OUFTI-1 Nanosatellite

University of Liège
Faculty of Applied Sciences

Aerospace and Mechanical Engineering Department
OUFTI-1 nanosatellite:
Dynamic analysis and qualification testing
Thesis submitted in partial fulfilment

of the requirements for the degree of
Master in Aerospace Engineering
by
BERTHA Mathieu
Supervisor: Prof. Jean-Claude GOLINVAL
Academic Year 2010–2011

Abstract
OUFTI–1, standing for Orbital Utility For Telecommunication Innovation,

is a nanosatellite in the CubeSat standard developed by a student team at
the University of Liège.
The first objective of the project is above all to provide a first experi-
ence to students in the design of a special mission. OUFTI–1 also carries
a principal and two secondary payloads. The first one is a repeater for the
D-STAR protocol, a new communication protocol for amateur radio. The
two secondary payloads are an innovative digitally-controlled power supply
developed with Thales Alenia Space ETCA, and new high-efficiency solar
cells developed by Azur Space.
To operate in orbit, OUFTI–1 has first to pass through the launch phase
and its harsh dynamic environment. The satellite must therefore prove its
capability to withstand quasi-static and dynamic loads encountered during
these few minutes.
Therefore, a finite element model of the current design of the satellite

is built. Quasi-static and sine vibrations analyses performed on this model
result in positive margin of safety concerning the quasi-static and sine vibra-
tions requirements.
Finally, a testing procedure is drawn up to further correlate with the

current model.
Contents
Acronyms xiii
Acknowledgements 1
Introduction 2
1 VEGA maiden flight 9
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.1 VEGA Launch Vehicle . . . . . . . . . . . . . . . . . . . . . . 10
1.2 Launch environment and ESA requirements . . . . . . . . . . 11
1.2.1 Lower limit in frequency requirement . . . . . . . . . . 13
1.2.2 Quasi-static loads requirements . . . . . . . . . . . . . 14
1.2.3 Sine vibration requirements . . . . . . . . . . . . . . . 15
1.2.4 Random vibration requirements . . . . . . . . . . . . . 19
1.2.5 Shock load requirements . . . . . . . . . . . . . . . . . 21
i
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2 Finite element modelling and updating 24
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.1 Previous finite element models . . . . . . . . . . . . . . . . . . 26
2.1.1 The CubeSat Kit structure . . . . . . . . . . . . . . . . 27
2.1.2 The battery support . . . . . . . . . . . . . . . . . . . 28
2.2 Finite element modelling of electronic cards . . . . . . . . . . 29
2.2.1 Choice of the element type in the modelling . . . . . . 30
2.2.2 Consideration of electronic components . . . . . . . . . 32
2.3 Experimental modal analysis and update process . . . . . . . 34
2.3.1 First step: tests . . . . . . . . . . . . . . . . . . . . . . 34
2.3.2 Second step: identification phase . . . . . . . . . . . . 40
2.4 Modelling of the OBC2 card . . . . . . . . . . . . . . . . . . . 41
2.4.1 Modelling of the PCB and PC104 . . . . . . . . . . . . 41
2.4.2 Experimental modal analysis of the unpopulated OBC2 44
2.4.3 Update of the unpopulated OBC2 model . . . . . . . . 47
2.4.4 Characterization of the PC104 connection . . . . . . . 50
2.4.5 Consideration of the electronic components . . . . . . . 52
2.5 Modelling of the EPS card . . . . . . . . . . . . . . . . . . . . 56
2.5.1 Modelling of the PCB . . . . . . . . . . . . . . . . . . 56
2.6 Modelling of the xEPS card . . . . . . . . . . . . . . . . . . . 62
2.6.1 Modelling of the PCB . . . . . . . . . . . . . . . . . . 62
ii
2.7 Modelling of the FM430 . . . . . . . . . . . . . . . . . . . . . 69
2.8 Global finite element model of the satellite . . . . . . . . . . . 74
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3 Numerical testing 80
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1 Modal analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1.1 Boundary conditions . . . . . . . . . . . . . . . . . . . 82
3.2 Quasi static accelerations . . . . . . . . . . . . . . . . . . . . . 85
3.3 Sinusoidal vibrations . . . . . . . . . . . . . . . . . . . . . . . 89
3.3.1 Boundary conditions and applied loads . . . . . . . . . 91
3.3.2 X direction loading . . . . . . . . . . . . . . . . . . . . 92
3.3.3 Y and Z direction loadings . . . . . . . . . . . . . . . . 95
3.4 Random vibrations . . . . . . . . . . . . . . . . . . . . . . . . 98
3.4.1 Transformation of the prescribed PSD . . . . . . . . . 99
3.4.2 Adding SPECTRAL commands . . . . . . . . . . . . . 100
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4 Testing procedure 102
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.1 Needed facilities . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.2 Test set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.2.1 Accelerometers placement . . . . . . . . . . . . . . . . 106
iii
4.3 Sinusoidal and random acceleration tests . . . . . . . . . . . . 106
4.3.1 Objectives of these tests . . . . . . . . . . . . . . . . . 106
4.3.2 Test sequence . . . . . . . . . . . . . . . . . . . . . . . 107
4.3.3 Pass criteria . . . . . . . . . . . . . . . . . . . . . . . . 107
4.4 Shock tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Conclusion 108
iv
List of Figures
1 Small satellites [1]. . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Pumpkin’s CubeSat structures corresponding to 1, 1.5, 2 and

3–units CubeSats, respectively for 10 × 10 × 15, 20 and 30 cm. 4
3 Illustration of the P–POD structure . . . . . . . . . . . . . . . 5
4 Exploded view of the OUFTI–1 CubeSat. . . . . . . . . . . . . 7
1.1 Vettore Europeo di Generazione Avanzata (VEGA) launch

profiles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2 Encountered types of loads [2]. . . . . . . . . . . . . . . . . . . 13
1.3 VEGA coordinate system . . . . . . . . . . . . . . . . . . . . 15
1.4 Acceleration envelope. . . . . . . . . . . . . . . . . . . . . . . 17
1.5 CubeSat coordinate system [3]. . . . . . . . . . . . . . . . . . 17
1.6 Qualification and acceptance X sine acceleration levels. . . . . 19
1.7 Qualification and acceptance Y sine acceleration levels. . . . . 20
1.8 Qualification ans acceptance Z sine acceleration levels. . . . . 21
v
1.9 Random vibration levels. . . . . . . . . . . . . . . . . . . . . . 22
1.10 Shock level for qualification. . . . . . . . . . . . . . . . . . . . 23
2.1 Model of the external structure by Gauthier Pierlot. . . . . . . 27
2.2 Correlation bewteen Finite Element Analysis (FEA) and Ex-

perimental Modal Analysis (EMA) mode shapes of the exter-
nal structure [4]. . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 FE model of the batery support. . . . . . . . . . . . . . . . . . 29
2.4 Example of an unpopulated and a populated electronic card. . 29
2.5 Geometrical dimensions of the electronic cards (in mm). . . . 30
2.6 Shell and volume models for comparison. . . . . . . . . . . . . 31
2.7 Good matching between shell and volume elements meshed

models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.8 Smearing methods [5]. . . . . . . . . . . . . . . . . . . . . . . 33
2.9 Last year’s experimental set-up [3]. . . . . . . . . . . . . . . . 35
2.10 Correlation with the last year’s experimental set-up [3]. . . . . 35
2.11 New free-free set-ups. . . . . . . . . . . . . . . . . . . . . . . . 36
2.12 Set-up with one card. . . . . . . . . . . . . . . . . . . . . . . . 37
2.13 Set-up with two cards. . . . . . . . . . . . . . . . . . . . . . . 37
2.14 Measurment points map. . . . . . . . . . . . . . . . . . . . . . 39
2.15 Measurement map used to point the laser beam. . . . . . . . . 40
2.16 Unpopulated and populated OBC2 card . . . . . . . . . . . . 41
2.17 Determination of masses by several weighing. . . . . . . . . . . 42
2.18 Finite element model of the PCB and the PC104 connector. . 44
2.19 Shading factor for the exponential moving average. . . . . . . 45
vi
2.20 Stabilisation diagram of the unpopulated OBC2. . . . . . . . . 46
2.21 Argand diagram of the firs mode of the first “double peak ”. . 46
2.22 Mode at ≈ 1400 Hz at the 30th order. . . . . . . . . . . . . . . 47
2.23 Correlation between numerical and experimental modes of the

OBC2’s PCB. . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

OBC2’s PCB (first update). . . . . . . . . . . . . . . . . . . . 49

OBC2’s PCB (second update). . . . . . . . . . . . . . . . . . . 50
2.26 Modelling of the PC104 connection pins. . . . . . . . . . . . . 51
2.27 Cerrelation for a couple of unpopulated OBC2 cards. . . . . . 52
2.28 Similar numerical mode shapes examples. . . . . . . . . . . . . 53
2.29 Stabilisation diagram of the SSI identification on the OBC2

card measurements. . . . . . . . . . . . . . . . . . . . . . . . . 54
2.30 Correlation between numerical and experimental results. . . . 55
2.31 3th and 5th experimental mode shapes. . . . . . . . . . . . . . 55
2.32 Unpopulated and populated xEPS card . . . . . . . . . . . . . 56
2.33 Correlation between numerical and experimental results for

the unpopulated EPS card . . . . . . . . . . . . . . . . . . . . 58
2.34 FE model of the EPS card . . . . . . . . . . . . . . . . . . . . 59
2.35 Stabilisation diagram from the EPS identification . . . . . . . 60
2.36 Correlation bewteen numerical and experimantal modes of the

EPS card . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.37 Unpopulated and populated xEPS card . . . . . . . . . . . . . 63
2.38 Stabilisation diagram for the PCB of the xEPS card . . . . . . 64
vii
2.39 Correlation between the initial of the PCB of the xEPS com-
pared to the experimental modes . . . . . . . . . . . . . . . . 65
2.40 Comparison between the 3th and 4th experimental mode shapes
with the 3th numerical one . . . . . . . . . . . . . . . . . . . . 65
2.41 Correlation between the updated model of the PCB of the

xEPS compared to the experimental modes . . . . . . . . . . . 65
2.42 Finite element model of the xEPS card . . . . . . . . . . . . . 67
2.43 Stabilisation diagram of the xEPS card . . . . . . . . . . . . . 67
2.44 MAC matrix between numerical and experimental modes of

the xEPS card . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.45 Nmerical modes involving only the bottom card . . . . . . . . 69
2.46 FM430 Flight Module . . . . . . . . . . . . . . . . . . . . . . 69
2.47 FE model of the FM430 card . . . . . . . . . . . . . . . . . . 72
2.48 Stabilisation diagram of the FM430 . . . . . . . . . . . . . . . 72
2.49 Correlation between the FM430 model and experimental results 73
2.50 Correlation between the updated FM430 model and experi-

mental results . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.51 Starting point for the global model . . . . . . . . . . . . . . . 76
2.52 Equivalent connection between cards and feet added to the

model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.53 Clearance between feet and the chassis . . . . . . . . . . . . . 78
2.54 Global model of the satellite . . . . . . . . . . . . . . . . . . . 79
3.1 Rails on the structure. . . . . . . . . . . . . . . . . . . . . . . 82
3.2 Graphical representation of the ten first modes. . . . . . . . . 84
3.3 Worst cases in longitudinal loading. . . . . . . . . . . . . . . . 86
viii
3.4 Load cases for the QSL study. . . . . . . . . . . . . . . . . . . 87
3.5 Stress concentration in the linked elements between the chassis

and the End plate. . . . . . . . . . . . . . . . . . . . . . . . . 89
3.6 Maximum displacements in the worst load case. . . . . . . . . 91
3.7 Maximum displacement during the sine vibration analysis in

the Xdirection. . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.8 Global displacements under sine vibrations along the X direc-

tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.9 Displacements of the EPS card under sine vibrations along the
X direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.10 Stress repartition inside the structure under sine acceleration

along the X direction. . . . . . . . . . . . . . . . . . . . . . . 94
3.11 Stress repartition inside the screws under sine acceleration

along the X direction. . . . . . . . . . . . . . . . . . . . . . . 95
3.12 Stress concentration in the battery support under sine accel-

eration along the X direction . . . . . . . . . . . . . . . . . . 96
3.13 Stress concentration in the PCBs under sine acceleration along

the X direction . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.14 Internal displacement under sine loading along the Y axis . . . 97
3.15 Internal displacement under sine loading along the Z axis . . . 97
3.16 Prescribed PSD and PSD required by SPECTRAL . . . . . . 99
4.1 Shaker installed in the laboratory . . . . . . . . . . . . . . . . 104
4.2 Jumper components to be removed . . . . . . . . . . . . . . . 105
4.3 Test-POD configuration for X, Y and Z directions testing . . 105
4.4 Accelerometers placement for X and Y axes loading . . . . . . 106
ix
List of Tables
2 Small satellites categories [1]. . . . . . . . . . . . . . . . . . . 2
1.1 Quasi-static loads during the launch. . . . . . . . . . . . . . . 16
1.2 Qualification and acceptance X sine acceleration levels. . . . . 18
1.3 Qualification and acceptance sine Y acceleration levels. . . . . 19
1.4 Qualification and acceptance Z sine acceleration levels. . . . . 20
1.5 Random vibration levels. . . . . . . . . . . . . . . . . . . . . . 21
1.6 Shock level for qualification. . . . . . . . . . . . . . . . . . . . 22
2.1 Numerical and experimental external structure frequencies. . . 28
2.2 Frequencies calculated with the two models. . . . . . . . . . . 32
2.3 First set of eigen frequencies. . . . . . . . . . . . . . . . . . . . 44
2.4 Natural frequencies and damping ratios of the unpopulated

OBC2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.5 OBC2’s PCB frequencies comparison . . . . . . . . . . . . . . 49
2.6 Measured frequencies and damping ratios. . . . . . . . . . . . 54
x
2.7 General comparison of the OBC2 results. . . . . . . . . . . . . 57
2.8 Measured frequencies and damping ratios of the EPS’s PCB . 58
2.9 List of considered components in the EPS model . . . . . . . . 59
2.10 Measured frequencies and damping ratios for the EPS card . . 60
2.11 Summary tabular of the EPS results . . . . . . . . . . . . . . 62
2.12 Electronic components considered in the xEPS model . . . . . 66
2.13 Frequencies and damping ratios from the identification process

of the xEPS card . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.14 Summary tabular of the xEPS results . . . . . . . . . . . . . . 70
2.15 Electonic components considered in the FE model of the FM430

card . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.16 Frequencies and damping ratios from the identification process

of the FM430 . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.17 Summary tabular of the FM430 results . . . . . . . . . . . . . 75
3.1 Ten first modes with their frequency and description. . . . . . 83
3.2 Yield limits and safety factors used in the quasi-static analyses. 88
3.3 Summarise of the margins of safety for the eight load cases
considered. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
xi
Acronyms
ASI Italian Space Agency.

AVUM Attitude and Vernier Upper Module.
Cal Poly California Polytechnic State University.

CDS CubeSat Design Specification.
CMIF Complex Mode Identification Function.
COM Communication.
COTS Commercially Off-The-Shelf.
CSK CubeSat Kit.
D–STAR Digital Smart Technology for Amateur Radio.

DoF Degree of Freedom.
EMA Experimental Modal Analysis.

EPS Electrical Power Supply.
ESA European Space Agency.
FEA Finite Element Analysis.

FM430 FM430 Flight Module.
FRF Frequency Response Function.
GFRP Glass Fibre Reinforced Plastic.
ICD Interface Control Document.
xii
Isp Specific Impulse.
ITD Ibrahim Time Domain.
LARES LAser RElativity Satellite.

LDV Laser Doppler Velocimeter.
LSCE Least Squares Complex Exponential.
LSFD Least Squares Frequency Domain.
LV Launch Vehicle.
MAC Modal Assurance Criterion.

MoS Marging of Safety.
OUFTI Orbital Utility For Telecommunication Inno-

vation.
P-POD Poly-PicoSatellite Orbital Deployer.

