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

net/publication/369624033

Satellite design optimization for differential lift and drag applications

Preprint · March 2023

CITATIONS READS

0 88

5 authors, including:

Constantin Traub Marcel Pfeiffer

Universität Stuttgart Universität Stuttgart
59 PUBLICATIONS 499 CITATIONS 67 PUBLICATIONS 610 CITATIONS

SEE PROFILE SEE PROFILE

Stefanos Fasoulas
Universität Stuttgart
399 PUBLICATIONS 2,589 CITATIONS

SEE PROFILE

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

Heat Shield Material: Ablators View project

Aerothermochemistry View project

All content following this page was uploaded by Constantin Traub on 30 March 2023.

The user has requested enhancement of the downloaded file.

 Satellite design optimization for differential lift and drag applications

C. Marianowskia , C. Trauba , M. Pfeiffera , J. Beyera , S. Fasoulasa

a Institute of Space Systems, University of Stuttgart, Pfaffenwaldring 29, Stuttgart, 70569, Germany

Abstract
Utilizing differential atmospheric forces in the Very Low Earth Orbits (VLEO) regime for the control of the relative motion within
arXiv:2303.16612v1 [physics.space-ph] 29 Mar 2023

a satellite formation is a promising option as any thrusting device has tremendous effects on the mission capacity due to the limited
weight and size restrictions of small satellites. One possible approach to increase the available control forces is to reduce the mass
of the respective satellites as well as to increase the available surface area. However, satellites of these characteristics suffer from
rapid orbital decay and consequently have a reduced service lifetime. Therefore, achieving higher control forces is in contradiction
to achieving a minimum orbital decay of the satellites, which currently represents one of the biggest challenges in the VLEO regime.
In this work, the geometry of a given reference satellite, a 3UCubeSat, is optimized under the consideration of different surface
material properties for differential lift and drag control applications while simultaneously ensuring a sustained VLEO operation.
Notably, not only the consideration of sustainability but also the optimization with regard to differential lift is new in literature. It
was shown that the advantageous geometries strongly depend on the type of gas-surface interaction and thus, two different final
designs, one for each extreme type, are presented. In both cases, improvements in all relevant parameters could be achieved solely
via geometry adaptions.
Keywords: VLEO, satellite aerodynamics, shape optimization, differential drag and lift

1. Introduction been assessed. In parallel efforts, optimized shapes for VLEO

satellites targeting a minimization of the atmospheric drag force
In the recent past, interest in operating satellites at much and thus extension of operational lifetime have been developed.
lower altitudes than before, in the so-called VLEO regime (i.e. Thereby, the satellite geometry has been optimized via a novel
the entirety of orbits with a mean altitude < 450 km [1]), has 2D profile optimization tool [5]. It was shown that the opti-
increased due to a variety of advantages, which have been sum- mized satellite geometries offer pure passive lifetime extensions
marized by Crisp et al. [1]. Similarly, the application of dis- of up to 46 % compared to a GOCE like reference body.
tributed satellite systems, which commonly are made up of sev-
eral small satellites working together to achieve a shared goal, is
nowadays prevalent [2]. Due to their stringent volume and mass In this article, which is based on the corresponding author’s
limitations, alternative solutions to the conventional propulsion master’s thesis [6] and builds upon said contributions, optimal
methods to exert control forces are of great interest. In the satellite designs for differential lift and drag applications are
VLEO regime, the use of the atmospheric forces from the pre- presented. To the best of the authors’ knowledge, such efforts
vailing residual atmosphere represents a promising solution. have never been discussed in the literature. The article elab-
At the Institute of Space Systems of the University of orates how the research objective leads to conflicting require-
Stuttgart, this methodology is actively researched since 2018 ments, which need to be prioritised accordingly to arrive at an
[3]. In the most recent publication, a planning tool for opti- overall optimal design. All optimization steps are presented in
mal three-dimensional formation flight maneuvers of satellites detail and the corresponding considerations are discussed. In
in VLEO using aerodynamic lift and drag via yaw angle devia- future efforts, it is foreseen to combine the optimal maneuver
tions has been presented [4]. As in the VLEO regime the large planning and the optimal satellite designs, which would repre-
levels of orbital decay represent the major challenge to be over- sent a more holistic optimization approach.
come for a sustained operation to become reality, the planned
trajectory is optimal in a sense that the overall decay during the
maneuver is minimized. Thereby, the remaining lifetime of the
satellites is maximized and the practicability and sustainability The necessary fundamentals are briefly presented in Sec-
of the methodology significantly increased. Additionally, ap- tion 2. In Section 3, the for the calculations and optimization
plying yaw angle deviations allows to simultaneously control set parameters and constraints as well as the applied optimiza-
the in- and out-of-plane relative motion via differential lift and tion steps are introduced. The design optimization for diffuse
drag. Throughout the work of Traub et al. [4], however, con- re-emission is conducted in Section 4, and the similar process
ventional 3UCubeSats augmented with two solar panels have for specular reflection can be found in Section 5.
Preprint submitted to Acta Astronautica March 30, 2023
 h [km] h0 [km] ρ0 [kg m−3 ] H0 [km]
Incident Flux Incident Flux
250-300 250 7.248·10-11 45.546
300-350 300 2.418·10-11 53.628
350-400 350 9.518·10-12 53.298
400-450 400 3.725·10-12 58.515

Table 1: Parameters for Eq. 3 [7].

Figure 1: Specular reflection (left) and diffuse re-emission (right) [12].

2. Satellite aerodynamics

2.1. Fundamentals particle, which is depending on the amount of exchanged en-

ergy between particle and wall, commonly described using en-
The forces acting on a satellite are the result of the inter-
ergy accommodation coefficients, and the angle of reflection/re-
change of momentum between the particles of the prevailing
emission, which is additionally depending on the type of gas-
atmosphere and the satellite. Whereas the atmospheric forces
surface-interaction (GSI). There are two extreme types of GSI;
attacking at the center of gravity are of interest within this work,
specular reflections, where the angle between surface and re-
the torque induced by the atmosphere was neglected and con-
flected particle is equal to the angle between surface and in-
signed to the attitude control system of the satellite. The spe-
coming particle, and diffuse re-emissions, with a particle re-
cific drag force f~D and specific lift force f~L for a satellite of emission according to a probabilistic velocity and direction dis-
mass m can be calculated as shown in Eq. 1 and Eq. 2: tribution (see Fig. 1).
1 C D · Are f ~vrel In general, specularly reflecting material are advantageous
f~D = · ρ · · |~vrel |2 · , (1) as they allow a deliberate deflection of the incoming particles.
2 m |~vrel |
However, specularly reflecting materials are not yet available
1 C L · Are f (~vrel × ~n) × ~vrel and remain subject of current research [8, 9]. Hence, the pre-
f~L = · ρ · · |~vrel |2 · , (2) vailing type of GSI in VLEO on the currently used materials is
2 m |(~vrel × ~n) × ~vrel |
a diffuse or quasi-diffuse gas-surface interaction [10, 11].
where ρ is the prevailing density, C D the drag coefficient, C L the
lift coefficient, Are f the reference area, which corresponds to the 2.2.1. Accommodation coefficients
projected area of the satellite perpendicular to the flow, and ~vrel The thermal energy accommodation coefficient
the relative velocity to the local atmosphere. In Eq. 2, ~n denotes
the surface normal vector of the surface under consideration. Ei − Er
αT = (4)
In order to investigate the sustainability of the different satel- Ei − Ew
lite designs, their lifetime was estimated. To ensure compara- is a measure for the adaption of energy from the impinging par-
bility with the previous study [5], the same simplified equation ticles Ei to the energy Ew , which the particles would have after
for the lifetime of a satellite in a circular orbit was applied [7]: reaching equilibrium with the satellite’s wall temperature. Er
corresponds to the actual energy of the reflected particles. Sim-
β · H0 h0 h0
! !!
tl = √ · 1 − exp − · 1 + . (3) ilarly, the momentum exchange between particle and surface is
ρ0 · µE · a H0 2a commonly described by the tangential momentum accommoda-
tion coefficient σt and the normal momentum accommodation
Here, tl is the lifetime of the satellite for a given orientation
coefficient σn as follows [13]:
and no additional use of propulsion. H0 is the atmospheric
scale height, h0 the base altitude and ρ0 the density as shown in τi − τr
σt = , (5)
Tab. 1. µE is the Earth’s gravitational parameter and a = h0 +RE τi − τw
with RE being the Earth’s radius. Although Eq. 3 is very sim- pi − pr
plified due to the neglected changes in density with time and σn = . (6)
pi − pw
location, it is well suited for the purpose of this work, since all
Here, τi is the tangential momentum carried to the surface by
atmospheric changes have to be eliminated to only account for
the incident particle and τr is the tangential momentum carried
the influence of the satellite’s design.
away from the surface by the reflected particle. The tangen-
tial momentum carried away from the surface by a diffusely re-
2.2. Aerodynamic coefficients and gas-surface-interaction flected particle after reaching thermal equilibrium with the wall
In Eq. 1 and Eq. 2, the parameters m, Are f , C D , and C L are is ascribed to τw and is per definition equal to zero. The nor-
characteristics of the satellite’s design. Whereas the first two mal momentum carried to the surface by the incident particle
can be directly derived and measured using the satellite or its and normal momentum carried away from the surface by the
model, the two aerodynamic coefficients are depending on the reflected particle are pi and pr respectively, and pw is the nor-
surface properties of the satellite and have to be calculated an- mal momentum carried away from the surface by a diffusely
alytically or numerically. The two main parameters influencing reflected particle after reaching thermal equilibrium with the
the aerodynamic coefficients are the velocity of the reflected wall.
2
 Drag Force • A nominal flight configuration experiencing as little drag
Flight Direction as possible and no/negligible lift ( f~D,min , see Fig. 3a left),

