Beruflich Dokumente
Kultur Dokumente
Orientador:
Setembro de 2008
Agradecimentos
A todos aqueles que, directa ou indirectamente, contriburam para a realizao da presente
dissertao, apresento os meus agradecimentos. Em especial:
Aos meus orientadores, Prof. Hlder Carrio Rodrigues e Prof. Joo Orlando Marques Gameiro
Folgado, apresento os meus sinceros agradecimentos pelo apoio e conhecimentos proporcionados,
fundamentais para a execuo deste trabalho.
Ao Prof. Lus Manuel Varejo Oliveira Faria apresento os meus sinceros agradecimentos pela sua
disponibilidade e preciosa colaborao prestada na clarificao de conhecimentos bem como pelas
sugestes propostas durante a realizao deste trabalho.
Prof. Maria de Ftima Reis Vaz agradeo a disponibilidade, os meios fornecidos e o auxlio
prestados na realizao dos ensaios experimentais.
Fundao para a Cincia e Tecnologia agradeo o apoio concedido atravs do Projecto FCT
PTDC/EME-PME/67658/2006 Projecto de materiais celulares elastomricos para controlo passivo
de vibraes.
A todos os meus colegas e amigos, gostaria de expressar o meu agradecimento pela amizade e
apoio que todos me deram ao longo deste trabalho.
Resumo
Actualmente, a competitiva indstria automvel obriga os construtores a prestar cada vez mais
ateno ao conforto dos passageiros e qualidade das deslocaes. Esta presso forou os
projectistas a focar a sua ateno no desenvolvimento de apoios de motor de elevada qualidade, em
que as tcnicas tradicionais de desenvolvimento e teste de prottipos fsicos vo sendo gradualmente
substitudas por prototipagem e testes virtuais. No seguimento desta tendncia, esta tese foi
elaborada por forma a criar um modelo de elementos finitos de um apoio de motor elastomrico
existente para que um conhecimento mais aprofundado acerca do comportamento dinmico destes
componentes pudesse ser obtido e assim formar uma base para o melhoramento futuro deste tipo de
apoios.
Neste trabalho, foi prestada especial ateno modelao precisa de efeitos no-lineares no
comportamento esttico do apoio tendo sido efectuada uma comparao entre ensaios experimentais
e numricos. Com base no modelo de elementos finitos criado, as curvas de resposta em frequncia
do apoio foram determinadas tendo sido observados efeitos multi-dimensionais nas mesmas.
Adicionalmente, foi obtida a resposta no tempo a cargas sbitas de amplitude moderada. Os
resultados globais indicam que o apoio analisado possui um comportamento satisfatrio tanto no
isolamento das vibraes produzidas pelo motor como na eliminao das oscilaes de grande
amplitude a que o motor est sujeito. Para concluir, algumas observaes so efectuadas acerca da
utilizao de borrachas colocadas em paralelo bem como de propriedades dinmicas no-lineares de
materiais como meio para o melhoramento do desempenho dinmico do apoio.
Palavras Chave: vibrao, isolamento, apoio de motor, borracha, mtodo dos elementos finitos.
iii
Abstract
Nowadays, the highly competitive automotive business industry requires manufacturers to pay more
attention to passenger comfort and riding quality. This has forced designers to direct their attention to
the development of high quality engine mounting devices, with traditional physical prototyping and
testing being gradually replaced by virtual prototyping and numerical simulations. Following this trend,
this thesis has been developed in order to create a finite element model of an available automotive
elastomeric engine mount so that a deeper understanding on the dynamics of these components can
be obtained to form a basis for improved design and development of future mounts.
In this work, special attention is given to the accurate modelling of nonlinear effects on the static
behaviour of the mount with a comparison being performed between experimental and numerical
tests. Based on the finite element model created, the mounts frequency response function curves are
determined and multi-dimensional effects in the mounts response are observed. Also, the mounts
time dependent response to step-like moderate amplitude loads is obtained. The overall results
indicate that the modelled mount has an acceptable performance in both isolating engine induced
vibration and suppressing high amplitude engine shake movement. Finally, some remarks are made
about the use of parallel associated rubbers and nonlinear dynamic properties as a mean for
improving the mounts dynamic behaviour.