PCB Printed Circuit Board.
PSD Power Spectral Density.
QSL Quasi Static Loads.
RMS Root Mean Square.
SF Safety Factor.
SRS Shock Response Spectrum.
SSI Stochastic Subspace Identification.
ULg University of Liège.
VEGA Vettore Europeo di Generazione Avanzata.
xEPS Experimental Electrical Power Supply.
xiii
Acknowledgements
First of all, I wish to acknowledge my supervisor, Professor Jean-Claude

Golinval, for its support throughout the realisation of this work.
I would also like to thank Professors Gaëtan Kerschen and Pierre

Rochus, Amandine Denis and Daniel Simon for the attention they will
bring to this work in the coming weeks.
Many thanks go to the OUFTI–1 System team for their involvement in

this project.
I would like to show my gratitude to Frédéric Marin for its precious

advices in the use of the Samcef software, and to Jean-Philippe Noël and
Ludovic Renson for their advices in modal identification.
Finally, I wish to acknowledge my nearest and dearest for their support

during my studies and the realisation of this work.
1
Introduction
Since Sputnik 1 in 1957, thousands of satellites have been launched and

are nowadays used in our everyday life. With time, some of them have
become increasingly large, massive, powerful, and sophisticated. However,
with the increase of these characteristics, the cost of a whole project (design,
launch, and orbital operations) risen too. Therefore, small satellites meet
some success for lower-cost spatial missions. According to [1], spacecraft for
which the mass is less than 1000 kg can be considered as small satellites.
Furthermore, this family of satellites can also be distinguished between four
categories reported in Table 2 and illustrated in Figure 1.
Class Mass (kg) Cost (£ M)

Conventional large satellite > 1000 > 100
Conventional small satellite 500 − 1000 25 − 100
Minisatellite 100 − 500 7 − 25
Microsatellite 10 − 100 1−7
Nanosatellite 1 − 10 0.1 − 1
Picosatellite <1 < 0.1
Table 2: Small satellites categories [1].
Small satellites have the advantage to reduce the part of the cost due to
launch operations, which is proportional to its mass. Indeed, they can be
launched besides a second payload in the launcher and so share the launch
expenses. The choice of a smaller Launch Vehicle (LV) can also be made in
the same perspective of cost reduction. Finally, for the smallest spacecrafts
2
Figure 1: Small satellites [1].
(nano- and picosatellites), the launch can be done beside a primary mission
that takes on the major part of the expenses.
In addition to the savings on the launch operation, the design phase is

also simpler and faster (12–18 months). Furthermore, the use of Commer-
cially Off-The-Shelf (COTS) components which are not initially designed for
applications in the harsh space environment can also dramatically reduce the
cost design.
The CubeSat standard
Designing a space mission, even in a small satellite approach, is still cost pro-
hibitive. Therefore, in order to facilitate the access to space to universities,
high schools and some private firms, the CubeSat standard was created. In
1999, California Polytechnic State University (Cal Poly) and the Space Sys-
3
tems Development Laboratory of the Stanford University created a standard
of a 10 cm cubic satellite. This is what is called a “One-Unit (1U)” CubeSat.
For slightly greater payloads, 1.5, 2, and 3-units CubeSat formats, also
exist as shown in Figure 2, representing the structures marketed by Pumpkin,
Inc. 1 . According to the previous classification, CubeSats are situated in the
bottom of the range of the nanosatellites family.
Figure 2: Pumpkin’s CubeSat structures corresponding to 1, 1.5, 2 and 3–

units CubeSats, respectively for 10 × 10 × 15, 20 and 30 cm.
The CubeSat standard is very suitable for educative projects with small
technical or scientific missions. Indeed, whereas the complete design of a
general space mission is long and expensive, the CubeSat standard allows
a significant time saving because of its association with standardised de-
ployment systems, such as the Poly-PicoSatellite Orbital Deployer (P-POD).
P-PODs are structures that can contain three 1-unit CubeSats, or the equiv-
alent (two 1.5 or one 3-units CubeSats), and put them into orbit. Figure 3
shows the structure of a P-POD.
Besides the compatible deployment system, a multitude of components,

equipments, or even hole subsystems marketed by various societies also exist.
In a such way, designer teams can concentrate themselves only on particular
subsystems instead of designing all the parts of the satellite.
To ensure the safety of the LV and the primary payload, the CubeSat
providers have to fulfil the CubeSat Design Specification (CDS) in addition
to the LV requirements.
1
http://www.cubesatkit.com
4
Figure 3: Illustration of the P–POD structure
OUFTI–1
Born in 2007, OUFTI–1 will be the first nano satellite designed at the Uni-
versity of Liège (ULg). Its name stand for Orbital Utility For Telecommuni-
cation Innovation (OUFTI) because of its principal payload which is a space
repeater for the Digital Smart Technology for Amateur Radio (D–STAR)
communication protocol.
The aim of this project is first of all to be an educative project. Through

their master thesis, several student teams have succeeded, and other will
follow, to lead to a functional satellite in orbit.
In addition to the main D–STAR payload, two secondary payloads will

take place aboard OUFTI–1. The first one is an innovative digitally-controlled
electrical power system developed with Thales Alenia Space ETCA2 (a major
electronic systems company for space application). The goal of this payload
is to study its behaviour in space environment.
The second secondary payload is a set of new high-efficiency solar cells

developed, by AzurSpace3 , shared on five solar panels. As for the Experi-
mental Electrical Power Supply (xEPS), the goal is to test this technology
in space environment.
2
www.thalesgroup.com
3
www.azurspace.com
5
Thanks to the first generation of students who defended the project,
OUFTI–1 is one of the nine CubeSats selected by the ESA Educational Of-
fice to take place aboard the VEGA maiden flight besides the Italian primary
payload : LAser RElativity Satellite (LARES).
General overview of the satellite
In order to help the reader, a non exhaustive list of particular elements

frequently brought up throughout the document is drawn up hereafter:
The CubeSat Kit (CSK) Structure: it is a cubic aluminium structure

supporting all the internal elements. It is made up of three parts:
• the End Plate: aluminium structure of the top face of the CSK
structure,
• the Chassis: aluminium structure forming the four lateral faces of
the CSK structure,
• the Base Plate: aluminium structure of the bottom face of the
CSK structure.
All these parts are soldered by screws.

Electronic cards and PC104 connector: aboard OUFTI–1, five electronic
cards are present. They are all designed on the same model:
• a PCB on which the elements of each card are soldered,
• a common connector: the PC104 connector. The PC104 connector
is a rectangular plastic volume in which 104 holes are dug and
connected to 104 connection pins under the component.
All the electronic are stacked up and interconnected through these
PC104 connectors.
The battery support: aluminium structure designed to support the two
batteries.
Endless screws: four screws that travel the structure along its height. They
are fixed directly on the Base Plate and maintained to the top of the
chassis with four Midplane Standoffs. These four screws cross all
the internal elements in their four corners.
6
Figure 4: Exploded view of the OUFTI–1 CubeSat.
Spacers: they are a series of aluminium cylinders slipped around screws in

order to fix the clearance between two superimposed elements.
All the latter components and additional ones are illustrated in Figure 4
representing an exploded view of the OUFTI–1 CubeSat.
7
Thesis outline
This thesis takes place in the [VIB] workpackage of the OUFTI–1 project
during the academic year 2010–2011. Its purpose is the quasi-static and
dynamic studies of OUFTI–1 in order to fulfil the mechanical requirements
provided by European Space Agency (ESA).
To this end, this work is divided in several parts.
First, the problem is introduced by a summary of the loads encountered

during the launch phase. The launch environment is also discussed in the
special case of the VEGA maiden flight.
Then, the building of a finite element model of the satellite is presented,

based on previous work. The aim of the work is the modelling of the electronic
cards aboard OUFTI–1.
After that, based on the finite element model, numerical testing will be
performed to verify if the structure is capable of withstand the launch loads.
Finally, testing procedures are exposed for further vibration tests.
8
VEGA maiden flight
1
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.1 VEGA Launch Vehicle . . . . . . . . . . . . . . . 10
1.2 Launch environment and ESA requirements . . 11
1.2.1 Lower limit in frequency requirement . . . . . . . . 13
1.2.2 Quasi-static loads requirements . . . . . . . . . . . 14
1.2.3 Sine vibration requirements . . . . . . . . . . . . . 15
1.2.4 Random vibration requirements . . . . . . . . . . . 19
1.2.5 Shock load requirements . . . . . . . . . . . . . . . 21
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 22
9
Introduction
In this chapter, the VEGA launcher is firstly presented.
Then, the dynamic loads that a payload encounters during its launch are
discussed. The several kinds of loads are studied, first in a general case, and
then in the particular case of the VEGA maiden flight.
Once the loading cases presented, they will be quantified by examination

of the mechanical requirements, prescribed by ESA, that the satellite must
withstand.
1.1 VEGA Launch Vehicle
VEGA, which is translated by “European Advanced

Generation Carrier”, is the new small launcher of
the ESA.
The development of VEGA began in 1998 in

close collaboration with the Italian Space Agency
(ASI) to provide ESA the ability to launch smaller
payloads (300 to 2000 kg) than conventional satel-
lites. Doing so, and taking advantage of the pre-
vious Ariane technologies, VEGA can offer lower
launch costs to small satellites providers.
Unlike Soyuz or Ariane launchers which use ex-

ternal boosters, VEGA is a single-body launcher.
It consists of three solid propellant stages with the
following characteristics in terms of thrust, Specific
Impulse (Isp) and burn time:
• 1st stage: P80FW Solid Rocket Motor:
– Thrust: 2261 kN (see level)

– Isp: 280 s
– Burn time: 106,8 s
• 2nd stage: ZEFIRO 23 Solid Rocket Motor:
10
– Thrust: 1196 kN (see level)
– Isp: 289 s
• 3rd stage: ZEFIRO 9 Solid Rocket Motor:
– Thrust: 225 kN (vacuum)
– Isp: 295 s
The last stage of VEGA, the Attitude and Vernier Upper Module (AVUM),
uses a liquid propellant engine :
• AVUM upper stage: RD-869 Liquid Motor:

– Thrust: 2,45 kN (vacuum)
– Isp: 315,5 s
– Burn time: 667 s (up to five controlled burns)
Figures 1.1(a) and 1.1(b) present the altitude and speed profiles of a
typical VEGA launch with the different flight operations. The last AVUM
stage allows the orbit and attitude control, the separation of the satellites
and the operation of desorbitation.
Details concerning VEGA can be found in [6].
1.2 Launch environment and ESA require-

ments
Before being operational in orbit, the spacecraft have to pass through the
launch phase. The LV provides to the satellite the adequate ∆v and orbital
parameters. From the dynamical point of view, the launch phase is certainly
the most critical part of the life of a satellite. During these few minutes, the
satellite will encounter several types of loads with their own frequencies and
acceleration levels.
11
450000
all-AVUM2 ext-AVUM2
400000
350000
300000
Altitude (m)
250000
ext-AVUM
200000
150000 Larg-Z9
coiffe
100000
Larg-Z23
50000
Larg-P80FW
0
0 200 400 600 800 1000 1200 1400 1600 1800
Temps (s)
(a) Altitude profile

9000
ext-AVUM
8000 Larg-Z9 ext-AVUM2
all-AVUM2
7000
6000
Vitesse relative (m/s)
coiffe
5000
Larg-Z23
4000
3000
2000
Larg-P80FW
1000
0
0 200 400 600 800 1000 1200 1400 1600 1800
Temps (s)
(b) Velocity profile
Figure 1.1: VEGA launch profiles.
12
During the launch phase, the spacecraft will encounter harsh quasi-static
and dynamic loads. These loads can be divided into
• quasi-static accelerations
• sinusoidal vibrations
• random vibrations
• shocks
All these loads have their own amplitude levels and frequency range.
Figure 1.2 illustrates these various loading cases.
Figure 1.2: Encountered types of loads [2].
A more detailed description of the launch behaviour can be found in

[1, 2, 7, 8] and is summarised hereafter.
Besides the description of the various loading cases, mechanical require-

ments to fulfil to be accepted aboard VEGA are quantified. These require-
ments are prescribed by ESA in an Interface Control Document (ICD) [9].
1.2.1 Lower limit in frequency requirement
The first dynamic requirement concerns the fundamental frequency of the

satellite, which has to be higher than a lower boundary value to avoid any
13
dynamic coupling between the launcher and the satellite. This limit corre-
sponds to the low frequency modes of the launcher.
The requirement stipulates that the CubeSat shall not have a fundamental
frequency below 120 Hz in hard-mounted configuration, that is to say inserted
into the P-POD.
1.2.2 Quasi-static loads requirements
Because of the thrust and roll motion of the launcher, the satellite will be
subjected to quasi-static or low frequency accelerations: the Quasi Static
Loads (QSL). The quasi-static assumption is made due to the small values
of excitation frequencies compared to the fundamental frequencies of the
LV and the satellite. The maximum amplitude of these loads is generally
encountered at the end of the first stage burn because of the weaker mass
of the launcher (the burned fuel mass being no longer present) for the same
amount of thrust.
The accelerations concerning the QSL requirement are referred in the

VEGA reference frame and are to be considered acting on the centre of
gravity of the CubeSat. The VEGA coordinate system is illustrated in Figure
1.3. In this coordinate system, the X axis correspond to the longitudinal axis
of the launcher. The lateral Y and Z axes are aligned in a such way that the
plane of separation of the shroud coincides with the plane XOZ.
d
It is important to note that the lateral loads can act simultaneously with
the longitudinal one. Concerning the latter, one will take the following sign
convention:
• a positive sign corresponds to a solicitation in tension in the axes of

VEGA
• a negative sign corresponds to a solicitation in compression in the axes

of Vega
Table 1.1 summarises all the quasi-static loads occurring during each
phase. From this table, we deduce the worst cases for each situation. Those
are:
14
X, Roll
Figure 1.3: VEGA coordinate system
• Maximum compression: 6.7 g
• Maximum in tension: 3.0 g
• Maximum in lateral direction: 1.2 g
Since lateral loads can act together with the longitudinal one, the space-
craft has to withstand any combination of these loading cases. Therefore, an
acceleration envelope is defined, based on the maximum values along each
axis taking the sign into account. Figure 1.4 illustrates this acceleration
envelope.
1.2.3 Sine vibration requirements
In addition to QSL, low frequency (f < 100 Hz) sine vibrations will ap-
pear. These one are mainly caused by the POGO effect. The POGO effect
is a coupling between the hydraulic system of the engines and the struc-
ture. Structural vibrations affect the supply of propellants what disrupts the
thrust. These disturbances in thrust will increase the structural vibrations
15
Event Longitudinal QSL [g] Lateral QSL [g]
min: −4.5
Lift-Off Phase ±1.2
max: +3.0
min: -3.0
P80 max. dyn. pressure ±1.2
max: -2.0
min: -5.0
P80 max. dyn. Acceleration ±0.7
max: -4.0
min: -5.0
Z23 ignition ±0.7
max: -3.0
min: -5.0
Z23 flight and Z9 ignition ±0.2
max: -3.0
min: -6.7
Z9 max. acceleration ±0.2
max: -6.3
min: -1.0
AVUM flight ±0.7
max: +0.5
Table 1.1: Quasi-static loads during the launch.
what will increases the disturbance of the fuel supply and so on. Fortunately,
this effect is well known and anti-POGO systems exist to reduce these vibra-
tions.
The sine vibration requirement imposes acceleration levels spectra that

CubSats must withstand in the P-POD. The following levels are obtained
by defining an envelope of the low frequency vibrations observed during the
LARES qualification tests.
This time, instead of defining the loading cases in the VEGA coordinate
system, they are directly given in the CubeSat one. The CubeSat coordinate
system is illustrated in Figure 1.5 in accordance with the CDS requirement
concerning the coordinate system [10]. It is centred at the geometrical centre
of the cube. The access panel and the base plate are located on the −X and
−Z faces, respectively. Then, the Y axis is defined to obtain a dextrorsum
coordinate system.
The prescribed acceleration levels are of two kinds: the qualification and
acceptance levels, the former being the most severe. These levels relate di-
rectly to the qualification and acceptance stage tests. Their respective defi-
nition can be found in [11] and are reported hereafter:
16
Quasi−Static Loads (QSL)
4
Longitudinal loads [g]

