Acta Astronautica 223 (2024) 77–97

Contents lists available at ScienceDirect

Acta Astronautica
journal homepage: www.elsevier.com/locate/actaastro

Temperature mapping methods for thermoelastic analyses of the ARIEL

spacecraft payload module
Andrés García-Pérez a, * , Alejandro Fernández-Soler a , Gianluca Morgante b , Javier Pérez-
Álvarez a , Gustavo Alonso a , Laura García-Moreno a , Antonio Scippa c , Daniele Gottini c ,
Riccardo Lilli c
a
Instituto Universitario de Microgravedad “Ignacio Da Riva” (IDR), Escuela Técnica Superior de Ingeniería Aeronáutica y del Espacio (ETSIAE), Universidad Politécnica
de Madrid (UPM), Pza. Cardenal Cisneros 3, 28040 Madrid, Spain
b
Istituto Nazionale di Astrofisica (INAF), Osservatorio di Astrofisica e Scienza dello Spazio (OAS), via Piero Gobetti 93/3, 40129 Bologna, Italy
c
Dipartimento di Ingegneria Industriale, Università degli Studi di Firenze (UniFI), via St Marta 3, I-50139 Florence, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: The ARIEL mission is a European space project that aims to detect exoplanets with a spacecraft orbiting around
ARIEL the L2 point of the Sun-Earth system. The main payload consists of a Cassegrain telescope composed of mirrors
Prescribed average temperature that reflect and concentrate the incoming light from the deep space observations to finally guide it to the de­
STOP analysis
tectors. As in many other space missions, a dedicated complex assessment is established during the design phase
Temperature mapping
Thermal analysis
to evaluate the impact on the optical performance caused by thermoelastic effects, which involves the coordi­
Thermoelastic analysis nated work of the thermal, structural, and optical engineers. Despite that well-known and standardized processes
and tools are established separately in each involved area, there is a lack of standardization about the way of
exchanging the data between them, where additional calculations are required in some cases. This work focuses
on the temperature mapping, which is the intermediate step between thermal and structural analyses, where
temperatures are transferred to the structural model. The main difficulty of this process is related to the dif­
ferences in modelling methods and approaches between both models, being necessary the development of an
adequate algorithm to find the most accurate transfer of temperatures. This paper shows two different options for
temperature mapping, detailing the proposed flowcharts. One of these methods requires the performance of an
additional thermal conductive analysis, where a new improved procedure has been implemented in this work to
solve some computational issues that made its application for large models difficult or even unfeasible. Both
temperature mapping methods have been applied to the payload module of the ARIEL spacecraft, comparing the
output results in terms of temperatures, stresses, forces, and displacements to evaluate their differences.

1. Introduction internal temperature. Space telescopes are one of the different types of
payloads on board spacecraft, which are present in scientific space
During space missions, spacecraft payloads are exposed to harsh missions for the distant observation of the stars, galaxies, and other
thermal environments, caused by the combination of their operations in celestial bodies [1–5]. The requirements for this type of payloads are
vacuum, the heat transfer by radiation from the different celestial bodies very stringent with regard to the dimensional stability, since small
and the variability of the internal temperature generated by the change changes in the mirrors or in their mechanical assembly in terms of dis­
of the incoming external heat fluxes during the different orbital phases. placements, angular distortions or deformations will cause a significant
In addition, the different operational modes of the internal equipment degradation in the optical performance.
(e.g. nominal, survival, idle, etc) lead to a variation of internal heat One of the main causes that may lead to the degradation of the
flows. Therefore, a crucial aspect during the design of spacecraft and dimensional stability of space systems is the change in temperature
their equipment is the evaluation of thermoelastic effects on the struc­ along the mission [6]. To evaluate the effects on the optical performance
ture and especially on the payloads caused by the variability of the of the thermoelastic phenomena [7,8], a design process denominated

* Corresponding author.
E-mail address: [email protected] (A. García-Pérez).

https://doi.org/10.1016/j.actaastro.2024.07.009
Received 22 March 2024; Received in revised form 6 May 2024; Accepted 3 July 2024
Available online 6 July 2024
0094-5765/© 2024 The Authors. Published by Elsevier Ltd on behalf of IAA. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Structural, Thermal and Optical Performance (STOP) analysis is estab­ 19]. Other more sophisticated search algorithms involve the calculation
lished for space projects during the development phase of these systems. of the shapes and sizes of both thermal and structural entities to estab­
This process involves a complex coordinated work between three lish with more accuracy the correct association between the geometries
different engineering disciplines, which usually belong to different or­ of the structural and the thermal models. Typically, the thermal model
ganizations. Therefore, it is recommended to define from the beginning mesh is much coarser than the structural FEM mesh. Therefore, there are
of the project a careful interrelation of the common aspects and the several structural nodes associated to the same thermal node. This
procedure for a better efficiency, focusing on the way of transferring method has the advantage of its simplicity, since only geometric com­
data between the different areas. Generally, the direct path in the STOP putations are performed to define the groups of structural nodes or el­
analysis process is defined in sequential order. The first step is to ements associated to each thermal node. The transferred temperature
perform thermal analyses using as input the different conditions of each map is visualized on the FEM as a set of zones or patches, each one with a
relevant thermal load case expected during the orbital phase to calculate constant temperature and geometrically similar to each corresponding
the temperature map on the analysed spacecraft. A subsequent step thermal node. The drawback of this method is that there are temperature
called temperature mapping consists of transferring these temperatures jumps between the patches that are not realistic and, in some cases, lead
to the structural model, which are used as input for the structural to conservative results in terms of local stresses or forces.
thermoelastic analyses to compute the displacements, rotations, and The second temperature mapping option is the Geometrical Inter­
deformations on the different elements of interest. Finally, these polation (GI) method, which is an approach that attempts to provide a
displacement fields are transferred to the optical team to evaluate smoother temperature distribution on the FEM, avoiding the patches of
whether the degradation on the optical performance is within the constant temperature generated from the previous method. This second
allowed limits or not. This general process has been implemented in option firstly transfers the temperature of each thermal node to a single
several space projects, each one with some particularities and small structural node, typically the closest one, and then calculates the tem­
differences. In Ref. [9], the STOP analysis is defined for LISA space perature field for the rest of structural nodes by employing an interpo­
mission, where together with the accurate calculation of the thermo­ lation algorithm, which only considers the geometric parameters such as
elastic effects on this spacecraft, the gravitational effects are also added the distances between the nodes. This option is used in some works
to the process, renamed as STOP-G, considering its relevancy for this related to the STOP analysis process [5,20]. The main drawback is the
project dedicated to study the gravitational waves. The same need for strong dependency of the resulting temperature maps with the selected
accurate numerical simulations of the STOP analysis is presented in interpolation algorithm and its corresponding input parameters, which
Ref. [10] for the design of future National Aeronautics and Space may lead to inaccurate maps if these input parameters are established a
Administration (NASA) missions for planet finder spacecraft. In priori arbitrarily.
Ref. [11], the STOP process was adopted to investigate the effects of The third method is called the Centre-Point Prescribed Temperature
various design modifications on the developed interferometer perfor­ (CPPT), which represents a more advanced approach in which the
mance to improve its design in the future. In Ref. [12], the application of temperature maps on the structural finite element model (FEM) are
the STOP analysis to the James Webb Space Telescope program is calculated in additional thermal analyses with the same FEM, but with
explained, showing the correlation of the predicted numerical analysis properties converted from structural to thermal. The thermal FEM
with the experimental results. The experimental verification of the nu­ analysis uses as boundary conditions the temperatures transferred from
merical results calculated by the STOP analysis in a small CubeSat is each node of the lumped parameter model (LPM) to its closest FEM
detailed in Ref. [13]. The STOP analysis is also adopted for bigger node, like in the previous method. In this case, the conversion of the
ground-based telescopes [14], where similar software tools are used for FEM from structural to thermal properties requires an important effort
the different main steps, but evaluating different thermomechanical involving the close collaboration between the thermal and the structural
environments according to the on ground conditions. engineers, being this aspect the main drawback of this method. This FEM
The three main tasks of the STOP process (thermal, structural and conversion mainly consists of the properties modification by setting the
optical analyses) are individually well established in the space industry conductive parameters of the materials and joining elements while
by the definition of general rules and guidelines [15,16] and by using maintaining the same nodes and most of the elements. The purpose is to
their corresponding commercial software (ESATAN-TMS, SINDA and refine the temperature distribution on the FEM by performing linear
Thermal Desktop for thermal analysis, NASTRAN, ANSYS and ABAQUS thermal conductive analyses. However, if this obstacle is overcome, in
for structural analysis and Zemax and CODE-V for optical analysis). principle the results are more reliable than those from the previous
However, there is a lack of standard procedures and commercial soft­ methods because the temperature distributions have been computed by
ware for the common data exchange between each involved area. The assuming the thermal equilibrium in these additional thermal conduc­
first transition step corresponds to the temperature mapping between tive analyses. This assumption is more realistic than considering con­
thermal and structural models. The main difficulty is the definition of stant temperature patches as in the first method, or an arbitrary
the data transfer methodology, since both the thermal and structural geometrical interpolation as in the second one. Some works about the
models may have been built based on different numerical methods STOP analyses of different space structures have used this method to
(lumped parameters or finite elements), with different modelling as­ transfer temperature maps from the thermal analyses to the FEM [12,
sumptions (detailed vs simplified) and with different mesh sizes. 21].
Four temperature mapping methods are exposed in Refs. [7,8], The fourth and most accurate method described in Ref. [8] is the
which have also been tested in other previous works [17,18]. These Prescribed Average Temperature (PAT) method, which is an improved
methods mainly focus on the transmission of temperatures between a version of the previous one. The PAT method is based on two
thermal model created by the lumped parameter method, being assumptions.
ESATAN-TMS the most used software in the European space industry,
and a structural model defined by the finite elements (FE) approach, • The temperature calculated at each thermal node of a lumped
usually in NASTRAN code. parameter model corresponds to the average temperature of the zone
The simplest temperature mapping method is the so-called Patch- represented by this node.
Wise Temperature (PWT) method, which consists of the direct transfer • The temperature map can be refined with the FEM assuming that it is
of the temperature calculated in each thermal node to the set of struc­ in a quasi-steady-balance equilibrium considering only the conduc­
tural nodes associated to it. This association is usually done by the tive properties of the structure (without any thermal radiative
nearest-node technique, which finds for each structural node the nearest exchange).
thermal node. This approach is extensively used in several projects [1,

78
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 1. CAD of the ARIEL spacecraft (left) and its Payload Module (right).

