jump in 0
s
from 6.8 to 18

. In general, [o[ is increasing throughout


this portion of the Pareto frontier, indicating higher [L,D[ is required
to produce additional cross range. At
xrs
=700 km, [o[ =16

, and
the geometric constraint [o[ [c 1

[ is active. The parametric


analysis in [17] (gure 19) indicates that, for specic combinations of
xed o and e (i.e., o =20

, e =0), an increase in 0
s
decreases the
magnitude [L,D[. Design A is in a similar part of the design space.
Since the [o[ is xed at 16

unless 0
s
> 15

and the optimizer is


attempting to increase [L,D[, it determines that a decrease in e along
with an increase 0
s
that allows for [o[ >16

would result in a
sufcient and incremental increase in [L,D[. The increase in 0
s
to
18

limits the reduction in radius of curvature, thus concurrently


minimizing Q
s.tot
. These necessary adjustments to 0
s
, e, and o
produce the sudden rise in Q
s.tot
on the Pareto frontier at

xrs
=700 km. Consequently, drag area is traded off with [L,D[ as
Table 3 Design variables with side constraints
Axial prole Prole-specic design variables Common design variables
SS 5.0

0
s
89.0

30

o 30

0.968 e 0.968
15.5

;
E
0.05


l.0
180

SC 55.0

0
c
89.0

0.15 r
n
,J 2.00
1.30 n
2
2.00
i
0
5 s i
1
7190 s
0


l.1
180


l.2
180

PL 0.900 A 10.000
0.200 l 0.650
i
1
5 s i
2
7190 s
i
2
5 s i
3
7190 s
0


l.3
180


l.4
180

Table 4 Trajectory and aerodynamic constraints


Optimization constraints
Trajectory Aerodynamic/geometric
i
]
7200 s M
o.]
5
h
i
3000 km C
m.cg.o
0.001
h
i.]
75 km C
n.cg.[
0.001
n
max
6g, V
EI
=11 km,s sign(C
L.V
)C
l.cg.[
0.01
n
max
12g, V
EI
=15 km,s [o[ [c 1

[
44 JOHNSON, LEWIS, AND STARKEY
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

xrs
increases. Similar behavior is noted for the other axial proles
and not shown here.
For
xrs
250 km, optimal solutions use direct entry trajectories.
To increase
xrs
, the banked lift vector must turn the vehicle further.
This is achieved by steepening ;
E
to travel deeper in the atmosphere
for a given V
o
, increasing q
o
and thus, the lift dedicated for turning.
As shown in Fig. 12b,
l.0
is adjusted, pointing the lift vector slightly
upward, to keep the vehicle slightly higher in the atmosphere for
minimizing Q
s.tot
. The change in behavior of ;
E
and
l
at
xrs
~
250 km indicates a switch from direct entry trajectories to skip
trajectories. This crossrange limit for direct entry trajectories is
consistent with the work of Putnam et al. [1]. For a skip trajectory, a
steeper ;
E
is used to dissipate sufcient energy to avoid skipping out
of the atmosphere. Minimizing heat load restricts low [;
E
[ and
trajectory duration, while maximizing
xrs
and the deceleration limit
restricts high [;
E
[. For skip trajectories, larger
xrs
requires
l
to
approach 90

. Since there is a smaller proportion of lift dedicated to


producing the skip as
l
approaches 90

, the vehicle requires a


steeper ;
E
, as indicated by Fig. 12b.
As seen in Fig. 11a, designs A, B, and C are optimal for

xrs
=520, 1010, and 1500 km, respectively. Design A represents
the SS geometry applied for
xrs
700 km. Its heat-shield axial
prole is illustrated in Fig. 9, and its base cross section is illustrated in
Fig. 10a. The results shown in Table 5 indicate that, by using
design B, a 54% increase in Q
s.tot
is required to achieve double
design As cross range. Not only does q
s.max
affect Q
s.tot
but so does
the change in individual contributions fromconvectionand radiation.
By halving the radius of curvature of design B from 6.3 to 3.15 m
using 0
c
=60.4

, e =0.682, and r
n
,J =0.615, the q
s.max
is
approximately unaffected with q
s.conv
=140 W,cm
2
and q
s.rad
=
160 W,cm
2
. However, Q
s.conv
increases by 40% while Q
s.rad
decreases by 20%. As a result, Q
s.tot
increases by 20%. For a design
similar to the Vikings SC, 0
c
=70

