Thesis Ozer Can

POLITECNICO DI MILANO
Polo Regionale di Lecco
Facoltà di Ingegneria Industriale
Master of Science in
Mechanical Engineering
Testing and optimization of the tunable damping system for

vibration control of a space devoted instrument
Relatore: Prof. Bortolino SAGGIN
Co-relatore: Ing. Diego SCACCABAROZZI
Master thesis of:
Can ÖZER
Matr. 722899
Anno Accademico 2009 - 2010

Acknowledgements
I owe my deepest gratitude to Prof Saggin and Diego; this thesis would
not have been possible without them.
Furthermore it is a great pleasure to thank everyone in the Measurements

Lab especially Marco (thanks for all the support!!) and Gerardo.
Special thanks to,
my wonderful, ever caring brother, Ceyhun,
my loving mother and father,
all my friends and family. Together we’ve been “there and back again”
Abstract
In this work the testing and the optimization of the vibration damping
system designed for a space application is performed. The damping
system consists of three versatile and adjustable dampers whose main
objective is to control the vibrations transmitted to the Mars Infrared
MApper (MIMA) through its mounting interface during its launch and
landing on Mars surface. This is essential in order to assure the survival
of the main instrument parts and especially of the brittle optics, which
has to overcome a random loading with an RMS acceleration of 150 m/s2
in a frequency range encompassing the main structure resonances.
The environmental conditions have been simulated through shaker tests

on a single individual damper to characterize its dynamic properties and
finally achieving its optimal tuning. Further shaker tests were
performed on an inertially equivalent mockup of the instrument with
the aim to investigate the performance of the full dampers system and
predict the instrument transmitted vibrations. Alongside the
experimental procedures, a computational model of the test setup is also
established and this model is validated with feedback from the mockup
tests at the expected vibration levels. Finally an optimized damper
configuration is conceived by minimization of the transmissibility in the
frequency regions of “dangerous random” evaluated on the instrument
optics.
Keywords: damping, vibration control, Finite Element methods
1
Sommario
Lo scopo del presente lavoro è la sperimentazione e l'ottimizzazione di
un sistema di smorzamento per impiego spaziale. Il sistema consiste di
tre "damper" con l'obiettivo di controllare i livelli di eccitazione
trasmessi allo spettrometro MIMA, durante le fasi di lancio e di
atterraggio su Marte. Tale sistema deve garantire il superamento degli
elevati livelli di accelerazione attesi per la missione, in particolare, le
ottiche dello strumento devono superare senza danneggiamento, un
eccitazione di tipo "random" che interessa un ampio intervallo di
frequenze e il cui valore RMS è pari a 150 m/s2.
I livelli di eccitazione previsti per la missione sono stati simulati

utilizzando un eccitatore elettrodinamico e i parametri dinamici di ogni
singolo damper sono stati definiti ed ottimizzati. La progettazione di un
modello dinamico equivalente dello strumento ha reso possibile l'analisi
delle prestazioni dei dampers attraverso la valutazione delle vibrazioni
trasmesse allo strumento. Parallelamente all'attività sperimentale, un
modello ad elementi finiti del mockup dello strumento è stato sviluppato
e validato in seguito al confronto con i risultati sperimentali ottenuti. La
configurazione ottimale del sistema di smorzamento è stata quindi
definita, minimizzando la trasmissibilità dell'eccitazione "random" nel
campo di frequenze in cui risulta essere particolarmente gravosa per lo
strumento.
Parole chiave: smorzamento, controllo delle vibrazioni, elementi finiti
2
CHAPTER 1: INTRODUCTION 11
1.1 Mars Infrared Mapper (MIMA) 11
1.2 MIMA environmental conditions 12

1.2.1 Sine Environment 13
1.2.2 Random environment 14
1.2.3 Pyroshocks 14
1.3 The need for a vibration damping system 15
1.4 The dampers 18
1.5 Objectives 19
CHAPTER 2: DESIGN AND MANUFACTURING OF A DUMMY MODEL FOR

TESTING 22
2.1 Necessity of a dummy model 22
2.2 Computer modeling of the prototype 22

2.2.1 Part design 22
2.2.2 Assembly 23
2.3 Modification of the dummy prototype to achieve an optimum model 25

2.3.1 The problem and the requirements 25
2.3.2 The method of optimization 26
2.4 Validity of the dummy model for testing purposes 28
CHAPTER 3: TUNING OF THE DAMPERS WITH ONE DEGREE-OF-FREEDOM

TESTING 30
3.1 The need for damper tuning 30
3.2 Test methodology 30

3.2.1 Experimental equipment 31
3.2.2 Experimental Modal Analysis techniques 36
3.2.3 Test setup 37
3.2.4 Vibration stimulus for the single degree-of-freedom tests 42
3.3 Tuning of the dampers 43

3.3.1 Objectives of the tuning procedure 43
3
3.3.2 Excitation amplitude and damper non-linearity 45
3.3.3 Damper response and tightening preload 47
3.3.4 Radial behavior of dampers 49
3.3.5 Damper Response Synchronization 51
3.3.6 Configurational modifications of damper assembly 54
3.4 Dampers in their final tuning and configuration 58
3.5 Results and discussion 60
CHAPTER 4: DAMPER PROPERTIES ESTIMATION 62
4.1 Excitation amplitude and linear modal parameters 62
4.2 Analytical formulation of the test setup 64
4.3 Estimating linear parameters from test data 66

4.3.1 Modal parameters from resonance 66
4.3.2 Extracting modal parameters by curve fitting 70
4.4 Estimating torsional properties from test data 74
4.5 Estimated properties of the final damper configuration 77

4.5.1 Axial stiffness coefficient, k1 77
4.5.2 Radial stiffness coefficients, k2 & k3 78
4.5.3 Rotational stiffness coefficient, kθ,1 78
4.5.4 Rotational stiffness coefficients kθ,2 & kθ,3 78
4.6 Results & Discussion 79
CHAPTER 5: FINITE ELEMENT ANALYSIS OF THE DAMPER – MOCKUP

SYSTEM 80
5.1 The need for a Finite Elements Analysis 80
5.2 Verification of the structural modes of the mockup using FEM 81

5.2.1 Finite Element Modeling 81
5.2.2 Meshing 83
5.2.3 Results and discussion 84
5.3 Analysis of the Damper-Mockup system using linear spring elements 87

5.3.1 Modeling of the dampers as spring elements 87
5.3.2 Defining Virtual Spring stiffness coefficients 88
4
5.3.3 Resulting behavior of the system 89
5.3.4 Evaluation of the results for the ultimate damper configuration 94
CHAPTER 6: TESTING OF THE DUMMY MODEL 97

6.1 Test methodology 97
6.1.1 Test setup 97
6.1.2 Reference stimulus 101
6.2 Mockup testing and results 105

6.2.1 Z axis testing 105
6.2.2 X axis testing 109
6.2.3 Y axis testing 112
6.3 Evaluation of the damping system characteristics from the test results 115
6.3.1 Performance in Z axis 115
6.3.2 Performance in X axis 116
6.3.3 Performance in Y axis 117
CHAPTER 7: FINITE ELEMENT METHODS FOR THE RESPONSE

IMPROVEMENT OF THE DAMPERS 120
7.1 Introduction 120
7.2 Objectives of response improvement 121
7.3 Response improvement by modifying the axial spring properties of the dampers 122
7.3.1 Initial damper tuning 122
7.3.2 Effects of individual dampers on the overall system response 123
7.3.3 Combination of different axial stiffnesses to achieve an optimum response (Opt 1) 124
7.4 Response optimization of the system with theoretical dampers 126

7.4.1 Effect of the radial stiffness coefficient 126
7.4.2 Effect of axial stiffness coefficient 127
7.4.3 Resulting modes of vibration of the optimal system (Opt 2) 128
7.5 Performance comparison of the damper tunings using FE methods 129

7.5.1 Method of performance assessment 129
7.5.2 Assessment of the test configuration 130
5
7.5.3 Assessment of the optimization performed through changing axial stiffnesses (Opt 1) 135
7.5.4 Assessment of the optimization performed through changing both the axial and the radial
stiffnesses (Opt 2) 139
7.5.5 Performance comparison of the optimal damper tunings 144
CHAPTER 8: CONCLUSION AND FINAL REMARKS 147
REFERENCES 150
APPENDIX A 152
APPENDIX B 159
6
List of Figures
Figure 1-1: The Rover Module equipped with the MIMA 12

Figure 1-2: Positioning of the MIMA on the rover and the defined axis syste 13
Figure 1-3: Weighing factor on the MIMA as a function of excitation frequency 15
2
Figure 1-4: FRF profiles of System 1 and System 2 16
2
Figure 1-5: Weighed FRF profiles of System 1 and System 2 17
Figure 1-6: Outside view of a tightly packed damper 19
Figure 1-7: The diagram representing the objectives of the study 21
Figure 2-1: Some of the components making up the model assembly 23
Figure 2-2: Exploded view of the model assembly 24
Figure 2-3: Dummy model created and assembled in CATIA 24
Figure 2-4: Local reference frame on the prototype 26
Figure 2-5: Modified final model of the MIMA 27
Figure 2-6 - Final up-to-date mockup to be tested 27
Figure 3-1: Dummy mass used for the single damper testing 32
Figure 3-2: Vibration generators (shakers) used in the tests performed 33
Figure 3-3: ENDEVCO 27A11 accelerometer used for vibration measurement 34
Figure 3-4: Data flow from the accelerometer to the PC 35
Figure 3-5: Data acquisition system 35
Figure 3-6: The basic Frequency Response Function used 36
Figure 3-7: The convention used for the directional parameters of the damper 37
Figure 3-8: Mechanical testing setup schematics for the single damper vertical direction 38
Figure 3-9: A fine measurement setup with four accelerometers 39
Figure 3-10: Mechanical testing setup schematics for the single damper radial direction 40
Figure 3-11: Accelerometer configuration used for the radial testing of the dampers. 41
Figure 3-12: A sample sweep sine with amplitude equal to 33g and increasing frequency 42
Figure 3-13: The critical sine and the random vibration environment 44
Figure 3-14: FRF comparison of the same damper response to a 0.5g sweep and a 5g sweep sine 45
Figure 3-15: FRF comparison of the same damper response to a 20g sweep and a 33g sweep sine 46
Figure 3-16: FRF comparison of the damper responses at different preloads. 48
Figure 3-17: Placement of the accelerometers on the mass for the transversal testing. 50
Figure 3-18: FRF of the three accelerometers versus the reference sweep sine 50
Figure 3-19: Comparison of the FRFs of the three fully closed dampers to a 20g sweep 52
Figure 3-20: The three dampers synchronized in terms of response. 53
Figure 3-21: Comparison of FRFs of the dampers in three different configurations. 55
Figure 3-22: Low level Resonance Search response of the damper in Config 2 and in Config 3. 56
Figure 3-23: Time-frequency plot for the damper channel in the testing of Config 2 57
Figure 3-24: Axial sweep sine test at 20g in the range 20-1000 Hz. 59
7
Figure 4-1: Formulation of a linear model 63
Figure 4-2: The actual damper-mass assembly 64
Figure 4-3: Identification of the natural frequency using the FRF from a sweep sine test at 5g. 67
Figure 4-4: FRF Magnitude and Phase plots comparison 69
Figure 4-5: FRF from 20g sweep sine test of a damper in Config 3. 70
Figure 4-6: Comparison of the mathematical and experimental FRF 72
Figure 4-7: The mathematical FRF fit with the adjusted modal parameters 73
Figure 4-8: The rotational motion at the tip of the damper. 75
Figure 4-9: Accelerometer configuration for an axial low level sweep sine test. 76
Figure 4-10: Low level sweep sine FRF in the region of the rotational resonance of the damper. 76
Figure 5-1: The CAD mockup assembled in CATIA. 81
Figure 5-2: Mockup CAD model with the necessary connections defined. 82
Figure 5-3: The model with fastened connection properties 83
Figure 5-4: The linear tetrahedron 83
Figure 5-5: A course mesh compared to a fine mesh. 84
Figure 5-6: Structural modes of vibration of the mockup 86
Figure 5-7: Spring Virtual Parts connected at three point on the mockup. 87
Figure 5-8: Clamped boundary condition for the spring elements. 88
Figure 5-9: Stiffness coefficients estimated from single degree of freedom tests 89
Figure 5-10: First three modes of vibration at (a) 131 Hz (b) 151 Hz (c) 264 Hz. 91
Figure 5-11: Displacement Vector plot of the vibration mode 1 at 131 Hz. 92
Figure 5-13: The axes of rotation for the first two modes of vibration. 93
Figure 5-15: Mockup COG position vs. Damper Triangle Moment Center. 95
Figure 6-1: The three testing directions shown on the mockup and damper designations. 98
Figure 6-2: Signal acquisition and feed schematics for the dummy model testing. 98
Figure 6-3: Z axis testing setup with the accelerometer positions. 99
Figure 6-4: X axis testing setup with the accelerometer positions. 100
Figure 6-5: Y axis testing setup with accelerometer positions. 101
Figure 6-6: Resonance search reference input profile plot. 102
Figure 6-7: Sweep sine reference input profile plot. 103
Figure 6-8: Random reference input profile plot. 104
Figure 6-9: Z axis sweep sine test time history. 106
Figure 6-10: Z axis FRFs 107
Figure 6-11: Z axis random test time history. 108
Figure 6-12: Z axis random FRF plot. 108
Figure 6-13: X axis sweep sine test time history. 109
Figure 6-14: X axis sweep sine FRF plot. 110
Figure 6-15: X axis random test time history. 111
Figure 6-16: X axis random FRF plot. 111
Figure 6-17: Y axis sweep sine test time history. 112
Figure 6-18: Y axis sweep sine FRF plot. 113
Figure 6-19: Y axis random test time history. 114
8
Figure 6-20: Y axis random FRF plot. 114
Figure 6-21: FRF Magnitude plot 116
Figure 6-22: FRF Magnitude plot for the Y axis testing 117
Figure 7-1: Damper name designation given. 121
Figure 7-2: Vibration mode shapes clockwise at (a) 101 Hz (b) 112 Hz (c) 175 Hz 128
Figure 7-3: Weighing profile plotted. 130
2
Figure 7-4: Evaluated FRF profile for the three axes in X loading 131
2
Figure 7-5: Weighed FRF profiles for X loading 131
2
Figure 7-6: FRF profile for the three axes in Y loading 132
Figure 7-7: Weighed FRF2 profiles for Y loading 133
2
Figure 7-8: Evaluated FRF profile for the three axes in Z loading 134
2
Figure 7-9: Weighed FRF profiles for Z loading 134
2
Figure 7-10: FRF profile measured for the three axes in X loading 135
2
Figure 7-11: Weighed FRF profile for X loading 136
2
Figure 7-12: FRF profile measured for the three axes in Y loading 137
2
Figure 7-13: Weighed FRF profile for Y loading 137
2
Figure 7-14: FRF profile measured for the three axes in Z loading 138
2
Figure 7-15: Weighed FRF profile for Z loading 139
2
Figure 7-16: FRF profile measured for the three axes in X loading 140
2
Figure 7-17: Weighed FRF profile for X loading 140
2
Figure 7-18: FRF profile measured for the three axes in Y loading 141
2
Figure 7-19: Weighed FRF profile for Y loading 142
2
Figure 7-20: FRF profile measured for the three axes in Z loading 143
Figure 7-21: Weighed FRF2 profile for Z loading 143
Figure B-1: Time history of out-of-plane vibrations for the Z sweep sine testing 159
Figure B-2: FRF of out-of-plane vibrations for the Z sweep sine testing 160
Figure B-3: Time history of out-of-plane vibrations for the Z random testing 161
Figure B-4: Time history of out-of-plane vibrations for the Z random testing 161
9
List of tables
Table 1-1: Sine environment profile acting on the MIMA. 13

