Applied Thermal Engineering 159 (2019) 113908

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

T
Analytical and numerical models for thermal related design of a new pico-
satellite
Mario Bonnicia, Pierluigi Molliconea, , Maurizio Fenechb, Marc Anthony Azzopardic
⁎

a
Department of Mechanical Engineering, Faculty of Engineering, University of Malta, Msida MSD 2080, Malta
b
Institute for Sustainable Energy, University of Malta, Triq il-Barrakki, Marsaxlokk MXK 1531, Malta
c
Department of Electronic Systems Engineering, Faculty of Engineering, University of Malta, Msida MSD 2080, Malta

HIG HL IG HTS

• Thermal response prediction of a pico-satellite is critical.

• ATwothermal modeling tool was developed and applied to a custom design pico-satellite.
• Maximumindependent modelling techniques were used and the two models agreed.
• temperatures can be controlled by adjusting thermal surface finish.
• Minimum temperatures are harder to control.

A R TICL E INFO A BSTR A CT

Keywords: Modelling, simulating and testing the on-orbit thermal behaviour of a pico-satellite before launch is critical to
PocketQube ensure component reliability and performance. Due to small thermal inertia and hence small time constant, a
Low Earth orbit thermal environment pico-satellite may undergo larger temperature swings than those experienced by larger satellites. Weight, size
Passive thermal control and power limitations are especially restrictive in pico-satellites and limit thermal control to passive systems
Lumped parameter thermal analysis
such as controlling conduction and radiation heat transfer paths. A lumped parameter thermal model was de-
Finite element analysis
veloped to study the on-orbit thermal response of the UoMBSat-1 PocketQube, Malta’s first pico-satellite project.
A numerical finite element model is presented in order to test and validate the thermal model and passive
thermal control. Small satellites are usually launched on rockets as piggyback satellites; therefore, orbit para-
meters are rarely known in the early stages of the project and various launch opportunities would need to be
evaluated and compared. This paper presents a parametric analysis where the effects of orbit parameters, such as
altitude and beta angle, on the thermal response are evaluated. We show that by controlling the surface finish
and beta angle, it is possible to place a pico-satellite in a thermal regime that is suitable for typical electronic
components, and batteries.

1. Introduction by studying the properties of the ionised region of Earth’s upper at-
mosphere (300–500 km altitude) and its real-time effects on radio
UoMBSat-1 [1] is a PocketQube (PQ)1 pico-satellite developed as a communications.
joint collaboration between the University of Malta and the University The project phases of a small satellite can be outlined as: concep-
of Birmingham. The PocketQube ‘standard’ traces its origins to a 2009 tion, technology development, design, fabrication, system assembly,
Cal Poly Developers Workshop proposal by Prof Robert Twiggs [2] and testing, launch and operations [4,5]. However, PQs are at the far end of
the details were further defined [3] by interested parties only recently. the scale in terms of size (an order of magnitude smaller than the
UoMBSat-1 will be launched into a Sun-synchronous orbit (SSO) and popular CubeSat format [6]) and present a new set of challenges that
will demonstrate the applicability of pico-satellites to scientific missions have yet to be comprehensively addressed within the spirit of a low cost

Corresponding author at: Engineering Building, Room 303, University of Malta, Msida MSD 2080, Malta.
⁎

E-mail addresses: [email protected] (M. Bonnici), [email protected] (P. Mollicone), [email protected] (M. Fenech),
[email protected] (M.A. Azzopardi).
1
See Nomenclature at end of paper

https://doi.org/10.1016/j.applthermaleng.2019.113908

Available online 10 June 2019

Received 3 October 2018; Received in revised form 25 April 2019; Accepted 2 June 2019

1359-4311/ © 2019 Elsevier Ltd. All rights reserved.

