Eth 41572 02

ETH Library
Electrostatic modification of the

bending stiffness of adaptive
structures
Doctoral Thesis
Author(s):
Bergamini, Andrea
Publication date:
2008
Permanent link:
https://doi.org/10.3929/ethz-a-005774538
Rights / license:
In Copyright - Non-Commercial Use Permitted
This page was generated automatically upon download from the ETH Zurich Research Collection.
For more information, please consult the Terms of use.
Diss. ETH No. 18159
Electrostatic Modification of the

Bending Stiffness of Adaptive
Structures
A dissertation submitted to the

SWISS FEDERAL INSTITUTE OF TECHNOLOGY
ZURICH
for the degree of

Doctor of Sciences ETH Zurich
presented by
ANDREA BERGAMINI
Dipl. Werkstoff-Ing. ETH
born 16 April 1969
citizen of Milan - Italy
accepted on the recommendation of

Prof. Dr. P. Ermanni, examiner
Prof. Dr. L. Schlapbach, co-examiner
Prof. Dr. M. Motavalli, co-examiner
2009
ii
Contents
1 Introduction 11
1.1 Adaptive Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.1.1 Material Based Actuators . . . . . . . . . . . . . . . . . . . . . . . 13
1.1.2 Host Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2 Approaches to the Modification of the Mechanical Properties of the Host
Structure, an Overview of the State of the Art . . . . . . . . . . . . . . . . 20
1.2.1 Modification of the Elastic Modulus . . . . . . . . . . . . . . . . . . 21
1.2.2 Modification of the Moments of Area . . . . . . . . . . . . . . . . . 23
1.2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.3 Goals and Structure of the Present Work . . . . . . . . . . . . . . . . . . 30
2 Working Principle of the Electrostatic Tuning of the Stiffness 31

2.1 Connectivity, Interfaces and Shear Stress Transfer . . . . . . . . . . . . . 31
2.2 Shear Stress Transfer and Stiffness . . . . . . . . . . . . . . . . . . . . . 32
2.3 Shear Stress Transfer in Crystalline solids . . . . . . . . . . . . . . . . . . 33
2.4 Creation and Cancellation of Interfaces . . . . . . . . . . . . . . . . . . . 34
2.5 Friction at Real Interfaces and Shear Stress Transfer . . . . . . . . . . . . 36
2.6 Electrostatic control of the Shear Stress Transfer . . . . . . . . . . . . . . 38
3 Modeling of Structures with Tunable Stiffness 41

3.1 Local Behavior: Shear Stress Transfer by Means of Friction vs. Shear
Stress Transfer in Solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
iii
3.1.1 Deflection and Shear Stress Distribution in a Multi-layer Beam . . 42
3.1.2 Numerical model of the friction based shear stress transfer at in-
terfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.1.3 Linear vs. Non-linear Behavior . . . . . . . . . . . . . . . . . . . . 47
3.2 Global Behavior: Discrete Stiffness States in Homogeneous Multi-Layer
Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4 Experimental Work 63
4.1 Inhomogeneous Tunable Structures: Sandwich Beams . . . . . . . . . . . 64
4.1.1 Mechanical properties of fully bonded sandwich structures . . . . 64
4.1.2 Complete debonding . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2 Proof of Concept with a 140 mm Cantilever Beam . . . . . . . . . . . . . 67
4.2.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2.2 Measurement Technique . . . . . . . . . . . . . . . . . . . . . . . 69
4.2.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3 Characterization of a Large (>2m) Tunable Glass Fiber Reinforced Beam 76
4.3.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3.2 Shift of the Natural Frequency . . . . . . . . . . . . . . . . . . . . 78
4.3.3 Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4 The Effect of the Interface Topography . . . . . . . . . . . . . . . . . . . . 89
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5 A Practical Application of Elements with Tunable Mechanical Properties 97

5.1 Suppression of the Vibrations of a GFRP Footbridge Deck . . . . . . . . . 97
5.2 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2.1 Undamped Behavior of the Beam . . . . . . . . . . . . . . . . . . . 101
5.2.2 Damped Behavior of the Beam . . . . . . . . . . . . . . . . . . . . 106
5.2.3 Bridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
iv
6 Conclusions and Outlook 115
6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
v
vi
Abstract
Adaptive structures are characterized by their ability to reversibly modify their response
to external mechanical stimuli so as to extend their operational space (i.e. the range of
conditions under which they can operate). This feature differentiates adaptive structures
from conventional ones, that are designed to maintain their properties invariant over
time. Typically, adaptive structures are comprised of a host structure, one or more
actuators, sensors to determine the actual state or trajectory of the structure and a
control system. The bulk of the research done in this field focuses on the development
of new transducing materials and systems to be used in material based actuators.
The present work focuses on the development of host structures with variable mechan-
ical properties. The ability to modify the bending stiffness or the damping properties of
a structure can be used to increase the functionality of the actuators in active structures
or to suppress high amplitude vibrations in passive ones. The suppression of vibrations
can be achieved either by shifting the eigenfrequency of the structure, thus avoiding
resonance in the case of narrow band excitation (energy rejection) or by introducing a
higher level of damping (energy dissipation).
The original contribution made by this work lies in the development and description of
a structure that has the ability to adaptively modify its behavior based on a change in
mechanical properties, rather than thanks to the action of one or more actuators. The
change of the structure’s properties is caused by an apparent change of its topology.
Very limited literature on the use of electrostatic forces to modify the interaction be-
tween the layers of a multi-layer structure and thus its bending stiffness is available.
This moderate level of knowledge on the working mechanism of multi-layer structures
with variable bending stiffness, warrants the global approach to the discussion of this
topic taken in this dissertation. After giving an overview on the current state of the art in
variable stiffness structures, this work first describes the working principle at the base
of the modification of the mechanical properties of simple structures, then it proceeds
to confirming the validity of the description with the help of simple models and experi-
1
ments. Finally, it shows a practical application of the system in calculations that show
the effect of the electrostatic modification of the mechanical properties of a Glass Fiber
Reinforced Polymer (GFRP) bridge deck on its dynamic properties.
The method proposed for the adaptive modification of the bending stiffness and damp-
ing ratio of simple structures such as beams, consists of preparing structural elements
made of multiple layers. These layers are provided with electrodes coated with a dielec-
tric material at their surfaces. By applying an electrostatic field between adjacent layers,
these are coupled by means of the electrostatic forces that accompany the field. If the
amount of shear stress that can be transferred at the interface is large enough relative
to the external transversal load that bends the structure, the structure will behave as if
no interfaces existed between the layers.
A linear elastic analytical model considering interfaces within the system and a numeri-
cal model considering normal interlaminar stress and implementing contact and friction
at the interfaces of the layers of a multi-layer beam are used to demonstrate the mech-
anism of the electrostatic stiffening of a multi-layer beam. A comparison between the
shear stress distributions calculated with the two models shows that up to the max-
imum shear stress that can be transferred by means of friction at the interfaces, no
difference can be made between true topology switch (i.e. an actual cancellation of the
solid-solid interfaces in the structure) and electrostatic coupling of the layers. Once the
shear stresses at the interface generated by external load exceed this limit, the system
softens and displays high damping due to the friction between the contact surfaces.
For small loads (i.e. below the elastic limit of the system), discrete stiffness states are
predicted. Their number increases rapidly with the increasing number of layers of the
system.
In the experimental section of this work, experiments on two different systems are pre-
sented. Both are sandwich beams with stiff faces and a compliant core. The first system
is used to verify the relationship between normal electrostatic stress, shear stress and
stiffness of the system. The mechanical properties of the second system used for this
work are comparable to the ones of Empa’s cable stayed pedestrian bridge. The effect
of the electrolamination of Carbon Fiber Reinforced Polymer (CFRP) strips on the dy-
namic behavior of a GFRP I-beam are investigated. As expected, the stiffening of the
system can be demonstrated for low vibration amplitudes. At large amplitudes damping
is the dominant effect of the electrolamination of the CFRP elements.
The change in mechanical properties due to the electrolamination of the CFRP ele-
ments onto the GFRP base structure measured in the experimental section is used to
extrapolate the expected behavior of the GFRP deck of the pedestrian bridge that will
2
be upgraded to create an adaptive system based on the present work. The main result
shown in the last section of this work is that the damping that is obtained through the
interaction of the elements of the electrostatically tunable system is expected to have
a more beneficial effect in terms of vibration damping than the stiffening of the bridge
deck.
3
Zusammenfassung
Adaptive Strukturen zeichnen sich dadurch aus, dass sie die Fähigkeit besitzen, ihre
Eigenschaften in Abhängigkeit von externen Bedingungen so reversibel anzupassen,
dass ihr Betriebsraum erweitert werden kann. Diese Eigenschaft unterscheidet adap-
tive Strukturen von konventionellen, die so konstruiert sind, dass ihre Eigenschaften
zeitinvariant sind. Adaptive Systeme bestehen typischerweise aus einer Struktur, einem
oder mehreren Aktuatoren, einem oder mehreren Sensoren, wodurch der Zustand oder
die Trajektorie der Struktur ermittelt wird, und einem Regelungssystem. Der Schwer-
punkt der Forschung auf diesem Gebiet liegt bei der Entwicklung von neuen Werk-
stoffen und Bauteilen, welche die Fähigkeit aufweisen, Energie von einer Domäne
der Physik in eine andere (z.B. von der elektrischen in die mechanische Domäne)
umzuwandeln und somit als Aktuatoren oder Sensoren eingesetzt werden können.
Die vorliegende Arbeit setzt sich mit der Entwicklung von Strukturen mit veränderbaren
mechanischen Eigenschaften auseinander. Die Fähigkeit, die Biegesteifigkeit oder die
Dämpfungseigenschaften einer Struktur kontrolliert zu verändern, kann dazu verwen-
det werden, die Funktionalität von Aktuatoren in aktiven Strukturen zu verbessern oder
um die Unterdrückung von Schwingungen grosser Amplitude in passiven Strukturen zu
erzielen. Bei engbandiger Anregung kann die Unterdrückung von Schwingungen durch
Verschiebung der Eigenfrequenzen (also durch Vermeidung von Resonanzbedingun-
gen) erreicht werden. Bei breitbandiger Anregung, ist die Erhöhung der Dämpfung und
somit der Energiedissipation von Vorteil.
Der Beitrag dieser Arbeit zum Gebiet der adaptiven Strukturen besteht aus der En-
twicklung und Untersuchung von Strukturen, welche Ihr Verhalten dank einer Modifika-
tion ihrer mechanischen Eingenschaften und nicht durch die Einwirkung von Aktua-
toren ändern können. Die Änderung der mechanischen Eigenschaften wird durch eine
scheinbare Änderung der Topologie der Struktur erziehlt.
Zur Verwendung von elektrostatischen Kräften zur Änderung der Wechselwirkung zwis-
chen Schichten einer Mehrschichten-Struktur ist wenig Literatur verfügbar. Dieser be-
5
grenzte Stand des Wissens zum Thema der Mehrschichten-Strukturen mit variabler
Biegesteifigkeit rechtfertigt den umfassenden Umgang in der Diskussion der Materie,
der für diese Dissertation gewählt wurde.
Nach einer Übersicht über die erwarteten Vorteile von adaptiven Strukturen gegenüber
herkömmlich passiven Strukturen wird das vorgeschlagene Funktionsprinzip der Steifig-
keitsänderung vorgestellt. Weiter werden wesentliche Aspekte der postulierten Funk-
tionsweise mittels Modellen und Experimenten bestätigt. Im letzten Kapitel wird eine
praktische Anwendung der elektrostatischen Änderung der Biegesteifigkeit und der
Dämpfungseigenschaften von Biegebalken vorgestellt. Anhand der numerischen Berech-
nung des dynamischen Verhaltens eines aus glasfaserverstärktem Kunststoff (GFK)
gefertigten Brückendecks, dessen Steifigkeit und Dämpfungseigenschaften geändert
werden können, gezeigt.
Die für die Veränderung der Biegesteifigkeit und des Dämpfungskoeffizienten vorgeschla-
gene Methode besteht darin, dass man mehrschichtige Strukturen erstellt, deren Gren-
zflächen jeweils mit einer Elektrode und einem darauf geschichteten Dielektrikum verse-
hen sind. Durch anbringen eines elektrischen Potentials zwischen benachbarten Schich-
ten, werden diese von den entstehenden elektrostatischen Kräften aneinander gekop-
pelt. Wenn die Kopplung im Vergleich zu den auf die Struktur wirkenden äusseren
Kräften ausreichend stark ist, wird sich die Struktur wie eine durchgehende (also nicht
mit Grenzflächen versehene) verhalten.
Der Vorgang, durch den ein Mehrschichten-System mittels elektrostatisch generierter
Normalkräfte versteift wird, wird anhand von numerischen Modellen, welche Kontakt
und Reibung berücksichtigen, und durch analytische linear-elastische Modelle dargestellt.
Die mit den zwei Modellen berechnete Schubspannungsverteilungen wurden verglichen.
Der Vergleich zeigt, dass wenn die maximale Schubspannung, die an den Grenzflächen
übertragen werden kann, nicht überschritten wird, die Schubspannungsverteilung in
einer elektrostatisch gekoppelten Mehrschichtenstruktur sich nicht von der Verteilung
in einer monolithischen Struktur unterscheiden lässt. Wenn die maximale Schubspan-
nung an den Grenzflächen überschritten wird, sinkt die Steifigkeit des Systems und
der Dämpfungskoeffizient steigt aufgrund der Reibung, wodurch Energie an den Gren-
zflächen dissipiert werden kann. Für ausreichend kleine externe Lasten, können durch
Aktivieren der Grenzflächen in unterschiedlichen Mustern zahlreiche, diskrete Steifigkeits-
werte realisiert werden. Die Anzahl der Werte nimmt mit der Anzahl der Grenzflächen
im System rasch zu.
Im experimentellen Teil dieser Arbeit werden zwei Mehrschichten-System untersucht.
In beiden Fällen handelt es sich um Sandwich-Balken mit steifen Aussenschichten
6
und einem nachgiebigem Kern. Die erste Struktur wird dazu verwendet, um den Ein-
fluss der Anlegung von Normalspannungen an den Grenzflächen auf die Steifigkeit
des Balkens zu zeigen. Die mechanischen Eigenschaften des zweiten Balkens, der
in dieser Arbeit präsentiert wird, entsprechen denen, der Bestandteile der Fussgänger
Brücke, die an der Empa errichtet wurde. Der Effekt der elektrostatischen Kopplung von
steifen kohlefaserverstärkten Kunststoff (CFK) Elementen an den weichen Kern aus
glasfaserverstärktem Kunststoff (GFK) auf die dynamischen Eigenschaften der Struk-
tur werden untersucht. Wie aufgrund der Modellierungsarbeit erwartet, kann die Ver-
steifung des Systems für kleine Schwingungsamplituden nachgewiesen werden. Bei
grossen Amplituden ist die Erhöhung des Dampungskoeffizient der dominante Effekt.
Zum Abschluss wird die im experimentellen Teil dieser Arbeit ermittelte Änderung der
mechanischen Eigenschaften einer elektrostatisch koppelbaren GFK-CFK Struktur ver-
wendet, um das erwartete Verhalten des GFK Decks einer Fussgängerbrücke zu ex-
trapolieren. Aus den Schätzungen, die mit Hilfe eines numerischen Modells durchgeführt
wurden, geht hervor, die Dämpfung, welche aufgrund der Wechselwirkung an den
Grenzflächen zwischen GFK Struktur und CFK Versteifungselementen zustande kommt,
erwartungsgemäss einen stärkeren Einfluss auf das dynamische Verhalten der Brücke
haben wird, als die Erhöhung der Steifigkeit und somit der Eigenfrequenzen der Struk-
tur.
7
List of Symbols
C capacitance of a parallel plate condenser

D bending stiffness of a beam
δ thickness of a dielectric layer
E elastic modulus (in mechanical equations), electric field (in electrostatic equations)
Eb breakdown electric field
G shear modulus
²0 permittivity of vacuum
²r dielectric constant of a material
²xy shear deformation
η damping ratio
λ logarithmic decrement
P mechanical load
Q shear force
σxy shear stress
f
σxy friction stress at an interface
i
σxy shear stress in the ith layer
int
σxy shear stress at an interface
σyy normal stress in y direction
el
σyy electrostatic normal stress in y direction
U electrostatic potential
Ui interaminar electrostatic potential
v(x) transversal displacement at the position x
W energy
9
Chapter 1
Introduction
This chapter gives the rationale for the work presented in this dissertation. First, it
shows where the contribution of structures with variable stiffness presented in this work
is situated in the context of the research on adaptive structures. Furthermore, a brief
overview of the state of the art in the field of variable stiffness host structures is pre-
sented in section 1.2, with a special attention to the energy requirements that are of
paramount importance for most applications of adaptive structures. Finally, an overview
of the goals of the present work is given.
1.1 Adaptive Structures
For a long time, engineering structures have been designed to be, ideally, invariant
over time. Hence, the global properties of the structure were determined with the goal
to satisfy at all times the extreme set of the requirements (marked in red in figure 1.1)
determined by the application. This set is determined based on the conditions that
are likely to be encountered during the service life of the structure, as shown in figure
1.1. Such requirements, that will be referred to as the design space, may include but
not be limited to properties such as high ultimate strength, high stiffness, low weight,
and many more. Some properties of the design space (the axes in figure 1.1) may
be in conflict with others. The development of low density, high strength materials and
composites (e.g. carbon fiber reinforced polymers) is a response to the need to fulfill
such conflicting requirements. In the case of the development of lightweight structures
with a high strength that are made by the use of high strength materials, the dynamic
behavior of the systems represents a limiting factor to their performance.
While covering the complete space of requirements simultaneously is a viable approach
11
Requirement 1
Ideal Design Envelope for Time Invariant Systems

Requirement 2
Requirement 7
Use
Case c
Requirement 6 Use Case b

Use Case a Requirement 3
Requirement 5 Requirement 4
Figure 1.1: A time invariant system is designed to fulfill the envelope space of require-
ments at all times. The use cases represent different operational conditions that the
structure will be exposed to during its operational life. Examples of load cases could be
’take off’, ’high altitude cruise’ or ’landing’, in the case of an aircraft.
for a great number of applications, this is not always feasible and is bound not to provide
optimum system properties in all situations. The need for lighter, more energy efficient,
better performing structures calls for an optimization of the covered design space. For
example, in many civil, automotive and aerospace engineering applications, the design
of light, slender structures is necessary to achieve better performance. The use of
high strength materials, such as fiber reinforced polymer composites makes it possi-
ble to satisfy the strength requirements of many such structures. Lightweight, slender
structures tend to be more prone to vibrate with large amplitudes than heavier, mas-
sive ones. Furthermore, operating loads have a more marked effect on the vibratory
properties of a lightweight system than they would have on heavy structures. This sit-
uation makes the implementation of time-invariant vibration mitigating measures more
challenging and constitutes a rationale for the development of adaptive systems for the
reduction of vibrations.
Adaptive structures are characterized by their ability to react to external stimuli (forces)
in a way that allows them to achieve a performance that cannot be achieved using
’passive’ materials, i.e. invariant structures.
A simple adaptive system is generally composed of a host structure, an actuator (more
or less integrated into the structure) and a control system, as schematically shown in
figure 1.2.
Through the action of the actuator, the control system minimizes the deviation of the
12
Desired Actual
Trajectory Plant Trajectory
Controller Actuator
(Structure)
+
Feedback
Sensor
Figure 1.2: Schematic representation of an adaptive mechanical system with its base
components: structure, actuator and controller. Adapted from [5].
actual state of the system from the desired one. The trajectory of the system, i.e. its
position as a function of time (thus the mechanical stress in the system) is one of the
most common examples of state that is controlled using an actuator. The introduction of
a time-dependent (controllable) component to the properties of the system (specifically
to its energy level) allows for new degrees of freedom in the design process.
Apart of the continuous development of powerful control systems that allow for the man-
agement of increasingly fast and complex processes, the introduction of transducer-
based actuators (as opposed to ’traditional’ pneumatic, hydraulic and electrical motor
based actuators, also referred to as geometric actuators) have given a strong stimulus
to the development of smart structures.
1.1.1 Material Based Actuators
Transducer materials are materials that exhibit the ability to transform energy from one
physical form to another by virtue of specific coupling mechanisms [5, 41] and are at
the core of transducer material based actuators. More generalized definitions include
the transformation of energy also within a domain [10]. Figure 1.3 outlines the best
known multi-physical interactions and gives examples of materials and systems that
exploit such couplings. Materials that are used to obtain specific effects thanks to the
coupling between different physical domains (e.g. mechanical and electrical, thermal
and electrical, etc.) are generally defined as ’smart’, ’active’ or ’adaptive’ materials.
Definitions of most of these terms are found in [41] and will be adopted for this work.
actuator An actuator can be seen as a system that establishes a flow of energy be-
tween an input (electrical) port and an output (mechanical) port. The actuator is
13
transducing some sort of input power into mechanical power. [41]
active In active actuators the work exchange can take any positive or negative value,
dW ≷ 0. For practical purposes, this means that active actuators can either
increase or decrease the energy level of the controlled system. [41]
semi-active In semiactive actuators the work exchange can only be negative, dW 6 0.

In practice, this means that semiactive actuators can only dissipate energy as a
consequence of mechanical interaction with the controlled system. [41]
smart A smart structure is comprised of transducers that are used both for sensing and
actuating, thus allowing for the concomitant implementation of two of the functions
needed for adaptive structures.
adaptive The term adaptive refers to structures and systems that are able to change
their properties or behavior in order to adjust themselves to new conditions. Struc-
tures or systems comprising actuators as the ones listed above and control sys-
tems can be expected to be adaptive.
passive adaptive The concept of passive adaptive structure applies to non-actuated

structures (dW = 0 or dW ≈ 0) that can nevertheless adjust their behavior to new
conditions.
In the case of mechanical smart structures, the output required from a transducer is
generally in the mechanical domain and can be expressed in the form of the power
conjugate variable force and velocity or torque and angular velocity [10]. Table 1.1
shows a comparison between transducer material based actuators and traditional actu-
ators. The use of actuators in a structure implies the presence of an appropriate source
of energy. While the nature and properties of such a source are not the subject of this
work, the need to minimize the energy requirements of the ’active’ components of an
adaptive structures is a parameter of paramount importance for any application.
The most noticeable difference between traditional and transduction based actuators
is that in the latter, the transformation of energy between domains takes place at the
material level. In many cases, this leads to a better scalability in the direction of smaller
systems. If the coupling between physical domains is bidirectional, the same device
can be used as both a sensor and an actuator. The use of transducer materials for
sensing and power generation thanks to this fact has also been reported, [4, 18, 29, 42].
The properties that characterize materials-based actuators make new applications that
could not be realized using traditional actuators more practicable or even possible. A
14
Q the
T→ rmo hν
T→ σ/ε mec
E ha
(N nic
al:
chemoluminescence:
iTi
x→hν
)
ph opoly ε
SMA semi-active
Az →σ/
oto
,S
hν
MP active
me ers
hν , Cd ctric
Me toele
→
ch
ph
passive
m
U, S, S :
an
o
ica
∆σ i
l:
mSMA: Gd alloys
Magnetostrictive: Terfenol D
H→σ/ε
H Magnetorheological: Fe-Suspensions, ferrofluids W
H→η , F
ct) ,
AP PVD
ire cle
thermoelectricity:
various perowskites,
pyroelectricity: SiO2...
T→E
: E , ns
ind us
ve ZT sio
cti c: P
σ/ MA ( al: m
a ri e n
o : sp
ctr lect ive Su
ele zoe strict
S nic
l :
pie ctro ica
ted ha
le ε o l og
σ/ rhe
oa ec
e ms
ε
E→ ctro ste
Pt- mom
s y
Ele η yer
z→
e
→
E lti- la yH
c
ch
Mu I z, ζ
xO
C
E→
E Electrochemical systems:
µ
Batteries, etc.
Figure 1.3: Chart representing the most common forms of interaction between different
physical domains, represented by the symbols in the circles: Q- Thermal energy, hν-
Energy in form of electromagnetic radiation, W - Mechanical energy, µ- Energy in form
of chemical potential, E- Energy stored in an electrical field, H- Energy stored in a
magnetic field.
short list of such applications includes among others the propulsion of vehicles [12],
pumps [51], active prosthetics [6, 27], prestressing of civil engineering structures and
building materials [28,39], the active suppression of vibrations [2,44–46], shape change
[15, 36].
Some of the main advantages claimed for the use of active materials based actuators
instead of more conventional electric or hydraulic actuators include lower weight and
better integrability. In perspective given the lower level of mechanical complexity of
transducer material based actuators, one could expect that eventually a high level of
reliability should be achieved.
15
Table 1.1: Comparison of traditional vs. material based actuators (adapted from [41])
Traditional actuators Transducer material based actuators

• Based on geometrical transducers† • Based on transducing materials (possi-
bly in combination with geometrical con-
cept)
• Off-the-shelf availability • Designed for the application
• Good performance at normal scale • Good for meeting miniaturization de-
mands
• Lumped approach: discrete compo- • Integrated and embedded approach:
nents in motion control systems. Used open to smart structure concepts. Pur-
in combination with external sensors suit of the smart actuator concept
• Conventional mechanical transmis- • New transmission designs based on
sions for (output) impedance matching hinges and friction
†
Geometrical transducers are devices that owe their ability to transform energy from one domain to
another to a geometrical feature. For example, a DC motor owes its ability to transform electrical energy
into rotational mechanical energy to the geometry of the interaction between current and magnetic fields.
1.1.2 Host Structure
A considerable effort is put in the development of novel transducer materials and actu-
ator systems for adaptive structures applications. Some of the couplings in figure 1.3
(marked in red) imply the modification of mechanical properties instead of transforma-
tion of energy from a non-mechanical domain into mechanical work and exhibit thus a
passive adaptive or semi-active behavior. These effects are substantially different from
the ones seen in active materials and can be used to develop passive adaptive host
structures. This less intensively investigated aspect of adaptive structures is neverthe-
less a promising approach to the extension of the functionality of adaptive system that
include actuators as well as for the development of purely passive adaptive systems.
In actuated systems, the control over the stiffness (i.e. the mechanical impedance)
of the host structure allows for an improved efficiency in the use of the actuators, as
schematically shown in figure 1.4. Only the selection of an appropriate transmission
ratio makes it possible to lift the weight attached to the pulley. If the ratio is too high for
a given actuator (e.g. an electrical motor), the energy will not flow into the mechanical
system but rather be dissipated in form of heat in the actuator. Similarly, if the stiffness
of a structure is too high, actuators integrated in the structure will not be able to carry
16
out their task in an efficient manner.
The schematic representation of an actuated system including a host structure with
tunable mechanical impedance equivalent to figure 1.2 is thus extended as shown in
figure 1.5.
Actuated structures with tunable mechanical impedance can be used for novel applica-
tions, such as the aeroelastically assisted shape change of structures [11]. A compre-
hensive presentation of the use of tunable stiffness elements in adaptive structures can
be found in [20, 35]. The modification of the stiffness of a structure as a method to its
morphing is presented in [25].
Also non-actuated systems can draw benefits from the adaptation of the host structure
(see figure 1.6) as a means to exert control over the flow of energy from the environment
(external perturbations) into the system. The primary goal is then not to change the
energy level of the system (dW ≈ 0) but rather to change its mechanical impedance.
Depending on the method chosen for the modification of the stiffness, a change in the
energy level may be a more or less marked side effect. A few examples of structures
with variable mechanical properties can be found in literature. One example of such
work is found in the development of joints with variable friction parameters reported
in [22–24] where the conditions at the boundaries of beam elements are modified by
modulating the friction in the joints between elements. Also the modification of the
stiffness of structures has been reported [17, 21, 31].
The coupling of damping devices to a structure such as a cantilever beam or the use of
active systems currently represent the methods of choice for the purpose of reducing
A A
a b
m
m
Figure 1.4: Mechanical impedance matching: (a) the actuator heats up as power is not
transferred to the load and (b) the power can be transmitted to the load after pulley’s
impedance is matched to the actuator. [41].
17
Desired
Impedance Structure
Controller
Desired Actual
Trajectory Actuator Plant Trajectory
Actuator
Controller (Structure)
Feedback
Feedback
Sensor
Figure 1.5: Schematic representation of an adaptive mechanical system with an ex-

tended set of components: structure, structure controller, actuator and actuator con-
troller.
Disturbance
Desired Actual
State + Plant State
Controller
(Structure)
-
Feedback
Sensor
Figure 1.6: Schematic representation of a passive adaptive mechanical system con-

sisting of structure and structure controller
the amplitude of its vibrations.