Table 1-2: Random environment plot 14
Table 1-3: Pyroshock profile 14
Table 2-1: Initial and desired properties of the model 25
Table 2-2: Properties of the final model versus those of the desired 28
Table 3-1: Comparison of damper modal parameters corresponding to different preloads 48
Table 3-2: Preload distances on the dampers 53
Table 4-1: Summary of the stiffness coefficients estimated for the final damper configuration. 79
Table 5-1: Comparison of the advantages and the disadvantages involved in experimental methods
of a shaker test to those of Finite Elements methods. 80
Table 5-2: Summary of the results of the FEM analysis for the structural modes of the mockup. 85
Table 5-3: Natural frequencies of the system corresponding to the first three modes of vibration. 90
Table 6-1: Resonance search profile properties. 102
Table 6-2: Sweep sine excitation profile. 103
Table 6-3: Random reference excitation 104
Table 6-4: Experimental modes compared to those of the FEM analysis. 119
Table 7-1: Computed first three natural frequencies of the FE model 121
Table 7-2: Stiffness values estimated and used in the FE model 123
Table 7-3: Results of the various FE analysis runs with different damper properties. 124
Table 7-4: Systematic study of the effect of damper stiffnesses and the results 125
Table 7-5: Various analyses changing the radial stiffness and the results 127
Table 7-6: Various analyses changing the axial stiffness for a given radial stiffness 127
Table 7-7: Weighing profile to be multiplied by the FRF squared profile 129
Table 7-8: RMS values for the weighed FRF2 profiles for the X loading condition 132
2
Table 7-9: RMS values for the weighed FRF profiles for the Y loading condition 133
2
Table 7-10: RMS values for the weighed FRF profiles for the Z loading condition 135
2
Table 7-11: RMS values for the weighed FRF profiles for the X loading condition 136
2
Table 7-12: RMS values for the weighed FRF profiles for the Y loading condition 138
2
2
Table 7-14: RMS values for the weighed FRF profiles for the X loading condition 141
Table 7-15: RMS values for the weighed FRF2 profiles for the Y loading condition 142
2
Table 7-17: Results of the performance assessment summarized 144
Table 7-18: Suggested optimum spring properties and the resulting modes of vibration. 146
10
Chapter 1: Introduction
1.1 Mars Infrared Mapper (MIMA)
The Mars Infrared Mapper (MIMA) is a miniaturized FT-IR

spectrometer which is being developed for the ESA ExoMars Pasteur
mission. MIMA has been designed to provide remote measurements of
mineralogy and atmosphere of the scene surrounding a Martian rover
and guide it to key targets for detailed in site measurements by other
rover experiments.[1] The MIMA is capable of measurements typically a
few tens of meters away searching for evidence of water and of water-
related processes like carbonates sulfates clay minerals and possibly,
organics.[2]
The design is based on the scheme of the double pendulum with corner
cubes which has been successfully implemented in similar instruments
(PFS) designed and qualified for Mars Express and Venus Express.[3]
The Descent Module of the ExoMars mission will carry a 210 kg mobile
Rover to the surface of Mars equipped with a drill, a sample preparation
and distribution system (SPDS) and the 16.5 kg Pasteur Payload of
remote-sensing, in-situ and analytical instruments, including the
MIMA.[4]
The rover payload, PASTEUR, with its 12 instruments mounted on it,

including the MIMA, is shown in Figure 1-1.
11
Figure 1-1: The Rover Module equipped with the MIMA designed for the ESA ExoMars
Mission[5]
1.2 MIMA environmental conditions
The MIMA has to cope with several vibration environments acting

during launch and landing. The instrument mounted on the rover with
the given axis system notation is shown below in Figure 1-2.
12
Figure 1-2: Positioning of the MIMA on the rover and the defined axis system (source:
MIMA testing documents)
The mechanical excitations acting on the MIMA are identified as mainly

the sine and the random vibration environment and the pyroshocks
which are explained in detail in the following sections. [6]
1.2.1 Sine Environment
There is a specific sine profile acting on the MIMA during launch based
on the defined axis system given as
Table 1-1: Sine environment profile acting on the MIMA.
Amplitude [g]
Frequency [Hz] X, Y axes Z axis
5 1 1
20 33 33
100 33 33
13
1.2.2 Random environment
The random vibrations acting on the MIMA are given in terms of Power
Spectral Density (PSD) [g2/Hz] defined for different directions as
Table 1-2: Random environment plot
PSD [g2/Hz]
Frequency [Hz] X, Y axes Z axis
20 0.007 0.0174
100 0.1720 0.43
400 0.1720 0.43
2000 0.007 0.0174
1.2.3 Pyroshocks
Pyroshocks are the excitations generated by the explosive devices that

allows the separation of different parts of assembly. The content of the
excitation is commonly given as Shock Response Spectrum (SRS), the
maximum of the decaying, oscillatory response of a structure to the high-
amplitude and high frequency of the mechanical excitation. The
frequencies that comprise this oscillatory response can extend to
thousands of Hertz and beyond. They are a subset of the resonant
frequencies of the structure. [7]
The pyroshock profile given in terms of Shock Response Spectrum (SRS)

is shown in the table below
Table 1-3: Pyroshock profile
Frequency [Hz] SRS [g]

100 25
300 400
2000 1500
10000 1500
14
1.3 The need for a vibration damping system
Having discussed the environmental conditions acting on the

instrument, here the necessity of implementing a vibration damping
system is addressed. The random environment that the MIMA is
subjected to, acts within a wide frequency band resulting in dangerous
excitations of the instrument resonances. The optics of the MIMA is the
most critical part and it becomes even more vulnerable at high
frequency excitations due to the presence of natural frequencies above
500 Hz. This phenomenon can be modeled using an appropriate
weighing function for the dynamic response as shown in Figure 1-3.
25
20
Weighing Factor
15
10
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 1-3: Weighing factor on the MIMA as a function of excitation frequency
This plot suggests that accelerations on the MIMA at frequencies above

500Hz are almost 20 times more effective than accelerations at lower
frequency. The maximum gain is the typical amplification of low damped
structures. To show the effect of such a function, a typical Frequency
Response Function (FRF) of a simple first order system is considered.
Assuming two systems with the given natural frequencies

, = 150
, = 250
15
and identical damping ratio
= 0.3
The FRF of a first order system is given by the mathematical relation

+ 2
=
− + 2
And so the FRF squared profiles of these two systems can be plotted as
(see Appendix A for the complete procedure)
FRF2 vs frequency
4
System 1
3.5 System 2
2.5
PSD
1.5
0.5
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
frequency [Hz]
Figure 1-4: FRF2 profiles of System 1 and System 2
From this plot it can be observed that the FRF2 profiles of the two
described systems are only shifted in the frequency domain with respect
to each other with more or less similar amplification trends.
However when the weighing function given in Figure 1-3 is applied to

these profiles, the plot becomes
16
Weighed FRF2 vs frequency
40
System 1
35 System 2
30
25
Weighed PSD
20
15
10
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
frequency [Hz]
Figure 1-5: Weighed FRF2 profiles of System 1 and System 2
If a uniform Power Spectral Density [m2s-4Hz-1] is applied to the

systems, the computed RMS values are given as
= 4.8
= 9.7
Clearly, now there is a significant difference between the behavior of

these two systems when the conditions of the MIMA are considered. As
the amplifications in the response are shifted to higher frequencies, their
effect on the MIMA becomes even more dramatic. This conclusion
emphasizes the necessity of implementing a damping system to optimize
the resonances of the system to achieve a certain cutoff frequency, above
which the amplifications of the response are dampened down to low
levels, thereby protecting the MIMA optics from high frequency
excitations.
17
1.4 The dampers
Due to the previously explained environmental conditions acting at

MIMA-Rover interface, the necessity of a damping system arises in order
to reduce the vibrations transmitted to the instrument internal
components and mostly to the optical elements.
The dampers were developed and manufactured by the MIMA team with
the help of the experience gained from the previous PFS dampers, which
utilized a somewhat similar concept.
The damping elements are Silicon rubber O-rings, held in place and
separated by plastic spacers. The principal elements housing the silicon
and nylon rings are made of titanium to minimize the heat exchange
between MIMA and the rover. There are basically three such titanium
parts including an inner element supporting the rings from the inside.
An outer element covers the periphery of the rings and a threaded
element is used to tighten the ring assembly, packing them closer.
By acting on the threaded cap, the damper characteristics can be

partially changed. Increasing the tightening of the cap, the ring
elements are packed closer and closer, thereby resulting in an increase
of the overall stiffness of the damper. On the other hand, loosening the
tightening torque, the ring elements are relieved and a less stiff damper
can be obtained. This is a great advantage since it allows the
optimization of the damper behavior to meet the mechanical behavior
objectives.
18
Figure 1-6: Outside view of a tightly packed damper
1.5 Objectives
The main objective of this study is to perform a complete analysis

including the optimization and verification of the MIMA–Damper
system in terms of mechanical vibrations it is subjected to, and to
suggest further improvements regarding this matter. In this scope, a
detailed work will be performed regarding:
• the modification and creation of the dummy model that is being

used as a replacement for the actual instrument for experimental
purposes. The modified model will account for the changes
occurred recently in the MIMA and provide a satisfactory
representation of the actual instrument. A computer model and
several software tools will be used to obtain the modified dummy.
19
• the testing of the dampers individually on the shaker in a single
degree of freedom setup in various orientations. The dampers will
be tested and then tuned with respect to the dynamic response by
changing the preload on them. Furthermore, the damper
configuration may be modified by adding or subtracting certain
components if the requirements cannot be met with only the
preload change.
• the modeling of the dummy inertial equivalent with FEM

software. The model should be consistent with the behavior of the
dummy in the testing conditions. The dampers must be
represented as proper elements having adjustable directional
parameters like the actual components.
• testing of the dummy in order to visualize the frequency response

of the actual dynamic system. Besides providing invaluable
information about the actual behavior of the MIMA, the results of
these tests will be used to check the accuracy of the Finite
Elements model.
• identifying the changes in the tuning of the dampers allowing to

achieve improvements of the dummy response, by using the
parametric analysis with the Finite Elements model.
• testing of the dummy - if possible - with the optimally tuned

dampers, verifying the behavior of the optimal system and thus
validating the numerical analysis.
20
Figure 1-7: The diagram representing the objectives of the study
21
Chapter 2: Design and manufacturing
of a dummy model for testing
2.1 Necessity of a dummy model
During the course of this work numerous tests will be performed, some
of which are quite strong and potentially dangerous to the subject of the
test. Therefore using the actual MIMA for all these testing purposes is
out of question. This results in the necessity of an appropriate substitute
that can be tested instead of the actual instrument.
In order to represent the inertial properties of the actual interferometer,

a dummy model of the instrument has been designed to explore the basic
size, feel and mechanics of the instrument without simulating the actual
function or exact visual appearance.
2.2 Computer modeling of the prototype
The design and the optimization of the dummy prototype is performed

using software tools. For this purpose, CATIA v5 is chosen for its
simplicity and powerful computational capabilities.
2.2.1 Part design
The CAD design is performed by first providing accurate dimensions of

the model and with respect to these, creating the individual parts
making up the complete dummy (see Figure 2-8).
22
Figure 2-8: Some of the components making up the model assembly
2.2.2 Assembly
Once all the components are modeled in CATIA, one by one they are
assembled together in a rigid fashion paying attention to the geometrical
requirements. Figure 2-9 shows an exploded view demonstrating the
assembly process. This procedure makes use of coincidence and offset
constraints that will result in the correct assembly representing the
dummy prototype (see Figure 2-10).
23
Figure 2-9: Exploded view of the model assembly
Figure 2-10: Dummy model created and assembled in CATIA
24
2.3 Modification of the dummy prototype to achieve an
optimum model
2.3.1 The problem and the requirements
Previous studies on the damper-MIMA system utilized a dummy model

made up of three aluminum plates bolted together in a rigid way.
However recent changes in the MIMA device calls for an update on the
dummy model.
The initial prototype, shown in Figure 2-10, must be modified to obtain

an up-to-date model with the new inertial characteristics to have a
behavior during experimentation similar to that of the actual device.
The most significant parameters that should be set for the model are its
mass, center of gravity and moments of inertia. These desired values are
listed in Table 2-4 along with the initial properties that the prototype
possesses with respect to a reference frame shown in Figure 2-11.
Table 2-4: Initial and desired properties of the model
Initial Prototype Desired Model

Mass [kg] 1.14 1.3
X Y Z X Y Z
Center of Gravity [mm] 33.5 26.3 12.5 57.2 35.9 19.0
Inertia Matrix [kgxmm2*10-3] x y z x y z
Ix_ 3.14 -0.93 -0.49 5.47 -2.32 -1.22

Iy_ -0.93 3.30 -0.78 -2.32 6.51 -0.71
Iz_ -0.49 -0.78 4.22 -1.22 -0.71 8.03
25
Figure 2-11: Local reference frame on the prototype
2.3.2 The method of optimization
There are numerous ways to achieve the desired values such as

translating, swapping or modifying the components and adding or
subtracting of masses. The most significant restriction is the mass,
which should be 1.3 kg, suggesting a total added mass of 0.16 kg. Since
there are infinitely many ways to do this, a systematic approach is
implemented in which certain parameters are modified in a trial-and-
error fashion keeping in mind the center of gravity and inertia
requirements.
After various attempts, a simple but yet effective approach is found in

which the thick vertical aluminum plate is replaced by a similar plate
made of steel. This change allows the desired increasing of 0.15 kg in the
model mass. Therefore the mass of the model is equal to that of the
desired model. Finally, swapping the vertical plates the center of mass
and the inertia matrix is improved. The resulting model is shown below
in Figure 2-5.
26
Figure 2-12: Modified final model of the MIMA
Figure 2-13 - Final up-to-date mockup to be tested
27
Once analyzed, it is seen that the final model decided upon has the
following properties listed in Table 2-5.
Table 2-5: Properties of the final model versus those of the desired
Final Model Desired Model
Mass [kg] 1.3 1.3
X Y Z X Y Z
Center of Gravity [mm] 46.8 35.8 19.1 57.2 35.9 19.0
Inertia Matrix [kgxmm2*10-3] x y z x y z
Ix_ 4.61 -2.22 -1.06 5.47 -2.32 -1.22
Iy_ -2.22 5.92 -1.41 -2.32 6.51 -0.71
Iz_ -1.06 -1.41 6.68 -1.22 -0.71 8.03
2.4 Validity of the dummy model for testing purposes
A close comparison of the properties of the dummy model and the

desired values is required in order to decide whether this model can be
used as a substitute for the simulation of the real instrument.
As previously stated, the most significant properties to be optimized

were the mass, center of gravity and the moments of inertia around a
certain axis. As seen from Table 2-5, with the help of the added mass,
now the model has exactly the same mass as that of the device. However
this criterion is not enough on its own since not only the mass but also
28
the shape of the model influences significantly the dynamical behavior.
Because of this, the center of gravity and the moments of inertia with
respect to the given reference frame must be compared. As seen from
Table 2-5, the center of gravity and the moments of inertia are quite
close to the desired values with some small deviations. The maximum
error in the cog position is 18% and it occurs in the X direction and
similarly for the moments of inertia the maximum deviation is 0.7
[kgxmm2*10-3] and is present in the term Iyz. These result are acceptable
for the purposes of this analysis. In the light of all these facts, this model
is said to be good enough and it can be implemented as an experimental
substitute of the MIMA instrument with expectations of similar dynamic
behavior.
29
Chapter 3: Tuning of the dampers with
one degree-of-freedom testing
3.1 The need for damper tuning
The dampers are by design adjustable and versatile elements since they
include modular internal elements which can be tightened together
variably with the help of the threaded damper cap. This specific
characteristic allows for the modifications of damper properties such as
stiffness and damping capabilities in a wide range.
This unique characteristic brings forth the task of tuning the dampers in
order to achieve an optimum dynamic behavior. This objective will be
accomplished mainly by adding (or subtracting) internal ring elements
to the assembly as necessary and also by determining the optimum
preload, that is the amount of closure of the damper, through a series of
tests.
The MIMA will be mounted on three such dampers and therefore an

optimal adjustment for each of them is required. The first approach for
this study was to tune identically the dampers in order to obtain the
same dynamical behavior. However this is not the only possible solution
as will be shown at the end of this work.
3.2 Test methodology
In order to characterize the mechanical properties of the dampers, a

mechanical testing procedure was followed using a single damper and an
appropriate dummy mass.
30
The design of the testing procedure includes the selection of the setup
facility and sensors and the decisions upon the methods that are
implemented. These decisions are made considering the objectives and
the requirements of the mechanical testing. The various components
used and the procedure followed are described in detail in the following
sections.
3.2.1 Experimental equipment
The mechanical testing setup is designed in order to assure a good

performance regarding the results with a feasible cost.
The fundamental components of the setup are the inertial mass, the
damper, the shaker, the sensors and the data acquisition system, some
of which require further attention.
3.2.1.1 The mass

The actual MIMA device is supported by three dampers mounted on
certain points. Such a model cannot be used to test the behavior of a
single damper. Thus a representative dummy mass is designed and
implemented in the tests. This mass is of 0.35kg, which is approximately
one-third of the actual MIMA, and is thus suitable for the testing of a
single damper instead of three. As seen on Figure 3-14, the geometry is
designed such that it is axisymmetrical and ideally it has the center of
gravity along the damper axis and close to the damper tip.
31
Figure 3-14:: Dummy mass used for the single damper testing and section view showing
its center of mass
3.2.1.2 Shaker
A shaker is an electrodynamic exciter that is used to generate certain
vibrations to produce the desired environmental conditions. For the
scope of this work, they are used to simulate the probable vibrations
that may act on the MIMA during its take
take-off
off and landing on the surface
of Mars.
Two different shakers will be used for the purposes of this analysis, the
decision being done on their simplicity versus their capabilities. The
most significant
ificant properties of these shakers are their force, acceleration,
32
velocity and displacement limits. The smaller TIRAvib vibration
generator will be used for the single-damper testing procedures that
include a smaller mass being excited. On the other hand for the full
model testing, a larger and more capable shaker will be used (see Figure
3-15).
Figure 3-15: Vibration generators (shakers) used in the tests performed
Both of these vibration generators are capable of producing the high

level and high frequency sine and random vibrations that will be present
in the environmental conditions of the MIMA and therefore are
appropriate for the testing of the dampers.
3.2.1.3 Sensors
The stimulus for the testing of the dampers is in terms of acceleration
and the response is measured using accelerometers. For this purpose,
very small accelerometers designed specifically for measuring vibration
on mini-structures and small objects are implemented (see Figure 3-16).
33
These accelerometers offer high resonance frequency and wide
bandwidth with their light weight eliminating mass loading.[8]
Figure 3-16: ENDEVCO 27A11 accelerometer used for vibration measurement
For the single degree of freedom case several accelerometers are

employed for each test with appropriate voltage sensitivity and dynamic
performance.
3.2.1.4 Data Acquisition System