Table of Contents
AGRADECIMENTOS ............................................................................................................................... I
RESUMO ................................................................................................................................................ III
ABSTRACT ............................................................................................................................................. V
LIST OF TABLES .................................................................................................................................. IX
LIST OF FIGURES ................................................................................................................................. IX
ABBREVIATIONS .................................................................................................................................. XI
LIST OF SYMBOLS ............................................................................................................................... XI
CHAPTER 1 - INTRODUCTION ............................................................................................................. 1
1.1. SCOPE AND OBJECTIVES ................................................................................................................. 1
1.2. AN OVERVIEW OF TYPICAL ELASTOMERIC ENGINE MOUNTS ................................................................ 2
1.3. OUTLINE OF THE DISSERTATION ....................................................................................................... 7
CHAPTER 2 - THEORETICAL CONSIDERATIONS ............................................................................. 9
2.1. RUBBER THEORY ............................................................................................................................ 9
2.1.1. Hyperelasticity ...................................................................................................................... 9
2.1.2. Viscoelasticity ..................................................................................................................... 13
2.2. VIBRATION THEORY ....................................................................................................................... 17
2.2.1. Free vibration ...................................................................................................................... 18
2.2.2. Forced vibrations ................................................................................................................ 20
2.2.3. Vibration of rubber-like materials ........................................................................................ 22
2.3. FINITE ELEMENT METHOD .............................................................................................................. 23
2.3.1. Brief historical background ................................................................................................. 23
2.3.2. General formulation ............................................................................................................ 24
2.3.3. Solving static structural problems ....................................................................................... 27
2.3.4. Solving dynamic structural problems .................................................................................. 28
CHAPTER 3 - STATIC BEHAVIOUR OF THE MOUNT ....................................................................... 31
3.1. EXPERIMENTAL SET-UP ................................................................................................................. 32
3.2. FINITE ELEMENT MODEL ................................................................................................................ 35
3.2.1. Finite element model simplifications ................................................................................... 35
3.2.2. Mesh characteristics ........................................................................................................... 36
3.2.3. Analysis type ....................................................................................................................... 39
3.2.4. Contact................................................................................................................................ 39
3.2.5. Boundary conditions ........................................................................................................... 40
3.3. MATERIAL PARAMETER DETERMINATION ......................................................................................... 41
3.4. STRESS ANALYSIS ......................................................................................................................... 45
3.5. DISCUSSION ................................................................................................................................. 46
CHAPTER 4 - FREQUENCY DOMAIN DYNAMIC BEHAVIOUR ........................................................ 49
4.1. MATERIAL PROPERTIES CONSIDERED ............................................................................................. 50
4.2. ONE-DIMENSIONAL MODEL............................................................................................................. 51
4.3. FINITE ELEMENT MODEL ................................................................................................................ 54
4.3.1. Boundary conditions ........................................................................................................... 54
4.3.2. Idling conditions .................................................................................................................. 56
vii
viii
List of Tables
Tab. 1 Mount materials typical properties .......................................................................................... 35
Tab. 2 Finite element mesh properties ............................................................................................... 38
Tab. 3 Best fitting Mooney-Rivlin parameters .................................................................................... 46