, e =0.682, r
n
,J =0.25, and
Table 5 Specic optimal designs from two multiobjective function optimizations, m
1
4
a
Minimizing Q
s.tot
and maximizing
xrs
Minimizing Q
s.tot
and q
s.max
V
E
=11 km,s (Fig. 11a) V
E
=15 km,s
(Fig. 14a)
V
E
=11 km,s
(Fig. 11b)
V
E
=15 km,s (Fig. 14b)
Design variables A B C E F D G H
SS SC PL SS SC SC SS SC
Axial prole 0
s
=6.80

0
c
=60.4

r
n
,J =1.26
l =0.34
A =5.25
0
s
=8.1

0
c
=84.3

r
n
,J =1.29
0
c
=84.4

r
n
,J =2.00
0
s
=10.2

0
c
=84.3

r
n
,J =1.30
Base cross section n
2
=1.99
e =0.968
n
2
=2.00
e =0.682
n
2
=1.96
e =0.003
n
2
=1.98
e =0.968
n
2
=2.00
e =0.968
n
2
=2.00
e =0.968
n
2
=2.00
e =0.968
n
2
=2.00
e =0.968
Trajectory
;
E
(i
0
.
l.0
)
. . .
(i
]
.
l.]
)
o =13.7

6.01

(0 s, 59.0

)
(1440, 76.0

)
o =23.8

6.14

(0 s, 75.9

)
(1530, 55.1

)
o =30.0

6.29

(0 s, 97.9

)
(1540, 135.7

)
o =14.8

6.60

(0 s, 43.7

)
(248.4, 161.5

)
(1450, 137.4

)
o =15.9

6.44

(0 s, 143.5

)
(190, 40.7

)
(870, 84.5

)
(1214, 85.8

)
o =15.8

5.37

(0 s, 144.0

)
(267, 66.1

)
o =9.00

6.41

(0 s, 1.93

)
(219, 6.14

)
o =15.9

6.18

(0 s, 177.0

)
(197, 84.0

)
Parameters Skip trajectories Direct trajectories
n
max
, g 6.0 5.9 6.0 12.0 11.8 5.9 11.9 11.7
Q
s.tot
, kJ,cm
2
(Q
s.conv
. Q
s.roJ
)
14.7
(7.1, 7.6)
22.6
(14.9, 7.7)
29.4
(19.8, 9.6)
82.4
(13.6, 68.8)
65.2
(18.5, 46.7)
12.4
(7.4, 5.0)
76.7
(12.9, 63.8)
63.6
(16.3, 47.3)
q
s.max
, W,cm
2
(q
s.conv
. q
s.rad
)
250
(50, 200)
300
(100, 200)
380
(130, 250)
1930
(150, 1780)
1400
(200, 1200)
160
(60, 100)
1500
(140, 1360)
1100
(180, 920)

xrs
, km 520 1010 1500 990 1000 120 10 100
C
D
1.62 1.32 1.17 1.57 1.56 1.57 1.60 1.56
[L,D[ 0.22 0.36 0.50
b
0.22
b
0.24 0.24 0.12
b
0.24
C
m.cg.o
. ,rad 0.18 0.15 0.10 0.20 0.19 0.19 0.27 0.20
BC, kg,m
2
130 220 350 130 100 110 120 100
h
l.HS
, m 3.4 5.0 5.0 3.5 4.0 3.8 3.6 4.0
S, m
2
36.9 27.1 19.0 38.1 49.0 45.8 41.0 49.0
j
:.HS
38.8% 60.6% 58.5% 43.0% 41.2% 40.5% 49.1% 41.3%
i
]
, s 1440 1530 1540 1450 1214 267 219 197
a
m
EV
=7800 kg
b
The sign of L,D is negative for the listed o
Fig. 9 Axial prole designs from Table 5. Fig. 10 Specic base cross sections from Table 5.
JOHNSON, LEWIS, AND STARKEY 45
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