M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

spacecraft. Thermal modeling runs through all phases and thermal

control (TC) can be a driver force when determining a pico-satellite’s
configuration, location of components, orbital parameters, attitude and
power requirements. The design costs of UoMBSat-1 are kept low by
using commercial off-the-shelf (COTS) components. Due to size and
weight limitations, redundancy is restricted to few cases. Similarly,
complex thermal control systems (TCS) such as those proposed for
larger spacecraft [7] are difficult to accommodate. Moreover, the power
budget of miniature satellites limits TC to passive systems. Reliability
can be increased by extensive modeling of the satellite’s on-orbit
thermal behavior. Therefore, thermal modeling of miniature satellites is
a research area of increasing importance [1,8,9,11,12].
Thermal analysis software dedicated to spacecraft allows the
transformation from detailed to reduced thermal mathematical models.
This procedure was employed by Gorlani et al. [10] where a detailed
Finite Difference (FD) lumped parameter model was used for design
optimisation. A reduction algorithm was then used to simplify the
thermal model making analysis at higher orders more efficient while Fig. 1. 3D CAD model of UoMBSat-1.
retaining accuracy. In the last few years, the popularity of lumped
parameter models for thermal analyses of satellites has increased due to 2. Methodology
simplicity and agility. The number of nodes in a thermal model depends
on the objectives of the analysis. Grande et al. [11] investigated how UoMBSat-1 (Fig. 1) measures 5 cm on each side and weighs 250 g.
solar absorptivity affects the general thermal response of a miniature The basis of the structure is an aluminum 7075-T6 chassis to which five
satellite; therefore, the satellite was represented using an external and solar panels are attached using a thermal adhesive. The chassis is
internal node, coupled by conduction and radiation. Corpino et al. [13] mounted on a base. Six printed circuit boards (PCBs), are inserted in
used fifteen nodes to evaluate the thermal response of the components corner slots machined in the chassis. Two cylindrical lithium-ion
on board PiCPoT. polymer batteries store excess power that is generated during sunlight.
An issue present in most satellite thermal analysis literature is the Attitude control is achieved using three 35.4 mm diameter reaction
selection of conductive and radiative parameters for the thermal net- wheels made of aluminum 7075-T6 oriented in the x, y and z axes.
work. Liu et al. [14] noted that simplifications in the mathematical These are connected to dc micro-motors attached to the ADCS (attitude
model and inaccuracies when selecting network coefficient parameters determination and control system) PCB. An internal heat dissipation
resulted in discrepancies between simulated and experimental results. profile was not used in this study. While the efficiency of the solar cells
This problem was resolved by adjusting the mathematical model until was taken into consideration when selecting thermo-optical properties
the numerical and experimental results agreed. for the solar panels, the internal heat generated by the battery was not
The objective of this paper is to provide an accurate and flexible tool considered. This, therefore, represents the worst-case cold scenario. In
which simplifies the thermal modeling process of UoMBSat-1. An ana- practice, waste heat such as that generated by the ADCS motors and
lytical thermal modeling tool which satisfies the requirements of small other devices can be used to raise the temperature of the most thermally
satellite programmes is implemented and validated. A small-satellite sensitive components - the batteries.
thermal modeling tool should give sufficiently accurate results, despite
having a limited level of detail, while reducing the time required to run
and maintain the model. Literature regarding the effects of orbit 2.1. Thermal control system (TCS) and thermal environment
parameters and surface finishes on satellites with small thermal inertia
is scant. Due to lower budgets, experiments on small satellites are not The TCS of a spacecraft ensures that the temperatures of all com-
readily available and it is often impractical to simulate the space en- ponents remain within predetermined operating and survival limits.
vironment in an adequately representative test environment. Therefore, The operating limit sets the temperature range in which a component
this paper serves to corroborate the work done by other researchers, should operate. The survival limit sets the temperature range outside
such as Kovács et al. [8] with the scope of increasing the reliability of which a component becomes inoperable. Since most of UoMBSat-1’s
lumped parameter modelling techniques to thermal analyses of pico- components are COTS their operating temperature limits are de-
satellites. Kovács et al. [8] simulated the thermal response of SMOG-1 termined by the industrial standard of −40 °C to 80 °C. The lithium
by applying a thermal network and finite element (FE) model. While polymer (LiPo) batteries have the narrowest operating temperature
the models employed by Kovács et al. [8] assumed constant thermal range. For longevity considerations, the battery operating range is
environment parameters, this paper uses the Simple Thermal Environ- narrowed down to −10 to 30 °C during discharge and 0 to 25 °C during
ment Model (STEM) to determine appropriate design values which charging. In SMOG-1, a restricted battery charging protocol [8] is
depend on both the orbit’s beta angle and the satellite’s altitude. A proposed to extend battery life. In UoMBSat-1, thermal control is
novelty of this paper is that it studies the effects of orbit parameters on achieved passively by controlling conduction and radiation heat
pico-satellites’ thermal response. This is important, especially for small transfer paths by configuring the geometry and position of components
satellites, since these are normally launched as piggyback satellites on as required, insulating components and selecting appropriate thermal
rockets carrying other satellites. Therefore, orbital parameters depend coatings [1,9], and if available, by selecting a launch opportunity that
on the launch vehicle provider and the primary payload and are not will place the satellite in a high beta angle SSO.
always known in the early stages of the project. Validation of the The environmental heat fluxes incident on an Earth-orbiting sa-
lumped parameter model is carried out using a numerical thermal finite tellite depend on the position of the satellite with respect to the Sun and
element method (FEM) approach. Finally, the effectiveness of thermal Earth. The beta angle, marked β in Fig. 2, is the smallest angle between
surface finishes to control the temperature variations of pico-satellites is the Sun’s position vector and the orbit plane. When β is equal to 0°, the
evaluated. satellite passes over the subsolar point (Sun is directly overhead) and
has the longest eclipse fraction. When β is equal to 90°, the satellite is
always close to the terminator and the eclipse fraction is zero [15].

2
M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

Fig. 2. The orbit beta angle measures the minimum angle between the orbit plane and Sun’s position vector (which in this case is [RA = 0°, Dec = 0°]).