The modification of the mechanical properties of structural elements, such as their
bending stiffness, can be of special interest for vibration mitigation applications. By
modifying the bending stiffness of a structural element it is possible to influence its vi-
18
bratory properties, such as the eigenfrequencies. This opens the way to application
of methods for the mitigation of vibrations by suppressing the adsorption of energy at
given frequencies.
The observation that high-amplitude vibrations are often caused by resonance phenom-
ena leads to the conclusion that, while the spectrum of the exciting forces can generally
not be influenced, it is in principle possible to shift the natural frequencies of a system,
so as to avoid the absorption of the high energy components of the exciting spectrum.
While this idea is almost trivial, its realization may be faced with serious technical diffi-
culties.
The vibratory properties of a body depend on its mass and its mechanical properties.
In the case of the bending natural frequencies of a beam, the eigenfrequencies are
described by:
s
D
ωk = βk2 · (1.1)
ρ
where ωk is the k th eigenfrequency of the system, βk is the k th solution of the equation

describing the vibration of the system, D is the stiffness of the system and ρ is its
density [52].
For the bending vibrations of a simple beam, the relevant mechanical property is the
bending stiffness D. Two parameters determine the bending stiffness of a structural
element, as shown in (1.2):
D = EIz (1.2)
Where E is the elastic modulus of the material the structure is made of and Iz is the
second moment of area of the structure, a geometric property of the structure. Since
the mass of the system is generally expected to be invariant, the vibratory properties
will be tuned by modifying of the stiffness of the system.
19
1.2 Approaches to the Modification of the Mechanical
Properties of the Host Structure, an Overview of the
State of the Art
In the case of structural elements made of one bulk material, such as the simple can-
tilever beam shown in figure 1.7, a modification of the bending stiffness of the system
can be achieved either by a modification of the elastic properties of the constituting
material or by a change of the geometry of the system.
Different approaches to the modification of the bending stiffness of simple structural
elements for passive adaptive structures will be presented and briefly discussed using
hypothetical examples and examples from literature.
As outlined in section 1.1, adaptive structures are chosen over time invariant structures
when the latter cannot fulfill the requirements set by a given number of use cases.
Weight is often one of the most stringent constraints, so that energy storage and thus
energy supply are likely to be among the critical factors that need to be addressed in
the development of adaptive structures.
In the considerations made in this section, we estimated the energy demand needed
to implement the solutions outlined in section 1.2.1 and 1.2.2. The energy to obtain a
z
F
l
B t
x
y
b
Figure 1.7: Simple cantilever beam with cross-section b · h subject to a load F at a

distance l from the origin.
20
certain increase in stiffness by modifying the elastic modulus or the second moment of
area is compared to the energy input needed to obtain the maximum possible stiffness
change in a cantilever beam using different methods.
The experiments are thought in comparison with a glass fiber reinforced polymer (GFRP)
I-beam used for the experimental work presented in section 4.3 and shown in figure 4.7.
The beams were thought to have the same initial bending stiffness as the GFRP beam.
The energy needed to achieve the maximum stiffness that can be realized for the de-
scribed stiffening methods was calculated. According to the manufacturer, the bending
stiffness of the reference GFRP beam is D=71.3 · 109 Nmm2 [16].
While the cases presented in this section are purely hypothetical and do not neces-
sarily represent engineering grade solutions to the problem of modifying the bending
stiffness of a structural element, they are meant to show the difficulty of developing
energy efficient passive adaptive host structures for adaptive systems.
1.2.1 Modification of the Elastic Modulus
At a fundamental level, the elastic properties of solids are governed by the electrostatic
attraction and repulsion forces between atoms of a material. The shape of the poten-
tials that determine the elastic properties of the material is mainly determined by the
nature of the bond (metallic, covalent, ionic etc.), the geometric distribution and the
electronic properties of the atoms [33]. For most engineering materials, the options for
a substantial modification of their elastic properties are extremely limited.
Significant changes in the elastic properties of solids are known mainly for certain solid-
solid phase transformations, such as the γ ↔ α0 phase transformation in NiTi shape
memory alloys. In this case, the high temperature (body centered cubic, bcc) γ phase
has an E-modulus of approximately 80 GPa, compared to 35 GPa for the low tempera-
ture (face centered cubic, fcc) α0 phase [7]. The noticeable change in elastic properties
in this case is not due to changes in the interaction potentials at an atomic level but
rather to the activation of twinning/de-twinning mechanisms in the fcc phase that are
not available in the bcc phase.
Also well known is a marked reduction of the shear modulus G of elastomers and par-
tially crystalline thermoplastic polymers in correspondence with an increase of the tem-
perature above the glass transition temperature Tg . Typical changes in shear modulus
G due to this second order transformation are of the order of 10 to 100 times. In both
cases, the change in material properties is due to temperature driven processes and is
thus subject to the laws of heat transfer.
21
Energy Considerations
In the case of the γ ↔ α0 phase transformation, the energy necessary to achieve a
stiffness increase of the order of three times can be estimated as follows:
W = Wth + W pt = m · ∆T · cp + ∆Gpt (1.3)
where Wth is the thermal energy needed to heat the material to the phase transfor-
mation temperature, Wpt is the phase transformation Enthalpy ∆Gpt for γ → α0 , m is
the amount of material to be heated and to undergo the phase transformation, T is the
temperature difference between the initial temperature (e.g. room temperature) and the
phase transformation temperature, cp and ∆Gpt are the specific heat capacity and the
phase transformation free enthalpy, respectively. Considering that cp is approximately
0.5 J/g·K [7] for a ∆T of 60 K, Wth is of the order of 30 J/g. Differential scanning
calorimetry measurements show that Wpt is also of the same order of magnitude (ap-
proximately 20 J/g) [7]. Assuming a total energy requirement of approximately 50 J/g,
the change in stiffness is energetically quite expensive and, given the nature of the
phenomenon, it is bound to be limited in speed by heat transfer processes.
A NiTi beam with an I cross-section and (54 mm wide, 108 mm tall, with a thickness
d=5.4 mm, as shown in Figure 1.9) has a cross sectional area of approximately 113.7·
103 mm2 and a bending stiffness D of approximately 71.3·109 Nmm2 , in its low temper-
ature phase α0 . The volume of such a 2500 mm long beam is 2.843· 106 mm3 and the
weight is approximately 18’340 g. The total energy needed to attain the stiffening of
the beam by a factor of approximately 2.8 thanks to the α0 → γ phase transformation is
thus of the order of 900 kJ.
In the examples outlined in this section, the change in elastic modulus takes place in
connection with a thermodynamic first or second order transition. The occurrence of
such phase transformations is thus mainly temperature controlled, although especially
in the case of the γ ↔ α0 phase transformation in NiTi alloys, mechanical stress also has
an influence. Heat transfer mechanisms limit the rate at which such desired temperature
driven changes in elastic properties can take place and thus the frequency response of
the system.
22
1.2.2 Modification of the Moments of Area
Shape Change
For a simple, solid, cantilever beam with a rectangular cross-section, the bending stiff-
ness is given by:
bh3
D = E · Iz = E · (1.4)
12
Where E is the elastic modulus of the material and Iz is the second moment of area
of a rectangular cross-section like the one shown in figure 1.7, where b and h are the
width and thickness of the cantilever, respectively.
Given an elastic modulus E for the beam material, the geometry of the cross-section of
a cantilever has a significant influence on its bending stiffness, which has been exten-
sively investigated.
The second moment of area Iz in 1.4 is calculated for an arbitrary cross-sectional ge-
0.1 a
0.2 a
2a 2a
a/ π
0.2 a
0.5 a
2a
A = A = A
1 2 3
I < I < I
z,1 z,2 z,3
Figure 1.8: Effect of the morphing of the cross section geometry on the second moment
of area of a beam. A comparison of the second moment of areas for the different cross
sections for a constant area A = a2 is given in (1.6).
23
ometry as: ZZ
Iz = y 2 dS (1.5)
A
with dS = dx · dy and the co-ordinate system with its origin in the center of gravity of the
cross-section, as depicted in figure 1.7. For a given cross-sectional area, an appropriate
selection of the geometry yields an increase in the stiffness of the beam. So, for the
cross-sections shown in figure 1.8 we will obtain using (1.5) that Iz,1 < Iz,2 < Iz,3 , while
A1 = A2 = A3 = a2 :
a4 a4
Iz,1 = ≈
4π 12.5
a4
Iz,2 =
3
5.824a4
Iz,3 = ≈ a4 (1.6)
6
A modification of the cross-sectional shape from cross-section 1 to cross-section 3
yields an increase in bending stiffness by a factor of approximately 12.5. This cir-
cumstance is well known and widely used to optimize the stiffness of structures while
maintaining their mass constant. Nevertheless the development of a cantilever beam
with variable stiffness based on the morphing of the geometry from cross-section 1 to
cross-section 3 presents a number of practical difficulties if the structure is made of a
commonly used engineering material.
More realistically, a fairly moderate geometry change can be achieved by modifying the
height of the web of an I-beam:
Similarly to (1.6), (1.8) shows the effect of a modification of the geometry on the second
moment of area of the I-beam shown in figure 1.9:
Aundef ormed = Adef ormed (1.7)
d + hweb 2 dh3web d + khweb 2 dk 2 h3web

Iz,undef ormed = 2bd( ) + < Iz,def ormed = 2bd( ) + (1.8)
2 12 2 12
The increase of the web height by a factor k leads to an increase in bending stiffness,
as shown in (1.8). Keeping the volume of the web constant, the width of the web will
then be reduced.
If the thickness of the flange is significantly smaller than the height of the web, the effect
can be considered of the order of k 2 for both terms of the right hand side of (1.8). The
remarkable material properties of NiTi alloys allow for large (up to approximately 7%)
24
d
d
hweb h k hweb k hweb+d
1/k d
d
b b
deform
Aundeformed, Iz,undeformed Adeformed, Iz,deformed
Figure 1.9: Effect of the modification of the cross section geometry on the second
moment of area of a beam. A comparison of the areas and second moment of areas
for the different cross sections is given in (1.7) and (1.8).
superelastic (i.e. reversible) deformations, if suitable forces are applied to the flanges.
Assuming k ≈ 1.07, an increase of the stiffness of the order of 15% can be obtained.
[8, 31] report about a method to modify the stiffness of a beam subject to bending by
modifying the geometry of the system, in order to create a tunable vibration absorber.
Figure 1.10 shows the stress-strain diagram of a suitably selected alloy. The energy
density needed for the superelastic deformation is approximately 17 MPa, based on
the area under the stress strain curve. The volume of the web is approximately 1.31·
106 mm3 (97 mm x 5.4 mm x 2500 mm). The total energy needed for the geometry
modification is given by the area under the stress-strain curve shown in Figure 1.10 is
then approximately 22 kJ. The obtained increase in bending stiffness is of the order of
14.5%.
25
400
300
200
100
0
0 1 2 3 4 5 6 7 8
Figure 1.10: Stress strain curve for a NiTi alloy adapted from [7], after [43]
Topology Modification
Considering the properties of multi-layer composite structures reveals immediately that

not only the overall geometry (shape) of a structure but also its topology has a substan-
tial influence on its elastic properties.
The starting point of an example to illustrate the effect of topology change shall be again
an I-beam. In its low stiffness configuration, the beam is divided in three layers (the two
flanges and the web), as shown in the left hand side of figure 1.11.
As long as no interaction between adjacent layers is allowed, the bending stiffness of
the described system can be written as the sum of the stiffness of the components:
b · d3 d · h3web b · d3
Ddisconnected = E · ( + + ) (1.9)
12 12 12
with b, d, h and hweb as shown in figure 1.11.

Assuming the proportions between the measurements of the beam to be the same as
26
d d
hweb hweb h
d d
b b
Adisconnected, Iz,disconnected connect Aconnected, Iz,connected
Figure 1.11: The modification of the topology of an I-beam causes a remarkable in-
crease of the stiffness.
for the one used in section 4.3:
d : b : h : hweb = 1 : 10 : 18 : 20 (1.10)
(1.9) becomes
5 5
Ddisconnected = E( d + 486d + d) = E · 487.7 · d4 (1.11)
6 6
Once the components are connected at the interfaces (highlighted in red in figure 1.11),
the contribution of the flanges becomes.
hweb 2
Df langes = E · d · b · ( ) = E · 1805d4 (1.12)
2
Thus the stiffness of the connected I-beam (on the righthand side of figure 1.11) is:
Dconnected = E · 2291d4 (1.13)
The ratio of the stiffness of the beam in the two states (disconnected and connected)
is:
Dconnected 2291
= ≈ 4.7 (1.14)
Ddisconnected 487.7
The difference between the two described cases is given by the ability of the interfaces
between layers of a laminated beam to transmit shear stresses from one layer to the
next.
27
The effect of topology on the stiffness of mechanical systems has already been re-
ported in [49]. Also, reports of an equivalent system to what previously described by
the authors can be found in [48], although the observed modification of the behavior of
the system is attributed entirely to dissipative friction processes, rather than the mod-
ification of the topology of the system, from a mechanical point of view. The results
reported in [3] indicate that the effects obtained by coupling layers of a beam exceed
the dissipative effects in [48].
The modification of the topology of a multi-layer system is achieved by creating or can-
celing the interfaces between layers of the system. The creation of new interfaces
requires energy, while the cancellation is an exothermal process. The order of magni-
tude of the energy needed for or released by the topology switching process equals, in
theory, the surface energy of the material (e.g. 1.5-2 Jm−2 for transition metals like Iron
or Titanium, [1]). The thickness d of the beam shown in figure 1.11 can be calculated
for the low stiffness state using (1.9). In the case of a steel beam (E = 210GP a) with
the proportions discussed in section 1.2.2 and the aforementioned stiffness of 71.3 ·
109 Nmm2 the thickness d will be 5.13 mm.
The energy needed or released for a connectivity change is then given by:
Econnectivity = 2 · A · esurf = 4 · 2.5m · 0.00513m · 1.5Jm−2 = 0.077J (1.15)
Where esurf is the surface energy of the material, A is the surface area of the interfaces
that are created or canceled.
Based on this estimate, only approximately 0.08 J would be needed to create new
surfaces and the same amount of energy would be released upon their cancellation.
A comparison with other methods for the modification of the mechanical properties
of structural elements, such as the ones outlined in this section shows that in theory
the topology switch promises a very high pay-off in terms of amplitude of the stiffness
variation in comparison with the energy needed to obtain it.
In reality, only upon cleavage of the material, new surfaces can be created at so lit-
tle energy expense. Similarly, the cancellation of surfaces is in reality a process that
requires a considerable amount of energy due to the need to provide the activation en-
ergy necessary to bond the surfaces. The energy necessary to bond metallic structural
elements is either provided in form of thermal energy (in the case of welding soldering
or brazing) or in form of chemical energy, if adhesive bonding is used. The separation
of the bonded layers is technically achieved by machining the structure or by other high
energy mechanical or thermal processes. These methods for the creation or cancella-
28
tion of interfaces are not compatible with the goal to adaptively modify the bending or
torsional stiffness of a structure.
1.2.3 Summary
The modification of the mechanical properties of the host structure offers an additional
degree of freedom in the design of adaptive systems. Structures with tunable me-
chanical properties can be used as such for vibration suppression applications or in
combination with actuators for enhanced shape control applications.
The previous sections have described possible ways to obtain a modification of the
bending stiffness of a simple structure.
Adaptive structures find often their application in systems where performance and weight
are of paramount importance. The energy demand to obtain a given effect (in this case
a modification of the stiffness) plays thus a central role in the assessment of the overall
effectiveness of a new system.
Table 1.2 gives an overview of the two main parameters that were calculated for the
presented examples: The increase in stiffness that can be achieved with each method
and the energy needed to achieve it, under ideal circumstances.
The parameters listed in the table clearly show the potential of topology switching as a
low energy approach to the modification of the bending of structural elements. Given its
sensitivity to the distribution of the cross-sectional area in space, the torsional stiffness
Table 1.2: Comparison of energy demand of the presented stiffening strategies
Method Obtained increase in stiffness Required energy [kJ]

Modification of the elastic 2.7 900
modulus of NiTi (section
1.2.1)
Modification of the height of 1.15 22
the web of a NiTi beam (sec-
tion 1.2.2)
Modification of the topology 4.7 80 * 10−6
of a steel beam (section
1.2.2)
29
of the structure can be modified according to the same principle, but is not discussed
here. The axial stiffness of structural elements cannot be modified using approaches
that leverage on a change of the area distribution, since this property only depends on
the cross-sectional area.
1.3 Goals and Structure of the Present Work
The primary goal of the present work is to present a method to realize the modification
of the bending stiffness of simple adaptive host structures based on the topology switch
approach.
The challenge of this work is to overcome the difficulties posed by the energy require-
ments set by the creation and cancellation of the interfaces within the system.
A method for the implementation of the stiffness modification equivalent to topology
switching will be presented in chapter 2.
In order to appreciate the limits and possibilities of the method devised for the modi-
fication of the bending stiffness, a quantitative understanding of the mechanisms that
govern it is necessary. The use of the analytical and numerical models presented in
chapter 3 allows for such understanding without the experimental difficulties posed by
the use of physical models.
The experiments described in chapter 4 will substantiate most of the aspects addressed
in the two previous, with special attention for the demonstration of the modification of the
vibrational properties of simple structures, and highlight the technical problems encoun-
tered in the implementation of the electrostatic modification of the bending stiffness.
An example of the application of multi-layer structures for vibration suppression will be
presented in chapter 5.
Finally, the meaning of the present work will be summarized and discussed in chapter
6.
30
Chapter 2
Working Principle of the Electrostatic

Tuning of the Stiffness
While true topology switching (i.e. the creation or cancellation of interfaces) cannot be
realized without the use of large amounts of energy per unit interface area, the mod-
ulation of the contact stress at the interfaces, and thus of the shear stress transfer by
means of friction, can yield an equivalent effect. In this chapter similarities between the
shear stress transfer at the atomic level and at the meso-level are outlined at first. Then,
the chapter shows how the application of electrostatic fields across the interfaces of a
multi-layer structure will be used to mimic the creation and cancellation of interfaces.
2.1 Connectivity, Interfaces and Shear Stress Transfer
The previous section outlined how a change in the connectivity (or topology) of a multi-
layer system can lead to a remarkably high change in mechanical properties for a nomi-
nally low energy input. In the context of this work, the term connectivity (of areas) refers
to the presence or absence of interfaces within a cross section. 1 In reality, the energy
needed to divide a structure in many layers or join the layers of a structure to a monolith
is much greater than the surface energy of the material the structure is made of. The
main obstacle to the use of topology switching for the modification of the mechanical
properties of a structural element is thus that there is no simple method to realize it
while fulfilling following requirements:
1
The connectedness of a space is a concept of topology. In this context the terms ’connectivity’ and
’topology’ of a section will be used interchangeably.
31
• Limited energy requirements
• Small weight penalty
• Short reaction time
These requirements are central to the application of adaptive systems in high perfor-
mance structures, as outlined in chapter 1.
In this chapter we will examine in more detail how the connectivity of a simple multi-
layer system influences its mechanical properties. An alternative approach to a true
topology switch for the modification of the properties of a structure will be presented.
2.2 Shear Stress Transfer and Stiffness
A closer look at the shear stress distribution in a multi-layer system shows that a change
in connectivity of the system essentially influences the transmission of shear stresses
at the interfaces between layers, as shown in figure 2.1.
Disconnected σxy (interface)=0

Adisconnected, Iz,disconnected
y σxy (y) P y
z
z x n·h x
n·h
L b
Connected
Aconnected, Iz,connected
y σxy (y) P y
z
z
x H=n·h x H=n·h
b
V L
Figure 2.1: Effect of connectivity switching on the shear-stress distribution in the cross-
section of a cantilever beam
32
In a multi-layer beam subjected to a bending load, as in the top drawing of figure 2.1,
the shear stress distribution is expected to have a zero value at the interfaces of the
layers of the beam i.e. at positions y = i · h with i = 1, 2, ...n, as detailed by the σxy
distribution. In the solid beam shown in the bottom figure, the σxy distribution reaches
a zero value only at y = ± n2 · h, i.e. at the outer faces. The difference between the two
distributions accounts for the different bending behaviors of the two systems. It should
be noticed that if subjected to a tensile stress in the x direction, the global behavior
of the two systems will be the same, independent of the number and position of the
interfaces in the system.
2.3 Shear Stress Transfer in Crystalline solids
The ability to transfer shear stresses across planes of a crystalline body is at the origin
of the shear stiffness of materials such as metals, as shown in figure 2.2.
In order to induce a shear deformation ²xy in a crystalline structure by applying a shear
stress σxy , the crystal planes have to be shifted along one another as shown in the
sequence (a) → (b) → (c) representing the idealized plastic deformation process2 by
one lattice constant d per plane. In order to do so, the distance d of the planes will be
increased up to d0 . The periodic potential shown in the graphic under (b) is given by
the electrostatic interaction between the atoms of the upper plane and the ones of the
lower plane. The system reaches its maximum energy level when the relative position
of neighboring planes is as shown in (b).
Small elastic shear deformations take place in proximity of the equilibrium state (a).
For these deformations, the stress is known to be approximately proportional to the
deformation, as indicated by the red line in the graphic under (b) [33]. Hence, the
stress-deformation relationship for small deformations is linear:
σxy = G · ²xy (2.1)
Once the distance between the two planes is sufficiently large (D, in (d)), due to the
creation of new surfaces within the material, the electrostatic interaction between the
atoms on originally contiguous planes becomes so small (as indicated by the horizontal
red line in the graphic under (e)) that no shear stress is needed in order to move the
planes past one another.
2
In reality the carriers of plastic deformation are dislocations that are activated at significantly lower loads
than would be needed for the homogeneous plastic shearing of the crystal.
33
Connected
σxy
d d’
σxy E
σxy I)
x
(a) (b) (c)

Disconnected
σxy σxy
D
E
σxy
II) x
(d) (e) (f )
Figure 2.2: The resistance to shear of crystalline solids is given by the energy difference
between the equilibrium state (a) and (c) and the state of maximum deviation of the
crystal lattice parameter d0 (b). Adapted from [33]
2.4 Creation and Cancellation of Interfaces
The amount of energy needed to create new surfaces is in theory very modest [1], as
shown in table 2.1. The reverse process, i.e. the joining of two surfaces is then an
exothermal process.
Table 2.1: Surface energy of some metals used in engineering applications [1]
Metal Atomic Number Surface Energy Surface Energy

