Sie sind auf Seite 1von 20

Viscoelastic Damping TechnologiesPart I:

Modeling and Finite Element


Implementation

C.M.A. Vasques

, R.A.S. Moreira

and J. Dias Rodrigues

Departamento de Engenharia Mecanica, Faculdade de Engenharia,


Universidade do Porto, Rua Dr. Roberto Frias s/n,
4200-465 Porto, Portugal
e-mail: cmav@fe.up.pt and jdr@fe.up.pt

Departamento de Engenharia Mecanica, Universidade de Aveiro,


Campus Santiago, 3810-193 Aveiro, Portugal
e-mail: rmoreira@ua.pt
Submitted: 05/03/2010
Accepted: 18/04/2010
Appeared: 29/04/2010
c HyperSciences.Publisher
Abstract: This is the rst of two companion articles addressing an integrated study on
the mathematical modeling and assessment of the eciency of surface mounted or embedded
viscoelastic damping treatments, typically used to reduce structural vibration and/or noise radi-
ation from structures, incorporating the adequate use and development of viscoelastic (arbitrary
frequency dependent) damping models, along with their nite element (FE) implementation,
and the experimental identication of the constitutive behavior of viscoelastic materials. This
rst article (Part I) is devoted to the development of mathematical descriptions of material
damping to represent the linear viscoelastic constitutive behavior, their implementation into
FE formulations and the use of the underlying dierent solution methods. To this end, internal
variables models, such as the Golla-Hughes-McTavish (GHM) and anelastic displacement elds
(ADF) models, and other methods such as the direct frequency response (DFR), based on
the complex modulus approach (CMA), iterative modal strain energy (IMSE) and an approach
based on an iterative complex eigensolution (ICE) are described and implemented at the global
FE model level. The experimental identication of viscoelastic materials properties and the
aforementioned viscoelastically damped FE modeling approaches are assessed and validated
in the companion article [Vasques, C.M.A. et al., Viscoelastic damping technologiesPart
II: Experimental identication procedure and validation, Journal of Advanced Research in
Mechanical Engineering 1(2): 96-110 (2010)].
Keywords: Finite element, damping, viscoelastic, complex modulus, internal variables.
1. INTRODUCTION
Since the seminal developments in the 1950s, viscoelastic
damping technologies are nowadays well established and
constitute widespread means of controlling the dynamics
of structures, reducing and controlling structural vibra-
tions and/or noise radiation. These technologies have been
widely used in several technological areas (e.g. automotive,
aeronautics, aerospace, acoustics). Several reasons could
explain its development and widespread use. However, the
most intuitive one relies on the evident growing interest in
the use of composite materials that have been intensively
applied to produce lighter and stier composite structures.

The joint funding scheme provided by the European Social Fund


and Portuguese funds from MCTES through POPH/QREN/Tipo-
logia 4.2 and project PTDC/EME-PME/66741/2006 are gratefully
acknowledged by the authors.
However, detrimentally, along with new and improved
assembly techniques, these new materials have also been
responsible for an important reduction of the inherent
material and joint damping characteristics of exible com-
posite structures.
In general, viscoelastic damping technologies have been
used mainly as distributed surface mounted or embed-
ded damping treatments, utilizing passive viscoelastic ma-
terials alone, the so-called passive damping treatments
[see, for example, Kerwin (1959), Mead and Markus
(1969), Nashif et al. (1985), Mace (1994)] or, more re-
cently, in an unied way making use also of active materi-
als such as piezoelectric polymers or ceramics, the so-called
hybrid damping treatments [see, for example, Plump and
Hubbard (1986), Baz (1993), Benjeddou (2001), Vasques
and Rodrigues (2008)]. When adequately designed, the
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
76 Copyright 2010 HyperSciences_Publisher. All rights reserved www.hypersciences.org
Viscoelastic Damping TechnologiesPart I: Modeling and Finite Element Implementation 2
use of viscoelastic materials in passive or hybrid damping
treatments furnishes structures with articial damping
mechanisms and might provide high damping capability
over wide temperature and frequency ranges. However,
their design and analysis (numerical and/or analytical) is
quite dicult and cumbersome. One of these diculties is
related with the development, application and implemen-
tation of ecient constitutive mathematical models able
to represent the frequency- and temperature-dependent
constitutive properties of general viscoelastic materials.
Another diculty is related with the reduced eorts that
have been made in the publication and dissemination of
databases with material data of general viscoelastic mate-
rials, making it sometimes dicult to nd accurate data
for the material under consideration.
Since a generalized and widespread used viscoelastic con-
stitutive model has not yet been recognized as such and
therefore is not available, the necessary viscoelastic mate-
rial information is usually identied from low resolution
and outdated nomograms published by the material man-
ufacturers in the form of frequency dependent complex
moduli data, which has been the most current and univer-
sal means of providing material data used by the manu-
facturers, or eventually obtained by adverse published data
related with fortuitous heuristic theoretical models and/or
by non-standardized experimental characterization proce-
dures. In fact, quite often, when dealing with new or
non-standard viscoelastic materials (e.g. cork compounds),
experimental identication tests for these materials need
to be carried out in order to furnish the designer with
the most important information at the start of the design
stage: the frequency- and temperature-dependent consti-
tutive law of the material.
In virtue of the aforementioned issues, the present work
presents an integrated study on the mathematical model-
ing and assessment of the eciency of surface mounted
or embedded viscoelastic damping treatments, incorpo-
rating the adequate use and development of viscoelastic
(arbitrary frequency dependent) damping models, along
with their nite element (FE) implementation, and the
experimental identication of the constitutive behavior of
viscoelastic materials. The work is presented in two com-
panion papers; the rst, Part I, addresses mathematical
modeling and implementation issues and the second, Part
II, experimental identication and modeling validation
ones.
The present article (Part I) is devoted to the development
of mathematical descriptions of material damping to repre-
sent the linear viscoelastic constitutive behavior, their im-
plementation into FE formulations and the use of the un-
derlying dierent solution methods. To this end, the main
diculties of the dierent approaches to modeling, FE
implementation and solution methods proposed in the last
decades are shortly reviewed and the new trends in both
time and frequency domain viscoelastic damping modeling
and their FE implementations are hereby presented and
discussed. Both time and frequency domain damping mod-
els, typically necessary in FE analysis of structural systems
with embedded or surface mounted hybrid or purely pas-
sive damping treatments, are presented. Internal variables
models, such as the Golla-Hughes-McTavish (GHM) and
anelastic displacement elds (ADF) models, and other
methods such as the direct frequency response (DFR),
based on the complex modulus approach (CMA), iterative
modal strain energy (IMSE) and an approach based on an
iterative complex eigensolution (ICE) are described and
implemented at the global FE model level. Additionally,
a state space representation of the integrated damped FE
models and model reduction procedures are discussed.
Regarding the experimental identication of viscoelastic
materials and the assessment and validation of the afore-
mentioned integrated FE damping modeling approaches,
they are presented in the companion article (Part II) by
Vasques et al. (2010) considering the well known viscoelas-
tic material 3M ISD112 and a sandwiched viscoelastic
plate.
2. VISCOELASTIC CONSTITUTIVE MODELING
2.1 Historical Developments of Viscoelasticity
In general, the constitutive behavior of viscoelastic mate-
rials might be said to depend upon the frequency, work-
ing temperature, amplitude and type of excitation [Ferry
(1980), Nashif et al. (1985), Riande et al. (2000), Jones
(2001)]. A mathematical model considering all these ef-
fects simultaneously has not yet been developed, is very
dicult to conceive and in practice has somewhat lim-
ited interest and applicability. Thus, for simplicity, since
the amplitude and type of excitation eects have been
reported to be of reduced importance, these parameters
are often overlooked. However, the temperature and fre-
quency dependent mechanical properties of the viscoelastic
materials still introduce serious diculties in the deni-
tion of an accurate mathematical model able to simulate
properly the dynamic behavior of the damped structure.
Therefore, for practical reasons, isothermal conditions are
usually assumed in the simulation conditions and merely
the frequency dependent constitutive behavior is directly
taken into account upon the constitutive mathematical
model. Following this assumption, the design of passive
viscoelastic damping treatments, for broad temperature
range applications, is usually conduced at several con-
stant temperature levels, selected within the temperature
range, considering isothermal conditions. A few excep-
tions, though, where the isothermal simplication is not
considered and where both the temperature and frequency
eects are captured in a coupled model able to account
for the the self-heating eects of the viscoelastic material,
were reported by Schapery (1964), who extended Biots
theory [Biot (1954, 1955)] in the application of the ther-
modynamics of irreversible processes to viscoelastic mate-
rials, yielding a unied theory of the thermomechanical
behavior of viscoelastic materials with an explicit tem-
perature dependence and a consistent thermodynamically
inclusion of the time-temperature superposition principle,
and by Lesieutre and his co-workers [Lesieutre and Govin-
dswamy (1996), Brackbill et al. (1996)], that extended the
ADF method to include thermal eects by means of the
use of the linear time-temperature superposition principle,
valid for thermorheologically simple materials, yielding a
nonlinear amplitude-dependent FE model which requires
specic nonlinear solution strategies. Other studies where
the temperature eects are also considered can be found
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
77
Viscoelastic Damping TechnologiesPart I: Modeling and Finite Element Implementation 3
in [Taylor et al. (1970), Zocher et al. (1997), Hammerand
and Kapania (1999), Demirdzic et al. (2005), Cliord
et al. (2006)]. The eect of the operating temperature
upon the performance of hybrid damping treatments and
viscoelastic damping eciency were analyzed, for example,
in [Baz (1998), Friswell and Inman (1998), Trindade et al.
(2000a), Silva et al. (2005), Pradeep and Ganesan (2006)].
Literature covering the three primary mechanisms of
damping which are important in the study of mechanical
systems, namely internal damping (or material ), struc-
tural damping (at joints and interfaces) and uid damping
(through uid-structure interaction), is very vast and it is
not the purpose here to survey it. The interested readers
are referred to some textbooks and articles on general
damping theory [Bishop (1955), Crandall (1970), Ewins
(2000), Lesieutre (2001), Inman (2001), Cremer et al.
(2005), de Silva (2005)] in order to become acquainted
with the developments on damping theories. Worth to
mention is a recent historical and contemporary critical
discussion on damping theory, also pointing out the per-
spective on how the treatment of damping is likely to
evolve in the future, performed by Peters (2005).
Directing now our attention to material damping, since
the rst experimental observation of material (or inter-
nal ) damping performed by Coulomb in the 1780s with
his memoire Sur la force de torsion et sur lelasticite
des ls de metal [Coulomb (1784)], where he not only
hypothesized regarding the microstructural mechanisms
of damping but also undertook experiments which proved
that the damping of torsional oscillations is not caused by
air friction but by internal losses in the material, over the
following centuries dierent methods to characterize fre-
quency dependent damping, in general, and the viscoelas-
tic materials constitutive behavior (viscoelastic damping),
in particular, were proposed.
As discussed by Snowdon (1968), the mechanical prop-
erties of rubber-like materials may be considered at two
dierent damping levels: (i) low-damping materials, which
have the dynamic modulus and damping factor (typically
with a low value of the order 0.1) varying slowly with fre-
quency, which enables them to be considered as constants
through the range of frequencies normally of concern in vi-
bration problems; and (ii) high-damping materials, where
both the dynamic modulus and damping factor vary sig-
nicantly with frequency. For the latter case, during the
transition zone, where the material undergoes a change
from rubber to glassy dominated behavior, the dynamic
modulus increases very rapidly with frequency and the
damping factor is large (typically with values of order 1)
and might vary slowly or more strongly with frequency.
Under the foregoing assumptions, Snowdon has considered
four dierent models of damping, which he called damping
of the solid type I and II, of the viscous type and of a
three-element spring and dashpot combination. The rst
damping type is nowadays referred to in the literature
as hysteretic damping [Banks and Pinter (2001), Inman
(2001)], where both the storage modulus and loss factor do
not depend on frequency and is typically applied to model
the small energy dissipation eects due to the internal
friction which occurs in metals in general. The second and
third damping types alleviate, respectively, the assumption
that the storage modulus or the loss factor do not depend
on frequency. The interest and applicability of them for
viscoelastic modeling is somewhat reduced since usually
models of rubber-like materials need to have explicit fre-
quency dependent mechanical properties in order to ac-
curately capture their real physical behavior. Regarding
the third damping type, it is considered by an equivalent
constitutive model comprising an Hookean spring in par-
allel with a Newtonian dashpot, which justies the term
viscous used for its designation, where the loss factor is
considered frequency-dependent (the well known viscous
damping [Gandhi (2001), Inman (2001)]). It should be
mentioned, though, that the viscous equivalent consti-
tutive model, also known as Kelvin-Voigt model, poorly
represents the dynamic mechanical properties of rubber-
like materials justifying the need for other type of models.
In view of that, the fourth model, known as standard linear
solid, considering a three-element combination of a spring
and dashpot in series (Maxwell model) placed in parallel
with another spring, allows both the dynamic modulus and
damping factor to depend on frequency. In fact, all the rst
three damping types can be seen as particular cases of this
more general last model.
The aforementioned four dierent types of damping illus-
trate well the real physical damping behavior observed in
the dierent kinds of materials that we have at our disposal
and it is up to the designer to choose the model that
best suits the real physical behavior of the material under
consideration. Usually, in the authors opinion, the desig-
nation of viscoelastic damping refers to the fourth case
mentioned above where both the storage modulus and loss
factor are physically observed to depend signicantly with
frequency. However, depending on the degree of frequency
dependence, some viscoelastic models might be preferred
over the others depending on how well the observed creep
and relaxation, in the time domain, and broadband, in
the frequency domain, behaviors are approximated by the
considered viscoelastic damping model.
The four models mentioned before can be regarded as
classical viscoelastic damping models. These models
were gradually developed since the 18th century in an at-
tempt to mathematically represent the real physically ob-
served behavior of a class of materials somewhat evidenc-
ing a mechanical behavior composed by a blend of pure
viscous uids and elastic solid materials constitutive char-
acteristics. These mechanical model analogies of viscoelas-
tic materials consider selected combinations of discrete
elastic and viscous damping elements, ranging from basic
discrete systems such as the Maxwell model, established by
Maxwell (1868), or Kelvin-Voigt model, rst established by
Kelvin (1875) and later by Voigt (1892) (sometimes also
referred to in the literature as only Kelvin or Voigt model),
to more complex models combining these basic elements
in parallel or series, namely, the linear standard solid
model, also known as Kelvin or Zener model [Zener (1948)],
and the Burgers model [Burgers (1935)], which represent
more realistic viscoelastic models with good complexity to
accuracy trade-o, or even generalized versions combining
these elements (e.g. the generalized Maxwell model and
generalized Kelvin model ). The latter generalized mod-
els yield frequency- and time-domain mathematical series
representations of the viscoelastic constitutive behavior in
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
78
terms of the so-called Dirichlet or, alternatively, Pronys
series [Prony (1795), Williams (1964)]; see also some works
using Pronys series in [Park and Schapery (1999), Slanik
et al. (2000)]. Lastly, all these models, though, can also
be seen as particular cases of a more general formulation,
namely, the so-called dierential equation form of the
linear viscoelastic constitutive behavior under isothermal
conditions. The interested reader can nd further details
on classical models and dierential equation forms using
mechanical analogies in [Alfrey and Doty (1945), Williams
(1964), Ferry (1980), Findley et al. (1989), Johnson
(1999), Fung and Tong (2001), Ottosen and Ristinmaa
(2005)] and in the references therein.
Certainly, a very comprehensive and accurate model may
be constructed by combining generalized models in an
arbitrary way. Irrespective of how we combine springs and
dashpots, it comes as no surprise that it is always possible
to write the constitutive relation in a dierential equation
form. An advantage of the dierential equation approach
is that each model is easy to physically understand and
interpret and models can be constructed in an intuitive
fashion. The drawback, however, is that more advanced
models soon become cumbersome since the initial con-
ditions become dicult to deal with when using higher-
order dierential equation in the time domain [Ottosen
and Ristinmaa (2005)].
Since to have good accuracy representing complex rate-
dependent materials, such as viscoelastic materials, more
complex networks of springs and dashpots are usually re-
quired, an alternative approach is to change the dierential
equation for the evolution of the dissipative stresses by
means of the tools of fractional calculus; see, for example,
Gemant (1936), Stiassnie (1979), Bagley and Torvik
(1983), Koeller (1984), Bagley and Torvik (1985), Koeller
(1986), Padovan (1987), Bagley (1989) and Pritz (1996).
In that case, the viscoelastic constitutive behavior is de-
ned by a fractional dierential equation form where the
time derivatives order are not integer numbers. As sug-
gested by Koeller (1984), the dashpots used in the classical
rheological models might be substituted by the so-called
spring-pot element, yielding a fractional dierential model
where the spring-pots are elements that, depending on the
non-integer order (or memory parameter) of the derivative,
allow a continuous transition of the constitutive behavior
from the solid state (full memory representing a spring
with complete elastic recovery) to the uid state (no mem-
ory representing a dashpot with unrecoverable behavior)
when the memory parameter varies from zero to one.
Using the fractional derivative model instead of the clas-
sical dierential approach, one advantage is the easiness
how with a reduced number of parameters, typically three
[Bagley and Torvik (1985)], four [Pritz (1996), Galucio
et al. (2004)] or ve [Bagley and Torvik (1983)], the vis-
coelastic material properties are modeled in the frequency
domain. However, its representation and use in the time
domain is far more complicated than in the frequency
domain requiring specic solution procedures to integrate
the governing equations; see, for example, Enelund and
Josefson (1997).
As discussed previously, there were several early contri-
butions to the theory of linear viscoelasticity stated in
terms of dierential equations, such as the ones given by
Maxwell, Kelvin and Voigt. However, integral equations
could be used as well, and in general terms, the integral
approach might be seen as the general approach and start-
ing point from which the viscoelastic theory should start
from. Boltzmann developed in 1874 the rst formulation
of a three-dimensional theory of linear isotropic viscoelas-
tic stress-strain relations while Volterra (1909, 1913) ob-
tained comparable forms for anisotropic solids in 1909.
Boltzmann initiated a linear hereditary theory of material
damping by formulating an integral representation of the
stress-strain relation which allows greater freedom when
constructing models than the dierential approach that
relies on the concepts of certain combinations of springs
and dashpots. Integrals are summing operations, and this
view of viscoelasticity takes the response of the material
at a certain instant of time to be the sum of the responses
to excitations imposed at all previous instants of time.
The ability to sum these individual responses requires the
material to obey a more general statement of linearity than
we have invoked previously, specically that the response
to a number of individual excitations be the sum of the re-
sponses that would have been generated by each excitation
acting alone.
Thus far we have devoted our attention to time-dependent
formulations by means of dierential or integral equations
forms where the typical creep and relaxation behavior
of viscoelastic materials were more directly observable
and identied from experimental measurements. However,
since one of the simplest ways of experimentally determine
the viscoelastic properties is to subject the material to
periodic dynamic oscillations, the dynamic representation
of the material properties by means of a complex modulus
became a current practice and is perhaps the most com-
monly used approach when modeling the material in the
frequency domain.
As far as material damping is concerned, and in par-
ticular the viscoelastic damping, the interested read-
ers in further details are referred to a report with
a unique comprehensive bibliography survey until the
early-mid of the 20th century of the material damping
eld compiled by Demer in the 1950s [Demer (1956)]
and to some articles [Alfrey and Doty (1945), Lazan
(1959), Ungar and Kerwin (1962), Williams (1964), Bert
(1973), Tschoegl (1997), Johnson (1999), Mead (2002)]
and textbooks [Zener (1948), Snowdon (1968), Chris-
tensen (1982), Nashif et al. (1985), Tschoegl (1989), Fung
(1993), Sun and Lu (1995), Mead (1998), Riande et al.
(2000), Fung and Tong (2001), Jones (2001), Goodman
(2002), Ottosen and Ristinmaa (2005)] which illustrate
well the later developments in the eld.
Within the scope of linear theory, time and frequency
domain constitutive models are presented in what follows.
2.2 Time Domain Representation: Hereditary Approach
Viscoelastic materials are sometimes called materials with
innite memory, in the sense that their actual mechan-
ical response is modulated by the past history. There-
fore, the viscoelastic constitutive behavior relies on the
assumption that the current value of the stress tensor
depends upon the complete past history of the strain
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
79
tensor components. This latter assumption therefore im-
plies that the behavior of any viscoelastic material may
be represented by a hereditary approach. Furthermore, if
the net strain response of the material to an arbitrary
sequence of stimulus can be determined through the sum
of the responses which would have been obtained if all
stimuli had act independently, time dependence linearity,
and if an increase in the strain stimulus by an arbitrary
factor increases the stress response by the same factor,
stress-strain linearity, the material is said to have a linear
viscoelastic behavior. These are the type of viscoelastic
materials discussed in what follows.
Considering an isotropic viscoelastic material under isother-
mal conditions and under small (innitesimal) deformation
conditions, the theory of linear viscoelasticity [Christensen
(1982), Fung and Tong (2001)] states that the constitutive
relationship for a generic one-dimensional isotropic stress-
strain system (e.g. in shear or extension) can be given by
a Riemann convolution integral,
(t) =
_
t