The right ascension of ascending node (RAAN), also marked in Direct solar radiation is also reflected off the Earth/atmosphere as
Fig. 2, measures the point on the equator, from the vernal equinox albedo. Albedo heat flux, qa , is calculated by [17]
[RA = 0°, Dec = 0°], where the spacecraft crosses from south to north.
Due to Earth’s equatorial bulge, the RAAN drifts with each orbit. This Gs a s Fe cos , if < 2
qa =
perturbation is exploited to place satellites in SSOs. For a SSO, the beta 0, if 2 (2)
angle is approximately constant since the orbital plane rotates east-
wards with the angular velocity of the Earth around the Sun (0.985° per where a is the albedo factor, Fe is a view factor of the surface relative
day). This can only be achieved by retrograde orbits (inclinations to Earth and is the angle between the solar rays and the surface’s
greater than 90°) [15,16]. position vector. The view factor depends on the satellite’s altitude and
A satellite in low Earth orbit (LEO) has three main sources of en- angle between the normal to the external surface and the surface
vironmental radiation. Direct solar irradiation is incident on the sur- position vector [18].
faces which are directed towards the Sun. The heat flux, qs , is calculated Earth-emitted outgoing longwave radiation (OLR) is incident on the
by [17], satellite’s external surfaces which face Earth. This heat flux is calculated
by [17]
qs = Gs s cos , (1)
qe = Fe IR OLR, (3)
where the solar constant, Gs , is equal to 1367 W/m2, s is the absorp-
tivity in the solar spectrum and is the angle between the normal to the where IR is the absorptivity of the surface in the infrared spectrum and
external surface and the solar rays (Fig. 3). OLR is the outgoing longwave radiation.

Fig. 3. Spacecraft in orbit showing the relevant orientations.

3
M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

The total environmental heat flux incident on a surface is glass and solar cells [24].
The components were represented using diffusion nodes where the
qext = qs + qa + qe . (4)
temperature changes according to the heat flowing into or out of the
Depending on its orbital beta angle, a satellite may spend a fraction node. Each component was assumed to be isothermal. Space was re-
of the orbital period in eclipse. A surface in eclipse will not receive presented using a boundary node where the temperature remained
direct solar irradiation or albedo radiation but may still receive Earth- constant, independent of heat flow into the node. The nodes were then
emitted longwave infrared radiation if oriented towards Earth [19]. connected using conductors to enable heat transfer by conduction and
The selection of a and OLR in (2) and (3) respectively is based on radiation. This was done by specifying 17 × 17 contact conductance
the NASA/MSFC Simple Thermal Environment Model (STEM). These and radiation network matrices. This completed the thermal network
values are derived from the Earth Radiation Budget Experiment (ERBE). model of UoMBSat-1 [9,25].
Highest OLR values occur where the satellite views cloud-free parts of The view factors between surfaces were calculated by considering
the Earth. Highest albedo factor values occur over cloud covered por- the satellite to be subdivided into a number of separate enclosures. For
tions of Earth or cloud-free areas with ice and snow cover. When se- simple enclosures, analytical equations were used. For more compli-
lecting a hot or cold extreme environment for a thermal analysis, a cated enclosures, such as the one containing the batteries, a Monte
conceptual problem arises since high OLR is associated with low albedo Carlo technique was implemented. These view factors were then com-
factors while high albedo factors are associated with low OLR (e.g. bined into a 17 × 17 matrix as defined in Equation (5) and selected
albedo factor increases with latitude but OLR decreases). Therefore, elements with significant values are shown in Table 2. The radiation
choosing a peak albedo factor and peak OLR results in a combination view factors were then checked to ensure that the summation and re-
far outside the observed data region. This problem is addressed by ciprocity properties were satisfied.
choosing an extreme albedo, OLR or combined environment, depending
on which extreme type yields the maximum (hot extreme) or minimum 1,1 1,17
(cold extreme) orbit-averaged environmental heat load [20,21]. = ,
17,1 17,17 (5)
2.2. Lumped parameter approach
The general heat transfer equation was discretised using the finite
The lumped parameter model of UoMBSat-1 was developed by re- difference method (FDM). The forward-explicit equation is [25],
ducing the surfaces and components of UoMBSat-1 to a set of 17 nodes.
n n
Each node is described by a single temperature since it is assumed that t
Ti' = Ti + Kji (Tj Ti ) + 4
ij Ai (T j Ti4 ) + Qi space + Qi ext + Qi int ,
each component will have a uniform temperature. Fig. 4 shows the Ci j =1 j =1
node numbering of components and the satellite’s local coordinate
(6)
system. Table 1 gives the thermal and surface-optical properties used to
determine the conductive and radiative couplings in the lumped para- where n is the total number of nodes, K is the absolute contact con-
meter model. The first six nodes are external and exposed to space ductance matrix, is the Stefan-Boltzmann constant, is the grey-body
environment. The remaining nodes are fully shielded from space en- shape factor matrix, Ai is the surface area of node i , Ti and Tj are the
vironment by the external nodes [22]. temperatures of nodes i and j at the current time, t is the time step, Ci
The aluminium chassis was not modelled using separate nodes; is the capacitance of node i , Ti ' is the temperature at the next time step,
however, mass-averaged thermal properties were used to combine the Qi space is the radiation from space to node i , Qi ext is the environmental
thermal properties of the chassis and solar panel into one node. The radiation absorbed by node i and Qi int is the internal heat generation of
total heat absorbed/emitted by each external surface was found by node i .
adding the heat absorbed/emitted by the solar panels and the heat The approximation of first and second partial derivatives in the heat
absorbed/emitted by the chassis. The longwave emissivity of the in- equation using the forward and central difference schemes causes
ternal PCBs was chosen assuming a matte copper surface without solder truncation errors. Errors also affect the stability of the solution. If the
mask. This prevents the build-up of static electricity [23]. The optical error is bounded as the solution progresses, then, the solution is stable.
properties of the solar panel assembly were determined by the cover However, if the error grows with time, the solution becomes unstable