Calculated, σc Measured, σexp
[J/m2 ] [J/m2 ]
Titanium 22 1.79 1.44
Chromium 24 1.93 1.59
Iron 26 2.22 -
Nickel 28 2.35 1.44
Copper 29 1.66 1.12
34
Nevertheless, it is known that in reality much higher amounts of energy are involved
in the switch from the connected to the disconnected state. Also, experience shows
that although the endothermic process in which surfaces are created is in principle
reversible, energy is needed both for dividing and for joining a body.
Observation of the geometric properties of real solid surfaces shown in figures 2.3
and 2.4 yields an insight into the reasons for the inapplicability of topology switching
to engineering grade surfaces. The presence of surface defects hinders the contact
between surface atoms [47] even on the smooth surfaces of cleaved single crystals.
Figure 2.3: At an atomic level, even the smoothest real surfaces present terraces, ad-
atoms, kinks, and other surface defects that make the real contact surface smaller than
shown in the idealization of figure 2.2. Adapted from [47]
Furthermore, almost every surface is known to be rough also on a larger scale than
the one shown in figure 2.3. This means that most parts of the surface are not flat but
form peaks or valleys. For engineering grade surfaces, the typical amplitude between
peaks and valleys is of the order of about 1µm. Under normal conditions, the profile of
the surface is essentially random, irrespective of its source and scale of size, as shown
in figure 2.4. Hence, when two solids contact each other, the actual contact surface is
substantially smaller than the apparent contact surface [47]. For common engineering
materials, the cancellation of an interface cannot be achieved by contact between the
surfaces.
The processes involved in the creation of new surfaces are generally more complex
than the mere cleavage of a single crystal, as they generally involve plastic deformation
of the material. This leads to a much higher energetic expenditure than needed just for
the creation of new surfaces. Generally speaking, a high fracture energy is a desirable
property of engineering materials, since the integrity of structures depends on it.
For the reasons outlined in this section, topology switching cannot be considered for
the stiffness tuning of engineering structures in real applications.
35
z
Average plane x
1mm of a roller-bearing race

z
Average plane x
30 mm of a ball-bearing raceway
z
Average plane x
110 mm of a lathe bed
Figure 2.4: Real surfaces, as the ones found in engineering structures are rough.
The surface profile is generally random and shows features at different scales of size.
Adapted from [47]
2.5 Friction at Real Interfaces and Shear Stress Trans-

fer
The contact interaction between solids is substantially different from the interaction be-
tween crystal planes moving relative to each other. The geometry of the surfaces that
are in contact is random, thus no periodic potential governs the relative displacement
between the surfaces. The actual contact area is generally so small that the electro-
static interactions between atoms of the two surfaces are not sufficient to make any
significant contribution to the transfer of shear stresses across the interface.
The contact between solid surfaces has been studied intensively at different levels of
detail, ranging from empirical observation to modeling of the processes at contact sur-
faces.
From a phenomenological point of view, it is known that when a normal force N is
applied to two solid surfaces that are in contact (see figure 2.5), friction between the
surfaces is experienced. Friction is the force Ff that is opposed to the relative motion
of the surfaces in the direction imposed by an external force F .
Four basic empirical laws of friction have been know for a long time [47]:
1. There is a proportionality between the maximum tangential force before sliding

and the normal force when a static body is subjected to increasing tangential load
36
F
N
Ff =µN
N
F
Figure 2.5: A simple representation of the friction between solid surfaces
F.
2. The tangential friction force is proportional to the normal force in sliding
3. Friction force is independent of the apparent contact area
4. Friction is independent of the sliding speed
If the forces shown in figure 2.5 are averaged over the apparent area of contact, points
1. and 2. can be written as:
f
σxy = µ · σyy (2.2)
f
Where σxy is the average shear (friction) stress at the contact surfaces and σyy is the
average normal stress at the contact surfaces.
The development of structures with tunable bending stiffness proposed in this work is
based on the assumption that for the purpose of calculating the bending stiffness of a
structural element, the connectivity of the element can be stated in terms of the ability
to transfer shear stresses across its cross-section independently of the mechanisms
involved in the process.
In this work, the shear stress transfer functionality that in a solid body is guaranteed
within the limits set by the strength of the material is replaced at selected locations
37
by shear stress transfer through friction between solid surfaces. Up to the limit set by
the static friction between the surfaces, the shear stress in the cross section of the
component (e.g. a beam) subjected to bending forces can be transferred in full. The
governing parameters are the normal stress and the friction coefficient at the interface.
The application of normal forces at the interfaces of a multi-layer structure can be
achieved in various ways and can be modulated at a fairly high speed. The first two
requirements stated at the beginning of this chapter (low weight penalty and limited
energy consumption) need to be addressed by a suitable choice of the method used to
apply the needed force, as will be described in section 2.6.
2.6 Electrostatic control of the Shear Stress Transfer
For the purpose of the adaptive modification of the bending stiffness, we have seen
that a sufficiently large normal stress needs to be applied at the interfaces between
the layers of a multi-layer structure, in order to enable shear stress transfer at those
locations.
The application of an electrical field between the contact surfaces of adjacent layers is a
practical way to generate the normal stresses needed for this purpose. In order to apply
an electrical field, the two interfaces have to be electrically insulated by a dielectric layer,
as shown in figure 2.6.
Load
Bearing
1.5x106 el
Element
σyy
Breakdown field 1x10 6 Electrode
σyy [Pa]
5x105 Ui Dielectric
2x10-5 thickness=δ
0 Electrode
el
4x10-5 1000 σyy Load
δ[m 6x10-5 Bearing
] 2000
V] Element
8x10-5 3000 U i[
Figure 2.6: The application of an electrostatic field between surfaces of the layers of
a multi-layer structure generates normal stress at the interface. Here, the layers are
assumed to be non-conductive, hence an electrode bonded to the layer is necessary at
the interface.
38
The use of strong electrostatic fields to generate normal stresses across thin dielec-
tric layer is known in the field of Dielectric Elastomer Actuators (DEA). One difference
between the system presented in this work and DEAs is that the dielectrics in DEAs
are highly compliant in order to allow for the deformation of the system that leads to
the actuation process. The electrodes have thus also to be compliant in order to fulfill
compatibility requirements with the deformation of the dielectric. [37, 38]
el
The normal stress σyy generated by the electrostatic potential U across the dielectric is:
el ²0 ²r U 2
σyy = (2.3)
2δ 2
Where ²0 is the permittivity of vacuum, ²r is the dielectric constant of the dielectric layer,
U is the applied potential, and δ is the thickness of the dielectric layer. (2.3) shows that
the stress increases proportionally to 1/2δ 2 , which indicates that thin dielectric layers
are highly desirable.
When multiple dielectric materials fill the space between the electrodes of a capacitor,
the system can be described as a series of capacitances, each with the thickness δi and
the dielectric constant ²ir . The effective dielectric constant of the system is calculated
via the calculation of the capacitance of the system:
1 X 1
=
Ctot i
Ci
1 X δi
=
Ctot i
²0 ²ir A
(2.4)
From (2.4) and the total thickness of the space between the electrodes, the equivalent
dielectric constant ²r of the system can be obtained:
P
i δi
P δi
i ² ²i A
²r = 0 r
P²0 A
δi
²r = P i δi (2.5)
i ²ir
The stress-potential relationship in a system with multiple dielectric layers can then be
calculated substituting ²r from (2.5) into (2.3).
39
The use of very thin dielectric layers is limited by the breakdown field Eb , a physical
property of the dielectric material indicating the maximum field strength above which
the insulating properties of the material are lost. For a material with a breakdown field
Eb the minimum thickness of the dielectric layer is given by:
U
δmin (U ) = (2.6)
Eb
Where U is the potential across between the electrodes.
Together with the dielectric constant, Eb determines the maximum normal stress that
can be generated across the insulator:
el ²0 ²r U 2 ²0 ²r Eb2
σyy,max = 2
= (2.7)
2δmin 2
It is noteworthy that (2.7) has the same form as the equation for the specific energy
stored in the electrical field of a capacitor. Using the physical properties of a dielectric
material commonly utilized for the production of capacitors , such as Al2 O3 , ²r ≈ 10 and
Eb ≈ 200 − 500kV /mm the maximum normal stress will be of the order of 2-11 MPa.
Assuming a fairly modest coefficient of friction µ of 0.2, the maximum stresses that can
be transferred through friction at the interface will be of the order of 0.4-2.2 MPa. This
compares to approximately 30 MPa shear strength for Aluminum or approximately 190
MPa for common stainless steel.
The energy needed to generate the estimated normal and friction stresses in the dielec-
tric material between load bearing layers is of the order of 2-11 mJ/mm3 . Thus, from
an energetic standpoint the use of electrostatic fields as a means to adaptively couple
the layers of a multi-layer structure is an appealing approach. In the following chapters,
numerical models and experiments will be used to show the potential and the limits of
this type of tunable stiffness structures.
40
Chapter 3
Modeling of Structures with Tunable

Stiffness
This chapter will deal with two aspects of the modeling of structures with tunable stiff-
ness. In the first part, the effect of the interfaces of a multi-layer system on the shear
stress distribution will be considered for bending loads. The system considered is a sim-
ple, homogeneous four-layer beam in which the shear stresses at all three interfaces
are transferred by means of friction. The results obtained are especially interesting
to show how friction-based shear stress transfer compares with shear stress transfer
within a solid material. The calculations performed with a numerical model will show
that once the maximum transferable shear stress at the interfaces is reached, the sys-
tem softens considerably and then behaves linearly again. In this second part of the
force displacement curve, the layers slip on one another. This is at the origin of a hys-
teretic behavior shown in the last section devoted to the local behavior of the system.
The second part of this chapter presents the global behavior of the system within the
elastic domain, i.e. before interface slipping starts occurring in a significant way. The
effect of the virtual cancellation or creation of interfaces on the stiffness of a homoge-
neous multi-layer beam is shown: As a result of different interface activation patterns,
the stiffness can be changed in discrete increments. The number of different stiffness
values that can be realized increases with the number of layers in the system.
41
3.1 Local Behavior: Shear Stress Transfer by Means of
Friction vs. Shear Stress Transfer in Solids
In this section, we present a numerical model of a homogeneous tunable multi-layer

beam that considers contact and friction to describe the behavior of the studied struc-
tures. We will consider beams with a homogeneous cross-section, in which all layers
are made of the same material and have the same width. The layers are the smallest
unit of the system. They can be connected to a bundle of multiple layers by removing
interfaces between them. The thickness of bundles is always a whole multiple of the
thickness of a layer. This choice makes the modeling work more clear and the results
more readily understood. The model is compared with analytical solutions for an equiv-
alent system in which the connectivity of the layers is switched. In the analytical model,
the layers are regarded as independent bodies that are subjected to loads such, that
their deflection lines are identical.
The obtained model is used to better understand the processes taking place at the
interfaces between layers, demonstrate the existence of discrete stiffness states and
as a base for the selection of suitable dielectric materials for the generation of the
electrostatic normal stresses needed for the shear stress transfer at the interface.
h y n.a. 1 y
h n.a. 2
x z 4h n.a.
h n.a. 3 x z
h n.a. 4
3
I~4h I~64h3
yi
x z n.a. i
Figure 3.1: Two cross sections with the same geometry and different topologies. The
topologically connected cross section (right) has one neutral axis (n.a.), while the dis-
connected one (left) has four (n.a.1 ...n.a.4 ). On the bottom left the local coordinates for
layer i are shown.
3.1.1 Deflection and Shear Stress Distribution in a Multi-layer Beam
For the analytical description of the bending behavior of a multi-layer beam with rect-
angular cross section of total height N · h and width b, consisting of k bundles of height
42
ni · h, following assumptions are made:
• The layers and the complete beam are modeled as Euler-Bernoulli beams, i.e.
no shear deformation is considered. This is justified by the slenderness of the
elements and the assumption that beams of an isotropic materials are considered.
Shear compliant beams will be considered in chapter 4.
• The bending stiffness Di of the ith layer is given by:

E·b
Di = E · Iz,i = · (ni · h)3 (3.1)
12
• The total stiffness of the beam is given by:
k
E·b X
Dtot = E · Iz,tot = · (ni · h)3 (3.2)
12 i=1
• The transversal displacement v(x) of the system is:
P
v(x) = · f (x) (3.3)
Dtot
Where f (x) is a function, typically a polynomial, that describes the shape of the
deflected beam and P is a load scaling factor.
• The transversal displacement vi (x) is the same for each layer:
v1 (x) = ... = vi (x) = ... = vk (x) (3.4)
From the previous assumptions, Pi , the load scaling factor used in the calculation of the
loads acting on each layer, can be obtained:
Di n3
Pi = P · = P · Pk i · (3.5)
Dtot i=1 ni
3
The system bends like an array of k parallel beams, each subject to a load proportional
to the cube of their height, where the sum of the loads acting on each layer corresponds
to the load applied to the component.
Using the shear force function Qi :
Qi (x) = Di · v 000 (x) = Pi · f 000 (x) (3.6)
43
The shear stresses in each layer become then:
Qi (x) · (( n2i h )2 − yi2 ) 6P · f 000 (x) · (( n2i h )2 − yi2 )

σxy,i (x, yi ) = = P (3.7)
2Iz,i bh3 ki=1 ni 3
yi is a local coordinate in each layer representing the distance in y-direction from the
neutral axis of each layer of k layers. The maximum shear stresses in each layer of the
multi-layer beam (σxy (x, yi = 0)) is given by:
3 n2i · f 000 (x)

σxy,i (x, yi = 0) = P P (3.8)
2 b · h · ki=1 ni 3
If all bundles in the system have the same thickness of n layers, then the sum in the
denominator of (3.8) can be written as s · n3 , where s is an integer thus yielding:
3 f 000 (x)
σxy,i (x, yi = 0) = P (3.9)
2 b·h·s·n
s · n · h is the total height H of the multi-layer beam. For a given load pattern, the
maximum shear stress in a multi-layer beam in which all bundles have the same height,
is independent of the height and number of the individual bundles.
The relations presented in this section are valid for a homogeneous multi-layer beam
composed of n layers of thickness h that interact only by transferring transversal forces
by contact. No other interaction (e.g. friction at the contact surface) is considered.
3.1.2 Numerical model of the friction based shear stress transfer

at interfaces
The analytical model presented in the previous section describes how the system be-
haves when full or no shear stress transfer is provided at the interfaces between layers.
A comparison between the electrostatically generated normal stresses (and the asso-
ciated maximum transferable shear stresses) and the shear stresses at the position of
the interfaces in a solid beam gives an indication whether the shear stress transfer de-
mand at the interface between layers can be satisfied or not. In this section, a numerical
model of the system is used to model the behavior of the system for the cases, where
the shear stress exceeds the maximum friction stress transfer capability σxy,max made
available through the application of normal stresses at the interface.
44
l
q
y
z x
Fixed boundaries
Distributed external load q
Interfacial normal stresses σyy
Figure 3.2: Model considered for the numerical calculation of the effect of shear stress
transfer control
For the purpose of investigating the bending behavior of a multi-layer system, a four
layer beam was modelled using the COMSOL 3.4 Multiphysics modeling package. The
modeled system shown in figure 3.2 had the properties listed in table 3.1.
Table 3.1: Main properties of the modeled system
Property Value
Number of layers 4
Material steel (E=205 GPa)
Friction coefficient µ between surfaces 0.2
Layer dimensions lxbxh [mm] 70x10x1
Load q distributed: 100N m−1 ...500N m−1
(i.e. 10 kPa...50 kPa) on the top face
Boundary conditions fixed on the left side of each layer,
free otherwise
Additionally, a positive normal stress σyy,el of 0 MPa or 3 MPa was applied to the contact
surface pairs to couple them in a way to simulate the effect of the electrostatic normal
stresses described above. A normal stress of 3 MPa corresponds to an electric potential
of approximately 75 V over a 400 nm thick Al2 O3 layer, i.e. a field of approximately 185
MV/m.
The system has 3 pairs of contact surfaces. Each of the pairs can be activated (σyy =
45
3MPa) or inactivated (σyy =0 MPa). The possible configurations of active and inactive
interfaces are summarized in table 3.2, where a ’0’ in the left hand column means that
the interface is not activated (i.e. no stress is applied to it) and a ’1’ means it is. In the
second column, the thickness of each bundle, expressed in number of layers.
The contact forces were modeled based on a penalty barrier approach [40]:
(
Tn − b n · g if g≤0
σyy,contact = − bTn g
(3.10)
Tn · e n otherwise
Where σyy,contact is the calculated contact pressure, g is the distance between the two
surfaces, and Tn and bn are, respectively, the initial contact pressure and the barrier
parameter with following values:
• Tn = 1M P a
• bn ≈ 1014 P a/m
The stress described by the penalty barrier function is represented graphically in figure
3.3
10
10
5
10
0
10
Contact stress σyy,contact [Pa]
−5
10
−10
10
−15
10
−20
10
Contact surface
−25
10
−30
10
−35
10
−40
10
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
gap size g [m] x 10
−6
Figure 3.3: Contact stress σyy,contact as a function of the gap size as calculated using
the penalty barrier approach described in (3.10)
The resulting contact stress was used to calculate the friction stresses. As for the
46
analytical model, no cohesion (i.e. friction at zero normal stress) was assumed for this
model.
Table 3.2 shows the convention used to describe the connection pattern of the inter-
faces of a 4 layer beam.
Table 3.2: Interface activation configurations and bundle thicknesses ni
Configuration ni
000 1,1,1,1
001,010,100 1,1,2 1,2,1 or 2,1,1
011,110 1,3 or 3,1
101 2,2
111 4
For the ’111’ beam (i.e. with fully connected layers), the behavior of the system was
calculated for loads increasing from q = 1N/m to q = 500N/m in several increments, in
order to investigate the slip behavior at the interface between layers.
3.1.3 Linear vs. Non-linear Behavior
Better understanding of the behavior of tunable multi-layer structures is obtained by

observing the load displacement diagram for a beam in ’111’ configuration. The dia-
grams shown in figure 3.4 were obtained using the analytical model (solid line) and the
numerical model presented in section 3.1.2 (markers).
The analytical model only describes the linear behavior of the system, since no slip at
the interfaces between layers is allowed for. Its validity is limited to the load ranges
where the shear stresses at the interfaces between layers of the beam are smaller than
the maximum transferable shear stress, as defined in (3.8). The main advantage of the
analytical model is that it gives a better understanding of the system and that it can
be easily adapted to different geometries and load cases. The described numerical
model accounts for slipping at the interfaces, as will be shown. This is a considerable
advantage, compared to the analytical model. The disadvantage is that new instances
of the model have to be set up, each time a new system is considered. The analytical
model is used to verify the reliability of the output of the numerical model in the linear
domain of the system.
47
For the analytical calculation of the deformation function v(x) = P · f (x) in the load case
described in the previous sections, the deformed shape of the beam is given by:
f (x) = (x4 − 4lx3 + 6l2 x2 ) (3.11)
The load scaling factor P is q/24.
−4
x 10
0
linear
−0.2 non−linear
non−linear w/ offset correction
−0.4
Displacement v(l) [m]
−0.6
−0.8
−1
−1.2
−1.4
−1.6
−1.8
−2
0 100 200 300 400 500
load q [N/m ]
Figure 3.4: Transversal displacement at x=l, as calculated using the linear (analytical)
model and the non-linear (numerical) model, as a function of the applied load q
The displacement calculated using the numerical model of the friction based system(’+’
symbol in figure 3.4) shows an offset compared to the linear calculation. This is inter-
preted as an effect of the singularity of the transversal load at the fixation point of the
cantilever at x=0 that leads to very high shear stresses in that region. If this offset is
subtracted from the results (’x’ symbol in figure 3.4), a very good agreement between
numerical and analytical model is found until values of q of approximately 300 N/m. For
loads exceeding this value, the deviation between results obtained from analytical and
the numerical behavior increases markedly.
48
The shear stress distributions σxy (x, y, q) calculated using the linear and the non-linear
models are shown in figures 3.5 to 3.9. A comparison of the distributions for different
levels of q confirms that the onset of the non-linear behavior of the beam coincides
with a marked deviation of the shear stress distribution in the numerical calculation
from the linear model at q ≈ 320N/m. The numerical calculations show that the shear
stress in the beam never exceed the value µ · σyy,int at the interfaces, even for values
of q exceeding 320N/m. As the load increases, the portion of each interface where
the actual shear stress deviates from the shear stress calculated with the linear model
increases, as seen comparing figures 3.8 and 3.9.
5
x 10
0 y
x=0.05m
−2 x
−4
σ [Pa]
−6
xy
x=0.01m σ [Pa]
xy,max
−8
x=0.01m
x=0.02m
−10 x=0.03m
x=0.04m
x=0.05m
−12
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
position on the y−axis [m] x 10
−3
Figure 3.5: Shear stress distribution in the cross section at different positions along the
length of the beam with a 111 interface activation pattern, for a load q=280 N/m. The
numerically calculated distribution (markers) at x=0.01 m shows a slight deviation from
the linear model (solid lines).
In figure 3.9 can also be noticed how the shear stress obtained from numerical calcu-
lations at x = 0.05m exceeds the one obtained from the analytical model calculations.
This is explained as a result of the redistribution of shear stresses along the interface
between the two central layers: The shear stress transfer demand near the fixation
point exceeds the stresses that can be transferred by the friction at the interface. Such
stresses are redistributed to locations at the interface where the demand is below the
critical value σxy,max = µ · σyy,el .
Figure 3.4 clearly shows that once the load q exceeds 320 N/m, the stiffness of the
system no longer is a constant. A comparison of the shear stress distributions shown
in figure 3.6 and 3.7 shows that the transition from full shear stress transfer to slipping
49
5
x 10
0 y
−2 x=0.05m x
−4
σ [Pa]
−6
xy
σ [Pa]
xy,max
x=0.01m
−8
x=0.01m
x=0.02m
−10 x=0.03m
x=0.04m
x=0.05m
−12
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
−3
numerically calculated distribution at x=0.01 m (markers) shows a minor deviation from
the linear model (solid lines).
5
x 10
0 y
−2 x=0.05m x
−4
σ [Pa]
−6
xy
σ [Pa]
xy,max
−8
x=0.01m
x=0.01m
x=0.02m
−10 x=0.03m
x=0.04m
x=0.05m
−12
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
−3
numerically calculated distributions at both x=0.01 m and x=0.02 m (markers) show a
clear deviation from the linear model (solid lines).
50
5
x 10
0 y
−2 x
x=0.05m
σ [Pa] −4
−6
xy
σxy,max [Pa]
−8
x=0.01m
x=0.02m x=0.01m
−10 x=0.03m
x=0.04m
x=0.05m
−12
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
−3
numerically calculated distributions at x=0.01 m and x=0.02 m (markers) show a re-
markable deviation from the linear model (solid lines). Also at x=0.04 m the distribution
does not correspond to the linear behavior .
5
x 10
0 y
−2 x
x=0.05m
−4
σ [Pa]
−6
xy
σxy,max [Pa]
−8
x=0.01m
x=0.02m
−10 x=0.03m
x=0.04m
x=0.05m x=0.01m
−12
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
−3
Figure 3.9: Shear stress distribution in the cross section at different positions along
the length of the beam with a 111 interface activation pattern, for a load q=500 N/m.
None of the shear stress distributions calculated numerically (markers) coincides with
the linear model (solid lines), although the results for x=0.04 m and x=0.05 m do not
exceed σxy,max .
51
is at the origin of the softening of the system. While this transition is not desirable for
a system that is required to work in an elastic manner, the ability to dissipate energy
through friction may be of interest for damping applications.
In order to show the hysteretic behavior of multi-layer systems when subjected to loads
exceeding its elastic limit, a cyclic force-displacement diagram for stresses larger than
shown in figures 3.5 to 3.9 was calculated using the numerical model and is shown in
figure 3.10.
2000
σ =0MPa σyy =6MPa
yy
1500 σyy =3MPa σ =9MPa
yy
1000
500
load q [N/m]
−500
−1000
−1500
−2000
−3 −2 −1 0 1 2 3
Displacement at x=0.07m [m] x 10
−3
Figure 3.10: The cyclic load-displacement diagrams calculated for three different levels
of interfacial stress (3 MPa, 6 MPa and 9 MPa) show different amounts of hysteresis in
the system. The load displacement diagram of the un-coupled system is drawn in black,
for comparison. The uncoupled system does not have hysteresis, since no interaction
between layers is allowed for.
As expected, the higher the interfacial stress, the higher is the average stiffness of the
system. Nevertheless it should be noted that the stiffness of the system in the first
section of the diagram (the new curve starting at the origin of the coordinates system)
is the same in all three cases. The onset of the softening is found at increasingly higher
loads as the interfacial stress increases, since the higher shear transfer rate at the
52
interfaces allows for an extended linear domain.
An additional effect of the modification of the interfacial stress on the behavior of the
change in the area contained within the hysteresis curves calculated for different levels
of interfacial stress. As the average stiffness of the system increases, the deformation at
the maximum load (q =1600 N/m) is substantially reduced, thus the ability to dissipate
energy by relative motion of the layers is limited.
3.2 Global Behavior: Discrete Stiffness States in Ho-

mogeneous Multi-Layer Beams
Section 3.1.3 showed that the amount of shear stress at the interfaces determines
the linearity of the behavior of the system. The activation of the interfaces (i.e. the
application of an electrical potential between two neighboring layers) determines the
amount of shear stress that can be transferred at the interface. In this section the effect
of different activation patterns on the stiffness of the system is presented, under the
assumption that the shear stresses do not exceed the maximum transferable stress
σxy,max = µ · σyy,el
Figures 3.11 to 3.16 show the transversal deformations and the shear stress distri-
butions in beams with different interface activation patterns in the linear domain. As
expected, with an increasing number of activated interfaces, the deformations of the
beam decrease. A quantitative description of the decrease in deformation as a function
of the interface activation pattern is contained in the ratio of the stiffness of the acti-
vated beam to the stiffness of the completely inactive beam (D/D000 ), shown in table
3.3. The indices used in the table and in the rest of this section indicate whether at an
interface shear stress transfer is permitted (’1’) or not (’0’). So, for example the index
’111’ indicates that all three interfaces of the four layer system are locked, whereas the
index ’100’) indicates that only one is locked. More examples are shown in figure 3.17.
For the numerical model results shown in table 3.3, the stiffness ratios are calculated
based on the displacement of the tip of the beams.
As expected, the shear stresses tend to zero in proximity of the non-active interfaces,
since no stress transfer is possible there. At the active interfaces, the shear stress is
given by the distribution calculated over a beam of the corresponding height, as if no
interface were present. Once the maximum transferable shear stress is reached at the
interface, the parabolic function described in (3.7) no longer applies.
53
−4
x 10
1
−1
deflection v(x) [m]
−2
−3
configuration:000
−4 configuration:010
configuration:101
configuration:011
configuration:111
−5
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
position on the x−axis [m]
Figure 3.11: Transversal deformation functions for the beams with different interface ac-
tivation patterns and a load q=100 N/m, as calculated analytically (line) and numerically
(markers)
5
x 10
0.5 y
0 x
-0.5
-1
-1.5
s [Pa]
xy
-2
-2.5
-3
x=0.01m
-3.5 x=0.03m
x=0.05m
x=0.07m
-4
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
position on the y-axis [m] x 10
-3
Figure 3.12: Shear stress distribution in the cross section at x=0.01 m,... 0.07 m for
the beams with ’000’ interface activation pattern and a load q=100 N/m, as calculated
analytically (line) and numerically (markers)
A comparison between the numerical and the analytical results shows a very good level
of agreement between the two models for the shear stress distribution as well as for the
transversal deformation, within the limit given by (2.7).
54
5
x 10
0.5 y
0 x
-0.5
-1
-1.5
s [Pa]
xy
-2
-2.5
-3
x=0.01m
-3.5 x=0.03m
x=0.05m
x=0.07m
-4
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
-3
5
x 10
0.5 y
0 x
-0.5
-1
-1.5
sxy [Pa]
-2
-2.5
-3
x=0.01m
-3.5 x=0.03m
x=0.05m
x=0.07m
-4
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
-3
The values for Dtot , Di , σxy,i (x, y) can be calculated using (3.1)...(3.7) under the as-
sumptions listed in section 3.1.1. The values for the maximum shear stress σxy,i (x, yi =
0) are calculated using f 000 (x) = 24(x − l).
As long as no slipping at the interfaces takes place, that is as long as the shear stress
55
5
x 10
0.5 y
0 x
-0.5
-1
-1.5
s [Pa]
xy
-2
-2.5
-3
x=0.01m
-3.5 x=0.03m
x=0.05m
x=0.07m
-4
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
-3
5
x 10
0.5 y
0 x
-0.5
-1
-1.5
s [Pa]
xy
-2
-2.5
-3
x=0.01m
-3.5 x=0.03m
x=0.05m
x=0.07m
-4
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
-3
transfer demand is satisfied by means of the friction at the interfaces, the stiffness of
the system is varied in a discrete manner, as shown in table 3.3.
According to the analytical model calculations, when the system is divided in bundles
of equal thickness, the maximum shear stress in each layer stays constant, as shown
56
by figures 3.12, 3.14 and 3.16. An uneven distribution of the bundle thickness across
the beam leads to an increase in the maximum value of σxy , as shown in figures 3.13
and 3.15.
Table 3.3: Interface activation configurations, stiffness increase compared to the 000
configuration, comparison between analytical and numerical results for q=100 N/m
Configuration D/D000 analytical D/D000 numerical