2
−2
−4
−6
−8
−1.5 −1 −0.5 0 0.5 1 1.5
Lateral loads [g]
Figure 1.4: Acceleration envelope.
Figure 1.5: CubeSat coordinate system [3].
Qualification stage: the objective is to demonstrate that the design fulfils

the applicable requirements including margins.
Acceptance stage: the objective is to demonstrate that the product is free

of workmanship defects, is in accordance with the qualified design, and
is ready for its intended use
If a protoflight approach1 is chosen, which is the case for OUFTI–1, the

qualification levels have to be considered for physical testing. In numerical
testing these levels will be applied too.
1
In a protoflight approach, a protoflight model is designed, assembled, tested and then
launched without passing by the step of an engineering model.
17
X-axis loading
The X-axis sine vibration levels for both qualification and acceptance levels
are listed and represented in Table 1.2 and Figure 1.6. The solicitation
frequency content covers the range from 10 to 200 Hz.
In addition to acceleration levels, the frequency sweep rate is also spec-

ified. In the case of the sine acceleration loading, it fixed at two octaves
per minute, which means that the frequency quadruples each minute until
reaching 200 Hz.
X direction
Freq. [Hz] Qualification level [g] Acceptance level [g]
10 1 0,8
25 1 0,8
30 22 17,6
40 22 17,6
45 16 12,8
55 16 12,8
60 6 4,8
85 6 4,8
95 28 22,4
110 28 22,4
120 8,5 6,8
160 8,5 6,8
200 4 3,2
Sweep rate : 2 oct / min
Table 1.2: Qualification and acceptance X sine acceleration levels.
Y and Z-axes loading
In the same manner than for the X-axis loading, the qualification and accep-
tance sine acceleration levels are presented in Table 1.3 and Figure 1.7 for the
Y direction, and in Table 1.4 and Figure 1.8 for the Z direction, respectively.
18
Sine vibration level (X direction)
2
10
Qualification sine
Acceptance sine
Frequency [Hz]
1
10
0
10
−1
10
0 1 2 3
10 10 10 10
Acceleration level [g]
Figure 1.6: Qualification and acceptance X sine acceleration levels.
Y direction
10 1 0,8
25 1 0,8
30 16 12,8
55 16 12,8
60 6 4,8
85 6 4,8
95 25 20
110 25 20
120 8,5 6,8
160 8,5 6,8
200 2 1,6
Table 1.3: Qualification and acceptance sine Y acceleration levels.
1.2.4 Random vibration requirements
Low frequency transient vibrations are also present during the launch phase.
These vibrations are encountered because of wind effects (gust) in the begin-
ning of the launch phase as well at the engines extinction (chugging).
The noisy environment is also very harsh. The acoustic noise (f < 10
kHz), mainly due to the engines but also to aerodynamic flows, can reach
150−180 dB. At the lift-off, the ground reflection increases its intensity. This
reverberation can be limited by injecting liquid water in the exhaust gases
of the engines. By its vaporisation, the water can absorb a fraction of the
19
Sine vibration level (Y direction)
2
10
Qualification sine
Acceptance sine
Frequency [Hz]
1
10
0
10
−1
10
0 1 2 3
10 10 10 10
Figure 1.7: Qualification and acceptance Y sine acceleration levels.
Z direction
10 1,5 1,2
25 1,5 1,2
30 10 8
55 10 8
60 3,5 2,8
85 3,5 2,8
95 17 13,6
110 17 13,6
120 13 10,4
160 13 10,4
200 4 3,2
Table 1.4: Qualification and acceptance Z sine acceleration levels.
noise energy.
Random vibrations (f < 2000 Hz) appear, mainly due to this noise en-
vironment. One may indeed refer to a “vibro-acoustic environment”. The
random vibrations induced to a spacecraft by the acoustic environment are
particularly sensitive for large flexible pieces like solar panels for example.
The random vibration requirement imposes that the CubeSat comply

with the random acceleration levels representative of the acoustic noise be-
haviour aboard VEGA during the launch. The specification to fulfil are pre-
scribed in terms of Power Spectral Density (PSD). As for the sine vibration
requirements, there exist qualification and acceptance levels. As previously,
20
Sine vibration level (X direction)
2
10
Qualification sine
Acceptance sine
Frequency [Hz]
1
10
0
10
−1
10
0 1 2 3
10 10 10 10
Figure 1.8: Qualification ans acceptance Z sine acceleration levels.
the qualification levels shall be withstood in a protoflight approach.
The PSD, expressed in g2 per hertz versus the frequency range is presented
in Table 1.5 and Figure 1.9, respectively. The Root Mean Square (RMS)
acceleration levels for both qualification and acceptance stages are equal to
the square root of the area under their respective curve.
3 - axes PSD
Freq. [Hz]
Qualification levels [g2 / Hz] Acceptance levels [g2 / Hz]
20 0,002 0,0008
70 0,035 0,014
900 0,035 0,014
2000 0,003 0,0012
Duration 2,5 min. per axis 2 min. per axis
Table 1.5: Random vibration levels.
1.2.5 Shock load requirements
The last loading case met is shocks (high frequency). Shocks are due to
stages and shroud separations that involve pyrotechnic devices like explosive
bolts. Shocks also appear at the ignition of the different stages.
The CubeSats shall withstand the shock loads simulated by the qualifi-
cation Shock Response Spectrum (SRS) represented in Table 1.6 and Figure
1.10. A SRS can be seen as the response of a series of one-degree-of-freedom
systems covering all the frequency range of the spectrum.
21
Random vibrations PSD
−1
10
−2
10
PSD [g2 / Hz]
−3
10
Qualification level (gRMS = 6.49)

Acceptance level (gRMS = 4.1)
−4
10
1 2 3 4
10 10 10 10
Frequency [Hz]
Figure 1.9: Random vibration levels.
Lower and upper tolerances are also provided, corresponding to −6 and

+6 dB. The definition of the decibel involves power ratios (proportional to
g2), subtracting or adding six dB corresponds approximately to halve or
double the values of the SRS.
Freq. [Hz] SRS input [g] Tolerance +6 dB Tolerance −6 dB
100 20 39.91 10.02
3225 5000 9976.31 2505.94
10000 5000 9976.31 2505.94
Table 1.6: Shock level for qualification.
Conclusion
All the loading cases occurring during the launch phase were examined and
quantified throughout this chapter. It was highlighted that the launch envi-
ronment is very harsh. Therefore, some numerical and experimental studies
are required to ensure that the CubeSat will be able to withstand the loads
encountered in these few minutes.
Next chapters present such studies. Firstly, in Chapter 2, the finite ele-
ment modelling of the satellite is presented. Then, based on the numerical
model, some simulations of these load cases are performed and discussed
in Chapter 3. Finally, a testing procedure is drawn up to test the actual
structure under sine and random vibrations.
22
Qualification SRS
4
10
3
10
SRS [g]
2
10
SRS input
SRS input upper tolerance (+6 dB)
SRS input lower tolerance (−6 dB)
1
10
2 3 4
10 10 10
Frequency [Hz]
Figure 1.10: Shock level for qualification.
23
Finite element modelling and updating
2
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.1 Previous finite element models . . . . . . . . . . 26
2.1.1 The CubeSat Kit structure . . . . . . . . . . . . . 27
2.1.2 The battery support . . . . . . . . . . . . . . . . . 28
2.2 Finite element modelling of electronic cards . . 29
2.2.1 Choice of the element type in the modelling . . . . 30
2.2.2 Consideration of electronic components . . . . . . 32
2.3 Experimental modal analysis and update process 34
2.3.1 First step: tests . . . . . . . . . . . . . . . . . . . . 34
2.3.2 Second step: identification phase . . . . . . . . . . 40
2.4 Modelling of the OBC2 card . . . . . . . . . . . . 41
2.4.1 Modelling of the PCB and PC104 . . . . . . . . . 41
2.4.2 Experimental modal analysis of the unpopulated
OBC2 . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.4.3 Update of the unpopulated OBC2 model . . . . . 47
2.4.4 Characterization of the PC104 connection . . . . . 50
2.4.5 Consideration of the electronic components . . . . 52
2.5 Modelling of the EPS card . . . . . . . . . . . . . 56
2.5.1 Modelling of the PCB . . . . . . . . . . . . . . . . 56
24
2.6 Modelling of the xEPS card . . . . . . . . . . . . 62
2.6.1 Modelling of the PCB . . . . . . . . . . . . . . . . 62
2.7 Modelling of the FM430 . . . . . . . . . . . . . . 69
2.8 Global finite element model of the satellite . . . 74
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 78
25
Introduction
This chapter is dedicated to the building of a finite element model of OUFTI–

1. First, thanks to Gauthier PIERLOT and Nicolas FRANÇOIS who have
already worked on it, the model does not start from zero. Their models are
described in the beginning of the chapter.
Then, more accurate models of electronic cards are created according to

simplification techniques that can be found in literature. Because of the
simplifications made in the modelling process, a validation step is required.
This is why experimental modal analyses are performed. The set-up for this
purpose is described and results are compared with the numerical ones.
Once the discrepancies between the model and the experimental results
are identified, an update process is performed to move the model as closer
as possible to the actual structure.
Finally, with all the electronic cards models, a global model of the satellite
is created.
2.1 Previous finite element models
The global finite element model of the satellite is divided in various parts
or sub-models. First, one may consider the CubeSat Kit (CSK) structure as
the main part of the model. Indeed, based on this structure, the final model
is built up by adding each element one by one.
Inside the structure, each electronic card of the satellite, as well as the
support of the batteries, will take place one above the other. Outside the
structure, solar panels will come in addition to the antennas support on each
of the six faces of the cube.
Some developments, [3, 4], have already contributed to build a global

model of OUFTI–1. Since the design of the antenna support has changed,
and taking into account that the Communication (COM) card does not exist
yet in the PC104 format and that the some changes on the Electrical Power
Supply (EPS) card have to be done, the model will not be frozen at the end
of this year.
26
Following sections present what already exist in terms of FE models.
2.1.1 The CubeSat Kit structure
In 2008–2009, Gauthier PIERLOT in its master’s thesis [4] began the mod-
elling of the satellite. He naturally started with the aluminium CSK struc-
ture. Due to the thin thickness of the walls (less than 2 mm) compared to
the two other dimensions (≈ 10 cm) the use of shell elements was well suited.
The structure is composed of three parts: the chassis, the base plate, and
the end plate. These three parts are held together by screws in the actual
structure which are modelled by rigid links. Figure 2.1 illustrate the FE
model of the external structure.
Figure 2.1: Model of the external structure by Gauthier Pierlot.
To validate its modelling, an EMA was performed. Two methods were

applied on the recorded data’s: Least Squares Frequency Domain (LSFD)
and Ibrahim Time Domain (ITD) which both gave similar results. Finally a
correlation was carried out to compare the FE model with the actual modal
parameters of the structure.
To summarise the results, the natural frequencies and the correlation

between numerical and experimental modes are tabulated and represented
in Table 2.1 and Figure 2.2. For the mode shapes correlation, the Modal
Assurance Criterion (MAC) is used.
Finally, G. PIERLOT has also modelled the satellite with simple plates
instead of the electronic cards to perform a first quasi-static loading case
but the configuration at this time was different, the design concerning the
battery has changed. The P-POD configuration was not the same too.
27
Modes FEA [Hz] EMA [Hz] (IDT) Relative error [%]
1 539 529.3 1.83
2 602 583.6 3.15
3 611 593.4 2.97
4 669 603.1 10.93
5 727 695.7 4.50
Table 2.1: Numerical and experimental external structure frequencies.
Figure 2.2: Correlation bewteen FEA and EMA mode shapes of the external
structure [4].
2.1.2 The battery support
In the initial design the two batteries aboard OUFTI were located on either
side of a dedicated Printed Circuit Board (PCB) screwed under the xEPS
card. Two problems occurred with this plan. The first one concerned the
available clearance between the batteries and the two cards around. The
smaller one had a value of 0.77 mm, which is insufficient to prevent collision
during the dynamic excitation of the CubeSat. The second trouble was due
to the behaviour of the batteries under vacuum which tend to bulge. These
two problems led to a new design of the batteries configuration. In the
current design, they are encapsulated into an aluminium box which restricts
the tendency to bulge and, by placing the batteries side by side, the clearance
between the support and the adjacent cards became acceptable.
Since the design has changed, this new component had to be modelled.
This was done last year by Nicolas FRANÇOIS in its master’s thesis [3]. The
model of the support he build is illustrated in Figure 2.3.
28
Figure 2.3: FE model of the batery support.
2.2 Finite element modelling of electronic cards
Populated electronic cards are very complex structures. They are made up of
a PCB on which all the electronic components take place. The PCB is a mul-
tilayer laminated substrate, generally a reinforced polymer, with connecting
holes and etched conductive tracks to connect the components together. The
Figure 2.4(a) illustrates the complexity of such supports. The complexity of
the cards also dramatically increases if the electronic components are taken
into account as illustrated in Figure 2.4(b).
(a) PCB of the EPS card (b) Populated EPS card
Figure 2.4: Example of an unpopulated and a populated electronic card.
In a finite element modelling approach, it is clear that all these details

will not be modelled, some simplifications are made.
29
2.2.1 Choice of the element type in the modelling
First, because of their thinness, shell elements are used to build the model
of the PCBs, instead of volume elements. The advantage of this choice is to
greatly reduce the number of Degree of Freedom (DoF) of the problem for an
assumed equal accuracy. Shell elements are well suited for the modelling of
structures whose one geometrical dimension is small compared to the other
two. Since the geometrical dimensions of the electronic cards aboard OUFTI–
1 are approximately 100 × 100 × 1.6 mm (the exact dimensions can be found
in Figure 2.5), the shell element assumption to describe the behaviour of the
PCB appears naturally.
Figure 2.5: Geometrical dimensions of the electronic cards (in mm).
To validate the use of shell elements, a simple modal analysis is performed.

Based on the geometry of the PCBs, two models are created, the first one
using shell elements and the second using volume elements. The mesh is
created by sharing the nodes along each edge and separating them by a
distance of 5 mm. The two meshed models are illustrated in Figures 2.6(a)
and 2.6(b).
For the test purpose, the following material properties representative of

a classical Glass Fibre Reinforced Plastic (GFRP) are applied on the two
PCBs:
• Young’s modulus: 20 GPa
• Poisson’s ratio: 0.3
30
(a) Shell elements mesh (b) Volume elements mesh
Figure 2.6: Shell and volume models for comparison.
• Density: 2200 kg/m3
Modal analyses with these two model performed, we may compare their
natural frequencies and mode shapes. To do so, the metric of comparison for
each case are the frequency deviation, measured by the relative error between
the sets of frequencies, and the MAC between each couple of mode shapes.
Graphically, the Figure 2.7 illustrates the good matching between the two
models.
Frequency deflection
2000
Calculated frequencies
1800
1600
Volume elements
1400
1200
1000
800
600
400
200
0 500 1000 1500 2000
Shell elements
(a) Frequency deflection (b) Mode shapes correlation
Figure 2.7: Good matching between shell and volume elements meshed mod-
els.
In Figure 2.7(a) frequencies of each model reported on the two axes lie
on a line for which the slope is close to the unit. This implies a very good
correlation in frequencies. Quantitatively, the frequencies and the relative
31
error between the two models are reported in Table 2.2, and we may observe
that the maximum relative error is less than 2.5%.
The correlation between the mode shapes of the two models is illustrated
in Figure 2.7(b) by the MAC matrix. Here again, the good correlation be-
tween the two model can be highlighted. Diagonal terms very close to the
unity besides values close to zero for the off-diagonal terms are representative
of a very good correlation between the two models.
Mode Shells [Hz] Volumes [Hz] Relative error [%]

1 369.84 376.07 1.69
2 531.16 541.52 1.95
3 676.57 683.69 1.05
4 923.20 937.90 1.59
5 994.15 1011.18 1.71
6 1584.68 1614.92 1.91
7 1744.11 1776.86 1.88
8 1781.04 1824.91 2.46
9 1825.17 1869.23 2.41
Table 2.2: Frequencies calculated with the two models.
Given these results, the assumption of shell elements to model the PCBs
can be adopted in the following.
2.2.2 Consideration of electronic components
The PCBs modelled, it remains to consider the electronic components pop-

ulating each card. Last year, Nicolas FRANÇOIS [3] gathered some refer-
ences about the modelling of electronic cards ([5, 12, 13, 14, 15, 16]) and
summarised them. Even if presented in [3], the methods to model electronic
cards are still shortly described here. Additional references were also con-
sulted [17, 7, 18].
An electronic component, when added on a PCB, locally increases both

the mass and the stiffness of the structure. As for the modelling of the PCBs,
electronic components will not be modelled in great detail and some degrees
of simplification can be applied.
32
The effect of electronic components on a PCB can be taken into account
according several methods based on averaging or “smearing” of material prop-
erties. These methods are described hereafter [5, 14]:
Simple method: All the components are neglected. The idea is that ne-
glecting stiffness decreases the natural frequencies of a structure and
conversely, ignoring mass increases them. The method is then based
on a balance between these two characteristics.
Global mass smearing: The mass of all the electronic components is smeared
over the whole surface of the PCB. No stiffness increase is done and
the actual mass of the card is conserved.
Global mass and stiffness smearing: The mass and the stiffness are smeared
on the PCB. To do so, the stiffness has to be calculated by testing or
detailed numerical studies of the component on the PCB.
Local smearing: Same as before, but instead of smearing the material
properties on the whole PCB, they are smeared over local areas at
the right position of each component.
These previous smearing methods are illustrated in Figure 2.8
Figure 2.8: Smearing methods [5].
The components themselves are sorted in three categories:
Light components: This category gathers very small and discrete compo-
nents as resistors or transistors with a negligible effect on the card
33
behaviour. They can generally be neglected or taken into account in a
global mass smearing method.
Surface mounted components (SMT): Flat components that increase

density and stiffness. They can be modelled by local or global smearing
methods.
Heavy components: Larger components with a great influence on the be-

haviour of the electronic card. They are modelled by discrete points
mass or by more detailed models attached to the PCB.
Note that this classification is subjective. A component could be classified as

heavy in a small electronic card like these aboard OUFTI–1, but as a light
component if mounted in a personal computer mother board for example.
2.3 Experimental modal analysis and update