Taking into account the above-mentioned assumptions, the PAT

method described in this work uses a thermal conductive analysis with
the FEM to calculate the final temperature map, similar to the CPPT
method. However, the CPPT method presents the inaccuracy of
assuming that the temperature calculated at each LPM node represents
the local value at the centre point of this node. Actually, the calculated
value represents the average temperature of the entire region covered by
the node. The PAT method considers this aspect by implementing an
additional condition to guarantee that the average temperature on each
FEM region associated to a single LPM node corresponds to the tem­
perature transferred from this LPM node. Therefore, the PAT method is
the option of the four methods described in Ref. [8] that provides the
most realistic temperature fields on the FEM. The PAT method has been
implemented in a temperature mapping software tool called SINAS [22],
which has been developed over the last decades and is currently being
improved by ESA in an updated tool called pysinas [23]. This software
tool has been designed to perform the temperature mapping between a
thermal LPM built in ESATAN-TMS and a structural FEM defined in
NASTRAN. For the geometric association between both models, the
software tool considers not only the position of the nodes, but also the
geometric shapes of the elemental entities. This process requires com­
plex calculations to evaluate the sizes and shapes of both the thermal
nodes and the structural elements in a step denominated “overlapping”
of models. For the old SINAS IV software, the overlapping process suc­
cess strongly depended on the similarity between the modelling ap­
proaches adopted for both models. Therefore, it was usually required
that both the thermal and the structural models were represented by
elements of the same dimensionality, mostly 2D elements, which limited
the versatility of this tool for models with different modelling ap­
proaches and mismatches in geometric definitions. For instance, struc­
tural models are continuously growing in quantity of nodes due to the
increasing use of 3D elements to obtain more accurate results, especially
required for the STOP analysis in several projects, while the thermal
models are still composed of 2D geometries. This is the handicap of the
analysed structure described in this paper, whose structural FEM is
mostly composed of 3D elements while the geometry of its thermal
model is defined by surfaces. The current pysinas software has solved Fig. 2. Cad of the ARIEL telescope assembly (top) and optical bench (bottom).
this limitation, being now versatile with any combination of geometries
between LPM and FEM. avionics and the communication equipment (see Fig. 1).
One of the purposes of this work is the development and application The main challenge of the Telescope Assembly (TA) of the ARIEL
of a new software tool for temperature mapping, which have been PLM (see Fig. 2) is the design of a full aluminium structure that will
successfully used for the Atmospheric Remote-Sensing Infrared Exo­ undergo a significant thermoelastic contraction from the room temper­
planet Large-survey (ARIEL) mission. This project is being carried out by ature to the operational cryogenic environment due to the relatively
the European Space Agency (ESA) with the objective of observing exo­ high coefficient of thermal expansion (CTE) [26]. This material was
planets while orbiting in a halo orbit around the L2 Lagrange point of the selected based on its lower price and better thermal conductivity that
Sun-Earth system [24,25]. The spacecraft is composed of two main minimizes the temperature gradients during observation. The pre­
modules, the Payload Module (PLM) consisting of a Cassegrain Tele­ liminary analysis from the optical point of view about these effects is
scope that will operate at cryogenic temperatures around 50 K and the depicted in Ref. [27]. It shows the great complexity of considering the
Service Module (SVM) containing the auxiliar subsystems such as different sources of misalignment degrading the optical performance

79
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

during operation. One of the main parts of the TA structure is the Optical can be applicable to the ARIEL PLM FEM for temperature mapping. The
Bench (OB), an aluminium part whose complex geometric design is the comparison between the PWT and PAT temperature mapping methods
result of a careful process involving the thermal and structural numer­ in terms of temperature distributions on the ARIEL PLM FEM is shown in
ical models [28] to achieve a final design compatible with a full set of section 5, together with the structural output (stresses, forces and dis­
requirements. Most of the requirements applied to the OB correspond to placements) calculated by the subsequent thermoelastic analysis.
the mechanical, thermal and electrical interfaces with most of the Finally, the conclusions are summarized in section 6.
equipment on board the telescope assembly, including the main mirror
M1. The M1 primary mirror is a monolithic stiffened part made of 2. Methodology
aluminium supported by three flexure hinges. Its design is also chal­
lenging [29] due to the combination of its large size (major axis of 1.1 m This work focuses on the transfer of temperature between a thermal
and minor axis of 0.7 m) with the high CTE value of the aluminium model built with the lumped parameter method and a structural model
material, which leads to a significant thermoelastic contraction that defined with the finite element method. The software widely used for
should be with the minimum angular and shape distortion. Different thermal analysis in the European space industry is ESATAN-TMS, spe­
tests were conducted on representative samples made of the same cifically designed by the European Space Agency (ESA) and supported
aluminium material [30,31] to find the best way to perform the heat by the company ITP Aero to calculate the temperature results on
treatment that achieves the mechanical stabilization of the primary spacecraft where the main heat transfer mechanism is by radiation. On
mirror. A detailed FEM of the primary mirror was initially created to the other hand, the most employed structural analysis code in the space
study the behaviour of this critical part under gravity environment in industry is NASTRAN, designed by the National Aeronautics and Space
Ref. [32], where numerical analyses were key in discovering the causes Administration (NASA), which remains popular due to its reliability for
of some failures during prototype tests, saving the costs associated with linear simulations.
the execution of additional tests. The lumped parameter method is based on the discretisation of a
The thermal architecture of the ARIEL PLM is described in Refs. [33, continuous medium as a discrete network of nodes representing the
34], where the stringent thermoelastic stabilization is achieved by a capacitance of the system linked by conductive (GL) and radiative (GR)
combination of passive and active cooling elements. Passive tempera­ couplings. The basic justification for this spatial discretisation is the
ture control is provided by three V-grooves and one radiator that allow isothermal assumption, meaning that each node is considered as
the telescope assembly to be in a stable temperature below 60 K. Active isothermal. Hence, in regions where spatial temperature gradient exists,
cooling control (Neon Joule-Thomson cryocooler [35]) is added to more nodes are needed to fulfil the isothermal assumption. Supposing a
decrease the temperature below than 42 K because the detectors of the spacecraft orbiting a planet, discretized as a model of N nodes, where Ci
ARIEL Infra-Red Spectrometer (AIRS) [36] must operate at such low is the heat capacity, Ti is the temperature, Q̇i is the internal heat load,
temperature during the operational observation periods. The numerical and Q̇load,i = Q̇P,i + Q̇albedo,i + Q̇Si is the environmental heat load absor­
calculation of the temperature on the entire ARIEL PLM is done by a bed by the node i, the energy balance is approached by Ref. [38]:
thermal model built in ESATAN-TMS. The temperature maps of different
N N
representative scenarios are transferred to the structural FEM of the dTi ∑ ( ) ∑ ( )
Ci = Kij Tj − Ti + Rij Tj 4 − Ti 4 + Q̇i + Q̇load,i , (1)
same assembly. Then, the resulting displacements and deformations of dt i∕
=j i∕
=j
the optical elements are evaluated through the corresponding thermo­
elastic analyses with the aim of assessing the impact on the optical where the linear couplings GL between node i and node j is Kij, and the
performance. This STOP analysis process should be done with the radiative exchange constant between those nodes is Rij. Therefore,
maximum possible accuracy considering the sensitivity of the optical solving the N first-order differential equations with enough boundary
results. With this purpose, a careful assessment was done in Ref. [37] to conditions, the temperature of the nodes and the heat fluxes exchange
evaluate the causes of the numerical inaccuracies in the structural between them are obtained. This approach is implemented in the soft­
thermoelastic analysis of the ARIEL PLM and define the way to improve ware ESATAN-TMS.
the precision of the calculated displacements. The finite element method is a numerical approach consisting of
The work presented in this paper describes the procedures and dividing a complex structure into several small elements with relatively
application of two different temperature mapping methods (PWT and simple geometric shapes to represent its behaviour by a discrete set of
PAT) performed for the ARIEL PLM in the framework of the STOP nodes. Unlike the lumped parameter method, each FEM node does not
analysis to compare the results and evaluate their differences. The have any associated volume or area, but they represent point locations of
procedures defined for both mapping methods are exposed in section 2, the structure. FEM nodes are connected by elements, which are the
while the STOP process used for the ARIEL project is explained in section entities with the associated dimensions. This is the most employed
3, together with the description of thermal (ESATAN-TMS) and struc­ method for structural analysis, where the simple geometries of the ele­
tural (NASTRAN) models of the ARIEL PLM. For the PAT method, an ments allow the easy definition of the stiffness, damping and mass ma­
additional thermal conductive analysis is required as an intermediate trixes for each element according to its corresponding formulation.
step to refine the temperature map on the structural FEM, where the These elemental matrices are assembled for the entire model considering
structural properties of the FEM are replaced by thermal properties to the full set of degrees of freedom (displacements and rotations of the
perform this thermal conductive analysis. In order to assure in this nodes), which are later used to solve the mechanical equations. For
analysis that the average temperature of each FEM region corresponds to linear static analysis, only the stiffness matrix [K] is used, together with
the value of the associated ESATAN-TMS node, extra entities, such as the vector of applied loads {P}, to calculate the vector of resulting dis­
multipoint constraints (MPC), are added in the FEM. However, some placements {u} by the following equation:
issues related to the required high computational resources have been
found in these analyses caused by MPC entities connected to a great [K]{u} = {P} (2)
number of nodes, making the application of the PAT method for large The rest of required results (stresses, forces, etc.) are later computed
models difficult or even unfeasible. Fortunately, this work proposes a according to the constitutive equations for each type of element.
solution consisting of dividing the original large MPCs into a set of Thermoelastic analysis is a particular option for linear static analysis
smaller MPCs. The computational assessment described in section 4 that evaluates the mechanical effect on a structure caused by the tem­
demonstrates the validation of the proposed solution, where the perature change. Therefore, the inputs are the initial or reference tem­
computational resources are appreciably reduced while maintaining the perature map (Tref) and final temperature map, which are initially used
same degree of accuracy in the results. Thanks to that, the PAT method

80
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

to define a set of equivalent nodal forces Pi by Ref. [7]:

N
∑ ( )
Pi = Ej Aij αj Tij − Tref , (3)
j=1

where N is the number of elements connected to the node i, Ej and αj are

the Young’s modulus and the coefficient of thermal expansion respec­
tively of the element j, Aij is a geometric parameter that depends on the
formulation of each type of element considering its shape and size, and
Tij and Tref are the final and initial (or reference) temperatures respec­
tively. Typically, the initial temperature map is constant and reflects the
stress-free or undeformed temperature field, while the final temperature
map required for a thermoelastic analysis has been derived by mapping
the temperature field produced by the lumped parameter thermal model
on the finite element model.
Two temperature mapping methods (PWT and PAT) have been
selected in this work to carry out the temperature transfer from the
thermal model to the structural model of the ARIEL PLM to evaluate the
thermoelastic effects within the framework of the STOP analysis. The
Patch-Wise Temperature (PWT) method is one of the most used tem­
perature mapping methods in the space industry due to its simplicity,
which consists of the direct definition of the temperature between each
thermal node and the associated group of structural nodes. Its main
drawback is that the temperature map generated on the structural FEM
Fig. 3. Flowchart of the temperature mapping process with the Patch-Wise
is composed of constant temperature patches (each patch corresponds to
method for the ARIEL PLM.
the influence zone of each thermal node), which is not completely
realistic. A way to improve the accuracy of the mapping process is by
used by SINAS IV software [22], denominated as “overlapping” of
using a more realistic approach, such as the Prescribed Average Tem­
models, where the geometric shapes of the thermal and structural
perature (PAT) method. This method provides more realistic tempera­
models are carefully considered to establish the classification of groups
ture distributions on the FEM, but with the need of performing
of structural elements that represent the same area as each corre­
intermediate thermal analyses with the same FEM to compute these final
sponding thermal node. Additionally, some extra operations are needed
temperature maps. The main efforts are focused on the conversion of the
to correctly classify the structural elements that are located on the edge
FEM, initially defined only for structural analyses, to be able to perform
of two adjacent thermal nodes. Therefore, the approach used by this
thermal conductive analyses. Additionally, a way to assure that the
software is the association of dimensional geometric entities instead of
average temperature of each FEM region corresponds to the temperature
points, what requires the complete definition of the geometry of both
transferred from the associated ESATAN-TMS node is necessary. This
models. One of its drawbacks is the difficulty of correctly establishing
can be done by adding some extra entities in the thermal FEM such as for
this association when there is a significant difference between both
instance multipoint constraint (MPC) elements. These are the main
models in terms of geometric accuracy, representativity and type of
difficulties of this method, especially the FEM conversion, which re­
mesh (2D mesh vs 3D mesh). This difference in geometric representation
quires the teamwork between the thermal and structural engineers.
of the structure in both models can be a source of errors in the mapped
Furthermore, it was found in this work that if MPC entities containing a
temperature fields and hence a source of errors in the final thermoelastic
great quantity of independent nodes, on the order of 10000 nodes, are
responses.
employed for temperature averaging, high computational resources are
For the temperature mapping of the ARIEL PLM, it was decided to use
required, which may make FEM thermal analysis difficult or even un­
a simpler algorithm for the structural and thermal entities association.
feasible. Due to the interesting advantages of the PAT method in terms of
Firstly, the structural entities associated in this process are the nodes
accuracy and realism, one of the objectives of this work is the assessment
instead of the elements. Secondly, the search algorithm only considers
of this method for the ARIEL project and the development of a meth­
coordinates referring to the same coordinate system for the nodes of
odology to overcome its difficulties.
both the thermal and structural models. The algorithm, based on the
A software tool called “ESATAN2NASTRAN” has been developed by
nearest-node approach, finds the closest thermal node for each struc­
IDR/UPM institute with the objective of performing temperature maps
tural node. Therefore, the shapes and sizes of the structural elements and
between a thermal model defined in ESATAN-TMS and a structural
thermal nodes are not considered, avoiding the problems related to the
model defined in NASTRAN. The tool was developed to provide tem­
complexity of evaluating a great variety of types and shapes of elements,
perature mapping in different methods, being the PWT and the PAT the
the possible mismatch between both models in some complex areas and
used ones for the ARIEL PLM. The detailed definition of both mapping
the uncertainty related to the final association of the border elements.
methods is explained below.
The only required data from the thermal model are the list of nodes
indicating their identification (ID) numbers, coordinates, and tempera­
2.1. Patch-Wise temperature method
tures for the analysed load case. An extra field is included to indicate the
name of the part or zone to which each thermal node belongs, named
The Patch-Wise Temperature (PWT) method has been extensively
label, which can be useful for the model partition explained below. All
used in several projects to perform in a relatively easy way the tem­
these data are compiled in a simple.csv file, which is imported in the
perature transfer from a thermal model to the structural model. As it was
mapping software tool. In a similar way, the data required from the
above-mentioned, the main difficulty in temperature mapping is the
structural model are the list of nodes indicating their ID numbers and
establishment of a process to link the nodes or elements of the structural
positions. The temperature mapping software can read all this data
model to the corresponding nodes of the thermal model. This process
directly in NASTRAN format (.bdf file), which can also include other
can be called “association” between structural and thermal entities and
data such as elements, properties, and coordinate systems, which in
typically involves geometric calculations. One option for this process is