The electrical signal coming from the accelerometers have to be
conditioned and transformed into meaningful data before it can be
useful. To achieve this goal, a data acquisition system is necessary.
The signal coming from the accelerometer cannot be acquired

immediately but needs to be conditioned into a signal easy to read by
conventional instrumentation. Therefore the use of a signal conditioner
is critical unless it is already mounted inside the data acquisition board.
The data acquisition board samples the signal and sends it to the PC
where it is stored for further processing. Such a flow of data is shown
schematically in Figure 3-17. The actual system involving these
elements is shown in Figure 3-18.
34
Figure 3-17: Data flow from the accelerometer to the PC
Figure 3-18: Data acquisition system involving the signal conditioner, data acquisition
board and the PC.
35
3.2.2 Experimental Modal Analysis techniques
Experimental Modal Analysis is a methodology that enables a

structure’s modal properties (natural frequencies, modal damping
factors and mode shapes) to be determined from experimental
observations of its dynamic response under test conditions.[9][10] The
main objective of experimental modal analysis is the construction of a
mathematical model. However, as far as this study is concerned,
construction of a Finite Elements Model is by far more critical.
Nowadays, the vast majority of practical applications of experimental

modal analysis provide the measured data in the form of Frequency
Response Functions (FRF). Frequency Response Functions are normally
used to describe the input-output relationship of any system. It is with
the help of these response functions that the modal parameters can be
determined.
The study of the damper dynamics is performed mostly using these

techniques that will be detailed further on. However the main idea is
always the same, using an FRF describing the input-output relationship.
#
=
$
Figure 3-19: The basic Frequency Response Function used throughout the study as a
ratio of measured accelerations.
36
3.2.3 Test setup
There are three significant physical properties of the damper, which are
the stiffness in the axial direction, stiffness in the radial direction and
the rotational stiffness. The sense of these directions is given below in
Figure 3-20.
Figure 3-20: The convention used for the directional parameters of the damper
In order to assess, quantify and tune these three directional properties

of the damper, reliable test methods are designed for each case
generating a one degree of freedom motion.
3.2.3.1 Axial stiffness testing

First of all the dampers must be tested along their axis to see if they
satisfy a certain behavior in the axial direction. For this purpose, a
single damper is mounted on the shaker and the mass is bolted on top of
the damper such that deformations other than the ones in the axial
direction are restraint as well as possible. Accelerometers measuring the
reference stimulus and the resulting response are placed and fixed along
with their cables to assure no interference with the dynamics of the
system. Schematics of such a setup can be seen in Figure 3-21.
37
Figure 3-21: Mechanical testing setup schematics for the single damper vertical
direction
As seen from the above schematics the stimulus, i.e. the reference
vibration that the shaker imposes on the damper, is created with the
help of the Control PC. The control action is a closed-loop feedback
system that utilizes the signal from the reference feedback
accelerometer. Simultaneously, another accelerometer measures the
same reference acceleration and feeds it to the Acquisition PC for late
signal processing tasks. Other sensors located at the top of the dummy
mass are used to measure the response.
Considering that ideally the dummy mass has its center of mass along
the axis of the damper and that the damper and the mass are fixed
together restraining any undesired motion, the only motion that can
occur is in the axial direction and is due to the deformation of the
damping elements. However this expectation is not so close to the
reality. In practice, it is wise to account for a small offset in the dummy’s
center of mass due to poor manufacturing tolerances. This fact may lead
38
to other non-restrained
restrained displacements, causing some rotation of the
mass. Therefore it is reasonable to use two accelerometers on top of the
dummy mass (see Figure 3-22)) in order to measure the axial
displacements as well as any undesired rotations.
Figure 3-22:: A fine measurement setup with four accelerometers including one for
feedback, one to measure the reference and two for measuring the response.
3.2.3.2
.2 Radial stiffness testing
Due to the environmental conditions, the dampers are strained not only
in their axial direction but also in their radial direction. These vibrations
acting radially are as critical as the axial ones and thus the dampers
must be tested in this axis and their corresponding radial stiffness must
be assessed and optimize
optimized.
The setup for this procedure is quite similar to that of the axial stiffness
tests. The main difference in this case is that the stimulus must be in
39
the direction to amplify the deformations straining the radial stiffness.
For this purpose a large metallic block is used and the damper-mass
assembly is mounted on it with the damper axis being horizontal (see
Figure 3-23 for schematics).
Figure 3-23: Mechanical testing setup schematics for the single damper radial direction
As seen from the above, the acquisition and the flow of data is similar to
the setup discussed before. One particular point to be noted is the
placing of accelerometers for the optimal analysis of the damper
response. Figure 3-24 shows the positions of the three response
accelerometers on the dummy mass and two reference accelerometers of
the metallic block, one for storage and one for control feedback.
40
Figure 3-24: Accelerometer configuration used for the radial testing of the dampers.
Due to the fact that the center of mass of the dummy mass is not at the
damper center, the expected response is a combination of radial
translation as well as some rotations. Therefore three accelerometers are
required to analyze the behavior correctly. The accelerometer on the top
circular edge of the dummy mass measures the response in the direction
of excitation while the other two measure the axial accelerations and
also account for various rotations of the mass.
3.2.3.3 Rotational stiffness testing

The final property that the dampers must be assessed for is the
rotational stiffness. This parameter accounts for the small rotations at
the damper tip that the translational stiffness parameters cannot
explain.
The method for the estimation of rotational stiffness is to check the

Frequency Response Function from the various single degree-of-freedom
tests on the damper. With the help of an accelerometer couple, rotational
resonances can be identified and the stiffness estimation can be
performed accordingly.
41
3.2.4 Vibration stimulus for the single degree-of-freedom tests
The testing and the characterization of the dampers require a

determined stimulus that is the foreknown reference input. Therefore a
signal driving the shaker is generated and controlled in a closed loop
feedback manner by the control PC.
The environmental conditions of the MIMA include vibrations that are

sinusoidal, random or shocks by nature. However for the
characterization of single dampers, only the sweep sine will be used it.
Sweep Sine Signal
40
30
20
10
Amplitude [g]
-10
-20
-30
-40
0 1 2 3 4 5 6 7 8 9 10
Time [s]
Frequency vs Time
3
2.8
2.6
2.4
Frequency [Hz]
2.2
1.8
1.6
1.4
1.2
1
0 1 2 3 4 5 6 7 8 9 10
Time [s]
Figure 3-25: A sample sweep sine with amplitude equal to 33g and increasing
frequency
42
As seen in the above figure, the sweep signal is a sine with constant
amplitude and increasing frequency at a constant logarithmic rate
(usually defined in octaves/min). This provides a good method to excite a
wide band of frequency in a small amount of time.
The environmental conditions suggest a sinusoidal excitation with

amplitude of 33g. Therefore the dampers must be tested and their
parameters characterized at this vibration level. Characterization at any
other level may lead to some errors due to the high non-linearity of the
dampers. However this level is quite high and repeated applications may
result in the damaging of the internal ring elements. Therefore alternate
signals with lower amplitudes are implemented as well, such as at 0.5g,
5g, 20g, etc…
3.3 Tuning of the dampers
The tuning procedure includes numerous tests performed on the

dampers in a trial and error fashion varying each time some certain
parameter, the testing orientation and even the damper configuration
until an acceptable dynamic behavior is achieved. The most significant
factors involved in the tuning process are the effect of closing torque on
the dampers, the effect of excitation amplitude and the different
behaviors associated with different configurations of damping ring
elements.
3.3.1 Objectives of the tuning procedure
The main objective of the damper tuning procedure is to test the

dampers to see if they satisfy the dynamic response requirements and if
not, modify them in some manner to achieve the desired response.
The MIMA environmental conditions suggest two types of vibrations

acting on the instrument, the sine and the random vibration
environment. From these, the sine environment is more significant at
43
low frequencies (below 100 Hz) while the random one is more critical
above 100 Hz (see Figure 3-26).
Figure 3-26: The critical sine and the random vibration environment
In the light of these considerations, it is desired that no resonance occurs

below 100 Hz where the sine vibrations are critical. The resonance
should preferably be above 100 Hz, but not so high as to cause suffering
from the strong random vibration environment. The safe area can be
identified within the 120-170 frequency range.
An important outcome of the single damper tuning tests is the

estimation of the damper stiffness parameters so that they can be used
further on in computational applications, such as the Finite Element
Model development.
The mounted MIMA has three such dampers and due to some
manufacturing tolerances and imperfections, they are not identical. As
previously stated, the final objective of the tuning process is to test each
damper and tune them with respect to each other to obtain three
identical dynamic responses.
44
3.3.2 Excitation amplitude and damper non-linearity
As previously mentioned, the dampers are highly non-linear elements.

This means that their behavior is strongly dependent on the magnitude
of the incoming stimulus and therefore a formulated linear model is only
valid for a certain level of excitation.
To demonstrate this effect, various tests can be performed on the same

damper with different levels of excitation. This is a critical phenomena
of the single damper testing and qualification since the actual high level
vibrations cannot be applied on a frequent basis due to damage-
prevention concerns.
Below, in Figure 3-27, are plotted two Frequency Response Functions

corresponding to the damper response for a given reference excitation.
These plots represent two different tests performed on the same damper,
one test with a sweep sine at a constant 0.5g level and the other with a
sweep sine at 5g level, i.e. 10 times stronger than the previous one.
Figure 3-27: FRF comparison of the same damper response to a 0.5g sweep and a 5g
sweep sine.
45
There are obvious differences between the two responses of the same
damper. One significant difference is the decreasing of the natural
frequency at the higher level excitation. This fact is apparent from the
shift of the resonant peak to the left in the magnitude plot and also the
shift of the phase fall in the phase plot. The fact that the natural
frequency is adversely affected by the excitation amplitude is a
phenomenon that must be taken into consideration throughout the
course of this study. Another obvious difference is the increase of
damping at the 5g test. In this test, the resonant peak is significantly
reduced and the phase fall is rotated unlike the steep fall at low
excitation level.
Another test performed at higher excitation levels but within a smaller

frequency band exhibit similar results (see Figure 3-28). At 100 Hz the
33g plot has a greater increase in magnitude and a larger fall in terms of
phase, which suggests the shift of the natural frequency to a lower
value.
1,07
1,06
1,05
1,04
1,03
1,02
1,01
0,99
0,98
0
-0,5
-1
-1,5
-2
-2,5
-3
-3,5
-4
-4,5
-5
30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Frequency [Hz]
Figure 3-28: FRF comparison of the same damper response to a 20g sweep and a 33g
sweep sine.
46
3.3.3 Damper response and tightening preload
The dampers are by design adjustable elements so that their dynamic

response can be modified by simply increasing or decreasing the preload
on the ring elements with the help of the threaded cap acting against the
inner element. This preload can be measured accurately with the help of
a micrometer, thus allowing a systematic study of the effect of
tightening or loosening the dampers. However this preload cannot be
extremely small in order to prevent any opening during operation and it
cannot be too large as there is a limit to this preload due to the limited
travel of the regulation screw.
The effect of preload has been studied by testing the same damper
configuration with a certain excitation profile, gradually increasing the
preload and comparing the damper response. The following figures and
tables show a comparison of four different tests, run with the same
damper that was gradually loosened. The excitation profile is a sweep
sine with a 5g amplitude within a frequency band 20-1000Hz.
47
1,6
1,55
1,5
1,45
1,4
1,35
1,3
1,25
1,2
1,15
1,1
1,05
1
0,95
0,9
10
0
-20
-40
-60
-80
-100
-120
20 50 100 150 200 250 300 350 400 450 500 550 600 650 700
Frequency [Hz]
Figure 3-29: FRF comparison of the damper responses at different preloads.
Table 3-6: Comparison of damper modal parameters corresponding to different

preloads
Preload Natural Frequency

Peak Magnitude
[mm] fo [Hz]
Preload 1 30.3 850 1.6
Preload 2 30.5 800 1.4
Preload 3 30.7 650 1.4
Preload 4 30.9 550 1.4
A close examination of the above test results reveals the obvious effect of
damper preload on the damper response and naturally, its dynamic
properties. The gradual loosening of the damper shifts the phase plot to
the left resulting from the lower natural frequency and thus indicating a
48
lower stiffness. Conversely increasing the preload will increase the
natural frequency and the stiffness of the damper. On the other hand a
look at the peak magnitudes and the phase slopes suggests that the
damping is not so significantly altered by the different preloads.
This aspect of the dampers, i.e. the possibility to vary its stiffness and
thus its dynamic response, is a key advantage and is utilized to achieve
the damper tuning objectives discussed previously.
3.3.4 Radial behavior of dampers
Besides all the previously investigated factors affecting the damper

response, there is another significant consideration, which is the
directional properties acting on the response of the dampers. Two
different stiffness parameters are defined to represent this directional
behavior, the axial and the radial stiffness (recall Fig 3.7).
The translational stiffness in the radial direction is tested and evaluated

using the setup described previously in Section 3.2.3.2. This system
basically provides an excitation of the damper-mass assembly causing
deformations that are mainly driven by this stiffness parameter.
For the purposes of this test, three accelerometers are used to measure
motion of the dummy mass in a complete way since the motion is a
combination of the axial and radial deformations (see Figure 3-30 for the
visualization of the accelerometer positioning).
The FRF given below in Figure 3-31 demonstrates the behavior of the
damper when the stimulus acts in the transversal direction.
49
Figure 3-30: Placement of the accelerometers on the mass for the transversal testing.
1,8
FRF Modulus
Axial Center m s-2 Mod
1,6
Axial Edge m s-2 Mod
Radial m s-2 Mod 1,4
1,2
0,8
0,6
0,4
0,2
0
180
FRF Phase
150
Axial Center m s-2 Ph
Axial Edge m s-2 Ph 100
Radial m s-2 Ph
50
-50
-100
-150
-180
20 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 450 475 500
Frequency [Hz]
Figure 3-31: FRF of the three accelerometers versus the reference sweep sine acting in
the transversal direction.
50
This FRF requires a close examination in order to deduct some useful
conclusions from it. The first observation would be the existence of two
resonance regions from the magnitude plot. However from the phase plot
it is apparent that the motion around these regions is of a completely
different nature. At first peak, the response of the two axial
accelerometers diverge and they become out-of-phase. This suggests a
rotational mode rather than the deformation of the damper in the radial
direction. Later on in the frequency range, around 250 Hz, a resonance
in the radial direction is observed while in the axial one the
accelerometers continue to behave the same (constantly out-of-phase).
This fact suggests the existence of a purely transversal mode at that
frequency.
All the results obtained from the numerous tests in the transversal
direction are of critical importance and will later on be used for the
characterization of the damper elastic parameters (see Chapter 4).
3.3.5 Damper Response Synchronization
The first approach to get the desired instrument dynamic response was
based on providing three equally tuned dampers. However to achieve
this is not so straightforward since the equal configuration and the equal
preload on the dampers does not necessarily guarantee an equal
response. This is due to the manufacturing tolerances of internal
elements of the damper. Therefore, if an equal tuning is required, this
must be performed through a trial-and-error procedure slightly varying
the preload on the dampers and comparing the resulting responses.
A procedure such as this one was performed to obtain three dampers

with equal stiffness parameters, i.e. similar natural frequencies.
Initially all the dampers were closed at the same preload distance and
were tested in this configuration to have some initial idea about the
internal elements’ differences. For the testing, a 20g Sweep Sine was
used. The Frequency Response Function was calculated between the
reference stimulus and the damper response. A comparison of the
results is given below in Figure 3-32.
51
2,1
2
1,9
1,8
1,7
1,6
1,5
1,4
1,3
1,2
1,1
1
0,9
0,8
0,7
0,6
0
-10
-20
-30
-40
-50
-60
-70
-80
-90
-100
20 50 100 150 200 250 300 350 400 450 500 550 600 650 700
Frequency [Hz]
Figure 3-32: Comparison of the FRFs of the three dampers to a 20g sweep in their fully
closed state.
Investigating the above plot, it is obvious that the components of the

three dampers are not identical due to some deviations within the
tolerances in manufacturing. While dampers 1 and 2 are acceptably
similar, Damper 3 is much more stiff with respect to the other and most
be loosened. Therefore a tuning session is required in which the
responses of the three dampers can be synchronized. By varying the
preload distance slightly a better combination can be achieved such as
the one in Figure 3-33.
52
2,1
2
1,8
1,6
1,4
1,2
0,8
0,6
0,5
0
-20
-40
-60
-80
-100
-120
-130
20 50 100 150 200 250 300 350 400 450 500 550 600 650 700
Frequency [Hz]
Figure 3-33: The three dampers synchronized in terms of response.
The preload on the dampers in terms of distances that result in the

response shown above is given below.
Table 3-7: Preload distances on the dampers
Preload [mm]
Damper 1 29.2
Damper 2 29.2
Damper 3 30.2
In this case the dampers are quite similar in terms of natural

frequencies and thus have similar stiffness parameters. However looking
at the peak magnitudes and the phase fall slopes it is apparent that they
have different damping ratios. Clearly Damper 3 has a lower damping
than the others. This phenomenon can be explained considering the
effect of several factors including the tolerances on the nylon ring
53
spacers and the rubber elements being forced against the titanium
producing friction. Unfortunately this behavior cannot be modified by
changing the preload and therefore it has been accepted as it is and the
tuning process is concluded.
3.3.6 Configurational modifications of damper assembly
The damper elements are, by design, adjustable elements made up of

small modular rings assembled together and tightened with a certain
preload. The fact that the ring elements are interchangeable allows for
the formation of different assembly configurations that might have some
advantages with respect to the others. This property is several times
taken advantage of throughout this study as it is considered necessary.
3.3.6.1 Changes to the damper assembly