• A maximum drag configuration for experiencing the high-

est possible drag ( f~D,max , see Fig. 3a right),

• A maximum lift configuration for experiencing the highest

possible lift ( f~L,max , see Fig. 3b).

With the aim of optimizing the satellite’s design for a sustain-

able use of differential lift and drag control methods in VLEO,
the specific optimization goals can be expressed as follows:

Figure 2: Depiction of a maximum differential drag force. • Drag Goal #1: Decreasing f~D,min
flight direction ys
vector of lift force
vector of drag force • Drag Goal #2: Increasing f~D,max
xs

• Lift Goal #1: Increasing f~L,max

xs
In order to derive design recommendations, the different de-
ys
sign characteristics are prioritized as they contribute to contra-
dictory optimization goals. Considering that aerodynamic force
control methods are not the means of choice for a time-critical
(a) Left: nominal flight configuration, right: maximum drag configuration maneuver due to the comparatively low absolute values of spe-
cific forces, the overall priority is to decrease the orbital decay.
It can be assumed that the satellite spends most of its service
ys xs life in the nominal flight configuration, and hence as little un-
wanted aerodynamic forces as possible are desired. Due to the
prevailing condition of the contamination with atomic oxygen
xs on currently used materials and a therefore diffuse gas-surface-
interaction with generally little C L values, increasing the lift
ys force is set to be the second priority within this work.
To summarize, the priorities underlying all decisions within
(b) Left: maximum lift configuration, right: opposite maximum lift configuration
this work are set as follows:
Figure 3: Depiction of the different possible orientations of a satellite depending
on the desired control forces. Here: reference satellite (Fig. 4) in top view. Priorities: ↑ tl > ↑ f~∆L > ↑ f~∆D ,

with ↑ indicating an increase. However, these priorities are

2.3. Differential lift and drag control method based on the considerations mentioned above, and naturally
In order to change the formation geometry, the method of may differ for individual missions.
differential lift and drag is generally understood to intention-
ally create differences in the acting aerodynamic forces [3], e.g.
by rotating only one of the two satellites of the formation (see 3. Optimization approach
Fig. 2). Whereas differential drag effects the relative movement
within the orbital plane, differential lift effects the out-of-plane The framework conditions set to pursue the defined goals
relative movement. The maximum achievable differential spe- alongside the utilized tools and optimization steps taken are pre-
cific forces are sented in the following sub-sections.

| f~∆D | = | f~D,max | − | f~D,min |, (7) 3.1. Boundary conditions of the optimization

| f~∆L | = | f~L,max,1 | + | f~L,max,2 | = 2 · | f~L,max |, (8) In this section, the chosen atmospheric conditions for the de-
termination of the forces, as well as the chosen reference satel-
if | f~L,max,1 | = | f~L,max,2 | and f~L,max,1 ↑↓ f~L,max,2 as shown in lite and the derived geometry constraints for the design varia-
Fig. 3b. tions are briefly presented.
Within this work, it is assumed that all satellites within the
formation share the same design. The following flight configu-
rations1 are defined, of which Fig. 3 gives an overview of: entation of a satellite, which can be further defined by the angle between the
satellite’s longitudinal axis and the orbital plane (angle of sideslip - AOS) as
well as the angle between the longitudinal axis and the plane formed by ~vrel
1 In this article, the term ”(flight) configuration” is used to describe the ori- and the orbit’s normal vector (angle of attack - AOA).

3
 Parameter Unit Value Diffuse Re-emission
−∆1
αT,1 = 1.00 −→ 0.91 −→ 0.70
−∆2

Ti K 1,056.6 +∆1 +∆2

ρ kg m−3 9.15 ·10-12 Specular Reflection αT,4 = 0.00 −→ 0.09 −→ 0.30
M kg mol−1 0.0174 Table 3: Gradation of the surface properties.
vrel m s−1 7,697.1
s - 7.66
nO m−3 2.64·1014 satellite (see Fig. 4). For compliance with the commonly uti-
nN2 m−3 4.18·1013 lized CubeSats [17], the mass of the satellite is set to be 5 kg.
nHe m−3 4.88·1012 The reference satellite has a maximum length lmax and height
nN m−3 4.44·1012 hmax of 0.3 m each as well as a main body volume V MB of
nO2 m−3 1.09·1012 0.003 m3 with main body height and width of both 0.1 m. The
nH m−3 8.53·1010 width of the panel is equal to 0.003 m.
nAr m−3 8.63·109 The dependence of the atmospheric forces on the surface
properties of the satellite is taken into account by consider-
Table 2: Atmospheric data used for all calculations in this work: temperature
of the impinging particle T i , molecular mass M, molecular speed ratio s, and
ing different energy accommodation coefficients αT and corre-
particle density ni of species i. sponding gas-surface-interaction models (GSIM). An overview
of the applied surface properties is given in Tab. 3. The cho-
sen gradation steps ∆1 and ∆2 are derived from the traditional
lm surface materials used for satellites. αT,1 = 1.00 represents the
ax
worst case scenario for a deliberate deflection of the particle.
The second value, αT,2 = 0.91 represents a value for the tra-
ditional surface materials [10, 11] and αT,3 = 0.70 represents
hmax

improved material properties for diffusely re-emitting materials

[18]. Therefore, ∆1 = 0.09 and ∆2 = 0.21, which is equally
V MB
applied on the energy accommodation coefficients for the con-
sideration of specular reflection as shown in Tab. 3. Addition-
ally, this choice of αT allows comparison to other literature with
similar gradation steps [19].

3.1.3. Geometry constraints

For comparability of the results within this work, all design
variations have to comply with the following constraints regard-
Figure 4: Reference satellite: 3UCubeSat with panels.
ing the geometry:

• Constant volume: Main body volume remains constant.

3.1.1. Atmospheric conditions
The atmospheric conditions are set constant for all consider- • Height and length limitations: Maximum height and
ations within this work in order to ascribe changes in the spe- length of the reference satellite must not be exceeded.
cific forces to the design variations. Therefore, moderate space
weather with a 10.7 cm flux F10.7 of 140 sfu and a geomagnetic Due to these geometric constraints, the mass of the satellite can
activity index of A p = 15 were chosen2 . The data shown in be considered constant and is equal to 5 kg according to the ref-
Tab. 2 was obtained by averaging the output of the NRLMSISE- erence satellite. Thus, the only variables in Eq. 1 and Eq. 2
00 model [15] of one day per month of 2004 over the year for are the aerodynamic coefficients and the reference area, and
the chosen altitude of 350 km. Additionally, the satellite’s wall changes in the forces can be directly attributed to the design
temperature is set to T w = 300 K as commonly assumed in the changes.
regarding literature [14, 16].
3.2. Utilized tools

3.1.2. Reference satellite The utilized open-source tools as well as an extension for a
Matlab 2D profile optimization tool for the derivation of satel-
All simulation results are related to a reference satellite un-
lite geometries are introduced in the following sections.
der the same conditions in order to evaluate different satellite
designs. Due to its universal applicability for various payloads
3.2.1. Determining the atmospheric forces
and the size suitable for satellite formations, a 3UCubeSat with
one panel each on the top and bottom is used as a reference Since numerous investigations of the influences of differ-
ent design parameters are necessary, the calculation time for
the aerodynamic forces per satellite model plays an important
2 Detailed information about solar proxies and indices for space weather de- role. Therefore, the Matlab toolkit ”ADBSat”, which was de-
scription can be found in the work of Doornbos [14]. veloped at the University of Manchester [21], was used to study
4
 Extended Volume Derivations