G
rel
(t )
()

d, (1)
where (t) and (t) are the time dependent stress and
arbitrary strain component histories and G
rel
(t) is called
the constitutive time varying (shear) characteristic relax-
ation function of the material (the stress response to a
unit-step strain input) which is also utilized in the litera-
ture under many dierent names, such as damping kernel,
retardation, hereditary or after-eect function. Equation
(1) expresses an essential feature of linear behavior of
viscoelastic materials known as Boltzmanns superposition
principle [Boltzmann (1874)]. Since the lower limit of inte-
gration is taken as , it is to mean that the integration
is to be taken before the very beginning of the motion.
Thus, if the motion starts at time t = 0, meaning that
the stress and strain are equal to zero up until time zero
where the loading begins, i.e., (t) = (t) = 0 for t < 0,
and discontinuous strain histories with a step discontinuity
at t = 0 are to be considered, Equation (1) reduces to
(t) = G
rel
(t)(0) +
_
t
0
G
rel
(t )
()

d, (2)
where (0) is the limiting value of (t) when t 0 from
the positive side. The rst term in Equation (2) gives the
eect of the initial disturbance and it arises from allowing
a jump of (t) at t = 0. Furthermore, when Equation (1)
was written it was tacitly assumed that (t) is continuous
and dierentiable [see Fung and Tong (2001) for further
details].
Considering nil initial conditions, i.e., (0) = 0, the La-
place transform of Equation (2) yields
(s) =

G(s)(s) s

G
rel
(s)(s), (3)
where

G(s) s

G
rel
(s) is a characteristic material func-
tion, which should be experimentally identied somehow.
Among all the possible identication tests [see Tschoegl
(1989) for further details], the relaxation test, where a
step relaxation strain stimulus with height
rel
is applied,
i.e. (s) =
rel
/s, is considered here. Substituting it into
Equation (3), the relaxation stress is given by

rel
(s) =
rel

G
rel
(s). (4)
Transforming the previous equation back to the time do-
main, the constitutive time varying characteristic relax-
ation function of the material is given by
G
rel
(t) =

rel
(t)

rel
. (5)
The previous equation allows us to directly determine its
value from the measured time varying relaxation stress
history to a step strain stimulus which is obtained from a
relaxation test applied to a viscoelastic material. In fact,
the typical relaxation behavior is well known and is usually
described by a time domain curve composed by the sum of
a constant step part and a time decaying counterpart with
the simplest relaxation function having the fading memory
characteristics being that of a single decaying exponential
[Christensen (1982)]. Therefore, in general, the relaxation
behavior might be expressed as
G
rel
(t) =
rel
(t)/
rel
G

+ r(t), (6)
which in the Laplace domain becomes
G
rel
(s) =
G

s
+ r(s), (7)
where G

is the so-called relaxed (also known as static)


modulus, which for an isotropic material might be the
shear or extension modulus, obtained after the material
relaxation, i.e., G

is the limiting value of G


rel
(t) when
t . The resultant relaxation modulus in Equation (7)
therefore has two components; since G

is constant, the
rst component, G

/s, represents the recoverable (elastic)


counterpart of the material stress, while the second com-
ponent, r(s), represents the non-recoverable (dissipated)
one. For that reason, r(s) is usually termed as dissipa-
tion, or more appropriately, relaxation function. These
two parameters, G

and r(s), are unique characteristic


properties of the viscoelastic material which are obtained
by curve tting experimental data, obtained from a relax-
ation identication test where a step function deformation
stimulus is utilized, according to the mathematical vis-
coelastic damping representation chosen by the analyst.
In fact, it is worthy to emphasize here that in order to
characterize the viscoelastic material behavior to any type
of strain stimulus, Equation (3) should be used; the nec-
essary characteristic material function,

G(s) s

G
rel
(s),
can readily be obtained after the relaxation stress has
been determined in a relaxation test. Depending upon
the nature of the viscoelastic material being considered,
as discussed for example by Brinson and Brinson (2008),
Equation (6) is pretty general being able to represent both
the behavior of thermosetting (crosslinked) polymers if
G