The initial assembly of the damper ring elements included four silicon
rings separated and supported by 4 nylon spacers (two small in between,
two large at the tips).
A more closely-packed damper configuration can be achieved by adding

some extra ring elements into the assembly. This results in a stiffer
damper. Besides increasing the stiffness, a tightly closed damper is
favorable since the chances of having some free motion before the actual
elastic deformation is eliminated significantly.
To achieve a tightly-packed damper, the first step was to add a small

nylon ring that is half as thick as the ones already present in the
assembly. This element is then inserted between the rigid damper
structure and the silicon ring to achieve the damper configuration called
Config 2.
The above configuration is definitely more packed than the initial one.
However, it is possible to obtain an even tighter assembly (Config 3)
within the damper’s mechanical limits by adding another half nylon ring
similar to the previous one on the opposite side.
54
3.3.6.2 Effects of the insertion of additional rings on the damper
response
The configurations described above were all tested on the shaker with
the one degree-of-freedom single-mass setup to compare the differences
in the response and thereby describe the effect of introducing additional
ring elements into the damper assembly.
The comparison was again performed in terms of FRFs between the

reference stimulus and the damper response. In Figure 3-34 below, three
such FRFs corresponding to the three configurations described
previously are plotted within a frequency band of 40-100 Hz as a
response to a 33g sweep excitation.
1,9
1,8
1,7
1,6
1,5
1,4
1,3
1,2
1,1
1
0
-2
-4
-6
-8
-10
-12
-14
-16
-18
-20
-22
-24
40 45 50 55 60 65 70 75 80 85 90 95 100
Frequency [Hz]
Figure 3-34: Comparison of FRFs of the dampers in three different configurations.
The above figure verifies all the expectations from the configurational
modifications on the dampers. Looking at the phase trend, it can be
55
stated that the natural frequency increases as additional elements are
inserted into the assembly, thus increasing its stiffness. On the other
hand, the FRF Magnitude plot reveals another useful information,
which is the increase in the damping caused by the presence of extra
rings. The rings which are fitted into the metallic damper structure
cause friction and dissipate energy. Therefore the increase in number of
such elements induces a higher damping ratio on the system.
The effect of the additional ring on the natural frequency of the damper
can be better visualized from the Figure 3-35, where a low level sine-
sweep (Resonance Search, 0.5g sweep sine) is performed within a large
frequency band.
10
9
8
7
6
5
4
3
2
1
0
0
-20
-40
-60
-80
-100
-120
-140
-160
-180
1
0,8
0,6
0,4
0,2
0
20 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400
Frequency [Hz]
Figure 3-35: Low level Resonance Search response of the damper in Config 2 and in
Config 3.
The above plot gives a comparison of the response of the two given
configurations of the dampers. The initial observation from the FRF of
Config 2 is the presence of periodic peaks in the spectrum which are not
56
actually evident in the time history. In order to investigate the nature of
these peaks, the time-frequency plot corresponding to the data obtained
with this configuration is given
Figure 3-36: Time-frequency plot for the damper channel in the testing of Config 2
Examining the above plot and the coherence function, it is concluded

that these peaks are in fact constant frequency incoherent components
in the signal caused by some electrical noise whose source has not been
identified, but that was present in this specific test run. The region of
interest of the FRF however is not affected dramatically and by
excluding the frequencies of the disturbances, the obtained test results
can be used for comparison.
As the above plot suggests, adding another nylon ring results in a stiffer
damper with a higher natural frequency (in this case almost 100 Hz
higher). Still there is a physical limit to the number of additional rings
that can be added to the damper assembly. With Config 3, i.e. with two
small half nylon rings included, the damper is in the maximum packing
configuration, no additional element can be added without changing the
design of the metallic parts.
57
3.3.6.4 Energy dissipation of spacers
The environmental conditions that the dampers face during the actual
operation are quite harsh and so must be the testing procedure for their
qualification. The sweep environment foresees excitations up to 33g in
amplitude while the random environment imposes vibrations with RMS
value of 16 g. Provided that the friction is one of the phenomena
exploited for the dissipation of the vibrations power, some wearing can
take place.
The dampers using nylon elements have, in some cases, shown a change
in performance that was not negligible after the high level sine test. For
this reason a different material has been adopted, being the wearing
related to the local temperature increase, a material with high operating
temperature such as the polyimide resins would have been the best
choice. A glass fiber reinforced nylon was eventually adopted because of
its immediate availability. This proposed change proved to solve the
wearing issue.
3.4 Dampers in their final tuning and configuration
After numerous tests performed on the dampers, trying over and over
changing several parameters and modifying the configurations as
described in detail in the previous sections, a satisfactory final damper
tuning is achieved.
This damper configuration has two additional nylon spacers in the

middle of the assembly with respect to the initial one in order to achieve
a more tightly closed damper. Moreover, one of the large nylon rings at
the tip of the assembly is replaced with a dimensionally identical one
made of glass fiber reinforced nylon to avoid the problem of melting due
to high energy dissipation.
The measured preload is around ℎ ≅ 29.2, almost fully-closed. With

The damper’s preload is adjusted so that the damper is tightly packed.
58
this approach, a quite stiff and stable (no free translation of inner
elements) damper is obtained.
Results in terms of FRF to sweep sine tests performed on this specific

final damper are given in the following figure below.
1,4
1,3
1,2
1,1
0,9
0,8
0,7
0,6
0
-20
-40
-60
-80
-100
-120
20 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz]
Figure 3-37: Axial sweep sine test at 20g in the range 20-1000 Hz.
The above plot demonstrates only the response of Damper 1 previously

shown on Figure 3-33 along with the other dampers. This single FRF
can be accepted as a representative behavior for all the identically tuned
dampers and used in the following analyses accordingly.
Recalling the MIMA environmental conditions, the behavior of this

single damper seems to be acceptable. The natural frequency seems to
be around 300-400 Hz which is well above 100 Hz but still not so high,
as to reduce the advantages for the random environment. In addition,
59
the overall FRF does not have a very high amplification (1.3 maximum).
This is also a positive result suggesting this damper configuration might
as well be the final one.
3.5 Results and discussion
As mentioned before the damper design is adjustable and quite versatile,

in the sense that it is possible to tune the dampers in order to obtain a
desired response for a given excitation. For this purpose, many
experiments were performed to test each and every aspect of the various
possibilities of the dampers.
All the tests were performed with a representative dummy mass one-
third of the original instrument to test a single damper each time, in a
one degree-of-freedom system. The setup and the procedures were
designed to comply with Experimental Modal Analysis techniques that
allow the characterization of the system through its modal parameters,
such as natural frequency, damping ratio and the vibration modes.
Three test setups were used to test and evaluate the three directional
properties of the dampers. Even though this procedure was quite
straightforward for the axial direction, evaluation of the radial
properties required transversal excitations which resulted in combined
modes. These modes had to be distinguished in the FRF by using several
accelerometers to visualize the motion completely.
The dampers had to be tuned in order to get the optimum tightening

preload and the best inner assembly. During this procedure numerous
tests were performed in all orientations, varying the parameters to be
optimized in a trial-and-error fashion, until finally a satisfactory damper
response is achieved.
Surely the decision upon the optimum damper cannot be made just by
looking at the single damper-single mass tests. This is a method for the
characterization and provides a good insight. The dampers should then
act satisfactorily when the three of them are used together with the
60
MIMA. Once the damper parameters are extracted from the single
degree-of-freedom tests, also the FEM analysis of the system can be a
resourceful and fast tool. All these steps to achieve the final system will
be explained in detail in the following chapters.
61
CHAPTER 4: Damper properties
estimation
The behavior of the dampers, as previously discussed, are highly

nonlinear. Their response is strictly dependent on the amplitude of
excitation applied on them. Even though for experimental purposes this
is not really significant, in order to make a computational analysis, be it
analytical or finite elements, this condition imposes some problems.
Therefore it is convenient to describe the dampers with linear modal
parameters.
4.1 Excitation amplitude and linear modal parameters
The conclusion reached before from the single degree-of-freedom testing

was that the natural frequency of the damper varied with the amplitude
of the incoming excitation (recall Figure 3-17). It is known that for a
linear single degree of freedom system the stiffness and the natural
frequency are correlated with the expression:
(
= '

where is the natural frequency in [rad/s], k the stiffness coefficient

and m the mass of the system connected to the spring. Therefore with
the mass being constant, a measured increase in the natural frequency
means that an increase of the stiffness occurred and vice versa. Finally
it can be said that the system acts in a less stiff way to a stronger
excitation while it appears to be stiffer when the excitation levels are
lowered.
62
From the previously obtained FRFs, magnitude levels of the resonant
peaks and the change in the slope of the phase plot suggest a variation
of the damping ratio with different excitation levels. As the result of the
experience gained from numerous tests, it can be stated that the
damping characteristics become more pronounced as the excitation level
is increased while the damping is greatly reduced at low levels.
Since all the properties associated to the experimental modal analysis

are dependent on the incoming excitation, defining a single linear model
describing every possible response of the dampers is not possible. A
single model describes only the excitations using which it has been
formulated, as described by the diagram below.
ωo,1 Linear
Response
Excitation 1 System
1
ζ1 Model
modes1 1
ωo,2 Linear
Response
Excitation 2 System
2
ζ2 Model
modes2 2
ωo,n Linear
Response
Excitation n System
n
ζn Model
modesn n
Figure 4-38: Formulation of a linear model with the linear modal parameters valid only
for a certain kind of excitation.
63
4.2 Analytical formulation of the test setup
The single damper – single mass test setup is a simple one, aimed at
achieving a one degree of freedom system with given mass m, stiffness
coefficient k and damping coefficient c. The equivalent system with these
properties representing the test setup is shown in Figure 4-39.
Figure 4-39: The actual damper-mass assembly represented by the linear one degree of
freedom system.
Here x(t) denotes the known displacement of the support while y(t)
denotes the absolute displacement of the mass. The equation of motion
governing such a system is then given by [11]
)* + +), + () = +-, + (-
Then in order to describe a relation between the input and the output of
the system, the frequency response must be defined [12]
#
=
$
To express the frequency response, the time variables are transformed
into frequency domain
). = #/ 012
64
), . = #/ 012
)* . = − #/ 012
-. = $/012
-, . = $/ 012
Replacing these into the equation of motion and simplifying
3− #/ 012 4 + +3#/012 4 + (3#/ 012 4 = +3$/ 012 4 + ( ($/ 012
− + + + (#/012 = + + ($/ 012
Finally the frequency response describing the relationship between the

input and the output of the system can be written as
# + + (
= =
$ − + + + (
The frequency response function is a complex function which can also be

described in terms of its magnitude and phase
+ + (
5 = | | = 7 7
− + + + (
+ + (
8 = arg3Hjω4 = arg
− + + + (

Here Y/X is calculated as the ratio of the absolute displacement of the

mass to displacement of the base. However, multiplying the numerator
and denominator of this ratio by ω2 shows that H(jω) also represents the
ratio of the absolute acceleration of the mass to the acceleration of the
base. This result is critical since for all the tests regarding this study,
the reference input and the measured output is in terms of acceleration
and the Frequency Response Function is calculated using these
measurements.
65
4.3 Estimating linear parameters from test data
The single damper one degree-of-freedom tests provide reference

stimulus and response acceleration measurements which then can be
used to calculate FRFs. These FRFs are useful tools in the extraction of
modal parameters from the test data.
The most straightforward method in estimating linear modal

parameters is to check for resonance and then use the values around
this region to calculate the desired parameters. However, this approach
is not always feasible since when strong excitation levels are involved, at
the resonance the response becomes dangerously critical and damage to
the dampers is probable. Therefore for such tests an alternative
approach is used in which the frequencies in a safe region are excited
and the dynamic properties are then extrapolated from the measured
FRF.
4.3.1 Modal parameters from resonance
If for any test with the single damper, the frequency band of the
excitation is large enough to cause the excitation of the natural
frequency, resonance occurs. Then this resonance can easily be identified
using the Frequency Response Function and the modal parameters can
be extracted with this information.
The stiffness coefficient k [N/m], for the damper for any test is simply
given by
( =
where is the natural frequency in rad/s and is the mass in kg

connected to the damper.
Below in Figure 4-40 is given the FRF of a sweep sine test at 5g within
the frequency band of 20-1000 Hz. An examination of this plot clearly
66
shows that the damper undergoes resonance between this frequency
range.
1,5
1,45
1,4
1,35
1,3
1,25
1,2
1,15
1,1
1,05
1
0,95
0,9
0,85
0,8
0
-20
-40
-60
-80
-100
-120
-140
20 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz]
Figure 4-40: Identification of the natural frequency using the FRF from a sweep sine
test at 5g.
Then the next step is to identify the modal properties of the damper.
This is not so straightforward due to the presence of damping and
therefore the resonance is not observed at the -90° crossing of the phase
but is shifted a slightly to lower frequencies.
Previously, a mathematical expression for the FRF was calculated as

+ + (
=
− + + + (
This expression can be arranged written in terms of the damping ratio
67
+ +
= =
+>? 2√(
+ 2
=
− + 2
Now, adjusting and the damping ratio , the mathematical model can
be tuned to match the FRF calculated from the experimental data given
in Figure 4-40. For this purpose, a MATHCAD script is used to iterate
for the unknown variables and plot the mathematical function against
the experimental FRF given above (details of this script can be found in
Appendix A).
Investigating the experimental FRF and making several iterations with

the model, the following parameters can be chosen,
= 450
= 0.4
The mathematical model with the chosen parameters results in the FRF
plot given below in Figure 4-41.
68
FRF MAGNITUDE
2
1.5
Magnitude
G ( f ⋅ 2π)
1
Mag_data
0.5
0
3
200 400 600 800 1× 10
f , f_data
Frequency [Hz]
FRF PHASE
0
− 30
− 60
Phase [deg]
ϕ ( f ⋅ 2π)
− 90
Pha_data
− 120
− 150
− 180
3
200 400 600 800 1× 10
f , f_data
Frequency [Hz]
Figure 4-41: FRF Magnitude and Phase plots comparing the mathematical model (red)
to the test results (blue).
Comparing the plotted mathematical model to the measured FRF, it can

be concluded that the estimated modal parameters are representative of
the system and from here, the stiffness coefficient can be calculated as
( = = 2.8 × 10B C/
69
4.3.2 Extracting modal parameters by curve fitting
If due to the large acceleration imposed, exciting the natural frequency

of the system is not possible, then this problem can be worked around by
exciting the frequency band lower than the natural frequency and then
using the mathematical model to extrapolate the rest of the response.
Below in Figure 4-42 is given an FRF of a single damper testing with a

20g sweep sine, a rather high level of excitation, within the frequency
range 20-100 Hz.
1,9
1,8
1,7
1,6
1,5
1,4
1,3
1,2
1,1
0,9
0
-10
-20
-30
-40
-50
-60
-70
-80
-90
20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Frequency [Hz]
Figure 4-42: FRF from 20g sweep sine test of a damper in Config 3.
It is apparent from this plot that the damper does not undergo
resonance and the natural frequency of the system is surely higher than
100 Hz. Then a solution of the modal parameters must be reached fitting
the acquired data to the mathematical model given as
70
+ 2
=
− + 2
At 100 Hz, the experimental FRF has the following values:

| 2E100| = 1.9
arg3Hj2S1004 = −50°
Using these two conditions, the mathematical model can be solved for
the modal properties utilizing the MATHCAD script (see Appendix A).
The script returns the following parameters as the result of the
mathematical model:

= = 105
2E
= 0.32
( = = 1.5 × 10F C/
Then the frequency response of the mathematical model can be plotted

together with the experimental FRF in order to compare the validity of
such an approach.
71
FRF MAGNITUDE
2
1.8
1.6
Magnitude
G( f ⋅ 2π )
1.4
Mag_data
1.2
0.8
50 100 150
f , f_data
Frequency [Hz]
FRF PHASE
0
− 30
− 60
Phase [deg]
φ ( f ⋅ 2π )
− 90
Pha_data
− 120
− 150
− 180
50 100 150
f , f_data
Frequency [Hz]
Figure 4-43: Comparison of the mathematical frequency response function (red) and
the experimental FRF (blue).
Investigating these two plots it can be seen that there is an almost

perfect representation of the FRF phase by the mathematical model
while it is acts poorly for the FRF Magnitude. Using just a single data
point of the experimental FRF for the construction of the whole curve is
therefore inefficient. To overcome this problem an adjustment of the
72
modal parameters can be performed to achieve a better curve fit.
Starting from the previously calculated set of modal parameters and
gradually adjusting them, the following fit can be obtained
FRF MAGNITUDE
2
1.8
1.6
Magnitude
G ( f ⋅ 2π )
1.4
Mag_data
1.2
0.8
50 100 150
f , f_data
Frequency [Hz]
FRF PHASE
0
− 30
− 60
Phase [deg]
φ ( f ⋅ 2π )
− 90
Pha_data
− 120
− 150
− 180
50 100 150
f , f_data
Frequency [Hz]
Figure 4-44: The mathematical FRF fit with the adjusted modal parameters
with the following parameters