Fig. 4. Geometric model of UoMBSat-1 and node numbering of external and internal components.

4
M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

Table 1
Node numbering, and thermal and surface optical properties of components [23,24,26].
Node Comp. ρ [kg/m3] cp [J/kgK] k [W/mK] αs εIR

External 1, 3, 4, 5, 6 Surface 1840 1150 108 0.62 (solar panel), 0.86 (solar panel)
0.15 (chassis) 0.05 (chassis)
2 Base 1840 1150 108 0.25 0.22

Internal 7, 8, 9, 10, 11, 12 PCB 1840 1150 108 n/a 0.22

13, 14 Battery 2118 1207 36 n/a 0.7
15, 16, 17 Reaction wheel 2180 960 130 n/a 0.05

Table 2 contact conductance values, based on literature, were used at the in-
12 of the most significant radiation view factors. terface between components: thermal contact conductance values
Radiating Node Receiving Node View Factor [unitless] across thermally conductive adhesives was obtained from Dow Corning
(12500 W/m2 °C) [28]; Mirmira and Fletcher [29] led a number of
Surface (3) PCB (7) 0.8101 studies with mixed surfaces contacts including contact of composites
Surface (5) PCB (12) 0.8016 with metals (200 W/m2 °C for e-glass-metal interference), and Fried and
PCB (7) PCB (8) 0.4365
Costello [30] studied this problem as it occurs in the space context.
PCB (11) PCB (10) 0.4051
Heat conduction models from Gomadam [31] were used to determine
PCB (11) PCB (12) 0.3879
Battery (14) PCB (8, 9) 0.3507 heat conduction in the batteries. Heat conduction models from Azar and
Battery (13) PCB (8, 9) 0.3481 Graebner [23] were used to obtain the normal and tangential thermal
Battery (14) Surface (4) 0.1230 conductivities of the PCBs.
Surface (4) PCB (9) 0.1199 The accuracy and efficiency of a numerical solution strongly depend
Base (2) PCB (12) 0.1178 on the number of elements and quality (mainly skewness, orthogonality
Surface (4) PCB (8) 0.1125 and Jacobian ratio) of the mesh. A structured grid is very space efficient
Battery (13) Battery (14) 0.0936
and therefore the geometry was sectioned to obtain, where possible,
mappable bodies. The mappable bodies were then meshed using hex-
ahedral elements (Fig. 5) [32,33].
and diverges [25].
A mesh independence test was carried out using four mesh sizes.
Eq. (7) is forward-explicit since temperatures at the next time step
The maximum and minimum temperatures, and the solution time were
depend on values from the previous time step. This system of equations
monitored for the different mesh cases. Fig. 6 shows that by decreasing
is simple to solve since each equation has only one unknown tem-
the mesh size from fine to superfine, the temperature difference was
perature, thus, avoiding the need of an iterative or matrix inversion
only 0.4%. However, solution time was negatively affected and in-
solution. However, if t is too large, the error will not be bounded and
creased by 300%.
the solution will be unstable. The stability criteria for each node i is,
The heat flux load history curves (Fig. 7) were then applied on the
1 FE model. All bodies started at a uniform temperature and a stabilized
ti 1
,
j Rij Ci (7) temperature response was obtained after a number of orbits. During
each orbit, the satellite passed through sunlight and eclipse periods.
where j Rij is the total thermal resistance, due to conduction or ra- These regions were separated by a transition region. Since an implicit
diation, between nodes i and j . Although the thermal capacity and total scheme was used, the solution was unconditionally stable. Automatic
thermal resistance between each node is different, it is necessary to time stepping was used where a conservative initial integration time
advance every node in the thermal system by the same time-step. step (ITS) was specified and the software was allowed to increase or
Therefore, the time step is determined by the node that requires the decrease the ITS, within predefined limits, as required (Table 3). This
smallest time step [9],
t = min( t1, t2, , tn). (8)
Since t is limited to very small values, a forward-explicit scheme is
only viable when the number of nodes is small. For a larger number of
nodes, a backward-implicit scheme could be more suitable since a
larger time step is allowed [9].