Original Volume Derivations

wN

h MB

h MB
Geometry A Geometry B Geometry C

B+ h MB B+ wN , h MB

Figure 5: Possible 3D volume derivations for the 2D optimization.

xlocal Θi xlocal
f (i)
→drag
f (i − 1)

ylocal
x
ylocal
dx
↓li f t
Figure 6: Definition of the 2D optimization tool function values according to Hild [20].

the effects of different designs on the aerodynamic forces. It etry option B is obtained by mirroring the profile at the x-axis
represents a means of calculating the acting forces with lit- and then extruding symmetrically with the value of the largest
tle computational effort. However, this method is not suitable cross-section width. Thereby, geometry option B has a squared
for concave surfaces, since multiple reflections of the incom- rear side by default. Geometry option C is obtained by inter-
ing particles cannot be taken into account. For selected and secting the extruded profile of option B with a cylinder with
promising geometries as well as geometries causing multiple the given length and a radius equal to the largest cross-section
reflections, their actual performance was determined using the width of the optimized profile. Additionally, the application of
more computationally intensive but proven Direct Simulation a tail geometry is possible, in which the height of the 2D pro-
Monte Carlo (DSMC) method, which is available within the file is reduced again from a certain longitudinal position. Each
gas-kinetic simulation framework ”PICLas”, developed at the individual profile optimization complies with the given volume
University of Stuttgart [22]. For the PICLas simulations, the restriction of V MB = 0.003 m3 . Within this work, the 2D profile
necessary variable hard sphere parameters for the species listed optimization tool was extended as presented in the following:
in Tab. 2 were taken from Bird [23]. A comparison of ADBSat Additional geometry options: The extrude option B was
and PICLas results can be found in Appendix A. modified to consider the following two additional user inputs
(see Fig. 5 right):
3.2.2. 2D profile optimization 1. Extrusion height h mb : For the application of the opti-
For optimizing and obtaining the satellite models, an already mization tool on a given structure such as the panels of
existing Matlab 2D optimization tool for reduced drag [5] was this work’s reference satellite, the height of the extrusion
utilized and adapted in the scope of this work. This tool was can be set as an input parameter.
developed for an analytical determination of a 2D profile with 2. Nose width w N : Whereas it might be aerodynamically
minimum drag for a given length and volume. A preceding de- advantageous to have a sharp or pointy nose geometry, it
termination of the type of volume derivation of the profile is could pose a problem for accommodating the payload and
necessary in order to comply with the given volume restrictions manufacturing of the satellite. Hence a definition of a nose
during the profile optimization. In the given optimization tool, width during the optimization process was implemented.
three different volume derivation options are possible as shown
on the left in Fig. 5. Whereas geometry option A is obtained by Lift optimization: Within the 2D optimization tool, a deriva-
rotating the profile around the x-axis (longitudinal axis), geom- tion of Sentman’s equation was used [5]. Since the Matlab tool
5
is aiming on optimizing a 2D profile, it followed for Sentman’s • Increasing number of area elements with C D,max :
equation that the two direction cosine η, t = 0. Considering the For an AOS of 90°, Are f cannot be further increased
definition of the area element on the profile as shown in Fig. 6, due to the given lmax and hmax and thus, the overall
the direction cosine between the local x and y axis and the di- C D of the whole body needs to be maximized. A
rection of the drag force is set as follows in the original tool: possible solution for increasing C D is implementing
the largest perpendicular area to the flow in the max-
kd = cos(Θi ) (9) imum drag configuration. Within this work, this was
ld = sin(Θi ) (10) achieved by changing the panel position and approxi-
mating the satellite shape with vertical and horizontal
with area elements (see Figs. 8a-8c).
f (i) − f (i − 1)
!
Θi = atan . (11) • Multiple reflections: For a consideration of the
dx
whole geometry and not only single area elements
By setting the two direction cosine according to the desired (as within ADBSat), a geometry promoting multi-
force direction of the lift force (see Fig. 6) to ple reflections poses another possible approach to
increase the drag. This approach was implemented
 π
kL = cos Θi + and (12) in the design by overlaying the optimized 2D profile
2 with a zigzag curve for particle-trapping concave ar-
 π
lL = sin Θi + , (13) eas (see Figs. 8d-8e).
2
an optimization of a given profile with regard to increasing the Opposite to the design optimization process for differential
lift force experienced by the profile is now possible as well. It drag, where a high difference of f~D,min and f~D,max but gener-
should be noted, that therefore the sign of the original optimiza- ally lower absolute values of f~D,min and f~D,max are desired, the
tion function has to be inverted, since originally a decrease in goal for optimizing the design for differential lift only consists
the acting force was desired. of increasing f~L,max . The process for optimizing the design for
differential lift was divided into the two steps:
3.3. Optimization steps
Lift 1a - Increasing maximum lift: For achieving higher lift
The optimization process presented in this subsection is forces, the profile obtained using the EPOT lift optimiza-
equally executed for the assumption of diffuse re-emission and tion function (see Fig. 8f and 8g) was implemented in the
specular reflection. In a first step, the design is varied with design.
the aim of optimizing the design for solely differential drag or
solely differential lift, in order to investigate the respective im- Lift 1b - Eliminating lift reducing areas: When considering
portant design characteristics. In a second step, the design char- the reference satellite with its frontal area it is apparent,
acteristics are combined for achieving a best possible trade-off that for an AOS ≈ 45° (i.e. maximum C L value for the
according to the set priorities. panels, see Fig. 7), the frontal area and the velocity vector
The optimization process for deriving a design optimal for form an angle ≈ −45°. Therefore, the frontal area experi-
differential drag is divided into the following steps: ences lift as well, but in the opposite direction as desired.
Hence, when dealing with different profiles than the opti-
Drag 1a - Decreasing minimum drag (main body): The mized one, areas experiencing lift in the opposite direction
main body is optimized for minimum drag in the nominal than desired must be decreased.
flight configuration utilizing the extended Matlab 2D
profile optimization tool (EPOT) while maintaining the Each design was generally tested in a variety of AOA and
maximum length of 30 cm and the main body volume of AOS in order to determine the orientation achieving the high-
0.003 m3 . est drag and lift force, referred to as the maximum drag and
maximum lift configuration, respectively. The same was ap-
Drag 1b - Decreasing minimum drag (panel): The panel is plied to the reference satellite. Therefore, in order to compare
optimized for minimal drag in the nominal flight con- the optimization in drag and lift forces, the respective config-
figuration using the EPOT, considering a given extrusion urations with the maximum achievable values are compared to
height of 30 cm. The trivial solution, omitting the panel, is each other.
not considered within this work as it would greatly reduce
the achievable control forces. Thus, the size of the panel
remains lmax · hmax . 4. Design optimization - diffuse re-emission

Drag 2 - Increasing maximum drag: The loss in maximum The results of a design variation are generally related to the
drag due to the first and second optimization steps has to be reference satellite with the same surface properties, if not stated
kept as small as possible. Considering the optimum angle otherwise. The gas-surface-interaction model applied for the
for the maximum of C D (see Fig. 7) and the set geometry ADBSat calculations within this section is Sentman’s model
restrictions, the following approaches can be derived: [24].
6
 3
0.4

2 0.3

C L [-]
C D [-]

0.2
1 αT = 1.0
αT = 0.9 0.1
αT = 0.8
αT = 0.7 0
0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
ψ [°] ψ [°]

Figure 7: Aerodynamic coefficients for a flat plate depending on the angle ψ between flow and surface normal according to Sentman’s model.