= 0 or thermoplastic (linear) polymers if G

= 0
assuming in both cases that r = r(t).
Many authors have proposed dierent mathematical rep-
resentations of the relaxation behavior given by r(s) in the
last decades, which in turn leads to dierent mathematical
models of the viscoelastic damping behavior and dierent
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
80
Table 1. Mathematical representations of eligible relaxation functions in Laplace and time domains.
Function r(s) Function r(t), t 0 Author and year
n

s=1
D
s
p +
s
+ D

n

s=1
D
s
e

s
t
+ (t)D

Biot (1954)
_

0
D()()
p +
d + D

_

0
D()()e
t
d + (t)D

Biot (1954)
Buhariwala and Hansen (1988)
E
1
s

E
0
bs

s(1 + bs

_
E
1
bE
0
b
_
E

[(t/b)

]

Bagley and Torvik (1983)
n

i=1

i
s + 2

i

i
s
2
+ 2

i

i
s +
2
i
n

i=1

i
b
2i
e
b
1i
t
b
1i
e
b
2i
t
b
2i
b
1i

Golla and Hughes (1985)


McTavish and Hughes (1993)
n

i=1

i
s +
i
n

i=1

i
e

i
t
Lesieutre (1992)
n

i=1
c
i

i
s + 1
+ c
0
n

i=1
(c
i
/
i
)e
t/
i
+ (t)c
0
Yiu (1994)
g
x
1 e
st
1
s
_
g
x
, t < t
1
0, t > t
1
Adhikari (1998)
g
x
s
1 e
st
3
+ 2(st
3
/)
2
1 + 2(st
3
/)
2
_
_
_
g
x
2
_
1 + cos
_
t
t
3
__
, t < t
3
0, t > t
3
Adhikari (1998)
e
s
2
4
2
erfc
_
s
2

2
_
2
_

2
t
2
Adhikari and Woodhouse (2001a)

In this expression a one-dimensional constitutive behavior was considered and the original notation used by Biot (1954)
was retained. Here p = d/dt is a time operator which in the Laplace domain corresponds to considering p = s and the
equality G

D was considered for the denition the relaxation function. Is is worthy to mention that, strictly speaking,
the second term of the relaxation function has no inverse in the time domain and therefore will not inuence the time domain
denition of the relaxation function, which is only applicable for a relaxation test considering a step strain stimulus. However,
for completeness, as the delta function is often used in mathematical physics as a generalized function, it can be formally
added to the time domain relaxation function, which will allow to represent the relaxation modulus, including already a pure
elastic term given by G

, also by an extra pure viscous term when the correspondent relaxation modulus is substituted into
Equation 2.

The ve parameter constitutive behavior considering the fractional derivative operator D

[ ], so that 0 < < 1 and


0 < < 1, with the property in the Laplace domain L{D

[x(t)]} = s

L{x(t)}, is given by (t) + bD

[(t)] = E
0
(t) +
E
1
D

[(t)]; the original notation in Bagley and Torvik (1983) was used; however, it is worthy to mention that E
0
in the
original notation represents the relaxed (static) modulus denoted in the present work as G

Since the fractional derivative identication of most viscoelastic materials has usually resulted in and that, in addition
to this experimental nding, it has been proved theoretically that for the model to be consistent with thermodynamic principles
= [Bagley and Torvik (1986)], yielding a four parameter model, the latter model is considered enough to describe the
dynamic behavior of real viscoelastic materials in a wide frequency range; the time domain relaxation function presented is
therefore dened under those assumptions, i.e., that = , and is expressed in terms of the one parameter Mittag-Leer
function E

(z) =

i=1
z
i
/(1 + i) [see Enelund and Olsson (1999), Welch et al. (1999) and Adolfsson et al. (2005) for
further details].

The constants b
1i
and b
2i
are given by b
1i
, b
2i
=
i

i

i
(

i
2
1)
1/2
.
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
81
Viscoelastic Damping TechnologiesPart I: Modeling and Finite Element Implementation
solution methods of the governing equations in the time
and frequency domains. As a result of a compilation from
several sources, some dierent relaxation functions were
summarized by Golla and Hughes (1985), Park et al.
(1999), Trindade and Benjeddou (2002) and Adhikari
and Woodhouse (2003), where the dierent models and
solution methods of the resultant mathematical models
are also thoroughly discussed. Some of the most well
known relaxation functions, frequently referred in the open
literature, are the ones used by the internal variables GHM
and ADF models, which will be considered in this work
and described in the following sections. The correspondent
relaxation functions along with some other mathematical
representations that have been used and presented in the
literature are summarized in Table 1.
Following the thermodynamics of irreversible phenomena
and applying it to viscoelasticity, the engineer, physicist
and applied mathematician Maurice Biot was a pioneer
rst giving the subject in the 1950s an original treat-
ment
1
based upon relaxation modes and hidden variables
and establishing a linear viscoelastic constitutive theory
[Biot (1954)]. It was shown that the proposed model was
very general in nature and that it could represent the
behavior of a wide variety of phenomena in the process
of their response to external actions which may involve,
for example, the application of external stresses, chemical
reactions, heat transfer, etc. In fact, Biots work comprised
an unusually broad range of scientic and technological
areas including applied mechanics, acoustics, heat transfer,
thermodynamics, aeronautics, geophysics and electromag-
netism, to which his contributions on the thermodynamics
of irreversible phenomena were extremely important. His
approach is presented in the rst two lines of Table 1.
As reported by Biot (1954), the summation presented in
the rst line of Table 1 is extended to all internal relaxation
constants,
s
. In order to fully represent the eects of all
the internal variables of a solid, since it has too many
degrees of freedom, there may be in some cases an almost
continuous distribution of relaxation modes. That will
correspond to a spectrum or spectral density distribution
of the relaxation constant D() with a density distribution
function (), as presented in the second line of Table
1. That constitutive model was later utilized to model
viscoelastic beams and plates as reported by Buhariwala
and Hansen (1988) and Biot (1972), respectively.
The exponential function introduced by Biot is proba-
bly the simplest physically realistic non-viscous damping
model [Adhikari and Woodhouse (2001b)] and has been
used extensively in the context of viscoelastic systems.
In fact, Biot has shown that, not only in the context
of mechanical models but also involving other coupled
phenomena such as chemical, thermodynamic, etc., any
relaxation phenomena may be represented by a spring, a
dashpot and a sum of a great many elements made up of a
Maxwell type material. As particular cases of Biots model
we have the double exponential relaxation function, known
1
The rst treatment given by Biot has been shown over the
following decades to include other theories and relaxation models,
as the ones presented in Table 1 which, with the exception of the
fractional calculus approach proposed by Bagley and Torvik (1983)
and presented in the third line of Table 1, correspond to particular
cases of Biots treatment.
as the GHM model, or the single exponential relaxation
function of the ADF model. Regarding Yius model, it
was proposed in the context of nite element solutions of
structures with viscoelastic treatments and is equal to the
model proposed by Biot, however with a slight change in
the constants of the coecients.
2.3 Frequency Domain Representation: Complex Modulus
There are practical situations in which structures with
viscoelastic materials may be subjected to steady-state
oscillatory forcing conditions. Under these conditions, the
characteristic (shear) material function previously dened
in the stress-strain relationship in Equation (3) is dened
by assuming s as a pure imaginary variable (or similarly
through its Fourier transformation), so that
G

(j) =

G(s) for s = j, (8)
where is the frequency, yielding the so-called complex
(shear) modulus in the form
G

(j) = G

() + jG

(), (9)
where G

() is the (shear) storage modulus, which accounts


for the recoverable energy, and G

() is the (shear) loss


modulus, which represents the energy dissipation eects
[Christensen (1982), Nashif et al. (1985), Jones (2001)].
The loss factor of the viscoelastic materials is dened as
() =
G

()
G

()
, (10)
which alternatively allows writing Equation (9) as
G

(j) = G

() [1 + j()] . (11)
For a linear, homogeneous and isotropic viscoelastic mate-
rial, equivalent representations of the previous equations
hold for the complex extensional modulus E(j) and the
relationship
G

(j) =
E

(j)
2 [1 + (j)]
, (12)
where (j) is the Poissons ratio, establishes a relation-
ship between the two. However, for simplicity, a real fre-
quency independent Poissons ratio (j) = is usually
assumed, leading to identical loss factors of the shear and
extensional complex moduli, i.e.,
E
() =
G
() = ().
Further details on the dynamic properties of the Poissons
ratio in linearly viscoelastic solids can be found in Chen
and Lakes (1993), Pritz (2000), Tschoegl et al. (2002)
and Lakes and Wineman (2006).
Before concluding this section on the constitutive mod-
eling of viscoelastic materials, it is worthy to empha-
size that in fact elastic materials are particular cases of
viscoelastic ones. Therefore, imposing the restrictions of
time-independent relaxation behavior, i.e. r(t) = 0,
or alternatively r(t) = const., whether in the time or
transformed (Laplace or Fourier) domains, a pure elastic
behavior is obtained where for an elastic isotropic material
we have G(t) = G

and E(t) = E

in the time domain.


As a consequence, in the frequency domain, the complex
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
82
modulus representation yields nil loss modulus and loss
factor without any time or phase lag between the stress
and the corresponding strain. In general, the material
properties may still be frequency dependent but without
any (or a weak) damping behavior, which corresponds to
a very restricted class of materials which have somewhat
reduced interest and applicability.
3. FINITE ELEMENT IMPLEMENTATION OF
VISCOELASTIC DAMPING
In the analysis of viscoelastically damped structures the
choice of the most adequate solution method is strongly
dictated by the applied temperature- and time-dependent
viscoelastic constitutive model and by the desired type and
accuracy of the system responses. This relation is strik-
ingly noticeable and both issues, viscoelastic model and
solution method, are often associated and no distinction
between them is usually reported in the open literature.
However, a clear distinction between the two must be
made, keeping still in mind that a close relation between
them exists. These methods are reviewed and presented in
what follows.
3.1 Damping Modeling and Solution Approaches
As previously referred, in general, the elastic and dissipa-
tive properties of viscoelastic materials may depend upon
the frequency, operating temperature, amplitude and type
of excitation [Ferry (1980), Nashif et al. (1985), Riande
et al. (2000), Jones (2001)]. These dependencies of the
viscoelastic material properties make a mathematical de-
scription of the viscoelastic constitutive behavior and ma-
terial damping more dicult and complicated to obtain,
and might turn the underlying FE implementations and
solution methods more troublesome and dicult endeav-
ors. Thus, for simplicity, the amplitude and type of exci-
tation eects are often overlooked, isothermal conditions
are usually assumed and only the frequency dependency of
the viscoelastic constitutive behavior is usually taken into
account.
The direct frequency response (DFR) method, which is
based upon the complex modulus approach (CMA), early
presented in the 1950s by Myklestad (1952) and subse-
quently further discussed and utilized by Snowdon (1968)
and Bert (1973), is a frequency domain method that
utilizes a time-domain based model which is limited to
steady state harmonic vibrations. This corresponds to
a simple way of modeling viscoelastic damping eects
where the material properties are continuously updated
for each discrete frequency value [Moreira and Rodrigues
(2004)]. However, the structural model can alternatively
be formulated in the frequency domain, yielding the so-
called wave models, which are based on the denition
of a dynamic stiness matrix. These wave methods can
also be extended to discontinuous structures and damp-
ing treatments, for example by the use of the so-called
spectral nite element method (SFEM) proposed by Doyle
(1997) which utilizes dynamic interpolation functions.
These wave-based methods, which also allow viscoelastic
damping eects to be accurately considered by dening
the material constitutive behavior for the current waves
specic frequency, have been used, for example, by Wang
and Wereley (1998), Baz (2000) and Wang and Wereley
(2002). This solution method is usually denoted as a wave
propagation method (WPM) and, according to Wang and
Wereley (2002), Douglas was the rst to explore wave
solutions in order to implicitly account for the frequency
dependent complex modulus of viscoelastic components in
the solution method [Douglas (1977), Douglas and Yang
(1978)].
The CMA is also the basis of the so-called modal strain
energy (MSE) method, rst derived and utilized by Mead
(1960) and later popularized by Johnson et al. (1980),
where the loss factor of each individual mode is determined
from the ratio between the dissipated modal strain energy
of the viscoelastic counterpart and the storage modal
strain energy of the whole structural system. However,
MSE based methods are known to lead to poor viscoelastic
damping estimation of highly damped structural systems;
iterative versions of the MSE have been successfully used,
though, only for moderate damping values [Trindade et al.
(2000b)].
In more recent years and in opposition to frequency do-
main based, i.e. CMA based, approaches, time domain
models, such as the Golla-Hughes-McTavish (GHM), after
Hughes and his colleagues [Golla and Hughes (1985), Mc-
Tavish and Hughes (1993)], anelastic displacement elds
(ADF), after Lesieutre and co-workers [Lesieutre and
Bianchini (1995), Lesieutre et al. (1996)], and others
presented by Yiu (1993) and more recently by Silva
(2003), utilizing additional internal (or dissipation) vari-
ables [Johnson (1999)], have been successfully utilized and
were shown to yield good damping estimates.
Alternatively, the use of fractional calculus (FC) models,
initially proposed by Bagley and Torvik (1983, 1985) and
recently revisited, among others, by Enelund and Lesieutre
(1999), Jones (2001), Schmidt and Gaul (2002) and Galu-
cio et al. (2004, 2005, 2006), provides a simpler and more
economic descriptor of the complex constitutive behavior
of viscoelastic materials, being able to represent the stor-
age and loss modulus of frequency dependent viscoelastic
materials using a low order parameter set formed by one
or two model parameter series and based upon the use
of fractional derivatives. It has the drawback, though, of
generating a non-standard FE time domain formulation,
with a more burdensome characteristic solution procedure
making use of some special solution artices, but which
still allows to obtain good damping estimates of the sys-
tem.
Other relevant contributions for damping, in general, and
viscoelastic damping modeling, in particular, were given,
among others, by Adhikari and Woodhouse (2003) and
the references therein on non-viscous damping in dis-
crete linear systems, Balm`es and his co-workers [Balm`es
(1997), Plouin and Balm`es (1998, 1999)], who have used
dierent variations of modal subspace models specially
dened and assumed to be representative of the damped
system response, and Kelly and Stevens (1989) and Lin
and Lim (1996), that have used perturbation method ap-
proaches. Since they fall out of the scope of the damping
strategies utilized in this work, it is not intended to review
those works here and the interested readers are referred to
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
83
the aforementioned references for further details on the
methods used.
To sum sup, the solution and analysis methods can
be divided into frequency and time domain based solu-
tion strategies, which clearly can admit distinct specic
constitutive models. Comparative analysis and surveys
of modern methods for modeling frequency dependent
damping in FE models were performed by Slater et al.
(1993), Trindade et al. (2000b), Benjeddou (2001) and
Trindade and Benjeddou (2002). The next sections address
the two alternative solution methods and describe dierent
alternatives as far as the FE implementation is concerned.
3.2 Frequency Domain Based Approaches
Direct Frequency Response (DFR) Analysis Model Con-
sidering the frequency dependent constitutive behavior of
viscoelastic materials discussed previously and adopting a
complex modulus representation, the general time depen-
dent FE equations of motion of a viscoelastically damped
general structural system can be written as
M u(t) +D u(t) + [K
E
+K
V
(j)]u(t) = f (t), (13)
where M and D are the global mass and viscous damping
matrices, K
E
and K
V
(j) are the elastic and complex
frequency dependent viscoelastic stiness matrices, and
u(t) and f (t) are the displacement degrees of freedom
(DoFs) and applied loads vectors. It is worthy to mention
that two types of damping models are considered in the
previous equation: (i) an arbitrary frequency dependent
hysteretic (or viscoelastic) damping type, represented by
Im[K
V
(j)], whose terms are frequency dependent, repre-
senting the viscoelastic dissipation (relaxation) behavior;
and (ii) a viscous damping type, which is described by D,
representing other general sources of damping (e.g. air-
based damping, energy dissipation in the supports, etc.),
which are assumed to be proportional to the velocity.
Using Equation (13) as it is, the frequency dependent
matrix denition implies that its use and analysis can
only be performed in the frequency domain, based on
the CMA, where the material properties of the stiness
matrix of the viscoelastic parts are dened for each discrete
frequency value [Vasques et al. (2004), Moreira and
Rodrigues (2004), Vasques et al. (2006)]. Thus, the
DFR is a frequency domain method where the frequency
response model [Ewins (2000)] can be generated in a
straightforward manner from the results of many discrete
frequency calculations of the equations of motion, in which
the complex stiness matrix of the viscoelastic parts is re-
calculated at each frequency value of the desired discrete
frequency range.
Considering simple harmonic excitation, with
f (t) =