2.3. Numerical approach

The results of the lumped parameter thermal model were compared

to results obtained using commercial FE software ANSYS®. The nu-
merical model served for benchmarking purposes and therefore it was
important to implement procedures such as mesh and time step in-
dependence tests, and heat energy conservation tests at steady state in
order to confirm reliability and accuracy of results [27].
The heat flux load history curves on the external surfaces of the
satellite were obtained using the theory outlined in Section 5. The
geometrical model of UoMBSat-1 was simplified by removing un-
necessary details such as the dc micro-motors attached to the EPS PCB
and the shafts connecting the reaction wheels to the micro-motors. The
reaction wheels were then connected to the EPS PCB using a thermal
contact pair (13000 W/m2 °C) across the gap. Appropriate thermal Fig. 5. FE discretization of UoMBSat-1’s components.

5
M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

Fig. 6. Mesh convergence study for steady-state case. The maximum (solid line) and minimum (dashed line) temperatures were monitored together with compu-
tational time (dotted line) for different mesh sizes.

Fig. 8. Time step independence test. The thermal response was monitored for
three different time steps.

were used when heat flux boundary conditions changed rapidly such as
during transition periods. However, ITS size was also conditioned by
truncation errors and the radiation-conduction solution. It was more
difficult to achieve convergence when using larger time steps since
Fig. 7. Example of a load history curve and thermal response.
temperatures changed more rapidly. Therefore, a time step in-
dependence test was carried out to ensure solution convergence as ITS
Table 3 decreased. Both time step 2 and time step 3 gave approximately the
Time stepping parameters used in time step independence test. same temperature response (Fig. 8). However, solution times varied
ITS 1 [s] ITS 2 [s] ITS 3 [s] substantially, with time step 2 taking 7 h and time step 3 taking 56 h
[34,35].
In sunlight Initial 100 0.5 0.1
Minimum 50 0.1 0.05
Maximum 1000 50 5 3. Results and discussion
Transition Initial 45 0.5 0.1
Minimum 50 0.1 0.05
3.1. Validation test
Maximum 100 10 1

Eclipse Initial 100 0.5 0.1

The validation case study presented is based on a satellite at a
Minimum 50 0.1 0.05 500 km SSO with an inclination of 97.4° (retrograde orbit) and a beta
Maximum 100 50 5 angle of 45°. The satellite’s orientation was fixed in inertial space by
setting the local + Z axis Fig. 4 to point along the direction vector
[RA = 45°, Dec = −35°]. The thermal and surface-optical parameters
avoided excessive CPU times without compromising accuracy. Larger are given in Table 2.
time steps were used when the heat flux boundary conditions changed The satellite spent 32% of its orbital period in eclipse. The orbit-
slowly, such as during sunlight and eclipse periods. Smaller time steps averaged albedo factor and OLR were 0.212 and 279 W/m2

6
M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

Fig. 11. Maximum and minimum temperatures of UoMBSat-1 as a function of

beta angle. In case 1, the satellite had no external surface finish. In case 2,
external surfaces were coated with a solar absorber coating.

Fig. 9. Heat fluxes absorbed by UoMBSat-1’s external surfaces in the analytical

period. The analytical model predicted the thermal response of com-
model (Fig. 4). ponents to within 2 °C of the numerical model. The temperature ranges
in the analytical model tended to be narrower due to approximations
and simplifications when calculating the absolute contact conductance
respectively. Both values were low since albedo reflection is more
values between components. The batteries experienced large tempera-
Lambertian (equal in all directions) at beta angles of 45° and the sa-
ture swings. By modeling insulation around the batteries, the new peak
tellite spent more time over cooler polar regions due to its high in-
battery temperatures (1.4 °C and 14.5 °C) turned out to be within op-
clination [20].
erating limits. An important requirement is that both batteries have the
Fig. 9 shows the environmental heat loads absorbed by the satellite’s
same temperature so as to charge and discharge at the same rate.
external surfaces, as computed by the analytical model (Fig. 4) during
one orbit. Surface 6, which was oriented towards the Sun during sun-
3.2. Parametric analyses
light periods, received the largest amount of direct solar irradiation.
The opposite face, surface 4, was oriented towards the Earth and
Changing the beta angle, led to different orbit plane orientations
therefore, was exposed to weaker albedo and Earth-emitted infrared
with respect to the Sun and Earth, resulting in different temperature
radiation. In eclipse, the surfaces oriented towards Earth (surfaces 1
ranges (Fig. 11). UoMBSat-1 was placed in SSO. The attitude was kept
and 3) continued to receive some Earth-infrared radiation.
fixed relative to the orbit plane, with the primary axis (+Z axis) always
The thermal responses computed analytically agreed with numerical
contained in the orbit plane and pointing along a direction vector with
results. Fig. 10 shows temperature range plots for the major compo-
a declination of −35° (Fig. 2). Peak maximum temperatures occurred at
nents on board UoMBSat-1 for one orbit. The numerical simulations
beta angles close to 30° and 150° where solar irradiation was dis-
confirmed the assumption that each component will have an isothermal
tributed better on the external surfaces of the satellite. At subsolar beta
temperature.
angles (0° and 180°), the maximum temperature decreased since solar
Components that were not connected directly to the chassis, such as
irradiation was incident on at most two surfaces (given the fixed re-
reaction wheels, experienced lower temperature oscillations with an
lative attitude) and albedo heat fluxes were weaker due to a more
out-of-phase thermal response. This response agrees with data in
Lambertian reflection. The minimum temperature of the pico-satellite
Corpino [13], Grande [11] and Fortescue [36]. Maximum and
largely depended on eclipse fraction. Therefore, minimum temperatures
minimum temperatures occurred during late sunlight and eclipse per-
occurred for satellites orbiting at beta angles of 0° and 180° where
iods and the period of the thermal response was equal to the orbital
eclipse periods are longest (38% of orbital period for a 500 km SSO).