Tab. 4 Natural frequencies for the idling state [Hz] ............................................................................ 56
Tab. 5 Percentage of the applied loads transmitted to the vehicles body in idling conditions .......... 61
Tab. 6 Natural frequencies for the acceleration state [Hz] ................................................................. 62
Tab. 7 Percentage of the applied loads transmitted to the vehicles body in acceleration conditions 65
Tab. 8 SDOF system transmissibility comparison between a filled butyl rubber mount and a parallel
anti-vibration mount working at distinct frequencies ............................................................................. 77
List of Figures
Fig. 1.1 Experimental spectral frequency content of a combustion engine fluctuating torque (from [2])
................................................................................................................................................................. 2
Fig. 1.2 Market of rubber vibration isolators (from [5]) ......................................................................... 4
Fig. 1.3 Typical elastomeric engine mounts (from [6]) ......................................................................... 6
Fig. 1.4 Commercially available passenger vehicle engine mount ...................................................... 7
Fig. 2.1 Frequency dependence of the shear dynamic properties for typical rubber-like materials
(from [14]) .............................................................................................................................................. 16
Fig. 2.2 Typical SDOF system ............................................................................................................ 17
Fig. 2.3 Effect of and in the transmissibility of a SDOF system ................................................... 21
Fig. 2.4 Effect of in the transient response of a SDOF system subjected to a step-like load ......... 22
Fig. 2.5 FEM historical development process (from [16])................................................................... 24
Fig. 3.1 Experimental set-up to obtain the static load-displacement curves ...................................... 33
Fig. 3.2 Experimental load-displacement curve for the vertical direction downward displacement 33
Fig. 3.3 Experimental load-displacement curve for the vertical direction upward displacement..... 34
Fig. 3.4 Experimental load-displacement curves for the transversal direction ................................... 34
Fig. 3.5 Three-dimensional model of the mount ................................................................................. 36
Fig. 3.6 Engine mount finite element model mesh ............................................................................. 37
Fig. 3.7 Nodal and face numbering of the element types used (from [25]) ........................................ 38
Fig. 3.8 Contact penetration allowances (from [25]) .......................................................................... 40
Fig. 3.9 Contact pairs definition .......................................................................................................... 40
Fig. 3.10 Finite element model static tests boundary conditions ....................................................... 41
Fig. 3.11 Load-displacement curves comparison (vertical direction) with deformed shapes.......... 43
Fig. 3.12 Load-displacement curves comparison (transversal direction) with deformed shapes ... 44
Fig. 3.13 Load-displacement curve (horizontal direction) with deformed shapes ........................... 44
Fig. 3.14 Von-Mises stresses for a load of -1500N in the vertical direction (values in MPa) ............. 45
Fig. 3.15 - Von-Mises stresses for a load of 1500N in the vertical direction (values in MPa) ............... 45
Fig. 4.1 Experimental results from an ON/OFF engine acceleration test (from 750 rpm to 3000 rpm)
............................................................................................................................................................... 49
Fig. 4.2 Rubber dynamic properties:
;
(from [14]) ................................................. 51
Fig. 4.3 Schematics of the one-dimensional model of the mount ...................................................... 52
Fig. 4.4 Equivalent SDOF system transmissibility plot for the vertical direction ................................ 52
Fig. 4.5 Equivalent SDOF system phase angle plot for the vertical direction .................................... 53
Fig. 4.6 Relative difference in transmissibility between the hysteretic material and the other rubbers
............................................................................................................................................................... 54
Fig. 4.7 Boundary conditions for the mounts frequency response determination ............................. 55
Fig. 4.8 Idling state vibration modes................................................................................................... 57
Fig. 4.9 Frequency dependence (in idling state) of the reaction forces - constant G' and material 58
Fig. 4.10 Frequency dependence (in idling state) of the reaction moments - constant G' and
material .................................................................................................................................................. 58
Fig. 4.11 Frequency dependence (in idling state) of the reaction forces unfilled NR material........ 59
Fig. 4.12 Frequency dependence (in idling state) of the reaction moments unfilled NR material... 59
Fig. 4.13 Frequency dependence (in idling state) of the reaction forces filled NR material............ 59
ix
Fig. 4.14 Frequency dependence (in idling state) of the reaction moments filled NR material....... 59
Fig. 4.15 Frequency dependence (in idling state) of the reaction forces filled BR material ............ 60
Fig. 4.16 Frequency dependence (in idling state) of the reaction moments filled BR material ....... 60