Fe
jt
, (14)
where

F is the amplitude vector of the applied mechanical
forces, the steady state harmonic response of the system
can be written as
u(t) =

U(j)e
jt
, (15)
where

U(j) is the complex response vector (displace-
ments phasor). Substituting Equations (14) and (15) into
Equation (13), yields
[K(j) + jD
2
M]

U(j) =

F, (16)
where K(j) = K
E
+ K
V
(j), from which the complex
solution vector

U(j) can be obtained.
For a force applied in the ith DoF and a displacement
measured in the oth DoF, the FRF (frequency response
function) can be obtained by solving Equation (16) for
dierent values of frequency,
[K(j
l
) + j
l
D
2
l
M]

U(j
l
) =

F
i
, (17)
where

F
i
denotes a force vector with a non-zero component
in the ith DoF and all other elements equal to zero, and

U(j
l
) is the resulting complex response vector (displace-
ments) solution at frequency
l
. Thus, the receptance FRF
at frequency
l
is given by
H
oi
(j
l
) =

U
o
(j
l
)

F
i
, (18)
where

F
i
is the amplitude of the force input and

U
o
(j
l
)
is the displacement response amplitude extracted from
the oth DoF of the vector

U(j
l
). Lastly, the frequency
response model (FRFs) can be generated from the results
of many discrete frequency calculations of Equations (17),
in which the complex stiness matrix of the viscoelastic
layers is re-calculated at each frequency value comprised
in the discrete frequency range =
0
, . . . ,
l
, . . . ,
f
, as
depicted in Figure 1.
E V
(j ) (j ) K K K
2
[ (j ) j ] (j )
i
K D MU F
l

(j )
oi
H
0
, , , ,
l f

Fig. 1. Generation diagram of the direct frequency respon-
se (DFR) analysis model.
Modal Strain Energy (MSE) Approach Assuming that
the normal modes obtained from the undamped system are
representative of the damped system, which in principle
is true only for lightly damped structures, and that a
frequency independent stiness matrix might be assumed,
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
84
an approximation to the modal loss factors can therefore
be obtained from the ratio between the loss and the
storage parts of the modal strain energy. Based on this
assumption, Johnson and Kienholz (1982) proposed a
direct method for the FE prediction of the modal damping
ratio in structures with viscoelastic damping layers. While
the storage energy is calculated from the modal strain
energy of the entire structure for each individual mode,
the dissipated energy is calculated by multiplying the
modal strain energy corresponding to the viscoelastic
material FEs by the respective material loss factor at
some arbitrarily chosen value of frequency, which ideally
would be the value of the natural frequency (not yet
known though) of the current mode. Thus, establishing
a relationship between the loss and storage modal strain
energies of a generic mode r, the modal loss factor of the
rth mode might be expressed as

r
=

r
T
Im[K(j)]
r

r
T
Re[K(j)]
r
, (19)
where
r
is the rth mode shape vector. In physical terms,
the previous relationship represents the ratio between the
dissipated energy, proportional to the imaginary compo-
nent of the complex stiness matrix, and the stored energy,
proportional to the real counterpart.
Conceptually, the MSE approach was initially introduced
by Mead (1960) and Ungar and Kerwin (1962) in the early
1960s and later popularized and applied to FE analysis
by Johnson and Kienholz (1982) and Soni and Bogner
(1982). This approximated method is simple to implement
in a FE simulation and is computationally cost-eective,
providing a good damping estimator in comparative design
analysis or optimization procedures [Hwang and Gibson
(1992), Moreira and Rodrigues (2006)]. However, its
usage is restricted to lightly damped structures wherein
the undamped real modal model can eectively represent
the damped structure. For highly damped structures,
where the modal coupling is noteworthy, and especially
for those where the added viscoelastic material induces a
considerable mass and stiness modication, the originally
proposed MSE approach usage may lead to considerable
representativeness errors. In view of this, modied versions
have been recently proposed in an attempt to nd more
accurate approaches. Among others, the one proposed by
Hu et al. (1995) is one of these modied MSE versions
and is based on the application of a modal vectors derived
from a modied eigenvalue problem which considers also
the imaginary counterpart of the stiness matrix.
Iterative Modal Strain Energy (IMSE) Approach As pre-
viously referred, the MSE approach is based on the princi-
ple that the undamped natural modes of vibration of the
viscoelastically damped structure are representative of the
damped modal model and that a frequency independent
stiness matrix might be assumed. However, and despite
the fact that this approximation might be valid for low to
moderate additions of damping materials, the application
of single or multi-layer viscoelastic sandwiched damping
treatments to structures might modify the stiness sub-
stantially due to the decoupling eects derived from the
viscoelastic core/layers. In view of this, the mechanical
properties may alternatively be assumed constant in the
neighborhood of each mode and the eigensolution of each
individual mode can be determined independently, which
corresponds to a more realistic approach. This method-
ology can, however, be also applied under an iterative
approach where the stiness matrix is updated iteratively
with the new corrected complex stiness of the calculated
mode. In general, this procedure allows to obtain more
realistic values of the modal loss factors and natural fre-
quencies with the drawback of requiring more processing
time.
Following this line of thought, a modication to the orig-
inal MSE algorithm is hereby proposed in order to intro-
duce the stiness changes of the structural damped system
through an iterative procedure which more appropriately
considers the eects of the variation of the storage and
loss moduli with frequency. As a consequence, an itera-
tive calculation of the real (undamped) eigensolution is
performed using the continuously iteratively updated real
part of the stiness matrix, which is updated according
to the viscoelastic material properties at the value of
frequency of the current iteration within the vicinity of
the natural mode being considered. Once the convergence
of the undamped natural frequency of the natural mode
under analysis is veried, the correspondent modal loss
factor is determined by means of the same methodology
used in the original MSE method but utilizing the cor-
respondent imaginary part of the stiness matrix dened
at the converged value of frequency. When necessary, this
method might be repeated according to the modal density
of the bandwidth of interest and employed to determine
the individual modal loss factors and natural frequencies
of the several modes within the considered bandwidth.
Some examples of the use of the IMSE in the context
of FE models can be found, for example, in [Trindade
et al. (2000b), Zhang and Chen (2006)]. A schematic of
the algorithm describing the IMSE approach is presented
in Figure 2 where it is assumed that the eigensolution
algorithm calculates the smallest eigenvalues.
Iterative Complex Eigensolution (ICE) Approach Re-
sembling the IMSE method, an alternative iterative com-
plex eigensolution (ICE) algorithm is proposed in this
section. This algorithm is exact in the sense that it does
not estimate the modal loss factor in the same way as
the MSE but, instead, it uses the determined complex
eigenvalue to calculate the exact modal loss factor value.
The resultant ICE method can therefore be seen as a more
accurate and generally applicable method. However, it has
the drawback of requiring an higher computational eort
since, for each mode, repeated complex eigensolutions are
sought before the convergence is achieved.
As the previous method it starts by determining the
real eigensolution of the frequency dependent undamped
FE model at a dened initial frequency. Based on that
preliminary analysis to give a rst guess of the value
of the damped natural frequencies, the proposed method
successively updates the complex stiness matrix and the
corresponding complex eigenvalues are obtained from a
complex eigensolution until the convergence satisfying the
necessary accuracy condition is achieved. A schematic of
the algorithm describing the ICE approach is presented in
Figure 3 where, in a similar way, it is assumed that the
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
85
IMSE algorithm
Step 1. Eigensolution with =
0
Re[K(j
0
)]
r
= (
0
r
)
2
M
r
Step 2. Loop for each eigenpair
_

0
r
,
r
_
with r = 1, . . . , p
i. Initial value

i
r
=
0
r
ii. Iterative loop for each natural frequency and mode shape
Eigensolution
_