Fig. 10. Analytical and numerical thermal responses of UoMBSat-1.

7
M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

Satellites at a 500 km SSO and beta angles between 68° and 112° do not satellite, maximum temperature did not vary with altitude.
have eclipse periods. Fig. 11 shows that these satellites did not ex- At orbit beta angles smaller than β* (marked by vertical dotted lines
perience large temperature oscillations since solar and albedo irradia- on Fig. 12 for different altitudes), the satellite passed through sunlight
tion were almost constant throughout the whole orbit. and eclipse periods. Satellites orbiting at higher altitudes had shorter
The lowest maximum temperatures occurred for satellites orbiting eclipse fractions and therefore, spent a lower percentage of their orbital
above the terminator (β = 90°). Although albedo reflectivity was period in Earth’s shadow. For example, at a beta angle of 65°, satellites
highest, the albedo heat fluxes incident on the pico-satellite’s surfaces orbiting at 500 km and 300 km had eclipse fractions of 17% and 26%
were negligible. Moreover, only two external surfaces received direct respectively. As a result, satellites at higher altitudes had higher
solar irradiation and Earth-emitted IR while the rest of the surfaces minimum temperatures.
acted as heat sinks by radiating heat to cold space. As the beta angle approached zero, eclipse times increased and the
Fig. 11 also shows the effect of coating UoMBSat-1’s external sur- difference between eclipse fractions at different altitudes became less
faces (base and chassis) with a solar absorber finish (such as TiNOX) significant. At a beta angle of 0°, satellites orbiting at 500 km and
characterized by a high ratio of solar absorptivity to infrared emissivity 300 km had eclipse fractions of 38% and 40% respectively. Therefore,
(αs/εIR ≈ 20) (case 2). The maximum temperature increased sig- at small beta angles, altitude had a negligible effect on the minimum
nificantly, especially when the satellite’s base was oriented towards the temperature.
Sun (90° < β < 180°). The effect on the minimum temperature, which Satellites orbiting at beta angles greater than β* have no eclipse
occurs during eclipse, was less pronounced. Infrared (IR) energy period. However, as the beta angle approaches 90°, both minimum and
emitted by the spacecraft is of the same wavelength as Earth-emitted IR. maximum temperatures decrease considerably off their β* peaks. At the
Decreasing the infrared emissivity, εIR, of the external surfaces did not 100 km SSO, the minimum temperature drops by about 14 °C. This is
increase the minimum temperature since the reduction in heat radiated due to the change in satellite attitude with respect to the sun position
to cold space during eclipse periods, due to lower infrared emissivity, vector, ultimately resulting in just once face facing the sun at a beta
was counteracted by lower absorption of Earth-emitted IR due to lower angle of 90°. The minimum temperature (at the respective β*) was
infrared absorptivity. (It was assumed that the spectral form of higher for lower orbits since the amount of incident albedo and Earth-
Kirchhoff’s law applies; therefore, εIR = αIR and εs = αs. This is valid if emitted radiation increased with decreasing altitudes.
irradiation or surface is diffuse.) These analyses showed that whereas
the maximum temperature of a spacecraft can be adjusted easily by 4. Conclusion
applying a particular surface finish, it is much more difficult to pas-
sively control the minimum temperature. Predicting the thermal response of a pico-satellite to the on-orbit
Changing the altitude of the pico-satellite also gave different tem- thermal environment is critical to ensure reliability and performance.
perature results. Fig. 12 gives the maximum and minimum tempera- An assessment of the thermal behavior of UoMBSat-1 was presented in
tures of the satellite as a function of beta angle and altitude. The sa- order to increase on-orbit reliability and optimize the TCS. Results were
tellite will orbit at an altitude between 300 and 500 km. The 100 and in line with a similar previous study [8].
1000 km altitudes were included to clarify the effect of altitude on A thermal modeling tool suitable for the thermal design of small
temperature. As altitude increased, the amount of albedo longwave and satellites was developed and applied to UoMBSat-1. This tool calculated
Earth-emitted shortwave radiation incident on the satellite decreased. the heating loads on the pico-satellite’s external surfaces due to en-
Direct solar irradiation heat flux was independent of altitude since vironmental sources such as direct solar, shortwave albedo and long-
changes in altitude were negligible compared to the mean Earth-Sun wave Earth-emitted radiation. The main inputs were the orbital, atti-
distance of 1 astronomical unit (1.496 × 108 km) [11]. Since solar ir- tude and environmental parameters. Selection of environmental
radiation is the greatest source of radiation on an Earth-orbiting parameters was standardised using the Simple Thermal Environment
Model (STEM) where thermal environment design points were chosen
consistently based on ERBE observations. A lumped parameter re-
presentation of the satellite was obtained by reducing the components
to a set of nodes connected by a thermal network. The heat load
boundary conditions were then used to predict the thermal response.
A numerical thermal FE model of UoMBSat-1 was developed and
transient heat load history curves on the spacecraft's external surfaces
were calculated. The numerical model was used for benchmarking
purposes; therefore, procedures were implemented to ensure both mesh
and time step independence of the solution.
A number of case studies were planned and the analytical model
was validated by comparing numerical and analytical results. The two
models agreed to within 2 °C. Analyses showed that whilst the max-
imum temperature of UoMBSat-1 can be controlled easily by adjusting
the thermal surface finish of external surfaces, it is much more difficult
to control the minimum temperature. Parametric analyses were con-
ducted to study the effects of beta angle and altitude on UoMBSat-1’s
thermal response. The maximum temperature was determined by the
distribution of solar irradiation on the pico-satellite’s external surfaces.
The minimum temperature decreased as the time the satellite spent in
eclipse increased.
The results therefore suggest that very small satellites, such as the
UoMBSat-1, may consider launches towards SSO having a high beta
angle (close to β*). When coupled with a suitable thermal surface finish,
Fig. 12. Maximum and minimum temperatures as a function of orbit altitude. such orbits may be sufficient means for bringing the thermal operating
The vertical dotted lines mark β* (the beta angles above which no eclipses envelope into the required regime. Mission-permitting, the benefits of
occur). such a high beta SSO also include much reduced thermal cycling, as