81
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 4. Flowchart of the temperature mapping process with the PAT method for the ARIEL Payload.

principle are not mandatory for the search algorithm, but they can help useful for the subsequent thermoelastic analysis when the input final
in classifying the structural nodes according to their represented part or temperatures need to be defined as deltas of temperatures (difference
subsystem for the model partition process explained below. Each between the initial temperature and the final temperature) instead of the
resulting patch may be very similar to the corresponding ESATAN-TMS corresponding absolute values.
node when the thermal mesh is regular (elements with the same shape The next step in the flowchart is the division of the analysed structure
and size), but it may be noticeable different for complex zones with into different parts or subsystems applied for both the thermal and
irregular mesh. However, this drawback is compensated by the versa­ structural models. This process is identified as “Partition” of models and
tility of this approach, compatible with any combination of different is a key step for the quality and the execution time of the mapping
types of elements used to model the analysed system. procedure. The objective is to perform a division in the maximum
The flowchart designed for the PWT mapping in the developed possible number of parts, the greater the number of divisions the better,
software is depicted in Fig. 3. The input data correspond to two files, a according to the representation of each part in both models and the
csv file with the required data from the thermal model, and a bdf file easiness of selecting the groups of nodes of each part. These parts are
with the definition of the structural FEM in NASTRAN format. Both files typically identified based on functional representation in both models.
are read to import all the necessary information from both models. In In many cases this approach also simplifies the selection of nodes in both
case that the position of some of the FEM nodes are referred to a different models into groups. Initially, the user should have a sufficient knowl­
coordinate frame than the one used by the thermal model, a coordinate edge about both models to identify the definition of each part in these
transformation process of these nodes is needed at the beginning. The models. Depending on the level of detail employed in the representation
mapping software has the possibility of reading the entries that define of the same part in each model, the modelling approach, and the
the coordinate frames from the bdf file and perform the corresponding geometrical simplifications, a part can be anything from a single piece to
coordinate transformation. To do this, it is necessary that all the entries a complex subsystem.
of the affected coordinate frames are defined in the bdf file, including at The main and largest structural parts of the ARIEL PLM, such as the
least one reference frame coincident in orientation and location with the Optical Bench, Metering Structure and the principal mirror M1 among
one used for the positions of the thermal nodes. This coincident coor­ others, are separately selected in this partition process as single pieces.
dinate system is selected by the user for the coordinate transformation On the other hand, smaller complex subsystems, such as the M2 mirror
step. mechanism, have all their internal components selected for the same
On the other hand, the temperatures read from the csv file may be in group due to the difficulty of individually selecting each internal
different unit system than the one required for the subsequent FEM component for the thermal and the structural models. The objectives of
analysis. Therefore, the software offers the possibility of selecting both the partition step are the following.
the imported and the final desired units for the temperatures to perform
this conversion before the mapping execution. Furthermore, a constant • Avoiding the unintended association between a structural node
value of temperature can be defined to be added to all the temperatures belonging to a subsystem and a thermal node of another different
to change the origin of the temperature scale if necessary. This can be subsystem.

82
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

• Reduction of the execution time during the subsequent search step.

This is because to find the closest thermal node for each structural
node, the search involves only the thermal nodes of the corre­
sponding part rather than all the thermal nodes of the entire model.

The mapping software tool provides two ways of performing the

definition of the set of nodes for each part for both the thermal and the
structural models. The first option is by indicating the numbering ranges
of the nodes, while the second option is by selecting the names of the
zones or regions belonging to each division. In case of the structural Fig. 5. Scheme of the extra entities (centre nodes, MPC, CELAS, SPC) added to
FEM, the regions are identified by the names of element properties, the FEM for the PAT mapping method.
while for the thermal model an extra field in the imported csv file
contains the name of the region associated to each node. FEM node with its closest thermal node, also defined for the PWT
After partitioning both models into different groups, the software method. However, this association is used for the PAT method to create
executes the temperature transfer to each structural node from the the aforementioned extra entities in the FEM to guarantee that the
corresponding closest thermal node. This process is done part-by-part, temperature for each thermal PLM node is transferred to the FEM.
that is, the search algorithm involves only the thermal nodes of the However, instead of directly applying the temperature to the associated
analysed part in that moment, reducing the execution time compared to set of FEM nodes, a multipoint constraint (MPC) is created for each zone
a process without this partition. In any case, this step is the most time- influenced by a thermal node connecting all the FEM nodes of this zone
consuming, where the distance between each FEM node and each with an extra node (see Fig. 5). By default, this extra node is located with
thermal node is calculated and compared to finally find the closest the same coordinates as the thermal node (in reality, the position of this
thermal node. Finally, this closest thermal node is associated to the extra node will not affect the calculated results). This extra node is the
structural node that is being evaluated. It is expected that each thermal dependent node of the MPC, allowing the temperature of this node to be
node is associated to several structural nodes due to the finer mesh the average temperature of all the independent nodes connected to this
resolution of the FEM compared to the thermal model. For the Patch- MPC. The problem is the incompatibility of directly setting the tem­
Wise Temperature method, the next step is the direct temperature perature of this dependent node by a single point constraint (SPC). To
transfer from each thermal node to all the structural nodes associated to overcome this inconvenient, an extra duplicate node is added (by default
it. Consequently, this generates a temperature map on the FEM in the same position as the dependent node) with the fixed input tem­
composed of patches, hence the name of this method, each one with a perature (by the SPC), which will be transferred to the dependent node
constant temperature that may be different from the temperatures of the in the analysis by an extra conductive element (CELAS) with a very high
adjacent patches. Once all the structural nodes of a part have their conductive value (for the ARIEL PLM simulations it was set to 1.0 × 105
corresponding temperature, the algorithm performs the same search W/K). In this way, both extra nodes will have the same input tempera­
algorithm for the next part. ture during the FEM thermal analysis, which corresponds to the tem­
Finally, the output data are the nodal temperature for each FEM node perature of the related thermal PLM node.
written by the NASTRAN entry “TEMP”. These results are written in The MPC establishes a linear relationship between the temperature
different bdf files, each one related to the corresponding part, which will of the dependent node, Td, and the temperatures of the N independent
be used as input for the subsequent structural thermoelastic analysis. nodes, Ti by the following equation:
N
∑
2.2. PAT method wd Td = wi Ti (4)
i=1

The implementation of the PAT method is significantly more

complicated than the previous PWT method. Essentially, the PAT For the definition of the MPC, it is necessary to configure the weighting
scheme (shown in Fig. 4) is an ampliation of the PWT scheme, where factor for each connected node, with the condition that the sum of all the
apart from the objective of reading the temperature from the thermal weighting factors for the independent nodes, wi, is equal to the absolute
model to be transferred to the FEM, it is necessary to include a second value of the weighting factor of the dependent node, wd, to correctly
path to convert the FEM from structural to thermal simulations. perform the averaging.
Therefore, extra activities are added to the flowchart to perform this N
∑
more sophisticated procedure. wi = wd (5)
Some of the steps described for the PWT method are still used for the i=1

PAT method, in particular, the importing of both models, the operations

In Ref [7], it is commented that one of the most realistic ways to set these
of selecting the desired units for temperatures, coordinate trans­
weighting factors is by considering the associated volume for each node.
formation of some of the FEM nodes and partition of both models remain
There are different options to calculate this parameter, but the used one
the same. One different aspect is related to the imported FEM. While for
for this work is by assuming that the influence zone of each PLM node
the PWT method it was not necessary to include in the input bdf file
corresponds to a portion, part or group of parts made of the same ma­
some typical entries such as the materials, for the PAT method, where
terial. If this assumption is correct, the associated volume for each FEM
the same FEM will be converted for thermal simulations, it is necessary
node is proportional to its associated mass, which in turn is proportional
to include all the relevant entries for the complete definition of this
to the resulting nodal force when the structure is subjected to a uniform
model. Another different aspect is the objective of the main path. While
gravity load in a static analysis. Therefore, it is necessary to previously
for the PWT it was the creation of the output files containing the tem­
run this static gravity load analysis with the same FEM to request the
peratures for all the FEM nodes, for the PAT method, this path finishes
nodal forces (by the NASTRAN command “GPFORCE”) and import these
with the creation of the boundary conditions (fixed temperatures
results (by a.pch file) into the software tool to automatically configure
defined by SPC entries for some extra nodes), together with extra nodes,
the weighting factor (wi) of each FEM node by the next equation:
conductive elements (CELAS) and multipoint constraints (MPC) that will
allow the calculation of the temperature map in a subsequent thermal
conductive analysis.
The objective of the main path is the same association between each

83
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Table 1
List of NASTRAN entries modified for thermal analysis.
Category Structural Thermal Description
entries entries

Materials MAT1 MAT4 Entry for isotropic material

definition
MAT8 MAT5 Entry for orthotropic and
anisotropic material definition
Element PCOMP PSHELL Entry for composite 2D
properties property
PBARL, PBAR, Entries for 1D bar and beam
PBEAML PBEAM properties
PBUSH PELAS Entry for property of the
joining 1D elements
Elements CBUSH CELAS Entry for the joining 1D
elements
Multipoint RBE2, RBE3, MPC or Entries for multipoint
constraints RBAR CELAS constraints

⃒ ⃒
⃒→ ⃒ Fig. 6. Typical scheme for the representation of the mechanical joints in a
⃒ F i⃒ structural FEM.
⃒ ⃒
wi = N ⃒ ⃒ (6)
∑ ⃒⃒→ ⃒⃒
⃒ F j⃒
j=1

⃒ ⃒
⃒→ ⃒
Where ⃒⃒ F i ⃒⃒ is the module of the resulting applied force to the node i.