Fig. 4.17 Estimate load transmissibility in the idling state .................................................................. 61
Fig. 4.18 Frequency dependence (in acceleration state) of the reaction forces - constant G' and
material .................................................................................................................................................. 62
Fig. 4.19 Frequency dependence (in acceleration state) of the reaction moments - constant G' and
material .................................................................................................................................................. 63
Fig. 4.20 Frequency dependence (in acceleration state) of the reaction forces unfilled NR material
............................................................................................................................................................... 63
Fig. 4.21 Frequency dependence (in acceleration state) of the reaction moments unfilled NR
material .................................................................................................................................................. 63
Fig. 4.22 Frequency dependence (in acceleration state) of the reaction forces filled NR material. 63
Fig. 4.23 Frequency dependence (in acceleration state) of the reaction moments filled NR material
............................................................................................................................................................... 64
Fig. 4.24 Frequency dependence (in acceleration state) of the reaction forces filled BR material . 64
Fig. 4.25 Frequency dependence (in acceleration state) of the reaction moments filled BR material
............................................................................................................................................................... 64
Fig. 4.26 Estimate load transmissibility in the acceleration state ....................................................... 65
Fig. 5.1 Boundary conditions for the time domain analysis of the mounts dynamic behaviour......... 68
Fig. 5.2 SLS material model dynamic properties ............................................................................... 69
Fig. 5.3 Transient modal decomposition analysis validation relative horizontal displacement (U1) 70
Fig. 5.4 Transient modal decomposition analysis validation relative vertical displacement (U2) ... 70
Fig. 5.5 Transient response to an instantaneously applied load of -250 N relative horizontal
displacement (U1) ................................................................................................................................. 71
Fig. 5.6 - Transient response to an instantaneously applied load of -250 N relative vertical
displacement (U2) ................................................................................................................................. 71
Fig. 6.1 SDOF system with parallel mounted rubbers ........................................................................ 73
Fig. 6.2 Dynamic properties of a SDOF system with an unfilled natural rubber mounted in parallel to
a filled butyl rubber ................................................................................................................................ 75
Fig. 6.3 Dynamic properties of a SDOF system with an unfilled natural rubber mounted in parallel to
an unfilled synthetic rubber Thiokol RD ............................................................................................... 76
Fig. 6.4 Transmissibility curves of a SDOF system with an unfilled natural rubber mounted in parallel
to an unfilled synthetic rubber Thiokol RD for various cross-sectional area ratios.............................. 76
Fig. 6.5 Nonlinear SDOF system with parallel mounted rubbers (nonlinearity caused by the
introduction of a gap) ............................................................................................................................. 77
Fig. 6.6 Dissipated energy in a loading cycle as a function of the strain amplitude for a nonlinear
SDOF system with parallel mounted rubbers ........................................................................................ 78
Abbreviations
DMA
DOF
Degree Of Freedom
FE
Finite Element
FEA
FRF
List of Symbols
Youngs modulus
Shear modulus
Strain invariant
Bulk modulus
Transmissibility
Element volume
Damping matrix
Stiffness matrix
Mass matrix
xi
Transformation matrix
Displacement array
Strain
Loss factor
Poissons ratio
Damping factor
Stress
Phase angle
Natural frequency
Frequency ratio
xii
Chapter 1
Introduction
1. Introduction
Nowadays, the highly competitive automotive business industry requires manufactures to pay more
attention to passenger comfort and riding quality. Resonant vibration of body panels arising from
unbalanced loads existing in the engine body, intensified by frameless or unitary chassis construction,
has forced designers to direct their attention to the development of high quality engine mounting
devices in order to ensure that improved comfort in riding and silencing shall not be offset by fatiguing
vibration effects. Furthermore, this demand for better performing mounts must not be offset by an
increase in the production costs and/or development cycle time.
As a response to this challenge, traditional physical prototyping and testing (highly time-consuming
and expensive processes) are gradually being replaced by virtual prototyping and simulations [1]. The
main advantages of this latter kind of approach is that a large number of virtual tests can be performed
instead of a few costly tests on physical prototypes, mainly because, once a simulation model has
been developed, the additional cost of further analysis is usually very low compared to building new
physical prototypes. However, this way of developing products has been more or less overlook in the
case of rubber components, partially due to the difficulty associated with modelling the complex
characteristics of rubber, but also due to a limited understanding of the mechanical properties of
rubber-like materials.