the q
s.max
increase is approximately 27% (to 380 W,cm
2
); at the
same time, the Q
s.tot
increase is 68% over that for design Bs heat-
shield geometry. The results in Fig. 12a indicate that the optimal heat-
shield geometry for skip trajectories at V
EI
=11 km,s is the SS with
nonzero eccentricity. Design C doubles the Q
s.tot
to triple the
xrs
over that of design A. Table 5 shows
xrs
, q
s.max
, and Q
s.tot
increasing
with BC as expected.
It is possible for Q
s.conv
> Q
s.rad
, even though the peak q
s.conv
is
less than the peak q
s.rad
. This situation occurs for designs B and C
when minimizing Q
s.tot
. Signicant convective heat transfer occurs
throughout the entire hypersonic trajectory, while radiative heat
transfer contributes signicantly only at the highest velocities for
V
o
7600 m,s. For V
o
< 7600 m,s, the q
s.rad
is less than
5 W,cm
2
.
2. Minimizing Q
s.tot
and q
s.max
Pareto frontiers are provided in Fig. 11b for q
s.max
ranging from
130 to 210 W,cm
2
, producing heat loads ranging from 11.8 to
19.3 kJ,cm
2
. UPTOP could not generate a Pareto frontier for the PL
form that followed behavior similar in form to the SSs and the SCs,
although the effective nose radius of the geometry was not
signicantly different. As a result, the output data of the PL form
were found to be inconclusive. This may be due to the lack of a
mathematical relationship that is sufciently accurate at determining
the effective nose radius for a given design with a PL axial prole.
The minimum Q
s.tot
decreases with increasing q
s.max
, as expected.
The trajectory designvariable distribution for the SCgiven in Fig. 13,
together with Fig. 11b, demonstrates that a shallower ;
E
yields a
smaller q
s.max
and larger Q
s.tot
. The optimal geometric congurations
are similar to those with low Q
s.tot
in Fig. 11a that use a direct entry
approach, as suggested by the shallower ;
E
and higher
l
, than those
reported in Fig. 12b.
Both the SS and SC geometries are relatively constant throughout
the Pareto frontiers. The SS geometry is 0
s
=6.83

, n
2
=2.00,
e =0.968, and h
l.HS
=3.2 m, and the blunted cone geometry is
listed in Table 5 as design D. The two Pareto frontiers are within
35 W,cm
2
of each other and within the calculated correlation
uncertainty. For altitudes from 66 to 72 km, stagnation-point
convective and radiative heat transfer calculations were judged
against FIRE II ight data from [46] and high-order computational
results from [37]. The uncertainty of peak q
s.conv
is 20 W,cm
2
,
Fig. 11 Pareto frontiers for Earth entry, V
EI
11 km=s.
Fig. 12 Design variable distribution for SS designs from Fig. 11a.
Fig. 13 Trajectory design variable distribution for SC designs from
Fig. 11b.
46 JOHNSON, LEWIS, AND STARKEY
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

while the uncertainty of the peak q
s.rad
is 20 W,cm
2
for
V
EI
=11 km,s, and 60 W,cm
2
for V
E
=15 km,s. Differences in
q
s
within these uncertainties were not considered discriminators for
the selection of specic optimal designs.
Both congurations, the SS detailed above and SC design D, are
equally optimal. Comparing the two geometries at the freestream
conditions of q
s.max
for design D, at h
i
=73.1 km, and V
o
=
10.5 km,s, the SS would produce a different combination of heat
uxes q
s.conv
=40 W,cm
2
and q
s.rad
=140 W,cm
2
(compared
with 60 and 100, respectively, for design D, as shown in Table 5) but
result in nearly the same Q
s.tot
, although the radius of curvature is
14.3 m for the SS and 7.5 m for SC. In this problem setup, a highly
eccentric base has an increase in drag area, which reduces BC, and
thus indirectly reduces q
s.max
and Q
s.tot
since larger drag area
provides deceleration at higher altitudes for a given m
EV
. However,
the skip entry of design B has a 132% increase in time for which q
s
exceeds 5 W,cm
2
over that for the direct entry of design D. This fact,
together with a 0.77

steeper ;
E
, makes radius of curvature assume
greater importance for skip entry.
B. Optimal Congurations for V
EI
15 Kilometers per Second
1. Minimizing Q
s.tot
and Maximizing
xrs
Entry at V
EI
=15 km,s represents a kinetic energy level 85%
greater than entry at V
EI
=11 km,s, resulting in a Q
s.tot
at least three
times greater. Radiative heat transfer produces a majority of Q
s.tot
and
can be minimized by decreasing the radius of curvature to reduce