= 115
73
= 0.35
and the stiffness coefficient can be estimated as
( = = 1.8 × 10F C/
After the suggested adjustment of the parameters, it is seen that the

mathematically estimated system represents the experimental results
more or less accurately in the excited region and moreover it gives a
rough estimation of the region beyond it. The maximum deviation in the
magnitude plot between the model and the experimental data points is
calculated as 0.2 and for the phase plot this value is 11°, which are
acceptable results.
4.4 Estimating torsional properties from test data
Besides the translational properties affecting the dampers, there is also

a rotational motion in particular at the tip of the damper that somewhat
originates from the very small deformations in the axial and the radial
directions (see Figure 4-45).
74
Figure 4-45: The rotational motion at the tip of the damper.
There is a rotational stiffness coefficient associated with this motion in

units of [Nm/rad] and it must be numerically defined to achieve a
complete model of the system together with the axial and the radial
stiffnesses.
The presence of a torsional deformation at the tip of the damper was

noted in most of the single degree of freedom test results. When
performing various tests on the single damper in the axial loading
configuration, a constantly appearing phenomenon was observed in most
of the FRFs in which there has been a short period of phase shift
between the accelerometers oriented in the same direction with a certain
offset (see Figure 4-46, Figure 4-47).
75
Figure 4-46: Accelerometer configuration for an axial low level sweep sine test.
3,5
2,5
1,5
0,5
0
150
125
100
75
50
25
0
-25
-50
-75
-100
-125
-150
100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400
Frequency [Hz]
Figure 4-47: Low level sweep sine FRF in the region of the rotational resonance of the
damper.
The spontaneous opposite phase shift of the axial accelerometers

suggests a resonance due to a rotation at the damper tip, a deformation
associated with the torsional stiffness parameter.
76
The torsional stiffness could be accurately determined if the eccentricity
of the mass was known however, investigating the response to a 0.5g
constant level loading FRF given in Figure 4-47, the torsional resonance
is observed around 250 Hz. This value will be tremendously lower when
the testing is performed at the actual test loading (20g), suggesting a
very low torsional stiffness. The parameter is in practice of no interest
for the modeling of the damper so, no additional effort was made to
determine it.
4.5 Estimated properties of the final damper

configuration
Based on the experiences gained from the numerous single damper

experiments, a final tuning and configuration of the damper was
achieved, which was described in detail in the previous chapter. In order
to model the behavior of the dummy mockup, the complete spring-like
properties of the dampers have to be estimated, in the same manner
discussed throughout this chapter.
4.5.1 Axial stiffness coefficient, k1
The axial stiffness coefficient is the property associated to the resistance

of the damper to the deformations purely along its axis. The FRF plot of
the axial sweep sine test at 20g suggested a natural frequency as
= 300
Then the axial stiffness can be estimated as
( = 2E ∙ = 1.24 × 10B C/
77
4.5.2 Radial stiffness coefficients, k2 & k3
The radial stiffness coefficient in the two transversal directions k2 & k3

correspond to translational deformations that are perpendicular to the
damper’s axis. The FRF plot of the transversal sweep sine test at 20g
evidenced a natural frequency associated to the radial deformation mode
as
= 350
Then the radial stiffness coefficient becomes
(,H = 2E ∙ = 1.70 × 10B C/
4.5.3 Rotational stiffness coefficient, kθ,1
The dampers are free to rotate around their own axis, therefore there is
no such stiffness regarding this motion. It is taken as 0, whenever it is
deemed necessary.
4.5.4 Rotational stiffness coefficients kθ,2 & kθ,3
For the purposes of modeling of the system, the torsional stiffness of the
dampers are taken as
(I,,H ≅ 250 C/JKL
This is an estimated value that is tuned and verified through the

computational model and further mockup tests that will be described in
detail in the following chapters.
78
4.6 Results & Discussion
The dynamic material properties of the dampers had to be described

numerically in order to be used in an analytical or a Finite Elements
model. With the help of the experimental modal analysis techniques, the
modal parameters are extracted from the test data in terms of natural
frequencies, damping ratios and mode shapes. From these modal
parameters, the necessary stiffness coefficients are estimated using a
suitable analytical model and mathematical formulas. The results of this
procedure is summarized in Table 4-8 below
Table 4-8: Summary of the stiffness coefficients estimated for the final damper
configuration.
Axial Stiffness, ( 1.24 x 106 N/m

Radial Stiffness, ( 1.70 x 106 N/m
Radial Stiffness, (H 1.70 x 106 N/m
At the end of all the single degree-of-freedom tests and the subsequent
calculations, some very useful results are obtained regarding the
characterization of the dampers. These results provide valuable insight
and are crucial for the construction of the Finite Elements model which
will be described in detail in the following chapter.
79
Chapter 5: Finite Element Analysis of
the Damper – Mockup system
5.1 The need for a Finite Elements Analysis
Even though experimental methods are accurate and reliable for the
study of the dynamics of a system, they also come with certain
restrictions that foster the use of computational methods along with
them. As far as the shaker tests of this work are concerned, a
comparison of the advantages and the disadvantages of the
Experimental methods to the computational ones are given in the table
below.
Table 5-9: Comparison of the advantages and the disadvantages involved in
experimental methods of a shaker test to those of Finite Elements methods.
Experimental Methods Finite Element Methods
Accuracy depends on the complexity and

Fast and accurate results
the validity of the model
Requires a data processing stage before Results are immediately clear and
meaningful results can be extracted visually appealing
Long setup change time, slow Once the FE model is created, allows
modification of test parameters quick changing of parameters
Repeated testing may cause the There is no limit to the number of runs
damaging of the setup resulting in time that can be performed. No risk of
and economic losses damage.
Such a comparison obviously exhibits the necessity of a Finite Elements

Model to be used together with the experiments, exchanging useful
information between them whenever possible.
80
5.2 Verification of the structural modes of the mockup
using FEM
Before including the dampers in the Finite Elements model, the mockup
structure should be checked in terms of its natural frequencies. This is
to assure that the deformation modes of the mockup are much higher
than the modes of the system when the three dampers are included. In
this way, it can be stated that the mockup is behaving in the region of
interest as an ideal rigid body.
5.2.1 Finite Element Modeling
In Chapter 2, the mockup to replace the actual MIMA for testing

purposes was modeled and assembled in CATIA. This CAD model (see
Figure 5-48) is identical to the mockup and therefore can be utilized for
the Finite Element Analysis using the Generative Structural Analysis
module of CATIA V5.
Figure 5-48: The CAD mockup assembled in CATIA.
81
The first step, once the CAD model is imported into the Generative
Structural Analysis, is to define the connections of the various
components making up the mockup assembly. This is done using
General Analysis Connections in CATIA. Once all the necessary
connections are defined, the model looks as depicted in the following
figure:
Figure 5-49: Mockup CAD model with the necessary connections defined.
The next step is to define the nature of the connection properties. This
function allows the user to define how the different components of the
assembly are fixed together. As far as the mockup is concerned, the
connections are Fastened Connections.
82
Figure 5-50: The model with fastened connection properties defined between the
different components.
Once all the connections are defined between the components making up
the mockup assembly, the model is ready to be meshed and computed.
5.2.2 Meshing
There are several factors that require special attention for the meshing
of the model. The first is the type of Finite Element to be used for the
meshing. Here, the simplest element, the Linear Tetrahedron is used
(see Figure 5-51).
Figure 5-51: The linear tetrahedron. Its geometry is fully defined by giving the location
of the four corner nodes with respect to the global coordinate system. [13]
83
Another important property of the mesh is the element size. This
parameter sets the overall size of a single finite element and therefore
directly affects the number of elements and the number of nodes in a
mesh. Increasing the number of nodes in a mesh may result in more
accurate results while on the other hand increasing the computation
time significantly. Below in Figure 5-52 are demonstrated two the
meshes, a course mesh with large element size and a fine one with
smaller elements.
Figure 5-52: A course mesh compared to a fine mesh.
Once the meshing procedure is completed, the governing equations of

the FE model can be computed to achieve the desired results.
5.2.3 Results and discussion
As previously discussed, the meshing done on the model can directly

influence the results. Therefore it is a common approach to perform
several computations gradually increasing the number of finite
elements, i.e. using a finer mesh, until a single solution is converged.
The decision on when to stop can be made by comparing the results of
each iteration and checking the relative error between them. When the
84
percentage difference falls below a certain criteria, the mesh is said to be
good enough and the results are assumed accurate.
The when performing the FEM analysis of the structural modes of the
mockup, this same method is used. To start with, large finite elements
are used (size 9mm) resulting in a course mesh. Then the mesh is made
finer and finer until convergence to a single result occurs. The results of
this procedure are summarized in Table 5-2 below.
Table 5-10: Summary of the results of the FEM analysis for the structural modes of the
mockup.
Element Size Frequency Computation

Mode Nodes Elements Modes
type [mm] [Hz] Time [s]
Linear 1 805
1 9 2555 8238 7
Tetrahedron 2 1205
Linear 1 751
2 7 3158 10262 10
Tetrahedron 2 1112
Linear 1 689
3 5 5291 17806 14
Tetrahedron 2 1007
Linear 1 601
4 3 20527 82636 53
Tetrahedron 2 875
Linear 1 570
5 2 44446 178817 149
Tetrahedron 2 832
Linear 1 561
6 1 203434 909090 1323
Tetrahedron 2 804
As it can be seen from the above table, the computation time increases
significantly as the element size is decreased beyond a certain value.
Therefore a mesh that is acceptable both in terms of accuracy and
computation time must be selected. Mesh 5 is such a mesh with only a
1.5% of relative error in terms of the change in the first natural
frequency compared to Mesh 6. Therefore it is said to be an optimum
mesh with accurate results.
The computed structural modes of vibration of the mockup are given in

the figure below.
85
Figure 5-53:: Structural modes of vibration of the mockup. (a) Undeformed mesh (b)
first mode at 570 Hz (b) second mode at 830 Hz.
The FEM analysis of the mockup suggest that the first structural
resonance of the mockup is quite high (above 550 Hz) with respect to the
frequency band of interest for the actual mockup
mockup testing. The vibration
modes related to the deformation of the dampers are expected to be
around 100 – 200 Hz and so the structural deformations of the mockup
should not interfere with these modes. Therefore at the end of the FEM
analysis, the conclusion
conclusion that the mockup is satisfactory for experimental
purposes can be reached.
86
5.3 Analysis of the Damper-Mockup system using linear
spring elements
The actual MIMA instrument is mounted on three dampers that are

fixed at designated locations. The mockup that was designed inertially
identical to the instrument requires a similar assembly for
experimentation and such a test setup can be modeled as a Finite
Elements model before any real experimentation is performed.
5.3.1 Modeling of the dampers as spring elements
Since the mockup itself was already modeled in the previous section,
here an appropriate representation of the dampers in the model must be
defined. Considering that only the natural properties of the system is of
interest, the dampers can be modeled as springs connected at the
designated locations on the mockup. Rigid Spring Virtual Part feature of
CATIA is used to model three springs with given stiffness coefficients as
shown in Figure 5-54.
Figure 5-54: Spring Virtual Parts connected at three point on the mockup.
87
As far as the boundary conditions are concerned, this time it is the
dampers that must be restraint. Therefore all the three Virtual Spring
elements are clamped (see Figure 5-55).
Figure 5-55: Clamped boundary condition for the spring elements.
5.3.2 Defining Virtual Spring stiffness coefficients
The Virtual Spring Parts have a set of directional stiffness coefficients

that must be defined before any computation can be made. Here, in
order to analyze the behavior of the system with the damper at their
ultimate configuration and tuning, the estimated stiffness parameters of
Chapter 4 are used. These stiffness properties are defined for all the
three dampers as shown in the below figure.
88
Figure 5-56: Stiffness coefficients estimated from single degree of freedom tests defined
for the virtual spring parts.
Once all three dampers are defined in this manner, the model is defined
completely and using the optimum mesh parameters of the previous
section, computations can be performed.
5.3.3 Resulting behavior of the system
There are several important results that can be obtained from this
analysis including the natural frequencies and the modes of vibration of
the system.
5.3.3.1 Natural frequencies

The computed first three natural frequencies of the system are reported
as follows
89
Table 5-11: Natural frequencies of the system corresponding to the first three modes of
vibration.
Mode Natural frequency [Hz]

1) 131
2) 151
3) 264
Here only the first three natural frequencies are given since the
remaining ones are of a much higher frequency and correspond to trivial
modes not significant for the scope of this analysis.
5.3.3.2 Mode shapes

Besides the natural frequencies of the model, the mode shapes at these
frequencies require special attention. A visualization of the first three
modes of the whole system is given in the next page.
90
(a) (b)
(c)
Figure 5-57:: First three modes of vibration at (a) 131 Hz (b) 151 Hz (c) 264 Hz.
A close investigation of Figure 5-57 reveals some important information

about the nature of these modes. The first two modes at low frequency
exhibit a rotational mot
motion
ion of the mockup while the third higher
frequency mode seems to be almost purely translational.
The rotational mode at low frequencies is primarily due to the out

out-of-
phase axial deformation of the spring elements. This phenomenon
results in a rotation ax
axis
is on the vertical mockup plate where there is
almost no displacement. This axes around which the mockup rotates can
be better visualized from the Translational Displacement Vector plot.
91
Figure 5-58:: Displacement Vector plot of the vibration mode 1 at 131 Hz.
Figure 5-59:: Displacement Vector plot of the vibration mode 2 at 151 Hz.
92
In these plots, there is a clear region (in shades of blue) where there is
almost no displacement while the areas on the opposite sides of this line
display large displacements. This fact proves the existence of a
rotational axis passing through this region. These rotational axes for the
first two modes are schematically illustrated below in Figure 5-60.
Figure 5-60: The axes of rotation for the first two modes of vibration.
Being this the case for the first two modes, the high frequency third
mode results from the in-phase deformation of the spring elements
producing an almost purely translational motion (see Figure 5-61).
93
Figure 5-61: Displacement Vector plot of the vibration mode 3 at 264 Hz.
5.3.4 Evaluation of the results for the ultimate damper

configuration
The FEM analysis of the system with the damper properties estimated
previously was performed and the returned results are positive in terms
of several aspects. To start with, there are no natural frequencies of the
damper-mockup assembly below 100 Hz. Furthermore, the first three
modes that are of interest do not occur at very high frequencies. These
are favorable results as far as the environmental conditions of the
MIMA are concerned.
As predicted, the first modes of vibration are rotations around a certain

axis. This expectation can be supported by comparing the COG position
of the mockup to the center of the Damper Moment Triangle, as shown
in Figure 5-62. The COG position is shifted due to the presence of the
perpendicular plate. In fact when superimposed, the rotational axes of
the first two modes intersect close to this center of moments, which is a
result of these two mentioned points not being coincident.
94
Figure 5-62: Mockup COG position vs. Damper Triangle Moment Center.
Unlike the first two, a third mode is observed at a higher frequency

which is almost purely translational in shape.
Considering all these results of the FEM analysis, the damper tuning
performed through the single DOF tests leads to an acceptable dynamic
behavior on the dummy mockup.
In this chapter, Finite Element Modeling techniques were used since

having a computational model alongside an experimental study has
numerous advantages, including the ease of parametric modifications,
visually appealing results and the elimination of failure risks.
As a first task, the mockup that was designed and manufactured for
testing purposes was verified in terms of its structural vibration modes.
Since the resonant deformations of the mockup did not occur until above
95
550 Hz, a frequency which is representative of the MIMA behavior if
fully constrained.
Furthermore the complete test setup with the damper-mockup assembly

was modeled and analyzed with the estimated stiffness properties of the
ultimate damper configuration. The resulting natural frequencies and
mode shapes were investigated and found to be acceptable when
compared to the environmental requirements.
Besides providing foresight for the mockup testing procedures, the

Finite Elements Model also has the purpose of giving tuning feedback
for the dampers. If a certain behavior of the FE model is stated, then the
stiffness parameters that produce this behavior is attempted to be
achieved using the previously described damper tuning methods. This is
a valuable tool for the mode optimization of the system and is utilized
many times during this study.
96
Chapter 6: Testing of the dummy
model
The MIMA instrument will be mounted through three dampers to the

rover. The dampers have the task of isolating and reducing the
transmitted vibrations arising during the launch and landing phases. In
this part of the study, a suitable shaker test setup and procedures were
designed to simulate the vibration environments expected at the
instrument-rover interface. The dummy mockup was designed and
described in chapter 2; it is used to represent the actual MIMA in these
tests to avoid any risk of damaging due to repeated testing. The
previously tuned and evaluated dampers are used to constraint the
mockup and reduce the transmitted vibrations.
6.1 Test methodology
Considering the environmental conditions of the MIMA instrument

given in Chapter 1, appropriate testing procedures are designed and
applied for the damper-mockup assembly in order to analyze its
behavior.
6.1.1 Test setup
Having studied the environmental conditions, it is apparent that all

three axes of the mockup need to be tested separately. Therefore three
different test setups are utilized to evaluate the three different
directional behaviors of the system as notated below in Figure 6-63.
97
Damper 2 Damper 3
Damper 1
Figure 6-63: The three testing directions shown on the mockup and damper
designations.
All the tests use a similar signal acquisition and processing layout
schematically described in Figure 6-64.
Figure 6-64: Signal acquisition and feed schematics for the dummy model testing.
98
6.1.1.1 Z axis
The Z axis testing primarily includes the axial excitation of the dampers
supporting the mockup. A typical test setup used for this axis is given
below in Figure 6-65.
Figure 6-65: Z axis testing setup with the accelerometer positions.
As seen in the above figure, several accelerometers are utilized for the
complete measurement of the stimulus and the response. Similar to the
single damper-mass test setup, one accelerometer is used to measure the
reference input and feed it back to the control PC and one accelerometer
is used to measure this same reference for storage. At least three
accelerometers are mounted immediately above each damper to measure
the response of the mockup, However it is reasonable to include more
accelerometers facing other directions to check for some transversal
vibrations of the mockup.
99
6.1.1.2 X axis
In X axis testing, the dampers are mainly excited in the transversal
direction, therefore the damper-mockup assembly must be rotated with
respect to the excitation direction of the shaker. A large metallic cube is
used for this purpose as shown in Figure 6-66.
Figure 6-66: X axis testing setup with the accelerometer positions.
Similar to the Z axis testing, two accelerometers are used to measure the
reference excitation, one for control and one for storage and further
analysis. Three accelerometers facing the horizontal are used to measure
the axial deformation responses of the dampers. Others measure the
accelerations of the mockup in the direction of excitation and also the
out-of-plane vibrations.
6.1.1.3 Y axis
The Y axis mockup testing is very similar to the X axis testing with the
only difference being a 90° rotation of the mockup. The dampers are
again stressed in the transversal direction (see Figure 6-67).
100
Figure 6-67: Y axis testing setup with accelerometer positions.
The configuration of the accelerometers is more or less the same as the

previously described test setup with enough attention paid to having
enough of them for a complete understanding of the system behavior.
6.1.2 Reference stimulus
For the testing of the mockup, some well-defined reference excitations

are required. These given excitation profiles are generated as signals
using the Control PC and are realized by the shaker as vibrations.
6.1.2.1 Resonance search

The resonance search is a low amplitude sweep sine in a wide frequency
band. The resonance search signal profile used for the mockup testing is
given below.
101
Table 6-12: Resonance search profile properties.
Frequency [Hz] Amplitude

20 0.5g
2000 0.5g
0,5
Amplitude [g]
0
20 200 2000
Frequency [Hz]
Figure 6-68: Resonance search reference input profile plot.
The low amplitude level of such an excitation allows the wide frequency
range testing of the mockup, even passing through its resonances.
However due to the high non-linearity of the dampers, the results of this
test is not significant for the dynamic parameter estimation. The main
purpose of such an excitation is to perform it before and after each
strong level test and compare the results of these two tests to state
whether the system was altered by the test performed in between.
6.1.2.2 Sweep sine