4.1. Differential drag optimization the lifetime, options for increasing the maximum drag while
Within the design considerations for the use of differential maintaining the geometry characteristics of geometry option A
drag, the influence of general shape and tail geometry, surface were investigated.
structures and panel position has been analyzed, and the results
are presented in the following sections. 4.1.2. Surface structures
Since the main body optimization with the EPOT led to
4.1.1. General shape and tail geometry side areas with different orientations to the flow and therefore
As investigated by Hild [20] and Walsh [25], it can be ad- smaller C D values, the maximum drag experienced is smaller
vantageous for the overall lifetime of the satellite to apply a tail compared to the reference satellite with completely vertical side
geometry on slender bodies. In the following, the effect of a tail surfaces. In order to increase the C D values of the side sur-
geometry for the three different volume derivation options A, faces in the maximum drag configuration while maintaining the
B and C is investigated in order to find the optimum combina- overall optimized profile shape obtained by the EPOT, a raster-
tion for an increased lifetime. At the same time, their influence ized cross section was added to the optimized profile in order
on the achievable maximum drag and thereby the overall differ- to approximate the curved surface by horizontal and vertical
ential drag is assessed. The influence of the tail length on the surface elements. Therefore, the frontal circular area, which
lifetime in nominal flight configuration and experienced drag was swept along the optimized profile, is rasterized using the
in maximum drag configuration as well as the resulting differ- midpoint circle algorithm3 to two different raster step sizes as
ential drag for different αT can be seen in Fig. 9, Fig. 10 and shown in Fig. 8a and Fig. 8b. Here, Raster 1 has a step size
Fig. 11. The data shown was obtained using ADBSat. of r f ront /14.3 , with r f ront being the radius of the frontal area,
It is apparent, that the optimum tail length generally depends which is a result of the EPOT for given volume restriction, lmax ,
on the volume derivation option and the surface properties. αT and geometry option. With this step size, the geometry con-
Whereas the geometry option A is promising for high αT , the straint of constant main body volume remains met. Raster 2
increase in lifetime for geometry options B and C generally in- has a step size of r f ront /10. Another possible means to increase
creases with decreasing αT . The three following geometry op- the maximum drag is to use surface structures promoting mul-
tions and tail lengths are optimal concerning the lifetime for the tiple reflections. Therefore, the geometries shown in Fig. 8d
given cases of surface properties: for αT = 1.00 and αT = 0.91 and Fig. 8e, which are derived by overlaying the optimized pro-
geometry A with 23 % tail length and for αT = 0.70 geometry file for geometry option A with a zigzag-curve with an extrema
A with 25 % tail length. However, as shown in Fig. 10, geom- distance dex of 5 mm (design MultRef 1) and 10 mm (design
etry option B experiences the higher drag forces in maximum MultRef 2) as well as extrema of ±5 mm, were investigated
drag configuration as expected due to its vertical side surfaces, further.
which are perpendicular to the flow in maximum drag config- The results for the above presented satellite designs are listed
uration and thus correspond to the highest C D values. Even in Tab. 4. Here, the lifetime was estimated in the nominal flight
though geometry option A can reach a lifetime increase of up configuration and the maximum drag in its respective config-
to 13%, the negative effect on the maximum achievable drag by uration. The data shown was obtained using PICLas and is
the rounded side surfaces leads to an overall loss in differential referring to the data from the reference satellite obtained us-
drag compared to the reference satellite as visible in Fig. 11. ing PICLas as well. It can be seen, that the loss in maximum
The increase in lifetime with geometry option B is smaller than
for A and C, but due to the generally higher maximum drag, an
increase in differential drag compared to the reference satellite 3 The midpoint circle algorithm is an algorithm used for determining the

is possible. However, as the priority of this work is to increase points of a rasterized circle.

7
 (a) Raster 1 (b) Raster 2 (c) Raster 3

(d) MultRef 1 (e) MultRef 2 (f) Lift 1

(g) Lift 2 (h) SpecOpt (i) SpecPrac

Figure 8: Selection of designs evaluated with PICLas and ADBSat.

8
 14
12
10
8
∆tl [%]

6
4 Geometry A
Geometry B
2 Geometry C

0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
lT [%] lT [%] lT [%]
(a) αT = 1.00 (b) αT = 0.91 (c) αT = 0.70

Figure 9: Dependence of the lifetime on general geometry and tail length.

2
Geometry A
1 Geometry B
0 Geometry C
∆ fD,max [%]

−1
−2
−3
−4

0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
lT [%] lT [%] lT [%]
(a) αT = 1.00 (b) αT = 0.91 (c) αT = 0.70

Figure 10: Dependence of maximum drag on general geometry and tail length.

0
Geometry A
Geometry B
∆ f∆D [%]

−1 Geometry C

−2

−3

0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
lT [%] lT [%] lT [%]
(a) αT = 1.00 (b) αT = 0.91 (c) αT = 0.70

Figure 11: Dependence of differential drag on general geometry and tail length.

9
 Design ∆tl ∆ fD,max ∆ f∆D geometry. Hence, the compared lift values represent the maxi-
Raster 1 +12.26 % -0.32 % +2.44 % mum possible lift, but may be experienced at different AOSs.
Raster 2 +11.81 % -0.31 % +2.34 % Optimizing the profile of the satellite using the EPOT’s lift
dex = 5 mm +3.27 % -1.55 % -1.13 % optimization function led to a simple geometry as presented in
dex = 10 mm +0.35 % -0.50 % -0.54 % Fig. 8f for a given main body height. However, here the side
surface of the main body and of the panel form a different angle
Table 4: Data for the tested designs with surface structures, αT = 1.00.
with the macroscopic velocity vector. Hence, the geometry can
be further optimized by increasing the main body height to the
drag is smaller for the design with additional vertical side sur- original height of the panel as shown in Fig. 8g. In order to vali-
faces than the designs promoting multiple reflections. The force date the EPOT’s output, other geometries (e.g. the designs built
distribution of the different designs compared to the reference for the investigations within the previous section) have been
satellite is shown in Fig. 12 for αT = 1.00 and obtained us- evaluated as well, but none of them were able to increase the lift
ing PICLas. As the area element’s C D value is the highest for such as design Lift 2, when compared to the reference satellite.
the vertical surfaces, these surfaces experience the highest drag However, since design Lift 2 contains a sharp nose geometry,
forces. However, the effect of multiple reflections for the tested possibly posing a problem towards the accommodation of the
design is lower than expected and the additional decrease in payload, the consequences of a mandatory frontal area were in-
vertical side surface area leads to an overall loss in experienced vestigated as well as shown in Fig. 15. In order to obtain the
drag. The expected effect of multiple reflections increasing the ADBSat results for the change in lift force, the utilized satellite
drag in maximum drag configuration for the tested design did models were obtained extruding the with the EPOT optimized
not appear. However, the results for the minimum drag in nom- profile considering the respective main body heights h MB and
inal flight configuration differed up to 25 % compared to the nose widths wN . It is apparent, that for a given nose width, the
results obtained using ADBSat. Therefore, the effect of multi- overall frontal area increases with increasing main body height,
ple reflections did occur, but not in the desired way. It can be and thus the lift opposite to the desired lift direction increases
assumed, that the utilization of structures promoting multiple as well. Hence, the greater the given nose width, the smaller
reflections for an increased maximum drag force is generally the optimum main body height and achievable increase in lift
possible, although the tested design is not suitable in this re- force. In order to achieve a high increase in lift force, a frontal
spect. The rasterized cross section on the other side did not nose area should be avoided. If not possible, the main body
lead to a loss in the generally increased lifetime by the opti- height has to be chosen according to the applied nose width for
mized profile, and additionally increased the effective side area a maximum increase in lift force.
and thereby the differential drag. In the case of a chosen geom-
etry option C, the rasterized cross section can be applied as well 4.3. Discussion on material as design factor
for increasing the vertical side surfaces (see Fig. 8c).
Within this section, the influence of improved surface prop-
erties under the assumption of diffuse re-emission is discussed.
4.1.3. Panel position
In Fig. 16, the data obtained for the considerations of a frontal
Taking up the theoretical approach for further increasing the
nose area, geometry option B, and a varying main body height
maximum drag, changing the panel position as shown in Fig. 13
as presented in Section 4.2 for different αT is shown with re-
poses another possibility to increase the surface perpendicular
spect to the data from the reference satellite with αT = 1.00.
to the flow, which is marked in orange. This parameter study
Here, the comparison to the reference satellite of αT = 1.00 was
of the panel position was performed on a main body of geom-
chosen in order to visualize the effect of varying αT and vary-
etry option A, since the geometry option C has limited width
ing the geometry. It can be seen, that changing the reference
to move the panel due to the sharp nose, and since changing
satellite’s αT from 1.00 to 0.91 increases the lift more than op-
the panel position on geometry option B does not change the
timizing the geometry with hN = 0 for αT = 1.00. So not only
amount of surface elements perpendicular to the flow. The in-
does the change from diffuse re-emission to specular reflection
fluence of the panel position on lifetime, maximum drag and
enable greater atmospheric forces than a design variation, but
differential drag can be seen in Fig. 14 and was estimated using
also a reduction of αT for diffusely re-emitting materials. Es-
ADBSat, by gradually moving the panel from the centered po-
pecially for high αT , adapting the material is therefore a more
sition to the edge of the frontal circular area. It can be seen, that
powerful design variation than optimizing the geometry with
as expected, the maximum drag can be increased, but simulta-
regard to increased control forces. As Fig. 16 shows, the effect
neously the lifetime decreases due to the greater Are f in nominal
of a change in geometry (e.g. from hN = 0 cm to hN = 2 cm)
flight configuration. Hence, changing the panel position is not
increases with decreasing αT .
a recommended design variation for the use of differential drag
control methods in the given case.
4.4. Performance of the optimized designs
4.2. Differential lift optimization The final design recommendation for the assumption of a dif-
The lift values compared and shown in this section are refer- fuse re-emission is presented in the following and its actual per-
ring to the optimum angle of attack for the respective satellite formance was evaluated using PICLas.
10
 dF/dA [N/m²] dF/dA [N/m²]
·10−4 ·10−4
Raster 1 7 MultRef 1 7
6 6
5 5
4 4
3 3
2 2
1 1
0 0