process
To validate the finite element models of each electronic card, an EMA is

performed. An EMA is composed of two steps. The first step consists in
taking a series of measurements in several points on the structure. The
second step consists in an identification phase based on the previously mea-
sured responses of the structure. Many identification algorithms exist in both
time and frequency domain depending on whether they work with tempo-
ral recorded data’s (displacements, velocities or accelerations) or frequency
data’s (Frequency Response Functions (FRFs)).
In the following, different configurations are discussed and the acquisition

system is presented.
2.3.1 First step: tests
Boundary conditions
To perform the measurements, the structure has to be set up. Last year, a
test on the OBC2 card was performed in free-free conditions. To simulate
34
the free-free conditions, the card was tied with wires at its four corners. This
set-up is illustrated in Figure 2.9.
Figure 2.9: Last year’s experimental set-up [3].
Accelerations were recorded through the use of a light accelerometer and

the excitation made by an adequate impact hammer. Results so obtained
were not very satisfying. Figure 2.10 shows the obtained correlation between
the identification based on this set-up and a free-free numerical model of the
studied card. We observe that only one mode is catch with a good matching
in frequency and mode shape.
Figure 2.10: Correlation with the last year’s experimental set-up [3].
The reasons of this bad correlation could come from too restrictive bound-
ary conditions and/or from a too much important perturbation due to the
presence of the accelerometer.
This year, a new free-free test was performed on the totally unpopulated
PCB of the EPS card. The applied boundary conditions were simplified to
35
just tie two of the four corners of the card. The card was suspended with
approximately 40 cm-long nylon wire with a diameter close to 1.5 mm. This
set-up is illustrated in Figure 2.11(a).
Given the obtained results, a new free-free set-up was carried out. This
time, in order to further minimize the influence of the boundary conditions,
sewing threads were used, always tying the same two corners (Figure 2.11(b)).
Once again, results were not satisfying and the free-free conditions for ex-
perimental analysis were abandoned. It is supposed that the cause of the
perturbations of measurements does not come directly from the accelerome-
ter, which can be considered with a 0.3 g point mass in the model, but from
the electric wires connecting it to acquisition system.
(a) Nylon wires (b) Sewing threads
Figure 2.11: New free-free set-ups.
Therefore, a new set-up was though. Free-free conditions were replaced by

fixed conditions at the four corners of the card. To do so, in the aluminium
interface plate to be mounted on a shake table, four threaded hole were
manufactured. The cards to test are now fixed above the plate through four
screws and maintained elevated with the use of aluminium spacers. In this
way, the test set-up get closer to the actual fixations that the electronic cards
will encountered once integrated into the satellite. Further, this configuration
permits to place a second card underneath the card to test in order to consider
the additional link through the PC104 connector. Figures 2.12 and 2.13 show
the configurations with one and two cards, respectively.
36
Figure 2.12: Set-up with one card.
Figure 2.13: Set-up with two cards.
37
Acquisition
Once the set-up built, the acquisition can be done. To this end, the needed
material is:
• the set-up itself with the card to be tested,
• an impact hammer to perform the excitation,
• a Laser Doppler Velocimeter (LDV) to record the response in terms of

velocities and some pieces of reflecting paper to ensure a good reception
of the Doppler shifted signal,
• the acquisition system (hardware and software).
The acquisitions parameters, in addition to the sensibilities of the hammer

and the LDV, have to be provided to the acquisition software. They are
chosen as :
• frequency range: [0 − 2000] Hz,
• sample time: 0.8 second,
• sample rate: 5120 Hz (corresponding to 2.56 × 2000 Hz to avoid alias-

ing),
• Force/Exponential windowing,
• average on three consecutive measurements for each point to minimize

errors due to the imprecision of, the orientation and/or the position
of the impact. Averaging also limits the effect of random noise on the
data’s.
In order to further compare experimental results with numerical ones, the

response of the structure has to be recorded in several points. First, for a
comparison in natural frequencies, just a few points are required. Theoreti-
cally, just one measurement may be sufficient if the measurement point does
not lie on a nodal point. Concerning the mode shapes comparison, more
measurement points are required in order to catch the exact deformation
38
91 92 95 96 97
93 94
89 90 98
80 81 82 83 84 85 86 87 88
79
69 70 71 72 73 74 75 76 77 78
Excitation point
59 60 61 62 63 64 65 66 67 68
50 51 52 53 54 55 56 57 58
41 42 43 44 45 46 47 48 49
31 32 33 34 35 36 37 38 39 40
21 22 23 24 25 26 27 28 29 30
11 12 13 14 15 16 17 18 19 20
1 2 3 4 5 6 7 8 9 10
Figure 2.14: Measurment points map.
of the structure. Not enough points may lead to spatial aliasing just as a
temporal signal may be not correctly recorded if the sample rate is too low.
To perform the acquisition, the electronic cards are spatially discretized

by ninety-eight points lying on an approximately 10 × 10 cm grid. These
points are referred to a local coordinates system on the card, centred at the
right-angle corner opposed to the PC104 connector. The x and y axes are
perpendicular and parallel to the principal direction of the PC104 component,
respectively. These mapping nodes are numbered in ascending order from left
to right and down to up. This measurement map is illustrated in Figure 2.14.
To measure properly the response on these points, they are reported on

an electronic card drawing printed on a 1 : 1 scale. To ensure the correct
positioning of the map on the card to test, the outline of the PC104 is cut
off. In that way, the map may be inserted above the card with the PC104
connector passing through it. The Figure 2.15 illustrates this map placed on
a card to point correctly the laser beam.
39
Figure 2.15: Measurement map used to point the laser beam.
2.3.2 Second step: identification phase
In the following, two methods are applied to process measurements: the

Stochastic Subspace Identification (SSI) method [19, 20] in time domain and
the Complex Mode Identification Function (CMIF) [21] in the frequency
domain.
The SSI method is chosen in the set of time domain methods because it
can be used on free responses of the structure and it does not require the
measure at the excitation point to calculate the mode shapes. This particu-
lar point on the diagonal of the FRF matrix is required in the Least Squares
Complex Exponential (LSCE) / LSFD method for instance. The determi-
nation of modal parameters with this method is based on a stabilisation
diagram. On such a diagram, the poles representing both the natural fre-
quencies and damping ratios for several model orders are plotted. In the set
of calculated poles, some are physical and stabilise themselves along their cor-
responding natural frequency. However, by increasing the order of the model,
the number of poles rises and the non-physical poles have to be eliminated.
To do so, some criteria such the stabilisation in frequency and damping ratio
are used.
40
The CMIF is one of the most used indicator function. It is based on
a singular value decomposition of a FRF matrix. The obtained result is a
single curve (if data’s are measured from only one excitation point) in which
natural frequencies are located by a series of peaks. Damping ratios cannot
be calculated but approximates of the mode shapes can.
2.4 Modelling of the OBC2 card
Figure 2.16: Unpopulated and populated OBC2 card
2.4.1 Modelling of the PCB and PC104
As for each electronic cards aboard OUFTI–1, the modelling of the OBC2
card begins by the modelling of its PCB. First, and for the need of experi-
mental tests, it was allowed to cut the connection pins of one of the available
unpopulated OBC2 cards. By several weighing, it was possible to determine
the weight of the different components of the card. Three weighing with the
element present in Figure 2.17 were performed with:
• the unpopulated OBC2 card with its PC104 connector
– Weight: 48.16 g
• the unpopulated OBC2 card with its PC104 connector for which the
connection pins are cut off
– Weight: 45.02 g
• the components forming the PC104 connector
41
– Weight: 15.83 g
Figure 2.17: Determination of masses by several weighing.
So, with these weights, it is possible to deduce the following masses by a

set of subtraction operations:
• the mass of the cut off pins is 3.14 g
• the mass of the PCB is 32.33 g
• the mass of the black plastic part of the PC104 connector is 12.69 g
By modelling the PCB by a face in Samcef Field, it is possible to have

access to its area which is
SP CB = 8.397 10−3 m2 (2.1)
Knowing the weight, the surface, and the thickness of the PCB it is
possible to calculate its density. But, with a thickness of 1.6 mm, the value
obtained exceeded the common value expected for a GFRP material which
should be close to 2200 kg/m3. After verification, it was observed that the
thickness of the OBC2 card was closer of 1.7 mm than 1.6 (by averaging of
several measurements in different points of the PCB). It should be noted that
it is not an exceptional manufacturing error on this particular PCB, the PCB
of the other OBC2 cards have the same thickness close to 1.7 mm. Similar
verifications for the PCB of the other cards led to values close to 1.6 mm as
expected.
42
We can now deduce the equivalent density of the PCB of the OBC2 card.
mP CBOBC2
ρP CBOBC2 =
SP CB × 1.6 10−3
32.33 10−3
=
1.4275 10−5
= 2264.8 kg/m3 (2.2)
Now that the density of the PCB is calculated, the same operation is
performed to calculate the equivalent density of the PC104 connector. In
the simplified finite element model, the 104 holes are not taken into account
and the connector is simply modelled by a rectangular volume with equivalent
material properties.
The product of its dimensions gives the volume of the PC104, that is
VP C104 = 66.55 10−3 × 10.16 10−3 × 11.05 10−3

= 7.47 10−6 m3 (2.3)
Dividing its mass by its equivalent volume, we get the equivalent density
of the PC104 connector :
12.69 10−3
ρpc104 =
7.47 10−6
= 1698.7 kg/m3 (2.4)
With these quantities already determined, the model of the PCB and
the PC104 connector can be built. For this purpose, we need a first guess
for the missing material properties which cannot be determined by testing.
The value of the Young’s modulus of the PCB is firstly estimated to 20 GPa
according to the CES software which provides values for E ∈ [15 − 28] GPa
for GFRP materials. Always according to CES, the Poisson’s ratio ν, is fixed
to 0.315.
The Young’s modulus for the PC104 connector is taken at 846 MPa as
first guess as chosen by Nicolas FRANÇOIS.
This model 1 is illustrated in Figure 2.18 . As boundary conditions, a

node is placed at the centre of each corner’s holes and are linked to them
1
This model will be used for the determination of the equivalent properties for each
electronic cards.
43
Figure 2.18: Finite element model of the PCB and the PC104 connector.
Mode FEA 1 2 3 4 5 6 7
Eigen fraquencies [Hz] 456.66 771.59 828.03 1111.60 1493.71 1612.07 1783.35
Table 2.3: First set of eigen frequencies.
by Mean2 connections. Then this central node is clamp to model the hard
mounted configuration. The natural frequencies obtained are listed in Table
2.3.
2.4.2 Experimental modal analysis of the unpopulated

OBC2
To validate the model, an experimental modal analysis is carried out on the

PCB of the unpopulated OBC2. The identification process is performed with
SSI and CMIF methods on the 98 impulse responses and FRFs, respectively.
To represent the results, the CMIF curve is used as indication in the

stabilisation diagram. To plot the latter, the SSI algorithm is launched with
a series of increasing system orders. To evaluate if a pole is physical or not, its
frequency and damping ratio are compared with an averaging of the nearest
pole for all the smaller orders. An exponential moving average involving
a shading factor is implemented to modify the weight to impose to poles of
decreasing orders. This moving average is calculated as follow and illustrated
2
[A definition of a mean assembly is that it determines the mean rotation and dis-
placement of the nodes concerned. The mean displacement and rotation of a set of master
nodes is given by a slave node]. [22]
44
in Figure 2.19.
pi−1 + (1 − α)pi−2 + (1 − α)2 pi−3 + · · · (1 − α)n pi−n−1

pi = (2.5)
1 + (1 − α) + (1 − α)2 + · · · + (1 − α)n
Pn j
j=0 (1 − α) pi−j−1
= Pn j
(2.6)
j=0 (1 − α)
A null shading factor α represents the simple arithmetic average. By increas-

ing α, greater importance is given to most recent poles.
Shading factor influence
0.35
α = 0.00
0.3 α = 0.05
0.25 α = 0.10
α = 0.15
α = 0.20
Weight
0.2
α = 0.25
0.15
α = 0.30
0.1
0.05
0
1 2 3 4 5 6 7 8 9 10
Previous data’s
Figure 2.19: Shading factor for the exponential moving average.
Figure 2.20 shows the stabilisation diagram obtained with the SSI algo-
rithm up to the 22th order. With this order, the principal peaks are caught
by the algorithm. The obtained frequencies and damping ratios are listed in
Table 2.4.
Mode Frequency [Hz] Damping ration [%]

1 547.50 0.53
2 906.30 0.49
3 1037.90 0.50
4 1583.64 0.52
5 1934.20 0.62
Table 2.4: Natural frequencies and damping ratios of the unpopulated OBC2.
On this diagram, we may see that some peaks seem to be doubled. This
may be caused by very close modes not caught in by the FE model because
of the applied simplifications, but also by errors due to noise during the test.
These “double peaks” make the algorithm also double its poles. In this case
45
Stabilisation diagram
9
22 10
8
10
20
7
10
18 6
10
Ordre oq
CMIF (f)
5
16 10
4
10
14
3
10
Not Stabilized
12 Stabilized in frequency (1 %) 2
Stabilized in damping (5 %) 10
Stabilized
1
10 10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz]
Figure 2.20: Stabilisation diagram of the unpopulated OBC2.
for example, if the algorithm is launched for higher orders, the first pole of
the first couple close to 900 Hz converge to a non real mode as it can be seen
in Figure 2.21 representing its Argand diagram. The same phenomenon does
not occur with the first pole of the second couple. For higher orders, it will
converge towards a similar mode than its neighbour.
Order 28: f = 881.41 Hz, ζ = 0.93
90 1
120 60
0.8
0.6
150 30
0.4
0.2
180 0
210 330
240 300
270
Figure 2.21: Argand diagram of the firs mode of the first “double peak”.
Finally, a small peak appears close to 1400 Hz, but when converged, it
46
seems to be a real mode. However the numerical model will not catch it even
after the update process. This mode is illustrated in Figure 2.22.
Mode 9 : f = 1410.8331
0
100
−1
0 80
60
50 40
20
100 0
Figure 2.22: Mode at ≈ 1400 Hz at the 30th order.
Figure 2.23 shows the correlation between the numerical and experimental
modes. We may observe that the mode shapes are well correlated, but the
frequency discrepancies are important, which will lead to an update process.
2.4.3 Update of the unpopulated OBC2 model
Because of the difference between numerical and experimental results, an up-

date process is needed. In this case, only the frequencies need to be adjusted.
In order to bring the numerical results closer to the experimental ones, the
BOSS-Quattro software was used. In the latter, design variables are pre-
scribed to be updated, and target values are given for a set of results. Based
on the design variables, the software perform a sensitivity analysis by finite
differences. Then a series of iterations on the design variable(s) is performed
to minimize the deviation between the calculated target variables and their
target values.
In a first step, only the Young’s modulus of the PCB was taken as design
variable. The goal of the update process was to minimize the frequency
deviation between the 1, 2, 3 and 6th numerical frequencies and the 1, 2 ,3
and 5th experimental ones by looking at the MAC matrix between the two
sets of modes.
47
Figure 2.23: Correlation between numerical and experimental modes of the
OBC2’s PCB.
This method led to poor results. A value of 29.71 GPa was found for the
Young’s modulus of the PCB but the frequency deviation, even if reduced,
was still too large (up to 22% if we consider the correlation between the
two fourth modes). Furthermore, the 906 Hz experimental mode still cor-
relates with two numerical modes as illustrated in Figure 2.24. By plotting
the experimental mode shapes, it appears that the numerical model of the
PC104 was too flexible and displayed a too large bending compared to the
experimental mode shapes of higher frequencies.
Therefore, a second update process was carried out. This time, the
Young’s modulus of the PC104 connector was added as design variable too
and just the four first frequencies are taken as target because of the poor
representation of higher modes. Not taking the higher modes into the up-
date process proved to be a good choice. The fact is that trying to move
the 6th numerical closer to the 5th experimental frequency based on their
good correlation in the MAC matrix was an error and blindly trust in the
update algorithm may be dangerous. In reality, during the update process,
mode shapes are modified because of the modification of the design variables.
Modes that appear to be paired with the initial set of parameters may no
longer be with updated ones. So, taking such modes in the update process
will not conduct to the optimal solution because these targets values are not
the good ones. This fact is illustrated by the second update process applied
on the PCB of the OBC2 card in which it is the two 5th are correlated instead
of the 5th with the 6th. The final values for the design variables are
48
OBC2’s PCB (first update).
Mode fF EA [Hz] fEM A [Hz] Relative error [%]