so
, which is smaller for a SC than a SS for a given L,D[
max
design.
The different thermal environment is expected to result in different
optimal congurations at 15 km,s than were found at 11 km,s.
Pareto frontiers are shown in Fig. 14a for
xrs
2200 kmwith Q
s.tot
in the range 60160 kJ,cm
2
compared with 1133 kJ,cm
2
for
V
EI
=11 km,s. The Pareto frontier of the PL is composed of
effective SS forms. For optimal designs with
xrs
of 500, 1000, and
1500 km, the values of Q
s.tot
are 64.5, 65.2, and 98.3 kJ,cm
2
,
respectively. All three axial proles followsimilar behavior: the Q
s.tot
increases ~12% from
xrs
=0 km to 1100 km.
The signicant difference in Q
s.tot
between the SS and SC Pareto
frontiers is caused primarily by differences in drag area. SS design E
has a 27.8% lower drag area than blunted cone design F because the
SS can maintain requirements with a smaller vehicle. Evidence of
this is shown in Table 5 where design E has a 12.5%lower h
l.HS
than
design F. Design E experiences q
s.max
at an altitude of 66.1 km, 2 km
deeper than design F, and sees a 26%larger Q
s.tot
than design F. Both
designs follow similar skip trajectories and experience q
s.max
at
V
o
=13.5 km,s. The trajectory for design F is illustrated in Fig. 15.
At 66.1 km, air density is 40% greater than at 68.1 km. Since the
normal-shock density ratio is relatively constant between the two
altitudes with the same V
o
, the
so
does not change signicantly.
This results in a 40% increase in power density E, and thus a 31%
increase in q
s.rad
. This demonstrates the high sensitivity of heat
transfer to density, and therefore altitude for travel at hyperbolic
speeds through the atmosphere.
A proper balance of radiative and convective heat loads is
important for minimizing total heat load, but the optimal SS and SC
geometries produce nearly the same minimum Q
s.tot
. To isolate the
effects of trajectory design and geometry, the heat uxes for both
designs are compared at each others q
s.max
freestream conditions.
Table 6 provides a comparison of q
s.max
and Q
s.tot
generated at the
q
s.max
freestream conditions of designs E and F, respectively. At the
q
s.max
conditions of design E (h
i
=66.1 km, V
o
=13.5 km,s),
Fig. 14 Pareto frontiers for Earth entry, V
E
15 km=s.
Fig. 15 Heat-shield skip trajectory of design F from Table 5.
Table 6 Max q
s
and Q
tot
comparison at q
s;max
freestream
conditions of designs E and F
Design E (SS) Design F (SC) Viking (SC) with
e =0.968
Freestream
Conditions
at q
s.max
q
s.max
, W,cm
2
(q
s.conv
. q
s.rad
)
q
s.max
, W,cm
2
(q
s.conv
. q
s.rad
)
q
s.max
, W,cm
2
(q
s.conv
. q
s.rad
)
Q
s.tot.
, kJ,cm
2
(Q
s.conv
. Q
s.rad
)
Q
s.tot.
, kJ,cm
2
(Q
s.conv
. Q
s.rad
)
Q
s.tot.
, kJ,cm
2
(Q
s.conv
. Q
s.rad
)
E: h
i
=66.1 km,
V
o
=13.5 km,s
1930
(150, 1780)
82.4
(13.6, 68.8)
1790
(230, 1560)

F: h
i
=68.1 km,
V
o
=13.5 km,s
1580
(120, 1460)
67.8
For h
l.HS
=4 m
1400
(200, 1200)
65.2
(18.5, 46.7)
1570
(470, 1100)
86
JOHNSON, LEWIS, AND STARKEY 47
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