Reference Reference

(a) Top: satellite design with additional vertical side surfaces (b) Top: satellite design promoting multiple reflections
Bottom: reference satellite Bottom: reference satellite

Figure 12: Force distribution shown in head on view in maximum drag configuration for αT = 1.00.

15
∆ fL,max [%]

10

wN = 0 cm
5 wN = 1 cm
wN = 2 cm
~vrel ~vrel wN = 3 cm

10 12 14 16 18 20 22 24 26 28 30
h MB [cm]

Figure 15: Influence of nose width wN and main body height h MB on the lift
Figure 13: Increasing vertical side surfaces by changing the panel position. force, αT = 1.00

1 4.4.1. Differential drag and lift - separate consideration

0 After investigating several design derivations of the refer-
ence satellite, a satellite geometry with a revolved lifetime-
−1
optimized 2D profile including an additionally added rasterized
∆ [%]

−2 cross-section swept along the 2D profile as a main body, and an

extruded lifetime-optimized 2D profile for the panel was shown
−3 to be advantageous for decreased minimum drag and increased
tl
−4 differential drag. The design (Raster 1) is shown in Fig. 8a with
fD,max
2D profiles optimized for αT = 1.00.
−5 f∆D
To increase the experienced lift force in the desired direction,
0.00 0.20 0.40 0.60 0.80 1.00
it was shown to be advantageous to increase the side surface
x panel /rN [-] area experiencing lift, as well as decreasing the frontal area,
which otherwise experiences a lift force in the opposite direc-
Figure 14: Influence of the panel position on lifetime, maximum drag and dif- tion than desired. The design recommendation (Lift 2), which
ferential drag referred to the panel position at x = 0, αT = 0.91.
applies for all the tested αT within the diffuse re-emission sec-
tion, is shown in Fig. 8g.
11
 103 Reference Raster 1 Lift 2
Reference
wN = 0 cm ∆tl % - +12.26 * -8.14
wN = 2 cm ∆ f∆D % - +2.44 * -4.17
αT = 1.00 ∆ f∆L % - +3.14 +16.03
αT = 0.91 tL d 157.09 176.35 * 142.27
αT = 0.70
| f∆D | m s−2 8.303·10−6 8.508·10−6 * 8.220·10−6
102 | f∆L | m s−2 6.221·10−7 6.416·10−7 7.218·10−7
∆ fL [%]

| fD,min | m s−2 2.157·10−6 1.922·10−6 * 2.380·10−6

| fD,max | m s−2 1.046·10−5 1.043·10−5 * 1.060·10−5
| fL,max | m s−2 3.110·10−7 3.208·10−7 3.609·10−7
| fD,L | m s−2 7.721·10−6 7.362·10−6 7.025·10−6

Table 5: Data for the reference satellite, design Raster 1 and Lift 2, αT = 1.00.
101 Data obtained using PICLas, with * marked data obtained using ADBSat.

5.1. Differential drag optimization

10 15 20 25 30 Within this section, the influence of general shape, tail ge-
h MB [cm] ometry, main body height, and frontal area on the experienced
drag forces is presented. Sections 5.1.2 and 5.1.3 differ from
Figure 16: Dependence of lift increase on the main body height and surface the respective sections under the design optimization process
properties compared to the reference satellite with αT = 1.00.
of diffuse re-emission, as they are linked to the results of the
investigation of general shape and tail geometry.
4.4.2. Differential drag and lift 5.1.1. General shape and tail geometry
The recommendation for differential drag and the recommen- In a first step, the best possible volume derivation of the op-
dation for differential lift do not share many geometric charac- timized 2D profile and the appropriate tail length (if any) were
teristics. Since a sustainable operation in VLEO is the goal, investigated. An overview of the lifetime in the nominal flight
design Lift 2 is not suitable for the general use for the relative configuration influenced by the three geometry derivation op-
motion control by means of the atmospheric forces. Only for tions A, B and C and the tail length is given in Fig. 18. The
αT = 0.70, no loss in lifetime compared to the reference satel- data shown was obtained using ADBSat. For very low energy
lite with αT = 0.70 is achieved with design Lift 2. However, accommodation coefficients, the geometry option B is the most
another solution for materials with higher αT is of interest. advantageous for a high increase in lifetime. With increasing
Raster 1 was shown to be the best trade-off, since the ver- energy accommodation coefficient, geometry option C is the
tical side surfaces for increasing the maximum drag simulta- more promising volume derivation option. From Fig. 18a and
neously have a positive effect on increasing the maximum lift Fig. 18b it can be seen, that for the mentioned advantageous
force and therefore, an improvement in all critical parameters volume derivation options the design with no tail geometry is
was achieved. The flow-fields in all three flight configura- the most promising design respectively.
tions for the reference satellite and design Raster 1 are given
in Figs. 17a-17f. It can be seen, that the 2D profile optimization 5.1.2. Main body height
of main body and panel reduces the shock occurring in Fig. 17b The volume derivation option B is promising regarding a
for x < 0 m. The reduction of opposed lift experienced by the long lifetime of the satellite. Since the original 2D optimiza-
frontal area is visible in Fig. 17f, where the shock preceding the tion tool does not investigate different main body heights for
frontal area of the recommended design is smaller compared to option B, their influence was investigated further. Figure 19
the reference satellite. The respective specific forces and life- gives an overview of the lifetime in nominal flight configura-
times for the designs Raster 1 and Lift 2 are given in Tab. 5. tion, maximum drag in maximum drag configuration and dif-
Due to the symmetry, the minimum lift is equal to zero. Since ferential drag depending on the main body height. The profiles
the altitude loss per maneuver plays an important role, the value used were obtained utilizing the EPOT with a given extrusion
for drag in the maximum lift configuration fD,L is given as well. height, which is set equal to the main body height h MB . The data
shown was obtained using ADBSat. It can be seen, that espe-
cially for very low energy accommodation coefficients, a high
5. Design optimization - specular reflection main body height has great potential for increasing the overall
lifetime. Additionally, an increased main body height is advan-
The satellite models within this section were generally eval- tageous for increasing the experienced drag in maximum drag
uated with the Cercignani-Lampis-Lord (CLL) model of ADB- configuration and thus also advantageous for a high differential
Sat. As the required input parameters are αn and σt and not drag force. The with regard to a long lifetime and high differ-
αT , they were derived from the chosen αT given in Tab. 3, as ential drag optimized satellite, obtained using geometry option
presented in Appendix B. B, is depicted as design SpecOpt in Fig. 8h.
12
 n [m−3 ] ·1014
0.8 0.8

0.6 0.6 6

0.4 0.4
4
0.2 0.2
y [m]

y [m]
2
0 0

-0.2 -0.2 0
-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.6 -0.3 0 0.3 0.6 0.9 1.2 -0.6 -0.3 0 0.3 0.6 0.9 1.2
x [m] x [m]
(a) Reference satellite - nominal flight configuration (b) Recommended design Raster 1 - nominal flight configuration
n [m−3 ] ·1014
0.8 0.8

0.6 0.6 6

0.4 0.4
4
0.2 0.2
y [m]

y [m]

2
0 0

-0.2 -0.2 0
-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.6 -0.3 0 0.3 0.6 0.9 1.2 -0.6 -0.3 0 0.3 0.6 0.9 1.2
x [m] x [m]
(c) Reference satellite - maximum drag configuration (d) Recommended design Raster 1 - maximum drag configuration
n [m−3 ] ·1014
0.8 0.8

0.6 0.6 6

0.4 0.4
4
0.2 0.2
y [m]

y [m]

2
0 0

-0.2 -0.2 0
-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.6 -0.3 0 0.3 0.6 0.9 1.2 -0.6 -0.3 0 0.3 0.6 0.9 1.2
x [m] x [m]
(e) Reference satellite - maximum lift configuration (f) Recommended design Raster 1 - maximum lift configuration

Figure 17: Simulated particle density n in the flow-field around the reference satellite and the recommended design Raster 1 for αT = 1.00.