1 557.91 547.50 1.90
2 961.67 906.30 6.11
3 992.47 1037.90 4.38
4 1529.70 1583.64 3.41
5 1923.80 1934.20 0.54
Table 2.5: OBC2’s PCB frequencies comparison
• RP CB = 28.5 GPa
• EP C104 = 3.05 GPa
and the correlation between simulation with the optimal parameters end the
experiment is illustrated in Figure 2.25. The two sets of frequencies and the
relative errors between them are listed in Table 2.5.
Remark: The optimal values found in the update process were 28.51 and
3.036 GPa for EP CB and EP C104 , respectively. But, the similar update
procedure for the xEPS card led to a value of 3.12 GPa for EP C104 .
So the intermediate value of 3.05 GPa was chosen and fixed for all the
connectors. There is now only one design variable.
49
OBC2’s PCB (second update).
2.4.4 Characterization of the PC104 connection
Before considering the electronic components of the OBC2 card, we may take
a look at the connection between each pair of cards by their PC104 connector.
Having a model of the unpopulated OBC2 card, and by superimposing two
identical models, it is possible to approximate the equivalent properties of
the connection through the PC104 connector.
Several ways to model this connection were explored. The first one was
to delete the inferior connector and to replace it by a new one with the
same base but its height was increased to go up to the superior card. The
density was modified the get the actual weight of an entire PC104 connector
(the plastic part plus the connection pins) into this new volume. The main
default of this modelling was that it implied to delete the part corresponding
to the PC104 connector in each electronic card model.
To avoid this change in models, a new model was thought. In this one, the
model of the PC104 plastic part on the card is not modified and an additional
rectangular piece of material is added to fill up the clearance between the
top of the inferior connector and the PCB of the superior card.
The third modelling is based on the same principle but instead of creating
a rectangular volume with the same base than the upper face of the connector,
50
the rectangular volume modelling the connection pins is designed with a
reduced base corresponding to the outline of the pin grid. The final model
is so a rectangular volume of dimensions 7.62 × 63.5 × 5.55 mm. To preserve
the actual weight of the hole connector, the mass of the cut off pins (3.14
grams) is concentrated in this volume. The density can be calculated and is
fixed at 1169.3 kg/m3. It remains to define the Young’s modulus to complete
the model, which is done by an updating process.
Figures 2.26(a) and 2.26(b) illustrate the connection between two cards
and the its model, respectively.
(a) Two connected unpopuleted OBC2 (b) Modelling of the connection

cards
Figure 2.26: Modelling of the PC104 connection pins.
As previously, an EMA is performed and a subsequent update process led

to a value for the Young’s modulus equal to
EConnectionP ins = 777.47 MPa (2.7)
with an assumed Poisson’s ratio equal to 0.3.
Figure 2.27 shows the obtained correlation between the numerical and
experimental results.
In Figure 2.27, we observe that experimental modes catch every times two
numerical modes. This fact will often appear in the following results. Due
to their high degree of similitude, the two cards have similar modes in close
frequency ranges. In most cases, because the measurement is performed on
the upper card, only one mode is viewed by the identification. These closed
modes are illustrated in Figure 2.28 for the four firsts numerical modes. Mode
shapes are similar for the two cards in a couple of modes, the difference lies on
the amplitude of the deflections. During the MAC correlation for instance,
the mode shapes are normalized and the two deflections of the upper card
51
Figure 2.27: Cerrelation for a couple of unpopulated OBC2 cards.
will be shown as the same mode shape when correlated with only DoF on
the upper card.
2.4.5 Consideration of the electronic components
Once the model of the PCB validated by measurements and the connection
between the cards being modelled, we may now model the components to
get the final model of the OBC2 card. The electronic components of the
OBC2 card may be the simplest to consider because of their small number
and sizes. Indeed, in the mass budget in [3], the card was weighed and its
mass is equal to 50.47 grams. By weighing an unpopulated OBC2, a mass of
48.16 grams was obtained. So, we can deduce that the hole set of electronic
components expected the PC104 connector have a mass of 2.31 grams which
represent less than 5 percents of the total mass of the card. Since there is
no heavy component, the global mass smearing method is chosen to consider
their effect.
The only parameter to be modified is the density of the PCB. The latter
is updated by considering a global repartition of the weight of the electronic
components over the whole surface. The new density is so calculated by
52
(a) Mode 1 : 624.11 Hz (b) Mode 2 : 703 Hz
(c) Mode 3 : 980.12 Hz (d) Mode 4 : 999.99 Hz
Figure 2.28: Similar numerical mode shapes examples.
Mcard − MPC104
ρupdt
P CB =
Volumecard
(50.57 − 15.83)10−3
=
1.43 10−5
= 2426.6 kg/m3 (2.8)
Once the final model of the OBC2 card built, an EMA on the populated
OBC2 card was performed. The measured frequencies and damping ratios
are listed in Table 2.6. The stabilisation diagram and the MAC matrix for
the correlation are illustrated in Figures 2.29 and 2.30, respectively.
In the MAC matrix in Figure 2.30, we observe that two modes are not
correlated with any numerical modes. The third experimental mode appears
on a peak of the CMIF curve, but this peak is close to a second one which is
well correlated. The fifth mode is located close to the real mode at 1092 Hz.
The 3rd and the 5th modes are represented in Figures 2.31. The 5th is clearly
non physical, but the 3rd mode could be a real deformation of the PCB even
if the model does no permit to catch it.
53
Mode Frequency [Hz] Damping ration [%]
1 601.68 1.21
2 695.10 0.46
3 876.27 0.81
4 928.11 0.45
5 1082.43 1.43
6 1092.26 0.49
7 1648.49 0.59
8 1736.87 0.48
Table 2.6: Measured frequencies and damping ratios.

10
22 10
20 8
10
18
6
10
Ordre oq
CMIF (f)
16
4
10
14
Not Stabilized 2
Stabilized in frequency (1 %) 10
12
Stabilized in damping (5 %)
Stabilized
0
10 10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz]
Figure 2.29: Stabilisation diagram of the SSI identification on the OBC2

card measurements.
Finally, the Table 2.7, whose the layout is inspired from [18], summa-
rizes the results obtained for the corresponding modes between the FE and
experimental models3 .
3
The mode shapes are represented in amplitude. The (dark) blue color corresponds
to a null amplitude of deformation and the (dark) red one to a unitary amplitude. The
deflection map is a top view of the card and its orientations is the same that was drawn
in Figure 2.14
54
Figure 2.30: Correlation between numerical and experimental results.
Mode 3 : f = 876.2652 Mode 5 : f = 1082.4317
1 1
0.5
0.5
0
0
−0.5
−0.5 −1
0 100 0 100
20 80 20 80
40 60 40 60
60 40 60 40
80 20 80 20
100 0 100 0
(a) 3th experimental mode (b) 5th experimental mode
Figure 2.31: 3th and 5th experimental mode shapes.
55
2.5 Modelling of the EPS card
Figure 2.32: Unpopulated and populated xEPS card
2.5.1 Modelling of the PCB
The PCB of the EPS card is the simplest one to model. Indeed, it is the only
one which is totally unpopulated. Without the PC104 connector, it looks like
a simple plate. A similar model to what was done for the acceptance of the
shell elements assumption was made with a refine mesh. The experimental
modal analysis was also simpler. A twenty-point measurement led to good
results.
The modelling step is the same as what is done for the OBC2 card, that
is to say, modelling, comparison with measurements and update process.
Details of these operations are described in the case of the OBC2 card, only
the results are discussed here.
The available PCB was first weighed and a mass of 29.61 grams was
found. Since the thickness of the PCB is no more 1.7 mm but 1.6 mm, a new
volume was calculated and led to the following equivalent density:
29.6050 10−3
ρP CBEP S = = 2203.5 kg/m3 (2.9)
1.3435 10−5
which is close to the expected value of 2200 kg/m3 found for GFRP.
The identified frequencies and the damping ratios are listed in Table 2.8.
An update process was performed with the first four frequencies as target
values, because of their good correlation in mode shapes with the numerical
model. A first guess of 20 GPa as Young’s modulus was considered.
56
Modes Frequencies Mode shapes
MAC
FEA EMA FEA EMA Rel. error [%] FEA EMA
1 1 652.12 601.68 8.38 0.98
2 2 707.45 695.10 1.78 0.99
4 4 1008.33 928.11 8.64 0.98
5 6 1025.07 1092.26 6.15 0.95
7 7 1683.17 1648.49 2.10 0.98
8 8 1785.59 1736.87 2.80 0.98
Table 2.7: General comparison of the OBC2 results.
57
Frequency [Hz] Damping ratios [%]
468.39 0.43
765.07 0.38
890.30 0.32
971.81 0.28
1550.42 0.40
1564.16 0.65
1904.17 0.30
Table 2.8: Measured frequencies and damping ratios of the EPS’s PCB
The optimal value for EP CB found in the frequency discrepancy minimi-

sation is 24.93 GPa. The Figure 2.33 represents the correlation between the
updated model and the identified modes. We may note that, expected for the
5th frequency for which the relative error is 6.69 %, the natural frequencies
are very well caught.
Figure 2.33: Correlation between numerical and experimental results for the
unpopulated EPS card
In this case, the electronic components may not be represented by a simple

or global mass method, because of their size and mass. To include their local
effects, some of them are modelled as rigid blocks with a given mass. The
choice of the modelled components is firstly based on their sizes. Then, their
58
masses are gathered either by looking in their data sheet or by weighing them
if an unmounted specimen is available. In these two cases, the task is difficult.
Indeed, we do not dispose of many specimens for larger components. Beside,
electronic components manufacturers do not often give the mass in the data
sheets. If none of these ways provides the mass information, it may still be
possible to find the weight on on-line providers as Farnell for instance.
The retained components with their dimensions and mass are listed in
Table 2.9 and the EPS model is shown in Figure 2.34. In the model, the
added components are rigidly fixed to the footprint they form on the PCB.
Component Dimensions [mm] Weight [g]

Inductor 10.4 × 10.4 × 4.8 2
Inductor 7.3 × 7.3 × 4.8 1.5
Self 6.86 × 6.47 × 3 1
Tantalium capacitor 7.3 × 4.4 × 4.2 1
Transistor φ = 8.51 ; h = 4.95 1
Table 2.9: List of considered components in the EPS model
Figure 2.34: FE model of the EPS card
An EMA is performed on the populated card in order to correlate the

model with the measurements. The obtained frequencies and the correspond-
ing damping ratios are listed in Table 2.10.
The stabilisation diagram issued from the identification process is illus-

trated in Figure 2.35. At the 24th order, we find that the main frequencies are
caught. The Figure 2.36 shows the MAC matrix obtained between numerical
and experimental modes.
In the MAC matrix, we may note that some modes are not well corre-
lated. Firstly, concerning the 5th and 6th numerical modes, it appears that
59
Frenquency [Hz] Damping ratios [%]
465.51 0.80
631.05 0.67
716.82 0.54
888.40 0.61
1193.56 0.48
1196.44 0.42
1662.00 0.98
1771.45 0.59
1803.64 0.94
Table 2.10: Measured frequencies and damping ratios for the EPS card
10
24 10
22
8
10
20
6
10
18
Ordre oq
CMIF (f)
16 4
10
14
Not Stabilized
2
12 Stabilized in damping (5 %)
Stabilized
0
10 10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz]
Figure 2.35: Stabilisation diagram from the EPS identification
Figure 2.36: Correlation bewteen numerical and experimantal modes of the

EPS card
60
they are mainly deflections involving the inferior card, the EPS card is still
undeformed. Since the measurements are only made on the superior card,
these modes cannot be caught. Two experimental modes, the 5th and 6th
ones, are also not well represented by the numerical model even with an
higher order in the SSI method. In this case it is certainly due to too many
approximations in the numerical model.
As for the OBC2 card, we may draw up a summary table of the obtained
results for the correlated modes. This is done in Table 2.11.
61
MAC
1 1 474.81 465.51 2.00 0.97
2 2 711.11 631.05 12.69 0.67
3 3 798.50 716.82 11.40 0.95
4 4 886.41 888.40 0.22 0.97
7 5 1328.69 1193.56 11.32 0.91
10 8 1832.90 1771.45 3.47 0.83
Table 2.11: Summary tabular of the EPS results
2.6 Modelling of the xEPS card
2.6.1 Modelling of the PCB
The methodology to model the PCB of the xEPS card is the same as what
was done for the OBC2 card, that is to say a model of the PCB in shell
elements and a model of the plastic part of the PC104 connector in volumes
elements. The thickness of the PCB is 1.6 mm and a weighing of the card
gave a value of 30.59 grams for the PCB after the subtraction of the mass
of the plastic part of the PC104 connector. The equivalent density was so
62
Figure 2.37: Unpopulated and populated xEPS card
calculated by:
30.59 10−3
ρP CB = (2.10)
1.3435 10−5
= 2276.9 kg/m3 (2.11)
The PCB of the xEPS is the most complicated one. Indeed, besides the
four holes that were already present in the PCB of the EPS card, three
additional ones are present to allow the fixation of the planar transformer
that pass through the thickness of the PCB. In the model, these holes are
not taken into account for simplification reasons.
As previously, an EMA is performed on the unpopulated xEPS for com-

parison purpose. The obtained stabilisation diagram shows the poles retained
by the SSI algorithm. Figure 2.38 represents this diagram.
On this diagram, we observe the main peaks are caught at the 24th system
order. The MAC matrix between numerical and experimental modes can
be calculated and is represented in Figure 2.39. We may observe that two
experimental modes correlate themselves with a single numerical mode (the
third one at 785.14 Hz).
Figure 2.40 represents the third numerical mode at 785.14 Hz, condensed
into the measurement points, surrounded by the two experimental modes at
877.36 and 1003.65 Hz. We observe that their mode shapes look quite similar.
By calculating the MAC value considering the two experimental modes, we
find a value of 0.7. The fact that they both correlate themselves with only
the third numerical mode with a MAC value of 0.89 indicates a limitation
of the model: this one is not complex enough to distinguish them. We
can see that the third experimental mode shape has a well deformed bulge
63
10
24 10
22
8
10
20
6
10
Ordre oq
18
CMIF (f)
16 4
10
14
Not Stabilized
2
12 Stabilized in damping (5 %)
Stabilized
0
10 10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz]
Figure 2.38: Stabilisation diagram for the PCB of the xEPS card
of the −x edge, but does not represent the deformation of the +y edge.
Conversely, the fourth experimental mode represents well the deformation
of the +y edge, but the bulge is quite less precise. Finally, if we look at
the third numerical mode shape, we may observe that it possess both these
characteristics and so represents an intermediate deformation between the
two consecutive experimental modes. The fact that the MAC value is equal
for each experimental mode, compared to the numerical one, indicates that
this deformation is equally different from each experimental mode shape.
Because of this inaccuracy in the numerical model, the third and fourth
experimental frequencies will not be considered in the update process. The
latter finally led to a value of 26.33 GPa for the Young’s modulus of the PCB
of the xEPS card.
The MAC matrix between the updated model and the experimental modes
is illustrated in Figure 2.41.
As previously made for the EPS model, some electronic components are mod-
elled by rigid blocks to take their local effects into account. The considered
components are listed in Table 2.12.
64
Figure 2.39: Correlation between the initial of the PCB of the xEPS com-
pared to the experimental modes
Mode 3 : f = 877.36 Hz Samcef Mode 3 : f = 785.14 Hz Mode 4 : f = 1003.65 Hz
1 1 1
0.5 0.5
0.5
0 0
0
−0.5 −0.5
−0.5 −1 −1
0 100 0 100 0 100
20 80 20 80 20 80
40 60 40 60 40 60
60 40 60 40 60 40
80 20 80 20 80 20
100 0 100 0 100 0
Figure 2.40: Comparison between the 3th and 4th experimental mode shapes
with the 3th numerical one
Figure 2.41: Correlation between the updated model of the PCB of the xEPS
compared to the experimental modes
65
Planar transformer 15 × 5 × 5.5 1.4
RJ11 connector 15.1 × 13.2 × 12.7 2
Two 1 × 14 connectors 2.54 × 2 ∗ (35.6) × 2.54 0.78
Pic 18F2331 18 × 7 × 2.5 1.33
Table 2.12: Electronic components considered in the xEPS model
The total mass of the considered components is 5.513 grams. Subtracting

this mass and the mass of the PC104 connector from the mass of the xEPS
card which is 42.87 grams lead to the calculus of the equivalent density of
the PCB:
30.357 10−3
ρP CB =
1.3435 10−5
= 2795.7 kg/m3 (2.12)
From the four considered components, the planar transformer and the
RJ11 connector are modelled as rigid blocks with the appropriate mass. This
is not the case for the double 1 × 14 connector and the PIC 18F2331. Indeed,
even if its mass is less than one gram, the geometry of the double 1 × 14
connector will have a sensitive effect on the behaviour of the card. It acts
like a stiffener in the y direction. The PIC 18F2331 is considered because its
mass is more important than those of the other surface mounted components.
To take its effect into account, the footprint it forms on the PCB is drawn
and an increment in density is brought to the PCB. This new additional
density is calculated by the mass of the component over the volume of the
PCB that support it:
1.333 10−3
ρadd =
(18 × 7 × 1.6) 10−9
= 6612.1 kg/m3 (2.13)
No increase in stiffness is done on the area delimited by the component, only