In consequence, this thesis has been developed in order to create a finite element (FE) model of an
available automotive engine mount component so that its current static and dynamic behaviour could
be analysed and that, in the future, several materials can be tested in this model and, perhaps,
improvements to the current performance specifications can be achieved.
The aim of this thesis is to provide a deeper understanding of the dynamics of automotive elastomeric
engine mounts to form a basis for improved design and development of future mounts. Computational
modelling and simulation of an available engine mount are performed to gain such an understanding
1
and to evaluate the effectiveness of current solutions available in the market. Special attention is given
to the correct modelling of nonlinear effects on the static behaviour of the mount. The FE analyses in
this thesis have been performed in ABAQUS. The choice of using a commercial FE code makes it
easier to focus on the engineering problem rather that a detailed description of complex FE models. It
also results in methods that can be put directly to use in industry.
Fig. 1.1 Experimental spectral frequency content of a combustion engine fluctuating torque (from [2])
Directly related to these reasons, several motives may be pointed for using flexible mounts for
supporting an automobile engine and its peripheral components (e.g. gearbox). Among them one may
refer:
i.
To prevent the fatigue failure (caused by small dynamic disturbances) of the engine and
gearbox support points which would occur if they were rigidly attached to the chassis or body
structure of the vehicle;
ii.
To reduce the amplitude of engine vibration which is transmitted to the body structure and,
hence, to reduce noise amplification which would occur if the engine vibration were allowed to
be transferred directly to the body structure;
2
iii.
To reduce human discomfort and fatigue by partially isolating the engine vibrations from the
body by means of an elastic media.
To accommodate engine block misalignment and to reduce residual stresses imposed on the
engine block and mounting brackets due to chassis or body distortion;
ii.
To prevent road wheel shocks (when driving over rough terrain) from causing excessive
rebound movement of the engine;
iii.
To prevent large engine-to-body relative movement due to torque reaction forces, particularly
in low gear, which would cause excessive misalignment and strain on components such as
the exhaust pipe and silencer system;
iv.
To restrict engine movement in the fore and aft direction of the vehicle due to the inertia of
the engine acting in opposition to the accelerating and braking forces.
From what was mentioned, one may already realize that the design of engine mounts must be
performed with care in order to ensure that each and every one of the engine mounts functions can
be accomplished successfully within some performance parameters ranges. One of such parameters
is the mounts transmissibility. To isolate the vibration caused by the engine unbalanced disturbances,
i.e. in order to ensure low values of vibration transmissibility, low elastic stiffness and damping are
needed as the forces transmitted to the structure are related to the stiffness and damping of the
mounts (a more detailed discussion is presented in 2.2.2). However, if the mounts elastic stiffness
and damping are too low, the transient response of the engine mount system can become problematic
in the case of shock excitations caused, for example, by sudden vehicle and engine acceleration and
deceleration, braking and riding on uneven roads. Hence, from this point of view, high elastic stiffness
and damping are required to minimize the engine motion and absorb engine shake. From this
discussion is easy to deduce that, in the development phase, there must usually be a compromise
between the fulfilment of some performance requirements.
The oldest engine mount type known is the elastomeric mount, developed mainly because their
characteristics make them very suitable for automotive applications. Among these characteristics one
may mention the fact that they are compact, cost-effective and maintenance free, which make them,
despite recent developments in hydraulic, magneto-rheological and active mounts for example, still the
most used engine mount type in the world and whose market is still growing Fig. 1.2 (more
information about engine mount types can be found in [3]).
Since the first elastomeric mount proposed in 1930 [4] to isolate vehicle structure from engine
vibration, many developments, such as improved adhesive and extended temperature range, have
improved the performance of the elastomeric mounts. However, with few exceptions, commercially
available passive vibration isolation devices have not changed much during the last 30 50 years.