With this equation, the sum of weighting factors of all the independent
nodes associated to the same MPC (N nodes) will give a unitary value.
The end step of this main path is the creation of output files con­
taining the definition of the boundary conditions (fixed temperatures by
SPC entries), together with the extra nodes, CELAS elements and MPC
entries, which are used in a subsequent thermal analysis of the FEM.
The second path is a new procedure added for the PAT method with
the aim of preparing the FEM for thermal analysis, where most of its
entities remain the same (nodes, coordinate systems, 2D and 3D ele­
Fig. 7. 2 options for FEM joining entities conversion: MPC approach (left) and
ments), but there are other entries (materials, joint elements, rigid ele­
CELAS approach (right).
ments) that need to be tailored with thermal parameters, mainly thermal
conductivities. The NASTRAN entries that shall be modified are indi­
cated in Table 1, where some particular aspects are clarified in the The key aspect in this FEM conversion is the definition of the joining
explanation below. entities, where the modelling of the corresponding thermal conductivity
Material definitions used for structural analysis, with parameters should be carefully performed. Modelling of the joints in the structural
such as the Young’s Modulus and Poisson’s Coefficient, among others, FEM is typically done by the combination of rigid elements (mostly
should be replaced by the material entries with thermal parameters such RBE2) and elastic 1D elements (CBUSH). The scheme for modelling
as the thermal conductivity. Isotropic structural materials (MAT1) are these joints is depicted in Fig. 6, where the CBUSH elements should be of
replaced by equivalent isotropic thermal materials (MAT4), while other zero length to avoid numerical errors for the thermoelastic analysis.
more complex materials definitions (orthotropic MAT8) are replaced by For the conversion of this set of entities, two options are available,
equivalent definitions for orthotropic and anisotropic thermal materials which are shown in Fig. 7. In both options, the original CBUSH element
(MAT5). PCOMP entries are used to define composite materials, as well is replaced by a CELAS element, which is the NASTRAN entry used for
as for 2D sandwich properties. This entry seems not compatible with generic 1D conductive elements for thermal analysis. The only needed
thermal analysis. Therefore, it should be substituted by an equivalent parameter of these conductive elements is the value of the thermal
monolithic PSHELL property whose thickness is the total thickness of the conductance, which is indicated in the associated property entry
original PCOMP property. Additionally, new MAT5 entries are created (PELAS). For the first option, each RBE2 is replaced by a thermal mul­
in this process to define the equivalent thermal materials associated to tipoint constraint (MPC), where the centre node is the dependent one,
these new shell entries, where the thermal conductivity in each direction whose temperature will be the average of the temperatures of the in­
is carefully calculated according to the replaced composite or sandwich dependent nodes. By default in this case, all the independent nodes have
structure. the same weighting factor, where the total sum of these factors is 1. The
On the other hand, two options are available in NASTRAN to define second option consists of replacing each original rigid element (RBE2)
the cross-section parameters for the 1D elements: by defining the section by a set of new CELAS elements, each of them connecting the centre
shape and its dimensions, or by directly provide its associated geometry node to each external node.
parameters such as the area, second moments of area and shear and The conductance values used for these joining elements have been
torsional modules. It seems that the first option is not compatible with extracted from the ESATAN-TMS model in their respective interfaces,
thermal analysis in NASTRAN versions up to 2021, so a conversion of the where the thermal conductance associated to a screw is the total
affected property entries (PBARL and PBEAML) to the equivalent entries conductance between the two connected parts divided by the number of
of the second option (PBAR and PBEAM) is necessary. For the thermal screws between both parts. For the option 1 (MPC approach), this
analysis, only the cross-sectional area is needed for the conductance of resulting thermal conductance is directly defined for the central CELAS
1D elements, which is calculated in this conversion considering the element. For the option 2 (CELAS approach), a more complex calcula­
different possible shapes and dimensions. tion is done, considering that there are 3 sets of conductive elements in

84
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 8. Scheme for the modification of MPC entities for PAT method: original MPC (left) and new set of MPCs and CELAS elements (right).

series (from replaced RBE2 (part A), CBUSH and RBE2 (part B)) and
assuming that the CELAS elements that replace each original rigid
element are quasi-parallel. Taking both assumptions into account, and
deciding by default that the thermal conductance is equally distributed
between the 3 sets of conductive elements, the resulting conductance
values are calculated by the following equations:
GLi = 3GLT (7)

GLi
GLij = (8)
Ni

where GLT is the total conductance of the represented joint and GLi is the
conductance of each set of conductive elements. GLi is directly the value
of the thermal conductance for the central CELAS element. Additionally,
GLi is also the resulting thermal conductance of each set of Ni conductive
elements that replace a single rigid element, considering that all these
CELAS elements are almost parallel and have the same individual
thermal conductance of GLij.
Once all the thermal parameters are configured for the FEM, the final
step of this path is the generation of a NASTRAN file (bdf) with the
converted FEM definition, which is used later as input file, together with
the other files with the additional entities, in the corresponding thermal
Fig. 9. Flowchart of the STOP analysis for ARIEL Payload.
analysis to refine the temperature map on its nodes. The results of this
analysis correspond with the temperature calculated to each structural
node by the PAT method, which are used as input in the subsequent maintain the same average temperature for the corresponding FEM zone
structural analysis to calculate the required thermoelastic results. as that provided by the original MPC, the absolute value of the weighting
One of the issues detected when performing the FEM thermal anal­ factor for the centre node of each secondary MPC is the sum of the
ysis in NASTRAN versions up to 2021 is the exigent computational re­ original weighting factors of its independent nodes. Furthermore, this
sources required for the simulations in terms of RAM memory and hard resulting weighting factor should be the same for the independent node
disk storage. Some rehearsal analyses were made with a simple model, of the primary MPC connected via a CELAS element to this centre node
and it was concluded that these computational issues were caused by the of the secondary MPC. This approach is also compatible for schemes
calculation of the dependency relationships between the nodes with 3 or more levels of MPCs.
belonging to the MPC entities, in particular for those MPCs with the This approach is validated in section 4 of this paper with a case study
greatest quantity of independent nodes. A new method to modify these with a simple FEM of a square plate modelled by 2D quadrilateral ele­
MPCs without altering the results was introduced into the procedure, ments to evaluate the influence of the maximum number of MPC inde­
where those MPC entities with more independent nodes than a given pendent nodes on the computational resources required to perform the
value are divided according to the scheme shown in Fig. 8. thermal analysis. Additionally, the same assessment is done with the
In the example shown in Fig. 8, the original MPC has 20 independent ARIEL PLM FEM to confirm the applicability of this method in a generic
nodes, and it is decided that the maximum allowable number of inde­ complex structure.
pendent nodes is 8, for instance. Therefore, this original MPC is replaced
by a set of MPCs defined in 2 levels, where the lower level MPCs (sec­ 3. STOP analysis in ARIEL mission
ondary MPCs), directly connected to the structure, have a maximum of 8
independent nodes (in the example there are 2 MPCs with 8 independent The Structural, Thermal and Optical Performance (STOP) analysis is
nodes each one, and an additional MPC with the remaining 4 nodes). a critical assessment for the success of the ARIEL mission. The impact of
These secondary MPCs are joined via secondary CELAS elements to one different thermoelastic and gravity environments on the ARIEL tele­
upper level MPC (primary MPC), which is in turn connected via the scope optical performance is evaluated. A set of different load cases is
original primary CELAS element to the node with the fixed temperature established to cover the different key scenarios, which are classified into
(SPC). In principle, secondary CELAS do not seem to be necessary since two groups: on-ground and in-flight load cases. The STOP analysis
the dependent nodes of the secondary MPCs can be at the same time flowchart for the ARIEL Payload is shown in Fig. 9. The first step is the
independent nodes of the primary MPC. However, it was checked that definition of the input conditions for each load case for both thermal and
the configuration without secondary CELAS elements requires the same structural analyses when applicable. The boundary conditions that
high computational resources as the original one. Therefore, the addi­ define the thermal analysis cases are typically the radiative environ­
tion of the secondary CELAS is fundamental to achieve the reduction in ment, the power mode of the instruments and/or by setting tempera­
the computational resources. Furthermore, they should have a very high tures and heat fluxes as boundary condition at particular thermal nodes.
thermal conductance. The weighting factors of the independent nodes of Then, to perform a thermal analysis, firstly, the Geometrical Mathe­
the secondary MPCs remain the same as in the original MPC. To matical Model (GMM) is built to calculate the thermal radiative

85
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

exchanges between the different thermal nodes and the incoming heat
fluxes from the thermal environment in each radiative case. Secondly,
the Thermal Mathematical Model (TMM) that includes all the conduc­
tive couplings, the power dissipation at each thermal case and the
radiative couplings obtained from the radiative case is generated. Then,
the thermal analysis can be run to obtain the conductive and radiative
heat flows and the temperature of each thermal node.
The temperature map from each thermal analysis load case is used as
input for the subsequent structural analysis. Before the corresponding
structural analysis, it is necessary to transform the temperature map
from the thermal model to the structural model, considering the diffi­
culties derived from the different methodologies employed for each type
of analysis (thermal analysis is based on lumped parameter method
while structural analysis is based on the finite element approach) and
the differences in the average sizes of the elements between both
models. This key aspect is evaluated in the next sections of this paper.
To calculate the deformation in the main elements of the Telescope
Assembly, the temperature maps from the thermal analyses, the defi­
nition of the gravity vector orientations and boundary conditions
(mostly with enforced motions at the IF nodes) are used as input con­
ditions for the structural analysis in each load case. Additionally,
stresses and forces are calculated for the structural evaluation in some of
the load cases. The displacements and rotations of the nodes that
represent the mirrors and other secondary optical elements are exported
to the optical teams. In this project there are two different teams of Fig. 10. Overview of the ARIEL GMM.
optical engineers, one dedicated to the evaluation of the ground cases
and the second one dedicated to the flight cases. For both teams, data
exportation is done in the same way, by using an intermediate software
(SIGFIT) [39] that receives as input the displacement fields of the ana­
lysed mirrors to generate the optical data that are used as input for the
optical analyses. Finally, each optical team evaluates these data to
calculate the final performance parameters that determines whether the
system is compliant or not with the established requirements.
In this work, three load cases of the ARIEL STOP analysis have been
used for the assessment of the temperature mapping method. The first
case corresponds to the “Operational Cold” load case, which is a steady-
state condition where the Telescope Assembly is at a nearly uniform
cryogenic temperature slightly above 50 K. It has been selected to be a
representative case for the optical performance. The second case is the
“Non-Operational Cold” load case, which is another steady-state con­
dition, but with lower quasi-uniform temperature. Therefore, being the
coldest steady-state scenario for the ARIEL mission, it is one of the worst
thermoelastic conditions from a stress point of view. The third case is the
“Max Gradient” load case, where after a transient thermal analysis
simulating the flight cool-down during the journey to L2 in the transfer
orbit [40] (similar to Planck [2] and Euclid [41] missions), the tem­ Fig. 11. M1, flexure hinges and OB GMM.
perature map corresponding to the instant with the maximum gradient
between the M1 mirror and the OB is used for the subsequent structural
telescope structure, which consists of the OB, the metering structure
analysis. Unlike the other two cases, this load case shows appreciable
(TMS) and the struts. In addition, the telescope structure is black pain­
temperature gradients between the parts, which will cause high shear
ted, achieving that the baffle and the OB, which are very large, act as
forces on the joint elements. The main issue related to these shear forces
radiating surfaces to enhance the PLM thermal performance.
is the possible slippage between the connected parts, which may have a
As previously commented, the ARIEL thermal model has been built
non-negligible negative effect on the optical performance due to
in ESATAN-TMS. The GMM and TMM is based on the CAD model and
micromotions of the optical elements with regard to their theoretical
represents the main units [34] to verify the thermal requirements. The
positions.
GMM consists of shell geometries as it is shown in Fig. 10, most of them
with quadrilateral shape. The whole ARIEL thermal model consists of
3.1. Thermal and structural models for ARIEL PLM 4905 thermal nodes.
As above mentioned, M1 is mechanically attached to OB via three
As above mentioned, the Payload Module of the ARIEL spacecraft is Flexure Hinges. Due to its surface finish (low IR emissivity), M1 is highly
designed to ensure that both the instruments and the telescope operate radiatively decoupled from its surrounding elements, and therefore, its
at temperatures below 60 K. This fact is achieved through a combination thermal behaviour is driven by OB through its I/F with the Flexure
of passive control elements: three V-grooves, the instrument radiator, Hinges. The GMM of the M1-Flexure Hinges-OB is shown in Fig. 11.
and by using insulating struts that mechanically attach the PLM to the In the ARIEL project, the use of two different finite elements (FE)
SVM. The SVM, which houses all the electronics, operates in the range models for the structural analysis at PLM level was decided. The first one
between 253 and 323 K. The TA consists of four mirrors, the M1 baffle, is the detailed PLM FEM used mainly for the structural simulations for
which blocks unwanted light rays from reaching the M1, and the

86
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 14. FE model for the secondary M2 mirror with the ID numbers of the joint
Fig. 12. Detailed FEM for ARIEL PLM. elements (CBUSH).