the local mass smearing is applied for the PIC 18F2331.
Figure 2.42 illustrates the finite element model of the xEPS card.
To validate the model, an EMA is performed ont the xEPS electronic

card. The resulting stabilisation diagram is shown in Figure 2.43. Note
66
Figure 2.42: Finite element model of the xEPS card
10
24 10
22
8
10
20
6
10
Ordre oq
18
CMIF (f)
16 4
10
14
Not Stabilized 2
10
Stabilized in frequency (1 %)
12
Stabilized 0
10 10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz]
Figure 2.43: Stabilisation diagram of the xEPS card
that this time, the four measurement points numbered 56, 74, 84 and 85
are not accessible because of the presence of electronic components. The
measurement is so performed on 94 points.
The order 24 catches the main peaks present on the CMIF curve. The
obtained results in terms of frequencies and damping ratios are listed in Table
2.13.
Figure 2.44 shows the correlation between numerical and experimental

modes. As previously seen in the case of the EPS, two adjacent numerical
modes (107.37 and 1084.08 Hz) do not correlate themselves with experimental
ones. The explanation is identical as previously, the response of the structure
at these frequencies is mainly a deformation of the unmeasured bottom card.
67
Frequency [Hz] Damping ratio [%]
518.98 0.87
662.47 0.98
752.71 0.45
936.17 0.73
940.38 0.33
1309.99 0.84
1311.68 0.56
1837.51 0.61
1916.02 0.66
Table 2.13: Frequencies and damping ratios from the identification process
of the xEPS card
Figure 2.44: MAC matrix between numerical and experimental modes of the
xEPS card
This is illustrated in Figure 2.45. Therefore, since no measurement on this

card is performed, the identification process does not match these two modes.
The second thing that can be observed is that the identification process
double two experimental modes. The identified poles are very close in fre-
quency : 0.45% for the modes at 936.14 and 940.38 Hz and 0.13% for the
modes 1309.99 and 1311.68 Hz. The crossing MAC between the two modes
of each couple are 0.85 and 0.98, repectively. These values indicate that both
their frequency and mode shape are similar and thus that they represent the
same mode.
Finally, as previously made in the other cases, a summary tabular is filled

with the numerical and experimental results in Table 2.14. Note that the nu-
68
(a) 1007 Hz mode (b) 1084 Hz mode
Figure 2.45: Nmerical modes involving only the bottom card
merical frequencies are always greater than the identified ones. The increase
in stiffness by rigidly fixing the planar transformer and the RJ11 connector
is an extreme case. So, a subsequent update process of the Young’s modulus
of the PCB could be made to compensate this simplification assumption in
order to get an equivalent global model of the card.
2.7 Modelling of the FM430
Figure 2.46: FM430 Flight Module
Now, pass to the last electronic card to model, the FM430 Flight Module
(FM430) bought with the CSK structure. This electronic card is the most
complex of all the electronic cards aboard OUFTI–1. It is the heaviest one
and it has the most components both above and below the PCB.
In this case we do not have at our disposal a specimen of the PCB of

this card. The PCB update step is no longer applicable. Therefore, the
69
MAC
1 1 532.21 518.98 2.55 0.96
2 2 706.17 662.47 6.60 0.73
3 3 787.81 752.71 4.66 0.98
4 4 906.70 936.17 3.15 0.96
4 5 906.70 940.38 3.58 0.92
7 6 1406.04 1309.99 7.33 0.95
7 7 1406.04 1311.68 7.19 0.97
Table 2.14: Summary tabular of the xEPS results
70
methodology to model this card is slightly different and it is made up with
the following step:
• select the components to model,

• weighing the card and the selected components to get a value of the
equivalent density of the PCB,
• include the retained components in the model with a first guess of 20
GPa for the Young’s modulus of the PCB,
• perform an update of the Young’s modulus of the PCB to get closer to
the identified experimental frequencies.
From all the electronic components populating the card, the retained ones
are listed in Table 2.15.
USB B 12 × 16.6 × 11 3.71
Remove Before Flight 20 × 12.5 × 6.7 5.07
Power Jack 9 × 14 × 11 2
1 × 17 connectors 2.54 × 43.69 × 8.51 1.6
1 × 13 connectors 2.54 × 33.53 × 2.54 1.2
Table 2.15: Electonic components considered in the FE model of the FM430

card
The total mass of the card is 66.85 grams. Subtracting the weight of the
considered components and of the plastic part of the PC104 connector, we
can calculate an equivalent density of the PCB in a global smearing approach
to consider all the other components. We have:
(66.85 − 12.69 − 13.58) 10−3

ρP CB =
1.3435 10−5
= 3020.5 kg/m3 (2.14)
The finite element model of the FM430 is build and is illustrated in Figure
2.47.
An EMA is then performed on the card to get its modal parameters. Note
that in this case, the FM430 being the bottom one, no connection pins are
71
Figure 2.47: FE model of the FM430 card
7
28 10
26
6
10
24
5
22 10
Ordre oq
CMIF (f)
20
4
10
18
3
16 10
Not Stabilized
14 Stabilized in frequency (1 %) 2
10
12
Stabilized
1
10 10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz]
Figure 2.48: Stabilisation diagram of the FM430
present under the PCB. Therefore, contrary to the other electronic cards, it
can be tested without a second card below it. The results of the experimental
and identification process are illustrated in the stabilisation diagram shown
in Figure 2.48. As an image of the complexity of the structure, we may note
that the obtaines CMIF curve is the most uneven one.
At the 26th system order, the main peaks are caught by the SSI algorithm.
For this order, the stabilized natural frequencies and damping ratios are listed
in Table 2.16.
Numerical and experimental modal parameters obtained, we can proceed

to the comparison between the two sets of modes. The cross correlation MAC
matrix between numerical and experimental modes is illustrated in Figure
72
Frequency [Hz] Damping ratio [%]
355.00 1.27
527.92 1.04
760.64 1.10
854.52 0.75
988.29 1.34
1218.94 0.88
1583.26 0.70
Table 2.16: Frequencies and damping ratios from the identification process
of the FM430
Figure 2.49: Correlation between the FM430 model and experimental results
2.49. In this figure, we may see the same phenomenon that was previously
observed during the study of the xEPS card, the 3th and 4th experimental
modes are approximated by only one numerical mode. Then if we look at
the frequency deviation between the fourth and third first experimental and
numerical modes, respectively, we get the following relative errors:
• EM A1 vs. F EA1 : 20.87 %,
• EM A2 vs. F EA2 : 29.11 %,
• EM A3 vs. F EA3 : 27.69 %,
• EM A4 vs. F EA3 : 13.66 %.
73
Figure 2.50: Correlation between the updated FM430 model and experimen-
tal results
An update process is therefore required which will results in a softening

of the PCB. As they are represented by only one numerical mode, the 3th
and 4th experimental modes are not taken as target values to not mislead the
optimisation.
After update, a value of the Young’s modulus of the PCB was found
equal to 19.2 GPa. The updated numerical modal parameters are obtained
and compared with the experimental results. This is illustrated in Figure
2.50. In this figure, we may see that it is now the fourth experimental mode
which is closer to the third numerical one. Further, the update procedure
increased the correlation level between the fifth experimental and the fourth
numerical modes to a value of 75%.
Finally, Table 2.17 summarises the obtained results concernig the FM430
modelling and testing.
2.8 Global finite element model of the satel-

lite
Now that the cards are modelled, we are able to construct a global model of
the satellite based on the one of N. FRANÇOIS. Its model was itself based
on the external aluminium structure and the four endless screws modelled by
74
MAC
1 1 338.85 355.00 4.55 0.98
2 2 578.84 527.927 9.64 0.98
3 3 805.61 760.64 5.91 0.79
3 4 805.61 854.52 5.72 0.95
4 5 945.98 988.29 4.28 0.75
Table 2.17: Summary tabular of the FM430 results
75
Figure 2.51: Starting point for the global model
G. PIERLOT. In that model, he added its new battery support, the antenna
support and the solar panels. In addition, he inserted simple models of
electronic cards too, using the global mass smearing to model the electronic
component expect for the PC104 connector. Note that these cards were
just fixed to the endless screws modelled by four beams and that they were
not connected together through their PC104 connector. From this model,
only the external structure, the screws and the battery model were kept, as
illustrated in Figure 2.51.
In this main model, each electronic card is imported as a Samcef Part and
positioned at its right place. Since the COM card does not exist yet in the
PC104 format, it is replaced by the model of the unpopulated OBC2 card.
Then, they are linked to the four beams modelling the endless screws. In
the last year’s model, these links were locking in the three directions. In this
model, only the lateral displacements along the X and Y axes are locked,
the vertical motion along the Z axis is left free expect for the two extreme
cards i.e. the FM430 and the card representing the COM card which are also
locked along the vertical direction.
The connection between each pair of cards is made by importing several

Samcef Parts modelling the PC104 connection pins, as what was illustrated
in Figure 2.26 on page 51. Note that a piece of material is added above the
76
PC104 connector of the EPS card with identical material properties in order
to fill the clearance due to the presence of the battery support. Indeed, in
the actual model, the PC104 of the EPS card must to be higher.
To consider the aluminium spacers, equivalent springs are added between

each “layer”. These spacers have the following properties:
• Height : 15 mm (12 spacers), 19 mm (4 spacers) and 4 mm (4 spacers),
• External diameter : φext = 4.5 mm,
• Internal diameter : φext = 3 mm,
• Young’s modulus : ∈ [68 − 74] GPa, 70 GPa is assumed.
From these characteristics, equivalent stiffness’s are calculated by

EA
keq = (2.15)
L
in which A is the cross section of the spacers and L is their length. After
calculus, we have:
k15 mm = 4.12 104 N/mm (2.16)

k19 mm = 3.26 104 N/mm (2.17)
k4 mm = 1.55 105 N/mm (2.18)
(2.19)
After, the eight feet are also added as rigid bodies and are meshed by
only one node. The equivalent connection between paired cards (equivalent
PC104 pins and equivalent spring for the spacers) is illustrated in Figure
2.52(a). Figure 2.52(b) illustrates an added foot.
Finally, a small clearance exist between the feet and the chassis. However,
applying a very light pressure (Figure 2.53) on the foot is sufficient to have
a contact between these two parts. Since the CubeSat will be compressed in
the P-POD two rigid links are made on each corner between two nodes on
the chassis and two on the Base/End plates.
The global finite element model of OUFTI–1 is illustrated in Figure 2.54.
77
(a) Equivalent PC104 pins (b) Added foot
and spacer
Figure 2.52: Equivalent connection between cards and feet added to the
model
Figure 2.53: Clearance between feet and the chassis
Conclusion
In this chapter, the already existing models were described. Then, to model
the electronic cards, a short review of the literature was summarized and a
test set-up for experimental modal analyses in order to validate the models
was presented.
To model the PCB of the electronic cards, a first simulation validated the
use of shell elements. On the PCBs , the electronic components were applied
using the simplification techniques previously presented. These techniques
have proven their usefulness because of the high complexity of the structure
of electronic cards, but also because of the important lack in mass informa-
tion concerning electronic components. An important simplification of the
connection through the PC104 connector has also been made.
The model being created and updated, the comparison with experimental
measurements provide a good matching between FEAs and EMAs. We may
note that the greatest discrepancy occurs in the EPS card with a relative
78
Figure 2.54: Global model of the satellite
error in frequency of 12.69% and a MAC value of 0.67.
With all the card models, a global finite element model of the satellite was
created. The latter will allow us to perform a series of numerical simulations
to study the effect of the prescribed external loads on the CubeSat. This
leads to the next chapter in which such studies are performed.
79
Numerical testing
3
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1 Modal analysis . . . . . . . . . . . . . . . . . . . . 81
3.1.1 Boundary conditions . . . . . . . . . . . . . . . . . 82
3.2 Quasi static accelerations . . . . . . . . . . . . . . 85
3.3 Sinusoidal vibrations . . . . . . . . . . . . . . . . 89
3.3.1 Boundary conditions and applied loads . . . . . . . 91
3.3.2 X direction loading . . . . . . . . . . . . . . . . . . 92
3.3.3 Y and Z direction loadings . . . . . . . . . . . . . 95
3.4 Random vibrations . . . . . . . . . . . . . . . . . 98
3.4.1 Transformation of the prescribed PSD . . . . . . . 99
3.4.2 Adding SPECTRAL commands . . . . . . . . . . . 100
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 101
80
Introduction
A global finite element model being created and described in the previous
chapter, we can now apply all the loading cases prescribed by ESA to verify
that the structure is strong enough to withstand the launch phase.
In the present case, the use of the FEA has several goals. Firstly, it
allows finding the natural frequencies of the CubeSat. Then, applying loads
representative of the launch environment will provide estimates of static and
dynamic stresses and strains inside the model.
Then, knowing the material properties, margins of safety can be calcu-

lated and provide information on how much greater loads the satellite can
undergo.
Finally, based on the obtained results the current design can be adopted
or modified.
3.1 Modal analysis
The modal analysis has two purposes. The first one is to verify the fulfilment
of the requirement which imposes a lower limit for the first natural frequency
of the structure to avoid dynamic coupling between the satellite and the LV.
This requirement was presented in Section 1.2.1 and imposes a minimal value
of 120 Hz for the first natural frequency in hard mounted conditions.
In addition to the study of the dynamic behaviour, the modal analysis is

used to construct a modal base for subsequent dynamic simulations as sine
and random dynamic simulations.
Based on the global model described in Section 2.8, the modal analysis
is performed. The results of this analysis are a set of eigen frequencies and
mode shapes, solutions of the eigen problem
(K − ω 2 M )q = 0 (3.1)
are obtained by resolving the motion equations
Kq + M q̈ = 0 (3.2)
with the assumption of synchronous motion.
81
3.1.1 Boundary conditions
In order to simulate the fixation of the CubeSat in the P-POD, boundary

conditions have to be applied. In this structure, the lateral faces of the chassis
will be in contact with rails close to the vertical edges of the cube. On the
actual structure, the portion of the lateral faces in contact with these rails
are hard anodized. In the finite element model of OUFTI–1, each lateral face
is divided into the central part of the face and two faces corresponding to
the hard anodized areas (Figure 3.1).
(a) Hard anodized zones on lateral faces (b) Subfaces equivalent to the rail’s areas
Figure 3.1: Rails on the structure.
Then, the nodes of these sub faces are gathered into Mesh Groups to
make their handling easier. For each lateral face, a group is made with the
nodes of its two rails. Altogether, six groups are built and they are referred
similarly to the names of the faces. For instance, the nodes of the rails of the
+X face are gathered in the mesh group “Rails +X”.
Concerning the lateral boundary conditions, only the nodes belonging

to the rails are constrained. For each one, the only fixed component is the
direction perpendicular to its plane. At this stage, the CubeSat is laterally
constrained, the only possible rigid motion is along the Z axis. In the same
manner than for the rails, the eight feet are also constrained. In the P-POD,
the feet of the satellite are in contact with either the bottom or the top of the
82
Mode Frequency [Hz] Description
1 330,82 Global lateral motion along the Y di-
rection plus a bulging of the xEPS card
2 363,32 Lateral translation and rotation along
the Z axis of the EPS card
3 377,88 Bulging of the EPS card
4 421,66 Bulging of the FM430
5 455,56 Bulging of the xEPS
6 526,07 Bulging of the OBC2 card
7 599,28 Second bulging of the FM430
8 644,87 Asymmetric bending of the EPS card
9 656,46 Bulging of the unpopulated OBC2
card replacing the COM card
10 669,75 Bulging of the +X face oof the chassis
Table 3.1: Ten first modes with their frequency and description.
structure, or with the feet of adjacent CubeSats. Therefore, the boundary

condition applied to the feet is simply a locking along the Z direction.
The boundary conditions fixed, the modal analysis can be performed.