8
M. Bonnici, et al. Applied Thermal Engineering 159 (2019) 113908

well as higher solar panel energy generation associated with the high for flight.
beta angle. The results also suggest that an attitude control system, if
available, will be able to significantly, and possibly sufficiently, actively Declaration of Competing Interest
control the mean temperature by varying the number of panels re-
ceiving direct solar radiation. Authors declare that there are no conflict of interest.
The UoMBSat-1 project is now entering its test phase. The lumped
parameter model developed in this paper can be useful when devel- Acknowledgements
oping test plans for thermal vacuum chamber testing. Since it is im-
practical to simulate the orbital environment, the model can be used to This research did not receive any specific grant from funding
design a testing environment that gives the same thermal response as agencies in the public, commercial, or not-for-profit sectors.

Appendix A. : Nomenclature

COTS Commercial off-the-shelf

ERBE Earth Radiation Budget Experiment
FE Finite Element
FD Finite Difference
FEM Finite Element Method
IR Infrared Radiation
ITS Integration Time Step
LEO Low Earth Orbit
OLR Outgoing Longwave Radiation
PQ PocketQube Satellite
RAAN Right Ascension of Ascending Node
SSO Sun-Synchronous Orbit
STEM Simple Thermal Environment Model
TC Thermal Control
TCS Thermal Control System
〈a〉 Albedo factor
qs Direct solar irradiation heat flux
qa Albedo radiation heat flux
qe Earth-emitted IR heat flux
αs Absorptivity in the solar spectrum
εIR Longwave infrared emissivity