below.
The detailed ARIEL PLM FEM used for the STOP analysis has been
used during the different design phases, increasing the level of detail in
the geometric representation by using mostly 3D elements to represent
the main structural parts of the Telescope Assembly (TA), including the
mirrors. The main structure of the TA was modelled by the team of the
Spanish research institute Ignacio Da Riva (IDR), which belongs to the
Universidad Politécnica Madrid (UPM). Additionally, the detailed FE
models of different subsystems created and delivered by the corre­
sponding responsible teams have been assembled into the detailed PLM
FEM: the FE models of the M1 mirror with its flexure hinges and the M2
mirror have been provided by the Instituto Italiano di Astrofisica (INAF),
the FEM of the mechanism of M2 mirror (M2M) was provided by the
Spanish company SENER, the FEM of the baffle B1 was provided by the
Portuguese company Active Space and the FEM of the bipods was pro­
vided by the Technical University of Denmark (DTU). Other subsystems
also considered as part of the PLM, such as the V-grooves, have not been
included in this PLM FEM because their mechanical contribution is
negligible, since they are not directly mechanically linked to the Tele­
scope Assembly. The version of the FEM used for the Cycle C1 (before
the Critical Design Review) of the STOP analysis is shown in Fig. 12 with
the PLM global coordinate frame, while the FE models of the Optical
Bench and Metering Structure are shown in Fig. 13.
This version of the PLM FEM is composed of more than two million
Fig. 13. Detailed FE models for the Optical Bench (top) and Metering Struc­ nodes, where the predominant type of elements is the tetrahedral with
ture (bottom). 10 nodes. The principal mirror M1 and its supports are mostly modelled
by linear hexahedral elements, while thin parts such as the Baffle B1 and
the STOP analysis, considering the need for high accuracy in terms of the mechanism that supports the secondary M2 mirror (M2M) are
structural results for a better optical performance evaluation during the modelled by 2D quadrilateral elements. The joints between the parts are
subsequent optical analysis. In addition, the structural simulations of represented by infinitely rigid elements (RBE2) and 1D elastic elements
STOP analysis are linear and static, which allows the use of a detailed of CBUSH type whose stiffness values can be adjusted for each interface,
FEM with several nodes and elements due to the lower computation following the scheme shown in Fig. 6. Some of the results of interest for
requirements in terms of execution time and memory storage. On the this project correspond to the interface forces on the joint elements
other hand, the second model is used for the structural simulations with (CBUSH). These forces are then used in a dedicated bolt analysis to
dynamic mechanical environments such as sine vibration or acoustic. It guarantee that all the bolts have positive design margins, being espe­
consists of a FEM created mostly by the 2D elements approach to cially critical the margins related to the slippage between the parts. This
represent the main structural parts of the entire ARIEL PLM. Conse­ work will show the calculation of the interface (IF) forces between the
quently, this second FEM is considered as the simplified PLM FEM, secondary M2 mirror and its mechanism (M2M) caused by thermoelastic
which is more suitable for dynamic analyses where the computation effects, where the ID numbers of the CBUSH elements are depicted in
requirements are much more severe than for static analyses. Further­ Fig. 14.
more, this simplified FEM is still adequate for stresses and forces For each mirror, there is an interpolation element (RBE3) that con­
calculation taking into account that the degree of accuracy is less nects all the nodes of the reflective surface to a single dependant node
stringent compared to the STOP analysis results. Therefore, the FEM located close to its centre. Its main objective is the computation of the
used in this work is the detailed one, whose characteristics are indicated rigid body motion of this reflective surface in terms of translational and

87
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Table 2
Computational results of the square plate FEM thermal analyses.
Secondary MPCs Primary MPCs Required Max Success
Open Core hard
Number Quantity Number Quantity
memory disk
of nodes of nodes
[GB] usage
[GB]

2000 500 125 4 15.0 39.4a No

1500 668 167 4 11.2 47.0 Yes
1000 1000 250 4 7.5 32.5 Yes
500 2000 500 4 3.8 22.2 Yes
a
This analysis was not successfully completed, and this value does not reflect
the maximum expected hard disk usage.

Fig. 15. Square plate with the identification of the four zones used for the
computational assessment of thermal analysis.

rotational displacements. These results are part of the inputs for the
subsequent optical performance analysis to measure the impact of the
different environments analysed in the STOP analysis.

4. Computational assessment of FEM thermal analysis in Fig. 16. Temperature map in ◦ C with the PAT method for the square plate FEM
NASTRAN – MPCs limited to 500 nodes.

In this section, an assessment is carried out to evaluate the impact on entities have been included.
the computational resources required for FEM thermal analysis caused
by the presence of MPCs with high number of independent nodes. In
NASTRAN, there are two solution sequences available for the steady- 4.1. Computational assessment of FEM thermal analyses with square
state thermal conductive analysis, the linear static SOL101 and the plate FEM
general-case steady-state thermal analysis SOL153, which takes into
account nonlinear features for radiation and convection. In this work, A representative FEM has been created for this case study, consisting
the first option was selected due to its simplicity and faster simulations. of a square plate of 1 m side and 10 mm of thickness made of aluminium
Two computational problems arise when performing a linear thermal (K = 167 W/(m K)) and modelled by one million 2D quadrilateral ele­
analysis in NASTRAN (versions up to 2021) with a FEM model that ments. The plate is divided into four zones as shown in Fig. 15, where
contains MPC entities with a high number of independent nodes. The each zone has a different known average temperature. That is, it is
first one is related to the large amount of disk storage required for a simulated that there exists a lumped parameter thermal model with 4
temporal file (scratch file), which may exceed the maximum computer elements for the square plate and it is desired to transfer the tempera­
capacity. The second problem is related to the use of the RAM memory. tures from this thermal model to the FEM by the temperature mapping
When running a NASTRAN analysis, the used RAM memory is distrib­ PAT method. In this particular case, the average temperatures are 10 ◦ C
uted into two main partitions, the Buffer Pool and the Open Core. It has for the zone 1, 20 ◦ C for the zone 2, 30 ◦ C for the zone 3 and 40 ◦ C for the
been observed that the required memory for the second partition (Open zone 4. Each zone has approximately 250000 nodes, which are initially
Core) increases when analysing a FEM with large MPC entities. If the connected to a single MPC. Therefore, four large MPCs of approximately
allocated memory for the Open Core partition is not enough for it, a fatal 250000 independent nodes each are included in the FEM to average the
error message appears in the output f06 file indicating that there is temperature for each region during the thermal conductive analysis in
insufficient memory, showing the estimated extra memory necessary to NASTRAN.
successfully finish the analysis. After the computational problems found in the initial configuration
To overcome these problems, a new method explained at the end of caused by the presence of these four large MPCs, the proposed method of
section 2.2 has been developed in this work, which consists of the MPC dividing them into smaller sets of MPCs was applied to overcome this
modification where each original MPC is substituted by a two-level set of issue. Different options were analysed by selecting different values for
smaller MPCs. This method is validated in this section with two different the maximum number of independent nodes for the secondary MPCs.
models. The first part of this assessment (subsection 4.1) is done in a When this number decreases, the quantity of secondary MPCs increases
simple representative FEM consisting of a square plate composed of one and vice-versa. The computational results of these analysed options are
million quadrilateral elements and slightly more than one million nodes. shown in Table 2.
The second part of this assessment (subsection 4.2) is done with the As can be seen in Table 2, the analysed options use MPCs with
ARIEL PLM FEM described in section 3.1, but converted for thermal significantly much less independent nodes than the original ones, which
analyses with the PAT method, where the corresponding extra MPC range from 500 to 2000 nodes. The maximum RAM memory assigned to

88
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

process to associate the FEM elements considering the shape geometry of

each thermal node. Another difference is the way to compute the
weighting factors for the averaging. While in the approach proposed in
this work is based on associated masses (through gravity load analysis),
in pysinas is based on associated volumes. The process to calculate the
temperature is also a bit different between both options. For the pysinas
software, the thermal conductive matrix together with the constraints
matrix are first assembled, and then the final distribution of temperature
in the FEM is calculated by solving the matrix system of linear equations.
When trying to perform the temperature mapping of this square plate by
the pysinas software, it was found that the process was not successfully
completed due to computational issues, possibly caused by the physical
memory limitation due to the in-core solver used by pysinas. In order to
compare the results between the proposed method and the pysinas
software, a new thermal model of the square plate with 8x8 thermal
nodes was created, whose temperatures can be seen in Fig. 17. The
resulting temperature maps from both options were very similar,
showing relative errors below 0.26 %, as can be appreciated in Fig. 18.
This great similarity in the resulting temperature maps for this simple
case demonstrates that despite the differences in the procedure between
Fig. 17. Temperature in ◦ C for an 8x8 thermal model of the square plate.
the PAT method proposed in this paper and the one implemented in the
pysinas software, they can give the same results in regions with uniform
these analyses was around 12 GB. Therefore, the option with secondary
thermal nodes. However, the differences between both options increase
MPCs of 2000 independent nodes each was not successfully executed.
when using a thermal model with non-uniform mesh because the more
However, when the maximum number of independent nodes is reduced,
sophisticated overlapping module of the pysinas software takes into
the computational resources are reduced as well, allowing the execution
account the geometry shape of each thermal node, unlike the
of these analyses. It can be observed that the relationship between the
nearest-node search algorithm implemented in the procedure explained
required RAM memory for the Open Core partition and the number of
in this paper.
independent nodes of the secondary MPCs is almost proportional. The
resulting temperature map calculated by the FEM thermal analysis (PAT
method) for the configuration of MPCs with 500 independent nodes is 4.2. Computational assessment of the thermal analyses of ARIEL PLM
shown in Fig. 16. The distributions of temperature are practically the FEM
same for all options, with slight differences (less than 0.26 %) between
them. It can be noticed that the range of the calculated nodal temper­ The approach proposed in this work consisting of dividing the orig­
atures is wider than the range of the input average temperatures, giving inal large MPCs into a set of smaller MPCs has also been tested in the
the coldest temperature of 2.5 ◦ C in the bottom-left corner and the ARIEL PLM model (described in section 3.1) to validate its application
hottest temperature of 47.4 ◦ C in the top-right corner. This is because the
set of input temperatures indicates the average value in each quadrant, Table 3
while in the FEM the calculated temperature distribution with gradients Classification of the MPCs according to the number of their nodes for the
is reproduced, providing values above and below the corresponding original ARIEL PLM thermal FEM.
average temperature. Number of independent nodes Quantity of MPCs
The PAT method proposed in this work presents some differences in
<300 470
the procedure compared to the PAT method described in Ref. [23] and 300–500 224
implemented in the pysinas software, which is currently being devel­ 500–1000 350
oped by ESA. While in the proposed method the FEM nodes are directly 1000–1500 126
associated to the corresponding thermal nodes using the nearest-node 1500–2000 70
215
search algorithm, the pysinas software uses a more sophisticated
>2000

Fig. 18. Temperature map in ◦ C with the PAT method for the square plate FEM (a) and relative error in temperatures between the PAT method described in this
paper and the pysinas software (b).

89
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Table 4 the differences between the amounts FEM nodes associated to their
Computational results of the ARIEL PLM FEM thermal analyses. corresponding thermal PLM nodes. There are 1455 MPCs, which are
Maximum Quantity Required Max Execution Success classified according to the number of independent nodes in Table 3. The
number of of MPCs Open Core hard time [min] largest single MPC has 81574 nodes.
nodes per memory disk Due to the large sizes of some of the MPCs, the thermal analysis in
MPC [GB] usage
NASTRAN of this original FEM exceeds the computational limits of the
[GB]
computer. Therefore, the process to transform the original MPCs into a
81574 1455 73.5 332.1a 76a No tree configuration of smaller MPCs is applied to this FEM to evaluate the
2000 2248 19.6 62.2a 8a No
1500 2592 15.6 50.8a 7a No
impact on the computational resources required by the corresponding
1000 3297 11.3 72.3 25 Yes thermal analysis. Different configurations are analysed, each one with a
500 5572 6.3 52.4 20 Yes defined value for the maximum allowable number of independent nodes
300 8426 4.0 40.6 15 Yes for the MPCs. Consequently, only those MPCs with more nodes than the
a
These analyses were not successfully completed. Therefore, the marked corresponding limit are replaced by smaller MPCs in the two-level tree
values do not reflect the maximum expected computation parameters. configuration explained at the end of section 2.2. The analysed config­
urations and their corresponding computational results are indicated in
for a complex FEM representative of a real structure. After the appli­ Table 4.
cation of the PAT method described in section 2.2 to the ARIEL PLM, the The first analysis corresponds to the original FEM, which after 76
original thermal FEM has a great variety of sizes for its MPCs, ranging min of execution stopped before finishing because the RAM memory
from a few nodes to tens of thousands of them, generated according to required by the Open Core partition was much higher than the available

Fig. 19. Temperature maps (in Kelvin) obtained from ESATAN-TMS model on the Telescope Assembly.