Among the proposed solutions one may refer the following (please refer to Fig. 1.3 for an illustration):
a) Double shear paired sandwich mounts: composed of a pair of rubber blocks bonded
between the jaws of a U shaped steel plate and a flat interleaf plate, this type of shear
mounts provides a large degree of flexibility in the upright direction;
b) Double inclined wedge mounts: when loaded in the vertical plane, the inclined wedge angle
forces the bonded rubber blocks downwards and outwards against the bent-up sides of the V
shaped lower steel plate;
c) Double inclined wedge with longitudinal control mounts: these mounts are designed to be
used when heavy vertical loads and large rotational reactions are to be absorbed. Their
configuration provides a combined shear and compressive strain to the rubber when there is a
relative fore and aft movement between the engine and chassis, in addition to that created by
the vertical loading of the mount;
d) Inclined interleaf rectangular sandwich mounts: this type of mounts is designed to be used
in pairs, i.e. the engine suspension system is composed of at least two of these mounts
placed on each side of the engine. To increase vertical elastic stiffness without compromising
engine roll flexibility, metal spacer interleafs are commonly used;
e) Metaxentric type bush mounts: composed of an inner and outer steel sleeves positioned
eccentrically in the unloaded state and bonded to a rubber core (which possesses a slot to
prevent tensile stresses from arising), this type of bush mounts provides a large amount of
vertical deflection with very little fore and aft movement;
f)
Metacone sleeve mounts: these mounts are formed from male and female conical sleeves
with rubber occupying the space between both surfaces. Depending on the type of
construction, there may exist an upper and lower plate bolted rigidly to the ends of the inner
sleeve which acts as both overload (bump) and rebound stops;
g) Double inclined rectangular sandwich mounts: providing a very good lateral stability
without impairing vertical deflection flexibility and progressive stiffness control, these mounts
4
use a physical principle similar to that of the inclined interleaf rectangular sandwich mounts
without the need for using the same mount type across the engine suspension system;
h) Flanged sleeve bobbin mounts with rebound control: these mounts are composed of a
rubber block moulded partially around an outer flange sleeve and in between this sleeve and
an inner tube. Rebound control can be attained through the reduction or increase of rubber
between the flanged sleeve and the lower plate;
i)
Hydraulic engine mounts: although not a truly elastomeric mount, these mounts make use of
rubbers elastic properties in their working principle. Composed of an upper and lower rubber
spring diaphragm whose space between is filled and sealed with fluid and is divided in two by
a separator plate with small transfer holes, these mounts use the damping created from the
passage of the fluid through the transfer holes as a mean for slowing down the up and down
motion of the engine support arm.
The report that follows presents the work done in order to create a FE model of a particular type of
engine mount used in a current commercially available passenger vehicle. In Fig. 1.4 a picture of this
mount is shown. This mount, as many others, is build upon a metal structure which is attached to the
body of the vehicle, an elastomeric nucleus that acts as a vibration isolator and a metal core that
provides the attachment support of the engine through a bracket type component. Overload and
rebound control is accomplished with two rubber blocks positioned in the lower and upper section of
the mount, respectively. As the vertical load increases and, consequently, as the middle V shaped
rubber block moves towards full distortion, contact between the deformable V shaped block and one
of these rubber stoppers eventually takes place. If the load continues to increase, these rubber blocks
are squeezed to each other increasing significantly the mounts stiffness and, thus, controlling the
overload and eventual rebound behaviour.
a)
d)
b)
c)
f)
h)
e)
g) Double inclined
rectangular sandwich
mount
The motivation and the aims of the research are introduced in this chapter together with the
background concerning the use of general elastomeric engine mounts in automotive applications,
particularly in the vibration isolation of engine disturbances. Finally, the mount studied in this
dissertation is presented.
Chapter 2 deals with the presentation of the theoretical backgrounds required to fully understand the
work developed in this dissertation. Starting with a short introduction about rubber-like materials, the
hyperelastic and viscoelastic material constitutive theories are presented. Then a short theoretical
introduction is made concerning the theory of vibrations. To conclude, the concepts of the FE method
are briefly explained and some particular analyses types of interest are mentioned.