design F would experience a 7%lower q
s.max
than design E yet result
in roughly the same Q
s.tot
as for design E. If design Ewere required to
have the same h
l.HS
as design F, h
l.HS
=4 m, then design E would
have drag area similar to design Fs (since they have nearly the same
C
D
), resulting in an equivalent BC. They would then y nearly the
same trajectories. At the q
s.max
conditions of design F(h
i
=68.1 km,
V
o
=13.5 km,s) and setting design Es h
l.HS
equal to design Fs,
design E would generate a 13% larger q
s.max
than design F, primarily
generated by the 22%increase in q
s.rad
. The resulting Q
s.tot
would be
approximately 4% greater than that of design F. At the q
s.max
conditions of design F, a geometry similar to Viking, except
e =0.968, would produce a 12% larger q
s.max
and a 32% increase
in Q
s.tot
over design F. This indicates that Q
s.tot
is sensitive to different
combinations of radius of curvature and
so
.
The design variable distributions for the blunted cone Pareto
frontier are illustrated in Fig. 16; the trends in these distributions are
consistent with those discussed for V
EI
=11 km,s (Fig. 12). Only
skip trajectories were captured for the SCas its Pareto frontier begins
at
xrs
=200 km. The SS switches from direct entry trajectories to
skip trajectories at
xrs
=170 km. As cross range is increased, the
sudden changes in geometry are produced by the optimizer to
incrementally increase L,D.
2. Minimizing Q
s.tot
and q
s.max
Pareto frontiers are shown in Fig. 14b for the SS and SC. The
results for the PL form were inconclusive for the same reasons as for
V
EI
=11 km,s (III.A.2). For the two Pareto frontiers, the geometries
are relatively constant, listed as designs Gand H in Table 5. Both the
SS and blunted cone geometries are very similar to designs E and F,
respectively. Similarly, the SS has a smaller h
l.HS
than the blunted
cone, thus having a smaller dragarea and resulting inhigher Q
s.tot
and
q
s.max
.
To isolate the effects of trajectory design and geometry, the heat
uxes for both designs are compared at each others q
s.max
freestream
conditions. Heat ux values are listed in Table 7. Design G
experiences q
s.max
at h
i
=68.7 km, V
o
=13.7 km,s, while
design H experiences q
s.max
at h
i
=70.4 km, V
o
=13.6 km,s. At
the q
s.max
freestream conditions of design G, design H would
experience a 26% increase in q
s.max
from its value at its own q
s.max
conditions. At design Gs q
s.max
conditions, this blunted cone
design Hexperiences an 8%lower q
s.max
than design G. At the q
s.max
conditions of design H, design G would experience a 21% decrease
in q
s.max
from its value at its own q
s.max
conditions; at design Hs
q
s.max
conditions, this SS design produces an 8% higher q
s.max
than
design H. The SC generates a lower q
s.max
at both the q
s.max
conditions of designs G and H in which q
s.max
is experienced.
Differences in Q
s.tot
between both cases are negligible. The q
s.rad
is
more sensitive than Q
s.tot
for these optimal designs. Since these
geometries are similar to E and F, the importance of balancing
radiative and convective heat transfer also holds.
The trajectory design variable distributions are given in Fig. 17 for
the blunted cone. For this objective function set, the aims of both o
and
l
are the same since
xrs
is not being optimized. For the blunted
cone case, o is constant, and the optimizer varies
l
to control how
much lift is applied to counteract gravity. For the SS case (not
shown),
l.0
is relatively constant at 0

, and o decreases with


increasing q
s.max
in order to lower the ight duration, and thus
minimize Q
s.tot
.
IV. Conclusions
Optimization has produced optimal heat-shield congurations for
Earth entry at V
EI
=11 and 15 km,s using two objective function
sets: 1) maximizing
xrs
and minimizing Q
s.tot
and 2) minimizing
Q
s.tot
and q
s.max
. The assumptions in this work are appropriate for
assessment of the trade space at the conceptual design level for heat-
shield shape optimization. For V
EI
=11 km,s with a 6g limit, the SS
is the optimal axial prole for maximizing
xrs
and minimizing Q
s.tot
.
Direct entry trajectories are best for
xrs
250 km, and skip
trajectories are preferred for larger
xrs
. For optimal designs with
xrs
of 500, 1000, and 1500 km, the values of Q
s.tot
are 14.7, 22.6, and
29.4 kJ,cm
2
, respectively. For
xrs
>750 km, the SC and PL
solutions are disguised SSs. The optimal designs for minimizing
Fig. 16 Design variable distribution for SC designs from Fig. 14a.
Fig. 17 Trajectory design variable distribution for SC designs from
Fig. 14b.
Table 7 Max q
s
comparison at q
s;max
freestream conditions
of designs G and H
Design G (SS) Design H (SC)
Freestream conditions at q
s.max
q
s.max
, W,cm
2
(q
s.conv
. q
s.rad
)
q
s.max
, W,cm
2
(q
s.conv
. q
s.rad
)
G: h
i
=68.7 km, V
o
=13.7 km,s 1500
(140, 1360)
1390
(210, 1180)
H: h
i
=70.4 km, V
o
=13.6 km,s 1190
(130, 1060)
1100
(180, 920)
48 JOHNSON, LEWIS, AND STARKEY
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