13
 800 Geometry A
Geometry B
600 Geometry C
∆tl [%]

400

200
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
lT [%] lT [%] lT [%]
(a) αn = 0.00 σt = 0.00 (b) αn = 0.09 σt = 0.0459 (c) αn = 0.30 σt = 0.1627

Figure 18: Increase in lifetime ∆tL depending on general shape and tail length lT .

·10−5
3,000 13
1.95
| fD,max | [m s−2 ]

∆ f∆D [%]
∆tl [%]

2,000 1.9 11
1.85 αn = 0.00
1,000 αn = 0.09
9
1.8 αn = 0.30
0
10 15 20 25 30 10 15 20 25 30 10 15 20 25 30
h MB [cm] h MB [cm] h MB [cm]
(a) Lifetime (b) Maximum Drag (c) Differential Drag

Figure 19: Increase in lifetime, drag, and differential drag depending on the main body height h MB .

5.1.3. Consequences of practical considerations comparing the scales of the y-axes of Fig. 19a and Fig. 20a, any
type of frontal area leads to a significant loss in lifetime and
As aerodynamically promising design SpecOpt is, the practi-
thereby also reduces the achievable increase in differential drag
cability however of accommodating the payload can be lowered
compared to the reference satellite.
due to the sharp nose. A more practicable solution would be to
include a frontal area. Therefore, the following nose geometries
were considered: a constant nose width of wN = 3 cm, a con- 5.2. Differential lift optimization
stant nose area equal to a third of the reference satellite’s frontal
area AN,0 , a constant nose area equal to the half of AN,0 as well The suggested profile by the 2D optimization tool is equal to
as a constant nose area equal to two thirds of AN,0 . An exem- the one as derived in Section 4.2 and thus, the influence of dif-
plary depiction of the tested satellite geometries can be found ferent frontal areas was investigated again for the respective αT .
as design SpecPrac in Fig. 8i. Therefore, the satellite models from Section 4.2 were evaluated
for αT = [0.00; 0.09; 0.30] using ADBSat. Figure 21 shows,
The dependence of lifetime in nominal flight configuration
that whereas the absolute value of specific lift force also de-
and differential drag for the considered nose geometries and
pends on the energy accommodation coefficient, no dependence
different αT is shown in Fig. 20. The utilized satellite mod-
of the percentage increase in lift with respect to the reference
els were derived using the EPOT considering h MB and wN , and
satellite can be seen. Additionally it is visible that the opti-
the data shown was obtained using ADBSat. It can be seen that
mal main body height is depending on the nose width similarly
for a given wN regardless of h MB , a smaller main body height is
to the equivalent investigation results in the diffuse re-emission
more advantageous regarding the lifetime, as an increase in h MB
section. For a high increase in experienced lift force due to
simultaneously increases the frontal area and therefore the drag.
solely geometry design factors, a frontal area should be avoided
If the frontal area remains constant for different h MB , i.e. a de-
or, if not possible, the nose width should be kept as small as
creasing nose width wN for increasing main body height h MB ,
possible.
the influence of αn and σt as well as the achievable improve-
ment in lifetime and differential drag decreases with increasing
frontal area. Therefore it is recommended to choose a main 5.3. Performance of the optimized designs
body height as small as possible for a mandatory given nose
width. However, if the value of the nose width is not of direct The final design recommendation for the assumption of a
importance, but a frontal area is required, it is recommended specular reflection is presented in the following and its actual
to keep the nose width as small as possible. As visible when performance was evaluated using PICLas.
14
 10 No Nose Area
wN = 1cm
wN = 2cm
wN = 3cm
8

∆ fL,max [%]
200 6

150 2 αn = 0.00; σt = 0.0000

αn = 0.09; σt = 0.0459
αn = 0.30; σt = 0.1627
∆tl [%]

10 15 20 25 30
100 h MB [cm]

Figure 21: Influence of the main body height h MB and different nose widths on
the lift.

50
5.3.1. Differential drag and lift - separate consideration
Especially the extruded optimized 2D profile with an extru-
sion height equal to the main body height of 30 cm was shown
to drastically increase the lifetime and thus the differential drag
10 15 20 25 30 for low αT . The resulting design (SpecOpt) is shown in Fig. 8h
with a 2D profile optimized for αT = 0.00. This design is pre-
(a) Lifetime
sented to increase the understanding of advantageous design
10
characteristics prior to deriving a realistic design recommen-
dation.
9 Similar to the case of diffuse re-emission, it was shown to
be advantageous to increase the total side surface area by incli-
8 nation, as well as eliminating a frontal area. Regardless of the
tested GSIM and αT , the design optimal for differential lift is
7 given in Fig. 8g, as presented in Section 4.2.
∆ f∆D [%]

6 5.3.2. Differential drag and lift

Opposite to the designs within the section for diffuse re-
5 emission, the two previous satellite design recommendations
share many geometry characteristics. The only difference be-
tween these two is the slightly curved profile, which is the result
4
of the 2D optimization tool with regard to an increased lifetime.
As design SpecOpt has a higher increase in lifetime (first prior-
3 ity) compared to design Lift 2, but almost the same increase in
wN = 3 cm
AN = 1/3 · AN,0
lift (second priority), design SpecOpt is also well suited for the
αn = 0.00; σt = 0.0000
2 AN = 1/2 · AN,0 αn = 0.09; σt = 0.0459 use of both, differential lift and drag control methods. How-
AN = 2/3 · AN,0 αn = 0.30; σt = 0.1627 ever, while design SpecOpt is the theoretical optimal design
for the use in VLEO in a satellite formation, it not only poses
10 15 20 25 30
difficulties for the accommodation of the payload, but also for
h MB [cm] other factors such as manufacturing or attitude control. Hence,
(b) Differential Drag a more practicable but less optimal solution is presented and
discussed. A minimum required frontal area is assumed with a
Figure 20: Lifetime and differential drag depending on the main body height nose width of wN = 3 cm4 . In order to keep the loss in lifetime
for different frontal areas.
as small as possible, the main body height should remain as

4 This value was adopted to primarily study performance impacts. It remains

to be decided which value is reasonable.