0
s
,
s
_
with s = 1, . . . , r
Re[K(j
i
r
)]
s
= (
i+1
s
)
2
M
s
Iterated natural frequency
_

i+1
s
: rejected, for s < r

i+1
r
=
_
(
i+1
s
)
2
, for s = r
Convergence condition test

i+1
r

i
r

i+1
r

max
iii. Modal loss factor

r
=

T
r
Im[K(j
i+1
r
)]
r

T
r
Re[K(j
i+1
r
)]
r
Fig. 2. Iterative modal strain energy (IMSE) algorithm.
ICE algorithm
Step 1. Eigensolution with =
0
Re[K(j
0
)]
r
= (
0
r
)
2
M
r
Step 2. Loop for each eigenpair (
r
,
r
) with r = 1, . . . , p
i. Initial value

i
r
=
0
r
ii. Iterative loop for each complex eigenvalue and eigenvector
Complex eigensolution (
s
,
s
) with s = 1, . . . , r
K(j
i
r
)
s
=
i+1
s
M
s
Iterated damped natural frequency
_

i+1
s
: rejected, for s < r

i+1
r
=
_
Re(
i+1
s
), for s = r
Convergence condition test

i+1
r

i
r

i+1
r

max
iii. Modal loss factor

r
=
Im(
i+1
r
)
Re(
i+1
r
)
Fig. 3. Iterative complex eigensolution (ICE) algorithm.
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
86
eigensolution algorithm calculates the smallest magnitude
eigenvalues.
When compared with conventional methods, the ICE
method improves considerably the accuracy while it some-
what maintains the computational eciency and simplic-
ity required for practical applications. Referring to the
two iterative approaches, namely the IMSE and ICE, in
both methods as the iteration continues, the estimated

i+1
r
,
i+1
r
and
i+1
r
will converge to more accurate solu-
tions,
r
,
r
and
r
. However, the computational eort
involved may increase with the modal density and/or
the bandwidth of interest, since, for each mode, several
real or complex eigensolutions are required before the
convergence is achieved. In opposition to that we have
the too simplistic MSE approach with the advantage that
these shortcomings are not an issue. However, the MSE
is an approximate method which considers some unre-
alistic simplications which make its accuracy strongly
dependent upon the level of frequency dependent damping
and stiness and also of the degree of isolation of the
mode to be analyzed. When compared with the DFR,
where the modal parameters are indirectly approximately
determined by modal identication procedures [Maia and
Silva (1997), Ewins (2000)], the IMSE and ICE allow
the modal parameters
r
,
r
, and
r
of the rth mode
of a generic viscoelastically damped structural system to
be obtained directly. However, with the aforementioned
advantages and disadvantages and modeling strategies, in
principle all the models can be used to build a modal
model of the damped structural system in a more or less
straightforward manner, which can be used to estimate
both frequency and time domain responses.
In conclusion, the proposed ICE method makes use of
classical complex eigensolution algorithms and represents
an extension and improvement to IMSE method, where
only real eigensolution algorithms are utilized. To the
authors best knowledge, this method has been seldom
used in the context of viscoelastically damped structures
and the discussion in the open literature regarding its
application and performance is very scarce [Lin and Lim
(1996)].
Methods based in the denition of time domain viscoelas-
tically damped spatial models which afterwards allow ob-
taining representative truncated modal models will be
discussed in the following sections.
3.3 Time Domain Based Approaches
Golla-Hughes-McTavish (GHM) Damping Model When
general transient responses are required, time domain
models are more suitable and versatile and they might rep-
resent better alternatives than the CMA-based frequency
domain methods, since they allow the reduction of the
computational burden due to the re-calculation of the
stiness matrix for each discrete frequency value (DFR)
and the use of iterative eigenproblem calculations (IMSE
and ICE). One alternative is the GHM (Golla-Hughes-
McTavish) model [Golla and Hughes (1985), McTavish
and Hughes (1993)], which assumes that the characteristic
material function

G(s) s

G
rel
(s) in Equation (3) may be
represented in terms of a series of mini-oscillator terms
(see Table 1),
s

G
rel
(s) =G

[1 + s r(s)]
=G

_
1 +
n

i=1

i
s
2
+ 2

i

i
s
s
2
+ 2

i

i
s +
2
i
_
, (20)
where G

is the relaxed modulus and each ith mini-


oscillator term is a second order rational function involving
three positive constant parameters,
i
,
i
and

i
. These
parameters govern the shape of the characteristic material
function (usually for isotropic materials the shear modu-
lus) over the complex plane (or over the frequency domain,
assuming only the imaginary part of the complex plane),
and depending upon the nature of the viscoelastic material
and the range of s (or frequency) over which it is to be
modeled, the number of mini-oscillator terms is dened
according to the required accuracy of the representation.
Considering the FE equations of motion of a viscoelas-
tically damped general structural system expressed in
Equation (13) with the shear modulus factored out of
the viscoelastic stiness matrix and assuming that the
structural system possesses only one type of viscoelastic
material, following the hereditary stress-strain law given
in (2), yields
M u(t) +D u(t) +K
E
u(t) + G
rel
(t)

K
V
u(0)
+
_
t
0
G
rel
(t )

K
V
u()

d = f (t), (21)
where

K
V
is the remaining viscoelastic stiness term after
factoring out the shear modulus; it is worthy to mention at
this point that both shear and extensional stiness terms
may be considered in the viscoelastic stiness matrix by
assuming a frequency independent Poissons coecient in
the extensional and shear modulus relationship dened
in Equation (12). Considering nil initial conditions and
transforming Equation (21) to the Laplace domain, we
have
_
s
2
M+ sD+K
E
_
u(s) + s

G
rel
(s)

K
V
u(s) =

f (s). (22)
Substituting the characteristic material function represen-
tation in Equation (20) into (22) yields
_
s
2
M+ sD+K
E
_
u(s) + G

K
V
_
1 +
n

i=1

i
s
2
+ 2

i

i
s
s
2
+ 2

i

i
s +
2
i
_
u(s) =

f (s). (23)
Then, introducing a set of n series of dissipation (or
internal ) variables u
D
i
(s), with i = 1, . . . , n, for each series,
we can establish the relationship
u(s) u
D
i
(s) =
s
2
+ 2

i

i
s
s
2
+ 2

i

i
s +
2
i
u(s). (24)
Substituting Equation (24) into (23) yields
_
s
2
M+ sD+K
E
+K
V
_
u(s)
+ K
V
n

i=1

i
_
u(s) u
D
i
(s)
_
=

f (s), (25)
where K
V
= G

K
V
is the relaxed (or static) stiness
matrix of the viscoelastic components. Next, in order to
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
87
describe the dissipative behavior of the internal DoFs,
after some algebra, Equation (24) can be rewritten as
u
D
i
(s) =

2
i
s
2
+ 2

i

i
s +
2
i
u(s). (26)
Considering Equations (25) and (26), an augmented cou-
pled system can be written as
_
s
2
M+ sD+K
E
+K
V0
_
u(s)
K
V
n

i=1

i
u
D
i
(s) =

f (s), (27a)
_
s
2
1

2
i
+ s
2

i

i
+ 1
_
u
D
i
(s) u(s) = 0, (27b)
where
K
V0
=
_
1 +
n

i=1

i
_
K
V
. (28)
In order to obtain the time-dependent behavior of the aug-
mented system, multiplying Equation (27b) by
i
K
V
,
and since all matrices are independent of s, a linear time
domain model is readily recovered by the inverse Laplace
transform of Equations (27), yielding
M u(t) +D u(t) +
_
K
E
+K
V0
_
u(t)
K
V
n

i=1

i
u
D
i
(t) = f (t), (29a)

i

2
i
K
V
u
D
i
(t) +
2
i

i

i
K
V
u
D
i
(t)
+
i
K
V
u
D
i
(t)
i
K
V
u(t) = 0. (29b)
The augmented coupled system in Equations (29) may still
be expressed in compact matrix form as

Mz(t) +

D z(t) +

Kz(t) =

f (t), (30)
where

M =
_
M 0
0 M
DD
_
,

D =
_
D 0
0 D
DD
_
,

K =
_
K
EE
K
ED
K
DE
K
DD
_
,
(31a-c)
z(t) = col
_
u(t), u
D
1
(t), . . . , u
D
n
(t)

f (t) = col [f (t), 0, . . . , 0] , (32a,b)


and
M
DD
=diag
_

1

2
1
K
V
, . . . ,

n

2
n
K
V
_
,
D
DD
=diag
_
2
1

1

1
K
V
, . . . ,
2
n

n

n
K
V
_
,
K
DD
=diag
_

1
K
V
, . . . ,
n
K
V

,
K
ED
=
_

1
K
V
, . . . ,
n
K
V

,
K
EE
=K
E
+K
V0
,
K
DE
= K
T
ED
. (33a-f)
Equation (30) represents a time domain mathematical
realization of a viscoelastically damped FE structural
system comprising viscoelastic and elastic materials. It
is important to emphasize that the derived model is
expressed in a second-order form. However, rst-order
state space formulations and model reduction techniques,
adapted to the problem and matrix topology in hands,
are often more convenient to solve the system equations.
Further details on these issues will be given in the next
sections.
Anelastic Displacement Fields (ADF) Damping Model A
time domain model based on a variation, or Laplace trans-
formed, formulation of the ADF (anelastic displacement
elds) model, originally proposed by Lesieutre and his co-
workers [Lesieutre and Bianchini (1995), Lesieutre et al.
(1996)], is presented in this work. It takes a denition
of the complex modulus in the frequency (or Laplace)
domain (see Table 1) and utilizes the so-called internal,
dissipation or anelastic (after Lesieutre) variables to sim-
plify and linearize the damped equations, although with
the drawback of increasing the size of the problem. Af-
terward, through an inverse Laplace transform we obtain
an amenable and computationally tractable augmented
system of linear ordinary dierential equations that can
be solved by standard numerical methods applicable to
rst-order linear systems. With this procedure, the FE
model implementation of the ADF model is more straight-
forward when compared with the Lesieutres original di-
rect time domain formulation, based upon the method
of irreversible thermodynamics and a decomposition of
the total displacement eld in an elastic and anelastic
counterpart. Thus, the process of deriving the augmented
coupled elastic-anelastic system (using the original desig-
nation of Lesieutre) is similar to what has been presented
for the GHM method. In this case, a dierent denition
of the characteristic material function G(s) is utilized. It
is also worthy to mention that Lesieutres denition of
the relaxation function is in fact a particular case of the
relaxation function derived by Biot and presented in Table
1, where the viscous term proportional to velocity is not
considered in the relaxation function denition. That is in
fact expected since both are derived from thermodynamic
principles. However, velocity proportional damping is still
considered here by means of a viscous damping model,
through the introduction of the viscous damping matrix
D, which is used here to model other general sources
of damping and not the viscoelastic material damping.
Furthermore, when compared with the GHM model, which
can also be seen as a particular case of Biots model
assuming a time-domain relaxation function as a series
of double exponential terms, the ADF approach has been
shown to yield the same damping capabilities with less
parameters per each series.
As reported by Lesieutre, the characteristic material func-
tion, which in this case corresponds to the frequency de-
pendent viscoelastic shear modulus, described by the ADF
model is represented by a series of functions in the Laplace
domain and is given by [Lesieutre (1992), Lesieutre and
Bianchini (1995); cf. Table 1]
s

G
rel
(s) = G

[1 + s r(s)] = G

_
1 +
n

i=1

i
s
s +
i
_
, (34)
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
88
where, again, G

is the relaxed (or static, low-frequency)


shear modulus, and
i
is the inverse of the characteristic
relaxation time at constant strain and
i
the correspon-
dent relaxation resistance. To take into consideration the
relaxation behavior, the entire ADF model itself may be
comprised of several individual elds, where n series of
ADF are used to describe the material behavior. Given
a set of measured values of the shear modulus in the
form of a frequency dependent complex modulus G

(j),
the relaxed shear modulus G

and the series of material


parameters
i
and
i
can be determined through curve t-
ting techniques. The number of series of ADF parameters
determines the accuracy of the matching of the measured
material data over the frequency range of interest.
Substituting Equation (34) into (22) yields
_
s
2
M+ sD+K
E
_
u(s)
+ G

K
V
_
1 +
n

i=1

i
s
s +
i
_
u(s) =

f (s). (35)
Then, in a similar way to the GHM, introducing a set
of n series of anelastic (or internal, dissipation) variables
u
A
i
(s), with i = 1, . . . , n, for each series, one can assume
the relationship
u(s) u
A
i
(s) =
s
s +
i
u(s). (36)
Substituting Equation (36) into (35) and considering the
dissipative behavior of the anelastic DoFs given from
Equation (36) as
u
A
i
(s) =

i
s +
i
u(s), (37)
after some algebra, we get the following augmented elastic-
anelastic coupled system,
_
s
2
M+ sD+K
E
+K
V0
_
u(s)
K
V
n

i=1

i
u
A
i
(s) =

f (s), (38a)
_
s

i
+ 1
_
u
A
i
(s) u(s) = 0, (38b)
where, in this case,
K
V0
=
_
1 +
n

i=1

i
_
K
V
. (39)
Multiplying Equation (38b) by
i
K
V
, the time-depen-
dent behavior of the augmented system is recovered by the
inverse Laplace transform of Equations (38), yielding
M u(t) +D u(t) +
_
K
E
+K
V0
_
u(t)
K
V
n

i=1

i
u
A
i
(t) = f (t), (40a)

i
K
V
u
A
i
(t) +
i
K
V
u
A
i
(t)
i
K
V
u(t) = 0.
(40b)
The augmented coupled system in Equations (40) can also
be expressed in compact matrix form, as in Equation (30),
where, for the ADF model, we get the following dierent
denitions,

M =
_
M 0
0 0
_
,

D =
_
D 0
0 D
AA
_
,

K =
_
K
EE
K
EA
K
AE
K
AA
_
,
(41a-c)
z(t) = col
_
u(t), u
A
1
(t), . . . , u
A
n
(t)