Q
s.tot
and q
s.max
have direct entry trajectories, and the resulting SS
and blunted cone geometries are equally optimal.
For V
EI
=15 km,s with a 12g limit, neither the optimal blunted
cone nor SS is signicantly better. Radiative heat transfer dominates
convection in both heat ux and heat load. For optimal designs with

xrs
of 500, 1000, and 1500km, thevalues of Q
s.tot
are 64.5, 65.2, and
98.3 kJ,cm
2
, respectively. For maximizing
xrs
and minimizing
Q
s.tot
, in general, the Q
s.tot
increases ~12% from
xrs
=0 km to
1100 km. For
xrs
>1200 km, although the optimal SS and blunted
cone designs produce nearly the same optimum Q
s.tot
with two
different sets of curvature and shock layer thicknesses, a proper
balance of convective and radiative heat transfer is necessary to
minimize Q
s.tot
.
As expected, it was observed that radiative heat transfer is more
sensitive to air density for entry at hyperbolic speeds than for entry at
lower velocities, including typical lunar return velocities of
~10.5 km,s. The blunted cone designs have higher drag area since
they are less volumetrically efcient, and thus decelerate 2 kmhigher
in altitude with 40% less air density. This reduces q
s.max
and
ultimately Q
s.tot
by 21%. The SScan satisfy the mission requirements
with a smaller-sized vehicle, which leads to the lower drag area and
higher Q
s.tot
. Due to the high sensitivity of Q
s.max
to drag area, vehicle
sizing and volumetric efciency can have important roles in
determining the required capability of the TPS.
For entries at 11 and 15 km,s, a highly oblate eccentricity e =
0.968 maximizes drag area, allowing deceleration at higher
altitudes, thus lowering both heat ux and heat load. As more
xrs
is
required, drag area is traded off with the need for larger L,D by
decreasing e. This behavior is consistent with previous work. An
elliptical cross section rather than a parallelogramformis optimal for
L,D 0.50, corresponding to
xrs
1500 and
xrs
2200 kmfor
V
EI
=11 and 15 km,s, respectively. The parallelogramcross section
could be applied to increase L,D beyond the capability of the
elliptical cross section, which is expected to increase
xrs
and
decrease peak g limits. The TPS would be required to handle the
expected higher Q
s.tot
.
This work has identied and demonstrated an approach for
optimizing multiple objectives in a broad design space using a
differential evolutionary algorithm. With the proper setup, it can be
used to automate the process of locating feasible solutions. This
approach can be used in a variety of disciplines to assess a broad
design space and focus it to achieve the desired objectives.
Acknowledgments
This research was supported by the Space Vehicle Technology
Institute (SVTI), one of the NASAConstellation University Institute
Projects (CUIP), under grant NCC3-989, with joint sponsorship
from the Department of Defense. Appreciation is expressed to
Darryll Pines, director of the SVTI at the University of Maryland,
Claudia Meyer, programmanager of CUIP, and Jeffry Rybak, deputy
program manager of CUIP of the NASA John H. Glenn Research
Center at Lewis Field, the support of whom is greatly appreciated.
Gratitude is expressed to Falcon Rankins for software training and
support of the University of Maryland Parallel Trajectory
Optimization Program code used for planetary entry trajectory
optimization.
References
[1] Putnam, Z. R., Braun, R. D., Rohrschneider, R. R., and Dec, J. A.,
Entry System Options for Human Return from the Moon and Mars,
Journal of Spacecraft and Rockets, Vol. 44, No. 1, 2007, pp. 194202.
doi:10.2514/1.20351
[2] Bertin, J. J., Hypersonic Aerothermodynamics, AIAA Education
Series, AIAA, Washington, DC, 1994, Chap. 3.
[3] Anderson, J. D., Hypersonic and High Temperature Gas Dynamics,
AIAA, Reston, VA, 2000, Chap. 6.
[4] Grifn, M. D., and French, J. R., Space Vehicle Design, 2nd ed., AIAA
Education Series, AIAA, Reston, VA, 2004, Chap. 6.
[5] Crowder, R. S., and Moote, J. D., Apollo Entry Aerodynamics,
Journal of Spacecraft and Rockets, Vol. 6, No. 3, 1969, pp. 302307.
doi:10.2514/3.29589
[6] Korkan, K. D., and Hanley, G. M., Apollo Command Module
Aerothermodynamic Characteristics at Hyperbolic Earth Entry