The ten first flexible modes are listed in Table 3.1 with their frequencies and
a short description of their mode shapes. The latter are illustrated in Figure
3.2.
After the 9th mode, a series of modes involving deflections of the external
structure follow, before a new series of electronic cards deflections.
Note that the first mode has a frequency at 330.82 Hz, that is, more
than 175% higher than the 120 Hz of the first requirement. The latter is so
fulfilled.
The modal analysis is performed up to the 150th mode to comply with

the sine and random analysis which need a solid modal basis. The sine
acceleration spectra and the random PSD go up to 2000 Hz, so the modal
basis has to include enough modes to sweep this frequency range. The 57th
mode has a frequency of 2013.14 Hz. Fifty-seven modes have to be calculated
at the minimum. However, to solve accurately the dynamic problem by mode
superposition, 90% of the structure mass should be include in the modal
basis. In the three directions, the cumulated effective modal masses with the
150 first modes are :
83
(a) Mode 1: f = 330,82 Hz (b) Mode 2: f = 363,32 Hz (c) Mode 3: f = 377,88 Hz
(d) Mode 4: f = 421,66 Hz (e) Mode 5: f = 455,56 Hz (f) Mode 6: f = 526,07 Hz
(g) Mode 7: f = 599,28 Hz (h) Mode 8: f = 644,87 Hz (i) Mode 9: f = 656,46 Hz
(j) Mode 10: f = 669,75 Hz
Figure 3.2: Graphical representation of the ten first modes.
84
• Cumulated effective modal masses in the X direction : 82.7%
• Cumulated effective modal masses in the Y direction : 79.9%
• Cumulated effective modal masses in the Z direction : 93.7%
The 90% are reached only in the Z direction but not in the X and Y
ones. However, going further gives very small increment of effective masses
by additional modes. Therefore, the modal basis is limited to these 150
modes.
3.2 Quasi static accelerations
Now, we consider the quasi-static loads requirement described in Section

1.2.2. For reminder, the QSL are prescribed in the VEGA coordinate system
and the maximum acceleration levels that the CubeSat will encounter are:
• Maximum compression: 6.7 g
• Maximum in tension: 3.0 g
• Maximum in lateral direction: 1.2 g
Aboard VEGA, OUFTI–1 will be placed in one of the three P-POD tilted
of 10˚to the longitudinal axis of VEGA. The CubeSats are placed in the P-
PODs by inserting them in such a way that the −Z face of the CubeSats
facing the +Z face of the P-POD [3].
To perform the quasi-static acceleration analysis and taking into account

that lateral loads can act simultaneously with longitudinal ones, a series of
critical load cases are studied and compared.
Because of the tilted orientation of the P-POD, the loads defined in the
VEGA frame have to be projected in the OUFTI–1 frame. The worst case
loads to apply in both longitudinal (in compression and tension) and lateral
directions are calculated by adding the projections of the loads in the VEGA
frame to the OUFTI–1 frame. In the OUFTI–1 frame, we have then:
85
(a) Worst case (b) Worst case
in tension in compression
Figure 3.3: Worst cases in longitudinal loading.
• Maximum longitudinal load in compression (along Z) :

acomp = 6.7 cos(10) + 1.2 sin(10) = 6.8066 [g] (3.3)
• Maximum longitudinal load in tension (along Z) :

atens = 3 cos(10) + 1.2 sin(10) = 3.1628 [g] (3.4)
• Maximum lateral load (along X and Y ) :

alat = 1.2 cos(10) + 6.7 sin(10) = 2.3452 [g] (3.5)
The external loads considered, it remains to take the effect of the two
other CubSats in the P-POD. They will apply their weight, under the longi-
tudinal acceleration levels, on OUFTI–1 by the means of their feet. Besides
considering the worst case in loading, we have also to consider the worst
position inside the P-POD. To do so, longitudinal loading in tension and
compression are divided into two sets of loading cases with two different
configurations. Figure 3.3 shows these two extreme positions.
In the tension case, the weight of the two other CubeSats is applied on
the four +Z feet, that is for each one:
2 × 1 × 3.1628 × 9.81
Wtens = = 15.5135 N (3.6)
4
In the compression case, the weight of the two other CubeSats is applied
on the four −Z feet, that is for each one:
2 × 1 × 6.8066 × 9.81
Wcomp = = 33.3864 N (3.7)
4
86
Tension in the VEGA frame
Load case 1 Load case 2 Load case 3
Load case 4
OUFTI-1
Load case 5
frame
Load case 6 Load case 7 Load case 8
Compression in the VEGA frame
Figure 3.4: Load cases for the QSL study.
Now that all the loads are defined, we may form the eight combinations
they can have and then perform a static analysis of each load case. These
eight load cases are illustrated in Figure 3.4 with the values previously cal-
culated for the acceleration levels. Concerning the boundary conditions to
take into account, in each case, the support facing each load is locked. For
instance, if the lateral load in the +X direction is active, the nodes of the
rails on the +X face are locked but not those on the −X face.
During each analysis, the equivalent Von Mises stresses are studied in
each part of the model and their amplitude are compared with the yield
limit of the different materials. A Safety Factor (SF) is also considered to
fill in the modelling assumptions and simplifications. The value of this SF is
prescribed in [9] and is equal to 1.5. The latter reference also gives the way
to calculate the Marging of Safety (MoS). Applied to the yield limit of the
87
Element Material Yield limit [Mpa] SF
Structure Chassis Al - 5052 H32 152,00 1,50

Base plate
End plate
Endless screws Screws Inox 170,00 1,50
PCB FR4 110,00 1,50
FM430
PC104 Nylon 86,40 1,50
Flexible connectors
PCB FR4 110,00 1,50
OBC2
PC104 Nylon 86,40 1,50
PCB FR4 110,00 1,50
EPS
PC104 Nylon 86,40 1,50
Battery support Support Al - 7075 T6 359,00 1,50
PCB FR4 110,00 1,50
xEPS
PC104 Nylon 86,40 1,50
Flexible connectors
Table 3.2: Yield limits and safety factors used in the quasi-static analyses.
material, it is equal to
σy
M oSy = − 1, (3.8)
SFy · σload
where
• σy is the yield strength of the material,
• SFy is the safety factor for yield strength,
• σload is the maximum Von Mises equivalent stress due to external loads.
The yield MoS indicates how many times the external loads can with-
stand by the structure before entering in plasticity. It should be noted that,
because of the use of shell elements, the equivalent Von Mises stresses have
to be studied in both lower and upper skin in addition to the neutral fibre.
Concerning the material properties, a conservative approach is also followed
by always taking the lowest values found, if intervals are available for the
yield strength of materials. Table 3.2 gives the used values for the calculus
of the MoS. The material considered for the PC104 and similar connectors is
Polyamide (Nylon) (Type 46, 15% Glass Fibre, V-0), based on [23] and CES
sofware.
88
Figure 3.5: Stress concentration in the linked elements between the chassis
and the End plate.
After the study of each load case and considering the values in Table 3.2
we can summarise all the results in a single tabular, which is done in Table
3.3. On this table, we find that the minimum margin of safety occurs in
the structure for the sixth load case. However, it should be noted that the
maximum Von Mises stresses appears at each corner of the chassis, exactly
where links were added to take into account the contact between the End
and Base plates with the chassis (Figure 2.53 on page 78). Figure 3.5 shows
this stress concentration.
Note that the maximum displacements occur already in the EPS card
and do not exceed 0.026 mm.
Concerning the displacements, Figure 3.6 illustrates the deflections en-

countered during the sixth load case where the minimum MoS was found.
Finally, because all the MoS are positive, we may conclude that the QSL
requirement is fulfilled.
3.3 Sinusoidal vibrations
The aim of the sinusoidal vibrations analysis is to verify if the structure can
withstand the sinusoidal acceleration levels prescribed in the sine vibrations
requirement. To this end, the REPDYN module of Samcef is used, preceded
by the DYNAM module for the calculus of the modal basis (150 modes
are retained). The sine vibrations analysis is composed of three harmonic
89
sidered.
FM430 OBC2 EPS Battery xEPS Max.
Structure Endless screws Min. M oS
PCB PC104 PCB PC104 PCB PC104 suppoort PCB PC104 displ. [mm]
σmax [Mpa] 8,73 10,90 1,39 0,19 0,28 0,06 1,14 0,13 1,30 0,83 0,23 0,012
LC1 9,40
M oS 10,61 9,40 51,76 302,16 260,90 959,00 63,33 442,08 183,10 87,35 249,43 (EPS)
σmax [Mpa] 9,27 10,91 1,89 0,32 0,71 0,2 1,21 0,07 1,16 0,51 0,1 0,013
LC2 9,39
M oS 9,93 9,39 37,80 179,00 102,29 287,00 59,61 821,86 205,32 142,79 575,00 (EPS)
σmax [Mpa] 8,47 11,05 1,54 0,18 0,3 0,06 1,31 0,14 0,99 0,83 0,24 0,013
LC3 9,26
M oS 10,96 9,26 46,62 319,00 243,44 959,00 54,98 410,43 240,75 87,35 239,00 (EPS)
σmax [Mpa] 8,55 10,87 2,03 0,32 0,72 0,19 1,31 0,07 1,11 0,53 0,1 0,013
LC4 9,43
90
M oS 10,85 9,43 35,12 179,00 100,85 302,16 54,98 821,86 214,62 137,36 575,00 (EPS)
σmax [Mpa] 25,82 21,73 2,5 0,42 1,1 0,3 2,77 0,18 2,08 0,87 0,09 0,026
LC5 2,92
M oS 2,92 4,22 28,33 136,14 65,67 191,00 25,47 319,00 114,06 83,29 639,00 (EPS)
σmax [Mpa] 28,85 29,56 2,27 0,3 0,56 0,3 2,69 0,24 1,93 1,16 0,34 0,025
LC6 2,51
M oS 2,51 2,83 31,31 191,00 129,95 191,00 26,26 239,00 123,01 62,22 168,41 (EPS)
σmax [Mpa] 25,84 20,77 2,35 0,41 1 0,29 2,61 0,18 2,2 0,94 0,1 0,026
LC7 2,92
M oS 2,92 4,46 30,21 139,49 72,33 197,62 27,10 319,00 107,79 77,01 575,00 (EPS)
σmax [Mpa] 27,66 27,9 2,06 0,33 0,54 0,13 2,5 0,24 2,37 1,19 0,33 0,025
LC8 2,66
M oS 2,66 3,06 34,60 173,55 134,80 442,08 28,33 239,00 99,98 60,62 173,55 (EPS)
Table 3.3: Summarise of the margins of safety for the eight load cases con-
Figure 3.6: Maximum displacements in the worst load case.
response analyses, one in each direction, with the qualification levels given
in Section 1.2.3.
3.3.1 Boundary conditions and applied loads
The boundary conditions applied on the structure are the same as for the
modal analysis, that is, locking of each rail in the direction orthogonal to it
and locking of the eight feet in the Z direction.
The applied loads are subsequently the qualification spectra represented

in Figures 1.6, 1.7 and 1.8, on pages 19, 1.7 and 1.8 for the loading along the
X, Y and Z axes, respectively.
Conversely to the quasi-static loading, the accelerations are this time

given in the OUFTI–1 coordinates system.
For all the studies, the assumption of viscous damping is made and a
1% modal damping ratio is applied to all the modes of the modal basis. As
expected for the FM430 all the measured damping ratios of the electronic
cards were under 1%. However these damping ratios were found for a light
impact solicitation. With the increase in acceleration levels injected to the
91
Maximum displacement in the structure (X direction loading)
0.1
0.08
Displacement [mm]
0.06
0.04
0.02
0
0 50 100 150 200
Frequency [Hz]
Figure 3.7: Maximum displacement during the sine vibration analysis in the
Xdirection.
structure, damping ratios will increase too. In order to stay in a conservative

approach the modal damping ratios are not increased above 1%.
3.3.2 X direction loading
Since the problem is linear and the higher frequency of the input spectrum is
far smaller than the fundamental frequency of the structure, the most impor-
tant response will appear at the maximum level with the higher frequency. In
each case, this appears at 110 Hz and in the case of X direction solicitation,
this occur with an acceleration level of 28 g’s.
To be convinced, Figure 3.7 represent the amplitude of the maximum

displacement in the structure versus the frequency. Note that all the other
quantities have the same comportment.
Saving all the results for the whole structure for multiple frequency steps
quickly becomes heavy in memory and causes problems. Therefore concen-
trates ourselves in just the 110 Hz frequency is sufficient.
Displacements
Under sine vibrations in the X direction, because of the additional fixation

caused by the PC104 connectors close to the −Y face, the greater displace-
92
Figure 3.8: Global displacements under sine vibrations along the X direction.
ment occur at the opposite, that is, close to the +Y face as illustrated in
Figure 3.8.
The EPS card undergoes the greatest displacement amplitudes. The main
effect is an in-plane rotation of the card, as shown in Figure 3.9 in which only
the EPS card and the endless screws are represented.
Stresses
The stresses are evaluated at a frequency of 110 Hz, whereas they are at their
maximum value.
Figure 3.10 shows the stress repartition in the external structure. We

may observe that the higher amplitudes are reached in the Base plate where
endless screws are fixed, and in the chassis where these same screws are
fixed by the midplanes standoffs. Concerning the screws, the same stress
concentration can be observed at their fixation points as illustrated in Figure
3.11. It is interesting to note that when G. PIERLOT preformed a sine
test, an unscrewing of midplane standoff occurred. So these zones should be
particularly inspected when performing new tests.
93
Figure 3.9: Displacements of the EPS card under sine vibrations along the
X direction.
Figure 3.10: Stress repartition inside the structure under sine acceleration
along the X direction.
94
Figure 3.11: Stress repartition inside the screws under sine acceleration along
the X direction.
In the PCBs and in the battery support, the stress concentrations are
located close to the fixation points. Figures 3.12 and 3.13 show these con-
centrations in the battery support and in PCBs, respectively.
3.3.3 Y and Z direction loadings
Displacements
The displacements occurring during the Y and Z loading cases are shown in
Figures 3.14 and 3.15.
The Y direction accelerations result in a global motion in the Y direction

of all the components. Conversely to the previous case, the PC104 connectors
follow the displacement with the same amplitude. Because of its central
position, it is the EPS which has the greater displacement amplitude.
As expected, the acceleration along the Z axis involves vertical deforma-

tions of the internal components except the battery support due to its more
important rigidity. Once again, we may remark that the EPS card is the
most deformed under loading.
95
Figure 3.12: Stress concentration in the battery support under sine acceler-
ation along the X direction
Figure 3.13: Stress concentration in the PCBs under sine acceleration along
the X direction
96
Figure 3.14: Internal displacement under sine loading along the Y axis
Figure 3.15: Internal displacement under sine loading along the Z axis
97
Stresses
Concerning the stress concentrations during the loadings, the same conclu-
sions as for the X direction case can be done, that is:
• In the structure: concentration around the fixations of the endless

screws on the base plate and around the midplanes standoffs which
hold the screws to the chassis.
• In the screws: concentrations at each extremity.
• In the battery support and the PCBs: concentrations around the fixa-
tion points on each corner.
3.4 Random vibrations
In order to fulfil the requirement relative to random vibrations, a numerical

random analysis has to be performed. Random vibration analysis are carried
out by the SPECTRAL module of the Samcef solver. Similarly to the calculus
of a harmonic response, the random analysis is based on a mode superposition
method.
The loading in a SPECTRAL analysis is given by an input PSD and

the results are also given in terms of PSD. In this case, the loading is the
qualification level random vibration PSD described in Section 1.2.4. The
PSD has to be applied in the three directions of the CubeSat.
Unfortunately, it is not possible to launch the simulations, apparently due

to memory problems, even by reducing the modal basis to 50 modes. The
followed procedure is still described.
Firstly, as for the harmonic response calculation, a modal analysis is

needed to build a modal basis. Here too, 150 modes are initially chosen
to cover the frequency range of the input PSD but also to have a sufficient
representation in terms of effective modal masses.
Then, a PSD is needed for the problem as loading. The prescribed PSD
is given in terms of g2/Hz versus Hz.
98
3.4.1 Transformation of the prescribed PSD
Before running the simulation, the input PSD has to be transformed. The
frequencies expressed in hertz’s have to be transformed in rad/s. The accel-
eration must also be modified from g’s to mm/s2. Therefore, the input PSD
to provide to the solver is expressed in terms of
2
mm
s2 rad
rad versus
s
s
Then, the PSD, which can be described with few points in log-log scale, has to
be interpolated in several other points for accuracy reasons since SPECTRAL
will interpolate the curve in a linear scale.
Between two points [x1 y1 ] and [x2 y2 ], the interpolation at an intermedi-

ate point x is given by
log(y2 /y1 )
x log(x2 /x1 )
y = y1 (3.9)
x1
Figure 3.16 illustrates the transformation of the given PSD the right form
demanded by the software.
Prescribed random PSD x 10