, (42)
where
D
AA
=diag
_

1
K
V
, . . . ,

n
K
V
_
,
K
AA
=diag
_

1
K
V
, . . . ,
n
K
V
_
,
K
EE
=K
E
+K
V0
,
K
EA
=
_

1
K
V
, . . . ,
n
K
V

,
K
AE
= K
T
EA
. (43a-e)
Regarding the augmented damped system in Equation
(30), we may notice from matrix

M that, when compared
with the GHM model, the anelastic DoFs of the ADF
model have no inertia and therefore the augmented mass
matrix

M is singular and hence is not positive-denite.
However, the singularity of the augmented mass matrix
can be overcome if, instead of solving the second-order
system (30) of the ADF, a state-space rst-order repre-
sentation with an adequate design of the state variables is
considered. While demanding rst-order solution methods,
an advantage that immediately results of the rst-order
representation is that the size of the ADF augmented
model is smaller than the one obtained with a state-space
rst-order representation of the augmented GHM model,
as will be presented in the following section. With the rst-
order representation, the number of exible modes is kept
the same and the dissipative modes, which correspond to
the internal relaxations of the viscoelastic material, are
overdamped with low observability.
3.4 Model Reduction and State-Space Representation
Associated with the FE discretization and in order to
account for the frequency dependence of the viscoelastic
material, as evidenced by Equations (29) and (40), the
main disadvantage of the GHM and ADF models is that
they require the use of additional DoFs which lead to the
denition of augmented systems with a higher number of
DoFs. In view of this, the initial system size is increased by
a number of additional DoFs equal to the initial number of
DoFs times the number of series utilized by the GHM and
ADF models to t the viscoelastic material constitutive
behavior. However, model reduction techniques might be
utilized in order to reduce the size of the augmented
problem and in part to circumvent this shortcoming.
As suggested by Trindade et al. (2000b), which follows
a similar treatment to the one given by Biot (1954) to
the unobservable variables of a generic irreversible physical
system, the matrices corresponding to the internal DoFs
might be reduced and diagonalized through a projection
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
89
in a suitable reduced modal basis, to reduce the com-
putational cost. Thus, dening I = (D, A) and using it
here to denote internal variables in general, which may be
both the dissipative or anelastic variables denitions of the
GHM and ADF approaches, respectively, and considering
the linear coordinate transformation
u
I
i
(t) =
I
u
I
i
(t), (44)
where
I
=
T
I
K
V

I
is a diagonal matrix composed
by the non-zero eigenvalues of K
V
and
I
is the cor-
respondent matrix of normalized eigenvectors, such that

T
I

I
= I, the vector z(t) in Equation (30) is alternatively
modied to
z(t) = col
_
u(t), u
I
1
(t), . . . , u
I
n
(t)

(45)
and the matrices M
DD
, D
II
, K
II
and K
IE
are modied to
M
DD
=diag
_

1

2
1

D
, . . . ,

n

2
n

D
_
,
D
DD
=diag
_
2
1

1

1

D
, . . . ,
2
n

n

n

D
_
,
K
DD
=diag [
1

D
, . . . ,
n

D
] ,
K
ED
=
_

1
K
V

D
, . . . ,
n
K
V

, (46a-d)
for the GHM model, and for the ADF model to
D
AA
=diag
_

A
, . . . ,

A
_
K
AA
=diag
_

A
, . . . ,
n

A
_
,
K
EA
=
_

1
K
V

A
, . . . ,
n
K
V

. (47a-c)
The advantages of this alternative (transformed) repre-
sentation are that in the case where only some part of the
structure has surface mounted or embedded viscoelastic
materials, only some FEs have viscoelastic components
and K
V
can have several rows and columns of zeros,
which in turn leads to some nil eigenvalues. Thus, the size
of u
I
i
(t) can be substantially smaller than that of u
I
i
(t).
Therefore, through an adequate coordinate transformation
based on the eigenvalues and eigenvectors of K
V
and
elimination of the nil (spurious) eigenvalues, the size of
the problem can be substantially reduced.
As previously mentioned, in comparison with the GHM
model, the ADF model due to the singularity of the
augmented mass matrix

M dened in Equation (41a),
demands a rst-order solution method of the ADF sys-
tem and therefore a state-space representation of the
augmented system which, following the forementioned
procedure, might have also the matrices associated with
the internal DoFs reduced and diagonalized. Therefore,
state-space model realizations are mandatory for the ADF
model and optional to the GHM model, which can also be
solved using classic second-order solution methods.
The state-space vector denition depends on the viscoelas-
tic damping model being used. For the GHM model, a
suitable and adequate design of the state variables would
yield a denition of the state vector x(t) given by
x
D
(t)=
_
z(t)
z(t)
_
, (48)
where the chosen state variables are the reduced aug-
mented vector z(t) given in Equation (45), which is com-
posed by the mechanical elastic DoFs vector u(t) and a set
of reduced dissipative (internal) variables vectors u
D
i
(t),
and its time derivative. However, for the ADF model, in
order to overcome the singularity of the augmented mass
matrix, a suitable and convenient denition of x(t) would
be given by
x
A
(t)=
_
z(t)
u(t)
_
, (49)
where the chosen state variables are the reduced aug-
mented vector z(t), composed by the mechanical elastic
DoFs vector u(t) and a set of reduced anelastic (internal)
variables vectors u
A
i
(t), and the time derivative of the
elastic DoFs vector, u(t). It is worthy to emphasize at this
point that the time derivatives of u
A
i
(t) are not considered
here since these variables are massless, i.e. the matrix

M
is singular.
Thus, taking the forementioned state-space vector deni-
tions into account and considering also the reduced and
diagonalized matrix denitions in Equations (46)-(47), the
augmented coupled system given in Equations (29) and
(40) for the GHM and ADF models, respectively, expressed
by the generic second-order system equation in compact
matrix form given in Equation (30), can be written in a
rst-order state-space form in terms of the generic state
variables vector x
I
(t), yielding
x
I
(t) = Ax
I
(t) +Bf (t), (50)
where the state-space system and input matrices, A and
B, for the GHM model are given by
A =
_