Velocities, Journal of Spacecraft and Rockets, Vol. 3, No. 8, 1966,
pp. 12741281.
doi:10.2514/3.28639
[7] Lee, D. B., and Goodrich, W. D., The Aerothermodynamic
Environment of the Apollo Command Module During Superorbital
Entry, NASATN D-6792, April 1972.
[8] Pavlosky, J., and L. Leger, Apollo Experience ReportThermal
Protection Subsystem, NASATN D-7564, Jan. 1974.
[9] Kirk, D. B., Intrieri, P. F., and Seiff, A., Aerodynamic Behavior of the
Viking Entry Vehicle: Ground Test and Flight Results, Journal of
Spacecraft and Rockets, Vol. 15, No. 4, 1978, pp. 208212.
doi:10.2514/3.57307
[10] Chrusciel, G. T., and Hull, L. D., Theoretical Method for Calculating
Aerodynamic Characteristics of Spherically-Blunted Cones, AIAA
Paper 1968-0674, June 1968.
[11] Sammonds, R. I., and Kruse, R. L., Viking Entry Vehicle
Aerodynamics at M=2 in Air and Some Preliminary Test Data for
Flight in CO
2
at M=12, NASATN D-7974, June 1975.
[12] Johnson, J. E., Starkey, R. P., and Lewis, M. J., Aerothermodynamic
Optimization of Reentry Heat Shield Shapes for a Crew Exploration
Vehicle, Journal of Spacecraft and Rockets, Vol. 44, No. 4, July
Aug. 2007, pp. 849859.
doi:10.2514/1.27219
[13] Whitmore, S. A., Banks, D. W., Anderson, B. M., and Jolley, P. R.,
Direct-Entry, Aerobraking, and Lifting Aerocapture for Human-Rated
Lunar Return Vehicles, AIAA Paper 2006-1033, Jan. 2006.
[14] Wooster, P. D., Braun, R. D., Ahn, J., and Putnam, Z. R., Trajectory
Options for Human Mars Missions, AIAA Paper 2006-6308,
Aug. 2006.
[15] Constellation Program: The Orion Crew Exploration Vehicle, NASA
Johnson Space Center, NASAfacts FS-2006-08-022-JSC, 2006.
[16] Gielis, J., A Generic Geometric Transformation that Unies a Wide
Range of Natural and Abstract Shapes, American Journal of botany,
Vol. 90, No. 3, 2003, pp. 333338.
doi:10.3732/ajb.90.3.333
[17] Johnson, J. E., Starkey, R. P., and Lewis, M. J., Aerodynamic Stability
of Reentry Heat Shield Shapes for a CrewExploration Vehicle, Journal
of Spacecraft and Rockets, Vol. 43, No. 4, JulyAug. 2006,
pp. 721730.
doi:10.2514/1.20044
[18] Larson, W. J., and Pranke, L. K. (eds.), Human Spaceight Mission
Analysis and Design, Space Technology Series, McGrawHill,
New York, 1999, Chaps. 10, 12.
[19] Johnson, J. E., Starkey, R. P., and Lewis, M. J., Multi-Objective
Optimization of Entry Vehicle Heat Shields with Reentry Trajectory
Analysis, 58th International Astronautical Congress, Hyderabad,
India, International Astronautical Federation IAC Paper 07-C2.7.09,
Paris, Sept. 2007.
[20] Darling, D., ASSET (Aerothermodynamic Elastic Structural Systems
Environmental Tests), The Complete Book of Spaceight: FromApollo
1 to Zero Gravity, Wiley, Hoboken, NJ, 2003.
[21] Starkey, R. P., Rankins, F., and Pines, D. J., Effects of Hypersonic
Cruise Trajectory Optimization Coupled with Airbreathing Vehicle
Design, AIAA Paper 2006-1036, Jan. 2006.
[22] Rankins, F., The University of Maryland Parallel Trajectory
Optimization Program Users Manual, Univ. of Maryland, College
Park, MD, 2007.
[23] Stevens, B. L., and Lewis, F. L., Aircraft Control and Simulation, Wiley,
New York, 1992.
[24] Department of Defense World Geodetic System 1984, Its Denition
and Relationships With Local Geodetic Systems, National
Imagery and Mapping Agency TR 8350.2, 3rd ed., Reston, VA,
July 1997.
[25] U.S. Standard Atmosphere, 1976, NASA TM-X-74335, U.S.
Government Printing Ofce, Washington, D.C, Oct 1976.
[26] Picone, J. M., Hedin, A. E., Drob, D. P., and Aikin, A. C., NRLMSISE-
00 Empirical Model of the Atmosphere: Statistical Comparisons and
Scientic Issues, Journal of Geophysical Research, Vol. 107, No. A12,
Dec. 2002, pp. 14681483.
doi:10.1029/2002JA009430
[27] Brauer, G. L., Cornick, D. E., Habeger, A. R., Petersen, F. M.,
and Stevenson, R., Program to Optimize Simulated Trajectories
(POST): Volume 1 Formulation Manual, NASA CR-132689,
April 1975.
[28] Powell, R. W., Striepe, S. A., Desai, P. N., and Braun, R. D., Program
to Optimize Simulated Trajectories (POST). Volume 2 Utilization
Manual, Version 5.2, NASA Langley Research Center, Oct. 1997.
JOHNSON, LEWIS, AND STARKEY 49
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5