90
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

one, which was about 12 GB for the used computer. The options with
2000 and 1500 for the maximum number of independent nodes also
overpassed the computational limits in the same aspect. The RAM
memory required for the Open Core partition can be known for those
non-successful analyses thanks to the estimation of this required mem­
ory indicated in the generated output files that appears when one
analysis fails. However, there is not any estimation about the maximum
hard disk usage required for the non-successful analyses. The values
marked with an asterisk in Table 4 are the computational results for the
corresponding failed analyses, but they are not representative of the
required maximum usage for the corresponding successful runs. On the
other hand, when the limit for the maximum number of nodes for the
MPCs was established equal or below 1000, the thermal conductive
analysis of the FEM could be successfully completed. As can be observed,
when reducing this limit, the quantity of MPCs increases due to the
greater number of affected original MPCs that are converted into smaller
MPCs and the greater number of these new smaller MPCs generated in
this conversion. Reducing the limit on the maximum number of nodes is
computationally beneficial because both the required RAM memory and
hard disk usage decrease quasi-proportionally. Finally, it was checked
that the output results in terms of temperature maps maintain the same
degree of accuracy between the different options, showing maximum
errors about 2 mK between the different successful analyses.
The conclusion of this assessment, done for both a simple square
plate FEM and the complex ARIEL PLM FEM, is that reducing the
maximum number of independent nodes of the MPCs is beneficial from
computational point of view, allowing the execution of the thermal
conductive analysis in NASTRAN in any FEM without loss of accuracy.
This fact facilitates the employment of the PAT method for the tem­
perature mapping for models with millions of nodes such as the ARIEL
PLM FEM.

5. Numerical results of the thermal and structural analysis of

the ARIEL PLM

This section describes the results calculated by the ESATAN-TMS

thermal model and the NASTRAN structural model for the ARIEL PLM
in the STOP analysis. Once the way to overcome the computational
drawbacks of thermal analyses for the PAT method in FE models with
many nodes has been determined in the previous section, the objective
of this section is to show the comparison of the temperature map and the
subsequent thermoelastic results (displacements, stresses and forces)
between the PWT and the PAT methods to determine the differences
between both methods in a real structure.
Three representative load cases from the STOP analysis of the ARIEL
project have been assessed to show in this work the results and the
comparison between the two studied temperature mapping methods:
The Operational Cold case, as the representative scenario for the optical
Fig. 20. Temperature maps (in Kelvin) obtained from ESATAN-TMS model on
performance, and the Non-Operational Cold case and the Max Gradient
different assemblies.
case (maximum gradient between M1 mirror and Optical Bench during
the cool-down), which are the worst cases regarding stresses and forces.
The cold operational and non-operational steady state cases correspond temperature maps focusing on some relevant parts are shown in Fig. 20.
to the nominal orbit at the Earth-Sun L2 point [34]. In both steady-state Once the temperatures for each load case have been calculated by the
cases, the SVM (PLM IF) thermal nodes are defined as boundary con­ thermal analyses in ESATAN-TMS, the next step in the procedure
ditions at 253 K. In addition, the instrument detectors and CFEEs are depicted in Fig. 9 is the transfer of these temperatures to the FEM
switched on in the cold operational steady-state case and switched off in (temperature mapping). Two different methods have been exposed in
the non-operational steady-state case. The Max Gradient transient case section 2 and have been applied to compare their differences for the
refers to the journey from the Earth parking orbit to the final orbit at the ARIEL PLM. For the PAT mapping method, the FEM thermal conductive
Earth-Sun L2 point. During the journey, all PLM units cool down to their analyses have been performed with a maximum of 500 independent
operational temperature. Therefore, as thermal gradients may occur at nodes for the averaging MPCs. Additionally, the option for converting
the IFs of different units, the effect on mechanical performance must be the structural rigid elements is the CELAS method (see Fig. 7 (right)).
assessed. The comparison of the resulting temperature maps transferred to the
The final temperatures for each load cases have been calculated with FEM between both temperature mapping methods for the three evalu­
the ESATAN-TMS model described in section 3.1 and the corresponding ated load cases is shown in Fig. 21 for the Telescope Assembly and in
input conditions. The temperature maps on the Telescope Assembly for Fig. 22 for some relevant parts. It can be noticed on the baffle B1 the
the three assessed load cases are shown in Fig. 19, while the same rectangular zones of constant temperature (patches) for the PWT

91
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 21. Temperature maps (in Kelvin) transferred to the ARIEL Telescope Assembly FEM.

method, with temperature jumps between two adjacent zones. However, appears along the bipods that support the mechanical interfaces with the
the PAT method provides smoother transitions of temperatures between M2 mirror, reducing considerably the temperature jump between the
the different zones. In general, the steady-state load cases (Operational M2 mirror and its interfaces, which in turn will reduce in the same order
and Non-Operational Cold cases) show a very uniform temperature map of magnitude the generated shear forces. This example shows one of the
on the telescope assembly, where the difference between both temper­ main differences between both mapping methods, which is the repre­
ature mapping methods is very small. However, the difference in the sentation of the temperature transition along the interface between two
resulting temperature maps between both methods is greater for the parts with a significant gradient of temperature.
transient load case (Max Gradient) due to its greater gradient. This can The next step in the STOP analysis procedure is the structural ther­
be appreciated in some local regions, such as for the M2 mirror and its moelastic analysis (see Fig. 9), which calculates the generated defor­
mechanism (see Fig. 23). mation, stresses and forces caused by the temperature change between a
Fig. 23 shows the comparison of the transferred temperature maps uniform room temperature (293 K for these analyses) and a final tem­
between PWT and PAT methods for the assembly of the mirror M2 and perature, previously calculated by the thermal model for the different
its mechanism for the Max Gradient case. For the partitioning of the load cases and transferred to the FEM by the two different options for the
models in the mapping process, M2 mirror and the M2M were in temperature mapping.
different groups. The high temperature jump between both parts (about The objective of the operational load cases, such as the Operational
53 K) is noticeable in this transient state during the cool-down. The Cold case, is the calculation of the deformations, translations, and ro­
resulting temperature map obtained with the PWT method shows a jump tations of the optical elements to evaluate the impact on the optical
just at the mechanical interface between the mirror and the mechanism, performance. One of the results directly calculated from the structural
which in turn will produce very high shear forces on the screws for this analysis that can give a quick assessment on the optical performance is
interface, as will be shown below. However, for the PAT method, the the rigid body motion (RBM) of the main optical elements. The obtained
thermal conductive analysis reveals that this temperature change RBM results for the Operational Cold case for the main mirrors M1 and

92
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 22. Temperature maps (in Kelvin) transferred to the FEM on different assemblies.

M2 are shown in Table 5, comparing the results calculated with the where the Non-Operational Cold case represents the steady-state situa­
input temperature maps from both mapping methods (PWT and PAT). tion with the coldest temperature, and the Max Gradient case represents
Due to the great uniformity in the temperature map for the telescope a transient situation with a great gradient of temperature across the
assembly, the differences in the motions of the mirrors between both telescope assembly. The resulting maximum stresses calculated from the
mapping methods are small. In this example, most of the relative dif­ respective structural analyses are depicted in Table 6 for each main
ferences are below 1 %, with the exception of the rotation of the M2 component. The stress distributions of the Max Gradient case are shown
around the longitudinal X axis, with a difference of 2 % between both in Fig. 24, comparing the differences between the results from the two
mapping methods. However, the very small absolute value of this mapping methods.
rotation implies that the impact of the selected mapping method is In general, the comparison of the maximum stresses for the steady-
practically negligible for this load case. state Non-Operational Cold case shows a great similarity between the
The main objective of the non-operational load cases is the calcula­ results calculated with the temperature maps from both mapping
tion of the maximum stresses and forces on the different parts, consid­ methods. The small gradient of temperature across the telescope as­
ering these cases as severe scenarios from the structural point of view, sembly for this steady-state case (see Fig. 21 (b)) explains these small

93
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 23. Temperature maps (in Kelvin) on the FEM of the M2 mirror and its mechanism for Max Gradient case: PWT (left) and PAT (right).

Table 5
Rigid body translations (Ti) and rotations (Ri) of the mirrors M1 and M2 for the Operational Cold case.
Displacements M1 mirror M2 mirror

PWT PAT Difference PWT PAT Difference

Tx [microns] 720.93 720.55 0.05 % − 3944.19 − 3944.36 0.00 %

Ty [microns] 22.37 22.40 − 0.10 % 29.81 29.81 0.00 %
Tz [microns] − 1381.56 − 1378.48 0.22 % − 1136.85 − 1134.20 0.23 %
Rx [μrad] 38.12 38.07 0.11 % 3.17 3.10 2.15 %
Ry [μrad] 1025.52 1023.17 0.23 % 594.45 594.64 − 0.03 %
Rz [μrad] − 14.69 − 14.72 − 0.18 % − 10.71 − 10.76 − 0.44 %

gradients, such as the transient Max Gradient case. In this example, the
Table 6
shear forces of the joining elements between the M2 mirror and its
Maximum stresses calculated from the thermoelastic analysis of the ARIEL PLM
mechanism are shown in Table 7 for the comparison of both mapping
FEM.
methods, where the numbering of the CBUSH elements can be seen in
Part Maximum stresses [MPa] Fig. 14. Fig. 25 shows the distribution of the shear forces at this interface
Non-Operational Cold case Max Gradient case for the Max Gradient case with the temperature map transferred to the
PWT PAT PWT PAT FEM by the PWT method. It is highlighted that the generated shear
forces appear like pairs of opposite vectors for each flange, which is a
Optical Bench 4.3 8.8 20.3 19.5
Metering structure 14.6 14.6 10.2 10.5
consequence of the different thermoelastic contraction between these
M1 19.3 19.3 26.2 27.2 flanges and the supports of the mechanism connected to them.
M1 Supports 25.2 25.2 57.1 41.1 This is the example found on the entire ARIEL PLM FEM with the
M2 0.8 0.8 53.6 0.3 maximum differences in the stresses and forces between the results of
both mapping methods. But in general, because with the PWT method
differences in results between both mapping methods, in a similar way temperature jumps appear at the interfaces between the different parts,
to the displacements results of the Operational Cold case. However, the usually associated with different thermal elements of the ESATAN-TMS
appreciable gradient obtained for the transient Max Gradient case leads model, high stresses and forces are generated precisely in these in­
to greater differences in the structural results between both mapping terfaces. This fact produces the highest difference with respect to the
methods in some regions, as can be appreciated for the supports of the results from the PAT method, where thanks to the FEM thermal
M1 mirror and for the M2 mirror. It is particularly noticeable the big conductive analysis, smoother transitions of temperature are calculated
difference in stresses generated on the M2 mirror, where with the tem­ in the same interfaces, generating lower stresses and forces. Therefore,
perature map transferred by the PWT method, the big temperature jump local results at the interfaces may be different depending on the selected
between this component and the mechanism that supports it (see Fig. 23 mapping method, which is more evident for load cases with high
(left)) provokes high stresses that reach around 54 MPa at the flanges gradient of temperature. Considering the better realism and accuracy of
(see Fig. 24 (e)). However, the thermal conductive analysis of the FEM the PAT method compared to the PWT method, it is recommended the
included in the PAT method calculates a smoother temperature distri­ first one for this type of situations.
bution, reducing considerably the temperature jump at the interface
between the M2 mirror and its mechanism (see Fig. 23 (right)), gener­ 6. Conclusions
ating negligible maximum stresses on the M2 (see Fig. 24 (f)), less than 1
MPa. This work has the purpose of showing the successful application of
Regarding the generated forces on the joining elements (screws) two different temperature mapping methods on the real case of the
represented by CBUSH elements, the results shown in Table 7 indicate Payload Module of the ARIEL spacecraft to demonstrate the feasibility of
the same consequences observed in the previous results, where the se­ performing the proposed procedures for both methods. Some drawbacks
lection of the mapping method is not critical when analysing a load case such as the computational limitation for the thermal conductive analysis
with low gradient temperature, such as the Non-Operational Cold case, in NASTRAN with a huge FEM with large MPCs, and the difficulty of
but has a significant impact on the results for load cases with high associating the nodes between the thermal and the structural models
created with different modelling approaches have been solved in this

94
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

Fig. 24. Stress distribution (in Pascals) for the thermoelastic Max Gradient case.