000 1 1
001, 100, 010 2.5 2.5
101 4 3.9
011, 110 7 6.7
111 16 14.3
Within the limits set by the ability of the interfaces to transfer the shear stress, each
bundle of layers that are coupled electrostatically can be considered as a solid beam of
thickness ni · h where ni is the number of layers in the bundle and h is the height of a
layer.
P P
y h k+k+ y
h h k+8k
x
k+k 2h +k
z h z x
h h
D000=4k D010=10k
P P
P
P
y y
8k+8k k+27k
2h 3h
x z x h
z 2h
D101=16k P D110=28k
P
Fixed boundaries P
y 64k
x 4h
z
D111=64k P
Figure 3.17: The stiffness of the beams with each connectivity pattern can be repre-
sented by an equivalent parallel connection of springs The stiffness D of each beam is
expressed in multiples of k, where k is the stiffness of one layer.
A total of 8 connectivity patterns can be obtained for the 4 layers beam in Figure 3.17.
The figure shows that only 5 different stiffness values can be realized. Patterns ’100’,
57
’010’ and ’001’ can be considered equivalent. Patterns ’110’ and ’011’ also correspond
to the same stiffness value.
If the shear stress transfer capability of each interface can be switched between 0 and 1,
the number N of different connectivity patterns that can be obtained for a given number
of layers n in the system is:
N = 2n−1 (3.12)
For a system with 4 layers, as the one shown in figure 3.17 there are 23 possible con-
nectivity combinations. The different stiffness levels for a 4 layers system is shown in
figure 3.17. The number of different stiffness values that can be achieved is smaller
than the number of connectivity patterns given by (3.12).
As outlined previously, the stiffness D of a multi-layer system can be varied in discrete
steps by modifying its connectivity pattern. The knowledge of the stiffness levels that
can be realized with a given multi-layer system is useful for the design of variable stiff-
ness devices. As shown in Figure 3.17, the achievable stiffness values can be calcu-
lated considering each bundle of layers as a spring connected in parallel with the other
bundles in the system. A bundle is a set of connected layers between two interfaces
with no shear stress transfer capability. The stiffness of each bundle is proportional to
its height cubed. The total stiffness is thus proportional the sum of the third power of
the bundle heights.
These assumptions are realistic for small transversal loads, for which the shear stress
at the interface, does not exceed σxy,max . Under such conditions, it is justified to assume
that a bundle of electrostatically coupled layers behaves as a solid bundle, as outlined
in the previous section.
The computational effort needed to calculate the distribution of stiffness values in-
creases with 2n .
Figure 3.18 shows the stiffness values that can be realized with a 20 layers system and
the frequency of each stiffness value. The diagram on the left hand side of the figure
shows the stiffness calculated for each of the 524288 (219 ) connectivity patterns that can
be realized in a system with 20 layers. If the state of the interface is represented with
’0’ for a non connecting interface (no stress transfer) and ’1’ for a connecting interface,
each connectivity pattern can be identified by a 19 bit binary number. The pattern
number on the x-axis of the diagram is the decimal equivalent of said binary number.
The diagram on the right hand side of Figure 3.18 shows how many patterns in a 20
layer system realize a particular stiffness value. While the points in this may intuitively
be interpolated to a curve somewhat resembling a normal distribution, it should be
58
104
103
Stiffness D[-]
102
101
x 104
1.8
1.2
1.4
1.6
0.4
0.8
2.0
0.2
0.6
1.0
0
x 105
Pattern Number [ ]
Number of patterns [-]
Figure 3.18: Stiffness distribution (left) and stiffness density (right) for n=20
noted that the number of patterns with stiffness values in between points drawn in the
graph is effectively zero. For the high and the low stiffness values, gaps are present in
which no connectivity pattern realizes certain values. For the high stiffness domain the
gaps span over very large number of values that are not represented in the set stiffness
states that can be realized by the system.
Figure 3.19 shows the stiffness value distributions of systems with 5, 10, 15, 20 and
25 layers, respectively. In order to make a comparison between the distributions for
different n-values possible, the number of patterns possible a double logarithmic repre-
sentation was chosen. The distributions show that the range over which the stiffness
can be modified increases very rapidly with n. From a 25-fold increase in stiffness for
n=5 the range grows to a 625-fold increase in stiffness for n=25. Also the number of
possible states increases remarkably (approximately as n3 /30). The high and low ends
of the distribution are more sparsely populated, both in terms number of patterns (y-
axis of the graph) and in terms of distance between stiffness states that can be realized
(x-axis of the graph).
59
6
10
n=5
n=10
5
10 n=15
n=20
n=25
Number of Patterns [−]
4
10
3
10
2
10
1
10
0
10 0 1 2 3 4 5
10 10 10 10 10 10
Stiffness D[−]
Figure 3.19: Stiffness states distributions for n=5, 10, 15, 20 and 25.
3.3 Conclusions
The goal of this section was to present simple analytical and numerical models to de-
scribe and understand various aspects of the behavior of tunable multi-layer beams.
The models provide useful information concerning the distribution of stresses in the
structure. The understanding of the local behavior of the system can be used to es-
timate the limits within which such a structure behaves approximately like an elastic
structure, without yielding due to slipping at the interfaces between layers. This can
be of special interest for applications in which the suppression of large amplitude vibra-
tions due to resonance phenomena is to be obtained by modifying the bending stiffness
and hence the natural frequencies of the system. In such cases, the primary goal of
the adaptation of the properties of a structure is to control its mechanical impedance,
rather than introducing high damping. Conversely, for other applications, such as the
harvesting of energy from structural vibrations, resonance effects are desirable, while
dissipation due to friction at the interfaces is not.
The behavior of the system is hysteretic when the loads it is subjected to exceed the
60
ability of the interfaces to transfer the shear stresses in full. The numerical model shows
that next to modifying the stiffness of the system, increasing the transfer of shear stress
at the interfaces also has an effect on the area enclosed by the hysteretic curve and thus
on the ability of the system to dissipate energy. This means that next to a modification
of the stiffness, also a tuning of the damping behavior is possible in multi-layer systems.
The models also give some insight into the effect of the homogeneity of the thickness of
the bundles of coupled layers on the value of the maximum shear stress in each cross
section of the system. This knowledge is valuable for the design of multi-layer systems.
The shear stresses that can be transferred at the interface between layers are limited by
the maximum energy density that can be stored in the used dielectric material. Hence,
interface activation patterns that lead to higher shear stress values for a given applied
load, are more likely to cause the interfaces to yield than others.
Both, the analytical and numerical models confirm that a finite number of discrete stiff-
ness levels can be achieved for each system. The number of stiffness levels increases
with the number of layers in the system. This prediction does not account for dissipative
effects such as inter-layer friction, that would lead to the coalescence of close states in
systems with a large number of layers. The softening predicted by the non-linear cal-
culations is expected to reduce the quality factor of the system and lead to a reduced
tuning effect for the natural frequencies.
The calculations of section 2.6 show that using commonly available dielectric materials,
such as alumina, titania and zirconia, the generated stresses are of a sufficient magni-
tude to effectively stiffen even in steel structures. Nevertheless the achievable interface
stresses are far below the shear strength of typical engineering materials. Hence, sim-
ple models as the ones presented here, can be useful tools to help users make suitable
design choices to obtain useful effects based on the electrostatic tuning of their bending
stiffness.
61
Chapter 4
Experimental Work
This chapter presents the experimental work performed on structures with variable stiff-
ness and variable damping. The examples chosen to demonstrate the effect of the
electrostatic coupling of elements of a structure are two sandwich structures. The first
structure considered in this chapter is a sandwich beam consisting of a highly compli-
ant core and stiff faces. This structure has mainly a proof of concept value, and is used
to show the similarity between the effect of lamination and electrostatic coupling on
the stiffness of the beam. The second structure investigated in this chapter is a sand-
wich beam made of a GFRP I-beam as the core of the sandwich and two unidirectional
CFRP bands as faces. The structure is used to demonstrate the effect of electrostatic
coupling on the eigenfrequency and the damping ratio of a sandwich beam. Except for
its size, this system is the same as the components that will be considered for the full
scale demonstrator described in the last chapter of this work.
In the models presented in the previous chapter, the contact surfaces are assumed to
be ’perfectly planar’ which means that the apparent contact area between two adjacent
layers extends to the whole length of the beam. This is possible if the individual lay-
ers do not present any curvature or other geometrical imperfections. In reality, most
commercially available sheet materials, such as the metal sheet that would be used to
prepare a set-up that replicates the modeled experiments present some level of curva-
ture. Typically, sheets and bands are wound on coils after production and do present
different amounts of curvature. The curvature is sufficient to create air gaps of the order
of several tenths of a millimeter between two layers that are fixed at one end so as to
create a cantilever beam. Under these conditions, the electrostatic stress between ad-
jacent layers is not sufficient to close the gap between layers, as the attraction between
electrodes decays in proportion to the square of the distance between them. Attempts
to eliminate the curvature by means of a stress relaxation thermal treatment were not
63
successful. The use of extremely thin, compliant layers were faced with the problem that
force needed to deform the structure in low stiffness state would be extremely small and
thus quite difficult to apply and measure in a reliable way.
To obviate these problems, the experiments presented in this chapters will demon-
strate the electrostatic coupling of the elements of a multi-layer structure using sand-
wich structures. These structures lend themselves for the demonstration of electro-
lamination for following reasons:
• Sandwich structures typically consist of 3 layers: 1 core and two faces. The limited
number of interfaces reduces the complexity of the experimental set-up and the
risk of electrical failure at the interfaces.
• The faces of a sandwich structure have generally a high tensile stiffness (thanks
to the high modulus of the materials they are made of) and a negligible bending
stiffness due to their very small thickness. This allows them to better conform to
the topography of the core surfaces.
• Based on the sandwich theory, the difference in bending stiffness between the
laminated and unlaminated sandwich structure is normally very large. Hence,
even if the irregular surface topography reduces the effectiveness of the electro-
lamination, an effect on the global properties of the structures can be clearly de-
tected.
In the following section the effect of electrolamination with respect to the global me-
chanical properties of sandwich beams are summarized. The experiments performed
on two different sandwich structures are then described in sections 4.2 and 4.3.
4.1 Inhomogeneous Tunable Structures: Sandwich Beams
4.1.1 Mechanical properties of fully bonded sandwich structures
The above considerations lead to the the idea to perform all experiments on sandwich
structures, a well known example of stiff, lightweight structural elements. In such struc-
tures, thin, axially stiff faces (typically made of CFRP or another high strength, high
modulus material) are glued on a lightweight, low modulus thick core to obtain a com-
posite structure with a high bending stiffness. The practical advantage of performing
64
the experiments on tunable stiffness sandwich structures is that the individual compo-
nents can be very compliant in bending direction, thus obviating the problem described
above.
The behavior of such structures has been extensively investigated. Some of the most
important relations describing the properties of such structures in relation to the con-
nectivity of the faces and the core are summarized in the following section. For common
engineering applications it is a requirement that the stiff faces be properly bonded to the
core, that means that a complete transfer of stresses between layers of the sandwich
structure takes places. The flexural rigidity of a sandwich structure is given by:
Z
Ef bf t3f Ef bf tf d2 Ec bc t3c
Dsandwich = Ez 2 dS = 2 ∗ + +
12 2 12
= 2Df + D0 + Dc (4.1)
Where Ef and Ec are the Young’s moduli of the faces and core, respectively, tf and tc
are the thicknesses of the faces and core, respectively and bf and bc the corresponding
widths, and d = tf + tc . [4] Under certain conditions, often met in sandwich structures,
the first and the third term in (4.1) can be considered negligible and the stiffness can be
approximated as:
Ef bf tf d2
Dsandwich ≈ (4.2)
2
The first term in (4.1) can be neglected (i.e. is less than 0.01 times the second term) if :
d 3
3( ) > 100 (4.3)
tf
The third term is less than 0.01 times the second term in (4.1) if
6Ef tf d2
> 100 (4.4)
Ec t3c
Additionally, if the core of the sandwich is ’soft’, a shear contribution to the transversal
deformation has to be considered. With a shear stiffness S
Gc bc d2
S= (4.5)
tc
the total deformation can be written as a sum of bending and shear deformation. In the
case of a discrete transversal force P acting on the beam at a distance l from the origin,
the total deformation vsandwich can be written as:
65
vsandwich = vbending + vshear
P l3 x x Px
= [3( )2 − )3 ] + (4.6)
6Dsandwich l l S
The shear stress between faces and core is found to be
int Tx E f bf tf d
σxy = · (4.7)
bf Dsandwich 2
with Tx = dMx /dx, where Mx is the bending moment acting on the structure. Combina-
tion with (4.1) yields
int P
σxy ≈ (4.8)
bf d
Local de-bonding between core and faces has also been studied as well as other modes
of failure that are common for sandwich panels [53]. For practical purposes, de-bonding
is regarded as a confined defect for which there is no adhesion between skin and core
int
only in delimited areas, hence σxy = 0. If the delamination area increases it has been
shown that the flexural stiffness of the sandwich will decrease [32].
4.1.2 Complete debonding
The extreme case of de-bonding is given when no shear stress transfer takes place be-
tween core and faces. Simple considerations and inclusion of the boundary conditions
show that for a loose bundle composed of the same faces and core as considered in
(4.1) and (4.2) the stiffness is
2Ef bf t3f Ec bc t3c

Dbundle = + (4.9)
12 12
Again, for a core with a sufficiently low bending stiffness compared to the faces, i.e.
Ec t3c ¿ 2Ef t3f (4.10)
the second term can be neglected in a first approximation, so that the ratio between the
stiffness of a sandwich and the stiffness of a loose bundle can be estimated as
66
Dsandwich 6Ef bf tf d2 tc
= 3 3
≈ 3( + 1)2 (4.11)
Dbundle 2Ef bf tf + Ec bc tc tf
Taking the shear contribution to the displacement into account, the ratio of the displace-
ments becomes a function of the position x:
vsandwich vbending,sandwich + vshear

=
vbundle vbending,bundle
Dbundle Dsandwich 6
= (1 + ) (4.12)
Dsandwich S 3lx − x2
and for x = l follows:
vsandwich Dbundle 3Dsandwich

= (1 + ) (4.13)
vbundle Dsandwich Sl2
Depending on the measurements and elastic properties of the individual components,
ratios Dsandwich /Dbundle of the order of 1000 can very easily be achieved, for the fully
bonded in comparison with the fully de-bonded sandwich layers.
4.2 Proof of Concept with a 140 mm Cantilever Beam
4.2.1 Experimental
The goal of the experiments presented in this section is to provide a demonstration

of the electrostatic stiffening of a simple structure in static tests. For this purpose a
sandwich beam (shown in figure 4.1) consisting of:
• two unidirectional (UD) carbon fiber reinforced polymer (CFRP) layers, coated on
one side with a ∼0.05 mm thick Poly(vinylidene fluoride) (PVDF) film, ²r = 8
• one electrically conductive, carbon-filled silicone elastomer core, , coated on both

sides with a 0.05 mm thick Poly(tetrafluoroethylene) (PTFE) film, ²r = 2
was chosen as a testing object.

From a mechanical point of view, the choice of the core and face materials was dictated
by the need to design a system in which the effect of changes in the shear stress trans-
fer at the interfaces between core and faces of the sandwich beam would be observed
67
unequivocally. Hence, the choice of a very low modulus material such as a silicone
elastomer for the core of the sandwich was made. UD CFRP bands as the ones utilized
for these experiment, have a high tensile stiffness and a very modest bending stiffness,
due to their very small thickness. This combination of mechanical properties is expected
to yield a markedly different behavior of the system in the bonded and unbonded state.
The main mechanical properties of the sandwich beam are listed in table 4.1.
Table 4.1: Properties of the mechanical components of the ’proof of concept’ sandwich
beam
Property Faces Core

Material Carbon Fiber Reinforced Carbon particle filled sili-
Polymer cone elastomer
Thickness [mm] 0.15 2.3
Width [mm] 12 10
Young’s Modulus [GPa] 120 ∼0.004
Free Length [mm] 140 140
From an electrical point of view, the use of electrically conductive structural elements
simplified the set-up of the experiment. The material originally chosen as the sole
dielectric between the structural elements was the ∼0.05 mm thick PVDF film. Due
to the tendency of the silicone elastomer to adhere to the PVDF film layers, additional
PTFE layers were applied to the surfaces of the core to obviate the problem. The thin
dielectric layers were considered to have a negligible contribution to the stiffness of the
system, due to their low elastic moduli.
With this configuration, an effective dielectric constant of ²r = 3.2 and an effective
dielectric layer thickness δ = 0.1 mm (sum of the thicknesses of the two layers) are
obtained, as calculated in (2.4) and (2.5).
The beam was positioned vertically and fixed 140 mm from the bottom end, where
two small FeNdB permanent magnets were positioned. Their function was to make it
possible to apply a transversal force by means of a time-variable magnetic field B(I(t)),
generated by a current I(t) that circulated in the coils. The coils were approximately in
a Helmholtz configuration, which means that the distance between the coils was equal
to their radius. In this configuration, a very homogeneous field in axial direction can be
obtained. The current I(t) was supplied by a Kepco bipolar operational power amplifier
BOP 20-20 M driven by the analog output tension of a National Instruments 6036 E
68
CFRP PVDF
x
PTFE
Si-Elastomer r
Coil Coil
LTD Coil
r Magnets
F(t)
B(t)
I(t) I(t)
Figure 4.1: Test set-up for the measurement of force-displacement diagrams of the
electrostatically tunable sandwich beam (left: schematic representation, right: photo-
graph).
DAQ board. The transversal displacement of the composite beam was measured at
two by means of two Micro-Epsilon optoNCDT laser displacement sensors (marked as
LDT in figure 4.1). The displacement was measured with a sampling frequency of 64
Hz. The components of the sandwich beam were connected to a high voltage power
supply (Stanford Research Systems PS 350), so that the necessary electrical potential
Ui could be applied between faces and core. The complete test setup was controlled
from a PC via a LabView interface developed for this purpose.
4.2.2 Measurement Technique
The behavior of the sandwich beam was observed in a step-wise quasi static cantilever
bending test. The force was applied at one position of the cantilever beam, 140 mm
from its fixation point. The force exerted on the magnets by the magnetic field B(I)
as a function of the coil current I(t) was measured by means of a load cell calibrated
in a separate experiment. The lateral displacement of the beam was measured in a
cantilever bending test, as a function of the position x along its main axis, the force P (I)
and of the high voltage setting Ui for the potential across the dielectric layer between
faces and core. After the initial loading to Pmax several cycles Pmax → −Pmax → Pmax
were measured. In addition, the transversal displacement of a similar beam with glued
faces was measured for comparison.
69
4.2.3 Results and Discussion
Behavior of the system
Figure 4.2 shows the force-displacement diagrams measured at a distance of x=125

mm from the fixation point of the cantilever beam for Ui settings between 0 V and 3000
V.
At lower voltage settings (0 V-1500 V) the curves can be roughly divided in three sec-
tions. The beam exhibits initially a high stiffness, while the force displacement behav-
ior of the beam shows a considerably lower stiffness for larger forces. Between the
high and the low stiffness portions, a transition domain can be observed. The transi-
tion between high and low stiffness becomes more diffuse with increasing voltage Ui .
The measurement at 500 V does not show a remarkable effect on the behavior of the
system. The slight reduction of the extension of the high stiffness portion of the force-
10
6
Force P [mN]
2 Measured U i Measured U i = 2000V

= 0V
Measured U i = 500V Measured U i = 3000V
1 Measured U i = 1000V Calculated (fully bonded)
Measured U i = 1500V Measured (fully bonded)
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Displacement v(x=125mm) [mm]
Figure 4.2: Force-displacement curves for the initial loading as a function of the elec-
trical potential across the face-core interface. Measured and calculated displacements
for the fully bonded sandwich are in good agreement and show higher stiffness than an
electrostatically stiffened bundle at high voltage (3000 V). The 2500 V measurement is
not shown as it is practically identical with the 3000 V measurement.
70
displacement curve should rather attributed to experimental artifacts. The extension
of the higher stiffness domain (between P = 0mN and the begin of the softening) in-
creases with further increasing voltage Ui . For the higher voltage curves (2000 V and
3000 V) only very limited softening is visible, starting approximately at 5 mN load. The
application of high voltage (3000 V) to the sandwich components, leads to an 18-fold
decrease in lateral displacement at x=125 mm and P = Pmax =9.2 mN in comparison
to the same system without the application of any electrical potential. The increase
in voltage does not change the slope of the high stiffness section of the curve but in-
creases the load at which softening can be observed. At each potential level Ui , after
the initial loading (shown in figure 4.2) two load cycles (from Pmax to −Pmax and back)
were performed.
Figure 4.3 shows an example of such a cycling procedure for Ui =1000 V. For better
clarity, a comparison of the central portions of such measurements (from −Pmax to Pmax
and back to −Pmax ) is shown in figure 4.4. As observed for the curves shown in figure
4.2, an approximately bi-linear behavior of the force-displacement curves is visible. The
approximated slopes ∆P ∆v
for the two sections of each curve are shown in each plot. The
intercept between the high stiffness branch and the low stiffness branch of the curves
(marked with a dot in figure 4.4) was calculated as the intersecting point of the linear
interpolation of the values measured for the two sections of the curves, hence com-
U hv= 1000V
10
2
Force P [mN]
-2
-4
-6
-8
-10
-4 -3 -2 -1 0 1 2 3 4
Displacement w(x=125mm) [mm]
Figure 4.3: Behavior of the system for two complete cycles (after the initial loading) at
potential Ui =1000 V.
71
pensating for the transition zone between high and low stiffness and providing an ideal
intersection point of the two lines. The linear interpolations were calculated assuming
that the slopes of the two branches (high and low stiffness) are approximately constant
over different Ui values. This could be confirmed for the slope of the upper branches
of the curves, where a ∆P ∆v
value of 3.1 Nm−1 could be calculated independently and
with good confidence for Ui values ranging from 0 V to 1500 V. The top section of the
2000 V curve seems to still be part of the transition to a lower stiffness value, which is
assumed to explain the higher ∆P ∆v
value.
The ∆P
∆v
value for the bottom section of the curves (70 Nm−1 ) could be calculated inde-
pendently for the 2500 V (not shown here as it is practically identical with the 3000 V
Uhv = 0V Uhv = 500V Uhv = 1000V

10 10 10
∆P ∆P ∆P
= 3.1 = 3.1 = 3.1
∆w ∆w ∆w
5 5 5
Force P [mN]
0 0 0
-5 -5 -5
∆P ∆P ∆P
= 70.0 = 70.0 = 70.0
∆w ∆w ∆w
-10 -10 -10
-4 -2 0 2 4 -4 -2 0 2 4 -4 -2 0 2 4
Uhv = 1500V Uhv = 2000V Uhv = 3000V
10 10 10
∆P ∆P
= 3.1 = 8.5
∆w ∆w
5 5 5
Force P [mN]
0 0 0
-5 -5 -5
∆P ∆P ∆P
= 70.0 = 70.0 = 70.0
∆w ∆w ∆w
-10 -10 -10
-4 -2 0 2 4 -4 -2 0 2 4 -4 -2 0 2 4
Displacement Displacement Displacement
w(x=125mm) [mm] w(x=125mm) [mm] w(x=125mm) [mm]
Figure 4.4: Behavior of the system as a function of the electrical potential across the
face-core interface. After the initial loading, the force-displacement diagrams show a
hysteretic behavior. The stiffness of the system increases with increasing potential Ui .
72
measurement) and 3000 V measurements. At lower voltages, this value is considered
to provide a reasonable interpolation of the bottom branch, but could not be confirmed
independently, due to the shortness of the interval on which the parameters for a re-
gression could be calculated. The calculated force difference ∆Ps between −Pmax and
the intersection point of the two interpolation lines described in the previous section is
plotted against the voltage Ui in figure 4.5. The magnitude of ∆Ps is interpreted as the
load at which the shear stresses between faces and core exceed the maximum shear
int
stress σxy,max that can be transferred across the face-core interface due to the electro-
int
static normal stress σyy . The corresponding shear stresses at the core-face interface
are calculated using (4.6).
According to (2.2) and (2.3), the shear stress at the face core interface is a quadratic
function in Ui . The measured force-displacement diagrams in figure 4.4clearly show a
hysteretic behavior in measurements at lower voltages. The force displacement curves
no longer pass through the origin of the coordinates, as shown in figure 4.4. This
behavior is attributed to friction related processes. The area contained within each
curve represents the energy dissipated by the friction processes in one load cycle. At
first, the area slowly increases from 18.8 J to 24.7 J with increasing Ui (0 V-1000 V)
and then it decreases rapidly to 0.6 J at 3000 V as shown in table 4.3.
Comparison with models
In addition to the force-displacement curves measured for the varying parameter Ui ,

figure 4.2 shows also the force displacement diagram for the case in which the faces
of the sandwich were glued to the core using epoxy resin. For comparison, the force
displacement curve for the glued sandwich calculated using (4.6) is also plotted. In this
case measured and calculated data are in good agreement. The displacements mea-
sured at 0 V are approximately 4 times smaller than predicted by (4.9), thus indicating
Table 4.2: Shear stresses calculated based on the sandwich theory

int
σxy,max as calculated using
Ui [V] ∆Ps [mN] ∆Ps for P in (4.8) [Pa]
0 5.1 209
500 6.2 254
1000 8.6 352
1500 12.3 504
73
Table 4.3: Energy dissipated by the sandwich beam in one loading cycle.
Tension across the Area of the

dielectric layer Ui [V] hysteresis curve [J]
0 18.8
500 21.1
1000 24.7
1500 22.3
2000 6.9
2500 0.8
3000 0.6
a higher stiffness than expected. The hysteresis in the curve indicates that already for
Ui =0 V, a certain level of interaction between faces and core is present, in spite of the
lack of an electrostatic normal force. To some extent, the difference can be explained by
the stiffer behavior observed at both ends of the force displacement diagrams (∆P > 0
in figure 4.4. Nevertheless, from the slope of the top section of the Ui =0 curve a bend-
ing stiffness of approximately 2.4*10−3 Nm2 was obtained, compared to 8.5*10−4 Nm2
as predicted by the model.
This difference between measured and calculated behavior can thus be partly explained
by an inaccuracy in the assumption that no interaction between faces and core takes
place for Ui =0 V. Even a small contribution of the bending of the faces about the neutral
axis of the sandwich would have a remarkable effect on the stiffness of the faces and
core bundle.
Figure 4.6 shows the normalized (i.e. divided by the value measured or calculated
at x=140 mm)displacements and fitted displacement functions for measurements per-
formed at Ui values of 0 V and 3000 V, respectively. For the measurements at 0 V,
the displacement function for a bundle was used to approximate the v(x) displace-
ment function (lateral displacement as a function of the position along the main axis),
whereas at 3000 V a sandwich behavior (i.e. also considering shear deformations) was
assumed. The displacement functions v(x) shown in figure 4.6 confirm an increase in
the contribution of shear to the displacement for the high voltage measurement, as the
reduced curvature of the beam indicates. With increasing voltage the overall stiffness
of the system increases. The maximum displacement of the sandwich subjected to high
voltage (3000 V) is approximately 0.12 mm at x = 125 mm, compared to 0.046 mm as
predicted by the model (4.6) and measured for the fully bonded sandwich. A stiffness of
74
35
Polynomial fit
Measured ∆P
30
25
20
∆P [mN]
15
10
5
-6 2
Polynomial fit: ∆P(Uhv )=3.2 .10 . U hv + 5.3
0
0 500 1000 1500 2000 2500 3000
High Voltage U hv [V]
Figure 4.5: Force at which the regression lines calculated for the force-displacement
diagrams shown in figure 4.4 intersect. The factor 3.2·10−6 mN V−2 is in fairly good
agreement with the value of 3.45·10−6 mN V−2 calculated on the basis of electrostatic
attraction forces using (4.8), (2.3) and (4.14).
approximately 1/3 of the maximum possible value for the described system could thus
int int
be achieved. Ideally no interaction between faces and core (i.e. σxy = σyy = 0P a) is
expected to take place at Ui =0. Nevertheless observation has shown that this does not
correspond to reality. The net contribution of electrostatic forces to the shear interaction
between faces and core was calculated by subtracting the ∆P offset at Ui =0 from ∆P :
∆Ps0 (Ui ) = ∆Ps (Ui ) − ∆Ps (0V ) (4.14)

int int
Table 4.4 lists values for σyy and an estimate of σxy based on the electrostatic attrac-
int
tion at different voltages using (2.2) and (2.3), respectively, as well as the σxy values
0
obtained by using the ∆Ps values from figure 4.4 in (4.8). A good degree of agreement
int
between σxy , as calculated based on the electrostatic forces between faces and core
int
and σxy , as obtained from the mechanical behavior of the system can be assumed.
75
U i = 3000V U i = 0V
1 1
Calculated Calculated
Displacement w(x) / w(l)