The environmental conditions of MIMA foresee a low frequency sine
environment with high acceleration amplitudes. Therefore an
appropriate reference excitation for the mockup testing must be defined
in order to test and characterize the system concerning such an
environment. The reference input used to simulate the sine environment
is given below.
102
Table 6-13: Sweep sine excitation profile.
Frequency [Hz] Amplitude [g]

5 1
20 20
100 20
25
20
Amplitude [g]
15
10
0
5 50
Frequency [Hz]
Figure 6-69: Sweep sine reference input profile plot.
Recalling the environmental conditions of the MIMA, the sweep

amplitude used here is different than the one reported. Here the
amplitude is lower since the aim of the study is to characterize and have
an understanding of the damper system response rather than to test the
endurance of the system. With this approach, failure risks due to
repeated testing are reduced significantly.
6.1.2.3 Random
The random vibrations of the MIMA environmental conditions are
simulated for the mockup shaker testing as follows
103
Table 6-14: Random reference excitation
Frequency PSD [g2/Hz]

[Hz] Z axis X,Y axes
20 0,0174 0,007
100 0,43 0,172
400 0,43 0,172
2000 0,0174 0,007
0,5
0,45
0,4
0,35
PSD [g2/Hz]
0,3
0,25
0,2 Z axis
0,15 X,Y axes
0,1
0,05
0
20 200 2000
Frequency [Hz]
Figure 6-70: Random reference input profile plot.
For the random test, the time of excitation can be more freely selected. A
too short excitation time may result in poor data because the number of
averages does not allow to well estimate the parameters means while
prolonging the excitation increases the risk of damaging. For the
purposes of characterization of the dampers supporting the mockup, an
excitation time of 45 seconds has been chosen as a compromise.
104
6.2 Mockup testing and results
This procedure aims to test with the shaker the behavior of the three
tuned dampers supporting the dummy mockup. The responses to the
previously defined excitations in all three axes will be measured using
systematically placed accelerometers. These measurement channels
then will be used for the evaluation of various performance criteria.
All the testing directions utilize the following excitation procedure:
1. Resonance search (0.5g)

2. Sweep sine
4. Random
So the basic idea is to perform the strong tests followed and preceded by
the resonance search tests. The resonance search provides some useful
insight about the behavior of the system and also it is used to prove that
the system at the start of the testing procedure has not been
dramatically altered at the end.
6.2.1 Z axis testing
The test setup shown in Figure 6-65 is excited using the given excitation
profiles and the following results are obtained.
105
6.2.1.1 Z axis sweep sine
The time histories of the acquired accelerations are given below.
Figure 6-71: Z axis sweep sine test time history.
A better visualization can be achieved using FRFs between the reference

acceleration and each response channel
106
1,12
FRF Modulus
Damper 1 m s-2 Mod 1,1
Damper 2 m s-2 Mod
1,06
1,04
1,02
0,98
0,96
0,94
0,92
5
FRF Phase
Damper 1 m s-2 Ph 2,5
Damper 2 m s-2 Ph 0
Damper 3 m s-2 Ph -2,5
-5
-7,5
-10
-12,5
-15
-17,5
-20
20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Frequency [Hz]
Figure 6-72: Z axis FRFs between the reference acceleration and the accelerations just
at the tip of each damper.
For this testing axis, the remaining out-of-plane vibrations (channels X

down left, X down right and Y) are very small and are not demonstrated
here (see Appendix B for these plots).
6.2.1.2 Z axis random

A random excitation 45 seconds of random vibration with the previously
defined profile is given to the system for a time of 45 s. The acquired
time history of data is given below in Figure 6-73.
107
Figure 6-73: Z axis random test time history.
As far as the random environment is concerned, having a good transfer

function is critical. Therefore the FRFs are also calculated.
1,6
FRF Modulus
Damper 2 m s-2 Mod
0,8
0,6
0,4
0,2
0
FRF Phase
Damper 1 m s-2 Ph -20
Damper 2 m s-2 Ph -40
Damper 3 m s-2 Ph
-60
-80
-100
-120
-140
-160
1
Coherence
Damper 1 m s-2 Coh 0,8
Damper 3 m s-2 Coh
0,4
0,2
0
20 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
Frequency [Hz]
Figure 6-74: Z axis random FRF plot.
108
6.2.2 X axis testing
The dampers for this testing direction are excited transversally using
the setup shown in Figure 6-66; with the given excitation profiles. In
this test, the X accelerometer measures the response in the same
direction as the excitation while the damper accelerometers measure the
horizontal acceleration at the tip of each damper.
6.2.2.1 X axis sweep sine

The 20g sweep sine excitation is given to the system within the defined
frequency band and the resulting time history data is given in Figure 6-
75.
Figure 6-75: X axis sweep sine test time history.
This data can also be evaluated in terms of FRFs between the reference
input and each response channel
109
1,3
FRF Modulus
Damper 3 m s-2 Mod
1
X m s-2 Mod
0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
180
FRF Phase
150
Damper 1 m s-2 Ph
Damper 2 m s-2 Ph 100
Damper 3 m s-2 Ph
X m s-2 Ph 50
-50
-100
-150
-180
20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Frequency [Hz]
Figure 6-76: X axis sweep sine FRF plot.
6.2.2.2 X axis random

The time history data for a 45 seconds random vibration test for the X
axis is given below in Figure 6-77.
110
Figure 6-77: X axis random test time history.
For the evaluation of the transfer function, the FRF plot is also given
1,2
FRF Modulus
Damper 1 m s-2 Mod
1
Damper 2 m s-2 Mod
X m s-2 Mod
0,6
0,4
0,2
0
200
FRF Phase
Damper 3 m s-2 Ph
50
X m s-2 Ph
0
-50
-100
-150
-200
1
Coherence
Damper 3 m s-2 Coh
0,4
X m s-2 Coh
0,2
0
20 50 100 150 200 250 300 350 400 450 500 550 600 650 700
Frequency [Hz]
Figure 6-78: X axis random FRF plot.
111
6.2.3 Y axis testing
Recalling the setup of Figure 6-67, the excitations along the Y axis are
acting transversally on the dampers, similar to the previous X axis tests.
Therefore a number of acceleration channels measuring the response
both in the direction of excitation (Y) and in the axial direction of the
dampers (Damper 1, Damper 2 and Damper 3) arenecessary.
6.2.3.1 Y axis sweep sine

The acquired time history data for the 20g sweep sine excitation in the Y
axis is shown below.
Figure 6-79: Y axis sweep sine test time history.
For a more detailed visualization, the acquired data is given in terms of

FRFs in Figure 6-80.
112
1,2
FRF Modulus
Damper 2 m s-2 Mod
1
Damper 3 m s-2 Mod
Y m s-2 Mod 0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
180
FRF Phase
150
Damper 1 m s-2 Ph
Damper 3 m s-2 Ph
Y m s-2 Ph 50
-50
-100
-150
-180
20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Frequency [Hz]
Figure 6-80: Y axis sweep sine FRF plot.
6.2.3.2 Y axis random

The time history data for a 45 seconds random vibration test for the Y
axis is given below.
113
Figure 6-81: Y axis random test time history.
As done previously, the FRF between the stimulus and the response is
calculated in order to assess the transfer function of the system.
1,4
FRF Modulus
Damper 2 m s-2 Mod
Damper 3 m s-2 Mod 1
Y m s-2 Mod 0,8
0,6
0,4
0,2
0
180
FRF Phase
150
Damper 1 m s-2 Ph
Damper 3 m s-2 Ph
50
Y m s-2 Ph
0
-50
-100
-150
-180
1
Coherence
Damper 3 m s-2 Coh
0,4
Y m s-2 Coh
0,2
0
20 50 100 150 200 250 300 350 400 450 500 550 600 650 700
Frequency [Hz]
Figure 6-82: Y axis random FRF plot.
114
6.3 Evaluation of the damping system characteristics
from the test results
With the results of the mockup testing acquired, an evaluation of the

damping system can be performed in each testing direction. This
evaluation will be performed in terms of the measured modal
parameters of the system, i.e. natural frequencies, damping and mode
shapes and the maximum amplifications present both for the sine and
the random environment.
6.3.1 Performance in Z axis
Recalling the sweep sine response FRF of the system given in Figure 6-
72, the first observation that can be made is the fact that there are no
large amplifications in the response (maximum amplification equal to
1.05 i.e. 5%). This is a positive result as far as the sine environment is
concerned. In fact, the dampers excited in this direction seem to be
acting quite stiff. To support this point, the phase diagrams of the three
accelerometers can be examined. The phase diagram of the signals from
the three accelerometers at the top of the dampers shows almost no
change. This suggests that the resonance is actually quite far away from
this region.
In the random environment, the system is expected to have a good

transfer function with a certain cutoff frequency, above which there are
no further large amplifications of the response.
Then from the calculated FRF of the random test, the cutoff frequency
can be derived as
> = 300
with an amplification of
| > | = 1.4
115
This given maximum amplification of the random transfer function is a
positive result since it is not so high.
6.3.2 Performance in X axis
The time history and the Frequency Response plots (Figure 6-75 &
Figure 6-76) for the sweep sine testing in the X axis do not exhibit high
amplifications of the response (maximum amplification 1.2) and there
are no significant resonances in this region of interest. These are
favorable results regarding the sine environment.
In order to evaluate the transfer function regarding the X axis random

behavior, the FRF between the stimulus and the response in the same
direction as the excitation is given below in Figure 6-83.
1,15
1,1
1,05
1
0,95
0,9
0,85
0,8
0,75
0,7
0,65
0,6
0,55
0,5
0,45
20 50 100 150 200 250 300 350 400 450 500 550 600 650 700
Frequency [Hz]
Figure 6-83: FRF Magnitude plot between the reference and the response in the same
direction as the excitation.
From this plot of the transfer function, the following cutoff frequency
and amplification can be defined as
> = 130
| > | = 1.15
116
The random transfer function of the system in the X axis is not as good
as the transfer function in the Z direction. Even though the presence of a
cutoff frequency is obvious, the amplifications beyond this value are not
reduced significantly. This is an unfavorable result in terms of the
damping performance. However, overall the amplifications in the FRF
are small, which should compensate this unfavorable condition.
6.3.3 Performance in Y axis
Similar to the X axis test results, investigating the FRF plot given in
Figure 6-80, it can be observed that the system does not have large
amplifications (maximum amplification is 1.2) and there is no significant
presence of a resonance within the sine frequency region. These are all
positive results concerning the sine environment.
Studying the transfer function shown in Figure 6-84, regarding the

random test, it can be stated that a slightly better performance is
observed compared to the X axis testing.
1,25
1,2
1,15
1,1
1,05
1
0,95
0,9
0,85
0,8
0,75
0,7
0,65
0,6
0,55
20 50 100 150 200 250 300 350 400 450 500 550 600
Frequency [Hz]
Figure 6-84: FRF Magnitude plot for the Y axis testing between the reference and the
response in the same direction as the excitation.
From this plot, the cutoff values are estimated as
> = 150
117
with,
| > | = 1.2
Beyond this point, the amplifications are significantly reduced.
Considering the mechanical environment conditions of the actual MIMA

instrument, appropriate test setup and procedures were applied to test
and evaluate the behavior of the previously designed and manufactured
mockup mounted on the tuned dampers in the ultimate configuration.
Due to the directional behavior of the system, each of the three axes was
tested with the sine and the random profiles defined.
The results of the mockup testing provided useful information for the
performance assessment of the tuned damping system. For all the
testing directions, the sweep sine test did not reveal large amplifications
of the response (maximum amplification of 20% in X and Y axes) and
there were no critical resonances within the frequency range of interest.
A wider frequency band could be tested using the random excitation and
the corresponding cutoff frequencies above which the amplifications are
dampened down were found. Overall, the transfer function for the
random testing was satisfactory. All these results of the mockup testing
suggest that the performance tuning performed on the dampers was on
the whole satisfactory.
Considering all the testing directions and excitations, a number of

critical resonances of the system were found. The first one occurred
during the Z axis testing at around 300 Hz with the in-phase axial
deformation of the three dampers. The second one was during X axis
transversal testing at 130 Hz with the dampers deforming out-of-phase
causing a rotational motion of the mockup. The third and the final one
118
was present during Y axis testing at 150 Hz again with the out-of-phase
deformation of the damper resulting in a rotational motion.
Besides assessing the performance of the tuned damper system with

three dampers in their ultimate configuration, the mockup testing also
has the objective of verifying the FE model and analysis performed in
Chapter 5. This can be done by comparing the results of the
experimental procedure to those of the FEM analyses.
Table 6-15: Experimental modes compared to those of the FEM analysis.
Experimental FEM
Mode Frequency [Hz] Mode Frequency [Hz] Mode
1 130 out-of-phase 131 out-of-phase
2 150 out-of-phase 151 out-of-phase
3 300 in-phase 265 in-phase
The above table clearly shows that the FEM model is accurate in
estimating the natural frequencies of the system, except for a slight
deviation in the third mode which is 13% lower than the measured one.
Besides the frequency values of the modes, examining the phase plots
from the tests show that the mode shapes are also similar to those of the
FEM analysis results. Therefore it can be stated that the created Finite
Elements Model is correct and accurate.
Having verified the computational model simulating a test setup, one

achieves a powerful tool that can be used for optimization purposes. In
the following chapter, such a study will be performed on improving the
system’s response using the validated FEM model.
119
Chapter 7: Finite Element Methods for
the response improvement of the
dampers
7.1 Introduction
In Chapter 5, a Finite Element model of the damper-mockup assembly

test setup was designed. Using this model, the natural frequencies and
the modes of vibration could be computed and visualized. Then in
Chapter 6, comparing the data from the mockup testing on the shaker to
the results of the FEM analysis, the accuracy of the model was verified.
Having an accurate and tuned computational model simulating the

characteristics of the real application, acts as a strong tool for the
performance improvement. Parameters of the computational model can
easily be modified, unlike the experimental setup, and the resulting
behavior of the system can be obtained. This procedure performed
systematically in a trial and error fashion allows for the optimization of
the response of the system in a fast and efficient way.
The damper name designation that will be used throughout this chapter
is shown in Figure 7-85 below.
120
Figure 7-85: Damper name designation given.
7.2 Objectives of response improvement
Before actually changing any parameters of the FE model, what is

meant by the possible improvements on the system response should be
clarified. Recalling the natural frequencies of the damper-mockup
assembly computed using the FE model
Table 7-16: Computed first three natural frequencies of the FE model
Mode Natural frequency [Hz]

1) 131
2) 151
3) 265
From Table 7-16, it can be seen that the third mode occurs at a higher
frequency than the other modes of vibration. Remembering the
environmental conditions of the MIMA, it is required to have a system
without any resonances below 100 Hz (for sine environment concerns)
121
and instead have them only slightly above this value (random
environment concerns). Therefore reducing the third natural frequency
can be stated as an improvement of the response. In fact having all these
three critical modes at around the same frequency around 150 Hz is the
ideal case and will be the main objective of the Finite Elements response
improvement procedure.
There are mainly two approaches for the response improvement using
the FE model depending on the parameters that will be modified to
obtain an optimum. The first approach is the experimentally feasible
one. Due to the limitation of the damper design, in this approach an
improvement of the response will be attempted by only changing the
axial stiffness of the three dampers. The second approach is more
theoretical in which all the properties, be it axial or radial, of the
dampers will be modified to obtain an ideal response. Of course the
resulting optimum dampers cannot achieved with this particular design,
so this approach will only provide insight for the future damper designs.
7.3 Response improvement by modifying the axial

spring properties of the dampers
As it was investigated in the single degree-of-freedom tests of Chapter 2,

the dampers’ axial stiffnesses can be changed by increasing or
decreasing the tightening preload on the dampers. Starting from this
point, the first idea is to use different axial stiffness coefficients for each
of the three dampers to obtain an improved response of the system. In
this method, only the axial stiffnesses will be varied while the other
radial and rotational parameters will be left untouched.
7.3.1 Initial damper tuning
In Table 7-17 is recalled the estimated spring properties of the dampers

in their ultimate configuration.
122
Table 7-17: Stiffness values estimated and used in the FE model
Axial Stiffness, ( 1.24 x 106 N/m

Radial Stiffness, ( 1.70 x 106 N/m
Radial Stiffness, (H 1.70 x 106 N/m
Rotational Stiffness, (I, 0 Nm/rad
Rotational Stiffness, (I, 250 Nm/rad
Rotational Stiffness, (I,H 250 Nm/rad
The FE model with these parameters for all the three dampers resulted
stiffness, ( , the third mode which is at a higher frequency with respect

in the modes given in Table 7-16. Now by only changing the axial
to the others will be decreased with the primary objective of having the
first three modes of vibration occurring around 150 Hz.
7.3.2 Effects of individual dampers on the overall system

response
In order to systematically analyze the outcome of modifying the

dampers’ stiffness parameters, initially the effect of having only one
damper different with respect to the initial configuration is studied.
Since the purpose is to decrease the occurring frequency of the third
phase axial deformations of the dampers, the ( value for each damper is
mode, which is almost a pure translation of the mockup due to the in-
decreased one at a time with the results summarized below in Table 7-

18.
123
Table 7-18: Results of the various FE analysis runs with different damper properties.
k1 [N/m] Natural Frequency [Hz]