In Chapter 3 the static behaviour of the mount is discussed. First, the experimental tests performed in
order to obtain the mounts static load-displacement curve are presented and its results are shown.
Then, the characteristics of the FE model created to simulate the experimental tests and, therefore,
determine the mounts rubber approximate hyperelastic parameters are mentioned. To conclude, the
numerical load-displacement curves for three loading directions are presented.
7
The frequency dependent dynamic behaviour of the mount is analysed in Chapter 4. First, a onedimensional model of the mount is analysed to determine its frequency response function (FRF) curve.
Next, the FE model created in Chapter 3 is adapted in order to perform a linear steady-state dynamic
analysis of the mount using the dynamic properties of four types of rubbers. The results for two
loading states (engine idling and acceleration) that result in different operating mounts stiffnesses are
then shown and analysed.
Chapter 5 presents the mounts transient dynamic behaviour when is subjected to high amplitude low
frequency loads. Starting with a presentation of the boundary conditions imposed upon the FE model,
a discussion about the validity of the analysis type chosen is then made. Afterwards, the mounts
response to a high amplitude step-like load is computed using four distinct dynamic rubber properties
sets and the results obtained are then analysed.
In Chapter 6 some possible approaches for improving the performance of engine mount components
are discussed. The advantages of parallel placement of low and high damping rubbers are discussed
and the usage of nonlinear material models as a mean to tune the dynamic behaviour of mounts is
introduced.
Finally, Chapter 7 presents the conclusions taken throughout this report and the suggestions for future
work.
Chapter 2
Theoretical considerations
Overview
This chapter deals with the presentation of the theoretical
backgrounds required to fully understand the work developed in
this dissertation. Starting with a short introduction about rubber-like
materials, the hyperelastic and viscoelastic constitutive theories are
presented. Then a short theoretical introduction is made
concerning the theory of vibrations. To conclude, the FE method is
briefly explained and some particular analyses types are
mentioned.
2. Theoretical considerations
2.1. Rubber theory
Every material can be described by a constitutive model. This model is classically a relationship
between stress and strain. Although this approximation is valid for some types of materials, in others
stress depends on more than just strain. This is the case of rubber-like materials, which have a very
complex material behaviour. Aside from the nonlinear elastic behaviour, most engineering rubber-like
materials also exhibit a considerable material damping which result in a hysteretic response to cyclic
loading. This dynamic response is, apart from the strain level, also dependent on the present strain
rate and strain history. For instance, if a harmonic load is applied to the material, this behaviour can be
observed through the dependence on frequency and amplitude respectively, dependence that can
lead to differences of several hundreds of percent in dynamic modulus and damping of typical
engineering rubber-like materials. In addition, rubber-like materials also show other important material
behaviours such as Mullins effect (Mullins 1969), temperature dependence, swelling and ageing,
among others. These latter effects, in addition to the amplitude dependence, were not however
considered in this thesis.
In the following topics a small introductory presentation is made on the general theory behind the
mathematical modelling of such behaviours, namely the hyperelastic and viscoelastic behaviours of
rubber-like materials.
2.1.1. Hyperelasticity
Hyperelasticity is used to describe the behaviour of rubber-like materials undergoing finite
deformations. This behaviour is characterized by a strain energy function which defines the strain
energy stored in the material per unit of reference volume (i.e. the volume in the undeformed
configuration) as a function of the strain at that point in the material.
9
The formulation of a strain energy function may be done by one of two theories: the phenomenological
theory and the statistical mechanical theory. The former is based on purely mathematical reasoning
and has the aim to find a general way of defining the mechanical properties, but not to pursue the
connection between the materials molecular and structural features and its mechanical description.
The latter theory bases itself on molecular or structural concepts when it aims to derive a description
directly, usually through the statistical treatment of elastomeric molecular networks.