Measured Measured
Displacement v(x) / v(l) 0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 50 100 0 50 100
Position x [mm] Position x [mm]
Figure 4.6: Normalized bending lines at Ui = 0 and 3000 V respectively. The 0 V line ap-
pears to have a higher curvature than the 3000 V line due to a higher shear contribution
(linear in x) to the displacement at higher voltages, as expected considering (4.6).
Table 4.4: Normal and shear stresses calculated using electrostatic attraction and sand-
wich theory.
int int
0 0 σxy , as calculated σxy , as calculated
int
Ui σyy , as calculated in (2.2), using ∆Ps0 for P
(V) in (2.3) (Pa) assuming µ=0.1, (Pa) in (4.8), (Pa)
0 0 0 0
500 354 35 45
1000 1417 142 143
1500 3187 319 295
4.3 Characterization of a Large (>2m) Tunable Glass

Fiber Reinforced Beam
The system presented in the previous section served the purpose of demonstrating the
effect of the electrostatic coupling of the faces and the core of a sandwich beam, but
can hardly be of any practical use, due to its low stiffness.
The application example that will be presented in the next chapter calls for the mod-
ification of the mechanical properties of the lightweight glass fiber reinforced polymer
(GFRP) elements that compose the deck of a cable stayed pedestrian bridge built in
Empa’s Structural Engineering Research Laboratory. The adaptive upgrade of the com-
76
plete bridge deck is expected to be a labor intensive and expensive undertaking. The
development of an intermediate size sample of tunable structural element deemed a
reasonable way to prepare for the construction of the full-size demonstrator. In addition
to characterizing the behavior of the tunable element, materials and methods for the
upgrading were tested.
4.3.1 Experimental
In the set up used for the experiments, the bond between additional stiffening elements
(CFRP plates, CarboDur M614, Sika, CH) and the core of the sandwich beam (GFRP
I-beam, Fiberline Composites, DK) is given by electrostatic forces generated by an elec-
trical field built up across a 12µm thick polyethyleneterephtalate (PET, ²r ≈ 1, Amcor
Flexibles, Burgdorf, CH) film that was coated with a thin aluminum film on one side
(coating thickness unknown). The PET film was applied to the flanges of the GFRP
beam using epoxide resin (Araldit Standard, Vantico AG, Basel, CH). The CFRP plates
served at the same time as second pair of electrodes and stiffening member. This setup
was chosen in order to optimize the electrostatic forces that could be obtained per unit
voltage.
The multi-layer beam was fixed on one end and excited at the free end by means of the
electromagnetic fields generated by a coil, acting on a 40 mm x 40 mm x 20 mm FeNdB
permanent magnet that was attached to the end of the beam, 2240 mm from its fixation
point. The brackets visible in the left hand side of Figure 4.7 were only used to keep the
loose CFRP plates in place when the electrostatic field was turned off, ideally without
exerting any normal forces. Two additional masses (1050 g each) were placed 2090
mm (position of the center of gravity) from the fixed end of the beam to help separate
the first bending and torsion modes of the beam.
The currents necessary to induce the exciting alternate magnetic field, were generated
by a bipolar operational power amplifier (BOP 20-20 M, Kepco, USA) driven in its current
controlled mode by an arbitrary signal generator (33120 A, Agilent, USA).
The transversal accelerations at the end of the composite beam were measured by
means of two accelerometers (PCB Piezotronic, Mod. 3701 G3 FA3 G, with a sensitiv-
ity of 1 V/g) positioned at the left and the right edges at the free end of the beam (sensor
1 and sensor 2, respectively). The current circulating through the coil was measured
using a shunt resistor connected in series with the coil. The shunt had a resistivity of
0.01Ω (60 mV/6 A). The acceleration and the current circulating through the coil were
measured with a sampling frequency of 512 Hz using an OROS 38 data acquisition sys-
77
Accelerometers
- Electrode (CFRP)
Dielectric layer - - - - - -
0...2000V
+Electrode
+ + + + + +
Fixation
CFRP Strip
GFRP I-beam
Excitation
coil 120mm
6mm
CFRP-Strip,
60mm
Figure 4.7: Overview of the test set-up and cross section of a GFRP I beam as used in
the experiments described in section 4.3.2
tem. The components of the sandwich beam were connected to a high voltage power
supply (PS 350, Stanford Research Systems, USA), so that the necessary electrical
potential Ui could be applied between faces and core.
The set-up could be prepared in a very short time. The application of the aluminized
PET film, that served at the same time as electrode and dielectric layer, required only
approximately half a day of the work of two people. The preparation of the electrical
set-up (contacting the elements, setting up the elements, etc.) also required approxi-
mately half a day of work by two people. The mechanical set-up (fixation of the beam,
preparation of the retaining brackets for the bottom CFRP UD band) was probably the
most labor intensive part of the preparation work.
4.3.2 Shift of the Natural Frequency
The beam was excited with a 7-17 Hz linear chirp signal at six different current ampli-
tudes (I0 =0.5 A ... 3 A). The vertical accelerations of the beam end and the coil current
(assumed to be directly proportional to the generated electromagnetic-field, hence to
the force exerted on the magnet at the end of the beam) were recorded as a function of
time.
78
Figure 4.8 shows the typical acceleration amplitudes measured for Ui =0 V. The curves
show that in the 7-17 Hz range the first bending and torsion modes are excited and are
fairly close. In order to simplify the interpretation of the measurements, the torsional
motion was filtered from the results, by averaging the signals of sensor 1 and sensor
2. As shown in the figure, this procedure effectively reduces the intensity of the tor-
sion signal. In the following sections, only the bending behavior of the system will be
discussed.
From the current and the acceleration data (average of sensor 1 and sensor 2), transfer
functions of the system were calculated for different excitation current I0 and potential
levels Ui , as shown in the plots of figure 4.9.
Figure 4.9 shows the behavior of the system at six excitation levels (I0 =0.5 A ... 3 A) for
different interlaminar potentials Ui . In the series of measurements performed at I0 =0.5
A, quite remarkable reduction of the vibration amplitudes of the beam (up to a factor
of 5 in the transfer function for Ui =300 V) is recorded when the electrostatic coupling
of stiffening elements to the core of the beam is activated. A widening of the transfer
function peak is also observed for intermediate levels of Ui (200 V and 300 V), while
the tendency is reversed for higher potential values, where the maximum value of the
transfer function also increases.
A similar pattern is observed in the spectra recorded at higher excitation amplitudes.
The most remarkable difference between the measurement series performed at differ-
-40 2
-50 bending sensor 1 sensor 1
sensor 2 1.5 bending sensor 2
-60
average average
-70 1
-80 0.5
|G| [dB]
torsion
φ [rad]
-90
0
-100
-110 -0.5
-120 torsion -1
-130
-1.5
-140
-150 -2
8 10 12 14 16 8 10 12 14 16
f [Hz] f [Hz]
Figure 4.8: Amplitude and phase spectrum of the accelerations (sensor 1, at the left
edge of the beam and sensor 2 at the right edge) measured at the free end of the beam
as a function of frequency for I0 = 2.6 A.
79
ent I0 values is that at higher I0 values, the peak values of the transfer functions |G|for
the activated beam are significantly lower than at low I0 values, as summarized in figure
4.10. At the same time the reversal of the trend, as observed in the series measured at
I0 =0.5 A is not observed for the highest excitation amplitudes.
Also, the maximum frequency shift for a given interlaminar voltage U tends to decrease
with increasing amplitude of the exciting force. This is caused by the softening of the
system due to the interlaminar slip at higher transversal loads. The transfer function
plots also show a clear difference between the lower and the higher excitation regimes
in terms of peak value of the transfer function |G|, as summarized in the right hand plot
of figure 4.10, where max(|G|) values are plotted as a function of interlaminar potential
Ui and excitation current amplitude. The resonance frequencies of the electrostatically
system are summarized for the measurement series at all excitation levels in the left
hand plot of figure 4.10.
The transfer function drawn in black in the I0 =1.0 A plot of figure 4.9 shows the behav-
ior of an equivalent GFRP-CFRP beam, in which the coupling between the elements
was achieved using an epoxy-type adhesive. This curve represents the behavior of
a system in which full shear stress transfer between the elements is obtained. The
transfer function is characterized by a well formed, sharp resonance peak, indicating
low energy loss due to damping and friction and by the significantly higher first bending
resonance frequency than the electrostatically coupled system. The considerably lower
eigenfrequencies registered in the electrostatically coupled system, compared to the
system with glued CFRP elements indicate that a large potential for the improvement of
the electrostatic coupling of the components of the beam is given. Ideally, it should be
possible to attain the same level of stiffening of the beam using electrostatic coupling
as with adhesive coupling. The reduction of the maximum values of the transfer func-
tions and thus of the energy content of the system are explained by losses due to the
damping behavior of the system, as detailed in the next section.
4.3.3 Damping
Section 4.3.2 shows that the application of an electrostatic field at the interface between
the GFRP core beam and the CFRP stiffening UD bands has an effect on the first
bending natural frequency of the structure. This property can be used to avoid the onset
of resonance situations if the average frequency of a narrow band excitation coincided
with the natural frequency of the unstiffened beam.
Figure 4.9 shows that due to the ability of the system to provide full shear stress trans-
80
Current Amplitude: Current Amplitude: Current Amplitude:
I0=0.5A I0=1.0A I0=1.5A
2 2 2
10 10 10
1 1 1
|G| [a.u.]
|G| [a.u.]
|G| [a.u.]
10 10 10
U =0V
i
U i=100V
0 0 0
10 10 10
U i=200V 5 10 15 20 5 10 15 20 5 10 15 20
U =300V Frequency [Hz] Frequency [Hz] Frequency [Hz]
i
81
U i=400V
U i=500V I0=2.0A I0=2.5A I0=3.0A
U i=600V
is plotted in black in the I0 =1 A graph for reference.

U i=800V
2 2 2
adhesive bond 10 10 10
1 1 1
|G| [a.u.]
|G| [a.u.]
|G| [a.u.]
10 10 10
0 0 0
10 10 10
5 10 15 20 5 10 15 20 5 10 15 20
Frequency [Hz] Frequency [Hz] Frequency [Hz]
terfaces of the system. The transfer function of a beam with laminated CFRP elements
excitation current amplitudes indicates a change in the shear stress transfer at the in-
Figure 4.9: The change of the transfer function measured at different Ui levels and
250
bending) [Hz]
12
200
max(|G|) [a.u.]
11.5
st
150
Eigenfrequency (1
11
100
10.5
50
800 800
600 3 600 3
U 400 2 400 2
i [V U
] 200 [V
1 [A] i ] 200 ]
0
I0 1 I 0 [A
0
Figure 4.10: Overview of the eigenfrequency values of the system at different interlam-
inar voltage levels and current amplitudes (left) and of the maximum values max(|G|)
(right). For low excitation currents the lowest value of max(|G|) is at intermediate voltage
levels, while for high excitation currents, the minimum value coincides with the highest
voltage levels that could be applied.
fer at the interfaces only within a small range, an approximately elastic behavior of
the stiffened structure is observed at only high tensions and low excitation amplitudes.
Specifically, only in the measurements with the lowest excitation amplitude (0.5 A) at the
highest interlaminar potential, the shape of the resonance peak (with a higher maximum
and sharper features) indicates a predominantly elastic behavior.
The energy dissipation in the system is due to material damping as well as to the friction
that occurs at the interfaces between the layers. Since the normal stress is proportional
to the square of the electrostatic field U 2 and the shear stress occurring at the interfaces
depends on the amplitude of the vibration, the damping of the system is expected to
vary depending on the amplitude of the excitation as well as on the voltage applied
between faces and core of the sandwich.
Variable damping systems based on the use of magneto-rheological fluids are in use
and under development for a variety of applications ranging from so called smart shock
absorbers in cars to the damping of stay cable vibrations [50]. The development of opti-
mal damping strategies using magneto-rheological dampers for stay cables is based on
the observation that a setting of the damper that is too stiff will reduce the effectiveness
of the device, while a setting that is too soft will not dissipate as much energy as would
be possible for a given vibration velocity and amplitude.
The damping behavior of the system was investigated in experiments in which the sys-
tem was excited with a given current amplitude I0 (varying between 0.5 A and 3.0 A) at
82
its measured eigenfrequency for that excitation level. The excitation was then switched
off and the acceleration at the free end of the cantilever was measured as it subsided.
The set up for the experiments was the same as shown in figure 4.7.
The plots in figure 4.11 show the displacement at the free end of the beam and the
envelope of the displacement curve for four of the experiments performed to investigate
the damping behavior of the beam. The envelope of the curve corresponds to the
instantaneous amplitude of the harmonic oscillation of the system. The envelope of
the time displacement curve is calculated as the absolute value of the Hilbert transform
H(v(t)) of the curve v(t). In figure 4.11 it is plotted against the time t. The portion
of the curve shown in the figure is the one used to study the damping behavior of the
system. In the experiments, the excitation current was switched off immediately before
the beginning of the shown plots.
The envelope of the amplitude-decay portion of the data was calculated for all the mea-
sured curves and is shown in figure 4.12, in its logarithmic form as a function of the
number of cycles, to compensate for the different frequencies at which the measure-
ment were performed. As expected based on the results shown previously, the curves
measured with no interlaminar stress (0 V curves) present the highest displacement
amplitudes. In figure 4.12 can be noticed how in the first graph on the top left (exci-
tation current amplitude 0.5 A) the displacement amplitudes of the curves with low to
intermediate interlaminar voltage (Ui =0 V ... Ui =300 V) show a faster decay than the
high voltage curves.
-4 Excitation Current Amplitude: 0.5A -3 Excitation Current Amplitude: 3A

1.5 x 10 1.5 x 10
Envelope 0V Envelope 0V
U i=0V U i=0V
1 1
Envelope 800V Envelope 800V
0.5 U i=800V 0.5 U i=800V
v(t) [m]
v(t) [m]
0 0
−0.5 −0.5
−1 −1
−1.5 −1.5
10 10.5 11 11.5 12 12.5 10 11 12 13 14 15
t [s] t [s]
Figure 4.11: Displacement as a function of time and in red its envelope (top) and ex-
citation current as a function of time for a weakly excited beam and no interlaminar
potential.
83
For the intermediate voltage values (Ui =200 V and Ui =300 V), the absolute value of
the amplitude at the beginning of the decay curve is only marginally higher than the
amplitude of the high voltage curves. Due to the higher slope of the former, the dis-
placement amplitudes drop to lower values than for the high voltage levels (Ui =400 V ...
Ui =800 V), within a few cycles from the beginning of the measurement. It follows that
for small excitation amplitudes, the system displays a more efficient damping behavior
at intermediate voltages than at the highest and lowest settings. At higher amplitudes,
this behavior changes and the high voltage curves gradually (as the excitation ampli-
tude grows) show a better performance in terms of overall reduction of the displacement
amplitude than the low voltage curves.
In order to quantify the amount of damping present in a structure, it is customary to
calculate the logarithmic decrement of the amplitude decaying vibration curve. This
parameter is calculated under the assumption that the envelope of the curve has an
exponential form, which is the commonly observed behavior for material damping. The
logarithmic decrement λ of a decaying vibration is given by quotient of the amplitude v0
of the vibration at any given cycle and the amplitude vn n cycles later as in (4.15).
1 v0
λ= ln (4.15)
n vn
In the plots shown in figure 4.12, an exponential decay of the vibration would be repre-
sented as a straight line with slope −λ. Figure 4.13 shows the effect of the interlaminar
voltage Ui on the shape of the decay curves. An exponential decay (plotted in red) rep-
resents the commonly measured material viscous damping behavior. A linear decay
(plotted in green) describes a system in which energy is dissipated by coulomb friction.
As it can be seen in examples shown in figure 4.13, the exponential decay curves
are in general a better approximation of the measured behavior than the linear decay.
However, for the higher ratios of exciting current amplitudes to interlaminar voltage
(namely I0 =0.5 A with Ui =0 V, I0 =3.0 A with Ui =0 V and I0 =3.0 A with Ui =300 V) it
can be seen that there is a higher level of disagreement between the exponential decay
approximation and the measured decay than for lower ratios. This points to a higher
contribution of friction damping to the overall damping of the system. Here, neither
a linear decay model nor an exponential decay model correctly describe the behavior
of the system. Nevertheless, there is a need to describe the damping behavior of
the presented structures in order to be able to calculate its dynamic behavior. Most
numerical packages for the mechanical modeling of structures support the calculation
of the damping in structures based on an exponential decay, so a global logarithmic
decrement will be calculated over a representative portion of the decay curves for use
84
I0=0.5A I0=1.0A I0=1.5A
−6 −6 −6
−7 −7 −7
−8 −8 −8
−9 −9 −9
ln (H(x(c))) []
ln (H(x(c))) []
ln (H(x(c))) []
−10 −10 −10
−11 −11 −11
−12 −12 −12

0 20 40 60 0 20 40 60 0 20 40 60
Cycles [ ] Cycles [ ] Cycles [ ]
85
interlaminar voltages between 0 V and 800 V
I0=2.0A I0=2.5A I0=3.0A
−6 −6 −6
−7 −7 −7
−8 −8 −8
−9 −9 −9
ln (H(x(c))) []
ln (H(x(c))) []
ln (H(x(c))) []
−10 −10 −10
−11 −11 −11
−12 −12 −12

0 20 40 60 0 20 40 60 0 20 40 60
a function of number of cycles c for excitation amplitudes between 0.5 A and 3.0 A and
Figure 4.12: Logarithm of the envelopes ln(H(v(c))) of the measured displacement as
I0=0.5A I0=0.5A I0=0.5A
Ui=0V Ui=300V Ui=800V
−4 −5 −5
x 10 x 10 x 10
1.5 6 6
5 5
1 4 4
3 3
0.5 2 2
Amplitude H(v(c)) [m]

1 1
0 0 0
100 110 120 130 140 150 160 170 100 110 120 130 140 150 160 170 100 110 120 130 140 150 160 170
Exponential decay
Linear decay
86
Measured decay
Start
at the lowest and at the highest excitation level.

End
I0=3.0A I0=3.0A I0=3.0A
Ui=0V Ui=300V Ui=800V
−3 −3 −4
x 10 x 10 x 10
1.5 3
1
1 2
0.5
0.5 1

0 0 0
100 110 120 130 140 150 160 170 100 110 120 130 140 150 160 170 100 110 120 130 140 150 160 170
linear approximation, between the start point and end point for decay curves measured
Figure 4.13: A comparison of the measured decay envelopes with their logarithmic and
in numerical models.
The relationship equivalent to 4.15, expressed for the continuously calculated envelope
(H(v(c))) is :
∆(ln(H(v(c))))
λ= (4.16)
∆(cycles)
Using the continuous envelope of the decay curves, it is possible to calculate a con-
tinuous logarithmic decrement curve that gives us more detailed information about the
damping behavior of the system, which is especially useful given the difficulty fitting the
decay curves with the commonly used approaches.
The logarithmic decrement λ was calculated for the beam in all test conditions (coil
current amplitude varying between 0.5 A and 3.0 A and interlaminar voltage varying
between 0 V and 800 V). The results are presented in figure 4.14, where λ is plotted
as a function of the number of cycles after the excitation current was switched off. The
circles at the right edge of each plot represent the global logarithmic decrement, for
comparison. The continuous logarithmic decrement curves shown in the figure are cal-
culated for the same portion of the decay curves used to calculate the global decrement
values. As it can be seen, the spread of the values increases remarkably for the higher
excitation current values.
Finally, figure 4.15 gives an overview of the effect of electrostatic coupling on the two
mechanical properties that are most relevant to the vibration behavior of the described
GFRP-CFRP beam. The figure shows how the eigenfrequency values display a low
domain, for the non activated system and after a fairly steep increase that begins at
Ui values between 200 V and 400 V they reach a high plateau value. Correspondingly
roughly at Ui values that mark the beginning of the increase in eigenfrequency, the
logarithmic decrement values show a maximum. This picture is compatible with the
general idea that the maximum damping is obtained when a good level of coupling
between the components of the system is given, but not sufficient shear stress transfer
is possible to obtain a full bond between them.
The values shown in figure 4.15 will be used as a base for the numerical calculations
of the behavior of a single GFRP-CFRP beam and of a GFRP-CFRP bridge deck pre-
sented in chapter 5.
87
40
40
I =0.75V
Cycles []
Cycles []
I =1.5V
20
20
0
0
0
0
0
0.4
0.2
0.4
0.2
λ[ ] 20 40 λ[ ]
20 40
I =1.25V
Cycles []
Cycles []
I =0.5V
0
0
0
0
0
0
0.4
0.2
0.4
0.2
λ[ ] λ[ ]
40
40
I =0.25V
Cycles []
Cycles []
I =1.0V
20
20
0
0
0
0
0
0.4
0.2
0.4
0.2
λ[ ] λ[ ]
decrement
U =100V
U =200V
U i=300V
U i=400V
U =500V
U i=600V
U i=800V
U i=0V
global
i
Figure 4.14: The graphs show the calculated instantaneous logarithmic decrement as
a function of the number of cycles after the begin of the amplitude decay. Results
for different excitation amplitudes I0 and interlaminar potentials Ui are shown. The
circles on the right hand side of the graphs represent the global logarithmic decrements
calculated based on the whole decay curves.
88
200 V
300 V
200 V
300 V
400 V
400 V
I 0 =0.5A
I 0 =1.0A
12.5 0.25
I 0 =1.5A
I 0 =2.0A
12
I 0 =2.5A 0.2
Eigenfrequency [Hz]
I 0 =3.0A
11.5
λ[ ]
0.15
11
0.1
10.5
10 0.05
0 200 400 600 800 0 200 400 600 800
U i [V] U i [V]
Figure 4.15: The effect of electrostatic coupling on the first bending eigenfrequency
and the logarithmic decrement of the described beam is a function of the interlaminar
potential Ui and the excitation amplitude I0 , as shown in the two plots of this figure.
4.4 The Effect of the Interface Topography
Sections 2.4 and 2.5 showed that the deviation of the surface topography from a per-
fect plane at an atomic level and at a microscopic level have noticeable consequences
on the interaction between interfaces. At a macroscopic level, the observation of the
interfaces of a GFRP-CFRP beam used for preliminary tests, showed that after several
million load cycles at varying amplitudes and interlaminar stress levels signs of wear
of the dielectric layer applied to the GFRP structures were very localized, as shown in
figure 4.16. In this section an estimation of the effect of the unevenness of the surfaces
of the GFRP base structure on the ability of the CFRP stiffening elements to conform
to the beam is made.
The beam shown in the figure was manufactured using a different technique than the
one used for the beam presented in the previous sections. The dielectric layer was a
0.08 mm thick Polyvinylidene Fluoride (PVDF) film that was chosen for its higher dielec-
tric constant (² ≈ 2...8, depending on the polarization of the film). Since no metallized
film was available, a separate electrode layer had to be applied between the GFRP host
structure and the PVDF film to obtain the desired experimental set up. In order to op-
timize the electrostatic performance of the system, an electrically conductive adhesive
was chosen to serve both as electrode and adhesive. This choice made it possible to
nominally reduce the gap between the electrode applied to the GFRP beam and the
CFRP elements to the thickness of the PVDF film. Due to the high viscosity of the sil-
ver particle loaded adhesive, it was not possible to obtain a sufficiently planar interface
89
Figure 4.16: A view of the top flange of a GFRP beam , used in a tunable GFRP CFRP
structure as described in the previous sections, after approximately two years of (non-
continuous) operation. The areas marked in black presented clear signs of wear due to
friction between the surface of the dielectric layer with the CFRP bands. All other areas
appeared to be in pristine condition.
between the elements of the structures.