Table 7
Shear forces generated at the M2 – M2M interface elements.
CBUSH ID Non-Operational Cold case Max Gradient case

Shear forces [N] Shear forces [N]

PWT PAT PWT PAT

5431 15.44 15.40 1863.92 9.40

5432 15.91 15.88 1864.23 9.47
5433 15.43 15.38 1861.62 9.96
5434 15.93 15.88 1861.92 10.01
5435 15.14 15.10 1881.34 8.60
5436 15.15 15.11 1881.33 8.58

work by the proposed methods.

Two existing temperature mapping methods (PWT and PAT) have
been evaluated in this work, explaining in detail the proposed schemes
for each one implemented in a dedicated software tool. Some particular
processes have been implemented to improve the versatility of their use Fig. 25. Distribution of shear forces for the M2 mirror IF for the Max Gradient
in complex models of any analysed system. One of the taken decisions case with PWT mapping.

95
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

was the selection of the nearest-node search method to associate each CRediT authorship contribution statement
structural node to its closest thermal node prior to the temperature
transfer between both models, which avoids the problems and diffi­ Andrés García-Pérez: Writing – review & editing, Writing – original
culties when working with models with different modelling approaches draft, Software, Resources, Methodology, Investigation, Formal anal­
and mismatching between their geometries. In the example of the ARIEL ysis, Data curation, Conceptualization. Alejandro Fernández-Soler:
PLM, the thermal model created in ESATAN-TMS is mainly composed of Writing – review & editing, Supervision, Software, Investigation, Formal
surface elements, while the structural FEM is created in detail with analysis. Gianluca Morgante: Software, Project administration, Inves­
volumetric elements to represent the main parts. Therefore, the over­ tigation, Formal analysis. Javier Pérez-Álvarez: Writing – review &
lapping process between both models would become very difficult or editing, Visualization, Resources, Project administration, Conceptuali­
even impossible when trying other more sophisticated methods that take zation. Gustavo Alonso: Supervision, Project administration, Funding
into account the geometry shape of each element. Therefore, the acquisition. Laura García-Moreno: Visualization, Resources. Antonio
selected nearest-node search technique is compatible with any differ­ Scippa: Software, Resources, Investigation. Daniele Gottini: Software,
ence in modelling approaches between both models. This allows struc­ Resources, Investigation. Riccardo Lilli: Software, Resources,
tural and thermal engineers to work independently on their respective Investigation.
models to optimally achieve the best representation of real physics, by
balancing adequately accuracy and computation resources. Another
Declaration of competing interest
interesting aspect implemented for the proposed flowchart of the tem­
perature mapping process is the possibility of dividing both models into
The authors declare that they have no known competing financial
different parts or groups, improving the temperature transfer between
interests or personal relationships that could have appeared to influence
the corresponding parts and reducing the execution time during this
the work reported in this paper.
process.
One of the key aspects studied in this work is the computational limit
Acknowledgements
found when performing thermal conductive analyses in NASTRAN with
models that contain several MPCs with a great quantity of independent
Authors would like to thank the overall ARIEL Consortium Team for
nodes. The results obtained from this study reveal that the computa­
this grateful collaborative work for the development of the Payload
tional resources such as the required RAM memory and the usage of hard
Module of the ARIEL spacecraft. We especially want to thank these
disk storage for the temporal files are practically proportional to the
partners (RAL, INAF, UniFI, Active Space, DTU, Sener, Admatis, CEA,
maximum number of independent nodes associated to each MPC in the
CBK and JPL) for providing us with the different models of their cor­
analysed model. This is particularly severe for FEMs with millions of
responding subsystems to integrate them into the overall thermal and
nodes, where all these nodes should be connected to MPCs necessary for
structural models of the ARIEL PLM, thereby achieving the completion
the correct calculation of the temperature map with the PAT method
of this work. Finally, we would like to recognise the excellent work
version proposed in this work. The solution found in this work is the
performed by the ESA and RAL teams in managing this project.
modification of the original set of large MPCs by set of smaller MPCs in a
two-level tree configuration, which achieves the same accuracy in the
References
results but with the significant reduction in the computational resources,
making available the successful application of the PAT method in a [1] M. Bourdeaud, J. Ponsy, S. Laborde, P. Corberand, Designing Satellites from the
complex FEM such as the ARIEL PLM. main thermo-elastic stability contributors quantification – process and tool applied
The comparison of the results in terms of temperature maps, as well on juice, in: Eur. Conf. Spacecr. Struct. Mater. Mech. Test., 2018.
[2] M. Burkhalter-Lindner, W.V. Davies, P.J. Schüngel, B. Schmid, Structural design
as the subsequent structural results such as displacements, stresses and
and verification of the Planck payload module structure, in: Eur. Conf. Spacecr.
forces between both temperature mapping methods for the ARIEL PLM Struct. Mater. Mech. Test., 2005. Noordwijk, The Netherlands.
concludes that the selection of the temperature mapping approach is not [3] V. Da Deppo, E. Pace, G. Morgante, M. Focardi, E. Pascale, G. Malaguti,
M. Terraneo, F. Zocchi, G. Bianucci, G. Micela, The primary mirror of the ARIEL
critical when analysing steady-state load cases with small gradients.
mission: study and development of a prototype, in: Eur. Planet. Sci. Congr., 2018.
Fortunately for the ARIEL mission, most of the load cases defined for the [4] S. Fransen, D. Doyle, B. Catanzaro, Opto-mechanical modeling of the herschel
STOP analysis that involve the evaluation of the optical performance space telescope at ESA/ESTEC, in: T. Andersen, A. Enmark (Eds.), Integr. Model.
correspond to situations where the excellent thermal design ensures a Complex Optomech, Syst., 2011 833604, https://doi.org/10.1117/12.915669.
[5] J.D. Johnston, J.M. Howard, G.E. Mosier, K.A. Parrish, M.A. McGinnis, A.M. Bluth,
very uniform temperature distribution throughout the entire telescope K. Kim, K.Q. Ha, Integrated modeling activities for the James Webb space
assembly, reducing possible thermoelastic distortions that can severely telescope: structural-thermal-optical analysis, in: J.C. Mather (Ed.), Opt. Infrared,
alter the correct operation of the astronomic observation. This level of Millim. Sp. Telesc, 2004, p. 600, https://doi.org/10.1117/12.551704.
[6] M. Herceg, P.S. Jørgensen, J.L. Jørgensen, Characterization and compensation of
uniformity of the temperature leads to a great similarity of the thermal thermo-elastic instability of SWARM optical bench on micro Advanced Stellar
and structural results between both mapping methods, preferring in Compass attitude observations, Acta Astronaut. 137 (2017) 205–213, https://doi.
these situations the PWT method due to its greater simplicity and speed. org/10.1016/j.actaastro.2017.04.018.
[7] S. Appel, J. Wijker, Simulation of Thermoelastic Behaviour of Spacecraft, Springer
However, the selection of the mapping method becomes more relevant International Publishing, 2022, https://doi.org/10.1007/978-3-030-78999-2_8.
when analysis a load case with significant temperature gradients be­ [8] TEC-M ESA/ESTEC, European Guidelines for Thermo-Elastic Verification, 2023.
tween some parts, causing remarkable differences between the calcu­ [9] S.M. Merkowitz, S. Conkey, W.B. Haile, W.R. KellyIII, H. Peabody, P.J. Dumont,
Structural, thermal, optical and gravitational modelling for LISA, Class, Quantum
lated results. For these situations, the greater sophistication of the PAT
Gravity 21 (2004) S603–S610, https://doi.org/10.1088/0264-9381/21/5/032.
method allows the calculation of more reliable results than those of the [10] M. Levine, J. Fanson, Advanced thermo-structural technologies for the NASA
PWT method, being consequently recommended the PAT option if the terrestrial planet finder mission, Struct. Control Heal. Monit. 13 (2006) 190–209,
https://doi.org/10.1002/stc.136.
time and computation constrains allow it.
[11] S.J. Scola, J.F. Osmundsen, L.S. Murchison, W.T. Davis, J.M. Fody, C.M. Boyer, A.
L. Cook, C.A. Hostetler, S.T. Seaman, I.J. Miller, W.C. Welch, A.R. Kosmer,
Funding Structural-Thermal-Optical-Performance (STOP) model development and analysis
of a field-widened Michelson interferometer, in: G.G. Gregory, A.J. Davis (Eds.),
Nov. Opt. Syst. Des. Optim., vol. XVII, 2014, p. 91930I, https://doi.org/10.1117/
This work has been funded by the Spanish ministry “Ministerio de 12.2061041.
Ciencia e Innovación” through project ARIEL_SCITECH with grant [12] R. Gracey, A. Bartoszyk, E. Cofie, B. Comber, G. Hartig, J. Howard, D. Sabatke,
number PID2021-125627OB-C33. G. Wenzel, R. Ohl, Structural, thermal, and optical performance (STOP) modeling
and results for the James Webb Space Telescope integrated science instrument
module, in: G.Z. Angeli, P. Dierickx (Eds.), Model. Syst. Eng. Proj. Manag. Astron,
VI, 2016, p. 99111A, https://doi.org/10.1117/12.2233641.