15
 Reference 4b 3b of differential lift and drag control methods was possible with
∆tl % - +200 * +2,464 the primary objective of increasing the lifetime. Even for the
∆ f∆D % - +4.01 * +13.52 case of complete accommodation αT = 1.00, differential drag
∆ f∆L % - +0.66 +10.03 and lift could be increased by 2 % and 3 % respectively, while
tL d 142.40 427.14 * 3745.19 at the same time increasing the lifetime under the set conditions
| f∆D | m s−2 1.722·10−5 1.792·10−5 * 1.973·10−5 by 12 %. Despite the possible increase of differential forces, the
| f∆L | m s−2 1.380·10−5 1.389·10−5 1.518·10−5 differential lift remains two orders of magnitude smaller com-
| fD,min | m s−2 2.380·10−6 7.934·10−7 * 0.905·10−7 pared to the achievable differential drag. Hence, a differential
| fD,max | m s−2 1.960·10−5 1.871·10−5 * 1.982·10−5 lift maneuver necessary for influencing the out-of-plane mo-
| fL,max | m s−2 6.900·10−6 6.945·10−6 7.592·10−6 tion cannot be performed without a significant loss in altitude.
| fD,L | m s−2 1.157·10−5 1.064·10−5 7.224·10−6 Generally, the design variations had more effect for specular
materials than for materials reflecting particles diffusely. How-
Table 6: Data for the reference satellite, design SpecOpt and Lift 2, αT = 0.00.
ever, the theoretical optimal design for a satellite with specular
Data obtained using PICLas, with * marked data obtained using ADBSat.
materials entails little practicability. A more practicable solu-
tion was presented, which still achieved an increased lifetime
small as possible as well and therefore it is set equal to the ref- of 200 %, increased differential drag by 4 % and differential
erence satellite with h MB = 10 cm. The new design (SpecPrac) lift by 1 % compared to the reference satellite with the same
is depicted in Fig. 8i with a 2D profile optimized for αT = 0.00. surface properties. The same optimized design compared to the
It was obtained using the EPOT geometry B under considera- reference satellite with a diffuse re-emitting material achieved
tion of given wN and h MB . The panel is optimized for an in- an increase of 172 % in lifetime, 116 % in differential drag and
creased lifetime as well using geometry option B and a given 2133 % in differential lift. Therefore, the more promising ap-
total height of hmax = 30 cm. proach for realizing relative motion control in a satellite forma-
The according flow-fields for the reference and the design tion by differential lift and drag for high αT is an improvement
SpecPrac are given in Fig. 22. In Tab. 6, the corresponding of the currently available surface properties, as only for spec-
data is given. The lifetime decreasing effect of a frontal area in ularly reflecting materials the differential lift and drag are of
the case of the practicable design recommendation can be seen same order of magnitude.
in Fig. 22b, where the deflection of the impinging particles by However, the optimized 3D designs presented and discussed
180° leads to an accumulation of particles ahead of the satellite. throughout the article are derived from optimal 2D profiles and
However, the reduction of the total frontal area compared to the thus, the direct optimization of the 3D bodies for possibly fur-
reference led to an increase in lifetime. In Fig. 22d the particles ther increasing the performance is currently still pending. Ad-
are reflected in a wider range with an (in the picture) downward ditionally, the combination of different surface properties in
shift due to the increased curvature of the main body’s side sur- one satellite design was not considered and represents a pos-
face. The overlapping region of particles reflected by the panels sible continuation of this work. Applying a diffusely reflecting
and particles reflected by the rearward part of the side surface frontal area and specular side surfaces for a combination of sus-
can be seen in the range around −0.5 m < y < 0.3 m. In this tainable application in orbit and increased forces for the use in
case, the non-symmetrical deflection in the maximum drag con- relative motion control within the formation is one example of
figuration leads to a lift force, which is comparatively small, but the possibilities for mixed surface properties. The improvement
nonetheless can build up to a considerable effect over time. As of the materials is the most influential design parameter in or-
the main body’s frontal surface area is smaller than the refer- der to make the sustainable implementation of thrust-free con-
ence’s, it was possible to increase the lift force (see Fig. 22f). trol methods possible and represents an important step towards
future space systems. Hence, advancing material research is of
great interest for the application of VLEO satellites and satellite
formations.
6. Conclusion
Appendix A. Comparison of ADBSat and PICLas results
In this work, the design of a given reference satellite has been
optimized with regard to the use of differential lift and drag con- In this section, the results obtained using ADBSat and PI-
trol methods and a sustainable application in orbit. The influ- CLas for the recommended designs presented in the previous
ence of design variations was generally evaluated using ADB- sections are compared. Figure 8 gives an overview over the
Sat, and the performance of the promising designs was verified designs whose results are listed in Fig. A.23.
using the DSMC code PICLas. It can be seen, that the results of ADBSat generally comply
For materials with diffuse re-emission, the design optimized with the ones obtained using PICLas. The biggest difference
for differential drag and the design optimized for differential is visible for design MultRef 1 and 2 for the minimum drag.
lift did not share many geometric characteristics. Hence, when However, design MultRef 1 and 2 are the designs created in
deriving a design optimized for both differential lift and drag, order to promote multiple reflections. Here, the minimum drag
a priority of either lift or drag has to be set. However, for all obtained using PICLas is 25 % higher for design MultRef 1 and
given surface properties, an optimization with regard to the use 12 % higher for design MultRef 2 than estimated with ADBSat.
16
 n [m−3 ] ·1014
0.8 0.8

0.6 0.6 6

0.4 0.4
4
0.2 0.2
y [m]

y [m]
2
0 0

-0.2 -0.2 0
-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.6 -0.3 0 0.3 0.6 0.9 1.2 -0.6 -0.3 0 0.3 0.6 0.9 1.2
x [m] x [m]
(a) Reference satellite - nominal flight configuration (b) Practical design SpecPrac - nominal flight configuration
n [m−3 ] ·1014
0.8 0.8

0.6 0.6 6

0.4 0.4
4
0.2 0.2
y [m]

y [m]

2
0 0

-0.2 -0.2 0
-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.6 -0.3 0 0.3 0.6 0.9 1.2 -0.6 -0.3 0 0.3 0.6 0.9 1.2
x [m] x [m]
(c) Reference satellite - maximum drag configuration (d) Practical design SpecPrac - maximum drag configuration
n [m−3 ] ·1014
0.8 0.8

0.6 0.6 6

0.4 0.4
4
0.2 0.2
y [m]

y [m]

2
0 0

-0.2 -0.2 0
-0.4 -0.4

-0.6 -0.6

-0.8 -0.8
-0.6 -0.3 0 0.3 0.6 0.9 1.2 -0.6 -0.3 0 0.3 0.6 0.9 1.2
x [m] x [m]
(e) Reference satellite - maximum lift configuration (f) Practical design SpecPrac - maximum lift configuration

Figure 22: Simulated particle density n in the flow-field around the optimal design SpecOpt and the more practicable version SpecPrac for αT = 0.00.

17
 ⋆ ADBSat
−5
⋆ PICLas
10
⋆ fD,min
□ fD,max
+ fD,L
| f | [m s−2 ]

⋄ fL,max
10−6 Multiple Reflections

10−7
αT = 1.00 αT = 0.00
r1

r2

ac
ce

ce

2
Op
ef

ef

ft

ft
Pr
en

en
ste

ste

Li

Li
ec
tR

tR

ec
fer

fer
Ra

Ra

Sp
ul

ul

Sp
Re

Re
M

Figure A.23: Comparison of ADBSat and PICLas results.