As a consequence of this situation, only a small fraction of the apparent area of contact
(i.e the length of the beam × the width of the flange) was in the condition to transfer
shear stress between the components. The localization of the wear signs on the PVDF
film surface points to the lack of planarity as a likely cause of decreased performance
of the electrostatic coupling between the GFRP beam and the CFRP elements.
The ability of the CFRP elements to conform to the topography of the GFRP beam
surface depends on their bending stiffness and on the radius of curvature of the topo-
graphical surface features. Although for the purpose of calculating the global stiffness
of the structure, the bending stiffness of the CFRP elements can be neglected, this
property is of interest in the local scale.
90
Peaks and valleys over a short distance (e.g. a few centimeters) are thus more likely to
reduce the area of contact between the GFRP beam and the CFRP stiffening elements.
The above observations indicate that local unevenness of the surface in this scale may
represent a major source of loss in contact area, as shown in figure 4.16: Due to its
stiffness, the CFRP elements cannot conform to such topographical features over a
short distance, while ’long wave’ features can be compensated for, as shown in figure
4.17.
In order to quantify the adverse effect of uneven surfaces on the contact stress between
elements of a multi-layer stiffness, we shall consider a simple situation, as the one
depicted in figure 4.18.
In figure 4.18, a CFRP element with bending stiffness EICF RP is bent to conform to a
local feature with radius ρ. The bending moment necessary to obtain the curvature κ of
the CFRP elements is given by the well known moment/curvature relation (4.17) [9]:
1 Mb (x)
κ(x) = = (4.17)
ρ(x) EICF RP
Where Mb is the bending moment acting on the element and EICF RP is the bending
Figure 4.17: Local unevenness vs. long range unevenness of the surface
91
Fr
Fr
CFRP stiffening
element y r(x)
Dielectric layer
z x
el
GFRP host σ yy
wmax
structure
Figure 4.18: The CFRP stiffening element must adapt to the topography of the host
structure, so that contact at the interfaces is possible and shear stress can be trans-
ferred between the two components by means of friction.
stiffness of the element and ρ is the radius of the curve described by the neutral axis
of the CFRP element. (4.17) means that for a given stiffness of the CFRP element, the
bending moment needed for it to conform to a feature is proportional to the curvature of
the feature.
From (4.17) and the situation depicted in figure 4.18, the order of magnitude of the
distributed load that is necessary to maintain the curvature κ can be easily calculated
as Mb = EICF RP /ρ. The load applied to the CFRP elements, so that they conform to the
surface of the GFRP beam and are pressed against it, is generated by the electrostatic
field applied between the electrodes of the system. In the following calculations it will be
assumed that the CFRP bands have already been deformed to conform to the surface
of the structure. The distributed force needed to keep the bands in their deformed state
is then calculated as a function of the bending stiffness of the CFRP elements and the
length of the topographic feature the band is adapting to, for a fixed depth of the feature
of 0.01 mm. Also, in order to eliminate unnecessary complexity, it is assumed that the
profile of the topographic feature is described by the equation:
y(x) = c · (x4 − 2Lx3 + L2 x2 ) (4.18)
which corresponds to the bending line of a homogeneously loaded beam fixed at both
92
ends, as is the case for the CFRP band. Specifically, the deformed shape of the CFRP
el
band subjected to distributed load σyy is given by:
el
σyy
v(x) = · (x4 − 2Lx3 + L2 x2 ) (4.19)
24 · EICF RP
el
A few transformations of (4.19) yield the relationship between the distributed load σyy
necessary to keep the CFRP element deformed so as to conform to the surface profile,
the length L of the feature and the stiffness EICF RP of the CFRP element, for a given
maximum depth of the feature:
el EICF RP
σyy = 384wmax (4.20)
L4
Figure 4.19 shows the load needed to maintain the contact between a CFRP element
with an indentation described by (4.18) and a depth vmax = 0.1mm, as described in
figure 4.18.
The above considerations show that topography of the contact surfaces is thus an
important parameter for the successful design of electrostatically coupled multi-layer
structures. In order to increase the smoothness of the contact surfaces following mea-
sures should be adopted:
• The use of more refined methods such as autoclave curing to temporarily reduce
the viscosity of the epoxide adhesive during curing (or the use of lower viscosity
materials at room temperature).
• The application of smooth molds to the surface during curing will improve the
quality of the surfaces, while applying a vacuum to remove excess resin.
• The use of metallized polymer films instead of a combination of plain polymer

film and a metal particle loaded adhesive, would eliminate the need of such high
viscosity adhesives.
4.5 Conclusions
The experiments presented in this chapter were designed so as to demonstrate the

most remarkable properties that were predicted by the models of multi-layer systems.
93
~σyy,el
100000
80000 3
F e [Pa]
60000
40000
20000 2
]
4
0
[Nm
0.05
RP
1
CF
EI
0.1
L [m
] 0.15
0.2
Figure 4.19: The distributed load needed for the CFRP element to conform to the to-
pographical features of the GFRP beam for a given feature depth increases steeply for
short length features. The red line shows the approximate electrostatic stress that can
be generated across an 80µm thick PVDF film applying a potential of 3000 V or across
a 12µm thick PET film applying a potential of 800 V.
In the first set of experiments (section 4.2 proof of concept), the ability to stiffen a simple
structure by coupling stiffening elements electrostatically was proven using a structure
with extreme properties (very compliant core and faces with very high axial stiffness).
Also, the hysteretic behavior of the system, when loads are applied beyond the ability
of the interfaces to transfer shear stresses was shown. This is of one of the properties
that were predicted by the numerical model of a multi-layer beam. Furthermore it was
possible to demonstrate the relationship that exists between maximum external load
under which the system behaves linearly and the interlaminar stress.
The experiments described in section 4.3 served various purposes:
The preparation of the test set-up was an opportunity to improve the techniques for the
manufacturing of electrostatically tunable devices. Next to the selection of appropriate
materials, The procedure showed some of the difficulties that are encountered in the
94
manufacturing of electrostatically coupled multi-layer beams. The main difficulty in the
fabrication by hand is to obtain a sufficiently planar surface of the interfaces. In the
case of the first beam prototype, an electrode needed to be applied to the surface of
the insulating GFRP beam. In order to minimize the thickness of the insulating layer
between the electrodes, a silver epoxide was chosen as an electrode that would serve
also as an adhesive to mount the dielectric layer on the beam (see figure 4.7). The
electrode was applied by hand, then the PVDF film was laid on it and pressed using a
rubber roller to remove air inclusions.
This procedure yielded a fairly smooth electrode/insulator layer, but the topography of
the electrode was still sufficiently pronounced. So, after about two years of operation it
became clear from the wear marks that only one relatively small portion of the area of
the dielectric layers was in contact with the CFRP elements, as shown in figure 4.16.
The second beam prototype that was used for the bulk of the presented experiments
was manufactured using a metallized PET film and epoxide resin. While no autoclaving
was used in the process, a qualitative improvement of the surface could be observed.
Quantitative methods to measure the topography of the surface have yet to be applied.
Two aspects of practical relevance were studied in the presented experiments:
• At sufficiently small excitation amplitudes, a modification of the natural frequency

of the element could be obtained while maintaining a similar quality factor as in
the case of the unstiffened beam. This indicates that the effect of dissipative
processes under these conditions is modest, as shown in figure 4.9.
• The dissipative effects suggested by the hysteretic behavior observed in the nu-
merical simulations and in the experiments presented in section 4.2 could be con-
firmed by the observation of the dynamic behavior of the CFRP-GFRP beam at
high excitation amplitude/interlaminar potential ratios. The measurements of the
decay of the vibration amplitude presented in section 4.3.3 showed the electro-
static coupling of the CFRP elements to the beam also leads to a remarkably
high damping of the system, under appropriate conditions. The effect of the fric-
tion damping is optimal only within a quite narrow window of vibration amplitudes,
appropriate control systems will be necessary to optimize the energy dissipation
over a wider window of amplitude, as shown by the results presented in figure
4.15.
The main goal of the experiments carried out on the CFRP-GFRP system was to
demonstrate that the successful use of electrostatic coupling to suppress vibrations
95
is possible on elements that are used in real structures, such as the beam presented
in the section. Based on the acquired information, the scaling up of this method will be
performed on Empa’s GFRP pedestrian bridge deck, as described in the next section.
96
Chapter 5
A Practical Application of Elements

with Tunable Mechanical Properties
The stiffness change and the damping properties determined for the CFRP-GFRP
sandwich beam in the previous chapter are at the basis of the estimation of the proper-
ties of a composite bridge deck presented in this chapter. The goal of the application of
electrostatic coupling of CFRP stiffening elements to the GFRP structure of a footbridge
deck is the reduction of the amplitude of the vibrations that such lightweight structures
are prone to. In the first section of this chapter, a homogenized finite element model of
the beam discussed in the previous chapter is set up. The elastic and damping prop-
erties of the orthotropic beam are set so that the frequency response of the modeled
beam corresponds to the measured frequency response. The second part of this chap-
ter shows, based on calculations performed using a numerical model of the bridge in
question, the effect that the modification of the elastic and damping is expected to have
on the frequency response calculated for the bridge deck.
5.1 Suppression of the Vibrations of a GFRP Footbridge

Deck
Light weight, slender bridges are prone to high amplitude vibrations due to their reduced
mass and high degree of slenderness. A full scale pedestrian cable stayed bridge with
a bridge deck composed of GFRP beams similar to the one described in section 4.3
was designed and built so as to recreate the conditions that are typically the origin of
considerable vibration problems in pedestrian bridges. The structure shown in figure
97
5.1 is described in detail in [26] and is intended as a research platform that can be
used in the investigation of vibration damping strategies and structural health monitoring
systems.
The experiments described in the previous chapter have demonstrated that a combi-
nation of stiffening and increased damping is achieved when an electric field is built
up between the interfaces. Under such conditions the system is stiffened. When the
interfacial shear stresses between layers exceed the maximum shear stress that can
be transmitted by means of friction, increased damping can be observed, as shown
by the plots of figure 4.9. Under these conditions, interfacial slip occurs and energy is
dissipated by Coulomb friction at the locations where high relative displacements are
found.
This combined approach to the reduction of structural vibrations presents a few interest-
ing advantages over traditional systems based on discrete devices such as conventional
viscous dampers or tuned mass dampers:
• Depending on the nature of the excitation, vibrations can be suppressed either by

a change of stiffness of the system (i.e. no energy transfer to or from the structure
is allowed for) or by the introduction of friction damping into the system.
• Since the electrostatically coupled elements can be distributed over the whole
length of the structure, a more efficient energy extraction is possible. The dis-
tributed nature of the system allows for a robust approach to damping that is less
sensitive to the location of the nodes of vibration modes.
Figure 5.1: Empa’s full scale lightweight pedestrian bridge is used as a test platform for
structural health monitoring and vibration mitigation developments. Left, general view.
Right, detail from the bottom
98
• Damping is only introduced as needed and the amount of damping can be con-
trolled, if an appropriate control system is implemented.
The application of electrolaminated stiffening elements for the damping of lightweight

structures such as the Empa laboratory footbridge is thus an appealing method for the
suppression of their vibrations.
The implementation of this method for the damping of the deck of the footbridge is
planned as a follow up to the work presented here. The plan is to embed the stiffening
CFRP elements into the flange of the longitudinal GFRP beams shown in the right hand
side picture of figure 5.1, as shown in the sketch of figure 5.2. Next to considerations
made in the previous chapters, practical issues need to be addressed in order to put
the plan into practice.
bc
dc
CFRP UD Element
CFRP UD Element
b Electrode
Dielectric
Figure 5.2: Embedded, electrostatically couplable CFRP elements for the stiffening and
damping of the GFRP bridge deck
99
The experience gathered with the adaptive retrofitting of the GFRP beam shows that
the quality of the interface between the host structure (GFRP beams with laminated
electrodes and dielectric layer) is of great importance, especially in terms of the local
planarity of the contact surfaces. Appropriate machining of the surfaces and lamination
of the electrodes and dielectric materials will be necessary to obtain the maximal con-
tact area (i.e. as close as possible to 100% of the apparent area of contact) between
GFRP beam and CFRP stiffening elements. All joining details of the GFRP host struc-
ture (e.g. the screw connections in the flanges highlighted in the righthand photograph
of figure 5.1) will have to be adapted to the new geometry of the beams (shown in figure
5.2), because of the space requirements set by the embedded CFRP elements.
In this pilot phase, the adaptive retrofitting of the bridge is expected to be labor intensive
(and thus expensive), because all modifications to the existing structural elements will
have to be performed by hand. Hence, it is necessary to estimate the benefits of the
stiffening and damping on the vibratory behavior of the bridge obtained through the
proposed retrofit prior to beginning its construction. A numerical model of the system is
a viable way to obtain this information. In a first step, a model of the beam presented
in the previous chapter is set up and used to estimate the mechanical parameters by
comparing the measured and calculated values for the first bending eigenfrequency
and adapting the mechanical parameters of the model accordingly. In a second step
the behavior of the bridge is calculated by applying the estimated parameters to the
numerical bridge model.
5.2 Models
As in section 3.1.2, the models presented in this section were calculated using Comsol
3.4. The package used for this work is not specialized for mechanical applications
and thus only offers limited freedom in the selection of element types and mechanical
material models. Since only the bending behavior will be considered for the calculations
presented in this work, two dimensional models are sufficient to investigate the behavior
of the structures under consideration. Given the geometry of the systems, plain strain
is assumed for the calculations.
In order to reduce the complexity of the model and hence the computing time, homog-
enized models are used. In order to render the shear compliance of the structure that
is derived from the shape of the cross section as well as the inherent properties of
the materials, an orthotropic material model had to be used, so that independent input
100
CFRP-Strips (faces)
Ifaces(active/inactive), Efaces Homogenized beam
with Ih, Eh, Ah, Gh
Shear compliant web

Aweb, Gweb
GFRP-I-beam (core)
Ibeam, Ebeam
Figure 5.3: Mechanical components of the switchable stiffness beam used in the ex-
periments (left) and homogenized equivalent used for the numerical model (right).
for the tensile and shear properties could be given. The essential properties of the
structures are given by two geometrical quantities (cross-sectional area A and second
moment of area I), their weight, and material properties of the orthotropic structure:
Ex , Ey , Ez and G (Gxy ), the elastic modulus in the direction of main axis of the beam,
the two moduli perpendicular to the axis and the shear modulus, respectively. Given
the bi-dimensional nature of the model, Gxz and Gyz are not considered in the model.
The model is hardly sensitive to changes in Ey , Ez for the load case considered here.
The Poisson ratios νxy , νxz , νyz were assumed to be 0.33.
5.2.1 Undamped Behavior of the Beam
In this section a numerical model of the beam presented in section 4.3 is described.
The model is meant to represent the dynamic behavior of the structure in its two ex-
treme states, i.e. when the stiffening CFRP elements are not bonded to the beam (low
stiffness state) and when they are laminated to the beam using epoxide adhesive (high
stiffness state). Additionally, in section 5.2.2 the effect of damping will be considered
based on the properties calculated in section 4.3.3.
The homogenized model of the beam has the same bending stiffness Ex ·Iz , core shear
stiffness G · Acore , outer dimensions and weight per unit length as the beam used in the
experiments, as shown in figure 5.3.
The eigenfrequencies calculated with this model are compared to the ones obtained
in the presented experiments. The discrepancies between the results of the numerical
101
calculation and the experiment are resolved by a simple model updating [19] procedure,
by which the elastic constants used in the model are modified so as to bring the model
results to coincide with the experimental results (see figure 4.9, I0 = 1A). Using this ap-
proach, the deviations of the behavior of the tested system from the expected behavior
are thus ’packed’ into the properties of the material defined in the model.
Additionally to the density of the beam, two discrete masses have to be considered in
the model:
• The permanent magnet attached to the free end of the beam, used to obtain the
excitation force in combination with the current coil placed under it (see figure 4.7).
• Two additional masses positioned near the web, 2090 mm from the fixation of the
beam, used to separate the first bending mode from the first torsional mode of the
beam, as explained in section 4.3.1.
At its fixation, the beam was clamped between two steel plates on a length of 250 mm.
In order to guarantee the electrical insulation of the system from the environment, two
thin PVC plates (2 mm thickness) were inserted between the beam and the clamping
plates. The low stiffness of the PVC plates increased the compliance of the system
thus contributing to the reduction of its bending eigenfrequencies. This property of the
experimental set-up was accounted for in the numerical model by introducing two thin
low stiffness plates (EP V C = 1GP a) in correspondence to the clamped area and setting
the appropriate boundary conditions, as shown in figure 5.4.
The updated model will subsequently be scaled up and extended to estimate the be-
havior of Empa’s footbridge after the planned adaptive upgrade, as described in the
previous section.
2240mm
Homogenized Beam
2000mm
PVC Plates
Discrete Masses
Figure 5.4: Geometry and boundary conditions of the homogenized beam model. The
mechanical properties assumed for the model components are listed in tables 5.1 and
5.2
102
The properties of the homogenized beam are rendered in the model as shown in ta-
ble 5.1. The measured and the calculated first bending eigenfrequencies f1bending are
shown at the bottom of the table the columns for the I-beam and the homogenized
beam, respectively.
The eigenfrequency of the system is initially calculated using the parameters shown
in table 5.1. The discrepancies between the calculated and the measured values for
the first bending eigenfrequencies in the low (non laminated) and the high stiffness
(laminated) states were resolved by updating the elastic constants (Exh and Gxy h) in
the model until an acceptable agreement between the measured and the calculated
eigenfrequencies was obtained. Both elastic constants were scaled by the same fac-
tor for the low stiffness state of the beam. To describe the stiffened state, only Exh
was modified, as the shear compliance of the beam is concentrated in the web. The
updated mechanical properties are listed in table 5.2. The corresponding calculated
first bending eigenfrequencies after the updating of the material properties are in good
agreement with the measured ones. Higher eigenfrequencies could not be determined
experimentally, so that no comparison was possible. The updated model is therefore
an acceptable representation of the actual beam.
The contribution of the beam and the CFRP elements to the overall stiffness of the
system is calculated as follows: In the low stiffness state, the contribution of the CFRP
elements is considered to be negligible, due to their very small second moment of area.
The stiffness of the system in this state is considered exclusively due to the GFRP host
structure. The difference between the frequency response in the low stiffness state and
in the high stiffness state is attributed to the contribution of the CFRP elements. The
updated elastic modulus of the CFRP bands is calculated dividing the stiffness change
by the second moment of area of the laminated bands.
For both stiffness states, the elastic properties are lower than the ones assumed initially.
In the low stiffness state, the difference between the elastic modulus stated by the man-
ufacturer and the updated modulus is in proportion smaller (approximately 13%) than
in the case of the CFRP bands (approximately 36%). In general, the deviations can be
attributed to the short length of the fixation of the cantilever, leading to a higher compli-
ance of the system, the inaccuracy of the stated material properties, and measurement
inaccuracies.
Finally, it is possible to estimate a ratio between the stiffening obtained by electro-
lamination of the CFRP elements and the maximum attainable stiffening based on the
homogenized model and the results shown in table 5.2. The maximal attainable eigen-
frequency by electrolamination is 12.30 Hz, this corresponds to an increase of only 31%
103
Table 5.1: Properties of the mechanical components of the investigated sandwich
beam, in the main axis, as declared by the suppliers and properties assumed in the
numerical model, also in the main axis. The elastic moduli in the transversal directions
(Eyh and Ezh were assumed to be 1/3 of the modulus in the main axis). f1bending is the first
bending eigenfrequency measured or calculated, for the I-beam or for the homogenized
beam, respectively.
Property I-beam, faces Model Beam, I-beam, lami- Model Beam,

not laminated faces not laminated faces laminated
nated faces
Length [m] 2.24 2.24 2.24 2.24
Height [m] 0.123 0.123 0.123 0.123
Width [m] 0.06 0.06 0.06 0.06
Efxaces [Pa] 210·109 - 210·109 -

Ebeam
x [Pa] 23·109 - 23·109 -
web
Gxy [Pa] 3·109 - 3·109 -
If aces [m4 ]1 2×1.37·10−11 - 2×3.10·10−7 -
Ibeam [m4] 3.1·10−6 - 3.1·10−6 -
Aweb [m2 ] 6.48·10−4 - 6.48·10−4 -
EItotal [N m2 ] 7.13·104 7.13·104 2.01·105 2.01·105

GAtotal [N] 1.95·106 1.95·106 1.95·106 1.95·106
Ih [m4 ] - 9.3·10−6 * 9.3·10−6

Ah [m2 ] - 7.38·10−3 - 7.38·10−3
Ehx [Pa] - 7.7·109 * 21.6·109
Ghx [Pa] - 263.4·106 - 263.4·106
Weight/unit 2.92 2.92 2.92 2.92

length[kg/m]
f1bending [Hz] 10.40(I0 = 1A) 11.36 15.33(I0 = 1A) 17.92
1
The contribution of the faces to the overall moment of area depends on the location of the neutral axis
(within the faces, in the inactive state or coinciding with the neutral axis of the beam in the activated
state.)
104
beam, as calculated from the updated numerical model. The elastic moduli in the
transversal directions (Eyh and Ezh were assumed to be 1/3 of the modulus in the main
axis). f1bending is the first bending eigenfrequency measured or calculated, for the I-
beam or for the homogenized beam, respectively.
Property I-beam, faces Model Beam, I-beam, lami- Model Beam,

not laminated faces not laminated faces laminated
nated faces
Length [m] 2.24 2.24 2.24 2.24
Height [m] 0.123 0.123 0.123 0.123
Width [m] 0.06 0.06 0.06 0.06
Ehx [Pa] - 6.6·109 * 15.6·109

Ghxy [Pa] - 223.9·106 - 223.9·106
Ih [m4 ] - 9.3·10−6 * 9.3·10−6
Ah [m2 ] - 7.38·10−3 - 7.38·10−3
EItotal [N m2 ] 6.14·104 6.14·104 1.45·105 1.45·105

GAtotal [N] 1.65·106 1.65·106 1.65·106 1.65·106
Ibeam [m4] 3.1·10−6 - 3.1·10−6 -

Aweb [m2 ] 6.48·10−4 - 6.48·10−4 -
If aces [m4 ]1 2×1.37·10−11 - 2×3.1·10−7 -
Ebeam
x [Pa] 19.8·109 - 19.8·109 -
web
Gxy [Pa] 2.6·109 - 2.6·109 -
Efxaces [Pa] 210·109 - 135·109 -
Weight/unit 2.92 2.92 2.92 2.92

length[kg/m]
f1bending [Hz] 10.40(I0 = 1A) 10.42 15.33(I0 = 1A) 15.43
1
state.)
105
of the difference in stiffness between the beam with non laminated CFRP elements and
the beam with laminated CFRP elements. This level of stiffening was attained at the
lowest level of exciting current (I0 = 0.5A). At high excitation levels, the shear stress
transfer at the interfaces is not sufficiently effective to obtain the same effect as at low
amplitudes. A more modest increase in stiffness is observed, while in return higher
damping levels could be attained.
5.2.2 Damped Behavior of the Beam
The previous section showed how the elastic parameters of the beam were updated in
the model, so as to obtain the same values for the 1st bending eigenfrequency as for the
investigated system. The parameters obtained through this procedure will be used to
estimate the behavior of the GFRP bridge deck in the next section. The previous chap-
ter also showed that the increase in the logarithmic decrement λ that can be achieved
by activating the interfaces is noticeable and that it contributes to the reduction of the
vibration amplitude of the beam in a significant way. The loss factor η is the commonly
used parameter to describe the damping behavior of a structure or material. The loss
factor is calculated from the logarithmic decrement λ as:
λ
η= (5.1)
π
Figure 4.9 shows that the strongest attenuation due to the application of the potential
Ui is observed for the highest excitation current I0 , with U0 = 500V . In this section, the
η and the stiffening values in the numerical model for an acceptable fit of the calculated
frequency response curves to the frequency responses measured at I0 = 3A with in-
terlaminar voltages U0 = 0V and U0 = 500V will be determined. These data will then
be used for calculating the effect of the activation of the GFRP-CFRP interfaces on the
behavior of the bridge deck.
The maximal attainable eigenfrequency by electrolamination at I0 = 3.0A is 11.35 Hz,
this corresponds to an increase of 17% of the difference in stiffness obtained by lami-
nation. Based on the estimates shown in figure 4.15, the loss factor of the GFRP-CFRP
beam can be varied approximately between 0.03 and 0.08, for I0 = 3.0A. These values
were used as a starting point for the calculation of the frequency response of the beam
at Ui = 0V and Ui = 500V , taking damping into account. The actual force acting on the
beam at the position of the coil/permanent magnet assembly used to excite the beam
is not known, as the system was not calibrated. This additional parameter was also
106
updated in a manual procedure. Figure 5.5 shows the frequency response curves cal-
culated based on the η values estimated in the previous section and an exciting force
amplitude F0 of 1 N. Based on the height and width of the low stiffess/low damping
curve the loss factor for the low stiffness system was reduced to η = 0.02 and the excit-
ing force to F0 = 0.53N . The resulting η value used to approximate the measurement
performed at Ui = 500V was 0.16.
As can be seen in figure 5.5, the frequency response curve calculated for the low
stiffness system is a quite good approximation for the system at Ui = 0V . Instead,
the high stiffness/high damping system is only partially approximated by the calculated
frequency response curve, because the overall shape of the measured frequency re-
sponse is non-linear. The reason for this behavior can be found in the onset of inter-
laminar slip once a certain amplitude is reached.
Nevertheless, the values for the increase in stiffness and the loss factor values found
with the described procedure give satisfactory approximations of the behavior of the
system at two different interlaminar voltage levels.
The values that will be used for the calculation of the frequency response of the GFRP
bridge deck described in the next section are listed in table 5.3.
101 101
a0 [ms-2]
Acceleration amplitude a0 [ms-2]

Acceleration amplitude
100 100
−
U =0V U =0V
i i
10-1 U i=500V
10-1
U =500V
i
η=0.08, ∆ EI=17% η=0.02, ∆ EI=0
η=0.03, ∆ EI=0 η=0.16 ∆ EI=17%
7 8 9 10 11 12 13 14 15 16 17 7 8 9 10 11 12 13 14 15 16 17
f[Hz] f[Hz]
Figure 5.5: Computed frequency response curves (dashed lines) before (left) and after
(right) updating of the exciting force and damping parameters, compared to the mea-
sured frequency response curves (solid lines)
107
beam, as calculated from the updated numerical model considering damping.
Ui = 0V Ui = 500V Laminated CFRP

∆EI 0% 17% 100%
η 0.02 0.16 0.02
5.2.3 Bridge
The goal of the present work is to devise a method to effectively attenuate the vibration
of structures by means of the electrolamination of stiffening elements. The numerical
calculations presented in this section will show the effect that the vibration suppres-
sion method is expected to have on the dynamic behavior of the bridge deck. Since
the calculations discussed in this section are meant as a preparation for the actual
experiments, it is necessary to make a prediction of the expected behavior based on
the experiments performed on the GFRP-CFRP beam. The extrapolations are based
on the assumption that the same scaling factors for the material properties that were
obtained through the model updating procedure used for the individual beam can be
applied to the bridge deck. The increase in bending stiffness that can be obtained by
laminating the CFRP elements to the GFRP structure is expected to have a smaller
impact on the dynamic behavior of the deck than it had on the behavior of the single
beam, due to the larger number of constraints given by the bearings and stay cables.
Figure 5.6 shows the simplified geometry used in the numerical model of Empa’s bridge.
The geometrical model captures the most important features of the structure such as
the overall size of the deck, the presence of three pairs of stays on the main span, the
angles between the stays and the deck and the pylon respectively. Other features were
rendered using equivalent properties, such as in the case of the three pairs of back
stays that were replaced by one pair with the triple stiffness. The pylon was modeled
only above the bridge deck, due to the limitations imposed by the two dimensional
character of the model. Also, details like the pivoting anchorages of the stays in the
bridge deck where disregarded. Especially the reduced length of the pylon and the
stiff connection of the stays to the bridge deck lead to a stiffer overall behavior of the
structure. The model is two-dimensional, since only the bending behavior of the bridge
deck is considered. Accordingly, the geometry was discretized using triangular and
rectangular 2D elements, as indicated in figure 5.6.
The bridge deck is made of two sets of longitudinal GFRP I-profiles similar to the beam
108
8m
-2m
-4m -2 0 2 4 6 8 10 12 14 16m
Figure 5.6: Simplified geometry used for the numerical modeling of Empa’s cable stayed
pedestrian bridge. The red arrow marks the position at which the displacement is sam-
pled to calculate the frequency response of the structure. In the red circles are exam-
ples of the 2D elements used to discretize the structures. Triangular and rectangular
2D elements were used.
described above and two sets longitudinal GFRP U-profiles. The joints along the lon-
gitudinal axis are not considered in the model, although they are expected to reduce
the overall stiffness of the deck. The total length of the deck is 19.2 m. The height
of the profiles is 200 mm and the width 100 mm for the I-profiles and 60 mm for the
U-profiles. The thickness of all profiles is 10 mm. Unlike the beam described in the
previous sections, the 1.4 mm thick CFRP elements are assumed to be embedded in
the GFRP profiles as shown in figure 5.2. The width of the CFRP elements is assumed
to be 60 mm. The homogenized bridge deck is assumed to be 1.6 m wide.
The mechanical properties of the GFRP bridge deck are summarized in table 5.4. The
total bending and shear stiffness values of the beams, calculated based on the me-
chanical properties stated by the manufacturers of the GFRP beams and CFRP bands
or found in literature are listed as Eth and Gth , respectively.
The homogenized values for E-moduli and G-modulus used for the calculations take
the updating factors calculated in sections 5.2.1 and 5.2.2 into account and are listed
109
1600
60 100
200
10
Figure 5.7: Cross-section of the load bearing elements of the bridge deck
in the table as Eupd. and Gupd. , respectively. The row marked as ∆ EI shows the change
in bending stiffness in relationship to the maximum increase that could be obtained by
laminating the CFRP bands: the stiffness of the deck with unlaminated CFRP elements
is defined as the base stiffness with a ∆ EI of 0%. 100% corresponds to the value calcu-
lated for the laminated elements. The stiffness for the deck with electrolaminated CFRP
elements is calculated as the base stiffness plus 17% of the difference between the two
extreme stiffness values, based on the results shown in section5.2.2. Additionally, the
loss factors used for the calculations are listed at the bottom of the table.
Figure 5.8 shows the frequency response functions calculated for the bridge deck (at
the position marked in figure 5.6)in three different states: unstiffened, with laminated
CFRP elements and with CFRP elements electrolaminated at Ui = 500V , based on
the properties extrapolated as per table 5.4. As expected, the frequency shift achieved
by the increase in bending stiffness is quite modest, due to the constraints set by the
bearings and the stay cables. A shift of the eigenfrequency of approximately 0.4 Hz
obtained by laminating the CFRP elements could only be an effective instrument for the
reduction of the vibration amplitude of the bridge deck, if the spectrum of the exciting
force is quite narrowly distributed.
While the response of the electrolaminated system shows an even more modest shift
of the eigenfrequency (0.2 Hz) than the laminated system, the maximum acceleration
amplitude is greatly reduced (by a factor of six to eight, compared to the other frequency
response curves). As expected, the introduction of a significantly higher damping has
a marked effect also on the highly constrained structure.
110
beam, as calculated from the updated numerical model The elastic moduli in the
transversal directions (Ey and Ez were assumed to be 1/3 of the modulus in the main
axis). f1bending is the first bending eigenfrequency calculated, for the homogenized
bridge deck at different stiffness states.
Property Profiles, Model Profiles, Model Model

faces not Deck, laminated Deck, Deck,
laminated faces not faces laminated electrolami-
laminated faces nated faces
(Ui 500 V)
Iprof iles [m4 ] 7.9·10−5 - 7.9·10−5 - -
Awebs [m2 ] 7.2·10−3 - 7.2·10−3 - -
If aces [m4 ]1 1.1·10−10 - 6.9·10−6 - -
Eprof iles [Pa] 28·109 - 28·109 - -
Gwebs [Pa] 3·109 - 3·109 - -
Ef aces [Pa] 210·109 - 210·109 - -
EIth. [N m2 ] 2.22·106 2.22·106 3.63·106 3.63·106 -