Iteration Damper 1 Damper 2 Damper 3 Mode 1 Mode 2 Mode 3
Initially 1,2E+06 1,2E+06 1,2E+06 130 150 264
1 5,0E+05 1,2E+06 1,2E+06 107 149 237
2 1,2E+06 5,0E+05 1,2E+06 103 143 243
3 1,2E+06 1,2E+06 5,0E+05 112 133 241
There are several conclusions that can be drawn from the above results.
The first observation is the fact that when the stiffness of any single
damper is decreased, all the resulting modes are reduced in terms of
their occurring frequencies. This poses some problems for the reducing of
a natural frequency with respect to the others.
Another observation would be the individual effects of the dampers on

the resulting modes. From Table 7-3, it can be seen that Mode 1 is most
dramatically affected by the change in Damper 2, Mode 2 by Damper 3.
This is a useful conclusion since it provides some instructions for the
modification of each natural frequency.
It is seen however from the above results that the response improvement
objective cannot be reached by modifying any single damper only. A
combination of these modifications on the dampers’ stiffness coefficients
must be attempted for the optimum outcome of the response.
7.3.3 Combination of different axial stiffnesses to achieve an

optimum response (Opt 1)
A systematic approach to modify the stiffnesses of individual dampers

must be formulated in order to find the optimum combination of the
values. The previous study showed the following results about the mode
characteristics
• Mode 1 most affected by Damper 2

• Mode 2 most affected by Damper 3
124
However, even though this is the case, all the modes are drastically
affected by the changes in the axial stiffness of an individual damper
since all the vibration modes are mostly axial deformations of the
dampers.
A systematic trial-and-error analysis is performed on the model with the

results summarized below in Table 7-19.
Table 7-19: Systematic study of the effect of damper stiffnesses and the results

Initially 1,2E+06 1,2E+06 1,2E+06 130 150 264
1 7,0E+05 7,0E+05 7,0E+05 108 123 205
2 6,0E+05 6,0E+05 6,0E+05 103 116 190
3 5,0E+05 5,0E+05 5,0E+05 97 109 174
4 5,0E+05 8,0E+05 5,0E+05 101 117 188
5 5,0E+05 8,0E+05 7,0E+05 103 125 198
of vibration as much as possible, all the ( values for the three dampers
To start with, in order to reduce the natural frequency of the third mode
are decreased equally until at some point the first mode slightly falls
below 100 Hz. This is an undesired condition due to the sine
environment conditions discussed before. Therefore this is a minimum
limit for the stiffnesses of the dampers. Then, to increase the first
natural frequency, Damper 2 is made stiffer, since this is the damper
most critical for the first mode. Furthermore the second natural
frequency is also increased by slightly increasing the stiffness of Damper
3. Doing this, the third natural frequency is also increased so this is said
to be a maximum limit for the stiffness of the dampers.
The damper tunings obtained at the end of Iteration 5 seem to give the
result that are the closest possible to the ones desired. All the natural
frequencies corresponding to the first three modes of vibration were
reduced, the third one more while the first two less, suggesting some
convergence. Still, the results of the suggested tuning of dampers are far
125
from the objectives since the first natural frequency became too low in
the process of optimization and the third mode is still too large.
Considering all these results, at the end it can be stated that the
response improvement objectives cannot be met only by changing the
axial stiffness of the dampers. The remaining radial and rotational
parameters of the dampers should also be modified, even though it
requires a principal design change.
7.4 Response optimization of the system with

theoretical dampers
In this section of the study, it is attempted to meet the response

improvement objectives using any possible modifications on the damper
stiffness parameters, not only the axial one but also the radial one.
7.4.1 Effect of the radial stiffness coefficient
Recalling the mode shapes of the original system utilizing the dampers
in their ultimate configuration, the modes of vibration obtained were
mostly due to the axial deformation of the dampers, either in-phase with
each other or out-of-phase. From the previous analyses, it was observed
that these modes were strictly dependent on the axial stiffnesses of the
damper and it was quite hard to achieve a perfect optimization
modifying only these values. Therefore in this section the effect of
changing the radial stiffnesses of the dampers are investigated.
Below in Table 7-20 are given some analyses performed with different
radial stiffnesses at a constant axial stiffness and the corresponding
results.
126
Table 7-20: Various analyses changing the radial stiffness and the results
k1 = 7e+05 N/m
k2,3 [N/m] Natural Frequency [Hz]
1 1.0E+06 1.0E+06 1.0E+06 105 119 203
2 5.0E+05 5.0E+05 5.0E+05 99 109 200
3 1.5E+06 1.5E+06 1.5E+06 109 123 204
From the results given above, it can be observed that changing the
radial stiffness mainly affects the natural frequencies that correspond to
the first two modes of vibration while the third natural frequency is not
so dramatically affected. The maximum deviations of the first, second
and the third natural frequencies between the 2nd and the 3rd iterations
are 9%, 11% and 2% respectively.
7.4.2 Effect of axial stiffness coefficient
Recalling the conclusions from the previous sections, it can be stated

that the radial stiffness mainly affects the first two modes while the
axial stiffness affects all three critical modes of the system. Therefore a
reasonable approach would be to increase the radial stiffness to increase
the first two natural frequencies and then decrease the axial stiffness to
reduce all the modes, resulting in some convergence. Such a procedure is
described with the obtained result in the table below.
Table 7-21: Various analyses changing the axial stiffness for a given radial stiffness
k2,3 = 6e+06 N/m

1 1.2E+06 1.2E+06 1.2E+06 136 159 268
2 9.0E+05 9.0E+05 9.0E+05 122 141 233
3 7.0E+05 7.0E+05 7.0E+05 111 127 206
4 5.0E+05 5.0E+05 5.0E+05 101 112 175
127
As seen from these results, choosing a high enough radial stiffness and
gradually decreasing the axial one, an improvement could be reached in
terms of the natural frequencies of the system.
7.4.3 Resulting modes of vibration of the optimal system (Opt 2)
At the end of the analysis, the following vibration modes shown in

Figure 7-86 are obtained from the FE model in CATIA.
Figure 7-86: Vibration mode shapes clockwise at (a) 101 Hz (b) 112 Hz (c) 175 Hz
128
7.5 Performance comparison of the damper tunings
using FE methods
In this section, more detailed FE analyses will be performed to compare

the performances of the damper tunings suggested previously.
7.5.1 Method of performance assessment
The objectives of the response improvement using FRM were defined

previously and the following analyses were conducted to achieve these
objectives. However, in order to give a more quantitative assessment on
whether the system response is improved or not, further simulations
have to be performed.
This method of performance assessment makes use of an advanced FE

model of the MIMA and the tuned dampers to simulate the system’s
random environment behavior. The details of the FE analysis of the
MIMA will not be included in the scope of this work, however what is
critical is that the tuned damper parameters can be input to this
advanced model to obtain the resulting FRFs at any critical point on the
MIMA (for this application, the optical elements).
Once the FRF is obtained from the FE analysis of the MIMA in terms
PSD (FRF squared), it is multiplied by a weighting profile to emphasize
the more critical portions of the frequency band. This profile is given
below in Figure
Table 7-22: Weighing profile to be multiplied by the FRF squared profile
Frequency [Hz] Factor

20 1.7
500 20
2000 20
129
25
20
15
10
5
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-87: Weighing profile plotted.
The result of the multiplication of the FRF2 profile with the weighing
function returns a Weighed FRF2 profile, whose RMS value can be used
to compare the performance of each tuning configuration. In principle
the function should be multiplied with the PSD one to get the
accelerations RMS. However, the PSD at this stage is still not well
defined, so it has been preferred not to link the result to this uncertain
input; the result will still be valid as long as the profile is quite flat in
the frequency range of interest.
7.5.2 Assessment of the test configuration
The above described procedure is applied using the dampers in their

testing configurations for loading conditions in all the three axes.
7.5.2.1 Test X loading

The computed FRF squared profile of the response to a random loading
in the X direction is given by
130
3,5
3
2,5
2
FRF2
1,5 FRF X^2

1 FRF Z^2
0,5 FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-88: Evaluated FRF2 profile for the three axes in X loading
Multiplying the above profile with the weighing function shown in

Figure 7-87, one gets
25
20
15
Mod FRF^2 X
10
Mod FRF^2 Z
5
Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-89: Weighed FRF2 profiles for X loading
Then the RMS value for each of these discrete profiles can be calculated
using the formula
131
∑ONP -N
= '
Q
Performing this simple calculation, the RMS values are found and are
summarized in the Table 7-23 below.
Table 7-23: RMS values for the weighed FRF2 profiles for the X loading condition
RMS X 7.234
RMS Y 0.341
RMS Z 0.988
7.5.2.2 Test Y loading

The FRF squared profile resulting from the loading in the Y direction is
shown below.
3,5
3
2,5
2
PSD
1,5 FRF X^2

1 FRF Z^2
0,5 FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-90: FRF2 profile for the three axes in Y loading
Once this profile is weighed by the weighing function it becomes
132
30
25
20
15 Mod FRF^2 X
10 Mod FRF^2 Z
5 Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-91: Weighed FRF2 profiles for Y loading
The RMS values calculated for the weighed FRF2 profile are given in the
table below.
Table 7-24: RMS values for the weighed FRF2 profiles for the Y loading condition
RMS X 0.213
RMS Y 8.570
RMS Z 2.880
7.5.2.3 Test Z loading

The FRF2 response to a random loading in Z direction is computed as
133
3,5
3
2,5
2
PSD
1,5 FRF X^2

1 FRF Z^2
0,5 FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-92: Evaluated FRF2 profile for the three axes in Z loading
Once multiplied by the weighing function given in Figure 7-87, the

weighed FRF2 profile becomes
40
35
30
25
20
15 Mod FRF^2 X
10 Mod FRF^2 Z
5 Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-93: Weighed FRF2 profiles for Z loading
Similarly, the RMS value is calculated for the three measurement axes
as
134
Table 7-25: RMS values for the weighed FRF2 profiles for the Z loading condition
RMS X 0.193
RMS Y 0.004
RMS Z 15.178
7.5.3 Assessment of the optimization performed through

changing axial stiffnesses (Opt 1)
The performance assessment procedure is applied using the suggested

dampers modified in terms of their axial stiffnesses for loading
conditions in all the three axes.
7.5.3.1 Opt 1 X loading

The FRF squared profile resulting from a loading in the X direction is
shown below in Figure 7-94
3
2,5
2
PSD
1,5 FRF X^2

1 FRF Z^2
0,5 FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-94: FRF2 profile measured for the three axes in X loading
135
When multiplied by the weighing function, this plot becomes
20
15
10 Mod FRF^2 X
5 Mod FRF^2 Z
Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-95: Weighed FRF2 profile for X loading
Then the RMS values corresponding to each measured axis can be

calculated. These results are given in the table below.
RMS X 6.872
RMS Y 0.336
RMS Z 1.006
7.5.3.1 Opt 1 Y loading

The computed FRF squared profile resulting from a loading in the Y
direction is given by
136
3
2,5
2
PSD
1,5 FRF X^2

1 FRF Z^2
0,5 FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-96: FRF2 profile measured for the three axes in Y loading
When weighed using the defined function, this profile becomes
20
15
10 Mod FRF^2 X
Mod FRF^2 Z
5
Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-97: Weighed FRF2 profile for Y loading
Then the corresponding RMS values to each measured axis can also be
calculated as
137
RMS X 0.283
RMS Y 6.668
RMS Z 3.086
7.5.3.3 Opt 1 Z loading

When the system is loaded in the Z direction, the resulting FRF squared
profile on the pint defined is computed and it is plotted below in Figure
7-98
4
3,5
3
2,5
PSD
2 FRF X^2
1,5
1 FRF Z^2
0,5 FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-98: FRF2 profile measured for the three axes in Z loading
Similarly, the weighed FRF2 profile can be plotted as
138
40
35
30
25
20 Mod FRF^2 X
15
Mod FRF^2 Z
10
5 Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-99: Weighed FRF2 profile for Z loading
And the corresponding RMS values for the three measurement axes are
given by
RMS X 0.012
RMS Y 0.064
RMS Z 11.221
7.5.4 Assessment of the optimization performed through

changing both the axial and the radial stiffnesses (Opt 2)
In section 7.4, an optimization of the response was suggested through

changing both the radial and the axial stiffnesses of the dampers in the
described manner. Here, these stiffness parameters are input to the FE
simulation for the assessment of the tuning performance.
139
7.5.4.1 Opt 2 X loading
The FRF squared profile computed at the designated location on the
MIMA for the X loading condition is given by
2,5
1,5
PSD
FRF X^2
1
FRF Z^2
0,5
FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-100: FRF2 profile measured for the three axes in X loading
Once this plot is multiplied by the weighing function, it becomes
16
14
12
10
8 Mod FRF^2 X
6
Mod FRF^2 Z
4
2 Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-101: Weighed FRF2 profile for X loading
The RMS values of each axis for this profile is calculated as
140
RMS X 7.334
RMS Y 0.279
RMS Z 1.985
7.5.4.2 Opt 2 Y loading

The FRF squared profile computed for the Y loading condition is shown
in the figure below.
2,5
1,5
PSD
FRF X^2
1
FRF Z^2
0,5
FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-102: FRF2 profile measured for the three axes in Y loading
The weighed form of this profile is shown in the figure below.
141
20
15
10 Mod FRF^2 X
Mod FRF^2 Z
5
Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-103: Weighed FRF2 profile for Y loading
And the corresponding RMS values are

RMS X 0.196
RMS Y 7.135
RMS Z 5.587
7.5.4.3 Opt 2 Z loading

The FRF squared profile from the simulation with a loading in the Z
direction is computed as
142
4
3,5
3
2,5
PSD
2 FRF X^2
1,5
FRF Z^2
1
0,5 FRF Y^2
0
0 500 1000 1500 2000
Fequency [Hz]
Figure 7-104: FRF2 profile measured for the three axes in Z loading
The above profile is multiplied by the weighing function to obtain the

weighed FRF2 profile as
30
25
20
15 Mod FRF^2 X
10 Mod FRF^2 Z
5 Mod FRF^2 Y
0
0 500 1000 1500 2000
Frequency [Hz]
Figure 7-105: Weighed FRF2 profile for Z loading
with the following calculated RMS values
143
RMS X 0.148
RMS Y 0.006
RMS Z 9.520
7.5.5 Performance comparison of the optimal damper tunings
Once all the calculations have been performed to assess the performance
of the initial test configuration and the improvements suggested, a
comparison between them is possible. The Table below summarizes all
the results of the previous performance computations for the test
configuration, the optimization performed by axial stiffness
modifications (Opt 1) and the radially and axially modified dampers (Opt
2).
Table 7-32: Results of the performance assessment summarized
RMS
Test Opt 1 Opt 2
X 7.234 6.872 7.334
X loading Y 0.341 0.336 0.279
Z 0.988 1.006 1.985
X 0.213 0.283 0.196
Y loading Y 8.570 6.668 7.135
Z 2.880 3.086 5.587
X 0.193 0.012 0.148
Zloading Y 0.004 0.064 0.006
Z 15.178 11.221 9.520
The first observation that can be made looking at these results is the
fact the both the optimizations resulted favorably as far as the RMS in
the Z axis is concerned (axial direction of the dampers). The critically
high initial value of 15.2 experienced a fall of 26% with the Opt 1
144
configuration and 38% with the Opt 2 configuration. This is an expected
result since the natural frequency corresponding to the Z translational
mode was the main objective of response improvement. Looking at this
data, it can be said that this objective was fulfilled.
However, there are other critical observations regarding these final RMS
results. The Opt 1 configuration obtained by only modifying the axial
stiffnesses did not cause much change in the transversal loading cases.
On the other hand the radially modified Opt 2 configuration suffers from
some increase in the final RMS for the transversal excitations up to 90%
in the Z axis when loaded in Y direction. This result can be attributed to
some high frequency torsional modes caused by the increased radial
stiffness of the dampers.
Finally it can be stated that both the optimizations suggested have their
advantages when compared to the testing configuration. However, there
is a certain trade-off between these two configurations. If the axial
performance of the system is desired to be increased further, a
compromise should be made from the transversal performance.
In this final part of the study, the designed and verified Finite Elements
model of the mockup testing setup was used to suggest some
improvements on the response of the system. This is done by getting the
optimal response from the FEM analysis by varying the individual
damper properties and using this information as damper tuning
feedback.
The objectives of response improvement were defined considering the

proximity of the first three critical modes of vibration of the system.. For
this purpose, the third mode of vibration at a higher natural frequency
compared to the others should be shifted to a lower frequency. This is
attempted by two different approaches.
145
The first approach to the response improvement was to use the
capabilities of the current damper design which allows changes in the
axial stiffness. Therefore just by varying the individual axial stiffness
coefficients of the three dampers, various runs of the FE model were
made to optimize the response. However in the end it was not actually
possible to meet the objectives of the response improvement since all the
modes were strictly dependent on the axial spring parameters.
The second approach was to assume a theoretical design of the dampers

that allows for any changes in the stiffness parameters. Through a
systematic modification procedure of the radial and the axial stiffnesses,
the optimum response of the system was obtained.
The two suggested damper configurations and the resulting modes of

vibration are summarized in Table 7-33.
Table 7-33: Suggested optimum spring properties and the resulting modes of vibration.
kaxial [N/m] kradial Natural Frequency [Hz]