_
0 0 I 0
0 0 0 I
M
1
K
EE
M
1
K
ED
M
1
D 0
M
1
DD
K
DE
M
1
DD
K
DD
0 M
1
DD
D
DD
_

_,
B =
_

_
0
0
M
1
0
_

_,
(51a,b)
and for the ADF model are modied to
A =
_
_
0 0 I
D
1
AA
K
AE
D
1
AA
K
AA
0
M
1
K
EE
M
1
K
EA
M
1
D
_
_
,
B =
_
_
0
0
M
1
_
_
. (52a,b)
The analysis of the previous matrix denitions, consistent
with the state-space vector denitions in Equations (48)
and (49), demonstrates that the rst-order form of the
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
90
GHM model has an higher number of DoFs than the ADF
model. In general terms, the number of total DoFs of the
GHM state-space model is given by 2n
E
+ 2nn
D
and of
the ADF model by 2n
E
+nn
A
, where n
E
is the number of
elastic DoFs, n is the number of series of model parameters
used to t the viscoelastic constitutive behavior, and n
D
and n
A
are the number of dissipative and anelastic DoFs
per series. In view of this, the dierence between the GHM
and ADF system sizes is equal to nn
I
with n
I
being the
number of internal (dissipative, anelastic) DoFs per series.
It is worth to emphasize that the alternative transformed
(reduced) denitions of the matrices presented in Equa-
tions (46)-(47) should be used in order to reduce the
size of the system which, as previously discussed, can
increase signicantly, sometimes dramatically aecting the
required computational time. This procedure has two main
advantages: (i) diagonalization of the matrices, reducing
the number of non-zero elements, and (ii) matrices reduc-
tion, only if some rows and columns of the matrices are
zeros. The latter advantage means that some of the elastic
DoFs may not be coupled with the internal DoFs, as is the
case for example when we have a structure only partially
covered by viscoelastic damping treatments. Therefore,
for a fully covered structure, i.e. with all the elastic and
internal DoFs coupled, with this basis transformation we
would only benet from advantage (i) since the matrices
do not have any null eigenvalue and therefore the size of
the matrices would be the kept the same. In those cases,
the general GHM and ADF denitions of the total number
of DoFs can be particularized since the number of elastic
DoFs is equal to the number of internal DoFs per series,
i.e. n
E
= n
I
. In view of this, the total number of DoFs of
the augmented fully coupled system is given by (2+2n)n
E
for the GHM and (2+n)n
E
for the ADF state-space model.
Even with a modal reduction of the DoFs of the viscoelas-
tic elements, the order of the system quickly increases as
the number of series of GHM or ADF parameters used in
the summation is increased. Larger order models makes
simulation and design more dicult. It is therefore ad-
vantageous to look at more model reduction techniques to
reduce the systems size even further. Model reduction in
structural dynamics, where the original state space system
is approximated by an equivalent system with a lower
dimension, can be achieved by a complex modal projection
of the original system and a subsequent truncation of the
number of modes considered. A study addressing reduced-
order FE models of viscoelastically damped beams, where
the selective signicance of the non-physical modes to
the net mechanical response is performed through internal
variables projection in order to signicantly reduce the
computational cost, was recently presented by Trindade
(2006), and the reader is also referred to Biot (1954), Yiu
(1993, 1994), Friswell and Inman (1999), Park et al.
(1999), Trindade et al. (2000b), Vasques and Dias Ro-
drigues (2008) and Vasques (2008) for further details about
state space design and model reduction techniques.
4. CONCLUSION
This article presents FE-based mathematical strategies to
model the damped constitutive behavior of viscoelastic
materials used as surface mounted, constrained or embed-
ded damping treatments in structures, in order to reduce
vibrations and/or noise radiation.
Both time and frequency domain based techniques were
considered to model the constitutive behavior and the
implementation of these approaches into FE solution pro-
cedures was presented and discussed. Time domain tech-
niques regard the use of the GHM and ADF internal
variables models. When a state-space rst-order represen-
tation with an adequate design of the state variables is con-
sidered, the ADF model is known to lead to an augmented
model of the damped structural system with a smaller size
than the GHM model. In the authors opinion the ADF
model represents the best internal variables alternative
to accurately model the damping behavior since it yields
good trade-o between accuracy and complexity. One ma-
jor disadvantage in using internal variables models, such as
the GHM or ADF, is the creation of additional dissipation
variables increasing the size of the coupled damped FE
model. Frequency domain techniques comprise the use of
CMA-based methods, namely the DFR, IMSE and ICE
(which is recalled in this work), where the FE spatial
model is used by re-calculating the complex viscoelastic
stiness matrix for each discrete frequency value, in the
case of the DFR, or during the iterative eigensolution pro-
cess, in the case of the IMSE and ICE. These approaches
are more straightforward to use and implement at the
global FE level. That is the reason why the CMA-based
method, DFR, is the most common approach implemented
in commercial FE codes incorporating viscoelastic damp-
ing modeling capabilities.
I
d
e
n
t
i
f
i
c
a
t
i
o
n
S
y
n
t
h
e
t
i
z
a
t
i
o
n
E
i
g
e
n
s
o
l
u
t
i
o
n
REQUENCY ESPONSE IME ESPONSE
Fig. 4. FE-based viscoelastic time and frequency domain
solution alternatives.
The ultimate aim of all these viscoelastic damping models
is to be able to simulate the time and frequency response
of viscoelastically damped structural systems. While the
frequency response is straightforward to obtain, whether
directly obtained (as is the case for the DFR approach) or
obtained through modal models derived from the spatial
model either by iterative frequency dependent eigenso-
lutions (IMSE and ICE) or from the augmented spatial
model eigensolution (GHM and ADF), the time domain
response can be obtained from the spatial model either
by direct integration methods or by the modal models
using the superposition principle (see Figure 4). With the
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
91
advantages and disadvantages mentioned thus far, all the
approaches can be used to build a truncated modal model
of the damped structural system, whether from the spatial
model or from a FRF model generated with the DFR
method, which can be used to estimate both frequency
and time domain responses. However, it is important to
emphasize that when considering structures with a high
modal density, possessing modes not well separated and
strongly damped, the modal identication methods may
not be accurate and ecient, rendering the identication
procedure more troublesome, if not impossible.
The experimental identication of viscoelastic materials,
with the 3M ISD112 being considered in this work, and
the aforementioned FE integrated damped modeling ap-
proaches are assessed and validated in the companion
article [Vasques et al. (2010)].
REFERENCES
Adhikari, S. (1998). Energy Dissipation in Vibrating Struc-
tures, PhD (First Year Report), Engineering Depart-
ment, University of Cambridge, Cambridge, UK.
Adhikari, S. and Woodhouse, J. (2001a). Identication
of damping: Part 1, viscous damping, Journal of Sound
and Vibration 243(1): 4361.
Adhikari, S. and Woodhouse, J. (2001b). Identication
of damping: Part 2, non-viscous damping, Journal of
Sound and Vibration 243(1): 6388.
Adhikari, S. and Woodhouse, J. (2003). Quantication of
non-viscous damping in discrete linear systems, Journal
of Sound and Vibration 260(3): 499518.
Adolfsson, K., Enelund, M. and Olsson, P. (2005). On the
fractional order model of viscoelasticity, Mechanics of
Time-Dependent Materials 9(1): 1534.
Alfrey, T. and Doty, P. (1945). The methods of specify-
ing the properties of viscoelastic materials, Journal of
Applied Physics 16(11): 700713.
Bagley, R. L. (1989). Power law and fractional calculus
model of viscoelasticity, AIAA Journal 27(10): 1412
1417.
Bagley, R. L. and Torvik, P. J. (1983). Fractional calculus
- A dierent approach to the analysis of viscoelastically
damped structures, AIAA Journal 21(5): 741748.
Bagley, R. L. and Torvik, P. J. (1985). Fractional calculus
in the transient analysis of viscoelastically damped
structures, AIAA Journal 23(6): 918925.
Bagley, R. L. and Torvik, P. J. (1986). On the fractional
calculus model of viscoelastic behavior, Journal of Rhe-
ology 30(1): 133155.
Balm`es, E. (1997). Model reduction for systems with
frequency dependent damping properties, 15th Interna-
tional Modal Analysis Conference - IMAC XV, Vol. 1,
Society for Experimental Mechanics, Orlando, FL, US,
pp. 223229.
Banks, H. T. and Pinter, G. A. (2001). Hysteretic damp-
ing, in D. J. Ewins, S. S. Rao and S. G. Braun (eds), En-
cyclopedia of Vibration, Academic Press, Oxford, UK,
pp. 658664.
Baz, A. (1993). Active constrained layer damping, Pro-
ceedings of Damping 93, Vol. 3, San Francisco, CA, US,
pp. IBB 123.
Baz, A. (1998). Robust control of active constrained layer
damping, Journal of Sound and Vibration 211(3): 467
480.
Baz, A. (2000). Spectral nite-element modeling of the
longitudinal wave propagation in rods treated with
active constrained layer damping, Smart Materials and
Structures 9(3): 372377.
Benjeddou, A. (2001). Advances in hybrid active-passive
vibration and noise control via piezoelectric and vis-
coelastic constrained layer treatments, Journal of Vi-
bration and Control 7(4): 565602.
Bert, C. W. (1973). Material damping: An introductory
review of mathematic measures and experimental tech-
nique, Journal of Sound and Vibration 29(2): 129153.
Biot, M. A. (1954). Theory of stress-strain relations in
anisotropic viscoelasticity and relaxation phenomena,
Journal of Applied Physics 25(11): 13851391.
Biot, M. A. (1955). Variational principles in irre-
versible thermodynamics with application to viscoelas-
ticity, Physical Review 97(6): 14631469.
Biot, M. A. (1972). Simplied dynamics of multilayered
orthotropic viscoelastic plates, International Journal of
Solids and Structures 8(4): 491509.
Bishop, R. E. D. (1955). The treatment of damping forces
in vibration theory, Journal of the Royal Aeronautical
Society 59(539): 738742.
Boltzmann, L. (1874). Zur theorie der elastischen
nachwirkung, Sitzungsberichte der Mathematisch-Na-
turwissenschaftlichen Klasse der Kaiserlichen Akademie
der Wissenschaften 70(2): 275306.
Brackbill, C. R., Lesieutre, G. A., Smith, E. C. and
Govindswamy, K. (1996). Thermomechanical modeling
of elastomeric materials, Smart Materials and Structures
5(5): 529539.
Brinson, H. F. and Brinson, L. C. (2008). Polymer En-
gineering Science and Viscoelasticity: An Introduction,
Springer, New York.
Buhariwala, K. J. and Hansen, J. S. (1988). Dynamics of
viscoelastic structures, AIAA Journal 26(2): 220227.
Burgers, J. M. (1935). First and Second Report on Viscos-
ity and Plasticity, Academy of Sciences at Amsterdam,
Amsterdam, NL.
Chen, C. P. and Lakes, R. S. (1993). Viscoelastic behaviour
of composite materials with conventional- or negative-
Poissons-ratio foam as one phase, Journal of Materials
Science 28(16): 42884298.
Christensen, R. M. (1982). Theory of Viscoelasticity: An
Introduction, 2nd edn, Academic Press, New York, US.
Cliord, S., Jansson, N., Yu, W., Michaud, V. and Manson,
J. A. (2006). Thermoviscoelastic anisotropic analysis
of process induced residual stresses and dimensional
stability in real polymer matrix composite components,
Composites Part A: Applied Science and Manufacturing
37(4): 538545.
Coulomb, C. (1784). Recherches theoriques et experimen-
tales: Sur la force de torsion et sur lelasticite des ls
de metal, Memoires de lAcademie Royale des Sciences
pp. 229269.
Crandall, S. H. (1970). Role of damping in vibration
theory, Journal of Sound and Vibration 11(1): 318.
Cremer, L., Heckl, M. and Petersson, B. A. T. (2005).
Structure-borne sound: structural vibrations and sound
radiation at audio frequencies, 3rd edn, Springer, Berlin,
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
92
DE.
de Silva, C. W. (2005). Vibration damping, in C. W.
de Silva (ed.), Vibration and Shock Handbook, CRC
Press, Boca Raton, FL, US.
Demer, L. J. (1956). Bibliography of the material damping
eld (with abstracts and punch card codings), Techni-
cal Report WADC-TR-56-180, Wright Air Development
Center, Wright-Patterson Air Force Base, Ohio, US.
Demirdzic, I., Dzaferovic, E. and Ivankovic, A. (2005).
Finite-volume approach to thermoviscoelasticity,
Numerical Heat Transfer, Part B: Fundamentals
47(3): 213237.
Douglas, B. E. (1977). The Transverse Vibratory Response
of Partially Constrained Elastic-Viscoelastic Beams,
PhD thesis, Department of Mechanical Engineering,
University of Maryland at College Park, Maryland, US.
Douglas, B. E. and Yang, J. C. S. (1978). Transverse
compressional damping in vibratory response of elastic-
viscoelastic-elastic beams, AIAA Journal 16(9): 925
930.
Doyle, J. F. (1997). Wave Propagation in Structures: Spec-
tral Analysis Using Fast Discrete Fourier Transforms,
2nd edn, Springer, New York, US.
Enelund, M. and Josefson, B. L. (1997). Time-domain
nite element analysis of viscoelastic structures with
fractional derivatives constitutive relations, AIAA Jour-
nal 35(10): 16301637.
Enelund, M. and Lesieutre, G. A. (1999). Time domain
modeling of damping using anelastic displacement elds
and fractional calculus, International Journal of Solids
and Structures 36(29): 44474472.
Enelund, M. and Olsson, P. (1999). Damping described by
fading memoryanalysis and application to fractional
derivative models, International Journal of Solids and
Structures 36(7): 939970.
Ewins, D. J. (2000). Modal Testing: Theory, Practice and
Application, 2nd edn, Research Studies Press, Baldock,
UK.
Ferry, J. D. (1980). Viscoelastic Properties of Polymers,
3rd edn, John Wiley & Sons, New York, US.
Findley, W. N., Lai, J. S. and Onaran, K. (1989). Creep
and Relaxation of Nonlinear Viscoelastic Materials:
With an Introduction to Linear Viscoelasticity, Dover,
New York, US.
Friswell, M. I. and Inman, D. J. (1998). Hybrid damping
treatments in thermal environments, in G. R. Tomlinson
and W. A. Bullough (eds), Smart Materials and Struc-
tures, IOP Publishing, Bristol, UK, pp. 667674.
Friswell, M. I. and Inman, D. J. (1999). Reduced-
order models of structures with viscoelastic components,
AIAA Journal 37(10): 13181325.
Fung, Y. C. (1993). Biomechanics: Mechanical Properties
of Living Tissues, 2nd edn, Springer, New York, US.
Fung, Y. C. and Tong, P. (2001). Classical and Computa-
tional Solid Mechanics, World Scientic, Singapore, SG.
Galucio, A. C., De u, J.-F., Mengue, S. and Dubois, F.
(2006). An adaptation of the gear scheme for fractional
derivatives, Computer Methods in Applied Mechanics
and Engineering 195(44-47): 60736085.
Galucio, A. C., De u, J.-F. and Ohayon, R. (2004). Finite
element formulation of viscoelastic sandwich beams us-
ing fractional derivative operators, Computational Me-
chanics 33(4): 282291.
Galucio, A. C., De u, J.-F. and Ohayon, R. (2005). A frac-
tional derivative viscoelastic model for hybrid active-
passive damping treatments in time domain Applica-
tion to sandwich beams, Journal of Intelligent Material
Systems and Structures 16(1): 3345.
Gandhi, F. (2001). Viscous damping, in D. J. Ewins, S. S.
Rao and S. G. Braun (eds), Encyclopedia of Vibration,
Academic Press, Oxford, UK, pp. 15481550.
Gemant, A. (1936). A method of analyzing experimental
results obtained from elasto-viscous bodies, Journal of
Applied Physics 7(8): 311317.
Golla, D. F. and Hughes, P. C. (1985). Dynamics of
viscoelastic structuresA time-domain, nite element
formulation, Jounal of Applied Mechanics 52(12): 897
906.
Goodman, L. E. (2002). Material damping and slip damp-
ing, in C. M. Harris and A. G. Piersol (eds), Harris
Shock and Vibration Handbook, 5th edn, McGraw-Hill,
New York, US.
Hammerand, D. C. and Kapania, R. K. (1999). Ther-
moviscoelastic analysis of composite structures using a
triangular at shell element, AIAA Journal 37(2): 238
247.
Hu, B. G., Dokainish, M. A. and Mansour, W. M. (1995).
A modied MSE method for viscoelastic systems: A
weighted stiness matrix approach, Journal of Vibration
and Acoustics 117(2): 226231.
Hwang, S. J. and Gibson, R. F. (1992). The use of
strain energy-based nite-element techniques in the
analysis of various aspects of damping of composite-
materials and structures, Journal of Composite Mate-
rials 26(17): 25852605.
Inman, D. (2001). Damping models, in D. J. Ewins, S. S.
Rao and S. G. Braun (eds), Encyclopedia of Vibration,
Academic Press, Oxford, UK, pp. 335342.
Johnson, A. R. (1999). Modeling viscoelastic materials
using internal variables, Shock and Vibration Digest
31(2): 91100.
Johnson, C. D. and Kienholz, D. A. (1982). Finite element
prediction of damping in structures with constrained
viscoelastic layers, AIAA Journal 20(9): 12841290.
Johnson, C. D., Kienholz, D. A. and Rogers, L. C. (1980).
Finite element prediction of damping in beams with
constrained viscoelastic layers, Shock and Vibration Bul-
letin (1): 7181.
Jones, D. I. G. (2001). Handbook of Viscoelastic Vibration
Damping, John Wiley & Sons, Chichester, UK.
Kelly, W. J. and Stevens, K. K. (1989). Application
of perturbation techniques to the modal analysis of a
shaft with added viscoelastic damping, 7th International
Modal Analysis Conference - IMAC VII, Society of
Experimental Mechanics, Las Vegas, NV, US.
Kelvin, L. (1875). Elasticity, Encyclopedia Britannica, 9th
edn.
Kerwin, E. M. (1959). Damping of exural waves by a
constrained viscoelastic layer, Journal of the Acoustical
Society of America 31(7): 952962.
Koeller, R. C. (1984). Applications of fractional calculus
to the theory of viscoelasticity, Journal of Applied Me-
chanics 51(2): 299307.
Koeller, R. C. (1986). Polynomial operators, Stieltjes
convolution, and fractional calculus in hereditary me-
chanics, Acta Mechanica 58(3-4): 251264.
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
93
Lakes, R. S. and Wineman, A. (2006). On Poissons
ratio in linearly viscoelastic solids, Journal of Elasticity
85(1): 4563.
Lazan, B. J. (1959). Energy dissipation mechanisms in
structures with particular reference to material damp-
ing, in J. Ruzicka (ed.), Structural Damping, ASME,
Atlantic City, US, pp. 134.
Lesieutre, G. A. (1992). Finite elements for dynamic
modeling of uniaxial rods with frequency-dependent
material properties, International Journal of Solids and
Structures 29(12): 15671579.
Lesieutre, G. A. (2001). Damping in FE models, in D. J.
Ewins, S. S. Rao and S. G. Braun (eds), Encyclopedia
of Vibration, Academic Press, Oxford, pp. 321327.
Lesieutre, G. A. and Bianchini, E. (1995). Time do-
main modeling of linear viscoelasticity using anelastic
displacement elds, Journal of Vibration and Acoustics
117(4): 424430.
Lesieutre, G. A., Bianchini, E. and Maiani, A. (1996).
Finite element modeling of one-dimensional viscoelastic
structures using anelastic displacement elds, Journal
of Guidance Control and Dynamics 19(3): 520527.
Lesieutre, G. A. and Govindswamy, K. (1996). Fi-
nite element modeling of frequency dependent and
temperature-dependent dynamic behavior of viscoelas-
tic materials in simple shear, International Journal of
Solids and Structures 33(3): 419432.
Lin, R. M. and Lim, M. K. (1996). Complex eigensensitivi-
ty-based characterization of structures with viscoelastic
damping, Journal of the Acoustical Society of America
100(5): 31823191.
Mace, M. (1994). Damping of beam vibrations by means of
a thin constrained viscoelastic layer: Evaluation of a new
theory, Journal of Sound and Vibration 172(5): 557
591.
Maia, N. M. M. and Silva, J. M. M. (eds) (1997). Theoret-
ical and experimental modal analysis, Research Studies
Press, Baldock, UK.
Maxwell, J. C. (1868). On the dynamical theory of gases,
Philosophical Magazine 35: 129145 and 185217.
McTavish, D. J. and Hughes, P. C. (1993). Modeling of
linear viscoelastic space structures, Journal of Vibration
and Acoustics 115(1): 103110.
Mead, D. J. (1960). The eect of a damping com-
pound on jet-eux excited vibration, Aircraft Engineer-
ing 32(1): 6472.
Mead, D. J. (1998). Passive Vibration Control, John Wiley
& Sons, Chichester, UK.
Mead, D. J. (2002). Structural damping and damped vi-
bration, Applied Mechanics Reviews 55(6): R45 (10 p.).
Mead, D. J. and Markus, S. (1969). The forced vibration
of a three-layer damped sandwich beam with arbitrary
boundary conditions, Journal of Sound and Vibration
10(2): 163.
Moreira, R. A. S. and Rodrigues, J. D. (2006). Partial con-
strained viscoelastic damping treatment of structures: A
modal strain energy approach, International Journal of
Structural Stability and Dynamics 6(3): 397411.
Moreira, R. and Rodrigues, J. D. (2004). Constrained
damping layer treatments: Finite element modeling,
Journal of Vibration and Control 10(4): 575595.
Myklestad, N. O. (1952). The concept of complex damp-
ing, Journal of Applied Mechanics 19(3): 284286.
Nashif, A., Jones, D. and Henderson, J. (1985). Vibration
Damping, John Wiley & Sons, New York, US.
Ottosen, N. S. and Ristinmaa, M. (2005). The Mechanics
of Constitutive Modeling, Elsevier, Amsterdam, NL.
Padovan, J. (1987). Computational algorithms for FE for-
mulations involving fractional operators, Computational
Mechanics 2(4): 271287.
Park, C. H., Inman, D. J. and Lam, M. J. (1999). Model
reduction of viscoelastic nite element models, Journal
of Sound and Vibration 219(4): 619637.
Park, S. W. and Schapery, R. A. (1999). Methods of in-
terconversion between linear viscoelastic material func-
tions. Part I A numerical method based on Prony
series, International Journal of Solids and Structures
36(11): 16531675.
Peters, R. D. (2005). Damping theory, in C. W. de Silva
(ed.), Vibration and Shock Handbook, CRC Press, Boca
Raton, FL, US.
Plouin, A.-S. and Balm`es, E. (1998). Pseudo-modal rep-
resentations of large models with viscoelastic behavior,
16th International Modal Analysis Conference - IMAC
XVI, Vol. 2, Society for Experimental Mechanics, Santa
Barbara, CA, US, pp. 14401446.
Plouin, A.-S. and Balm`es, E. (1999). A test validated
model of plates with constrained viscoelastic materi-
als, 17th International Modal Analysis Conference -
IMAC XVII, Vol. 1, Society for Experimental Mechan-
ics, Kissimmee, FL, US, pp. 194200.
Plump, J. M. and Hubbard, J. E. (1986). Modeling
of active constrained layer damped, 12th International
Congress on Acoustics, Toronto, CA, Paper no. D4-1.
Pradeep, V. and Ganesan, N. (2006). Vibration behavior
of ACLD treated beams under thermal environment,
Journal of Sound and Vibration 292(3-5): 10361045.
Pritz, T. (1996). Analysis of four-parameter fractional
derivative model of real solid materials, Journal of
Sound and Vibration 195(1): 103115.
Pritz, T. (2000). Measurement methods of complex Pois-
sons ratio of viscoelastic materials, Applied Acoustics
60(3): 279292.
Prony, R. (1795). Essai experimentale et analytique, Jour-
nal de lEcole Polytechnique de Paris (1): 2776.
Riande, E., Calleja, R. D., Prolongo, M. G., Masegosa,
R. M. and Salom, C. (2000). Polymer Viscoelasticity:
Stress and strain in practice, Marcel Dekker, New York,
US.
Schapery, R. A. (1964). Application of thermodynam-
ics to thermomechanical, fracture, birefringent phenom-
ena in viscoelastic media, Journal of Applied Physics
35(5): 14511465.
Schmidt, A. and Gaul, L. (2002). Finite element for-
mulation of viscoelastic constitutive equations using
fractional time derivatives, Nonlinear Dynamics 29(1-
4): 3755.
Silva, L. A. (2003). Internal Variable and Temperature
Modeling Behavior of Viscoelastic Structures A Con-
trol Analysis, PhD thesis, Virginia Tech, VA, USA.
Silva, L. A., Austin, E. M. and Inman, D. J. (2005). Time-
varying controller for temperature-dependent viscoelas-
ticity, Journal of Vibration and Acoustics 127(3): 215
222.
Slanik, M. L., Nemes, J. A., Potvin, M. J. and Piedboeuf,
J. C. (2000). Time domain nite element simulations
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
94
of damped multilayered beams using a Prony series
representation, Mechanics of Time-Dependent Materials
4(3): 211230.
Slater, J. C., Belvin, W. K. and Inman, D. J. (1993).
Survey of modern methods for modeling frequency de-
pendent damping in nite element models, 11th Inter-
national Modal Analysis Conference - IMAC XI, Vol.
1923, Kissimmee, FL, US, pp. 15081512.
Snowdon, J. C. (1968). Vibration and Shock in Damped
Mechanical Systems, John Wiley & Sons, New York, US.
Soni, M. L. and Bogner, F. K. (1982). Finite-element
vibration analysis of damped structures, AIAA Journal
20(5): 700707.
Stiassnie, M. (1979). Application of fractional calculus for
the formulation of viscoelastic models, Applied Mathe-
matical Modelling 3(4): 300302.
Sun, C. T. and Lu, Y. P. (1995). Vibration Damping
of Structural Elements, Prentice Hall, Englewood Clis,
NJ, US.
Taylor, R. L., Pister, K. S. and Goudreau, G. L. (1970).
Thermomechanical analysis of viscoelastic solids, Inter-
national Journal for Numerical Methods in Engineering
2(1): 4559.
Trindade, M. A. (2006). Reduced-order nite element
models of viscoelastically damped beams through in-
ternal variables projection, Journal of Vibration and
Acoustics 128(4): 501508.
Trindade, M. A. and Benjeddou, A. (2002). Hybrid
active-passive damping treatments using viscoelastic
and piezoelectric materials: Review and assessment,
Journal of Vibration and Control 8(6): 699745.
Trindade, M. A., Benjeddou, A. and Ohayon, R. (2000a).
Finite element analysis of frequency- and temperature-
dependent hybrid active-passive vibration damping, Re-
vue Europeenne des