[29] Rasmussen, M., Hypersonic Flow, Wiley, NewYork, 1994, Chaps. 8, 10.
[30] Moseley, W. C., Jr., Graham, R. E., and Hughes, J. E., Aerodynamic
Stability Characteristics of the Apollo Command Module, NASA
TN D-4688, 1968.
[31] Tauber, M. E., and Menees, G. P., Aerothermodynamics of
Transatmospheric Vehicles, Journal of Aircraft, Vol. 24, No. 9,
1987, pp. 594602.
doi:10.2514/3.45483
[32] Ried, R. C., Jr., Rochelle, W. C., and Milhoan, J. D., Radiative Heating
to the Apollo Command Module: Engineering Prediction and Flight
Measurement, NASATM X-58091, April 1972.
[33] Kaattari, G. E., A Method for Predicting Shock Shapes and Presssure
Distributions for a Wide Variety of Blunt Bodies at Zero Angle of
Attack, NASATN D-4539, April 1968.
[34] Kaattari, G. E., Shock Envelopes of Blunt Bodies at Large Angles of
Attack, NASATN D-1980, Dec. 1963.
[35] Bertin, J., Hypersonic Aerothermodynamics, AIAA Education Series,
AIAA, New York, 1993, p. 271.
[36] Lovelace, U. M., Charts Depicting Kinematic and Heating Parameters
for a Ballistic Reentry at Speeds of 26,000 to 45,000 Feet Per Second,
NASATN D-968, Oct. 1961.
[37] Tauber, M. E., and Sutton, K., Stagnation-Point Radiative Heating
Relations for Earth and Mars Entries, Journal of Spacecraft and
Rockets, Vol. 28, No. 1, 1991, pp. 4042.
doi:10.2514/3.26206
[38] Cauchon, D. L., Radiative Heating Results from the FIRE II Flight
Experiment at a Reentry Velocity of 11.4 Kilometers per Second,
NASATM X-1402, July 1967.
[39] Madavan, N. K., On Improving Efciency of Differential Evolution
for Aerodynamic Shape Optimization Applications, AIAA
Paper 2004-4622, Aug. 2004.
[40] Deb, K., Multi-Objective Optimization Using Evolutionary Algorithms,
Wiley, Chichester, England, U.K., 2001.
[41] Rai, M. M., Robust Optimal Aerodynamic Design Using Evolutionary
Methods and Neural Networks, AIAA Paper 2004-778, Jan. 2004.
[42] Johnson, J. E., Lewis, M. J., and Starkey, R. P., Coupled Entry Heat
Shield/Trajectory Optimization for Lunar Return, AIAA Atmospheric
Flight Mechanics Conference, AIAA Paper 2008-6557, Aug 2008.
[43] Jones, E., and Shiva, K. (eds.), NASATechnical Standards Program,
NASASTD-3001, Vol. 2, https://standards.nasa.gov/documents/detail/
3315785 [retained 2011].
[44] Graves, C. A., and Harpold, J. C., Re-Entry Targeting Philosophy and
Flight Results from Apollo 10 and 11, AIAA Paper 1970-0028,
Jan. 1970.
[45] Young, J. W., and Smith, R. E., Trajectory Optimization for An
Apollo-Type Vehicle Under Entry Conditions Encountered During
Lunar Return, NASATR R-258, May 1967.
[46] Scott, C. D., Ried, R. C., Maraia, R. J., Li, C. P., and Derry, S. M., An
AOTVAeroheating and Thermal Protection Study, Thermal Design of
Aeroassisted Orbital Transfer Vehicles, edited by H. F. Nelson, Vol. 96,
of Progress in Astronautics and Aeronautics, AIAA, New York, 1995,
pp. 309337.
K. Wurster
Associate Editor
50 JOHNSON, LEWIS, AND STARKEY
D
o
w
n
l
o
a
d
e
d

b
y

U
n
i
v
e
r
s
i
t
y

o
f

S
u
r
r
e
y

o
n

N
o
v
e
m
b
e
r

1
,

2
0
1
2

|

h
t
t
p
:
/
/
a
r
c
.
a
i
a
a
.
o
r
g

|

D
O
I
:

1
0
.
2
5
1
4
/
1
.
4
2
5
6
5