GAth. [N] 2.16·107 2.16·107 2.16·107 2.16·107 -
EIupd. [N m2 ] - 1.88·106 - 2.80·106 2.04·106
GAupd. [N] - 1.8·106 - 1.8·106 1.8·106
Ih [m4 ] - 1.07·10−3 - 1.07·10−3 1.07·10−3

Ah [m2 ] - 0.32 - 0.32 0.32
Eh [Pa] - 1.78·109 - 2.63·109 1.91·109
Gh [Pa] - 5.74·107 - 5.74·107 5.74·107
Weight/unit 66.6 66.6 66.6 66.6 66.6

length[kg/m]
∆ EI [ ] - 0% - 100% 17%
η[] - 0.02 - 0.02 0.16
f1bending [Hz] - 7.34 - 7.78 7.54
a0,max - 13.6 - 15.8 1.9
[ms−2 ]
1
state.)
111
2
10
Unlaminated CFRP
Electrolaminated CFRP,
U =500V
i
Acceleration amplitude a0 [ms-2]
Laminated CFRP
1
10
0
10
−1
10
6 6.5 7 7.5 8 8.5 9

f [Hz]
Figure 5.8: Frequency response functions calculated based on the properties listed in
table 5.4, for the bridge deck with unlaminated, electrolaminated and laminated CFRP
stiffening elements, respectively. For the laminated and unlaminated cases, the loss
factor η was assumed to be 0.02 while for the electrolaminated case, the loss factor
was assumed to be 0.16, as determined in the previous section.
5.3 Conclusions
In this chapter, two simple numerical models of the GFRP-CFRP cantilever beam in-
vestigated in chapter 4 and of Empa’s pedestrian bridge, respectively were presented.
By comparing the response of the modeled beam with the measured response, the rel-
evant mechanical properties of the materials implemented in the numerical model were
adapted so as to obtain a good agreement between model and experiment. Subse-
quently, the properties were extrapolated for the GFRP bridge deck. The calculated first
bending eigenfrequency for the unstiffened system (7.34 Hz) is in fairly good agreement
112
with the value of 6.59 Hz stated in literature [26], especially considered the simplifica-
tions made in the model that are expected to lead to a stiffer behavior than in the real
structure.
As with any extrapolation, the correctness of the predicted behavior can only be con-
firmed by experiments. Nevertheless, the results shown in figure 5.8 are in agreement
with the expectation that the frequency shift obtained by lamination of the CFRP el-
ements would be modest. The results obtained from the calculations for the system
equivalent to the electrolaminated structure give reason to expect that the damping
introduced by the interaction between the GFRP structure and the CFRP stiffening el-
ements will have a substantial and beneficial effect on the dynamic behavior of the
bridge.
113
Chapter 6
Conclusions and Outlook
6.1 Conclusions
The development of low density, high strength materials and composites (such as car-
bon fiber reinforced polymers) has made the design and construction of structures with
increasingly high performance possible. Due to their reduced mass and intrinsic damp-
ing, such systems are more prone to vibrations than before, thus making their dynamic
behavior likely to be an important limiting factor for their operations. The goal of the
present work was to demonstrate the use of electrostatic forces between layers of a
multi-layer structure as a means to modify its mechanical properties and thus its dy-
namic behavior.
Very limited work has been published on the topic of multi-layer system with variable
mechanical properties, to date. At the time when the present project was being de-
fined, only the work of Tabata [48] explicitly mentioned the use of electrostatic fields in
a multi-layer system as a method to modify its mechanical properties. Later, the work
of Kornbluh [35] generally described the possibilities that might be opened by the de-
velopment of materials or structures with variable mechanical properties. Only through
personal communication with the author was it possible to confirm that work in the same
direction as what is described in this report was at the base of the published report. At a
later stage in the project two patents by the same authors were found, describing some
practical aspects of the modification of the mechanical properties of structures [30, 34].
Lately, some additional work in the field of the modification of the bending stiffness of
simple structures, typically based on the modification of the elastic modulus of this poly-
mer layers intercalated between the stiffer load bearing layers of a structure [25] has
been published.
115
The lack of any substantial information about the function of the electrostatic modifica-
tion of the mechanical properties of multi-layers and experiments to support it, as well
as of any description of a practical application of such systems, warrants the investi-
gation described in this work. The main points of the work are addressed in the four
central chapters:
• Working Principle of the Electrostatic Tuning of the Stiffness:

In this chapter, a method for the modification of the mechanical properties of adap-
tive structures based on the electrostatic coupling of the layers of a multi-layer
system is described.
• Modeling of Structures with Tunable Stiffness:

In this chapter, an explanation of how the system works is given based on calcu-
lations made with analytical and numerical models.
• Experimental Work:
In this chapter proof of the functionality of the electrostatic modification of the me-
chanical properties of a structure is given. Indirect confirmation of the predictions
made based on the numerical and analytical models is given.
• A Practical Application of Elements with Tunable Mechanical Properties:

Finally, in this chapter an outlook on the applications of the described method is
given by making an estimate of the benefits that could be obtained thanks to the
retrofitting of the GFRP deck of a pedestrian cable stayed bridge.
In chapter 2 the fundamental idea is presented, that if it were possible to introduce

or remove interfaces in the cross section of a simple structure subjected to a bending
load, it would be possible to obtain a marked change of the bending stiffness of the
structure itself. This procedure would be equivalent to laminating the layers of a multi-
layer structure and vice versa. If only the energy needed for the creation of new surfaces
is considered, with surface energies of the order of 1 or 2 J/m2 the process is also
energetically very inexpensive. Furthermore, the reverse process (cancellation of the
interfaces) is in principle exothermic. The amplitude of the range of stiffness values that
can be realized by the switching of the topology can easily range in the order of a factor
of 102 . The reality is that processes that involve the creation or cancellation of interfaces
in a solid have generally an endothermic global balance and are ill suited for the kind
of adaptive applications envisioned in this work. The modulation of the shear stress
transfer at the interface of a multi-layer system is proposed as an alternative to true
topology switching. This assumption means that under ideal conditions it is possible
116
to replicate the behavior of a system in which interfaces are created and canceled
by switching the normal stress across interfaces on or off. The use of electrostatic
fields for the generation of the necessary normal stresses is a logical choice thanks to
the very limited weight penalty that has to be accepted and the ease of interfacing an
electrostatic system with a control system. This formulation of the working principle of
a structure with adaptable mechanical properties is the result obtained within this work
and represent the starting point for the work presented in the other chapters.
The calculations described in chapter 3 are primarily meant to demonstrate, without the
interference of unwanted effects that experimental data are inevitably affected by, that
the modulation of the shear stress transfer at the interfaces of a multi-layer structure
is equivalent to the switching of the topology of the structure from monolithic to multi-
layer and vice versa. A numerical model is used to take the effect of normal and shear
stresses at the interfaces into account. An analytical model shows the effect of the
interfaces on the shear stress distribution in the structure. A comparison of the calcu-
lated shear stress distributions shows that as long as the shear stress at the interfaces
is smaller than maximum transferable shear stress σxy,max = µ · σyy,el , the behavior of
the interfaces at which a normal stress is applied is equivalent to the behavior of a sys-
tem in which no interfaces are present. Hence, the hypothesis that the modification of
shear stress transfer is mechanically equivalent to a modification of the topology of the
system is confirmed.
At higher load levels, the presence of an interface subjected to a normal stress can be
seen again, in terms of shear stress distribution as the shear stress cannot be trans-
ferred in full and slipping at the interfaces starts occurring. This situation is at the origin
of the force-displacement hysteretic behavior of the system that causes energy dissi-
pation to take place. The system chosen to show the effect of topology switching on
the global and local behavior of a structure was the simplest possible in terms of geom-
etry and selection of materials, since all layers of the system had the same geometry
and were made of the same material. Electrostatically couple multi-layer structures are
thus expected to display two interesting properties: the ability to modify their bending
stiffness and, at high load levels, the ability to increase the damping ratio of the system.
The experimental verification of the proposed method for the modification of the me-
chanical properties of structural elements was focused on the investigation of sandwich
beams due their higher sensitivity to the shear stress transfer between the core and the
faces of the structure. Also, the high length to stiffness ratio of the faces makes the
system fairly insensitive to the adverse effect of the lack of planarity of the interfaces
presented in section 4.4. While it was not possible to determine directly the effect of the
117
face-core coupling on the shear deformation field, the results of the experiments per-
formed on the CFRP-Silicone-CFRP sandwich could confirm the relationship between
the applied potential Ui and the external load P (and thus the interfacial shear stress
σxy,int ) at which a softening of the system could be observed. This is interpreted as
a reasonable indirect confirmation of the mechanism and calculation presented in the
previous sections. The second set of experiments, performed on a fairly large CFRP-
GFRP-CFRP sandwich beam, were geared towards the estimation of the impact of the
electrolamination of stiffening elements onto a fairly rigid structure on its dynamic be-
havior. Additionally, this portion of the work yielded a useful base for the estimation
of the range in stiffness and damping ratio that can reasonably be achieved with the
chosen combination of core and face materials.
The system properties determined in the experimental part of this work, is finally used in
the last part of this work that describes a real application of the developed method: the
implementation of a novel vibration suppression method for the GFRP deck of Empa’s
pedestrian bridge. In the first place, the stiffness and damping properties determined in
the experiments performed on the CFRP-GFRP-CFRP beam were used in a homoge-
nized numerical model to estimate the frequency response of the beam. Subsequently
they were updated so as to resolve the discrepancies between the measured system
and the original numerical model. Finally based on the updated properties of the model
beam, the properties of the bridge deck model were calculated for three conditions of
the target system: the unstiffened bridge deck, the bridge deck stiffened by laminating
CFRP bands to it and finally the bridge deck with electrolaminated CFRP bands at con-
ditions that yield the highest damping factor. Due to the fairly high level of constraint
of the deck, the frequency shift obtained by laminating the CFRP bands is very limited.
Depending on the band width of the exciting spectrum, the predicted frequency shift of
approximately 0.5 Hz is likely not to be sufficient to suppress a high amplitude reso-
nant vibration by disrupting the resonance situation. The ability of the system to realize
quite high levels of damping represents a more viable approach to the suppression of
structural vibrations.
The use of a coulomb friction based damping system leaves the method open to wear
related issues, especially in view of the fact that the element subject to dry friction is
also subject to a considerable electrostatic load. The danger of damaging the dielectric
layer by the combined action of mechanical and electrostatic loads should be carefully
considered, especially if thin dielectric layers are used to optimize the normal stress
generated by the electrostatic field. The durability of the system, the effect of heat de-
velopment as a consequence of the friction on the reversibility of the property changes
of the interface system shall be taken in serious consideration in view of real life applica-
118
tions. Nevertheless, the behavior of the first of the two CFRP-GFRP-CFRP beams was
operated for several months under fairly severe conditions and gave only marginal signs
of deterioration, in spite of the unfavorable surface topography that lead to non-uniform
contact between the elements. Furthermore, the planarity of the contact surfaces of the
elements of the system is a property that has a strong influence on its performance.
In summary, the present work showed the development of a novel method for the mod-
ification of the stiffness and the damping ratio of structural elements. The work spans
from the description of a concept for the modification of the mechanical properties of
simple structures, to its experimental demonstration to the presentation of a real appli-
cation of the system. Its working principle could be explained using simple models that
describe the effect of normal stresses at the interfaces of the system. The performed
experiments indicate that the proposed principle and model can describe the observed
system behavior. Calculations performed to estimate the behavior of a structure that
makes use of the proposed vibration suppression method show that the obtained in-
crease of the damping ratio is more likely to have a remarkable effect on the vibration
amplitude of the structure than the increase in stiffness, due to the high number of
constraints in the structure.
6.2 Outlook
The present work has shown that the electrostatic modification of the stiffness and the
damping properties of adaptive structures has the potential to become an effective,
robust and relatively inexpensive method for the suppression of structural vibrations.
Next to the use of polymeric films as dielectric layers, as in this work, there are many
different materials and methods to exploit electrostatic stress for the implementation of
this method. This leaves a great deal of freedom in the optimization of performance
and costs for the realization of adaptive structures with variable stiffness and damping.
The issues that will need to be addressed in the development of electrostatically tunable
structures span from materials optimization to the design of tunable structures on the
macro scale.
At the smallest end of the scale is the optimization of dielectric layers for high energy
density (and thus high normal stress) and resistance to wear. Possible approaches will
include the use of particle toughening (e.g. with diamond particles that also have good
dielectric properties or with layered alumino-silicate nanoparticles).
The micro- and nanostructuring of the contact surfaces promises to make a significant
119
contribution to the ’conversion’ of normal stress into transferrable shear stress. Here the
hope is that by obtaining mechanical interlocking of surface features such as ridges, an
increased amount of shear stress can be transferred per unit normal stress. This will
extend the amount of external load under which the system can be expected to behave
elastically, when the interfaces are activated. Also in this case, both the mechanical
and dielectric properties of the materials used to structure the surfaces will determine
the usability and reliability of the system.
On a larger scale, the planarity of the contact surfaces has a strong influence on the
effective area of contact between the surfaces of the system. The requirements set
to the planarity of the interfaces increase as the stiffness of the layers increases. The
use of more refined manufacturing processes and of dielectric materials specifically
designed for applications with high field strengths will lead to an improved performance
of the system.
The above mentioned steps towards an optimization of the interfaces system used to
implement variable stiffness and damping structures are necessary steps towards a
real life use of this type of novel intrinsically adaptive structures.
Especially for the implementation of the friction damping in connection with the interlam-
inar slip, the performance of the system could probably be enhanced by appropriately
modulating the electrostatic slip so as to maximize the dissipation of energy. A model
describing the quantitative relationship between dissipated energy, interlaminar stress
and deformation of the system will be needed, in order to implement a suitable control
system.
Furthermore, direct confirmation of the influence of the modulation of the interface in-
teractions on the shear stress distribution in the structure will be sought using methods
such as speckle interferometry or other image correlation methods, once the techno-
logical issues limiting the domain within which the multi-layer systems behave as a
monolithic element will be addressed.
Finally, in the field of basic considerations, the investigation of the use of the elec-
trostatic modification of shear stress transfer at a low scale should be considered. If
applied on structures consisting of a large number of layers, each a few µm thick, the
concepts presented in this work are expected to lead to novel devices with interesting
properties. High damping and high stiffness changes are expected from these devices
that could provocatively be called ’matrixless composite materials’. An impression of
the possibilities of such materials is given in figure 6.1.
While these aspects are considered, the design and demonstration of structures that
120
Deform,
Activate
Interfaces
Low Stiffness High Stiffness

Figure 6.1: A proof of concept experiment showing the behavior of a structure made
of approximately 50 layers of polymer film (coated on one side with aluminum) before
and after an electrostatic field is applied. Such structures can potentially be used for
morphing applications.
take advantage of the variable stiffness and variable damping concept can be carried
out, based on currently available materials and processes.
The first step towards the application of the electrostatic modification of the mechanical
properties of structures will be made with the implementation of the method on Empa’s
pedestrian bridge. Within this project, technological aspects of the manufacturing of
electrostatically tunable structures, such as the production of sufficiently planar contact
surfaces and the application of suitable electrode/dielectric assemblies will be of special
interest. It is difficult to predict, how closely related to a real life application this project
is. Nevertheless, the construction of a demonstrator in the scale of Empa’s pedestrian
bridge is a necessary step towards the development of ’real life’ applications.
In the large scale, the use of variable damping to optimize the structural damping of
wind turbine rotorblades is currently being considered. The wind turbine industry is
potentially a good target for this application, as it makes use of technologically very
advanced materials and solutions and is not yet as strictly regulated as, for example,
the aero-space industry. Some efforts have already been made to increase the amount
of structural damping of rotorblades [14] by optimizing the materials used for these
structures. Possibly the additional damping that can be obtained with the method pre-
sented in this work could offer additional advantages in terms of fatigue resistance of
the structure.
At a considerably smaller scale, some considerations have been made about the use
of variable stiffness and variable damping in the cantilevers Atomic Form Microscopes.
Considerations on the effect of variable damping are presented by Chang et. al [13].
121
The challenges posed by the fabrication of such devices as multı̂-layer structures will
have to be assessed in the first place and put in relationship to the potential benefits.
122
Acknowledgements
The author wishes to thank:

Prof. Dr. Paolo Ermanni, for accepting to be the Examiner of this work and for his
support and interest.
Prof. Dr. Louis Schlapbach, Co-Examiner, for his interest in this work already in its
early stages and for his support.
Prof. Dr. Masoud Motavalli, Co-Examiner, for his support, for the encouragement to
pursue this work from a very early stage and for allowing for the time to do so.
The Structural Engineering Research Laboratory for kindly funding the work that is at
the origin of this thesis.
Mr. Rouven Christen, colleague and friend, for his invaluable help with discussions and
experimental work.
Dr. Felix Weber, for the interesting and useful discussions about damping and friction
in multi-layer systems and magnetorheological dampers.
Dr. Mark Melnykowycz, for so many good chats about engineering and photography
matters.
The Colleagues in the ’Bauhalle’ at the Structural Engineering Research Laboratory,
for their help preparing devices and experimental set-ups from small to big, so that
everything worked at its best.
Empa, for making it possible to do this work next to performing my regular duties in the
Structural Engineering Research Laboratory.
My family, for putting up with me in times of difficulty and doubt and for all the ’family
time’ that became ’research time’: I look forward to catching up!
Thank you to all of you, I could not have done this without your help!
123
List of Tables
1.1 Comparison of traditional vs. material based actuators (adapted from [41]) 16
1.2 Comparison of energy demand of the presented stiffening strategies . . . 29
2.1 Surface energy of some metals used in engineering applications [1] . . . 34
3.1 Main properties of the modeled system . . . . . . . . . . . . . . . . . . . 45
3.2 Interface activation configurations and bundle thicknesses ni . . . . . . . 47
3.3 Interface activation configurations, stiffness increase compared to the

000 configuration, comparison between analytical and numerical results
for q=100 N/m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1 Properties of the mechanical components of the ’proof of concept’ sand-

wich beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.2 Shear stresses calculated based on the sandwich theory . . . . . . . . . 73
4.3 Energy dissipated by the sandwich beam in one loading cycle. . . . . . . 74
4.4 Normal and shear stresses calculated using electrostatic attraction and
sandwich theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.1 Properties of the mechanical components of the investigated sandwich

beam, in the main axis, as declared by the suppliers and properties as-
sumed in the numerical model, also in the main axis. The elastic moduli
in the transversal directions (Eyh and Ezh were assumed to be 1/3 of the
modulus in the main axis). f1bending is the first bending eigenfrequency
measured or calculated, for the I-beam or for the homogenized beam,
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
1
beam, as calculated from the updated numerical model. The elastic mod-
uli in the transversal directions (Eyh and Ezh were assumed to be 1/3 of the
measured or calculated, for the I-beam or for the homogenized beam,
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
beam, as calculated from the updated numerical model considering damp-
ing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
beam, as calculated from the updated numerical model The elastic mod-
uli in the transversal directions (Ey and Ez were assumed to be 1/3 of the
calculated, for the homogenized bridge deck at different stiffness states. . 111
2
List of Figures
1.1 A time invariant system is designed to fulfill the envelope space of re-
quirements at all times. The use cases represent different operational
conditions that the structure will be exposed to during its operational life.
Examples of load cases could be ’take off’, ’high altitude cruise’ or ’land-
ing’, in the case of an aircraft. . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2 Schematic representation of an adaptive mechanical system with its base

components: structure, actuator and controller. Adapted from [5]. . . . . 13
1.3 Chart representing the most common forms of interaction between dif-
ferent physical domains, represented by the symbols in the circles: Q-
Thermal energy, hν- Energy in form of electromagnetic radiation, W -
Mechanical energy, µ- Energy in form of chemical potential, E- Energy
stored in an electrical field, H- Energy stored in a magnetic field. . . . . . 15
1.4 Mechanical impedance matching: (a) the actuator heats up as power is

not transferred to the load and (b) the power can be transmitted to the
load after pulley’s impedance is matched to the actuator. [41]. . . . . . . . 17
1.5 Schematic representation of an adaptive mechanical system with an ex-

tended set of components: structure, structure controller, actuator and
actuator controller. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Schematic representation of a passive adaptive mechanical system con-

sisting of structure and structure controller . . . . . . . . . . . . . . . . . . 18
1.7 Simple cantilever beam with cross-section b · h subject to a load F at a

distance l from the origin. . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.8 Effect of the morphing of the cross section geometry on the second mo-
ment of area of a beam. A comparison of the second moment of areas
for the different cross sections for a constant area A = a2 is given in (1.6). 23
3
1.9 Effect of the modification of the cross section geometry on the second
moment of area of a beam. A comparison of the areas and second mo-
ment of areas for the different cross sections is given in (1.7) and (1.8). . 25
1.10 Stress strain curve for a NiTi alloy adapted from [7], after [43] . . . . . . . 26
1.11 The modification of the topology of an I-beam causes a remarkable in-

crease of the stiffness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1 Effect of connectivity switching on the shear-stress distribution in the

cross-section of a cantilever beam . . . . . . . . . . . . . . . . . . . . . . 32
2.2 The resistance to shear of crystalline solids is given by the energy differ-
ence between the equilibrium state (a) and (c) and the state of maximum
deviation of the crystal lattice parameter d0 (b). Adapted from [33] . . . . . 34
2.3 At an atomic level, even the smoothest real surfaces present terraces,
ad-atoms, kinks, and other surface defects that make the real contact
surface smaller than shown in the idealization of figure 2.2. Adapted
from [47] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4 Real surfaces, as the ones found in engineering structures are rough.
The surface profile is generally random and shows features at different
scales of size. Adapted from [47] . . . . . . . . . . . . . . . . . . . . . . . 36
2.5 A simple representation of the friction between solid surfaces . . . . . . . 37
2.6 The application of an electrostatic field between surfaces of the layers of

a multi-layer structure generates normal stress at the interface. Here, the
layers are assumed to be non-conductive, hence an electrode bonded to
the layer is necessary at the interface. . . . . . . . . . . . . . . . . . . . . 38
3.1 Two cross sections with the same geometry and different topologies. The
topologically connected cross section (right) has one neutral axis (n.a.),
while the disconnected one (left) has four (n.a.1 ...n.a.4 ). On the bottom
left the local coordinates for layer i are shown. . . . . . . . . . . . . . . . 42
3.2 Model considered for the numerical calculation of the effect of shear
stress transfer control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.3 Contact stress σyy,contact as a function of the gap size as calculated using
the penalty barrier approach described in (3.10) . . . . . . . . . . . . . . . 46
4
3.4 Transversal displacement at x=l, as calculated using the linear (analyt-
ical) model and the non-linear (numerical) model, as a function of the
applied load q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.5 Shear stress distribution in the cross section at different positions along
the length of the beam with a 111 interface activation pattern, for a load
q=280 N/m. The numerically calculated distribution (markers) at x=0.01
m shows a slight deviation from the linear model (solid lines). . . . . . . . 49
q=320 N/m. The numerically calculated distribution at x=0.01 m (mark-
ers) shows a minor deviation from the linear model (solid lines). . . . . . . 50
q=360 N/m. The numerically calculated distributions at both x=0.01 m
and x=0.02 m (markers) show a clear deviation from the linear model
(solid lines). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
q=400 N/m. The numerically calculated distributions at x=0.01 m and
x=0.02 m (markers) show a remarkable deviation from the linear model
(solid lines). Also at x=0.04 m the distribution does not correspond to the
linear behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
q=500 N/m. None of the shear stress distributions calculated numeri-
cally (markers) coincides with the linear model (solid lines), although the
results for x=0.04 m and x=0.05 m do not exceed σxy,max . . . . . . . . . . 51
3.10 The cyclic load-displacement diagrams calculated for three different lev-
els of interfacial stress (3 MPa, 6 MPa and 9 MPa) show different amounts
of hysteresis in the system. The load displacement diagram of the un-
coupled system is drawn in black, for comparison. The uncoupled system
does not have hysteresis, since no interaction between layers is allowed
for. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5
3.11 Transversal deformation functions for the beams with different interface
activation patterns and a load q=100 N/m, as calculated analytically (line)
and numerically (markers) . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.12 Shear stress distribution in the cross section at x=0.01 m,... 0.07 m for
the beams with ’000’ interface activation pattern and a load q=100 N/m,
as calculated analytically (line) and numerically (markers) . . . . . . . . . 54
3.17 The stiffness of the beams with each connectivity pattern can be repre-
sented by an equivalent parallel connection of springs The stiffness D of
each beam is expressed in multiples of k, where k is the stiffness of one
layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.18 Stiffness distribution (left) and stiffness density (right) for n=20 . . . . . . 59
3.19 Stiffness states distributions for n=5, 10, 15, 20 and 25. . . . . . . . . . . 60
4.1 Test set-up for the measurement of force-displacement diagrams of the