5 Random PSD for SPECTRAL
−1
10 6
PSD [(mm/s2)2 / (rad/s)
5
PSD [g2 / Hz]
4
−2
10 3
1
−3
10 0
1 2 3 4
10 10 10 10 0 5000 10000 15000
Frequency [Hz] ω [rad / s]
Figure 3.16: Prescribed PSD and PSD required by SPECTRAL
The PSD has to be written in a specific format and it must have the same
name than the problem with the .psd extension.
99
3.4.2 Adding SPECTRAL commands
To perform the random analysis, the DYNAM data code has to be filled
with SPECTRAL commands. This can be done in the Epilogue of the solver
menu. These commands describe the analysis and they provide information
such as:
• The type of load
• Boundary conditions
• Nodes on which the excitation is applied
• Frequency range and number of control frequencies
• Time of simulation
• Damping ratio
• Results to archive
100
Conclusion
In this chapter, the use of the finite element model of OUFTI–1 under several
studies has provided results overcoming the requirements. The first require-
ment concerns the fundamental frequency of the CubeSat. This frequency
shall be greater than 120 Hz in hard-mounted conditions. The modal study
in the DYNAM solver led to a fundamental frequency equal to 330.82 Hz.
This dynamical requirement is therefore fulfilled.
The second requirement concerned quasi-static loads encountered during

the flight because of the thrust and roll motion of the launcher. Eight cases
representing all the possible configurations of loading have been studied. In
each case, the most critical amplitude for each acceleration was considered.
The criteria to deduce if the structure passes or fails the test is the calculus
a margin of safety. The latter involves material properties and safety factors.
Margins were calculated with the yield limit of the different materials. To
stay in a conservative approach, when this property was given in an interval of
values, the lowest one was always considered. Concerning the safety factor, it
is prescribed equal to 1.5 in the ICD. For all the cases, it was in the sixth one
that the lowest margin was found. The latter being positive, the quasi-static
loads requirement is therefore fulfilled.
Then, the following requirement concerns sine acceleration loadings. The

simulation was performed taking into account the qualification levels pre-
scribed in the ICD. Here again, all the margins of safety were found positive,
the requirement is then fulfilled. Note that the greatest stress concentration
was this time found in the endless screws, at the level of their connection to
the chassis via the midplane standoffs.
Unfortunately, concerning the random vibration requirement, it is not

possible to run the models, the problem remains unresolved.
The numerical results have now to be confirmed by testing. With the

little time remaining, tests could not be performed, but testing procedures
are already drawn up in the next chapter.
101
Testing procedure
4
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.1 Needed facilities . . . . . . . . . . . . . . . . . . . 103
4.2 Test set-up . . . . . . . . . . . . . . . . . . . . . . 104
4.2.1 Accelerometers placement . . . . . . . . . . . . . . 106
4.3 Sinusoidal and random acceleration tests . . . . 106
4.3.1 Objectives of these tests . . . . . . . . . . . . . . . 106
4.3.2 Test sequence . . . . . . . . . . . . . . . . . . . . . 107
4.3.3 Pass criteria . . . . . . . . . . . . . . . . . . . . . . 107
4.4 Shock tests . . . . . . . . . . . . . . . . . . . . . . 107
102
Introduction
The aim of this chapter is to draw a test procedure for intermediate testing.
Note that operations that have to be followed for the qualification tests on the
flight model are described in [24]. For the qualification tests, an official Test-
POD must be requested to ESA or directly to Cal Poly. In the following, the
available Test-POD built by Artur Scholz (COMPASS–1) for intermediate
tests is sufficient.
Hereafter, the needed facilities to perform the tests are first listed and
then objectives and step by step procedure to perform sinusoidal and random
vibrations tests are described.
4.1 Needed facilities
To perform the vibration tests, the following facilities are needed:
A shaker: we may have at our disposal the electrodynamic shaker installed

in the vibration laboratory at the University (B52). This shaker is a
Gearing & Watson V2664 with the following characteristics:
• Maximum force: 26 kN (sinus and random)

• Maximum displacement: 50 mm (peak to peak)
• Maximum velocity: 1.52 m/s
Figure 4.1 illustrates this shaker.
The interface aluminium plate: this plate is needed to interface the the
shake table and the Test-POD.
The Test-POD: to simulate the actual behaviour of the CubeSat in the

P-POD.
The test article: the CubeSat in its current configuration.
Accelerometers: to measure the response of the CubeSat and control the

input acceleration. Two accelerometers are used:
103
Figure 4.1: Shaker installed in the laboratory
• The first one is used to control the input acceleration level (con-
trol accelerometer). Several PCB Piezotronics accelerometers are
available and, in particular, an accelerometer Model 352C661 can
be used as control accelerometer:
– Frequency range: 0.5 to 10000 Hz
– Measurment range: ±50 g’s
• The measurement accelerometer is also a PCB Piezotronics one,
but for weight and cluttering reasons, it should be as small as
possible. An accelerometer Model 352A732 will be used:
– Frequency range : 2 to 10000 Hz
– Measurment range : ±1000 g’s
Note that only one specimen is available, so it must be moved
between each direction loading.
The acquisition system: the hardware and software. The LMS software
is used.
4.2 Test set-up
The test set-up is the same for the sine and random accelerations tests for
each direction.
1
http://www.pcb.com/spec_sheet.asp?model=352C66&item_id=11988
2
http://www.pcb.com/spec_sheet.asp?model=352A73&item_id=5963
104
1. Follow the integration procedure [25] to assembly the CubeSat up to the
positioning of the accelerometer, place the accelerometer (see below)
and continue the integration procedure.
Important remark: Some electronic components, precisely the jumpers

are not hardly fixed on their cards (EPS and xEPS). They must
be removed to prevent their ejection and collision with other com-
ponents. Figure 4.2 shows this kind of components.
Figure 4.2: Jumper components to be removed
2. Introduce the CubeSat in the Test-POD. The integration procedure

concerning the integration of a CubeSat in a Cal Poly Test-POD could
be used as a basis to adapt to our Test-POD [26].
3. Fix the interface plate on the shaker table, parallel to the loading direc-
tion for the X and Y axes testing and perpendicular to the loading di-
rection for the Z axis testing. Fix the Test-POD on the interface plate.
Figure 4.3(a) and 4.3(b) illustrates the configuration of the Test-POD
for X and Y axes, and Z axis testing, respectively.
Excitation direction
Excitation direction
(a) Set-up for X and Y axes testing (b) Set-up for Z axis testing
Figure 4.3: Test-POD configuration for X, Y and Z directions testing
105
4.2.1 Accelerometers placement
First, the control accelerometer can be glued on the interface plate in a

direction which is aligned with the loading direction.
Concerning the measurement accelerometer, it must be glue on a facet

perpendicular to the loading direction.
• For the X and Y loading directions, the suggested positioning is illus-

trated in Figure 4.4, where vertical facets of the battery support are
used.
Figure 4.4: Accelerometers placement for X and Y axes loading
• For the Z loading direction, since the greater amplitude is reached on

the EPS card, the point number 41 (Figure 2.14 on page 39) is suggested
as position for the accelerometer.
• Strips of Kapton tapes should be adhered on faces at these positions

before gluing the accelerometer, in order to avoid glue residues after
testing.
4.3 Sinusoidal and random acceleration tests
4.3.1 Objectives of these tests
The goal of the vibration tests is to prove that the structure is able to with-
stand the acceleration levels prescribed in the ICD [9] and presented in Sec-
tions 1.2.3 and 1.2.4.
106
4.3.2 Test sequence
For each test, the following sequence is to be applied:
1. Perform a low level sine sweep test in the range [0.5 2000] Hz with a
sweep rate of 2 octaves per minute and an amplitude included between
0.2 and 0.5 g. Perform this sweep up and down and identify the natural
frequencies ([24]).
2. Perform the test with the qualification level and the sweep rate pre-
scribed in the requirements (Sections 1.2.3 and 1.2.4).
3. Perform a new low sine sweep and check if there is a deviation com-
pared with the first one. A discrepancy may indicate some damage or
modifications in the fixing conditions of the internal elements.
4. Perform a visual inspection. Pay a particular attention to the fasteners

and electric cables connections.
4.3.3 Pass criteria
The test is considered successful if:
• No damage or default is identified by visual inspection.
• Discrepancies between the surrounding low sine sweep are less than 5%
in frequency and 3 dB in amplitude ([24]).
4.4 Shock tests
Shock tests have to be performed along each axis of the CubeSat considering
the “plus” and “minus” directions.
Currently, the test facilities are not yet identified. The V2i’s shakers are
not sufficient to simulate the SRS of the shock load requirement. Pyrotechnic
or free fall shock tests facilities have to be found to performed these tests.
Pass criteria are also visual inspection and functional test.
107
Conclusion
In this master thesis, several aspects of the qualification process of OUFTI–1

were swept through dynamic analyses.
In the first part, the problem to be encountered, i.e. the launch phase
and its various load cases, were exposed and quantitative mechanical re-
quirements imposed by ESA in an in the Interface Control Document were
reported.
Then, based on the previous work of Gauthier PIERLOT (2008–2009)

and Nicolas FRANÇOIS (2009–2010), the building of the global finite ele-
ment model of the CubeSat continued. Starting with the external structure,
the endless screws and the battery support, the model was completed by
importing models of each card.
In Chapter 2, the complete procedure to model electronic cards was pre-

sented and comparison with experimental measurement led to acceptable
correlation between numerical and experimental results. The global model
built up, numerical tests were launched.
In chapter 3, these numerical testing led to a fulfilment of all the require-

ments except the random vibration requirement for which it is unfortunately
not yet possible to run the analysis.
Finally, in Chapter 3, testing procedures were drawn up, but facilities to

perform the shock test still need to be found to fulfil this requirement.
108
The qualification process is obviously not yet finished. First, modelling
both the COM card which is not already built, and the new antenna support
still remain. The latter and the five solar panels will have to be taken into
account in the last model, when the flight model of OUFTI–1 will be as-
sembled and tested. Correlation between measurement and the final model
should be done to validate it.
109
Bibliography
[1] Fortescue Peter, John Stark, and Graham Swinerd, editors. Spacecraft
Structures Engineering. Wiley edition.
[2] Adriano Calvi. Spacecraft structural dynamics & loads. an overview,

November 2009. Lecture notes.
[3] Nicolas FRANÇOIS. Dynamic analysis and launch qualification of oufti-

1 nanosatellite. Master’s thesis, University of Liège, 2010.
[4] Gauthier Pierlot. Oufti-1 : flight system configuration and structural

analysis. Master’s thesis, University of Liège, 2009.
[5] Alastair Amy Robin, Aglietti G.S., and Richardson Guy. Sensitivity
analysis of simplified printed circuit board finite elements models. EL-
SEVIER, 2009.
[6] Arianespace. Vega user’s manual, March 2006.
[7] DREVON Claude and ALDEGUER Jodé. Conception de cartes pour

équipements spacialisables. Techniques de l’ingénieur.
[8] Pierre Rochus. Conception d’expériences spatiales. Lecture notes.
[9] ESA. CubeSat Educational Payload on the Vega Maiden Flight. Interface
Control Document, March 2011. Issue 0 - Revision 7.
[10] Cal Poly. CubeSat Design Specification, Jan 2009.
110
[11] ESA-ESTEC. ECSS-E-ST-10-02C : Space engineering-Verification,
March 2009.
[12] Alastair Amy Robin, Aglietti G.S., and Richardson Guy. Accuracy of
simplified printed circuit board finite element models. ELSEVIER, 2009.
[13] D.WATTIAUX, O.VERLINDEN, and Ch. DE FRUYTIER. Modelling

of the dynamic behaviour of modular equipments designed by thales
alenia space etca.
[14] Alastair Amy Robin, Aglietti G.S., and Richardson Guy. Simplified
modelling of printed circuit boards for spacecraft applications. Acta
Astronautica, 2009.
[15] Alastair Amy Robin, Aglietti G.S., and Richardson Guy. Analysis of
simplified fe models of pcbs exposed to random vibrations.
[16] J. Pitarresi, Phil Geng, Beltman Willem, and Yun Ling. Dynamic mod-
eling and measurement of personal computer motherboards. Electronic
Components and Technology Conference, 2002.
[17] Zhang Bo, Liu Pin-kuan, Ding Han, and Cao Wenwu. Modeling of
broad-level package by finite element analysis and laser interferometer
measurements. ELSEVIER, 2008.
[18] Lee Ying-Chih, Wang Bor-Tsuen, Lai Yi-Shao, Yeh Chang-Lin, and
Chen Rong-Sheng. Finite element model verification for packaged
printed circuit board. ELSEVIER, 2008.
[19] J.-C. Golinval. Vibration testing & modal identification. Lecture notes.
[20] J.-C. Golinval and Gaetan Kerschen. Experimental modal analysis. Lec-
ture notes.
[21] Ewins D. J. MODAL TESTING theory,practice and application. Re-

search Studies Press, john wiley & sons, inc. edition, 2000. second edi-
tion.
[22] Samcef Field html documentation.
[23] Tyco electronics. PC/104 and PC/104-Plus Connectors.
[24] Galeone P. C. Approach for the structural tests for the qualification of
the CubeSats/P-PODs, April 2009.
111
[25] Pierre Martinelli. Integration procedure for the engineering model of
OUFTI–1, 2011.
[26] California Polytechnic State University, San Luis Obispo, CA 93407.

Test Pod User’s Guide, June 2006. Revision 6.
112

OUFTI-1 Nanosatellite

Uploaded by

Copyright:

Available Formats

OUFTI-1 Nanosatellite

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

OUFTI-1 Nanosatellite

Uploaded by

Copyright:

Available Formats

University of Liège

Faculty of Applied Sciences

Thesis submitted in partial fulfilment

Supervisor: Prof. Jean-Claude GOLINVAL

Academic Year 2010–2011

OUFTI–1, standing for Orbital Utility For Telecommunication Innovation,

Therefore, a finite element model of the current design of the satellite

Finally, a testing procedure is drawn up to further correlate with the

1 VEGA maiden flight 9

1.1 VEGA Launch Vehicle . . . . . . . . . . . . . . . . . . . . . . 10

1.2 Launch environment and ESA requirements . . . . . . . . . . 11

1.2.1 Lower limit in frequency requirement . . . . . . . . . . 13

1.2.2 Quasi-static loads requirements . . . . . . . . . . . . . 14

1.2.3 Sine vibration requirements . . . . . . . . . . . . . . . 15

1.2.4 Random vibration requirements . . . . . . . . . . . . . 19

1.2.5 Shock load requirements . . . . . . . . . . . . . . . . . 21

2 Finite element modelling and updating 24

2.1 Previous finite element models . . . . . . . . . . . . . . . . . . 26

2.1.1 The CubeSat Kit structure . . . . . . . . . . . . . . . . 27

2.1.2 The battery support . . . . . . . . . . . . . . . . . . . 28

2.2 Finite element modelling of electronic cards . . . . . . . . . . 29

2.2.1 Choice of the element type in the modelling . . . . . . 30

2.2.2 Consideration of electronic components . . . . . . . . . 32

2.3 Experimental modal analysis and update process . . . . . . . 34

2.3.1 First step: tests . . . . . . . . . . . . . . . . . . . . . . 34

2.3.2 Second step: identification phase . . . . . . . . . . . . 40

2.4 Modelling of the OBC2 card . . . . . . . . . . . . . . . . . . . 41

2.4.1 Modelling of the PCB and PC104 . . . . . . . . . . . . 41

2.4.2 Experimental modal analysis of the unpopulated OBC2 44

2.4.3 Update of the unpopulated OBC2 model . . . . . . . . 47

2.4.4 Characterization of the PC104 connection . . . . . . . 50

2.4.5 Consideration of the electronic components . . . . . . . 52

2.5 Modelling of the EPS card . . . . . . . . . . . . . . . . . . . . 56

2.5.1 Modelling of the PCB . . . . . . . . . . . . . . . . . . 56

2.5.2 Consideration of the electronic components . . . . . . . 58

2.6 Modelling of the xEPS card . . . . . . . . . . . . . . . . . . . 62

2.6.1 Modelling of the PCB . . . . . . . . . . . . . . . . . . 62

2.7 Modelling of the FM430 . . . . . . . . . . . . . . . . . . . . . 69

2.8 Global finite element model of the satellite . . . . . . . . . . . 74

3.1 Modal analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 81

3.1.1 Boundary conditions . . . . . . . . . . . . . . . . . . . 82

3.2 Quasi static accelerations . . . . . . . . . . . . . . . . . . . . . 85

3.3 Sinusoidal vibrations . . . . . . . . . . . . . . . . . . . . . . . 89

3.3.1 Boundary conditions and applied loads . . . . . . . . . 91

3.3.2 X direction loading . . . . . . . . . . . . . . . . . . . . 92

3.3.3 Y and Z direction loadings . . . . . . . . . . . . . . . . 95

3.4 Random vibrations . . . . . . . . . . . . . . . . . . . . . . . . 98

3.4.1 Transformation of the prescribed PSD . . . . . . . . . 99

3.4.2 Adding SPECTRAL commands . . . . . . . . . . . . . 100

4 Testing procedure 102

4.1 Needed facilities . . . . . . . . . . . . . . . . . . . . . . . . . . 103

4.2 Test set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

4.2.1 Accelerometers placement . . . . . . . . . . . . . . . . 106

4.3.1 Objectives of these tests . . . . . . . . . . . . . . . . . 106

4.3.2 Test sequence . . . . . . . . . . . . . . . . . . . . . . . 107

4.3.3 Pass criteria . . . . . . . . . . . . . . . . . . . . . . . . 107

4.4 Shock tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

1 Small satellites [1]. . . . . . . . . . . . . . . . . . . . . . . . . 3