96
A. García-Pérez et al. Acta Astronautica 223 (2024) 77–97

[13] F. Navarro-Medina, A.E. Oudijk, M.B. Henriksen, U. Garcia-Luis, A. Gomez-San Sci. Congr. 2022, 2022, pp. 16–18, https://doi.org/10.5194/epsc2022-861.
Juan, T.A. Johansen, Structural thermal optical performance (STOP) analysis and Granada, Spain.
experimental verification of an hyperspectral imager for the HYPSO CubeSat, Opt. [29] A. Tozzi, A. Brucalassi, R. Canestrari, P. Chioetto, C. Del Vecchio, L. Carbonaro,
Lasers Eng. 173 (2024) 107919, https://doi.org/10.1016/j. F. Cortecchia, E. Diolaiti, P. Eccleston, G. Falcini, D. Ferruzzi, D. Gottini,
optlaseng.2023.107919. E. Guerriero, M. Iuzzolino, R. Lilli, M. Lombini, G. Malaguti, G. Micela, F. Miceli,
[14] M. Cho, A. Corredor, K. Vogiatzis, G. Angeli, Thermal modeling of the TMT G. Morgante, E. Pace, E. Pascale, R. Piazzolla, G. Preti, M. Salatti, A. Scippa,
telescope, in: T. Andersen, A. Enmark (Eds.), Integr. Model. Complex Optomech. G. Tinetti, E. Tommasi, D. Vernani, P. Zuppella, Toward ARIEL’s primary mirror,
Syst., 2011, p. 83360V, https://doi.org/10.1117/12.915634. in: L.E. Coyle, M.D. Perrin, S. Matsuura (Eds.), Sp. Telesc. Instrum. 2022 Opt.
[15] ECSS, ECSS-E-ST-31-04C, Space engineering, Exchange of thermal analysis data, Infrared, Millim. Wave, SPIE, 2022, p. 193, https://doi.org/10.1117/12.2628906.
ESA-ESTEC Requirements and Standards Division, Noordwijk, The Netherlands [30] P. Chioetto, P. Zuppella, V. Da Deppo, E. Pace, G. Morgante, L. Terenzi, D. Brienza,
(2018). N. Missaglia, G. Bianucci, S. Spinelli, E. Guerriero, M. Rossi, C. Grèzes-Besset,
[16] ECSS, ECSS-E-ST-32-03C, Space engineering, Structural finite element models. C. Bondet, G. Chauveau, C. Porta, G. Malaguti, G. Micela, Qualification of the
ESA-ESTEC Requirements and Standards Division, 2008. Noordwijk, The thermal stabilization, polishing and coating procedures for the aluminum telescope
Netherlands. mirrors of the ARIEL mission, Exp. Astron. 53 (2022) 885–904, https://doi.org/
[17] M. Koot, S. Appel, S. Simonian, Temperature mapping for structural thermo-elastic 10.1007/s10686-022-09852-x.
analyses ; method benchmarking, in: Proc. Eur. Conf. Spacecr. Struct. Mater. Mech. [31] E. Guerriero, P. Chioetto, A. Tozzi, P. Zuppella, R. Canestrari, A. Brucalassi,
Test., 2018. Noordwijk, The Netherlands. M. Iuzzolino, D. Ferruzzi, A. Scippa, C. Del Vecchio, G. Falcini, L. Carbonaro,
[18] M. Koot, S. Appel, S. Simonian, Thermal conductor generation for thermal and G. Morgante, F. Cortecchia, E. Diolaiti, P. Eccleston, M. Lombini, G. Malaguti,
thermo-elastic analysis using a finite element model and sinas, in: Proc. Eur. Conf. G. Micela, E. Pace, E. Pascale, R. Piazzolla, G. Preti, M. Salatti, G. Tinetti,
Spacecr. Struct. Mater. Mech. Test., Noordwijk, The Netherlands, n.d. E. Tommasi, Heat treatment procedure of the aluminium 6061-T651 for the Ariel
[19] L.A. Tse, Z. Chang, R.P. Somawardhana, E. Slimko, Structural, thermal, and optical telescope mirrors, in: L.E. Coyle, M.D. Perrin, S. Matsuura (Eds.), Sp. Telesc.
performance (STOP) modeling and analysis for the surface water and ocean Instrum. 2022 Opt. Infrared, Millim, Wave, SPIE, 2022, p. 39, https://doi.org/
topography mission, in: 48th Int. Conf. Environ. Syst., 2018. Alburquerque, NM. 10.1117/12.2628178.
[20] E. Heesel, T. Weigel, P. Lochmatter, E. Rugi Grond, Coupled thermo-elastic and [32] D. Gottini, E. Pace, A. Tozzi, G. Bianucci, A. Bocchieri, D. Brienza, A. Brucalassi,
optical performance analyses of a reflective baffle for the BepiColombo laser R. Canestrari, L. Carbonaro, P. Chioetto, F. Cortecchia, C. Del Vecchio, E. Diolaiti,
altimeter (BELA) receiver, in: J. Costeraste, E. Armandillo, N. Karafolas (Eds.), Int. P. Eccleston, S. Fahmy, D. Ferruzzi, C. Galy, G. Grisoni, E. Guerriero, J.-P. Halain,
Conf. Sp. Opt. — ICSO 2008, SPIE, 2017, p. 68, https://doi.org/10.1117/ M.-L. Hellin, M. Iuzzolino, D. Jollet, M. Lombini, G. Malaguti, G. Micela,
12.2308252. N. Missaglia, G. Morgante, L. Mugnai, L. Naponiello, E. Pascale, R. Piazzolla,
[21] C. Blaurock, M. McGinnis, K. Kim, G.E. Mosier, Structural-thermal-optical G. Preti, S. Roose, M. Salatti, J.-C. Salvignol, A. Scippa, L. Terenzi, G. Tinetti,
performance (STOP) sensitivity analysis for the James Webb space telescope, in: M. E. Tommasi Di Vigano, P. Zuppella, FEA testing the pre-flight Ariel primary mirror,
A. Kahan (Ed.), Opt. Model. Perform. Predict, II, 2005, p. 58670V, https://doi.org/ in: L.E. Coyle, M.D. Perrin, S. Matsuura (Eds.), Sp. Telesc. Instrum. 2022 Opt.
10.1117/12.618697. Infrared, Millim, Wave, SPIE, 2022, p. 195, https://doi.org/10.1117/12.2629815.
[22] S. Appel, SlNAS IV User ’ S Manual, Fokker Space B.V, Leiden (The Netherlands), [33] G. Morgante, L. Terenzi, D. D’Ascanio, P. Eccleston, M. Crook, T. Hunt, V. Da
1998. Deppo, M. Focardi, G. Micela, G. Malaguti, E. Pace, G. Tinetti, Thermal
[23] S. Appel, Using Pysinas, 2024. architecture of the ESA ARIEL payload, in: H.A. MacEwen, M. Lystrup, G.G. Fazio,
[24] V. Da Deppo, M. Focardi, K. Middleton, G. Morgante, E. Pascale, S. Grella, E. Pace, N. Batalha, E.C. Tong, N. Siegler (Eds.), Sp. Telesc. Instrum. 2018 Opt. Infrared,
R. Claudi, J. Amiaux, J. Colomé Ferrer, T. Hunt, M. Rataj, C. Sierra-Roig, I. Ficai Millim, Wave, SPIE, 2018, p. 154, https://doi.org/10.1117/12.2313153.
Veltroni, P. Eccleston, G. Micela, G. Tinetti, An afocal telescope configuration for [34] G. Morgante, L. Terenzi, L. Desjonqueres, P. Eccleston, G. Bishop, A. Caldwell,
the ESA ARIEL mission, CEAS Sp. J. 9 (2017) 379–398, https://doi.org/10.1007/ M. Crook, R. Drummond, M. Hills, T. Hunt, D. Rust, L. Puig, T. Tirolien, M. Focardi,
s12567-017-0175-3. P. Zuppella, W. Holmes, J. Amiaux, M. Czupalla, M. Rataj, N.C. Jessen, S.
[25] E. Pascale, N. Bezawada, J. Barstow, J.-P. Beaulieu, N. Bowles, V. Coudé du M. Pedersen, E. Pascale, E. Pace, G. Malaguti, G. Micela, The thermal architecture
Foresto, A. Coustenis, L. Decin, P. Drossart, P. Eccleston, T. Encrenaz, F. Forget, of the ESA ARIEL payload at the end of phase B1, Exp. Astron. 53 (2022) 905–944,
M. Griffin, M. Güdel, P. Hartogh, A. Heske, P.-O. Lagage, J. Leconte, P. Malaguti, https://doi.org/10.1007/s10686-022-09851-y.
G. Micela, K. Middleton, M. Min, A. Moneti, J.C. Morales, M. Ollivier, E. Pace, [35] J. Lee, H. Oh, S. Jeong, Investigation of neon–nitrogen mixed refrigerant
G. Pilbratt, L. Puig, M. Rataj, T. Ray, I. Ribas, M. Rocchetto, S. Sarkar, F. Selsis, Joule–Thomson cryocooler operating below 70K with precooling at 100K,
W. Taylor, J. Tennyson, G. Tinetti, D. Turrini, B. Vandenbussche, O. Venot, I. Cryogenics 61 (2014) 55–62, https://doi.org/10.1016/j.cryogenics.2014.02.006.
P. Waldmann, P. Wolkenberg, G.S. Wright, M.-R. Zapatero Osorio, T. Zingales, [36] J. Martignac, J. Amiaux, M. Berthé, C. Cara, C. Delisles, A. Direk, L. Dumaye,
A. Papageorgiou, L. Mugnai, The ARIEL space mission, in: H.A. MacEwen, J. Fontignie, A. Goestschy, B. Horeau, N. Hurtado, D.-D. Huyn, G. Kaszubiak, P.-
M. Lystrup, G.G. Fazio, N. Batalha, E.C. Tong, N. Siegler (Eds.), Sp. Telesc. Instrum. O. Lagage, I. Le Mer, M. Lortholary, V. Moreau, S. Mouzali, P. Mulet, F. Nico,
2018 Opt. Infrared, Millim. Wave, SPIE, 2018, p. 16, https://doi.org/10.1117/ T. Pichon, L. Provost, D. Renaud, V. Schwartz, M. Talvard, T. Tourrette, F. Visticot,
12.2311838. A. Arhancet, D. Bachet, N. Berton, M. Lacroix, H. Le Provost, O. Tellier,
[26] E. Pace, A. Tozzi, M. Adler Abreu, G. Alonso, B. Barroqueiro, G. Bianucci, A. Philippon, C. De Jabrun, J.-P. Dubois, F. Langlet, D. Le Claire, B. Lecomte,
A. Bocchieri, D. Brienza, A. Brucalassi, M. Burresi, R. Canestrari, L. Carbonaro, M. Ollivier, S. Tosti, V. Lapeyrère, M. Bonafous, J. Parisot, J.-M. Réess,
J. Castanheira, P. Chioetto, J. Colomé Ferrer, C. Compostizo, F. Cortecchia, D. Zegadanin, J.-P. Beaulieu, V. Batista, P. Drossart, S. Fahmy, D. Jollet, L. Puig,
F. D’Anca, C. Del Vecchio, E. Diolaiti, P. Eccleston, S. Fahmy, A. Fernandez Soler, T. Tirolien, P. Danto, G. Hervet, O. Maisonnave, P. Eccleston, G. Bishop,
D. Ferruzzi, M. Focardi, S. Freitas, C. Galy, A. Garcia Perez, D. Gottini, S. Grella, R. Drumond, A. Caldwell, M. Caldwell, L. Desjonqueres, M. Whalley, E. Pascale,
G. Grisoni, E. Guerriero, J.-P. Halain, M.-L. Hellin, L. Ianni, M. Iuzzolino, D. Jollet, G. Morgante, M. Focardi, E. Pace, A.-M. Di Giorgio, AIRS: ARIEL IR spectrometer
M. Lombini, R. Machado, G. Malaguti, A. Mazzoli, G. Micela, F. Miceli, status, in: L.E. Coyle, M.D. Perrin, S. Matsuura (Eds.), Sp. Telesc. Instrum. 2022
G. Mondello, G. Morgante, L. Mugnai, L. Naponiello, V. Noce, E. Pascale, J. Perez Opt. Infrared, Millim, Wave, SPIE, 2022, p. 37, https://doi.org/10.1117/
Alvarez, R. Piazzolla, C. Pompei, G. Preti, S. Roose, M. Salatti, J.-C. Salvignol, 12.2628920.
A. Scippa, C. Serre, C. Simoncelli, F. Teixeira, L. Terenzi, G. Tinetti, L. Tommasi, [37] A. García-Pérez, G. Alonso, A. Gómez-San-Juan, J. Pérez-Álvarez, Thermoelastic
E. Tommasi Di Vigano, B. Vandenbussche, D. Vernani, P. Zuppella, The telescope evaluation of the payload module of the ARIEL mission, Exp. Astron. 53 (2022)
assembly of the Ariel space mission, in: L.E. Coyle, M.D. Perrin, S. Matsuura (Eds.), 831–846, https://doi.org/10.1007/s10686-021-09704-0.
Sp. Telesc. Instrum. 2022 Opt. Infrared, Millim. Wave, SPIE, 2022, p. 36, https:// [38] J.N. Loughhead, G.A. Miles, J.R. Turner, The ESATAN thermal network analysis
doi.org/10.1117/12.2629432. computer program. https://doi.org/10.4271/840967, 1984.
[27] P. Chioetto, A. Tozzi, A. Brucalassi, D. Ferruzzi, A. Caldwell, M. Caldwell, [39] V.L. Genberg, G.J. Michels, Integrated optomechanical STOP analysis linking
F. Cortecchia, E. Diolaiti, P. Eccleston, E. Guerriero, M. Lombini, G. Malaguti, nastran to Zemax, CODEV, and OSLO with SigFit (2017) 1–27.
G. Micela, E. Pace, E. Pascale, R. Piazzolla, G. Preti, M. Salatti, G. Tinetti, [40] A. Fernández-Soler, G. Morgante, A. García-Pérez, J. Pérez-Álvarez, Preliminar
E. Tommasi, P. Zuppella, Preliminary analysis of ground-to-flight mechanical ariel decontamination thermal analysis, in: Eur. Sp. Therm. Eng. Work., 2023.
tolerances of the Ariel mission telescope, in: L.E. Coyle, M.D. Perrin, S. Matsuura [41] S. Tuttle, L. Vaillon, U. Johann, O. Wallner, K. Ergenzinger, EUCLID mission
(Eds.), Sp. Telesc. Instrum. 2022 Opt. Infrared, Millim. Wave, SPIE, 2022, p. 220, design, in: N. Kadowaki (Ed.), Int. Conf. Sp. Opt. — ICSO 2010, 2017, p. 90,
https://doi.org/10.1117/12.2628900. https://doi.org/10.1117/12.2309226. SPIE.
[28] G. Alonso, A. Garcia-Perez, J. Perez-Alvarez, A. Fernandez-Soler, The mechanical ,
structural and thermal design of the Telescope Assembly of ARIEL, in: Proc. Eur.

97