References 1971.
[17] T. Sarafin, Spacecraft Structures and Mechanisms: from Concept to Launch,
Microcosm Press, 1995.
[1] D. Cachia, J. Camilleri, M.A. Azzopardi, M. Angling, A. Sammut, Feasibility Study of [18] J. Cheng, The Principles of Astronomical Telescope Design, Springer, 2009.
a PocketQube Platform to Host an Ionospheric Impedance Probe, Proceedings of the [19] D. Vallado, Fundamentals of Astrodynamics and Applications, McGraw Hill
4S Symposium: Small Satellites go Viral, Valletta, Malta, 30 May - 3 June, 2016. Companies Inc, 1997.
European Space Agency (ESA), (2016). [20] B.J. Anderson, C.G. Justus and G.W. Batts, Guidelines for the Selection of Near-
[2] R.J. Twiggs, Making it Small, Proceedings of the Cal Poly Developers’ Workshop, Earth Thermal Environment Parameters for Spacecraft Design, NASA/TM-2001-
California State Polytechnic University, San Luis Obispo, CA, 2009. xxxx, 2001.
[3] Alba Orbital, TuDelft, Gauss, The PocketQube Standard, Issue 1, 7 June, 2018, Link: [21] C.G. Justus, G.W. Batts, B.J. Anderson and B.F. James, Simple Thermal
https://dataverse.nl/api/access/datafile/11680, Accessed: September 2018. Environment Model (STEM), NASA, October 2001.
[4] W.J. Larson, J.R. Wertz, Space Mission Analysis and Design, Microcosm Press and [22] National Aeronautics and Space Administration (NASA), Guidelines for Thermal
Kluwer Academic Publishers, (1999). Analysis of Spacecraft Hardware, Guideline no. GD-AP-2302. NASA, United States
[5] B. Shiotani, N. Fitz-Coy, S. Asundi, An End-to-End Design and Development Life- of America.
Cycle for CubeSat class Satellites, AIAA SPACE 2014 Conference and Exposition, [23] K. Azar, J.E. Graebner, Experimental Determination of Thermal Conductivity of
(2014). Printed Wiring Boards, 12th Annual IEEE Semiconductor Thermal Measurement
[6] The CubeSat Program, California Polytechnic State University, CubeSat Design and Management Symposium, Austin, Texas, 5-7 March 1996. IEEE, (1996).
Specification, Rev. 13, http://www.cubesat.org/s/cds_rev13_final2.pdf, Accessed: [24] H.K. Kim, C.Y. Han, Analytical and numerical approaches of a solar array thermal
April 2019. analysis in a low-earth orbit satellite, Adv. Space Res. 46 (2010) 1427–1439.
[7] W. Guo, Y. Li, Y.Z. Li, M.L. Zhong, S.N. Wang, J.X. Wang, J.X. Zhang, A self-driven [25] A.J. Chapman, 1987, Fundamentals of Heat Transfer, Macmillian, 1987.
temperature and flow rate co-adjustment mechanism based on Shape-Memory- [26] M.S. Rad, D.L. Danilov, M. Kazemenini, P.H. Norton, Thermal modelling of cy-
Alloy (SMA) assembly for an adaptive thermal control coldplate module with on- lindrical lifepo4 batteries, J. Modern Phys. 4 (2013) 1–7.
orbit service characteristics, Appl. Therm. Eng. 114 (2017) 744–755. [27] ANSYS® Academic Research, Release 17.2.
[8] R. Kovács, V. Józsa, Thermal analysis of the SMOG-1 PocketQube satellite, Appl. [28] Dow Corning, Product Information, Product Selection Guide: Dow Corning TC-2030
Therm. Eng. 139 (2018) 506–513. Thermally Conductive Adhesive.
[9] D.G. Gilmore, Spacecraft Thermal Control Handbook: Fundamental Technologies, [29] S.R. Mirmira, L.S. Fletcher, Comparison of effective thermal conductivity and
vol. 1, 2nd ed., The Aerospace Press, 2002. contact conductance of fibrous composites, J. Thermophys. Heat Transf. 13 (2)
[10] M. Gorlani, M. Rosi, Thermal model reduction with stochastic optimisation, SAE (1999) 272–276.
Int. (2004) 120125. [30] E. Fried, F.A. Costello, Interface thermal contact resistance problem in space ve-
[11] I. Grande, A. Sanz-Andrs, C. Guerra, G. Alonso, Analytical study of the thermal hicles, ARS J. 32 (1962) 237–240.
behaviour and stability of a small satellite, Appl. Therm. Eng. 29 (2009) [31] P.M. Gomadam, R.E. White, J.W. Weinder, Modeling heat conduction in spiral
2567–2573. geometries, J. Electrochem. Soc. 150 (10) (2003) A1339–A1345.
[12] M. Bulut, N. Sozbir, Analytical investigation of a nanosatellite panel surface tem- [32] J.N. Reddy, D.K. Gartling, The Finite Element Method in Heat Transfer and Fluid
peratures for different altitudes and panel combinations, Appl. Therm. Eng. 75 Dynamics, 3rd ed., CRC Press, 2010.
(2015) 1076–1083. [33] ANSYS® Academic Research, Release 17.2, Help System, Meshing User’s Guide,
[13] S. Corpino, M. Caldera, F. Nichele, M. Masoero, N. Viola, Thermal design and ANSYS, Inc.
analysis of a nanosatellite in low Earth orbit, Acta Astron. (2015) 247–261. [34] ANSYS® Academic Research, Release 17.2, Help System, Theory Reference, ANSYS,
[14] Y. Liu, X. Pan, G. Li, X. Liu, L. Jiang, Investigations on temperature distribution of Inc.
satellite surfaces affected by solar absorptivity, Adv. Space Res. (2010) 593603. [35] ANSYS® Academic Research, Release 17.2, Training Material, ANSYS Mechanical
[15] H. Curtis, Orbit Mechanics for Engineering Students, Elsevier Butterworth- Heat Transfer 17.0, Nonlinear and Transient Thermal Analysis, ANSYS, Inc.
Heinemann, 2005. [36] P. Fortescue, J. Stark, G. Swinerd (Eds.), Spacecraft Systems Engineering, 3rd ed.,
[16] M. Bate, D. Mueller, J. White, Fundamentals of Astrodynamics, Dover Publications, Wiley, 2003.