Elements Finis 9(1-3): 89111.
Trindade, M. A., Benjeddou, A. and Ohayon, R. (2000b).
Modeling of frequency-dependent viscoelastic materials
for active-passive vibration damping, Journal of Vibra-
tion and Acoustics 122(2): 169174.
Tschoegl, N. W. (1989). The Phenomenological The-
ory of Linear Viscoelastic Behaviour: An Introduction,
Springer-Verlag, Berlin, DE.
Tschoegl, N. W. (1997). Time dependence in material
properties: An overview, Mechanics of Time-Dependent
Materials 1(1): 331.
Tschoegl, N. W., Knauss, W. G. and Emri, I. (2002).
Poissons ratio in linear viscoelasticity - a critical review,
Mechanics of Time-Dependent Materials 6(1): 351.
Ungar, E. E. and Kerwin, E. M. (1962). Loss factors of
viscoelastic systems in terms of energy concepts, Journal
of the Acoustical Society of America 34(7): 954957.
Vasques, C. M. A. (2008). Vibration Control of Adaptive
Structures: Modeling, Simulation and Implementaion
of Viscoelastic and Piezoelectric Damping Technologies,
PhD thesis, Faculdade de Engenharia, Universidade do
Porto, Porto, Portugal.
Vasques, C. M. A. and Dias Rodrigues, J. (2008). Com-
bined feedback/feedforward active control of vibration
of beams with ACLD treatments: Numerical simulation,
Computers and Structures 86(3-5): 292306.
Vasques, C. M. A., Mace, B. R., Gardonio, P. and Dias
Rodrigues, J. (2004). Analytical formulation and nite
element modelling of beams with arbitrary active con-
strained layer damping treatments, Technical Memoran-
dum no. 934, Institute of Sound and Vibration Research,
Southampton, UK.
Vasques, C. M. A., Mace, B. R., Gardonio, P. and Ro-
drigues, J. D. (2006). Arbitrary active constrained layer
damping treatments on beams: Finite element modelling
and experimental validation, Computers and Structures
84(22-23): 13841401.
Vasques, C. M. A., Moreira, R. A. S. and Dias Rodrigues,
J. (2010). Viscoelastic damping technologiesPart II:
Experimental identication procedure and validation,
Journal of Advanced Research in Mechanical Engineer-
ing 1(2): 96110.
Vasques, C. M. A. and Rodrigues, J. D. (2008). Numerical
and experimental comparison of the adaptive feedfor-
ward control of vibration of a beam with hybrid active-
passive damping treatments, Journal of Intelligent Ma-
terial Systems and Structures 19(7): 805813.
Voigt, W. (1892).

Uber innere reibung fester koper, ins-
besondere der metalle, Annalen der Physik und Chemie
47: 671693.
Volterra, V. (1909). Sulle equazioni integro-dierenziali
della teoria dellelasticita, Atti della Reale Academia dei
Lecei 18(2): 295301.
Volterra, V. (1913). Lecons sur les Fonctions de Lignes,
Collection De Monographies Sur La Theorie des Fonc-
tions, Gauthier-Villars, Paris, FR.
Wang, G. and Wereley, N. M. (1998). Frequency response
of beams with passively constrained damping layers and
piezo-actuators, in L. P. Davis (ed.), Smart Structures
and Materials 1998: Passive Damping and Isolation,
SPIE, Vol. 3327, Bellingham, WA, US, pp. 4460.
Wang, G. and Wereley, N. M. (2002). Spectral nite
element analysis of sandwich beams with passive con-
strained layer damping, Journal of Vibration and Acous-
tics 124(3): 376386.
Welch, S. W. J., Rorrer, R. A. L. and Duren Jr., R. G.
(1999). Application of time-based fractional calcu-
lus methods to viscoelastic creep and stress relaxation
of materials, Mechanics of Time-Dependent Materials
3: 279303.
Williams, M. L. (1964). Structural analysis of viscoelastic
materials, AIAA Journal 2(5): 785808.
Yiu, Y. C. (1993). Finite element analysis of struc-
tures with classical viscoelastic materials, 34th AIAA/-
ASME/ASCE/AHS/ASC Structures, Structural Dy-
namics and Materials Conference et al., Vol. 4, La Jolla,
CA, US, pp. 21102119.
Yiu, Y. C. (1994). Substructure and nite element for-
mulation for linear viscoelastic materials, 35th AIAA/-
ASME/ASCE/AHS/ASC Structures, Structural Dy-
namics and Materials Conference et al., Vol. 3, Hilton
Head, SC, US, pp. 15851594.
Zener, C. (1948). Elasticity and Anelasticity of Metals,
University of Chicago Press, Chicago, US.
Zhang, S. H. and Chen, H. L. (2006). A study on
the damping characteristics of laminated composites
with integral viscoelastic layers, Composite Structures
74(1): 6369.
Zocher, M. A., Groves, S. E. and Allen, D. H. (1997). A
three-dimensional nite element formulation for thermo-
viscoelastic orthotropic media, International Journal for
Numerical Methods in Engineering 40(12): 22672288.
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2)
Vasques et al. / Viscoelastic Damping TechnologiesPart I : Modeling and / pp. 76-95
95

Das könnte Ihnen auch gefallen