Damper 1 Damper 2 Damper 3 [N/m] Mode 1 Mode 2 Mode 3
6 6 6 6
Init 1.2x10 1.2x10 1.2x10 1.7x10 131 151 265
5 5 5 6
Opt 1 5.0x10 8.0x10 7.0x10 1.7x10 103 125 198
5 5 5 6
Opt 2 5.0x10 5.0x10 5.0x10 6.0x10 101 112 175
Furthermore, the performance study of the two suggested optimizations

in the damper configuration is performed using an advanced FE model
of the MIMA instrument. After various calculations, the load
transmissibility on the instrument was obtained and its RMS value was
used as the basis of comparison for performance. At the end of this
study, both the suggested optimizations were proved to be advantageous.
Opt 2 configuration was more efficient than Opt 1 in reducing the
excitations in the axial direction while it had poorer performance in the
transversal loading conditions. The choice between these two
optimizations involve some compromise and is strictly application
dependent.
146
Chapter 8: Conclusion and final
remarks
In this study, the characterization and optimization process for a

vibration damping system designed for an instrument, devoted to a
space mission, was realized through experimental and computational
methods. In order to achieve this goal, several integrated objectives were
pursued simultaneously.
The first objective of the study was to design and manufacture a test
mockup to be used in place of the actual instrument for vibration testing
purposes. To achieve this, software tools were used to design an
inertially equivalent dummy simple enough for easy and cheap
manufacturing.
The second task to be performed was the individual testing and tuning
of the dampers to be used in the mission. For this purpose, a shaker
based test setup was implemented in which the single damper and a
representative dummy mass were assembled together as a single degree
of freedom system. Using this test setup, several characteristics of the
dampers were investigated and some important conclusions were
reached. To start with, the non-linear characteristics of the dampers
were investigated, observing different responses of the same damper to
different types of input or amplitudes of the excitation. Furthermore, the
effect of changing the tightening preload on the response of the dampers
was studied extensively and it was concluded that a wide range of
stiffness values could be achieved with a single damper design by
utilizing this regulation capability. Another useful study was performed
to characterize the effect of including additional ring elements in the
damper assembly and the result was found to be positive due to an
increased stiffness and stability of the dampers. Finally having tested
and investigated all these aspects, an ultimate damper tuning could be
achieved to achieve the optimum performance with respect to the given
environmental conditions of the mission.
147
The shaker testing procedures were adopted inspired from Experimental
Modal Analysis techniques which allow the description of a system
through its modal properties such as the natural frequencies, damping
ratio and mode shapes. Having obtained these parameters for the
ultimate damper configuration through single DOF testing, various
analytical calculation methods and mathematical models were utilized
to estimate the numerical properties of the individual dampers.
One of the most important objectives of this study was the

implementation of a Finite Elements model of the experimental setup
involving the damper-mockup assembly. A FE modeling was established
in CATIA software using spring elements to replace the dampers in
order to analyze the natural properties of the system. The various
directional stiffness coefficients of these virtual springs were estimated
using the previously calculated modal parameters from the single DOF
tests. Running the FE model, three significant modes of vibration were
observed at frequencies above 100 Hz, a positive result regarding the
environmental conditions. Besides the natural frequencies, the mode
shapes of the system were also investigated, showing two rotational
modes at lower frequencies caused by the out-of-phase deformation of
the springs and an almost purely translational mode at higher frequency
caused by the in-phase deformation of all the three springs.
With a mockup inertially equivalent to the actual MIMA instrument and

three dampers tuned at their ultimate configuration, the complete
system was assembled and tested on the shaker along the three axes
using sweep sine and random excitations considering the environmental
conditions of the real application. The response of the system was
measured and the results were analyzed using various signal processing
techniques. Considering the desired response of the system, the
resulting performance of the dampers was found to be satisfactory.
There were no significant resonances or amplifications in the sweep sine
region (5-100 Hz) that is a requirement for the sine environment.
Furthermore, the resulting transfer function in the random environment
was satisfactory with low amplifications that were cutoff after a certain
frequency. From the dummy model test results, this cutoff frequency
was found to be 300 Hz for the Z axis loading and around 150 Hz for the
148
X and Y axis loadings. All these results suggest that the damper tuning
performed initially was successful.
Besides providing invaluable information about the working behavior of

the system, the mockup testing also provided feedback data to check the
accuracy of the FE model implemented. The natural frequencies and the
mode shapes of the mockup tests were compared to the results of the
FEM analysis and the model was found to be acceptable because the
errors in the predicted natural frequency was less than 12%.
The final objective of the study was to suggest some improvements on

the response of the system rather than accepting the results coming out
of the tests. For this purpose, the previously verified FE model has been
used to find the optimal damper parameters that result in an improved
response in terms of the modes of vibration. Two optimizations were
suggested for the damper parameters, one including changes in the axial
stiffness therefore being applicable for the current damper design, the
second assuming a theoretical design which allows changes both in the
axial and the radial parameters. In order to support these optimizations
with a more quantitative approach, an advanced FE model of the actual
MIMA was used to calculate the FRFs and the corresponding RMS
values resulting from a random excitation on the most critical part of the
instrument. In the end, both these optimizations were found to be
advantageous, one having better performance in axial loading conditions
and the second acting more favorably when loaded transversally. The
choice of the optimization to be used surely requires some compromise
and is strictly ‘application dependent’. This could be the objective of a
further study regarding the optimization of the damping system as well
as the mechanical design of new damper elements having more versatile
stiffness parameters.
149
References
[1] Bellucci G.; Saggin B.; Fonti S.; Biondi D.; Cerulli P.; De Luca M.;
Altieri F.; Mattana A.; Alberti E.; Marzo G.; Zasova L.
MIMA, a miniaturized Fourier infrared spectrometer for Mars ground
exploration: part I, concept and expected performance, Proceedings of
SPIE, the International Society for Optical Engineering 2007
[2] Marzo, G. A.; Bellucci, G.; Fonti, S.; Saggin, B.; Alberti, E.; Altieri, F.;
Politi, R.; Zasova, L.; The Mima Team
MIMA: Mars Infrared MApper - The Fourier spectrometer for the ESA
Pasteur/ExoMars rover mission, 36th COSPAR Scientific Assembly 2006,
Beijing, China
[3] ESA – ExoMars – The ExoMars Instruments

http://www.esa.int/SPECIALS/ExoMars/SEMSZIAMS7F_0.html
Last visited on: April 2010
[4] ESA ExoMars Mission Information Sheet
[5] CNES, Centre National d'Études Spatiales, ExoMars Rover

http://smsc.cnes.fr/EXOMARS/GP_rover.htm
Last visited on: April 2010
[6] The MIMA Team – POLIMI, MIMA Environmental Conditions,

MIMA Thermomechanical Design documents 2009
[7] Patrick L. Walter, Pyroshock Explained, Technical Note PCB

Piezotronics, Inc.
150
[8] ENDEVCO 27A11 Product Data Sheet
[9] David J. Ewins - Modal Analysis And Modal Testing

Handbook of Noise and Vibration Control - Chapter 47, Wiley 2007
[10] Peter Avitabile, Experimental Modal Analysis (A Simple Non-

Mathematical Presentation), University of Massachusetts Lowell
[11] Kelly, S. Graham, Fundamentals of Mechanical Vibrations, 2nd ed.

McGraw-Hill 2000
[12] Robert B. Randall, Noise and Vibration Data Analysis, Handbook of

Noise and Vibration Control - Chapter 46, Wiley 2007
[13] Carlos A. Felippa, Advanced Finite Element Methods, Aerospace

Engineering Sciences - University of Colorado at Boulder, Spring 2009
151
Appendix A
A.1 MATLAB scripts

1.3 The need for a damping system
% Can Ozer
% The need for a damping system
% Parameters
f01 = 150; %[Hz]
f02 = 250; %[Hz]
w01 = 2*pi*f01;
w02 = 2*pi*f02;
xsi = 0.3;
f = 20:2000;
w = 2*pi*f;
H1 = (w01^2+j*2*w.*xsi*w01)./((w01^2-w.^2)+j*2*w.*xsi*w01);
G1 = abs(H1);
Phi1 = angle(H1)*180/pi;
H2 = (w02^2+j*2*w.*xsi*w02)./((w02^2-w.^2)+j*2*w.*xsi*w02);
G2 = abs(H2);
Phi2 = angle(H2)*180/pi;
% FRF squared
PSD1=G1.^2;
PSD2=G2.^2;
% Weighing profile
for i=20:500
factor(i-19)=(20-1.7)/(500-20)*i+1.7-(20.17)/(500-20)*20;
end
for i=501:2000
factor(i-19)=20;
end
% Weighed PSD
PSD_w1 = PSD1.*factor;
PSD_w2 = PSD2.*factor;
RMS_PSD1=norm(PSD1)/sqrt(length(PSD1))
RMS_PSD2=norm(PSD2)/sqrt(length(PSD2))
152
RMS_PSD_w1=norm(PSD_w1)/sqrt(length(PSD_w1))
RMS_PSD_w2=norm(PSD_w2)/sqrt(length(PSD_w2))
figure(1)
subplot(2,1,1)
plot(f,G1,f,G2);
title('FRF Magnitude');
xlabel('frequency [Hz]');
ylabel('Magnitude');
grid on
legend('System 1','System 2')
subplot(2,1,2)
plot(f,Phi1,f,Phi2);
title('FRF Phase');
ylabel('Phase [deg]');
grid on
figure(2)
plot(f,factor);
title('Weighing Profile');
ylabel('Factor');
grid on;
figure(3)
plot(f,PSD1,f,PSD2);
title('FRF^2 vs frequency');
ylabel('PSD');
grid on;
figure(4)
plot(f,PSD_w1,f,PSD_w2);
title('Weighed FRF^2 vs frequency');
ylabel('Weighed PSD');
grid on;
153
Mathcad Scripts
4.3.1 Estimation of the Modal Parameters from the

Experimental Data
j := −1
m := 0.35 [kg]
f0 := 450 [Hz]
3
ω0 := f0 ⋅ 2π = 2.827 × 10 [rad/s]
2 6
k := ω0 ⋅ m = 2.798 × 10 [N/m]
ζ := 0.4
2
ω0 + j ⋅ 2⋅ ω⋅ ζ⋅ ω0
H( ω) :=
 ω 2 − ω2 + j ⋅ 2⋅ ω⋅ ζ⋅ ω
 0  0
G( ω) := H( ω)
180
φ ( ω) := arg ( H( ω) ) ⋅
π
154
f_data := Mag_data := Pha_data :=
0 0 0
0 20.009 0 0.976 0 1.201
1 20.083 1 0.975 1 1.021
2 20.157 2 0.975 2 1.029
3 20.231 3 0.974 3 1.199
4 20.305 4 0.973 4 1.189
5 20.379 5 0.975 5 1.136
6 20.454 6 0.977 6 1.22
7 20.528 7 0.978 7 1.244
8 20.602 8 0.978 8 1.135
9 20.676 9 0.978 9 1.049
10 20.75 10 0.977 10 1.004
11 20.824 11 0.976 11 0.97
12 20.898 12 0.975 12 0.951
13 20.972 13 0.975 13 0.943
14 21.046 14 0.974 14 0.954
15 ... 15 ... 15 ...
FRF MAGNITUDE
2
1.5
Magnitude
G ( f ⋅ 2π)
1
Mag_data
0.5
0
3
200 400 600 800 1×10
f , f_data
Frequency [Hz]
FRF PHASE
0
− 30
− 60
Phase [deg]
ϕ ( f ⋅ 2π)
− 90
Pha_data
− 120
− 150
− 180
3
200 400 600 800 1× 10
f , f_data
Frequency [Hz]
155
4.3.2 Estimation of the Modal Parameters by Curve Fitting
j := −1
m := 0.35
[kg]
From FRF calculated using test data:

f1 := 100
[Hz] [Hz]
Mag 1 := 1.9
Pha1 := −50 [deg]
ω1 := f1 ⋅ 2 π = 628.319 [rad/s]
Initial guess:
f0 := 150 [Hz]
ζ := 0.3
ω0 := f0 ⋅ 2 π = 942.478
Solving the mathematcial model:

Given
2
ω0 + j ⋅ 2 ⋅ ω1 ⋅ ζ⋅ ω0
= Mag 1
 ω 2 − ω 2 + j ⋅ 2⋅ ω ⋅ ζ⋅ ω
 0 1  1 0
 2
ω0 + j ⋅ 2 ⋅ ω1 ⋅ ζ⋅ ω0 
arg   ⋅ 180 = Pha
  ω 2 − ω 2 + j ⋅ 2⋅ ω ⋅ ζ⋅ ω  π 1
 0 1  1 0 
 ω0   659.615 
  := Find( ω0 , ζ) =  
 ζ   0.32 
156
Getting the modal parameters
ω0 = 659.615
ζ = 0.32
ω0
f0 := = 104.981
2π
2 5
k := ω0 ⋅ m = 1.523 × 10
Mathematical Model:
2
ω0 + j ⋅ 2 ⋅ ω⋅ ζ⋅ ω0
H( ω) :=
 ω 2 − ω2 + j ⋅ 2 ⋅ ω⋅ ζ⋅ ω
 0  0
G( ω) := H( ω)
180
ϕ( ω) := arg ( H( ω) ) ⋅
π
Maximum Deviation
max_dev := max ( Pha_data − ϕ( f_data⋅ 2 π) ) = 3.253
→
maxdev_pha := max ( Pha_data − φ ( f_data ⋅ 2π ) )  = 11.649
157
f_data := Mag_data := Pha_data :=
0 0 0
0 20.008 0 0.975 0 0.033
1 20.057 1 0.974 1 0.017
2 20.107 2 0.973 2 -2.08·10-3
3 20.157 3 0.974 3 -0.043
4 20.207 4 0.974 4 -0.067
5 20.256 5 0.975 5 -0.235
6 20.306 6 0.973 6 -0.052
7 20.356 7 0.975 7 -0.174
8 20.406 8 0.975 8 -0.182
9 20.455 9 0.976 9 -0.184
10 20.505 10 0.975 10 -0.117
11 20.555 11 0.976 11 -0.098
12 20.605 12 0.976 12 -0.13
13 20.655 13 0.977 13 -0.129
14 20.704 14 0.977 14 -0.119
15 ... 15 ... 15 ...
FRF MAGNITUDE
2
1.8
1.6
Magnitude
G ( f ⋅ 2π )
1.4
Mag_data
1.2
0.8
50 100 150
f , f_data
Frequency [Hz]
FRF PHASE
0
− 30
− 60
Phase [deg]
φ ( f ⋅ 2π )
− 90
Pha_data
− 120
− 150
− 180
50 100 150
f , f_data
Frequency [Hz] 158

Appendix B
Further Dummy Testing Plots
B.1 Z axis sweep
The out-of-plane vibrations given as time history and FRFs
Figure B-106: Time history of out-of-plane vibrations for the Z sweep sine testing
159
0,1
FRF Modulus
X down left m s-2 Mod 0,09
X down right m s-2 Mod
Y m s-2 Mod 0,08
0,07
0,06
0,05
0,04
0,03
0,02
0,01
0
200
FRF Phase
X down left m s-2 Ph 150
X down right m s-2 Ph
100
Y m s-2 Ph
50
-50
-100
-150
-200
20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Frequency [Hz]
Figure B-107: FRF of out-of-plane vibrations for the Z sweep sine testing
B.2 Z axis random
The out-of-plane vibrations for the random testing in Z are as follows
160
Figure B-108: Time history of out-of-plane vibrations for the Z random testing
0,9
FRF Modulus
X down left m s-2 Mod
0,8
X down right m s-2 Mod
Y m s-2 Mod 0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
180
FRF Phase
150
X down left m s-2 Ph
X down right m s-2 Ph 100
Y m s-2 Ph
50
-50
-100
-150
-180
20 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz]
Figure B-109: Time history of out-of-plane vibrations for the Z random testing
161

Thesis Ozer Can

Uploaded by

Copyright:

Available Formats

Thesis Ozer Can

Uploaded by

Document Information

Original Description:

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

Thesis Ozer Can

Uploaded by

Copyright:

Available Formats

POLITECNICO DI MILANO

Polo Regionale di Lecco

Facoltà di Ingegneria Industriale

Testing and optimization of the tunable damping system for

Relatore: Prof. Bortolino SAGGIN

Co-relatore: Ing. Diego SCACCABAROZZI

Master thesis of:

Anno Accademico 2009 - 2010

Furthermore it is a great pleasure to thank everyone in the Measurements

Special thanks to,

my wonderful, ever caring brother, Ceyhun,

my loving mother and father,

The environmental conditions have been simulated through shaker tests

Keywords: damping, vibration control, Finite Element methods

I livelli di eccitazione previsti per la missione sono stati simulati

Parole chiave: smorzamento, controllo delle vibrazioni, elementi finiti

1.1 Mars Infrared Mapper (MIMA) 11

1.2 MIMA environmental conditions 12

1.3 The need for a vibration damping system 15

1.4 The dampers 18

CHAPTER 2: DESIGN AND MANUFACTURING OF A DUMMY MODEL FOR

2.1 Necessity of a dummy model 22

2.2 Computer modeling of the prototype 22

2.3 Modification of the dummy prototype to achieve an optimum model 25

2.4 Validity of the dummy model for testing purposes 28

CHAPTER 3: TUNING OF THE DAMPERS WITH ONE DEGREE-OF-FREEDOM

3.1 The need for damper tuning 30

3.2 Test methodology 30

3.3 Tuning of the dampers 43

3.4 Dampers in their final tuning and configuration 58

3.5 Results and discussion 60

CHAPTER 4: DAMPER PROPERTIES ESTIMATION 62

4.1 Excitation amplitude and linear modal parameters 62

4.2 Analytical formulation of the test setup 64

4.3 Estimating linear parameters from test data 66

4.4 Estimating torsional properties from test data 74

4.5 Estimated properties of the final damper configuration 77

4.6 Results & Discussion 79

CHAPTER 5: FINITE ELEMENT ANALYSIS OF THE DAMPER – MOCKUP

5.1 The need for a Finite Elements Analysis 80

5.2 Verification of the structural modes of the mockup using FEM 81

5.3 Analysis of the Damper-Mockup system using linear spring elements 87

5.4 Results and discussion 95

CHAPTER 6: TESTING OF THE DUMMY MODEL 97

6.2 Mockup testing and results 105

6.4 Results and discussion 118

CHAPTER 7: FINITE ELEMENT METHODS FOR THE RESPONSE

7.1 Introduction 120

7.2 Objectives of response improvement 121

7.4 Response optimization of the system with theoretical dampers 126

7.5 Performance comparison of the damper tunings using FE methods 129

7.6 Results and discussion 145

CHAPTER 8: CONCLUSION AND FINAL REMARKS 147

Figure 1-1: The Rover Module equipped with the MIMA 12

Table 1-1: Sine environment profile acting on the MIMA. 13

1.1 Mars Infrared Mapper (MIMA)

The Mars Infrared Mapper (MIMA) is a miniaturized FT-IR

The rover payload, PASTEUR, with its 12 instruments mounted on it,

1.2 MIMA environmental conditions