If only geometries suitable for the panel method are regarded, The normal and tangenatial energy accommodation coefficient
the values for the specific forces obtained using PICLas are gen- αn and αt hence can be defined as
erally smaller than the ones calculated with ADBSat. However,
p2i − p2r
the biggest difference is only a 4 % smaller maximum drag αn = , (B.3)
of PICLas compared to ADBSat. The generally smaller spe- p2i − p2w
cific forces within PICLas can be explained due to small dif-
τ2i − τ2r
ferences in the edges of the models. The surface mesh used αt = . (B.4)
within ADBSat is better suited to model complex surfaces as τ2i − τ2w
the triangular shape of the mesh elements allows more narrow The parameter τi and pi as properties of the impinging particle
and sharp edges. On the other side, sharp edges or small radii depend on the properties of the atmosphere and are proportional
within the mesh for PICLas can pose a problem for the 3D mesh to the following terms [20]:
generation. Here, rounded and sharp edges are approximated
1 −γ2 s2
" #
with a greater radius, as a smaller radius increases the cells and τi ∝  γ(1 + erf(γs)) + √ e , (B.5)
thereby the necessary amount of simulated particles and thus s π
simulation time. Altogether it can be said, that for geometries
1 1
" #
2 2
without multiple reflections the results of ADBSat and PICLas pi ∝ γ γ(1 + erf(γs)) + √ e−γ s + 2 (1 + erf(γs)). (B.6)
do not differ more than 4 % and the differences can be traced s π 2s
back to differences in the mesh. τw is per definition equal to zero, and the normal momentum pw
carried away after reaching thermal equilibrium with the wall is
Appendix B. Derivation of the ADBSat input parameters proportional to
" √
1 Tw γ π 1 −γ2 s2
r #
For the application of the CLL model in ADBSat, the normal pw ∝ (1 + erf(γs)) + 2 e , (B.7)
energy accommodation coefficient αn and the tangential mo- 2 Ti s s
mentum accommodation coefficient σt are required and there- where T w is the wall temperature of the satellite and T i is the
fore are derived from the given αT in Tab. 3. The following temperature of the incident particles [20]. Finally, the momen-
derivation is based on transformations of Sentman’s equation tum components of the reflected particle have to be expressed.
by Hild [20] and the consideration of a 2D profile. Therefore, Eq. B.1 can be transformed into
Since the kinetic energy E = 12 mv2 is proportional to the
squared momentum I 2 = m2 v2 under the assumption of constant
q
Ir = Ii2 − αT (Ii2 − Iw2 ). (B.8)
mass, Eq. 4 can be also expressed as
Ii2 − Ir2 Using Eq. B.2 on Ii leads to
αT = , (B.1)
Ii2 − Iw
2
q
Ii = p2i + τ2i , (B.9)
with the momentum I consisting of its normal and tangential
which now can be used to express Eq. B.8 as
components p and τ
q
I 2 = p2 + τ2 . (B.2) Ir = p2i + τ2i − αT (p2i + τ2i − p2w ). (B.10)
18
 αT [-] αn [-] σt [-] [3] C. Traub, F. Romano, T. Binder, A. Boxberger, G. H. Herdrich, S. Fa-
0 0 0 soulas, P. C. E. Roberts, K. Smith, S. Edmondson, S. Haigh, et al., On
0.09 0.09 0.0459 the exploitation of differential aerodynamic lift and drag as a means to
control satellite formation flight, CEAS Space Journal 12 (2020) 15–32.
0.30 0.30 0.1627 [4] C. Traub, S. Fasoulas, G. H. Herdrich, A planning tool for optimal three-
dimensional formation flight maneuvers of satellites in vleo using aero-
Table B.7: Derived CLL input parameters according to the specified αT grada- dynamic lift and drag via yaw angle deviations, Acta Astronautica 198
tion in Tab. 3. (2022) 135–151. doi:10.1016/j.actaastro.2022.04.010.
[5] F. Hild, C. Traub, M. Pfeiffer, J. Beyer, S. Fasoulas, Optimisation of satel-
lite geometries in very low earth orbits for drag minimisation and lifetime
Inserting the above derived parameters into the following equa- extension, Acta Astronautica 201 (2022) 340–352. doi:10.1016/j.
actaastro.2022.09.032.
tions from Hild for τr and pr [20] [6] C. Marianowski, Satellite design considerations and analysis for differen-
tial lift and drag control methods, Master thesis, University of Stuttgart
Ir (2022).
τr = (1 − g)τi (B.11)
Ii [7] U. Walter, Astronautics: The Physics of Spaceflight, third edition Edition,
2018. doi:10.1007/978-3-319-74373-8.
Ir [8] P. C. E. Roberts, N. H. Crisp, F. Romano, G. H. Herdrich, V. T. A.
pr = g · Ir + (1 − g) · pi , (B.12) Oiko, S. Edmondson, S. J. Haigh, C. Huyton, S. Livadiotti, R. E. Lyons,
Ii
Discoverer-making commercial satellite operations in very low earth or-
the tangential and normal momentum carried away from the bit a reality, in: 70th International Astronautical Congress, 2019.
wall by the particle can be described using the parameters αT URL https://discovery.ucl.ac.uk/id/eprint/10120602
[9] N. H. Crisp, P. C. E. Roberts, S. Livadiotti, A. M. Rojas, V. T. A. Oiko,
and g: S. Edmondson, S. J. Haigh, B. E. A. Holmes, L. A. Sinpetru, K. L. Smith,
s In-orbit aerodynamic coefficient measurements using soar (satellite for
p2i + τ2i − αT (p2i + τ2i − p2w ) orbital aerodynamics research), Acta Astronautica 180 (2021) 85–99.
τr = (1 − g)τi , (B.13) [10] K. Moe, M. M. Moe, Gas-surface interactions and satellite drag coef-
p2i + τ2i ficients, Planetary and Space Science 53 (8) (2005) 793–801. doi:
10.1016/j.pss.2005.03.005.
[11] B. M. Pilinski, M. D.; Argrow, S. E. Palo, Semiempirical model for satel-
q lite energy-accommodation coefficients, Journal of Spacecraft and Rock-
pr = p2i + τ2i − αT (p2i + τ2i − p2w ) ets 47 (6) (2010) 951–956. doi:10.2514/1.49330.
  [12] P. Walker, A.; Mehta, J. Koller, Drag coefficient model using the
cercignani–lampis–lord gas–surface interaction model, Journal of Space-
pi
 
· g + (1 − g) q  , (B.14) craft and Rockets 51 (5) (2014) 1544–1563. doi:10.2514/1.A32677.
 
[13] P. A. Chambre, S. A. Schaaf, Flow of Rarefied Gases, Princeton legacy
 p2i + τ2i 

library, Princeton University Press, 1958.
[14] E. Doornbos, Thermospheric Density and Wind Determination from
and thus, αn and σt can be obtained utilizing Eq. 5 and Eq. B.3 Satellite Dynamics, Springer Theses, Springer Berlin Heidelberg, Berlin,
for given αT and g. However, it should be noted that τi , pi and Heidelberg, 2012.
[15] J. M. Picone, A. E. Hedin, D. P. Drob, A. C. Aikin, Nrlmsise-00 empirical
pw are determined using Sentman’s equation in a 2D simplified model of the atmosphere: Statistical comparisons and scientific issues,
version, which is valid for one area element with a given orien- Journal of Geophysical Research: Space Physics 107 (A12) (2002) SIA
tation regarding the velocity vector. Hence, the values for τi , pi 15–1–SIA 15–16. doi:10.1029/2002JA009430.
and pw as well as τr and pr change with different orientation of [16] J. V. Llop, Spacecraft flight in the atmosphere, Ph.D. thesis, Cranfield
University (2014).
the considered area element. Assuming a constant energy and [17] H. Heidt, J. Puig-Suari, A. Moore, S. Nakasuka, R. Twiggs, Cubesat: A
momentum accommodation over the satellite, the respective pa- new generation of picosatellite for education and industry low-cost space
rameters may still be used on one representative area element experimentation, Proceedings of the Small Satellite Conference Technical
in order to derive the normal energy accommodation coefficient Session V: Lessons Learned - In Success and Failure (SSC00-V-5) (2000).
[18] C. Pardini, L. Anselmo, K. Moe, M. M. Moe, Drag and energy accom-
and tangential momentum accommodation coefficient. modation coefficients during sunspot maximum, Advances in Space Re-
The derived input parameters for the use of the CLL model search 45 (5) (2010) 638–650. doi:10.1016/j.asr.2009.08.034.
according to the given αT in Tab. 3 can be found in Tab. B.7 [19] C. Traub, G. H. Herdrich, S. Fasoulas, Influence of energy accommoda-
and were used for all the CLL ADBSat calculations within this tion on a robust spacecraft rendezvous maneuver using differential aero-
dynamic forces, CEAS Space Journal 12 (2020) 43–63.
work. [20] F. Hild, Adaptation and application of piclas for the design of vleo satel-
lite geometries, Master thesis, University of Stuttgart (2021).
[21] L. A. Sinpetru, N. H. Crisp, D. Mostaza-Prieto, S. Livadiotti, P. C. E.
References Roberts, Adbsat: Methodology of a novel panel method tool for aero-
dynamic analysis of satellites, Computer Physics Communications 275
[1] N. H. Crisp, P. C. E. Roberts, S. Livadiotti, V. T. A. Oiko, S. Edmond- (2022) 108326. doi:10.1016/j.cpc.2022.108326.
son, S. J. Haigh, C. Huyton, L. A. Sinpetru, K. L. Smith, S. D. Worrall, [22] S. Fasoulas, C.-D. Munz, M. Pfeiffer, J. Beyer, T. Binder, S. Cop-
J. Becedas, R. M. Domı́nguez, D. González, V. Hanessian, A. Mølgaard, plestone, A. Mirza, P. Nizenkov, P. Ortwein, W. Reschke, Combining
J. Nielsen, M. Bisgaard, Y.-A. Chan, S. Fasoulas, G. H. Herdrich, F. Ro- particle-in-cell and direct simulation monte carlo for the simulation of
mano, C. Traub, D. Garcia-Almiñana, S. Rodriguez-Donaire, M. Sureda, reactive plasma flows, Physics of Fluids 31 (7) (2019) 072006. doi:
D. Kataria, R. Outlaw, B. Belkouchi, A. Conte, J. S. Perez, R. Vil- 10.1063/1.5097638.
lain, B. Heißerer, A. Schwalber, The benefits of very low earth orbit for [23] G. A. Bird, Molecular gas dynamics and the direct simulation of gas
earth observation missions, Progress in Aerospace Sciences 117 (2020) flows, repr Edition, Vol. 42 of Oxford science publications, Clarendon
100619. doi:10.1016/j.paerosci.2020.100619. Press, Oxford, 1994.
[2] S. D’Amico, Autonomous formation flying in low earth orbit, Ph.D. thesis [24] L. H. Sentman, Free molecule flow theory and its application to the deter-
(2010).

19
 mination of aerodynamic forces: Technical Report, 1961.
URL https://apps.dtic.mil/sti/citations/ad0265409
[25] J. A. Walsh, L. Berthoud, Reducing spacecraft drag in very low earth or-
bit through shape optimisation, 7th European Conference for Aeronautics
and Aerospace Sciences (2017). doi:10.13009/EUCASS2017-449.

20

View publication stats