electrostatically tunable sandwich beam (left: schematic representation,
right: photograph). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2 Force-displacement curves for the initial loading as a function of the elec-
trical potential across the face-core interface. Measured and calculated
displacements for the fully bonded sandwich are in good agreement and
show higher stiffness than an electrostatically stiffened bundle at high
voltage (3000 V). The 2500 V measurement is not shown as it is practi-
cally identical with the 3000 V measurement. . . . . . . . . . . . . . . . . 70
6
4.3 Behavior of the system for two complete cycles (after the initial loading)
at potential Ui =1000 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.4 Behavior of the system as a function of the electrical potential across the
face-core interface. After the initial loading, the force-displacement dia-
grams show a hysteretic behavior. The stiffness of the system increases
with increasing potential Ui . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.5 Force at which the regression lines calculated for the force-displacement
diagrams shown in figure 4.4 intersect. The factor 3.2·10−6 mN V−2 is in
fairly good agreement with the value of 3.45·10−6 mN V−2 calculated on
the basis of electrostatic attraction forces using (4.8), (2.3) and (4.14). . . 75
4.6 Normalized bending lines at Ui = 0 and 3000 V respectively. The 0 V line

appears to have a higher curvature than the 3000 V line due to a higher
shear contribution (linear in x) to the displacement at higher voltages, as
expected considering (4.6). . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.7 Overview of the test set-up and cross section of a GFRP I beam as used
in the experiments described in section 4.3.2 . . . . . . . . . . . . . . . . 78
4.8 Amplitude and phase spectrum of the accelerations (sensor 1, at the left
edge of the beam and sensor 2 at the right edge) measured at the free
end of the beam as a function of frequency for I0 = 2.6 A. . . . . . . . . . . 79
4.9 The change of the transfer function measured at different Ui levels and
excitation current amplitudes indicates a change in the shear stress trans-
fer at the interfaces of the system. The transfer function of a beam with
laminated CFRP elements is plotted in black in the I0 =1 A graph for
reference. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.10 Overview of the eigenfrequency values of the system at different inter-

laminar voltage levels and current amplitudes (left) and of the maximum
values max(|G|) (right). For low excitation currents the lowest value of
max(|G|) is at intermediate voltage levels, while for high excitation cur-
rents, the minimum value coincides with the highest voltage levels that
could be applied. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.11 Displacement as a function of time and in red its envelope (top) and
excitation current as a function of time for a weakly excited beam and no
interlaminar potential. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7
4.12 Logarithm of the envelopes ln(H(v(c))) of the measured displacement
as a function of number of cycles c for excitation amplitudes between 0.5
A and 3.0 A and interlaminar voltages between 0 V and 800 V . . . . . . 85
4.13 A comparison of the measured decay envelopes with their logarithmic

and linear approximation, between the start point and end point for decay
curves measured at the lowest and at the highest excitation level. . . . . . 86
4.14 The graphs show the calculated instantaneous logarithmic decrement as

a function of the number of cycles after the begin of the amplitude decay.
Results for different excitation amplitudes I0 and interlaminar potentials
Ui are shown. The circles on the right hand side of the graphs represent
the global logarithmic decrements calculated based on the whole decay
curves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.15 The effect of electrostatic coupling on the first bending eigenfrequency

and the logarithmic decrement of the described beam is a function of the
interlaminar potential Ui and the excitation amplitude I0 , as shown in the
two plots of this figure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.16 A view of the top flange of a GFRP beam , used in a tunable GFRP CFRP
structure as described in the previous sections, after approximately two
years of (non-continuous) operation. The areas marked in black pre-
sented clear signs of wear due to friction between the surface of the
dielectric layer with the CFRP bands. All other areas appeared to be in
pristine condition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.17 Local unevenness vs. long range unevenness of the surface . . . . . . . . 91
4.18 The CFRP stiffening element must adapt to the topography of the host
structure, so that contact at the interfaces is possible and shear stress
can be transferred between the two components by means of friction. . . 92
4.19 The distributed load needed for the CFRP element to conform to the
topographical features of the GFRP beam for a given feature depth in-
creases steeply for short length features. The red line shows the approx-
imate electrostatic stress that can be generated across an 80µm thick
PVDF film applying a potential of 3000 V or across a 12µm thick PET
film applying a potential of 800 V. . . . . . . . . . . . . . . . . . . . . . . . 94
8
5.1 Empa’s full scale lightweight pedestrian bridge is used as a test platform
for structural health monitoring and vibration mitigation developments.
Left, general view. Right, detail from the bottom . . . . . . . . . . . . . . . 98
5.2 Embedded, electrostatically couplable CFRP elements for the stiffening
and damping of the GFRP bridge deck . . . . . . . . . . . . . . . . . . . . 99
5.3 Mechanical components of the switchable stiffness beam used in the
experiments (left) and homogenized equivalent used for the numerical
model (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.4 Geometry and boundary conditions of the homogenized beam model.
The mechanical properties assumed for the model components are listed
in tables 5.1 and 5.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.5 Computed frequency response curves (dashed lines) before (left) and
after (right) updating of the exciting force and damping parameters, com-
pared to the measured frequency response curves (solid lines) . . . . . . 107
5.6 Simplified geometry used for the numerical modeling of Empa’s cable
stayed pedestrian bridge. The red arrow marks the position at which
the displacement is sampled to calculate the frequency response of the
structure. In the red circles are examples of the 2D elements used to
discretize the structures. Triangular and rectangular 2D elements were
used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.7 Cross-section of the load bearing elements of the bridge deck . . . . . . . 110
5.8 Frequency response functions calculated based on the properties listed
in table 5.4, for the bridge deck with unlaminated, electrolaminated and
laminated CFRP stiffening elements, respectively. For the laminated and
unlaminated cases, the loss factor η was assumed to be 0.02 while for
the electrolaminated case, the loss factor was assumed to be 0.16, as
determined in the previous section. . . . . . . . . . . . . . . . . . . . . . . 112
6.1 A proof of concept experiment showing the behavior of a structure made

of approximately 50 layers of polymer film (coated on one side with alu-
minum) before and after an electrostatic field is applied. Such structures
can potentially be used for morphing applications. . . . . . . . . . . . . . 121
9
10
Bibliography
[1] S. E. Andryushechkin and M. G. Karpman. Method for processing calculated and

expemental values of the specific surface energy of transition metals. Metal Sci-
ence and Heat Treatment, 37:308–311, 1995.
[2] A. Belloli and P. Ermanni. Optimum placement of piezoelectric ceramic modules

for vibration suppression of highly constrained structures. Smart Materials and
Structures, 16:1662–1671, 2007.
[3] A. Bergamini, R. Christen, and M. Motavalli. Electrostatic tuning of the bending

stiffness of a large scale gfrp-cfrp beam. In Proceedings of the SPIE - The In-
ternational Society for Optical Engineering, volume vol.6169, pages 616919/1–8,
San Diego, Ca, 2006. SPIE - The International Society for Optical Engineering.
Conference Paper.
[4] M. Birchmeier, Brunner A., Paradies R., and J. Dual. Active fiber composites as
orthotropic piezoelectric transducers for structural health monitoring applications:
experimental approach. In Proc. Intern. Workshop Smart Materials and Structures
- Cansmart, Toronto, Canada, volume Proc. Intern. Workshop Smart Materials and
Structures - Cansmart, Oct. 2006.
[5] Robert H. Bishop, editor. Mechatronics : an introduction. Taylor & Francis, Boca
Raton, FL, 2006.
[6] J. A. Blaya and H. Herr. Adaptive control of a variable-impedance ankle-foot or-

thosis to assist drop-foot gait. Ieee Transactions on Neural Systems and Rehabili-
tation Engineering, 12(1):24–31, 2004. IEEE Trans. Neural Syst. Rehabil. Eng.
[7] V. Brailovski, S. Prokoshkin, P. Terriault, and F. Trochu. Shape Memory Alloys:

Fundamentals, Modeling and Applications. Université du Québec, Ecole de tech-
nologie Supérieure, 1st edition, 2003.
11
[8] M. J. Brennan. Actuators for active vibration control-tunable resonant devices. In
4th Euro Conf on Smart Structures and Materials,, pages 41–48, 1998.
[9] L.L. Bucciarelli. Engineering mechanics for structures. Online

course material, http://dspace.mit.edu/bitstream/handle/1721.1/36381/1-
050Fall-2002/OcwWeb/Civil-and-Environmental-Engineering/1-050Solid-
MechanicsFall2002/Readings/index.htm, Fall 2002.
[10] Ilene J. Busch-Vishniac. Electromechanical sensors and actuators. Springer,

1999.
[11] L. F. Campanile, R. Paradies, and A. Bergamini. Formadaption an aktiven ma-

terialsystemen unter nutzung aeroelastischer koppeleffekte, empa internal project
2006. -, 2006.
[12] F. Carpi, A. Tralli, D. De Rossi, and P. Gaudenzi. Martian jumping rover equipped
with electroactive polymer actuators: A preliminary study. Ieee Transactions on
Aerospace and Electronic Systems, 43(1):79–92, 2007.
[13] Win-Jin Chang, Te-Hua Fang, and Huann-Ming Chou. Effect of interactive damp-
ing on sensitivity of vibration modes of rectangular afm cantilevers. Physics Letters
A, 312:158–165, 2003.
[14] P. K. Chaviaropoulos, E. S. Politis, D. J. Lekou, N. N. Sørensen, M. H. Hansen,

B. H. Bulder, C. Winkelaar, D.and Lindenburg, D. A. Saravanos, T. P. Philippidis,
C. Galiotis, M. O. L. Hansen, and T. Kossivas. Enhancing the damping of wind
turbine rotor blades, the dampblade project. WIND ENERGY, 5:163–177, 2006.
[15] M. Cho and S. Kim. Structural morphing using two-way shape memory effect of
sma. International Journal of Solids and Structures, 42(5-6):1759–1776, 2005. 21
PERGAMON-ELSEVIER SCIENCE LTD.
[16] Fiberline Composites. Catalog of standard profiles,. Online Profile Catalog,

http://www.fiberline.com/d/products/profileDetails.asp?prodID=18025&context=1,
2007.
[17] C. Czaderski, B. Hahnebach, and M. Motavalli. Rc beam with variable stiffness

and strength. Construction and Building Materials, 20(9):824–833, 2006.
[18] W. W. Clark (Editor). Special issue on energy harvesting. Journal of Intelligent

Material Systems and Structures, 16(10):783 – 901, 2005. J. Intell. Mater. Syst.
Struct.
12
[19] M. Friswell and J.E. Mottershead. Finite Element Model Updating in Structural
Dynamics. Springer, 1995.
[20] F. Gandhi and S.G. Kang. Beams with controllable flexural stiffness. Smart Mate-
rials and Structures, 16:1179–1184, 2007.
[21] X. J. Gao, D. Burton, T. L. Turner, and L. C. Brinson. Finite element analysis

of adaptive-stiffening and shape-control sma hybrid composites. Journal of En-
gineering Materials and Technology-Transactions of the Asme, 128(3):285–293,
2006.
[22] L. Gaul, J. Lenz, and D. Sachau. Active damping of space structures by contact
pressure control in joints. Mechanics of Structures and Machines, 26(1):81–100,
1998. MARCEL DEKKER INC.
[23] L. Gaul and R. Nitsche. Friction control for vibration suppression. Mechanical
Systems and Signal Processing, 14(2):139–150, 2000. ACADEMIC PRESS LTD.
[24] L. Gaul and R. Nitsche. Vibration control by interface dissipation in semi-active

joints. Zeitschrift Fur Angewandte Mathematik Und Mechanik, 80:S45–S48, 2000.
WILEY-V C H VERLAG GMBH Suppl. 1.
[25] B. O. Gordon and W. W. Clark. Morphing structures by way of stiffness varia-

tions. In 49th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and
Materials Conference, volume 1, pages 183–191, 2007.
[26] D. Gsell and M Motavalli. Indoor cable stayed gfrp-bridge at empa, switzerland.
In 4th International Conference on Advanced Composite Materials in Bridges and
Structures, Calgary, AL, Canada, 2004.
[27] H. Herr. Presentation highlights: Prosthetic and orthotic limbs. Journal of Reha-
bilitation Research and Development, 39(3):11–12, 2002.
[28] L. Janke, C. Czaderski, M. Motavalli, and J. Ruth. Applications of shape memory

alloys in civil engineering structures - overview, limits and new ideas. Materials
and Structures, 38 (5):578–592, 2005.
[29] T. J. Johnson and W. W. Clark. Harvesting energy from piezoelectric material.

Ieee Pervasive Computing, 4(1):70–71, 2005. IEEE Pervasive Comput.
[30] S. Kawamura, O. Tabata, and S: Konishi. Motion restraing tool and simulator,
2002.
13
[31] M.R.F Kidner and M.J. Brennan. Varying the stiffness of a beam-like neutralizer
under fuzzy logic control. Transactions of the ASME, 124:90–99, 2002.
[32] H. Y. Kim and W. Hwang. Effect of debonding on natural frequencies and fre-
quency response functions of honeycomb sandwich beams. Composite Struc-
tures, 55(1):51–62, 2002. 508QN COMPOS STRUCT.
[33] C. Kittel. Einführung in die Festkörperphysik. R. Oldenbourg Verlag, München,

12th edition, 1999.
[34] R. Kornbluh and al. Mechanical meta-materials, patent us 2006/0192465a1, 2006.
[35] R. D. Kornbluh, H. Prahlad, R. Pelrine, S. Stanford, M. A. Rosenthal, and P. A. von

Guggenberg. Rubber to rigid, clamped to undamped: toward composite materials
with wide-range controllable stiffness and damping. In Proceedings of the SPIE
- The International Society for Optical Engineering, volume vol.5388, no.1, pages
372–386, 2004. Conference Paper.
[36] S. C. Lim and S. B. Choi. Vibration control of an hdd disk-spindle system utilizing
piezoelectric bimorph shunt damping: I. dynamic analysis and modeling of the
shunted drive. Smart Materials & Structures, 16(3):891–900, 2007.
[37] G.; Ermanni P. Lochmatter, P.; Kovacs. Design and characterization of shell-like
actuators based on soft dielectric electroactive polymers. Smart Materials & Struc-
tures, 16:1415–1422, 2007.
[38] S.; Kovacs G. Lochmatter, P.; Michel. Electromechanical model for static and dy-
namic activation of elementary dielectric elastomer actuators - art. no. 61680f. In
Y. BarCohen, editor, Smart Structures and Materials 2006 Conference, 2006.
[39] K. Moser, A. Bergamini, R. Christen, and C. Czaderski. Feasibility of con-

crete prestressed by shape memory alloy short fibers. Materials and Structures,
38(279):593–600, 2005. 27 R I L E M PUBLICATIONS.
[40] Comsol Multiphysics. Structural Mechanics Module User’s Guide. COMSOL, AB,
version 3.3 edition, 2006.
[41] José L. Pons. Emerging Actuator Technologies: A Micromechatronic Approach.

Wiley, 2005.
[42] K. Ren, Y. M. Liu, H. Hofmann, Q. M. Zhang, and J. Blottman. An active en-

ergy harvesting scheme with an electroactive polymer. Applied Physics Letters,
14
91(13):3, 2007. Ren, Kailiang Liu, Yiming Hofmann, H. Zhang, Q. M. Blottman,
John 15 AMER INST PHYSICS.
[43] T. Saburi, M. Yoshida, and S. Nenno. Deformation behavior of tini alloy crystals.
Scripta Metallurgica, 18(9):363–366, 1984.
[44] G. Schandl, P. Lugner, C. Benatzky, M. Kozek, and A. Stribersky. Comfort en-

hancement by an active vibration reduction system for a flexible railway car body.
Vehicle System Dynamics, 45(9):835–847, 2007.
[45] R.; Schirmacher. The use of active vibration control for the reduction of ice interior
noise. Zeitschrift Fur Angewandte Mathematik Und Mechanik, 81:S41–S44, 2001.
4 WILEY-V C H VERLAG GMBH Suppl. 1.
[46] R. Schutze, H. C. Goetting, E. Breitbach, and T. Grutzmacher. Lightweight

engine mounting based on adaptive cfrp struts for active vibration suppres-
sion. Aerospace Science and Technology, 2(6):381–390, 1998. 11 GAUTHIER-
VILLARS/EDITIONS ELSEVIER.
[47] A.W. Stachowiak, G.W. Batchelor. Engineering Tribology. Elsevier Butterwirth

Heinemann, third edition edition, 2005.
[48] O. Tabata and K. et al. Konishi. Micro fabricated tunable bending stiffness devices.
Sensors and Actuators A, 89:119–123, 2001.
[49] M. Tanaka, M. Todoh, and Naomi. A. Adaptive stiffness design for multimaterial
structural system. Journal of Intelligent Material Systems and Structures, 14:323–
329, April-May 2003.
[50] F. Weber, G. Feltrin, and M. Motavalli. Passive damping of cables with mr dampers.
Journal Material and Structures, 38(279):595–600, June 2005.
[51] F. Xia. Electroactive polymer based microfluidic pump. Sensors and Actuators
a-Physical, 125(2):346–352, 2006. 28 ELSEVIER SCIENCE SA.
[52] Bingen Yang. Stress Strain and Structural Dynamics. Elsevier Academic Press,
2004.
[53] D. Zenkert. An Introduction to Sandwich Construction. The chamaleon Press,

Ltd., London, UK, 1995.
15
Curriculum Vitae
Personal
Name: Andrea Bergamini
Born: April 16 1969
Nationality: Italian
Status: Married, 1 child
Education
2005-2008 EPFL/ETHZ Lausanne/Zurich, CH Doctoral Degree in Mechanical Engi-

neering
1988-1994 ETH Zurich, CH Degree in Materials Engineering, Dipl. Ing. ETH, With
working experience at IGI, Inc. Nashua, NH, USA, Centre for Biomimeitics, Uni-
versity of Reading, UK; and Schwarz PharmaAG, Mohnheim, D.
Working Experience
1998-present EMPA, Materials science and technology, Dübendorf, CH Test and De-
velopment Engineer
1996-1998 ETH, Institute for Building Materials, Zurich, CH Research Assistant
1995-1996 NB&A, Paris, France Technical Consultant
17
Publications with peer review
1. Christen, R., Bergamini, A., Motavalli, M.,Influence of steel wrapping on magneto-

inductive testing of the main cables of suspension bridges, NDT&E Int (2008), in
press: doi.10.1016/j.ndteint.2008.08.003
2. Christen, R., Bergamini, A., Motavalli M., Three-Dimensional Localization of De-

fects in Stay Cables Using Magnetic Flux Leakage Methods. Journal of Non de-
structive Evaluation, September 2004. 22(3): p. 93-101
3. Christen, R., Bergamini, A., Automatic flaw detection in NDE signals using a panel
of neural networks. NDT&E International, October 2006. 39(7): p. 547-553
4. Bergamini, A., Christen, R., Motavalli, M., Sandwich beam with electrostatically
tunable bending stiffness. Smart Materials and Structures, 15 (2006): p. 678-686
5. Bergamini, A., Christen, R., Motavalli, M., Electrostatically tunable bending stiff-
ness in a GFRP-CFRP composite beam. Smart Materials and Structures, June
2007. 16(3): p. 575-582
6. Moser, K., Bergamini,A., Christen, R., Czaderski, C., Feasibility of concrete pre-
stressed by shape memory alloy short fibers. Materials and Structures, June
2005. Vol.38-No.279: p. 593-600
7. Christen, R., Bergamini, A., Wakili, K.G., Full scale, low temperature mechanical
testing of pre-stressing systems. Experimental Mechanics, February 2005. 45(1):
p. 96-100
8. Christen, R., Bergamini, A., Motavalli M., Three-Dimensional Localization of De-

fects in Stay Cables Using Magnetic Flux Leakage Methods. Journal of Non de-
structive Evaluation, September 2004. 22(3): p. 93-101
Publications without peer review / Conference Proceed-

ings
1. Bergamini, A., Christen, R., Motavalli, M. Tunable stiffness structures for vibration
suppression and other applications, in ICAST 2007, October 2007. Ottawa, CA
18
2. Christen, R., Bergamini, A., Novakovic, V., Motavalli, M. Filtering of NDT signals
obtained from wrapped steel cables. in SPIE Conference 2007, Progress in Non-
destructive Evaluation. March 2007. San Diego CA, USA.
3. Bergamini, A., Christen, R., Motavalli, M. A numerical model for the prediction of
the stiffness of a homogeneous electrostatically tunable multilayer beam, in SPIE
Conference 2007, Damping and Isolation. March 2007. San Diego CA, USA.
4. Bergamini, A., Christen, R., Motavalli, M. Electrostatic Tuning of the Bending Stiff-
ness of a large scale GFRP-CFRP Beam. in SPIE Conference 2006, Damping
and Isolation. February 2006. San Diego CA, USA.
5. Bergamini, A., Christen, R., Motavalli, M. Electrostatic tuning of the bending stiff-
ness of simple, slender multi-layer composite structures. in SPIE Conference
2005, Smart Structures and Materials. February 2005. San Diego CA, USA.
6. Bergamini, A., Feltrin, G., ”Non Destructive Inspection of Stay cables / Structural
Engineering Research Laboratory at Empa”, SAMCO Newsletter, 2004, pp. 4-8
7. Bergamini, A., Christen, R. Non-destructive inspection of steel cables with large

diameters for cable-stayed bridges. May 2004. ASTRA Report No 574
8. Bergamini, A., Christen, R., Motavalli, M. A simple approach to the automatic

recognition of flaws in large diameter steel cables. in International Symposium
NDT-CE 2003. September 2003. Berlin
9. Bergamini, A., Christen, R. Damage Detection in Cables. in SAMCO Summer

Academy. June 2003. Robinson College Cambridge University U.K.
10. Bergamini, A., Christen, R., Motavalli, M. A Simple Approach to the Localization
of Flaws in Large Diameter Steel Cables. in SPIE Conference 2003, Smart Non-
destructive Evaluation and Health Monitoring of Structural and Biological Systems
II. February 2003. San Diego CA.
11. Bergamini, A. Nondestructive Testing of Stay Cables - Field Experience in South

East Asia, Proceeding 3rd World Conference on Structural Control, Como, Italy,
Vol. II, p. 1057-1064, April 2002
12. Bergamini, A./ Motavalli, M. ”Laboratory and Field Investigation of Cable Systems
at EMPA, From System Acceptance Testing to the In Situ Non-destructive Eval-
uation of Stay Cables”, fib-CH Publication, The First fib congress 2002, Osaka,
Japan
19
13. Bergamini, A. Nondestructive Testing of Stay cables, Proc. IABSE Conference on
Suspended Bridges, Seoul, Korea, June 8-14, 2001
14. Bergamini, A. High Fracture Energy, High Ductility, Proceeding RILEM-WTA Sem-
inar on High Performance Cement-Based Materials, Zürich, Switzerland, Aedifi-
catio Verlag - Fraunhofer IRB Verlag
Patents
• CH Patent. International Publication No. WO 02/06812 A1 (2001): Method and

Device for the non-destructive examination of steel cables in anchorages
• CH Patent, Novel cement based material, Application Nr. 01935/04, Swiss Fed-
eral Institute for Intellectual Property, Nov 24th 2004
• CH Patent, Auf Eisen, Mangan und Silizium basierende Formgedächtnislegierung,

Application Nr. 00968/08, Swiss Federal Institute for Intellectual Property, June
25th 2008
20

Eth 41572 02

Uploaded by

Copyright:

Available Formats

Eth 41572 02

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Eth 41572 02

Uploaded by

Copyright:

Available Formats

ETH Library

Electrostatic modification of the

Electrostatic Modification of the

A dissertation submitted to the

for the degree of

accepted on the recommendation of

Prof. Dr. L. Schlapbach, co-examiner

Prof. Dr. M. Motavalli, co-examiner

2 Working Principle of the Electrostatic Tuning of the Stiffness 31

3 Modeling of Structures with Tunable Stiffness 41

5 A Practical Application of Elements with Tunable Mechanical Properties 97

C capacitance of a parallel plate condenser

1.1 Adaptive Structures

Ideal Design Envelope for Time Invariant Systems

Requirement 6 Use Case b

1.1.1 Material Based Actuators

semi-active In semiactive actuators the work exchange can only be negative, dW 6 0.

passive adaptive The concept of passive adaptive structure applies to non-actuated

Traditional actuators Transducer material based actuators

1.1.2 Host Structure

Figure 1.5: Schematic representation of an adaptive mechanical system with an ex-

Figure 1.6: Schematic representation of a passive adaptive mechanical system con-

the amplitude of its vibrations.

where ωk is the k th eigenfrequency of the system, βk is the k th solution of the equation

Figure 1.7: Simple cantilever beam with cross-section b · h subject to a load F at a

1.2.1 Modification of the Elastic Modulus

W = Wth + W pt = m · ∆T · cp + ∆Gpt (1.3)

Aundef ormed = Adef ormed (1.7)

d + hweb 2 dh3web d + khweb 2 dk 2 h3web

hweb h k hweb k hweb+d

Aundeformed, Iz,undeformed Adeformed, Iz,deformed

Considering the properties of multi-layer composite structures reveals immediately that

with b, d, h and hweb as shown in figure 1.11.

for the one used in section 4.3:

Dconnected = E · 2291d4 (1.13)

Econnectivity = 2 · A · esurf = 4 · 2.5m · 0.00513m · 1.5Jm−2 = 0.077J (1.15)

Table 1.2: Comparison of energy demand of the presented stiffening strategies

Method Obtained increase in stiffness Required energy [kJ]

1.3 Goals and Structure of the Present Work

Working Principle of the Electrostatic

2.1 Connectivity, Interfaces and Shear Stress Transfer

• Small weight penalty

• Short reaction time

2.2 Shear Stress Transfer and Stiffness

Disconnected σxy (interface)=0

2.3 Shear Stress Transfer in Crystalline solids

σxy = G · ²xy (2.1)

(a) (b) (c)

2.4 Creation and Cancellation of Interfaces

Metal Atomic Number Surface Energy Surface Energy

1mm of a roller-bearing race

110 mm of a lathe bed

2.5 Friction at Real Interfaces and Shear Stress Trans-

1. There is a proportionality between the maximum tangential force before sliding

Figure 2.5: A simple representation of the friction between solid surfaces

2. The tangential friction force is proportional to the normal force in sliding

3. Friction force is independent of the apparent contact area

4. Friction is independent of the sliding speed

2.6 Electrostatic control of the Shear Stress Transfer