In the early phenomenological theories addressing large deformations, the approach that obtained the
greater recognition and played an important role in the development of this field was Mooneys theory
[7] developed in 1940. Mooneys theory is based on the assumptions that the elastomer is
incompressible and isotropic in the undeformed configuration, and that the Hookes law is obeyed in
simple shear. The first assumption describes with a fair amount of precision the reality, as most
elastomers are considered to be nearly incompressible. As to the second assumption, Hookes law
describes rather well the simple shear up to moderate deformations. Based on these assumptions,
and using purely mathematical arguments involving material symmetry, Mooney has derived the
following strain energy function:
1
1
1
+ + 3
12 22 23
= 1 12 + 22 + 23 3 + 2
( 2.1 )
Where 1 and 2 are material constants. The 1 , 2 , 3 refer to the three principal extension ratios1 or
stretches which, when measured from a given configuration together with their directions, can
completely specify the state of strain of the material.
The phenomenological approach to describe the hyperelastic behaviour of rubber-like materials was
further developed by Rivlin [8]. Rivlin argued that the strain energy function for an isotropic material
should be approximated by the power series:
1 3
2 3
3 3
( 2.2 )
+ +=1
Where are a set of material constants and 1 , 2 and 3 are the strain invariants given by:
1 = 12 + 22 + 23
2 = 1 2
+ 2 3
3 = 1 2 3
+ 3 1
2
( 2.3 )
2
( 2.4 )
( 2.5 )
A principal extension ratio is defined as the ratio of the final length to the initial one in the direction of the
corresponding principal strain axis, such that subscripts 1 to 3 denote three mutually orthogonal directions.
10
In order to guarantee, by convention, that the strain energy function has a value of zero in the
unstrained state, the quantities 1 3 , 2 3 and 3 3 are used rather than 1 , 2 and 3 .
Similarly, the parameter 000 is taken as zero in the same state. The third invariant 3 can be
interpreted in physical terms as the square of the ratio between the volumes of a material element in
the deformed and undeformed states. Consequently, it follows that, for incompressible materials
1 2 3 = 1 and 3 = 1. Hence, for these materials:
1
12 22
( 2.6 )
1
1
2 + 2
1 2
( 2.7 )
1 = 12 + 22 +
2 = 12 22 +
1 3
2 3
( 2.8 )
+ =1
If one only considers the first term of the series, the neo-Hookean model is obtained:
= 10 1 3
( 2.9 )
And using two terms one obtains the commonly-known Mooney-Rivlin2 form:
= 10 1 3 + 01 2 3
( 2.10 )
Equation ( 2.9 ) describes the simplest hyperelastic model available and often serves as a prototype
for elastomeric materials in the absence of accurate material data. It also has some theoretical
relevance since the mathematical representation is analogous to that of an ideal gas: the neoHookean strain energy function represents the Helmholtz free energy of a molecular network with
Gaussian chain-length distribution [9].
Rivlins theory played an important role in the development of nonlinear elasticity for describing
rubber-like solids, given that most of the strain energy functions proposed to date are also expressed
in terms of the strain invariants (e.g. [10]).
Please note that ( 2.10 ) is identical to the strain energy function ( 2.1 ) obtain by Mooney if one substitute
10 = 1 and 01 = 2 .
11
Regardless of the value of the number of terms chosen in ( 2.8 ), the initial shear modulus 0 depend
only on the polynomial coefficients of order + = 1:
0 = 2 10 + 01
( 2.11 )
The strain energy functions above are derived with the assumption of incompressible material
behaviour. Although rubber-like material exhibit, in general, nearly incompressible behaviour, the use
of an incompressible formulation puts some problems in terms of its numerical implementation in a
finite element code. Consequently, these models are available in ABAQUS in a modified way to
account for material compressibility. This can be done if the deformation is divided into one isochoric
and one dilatational part, the latter being a purely hydrostatic part. In incompressibility, the relative
volume change is equal to 1, i.e.
1 2 3 = 1
( 2.12 )
For the compressible case, the relative volume change is equal to 1 2 3 . To be able to use the
earlier functions without any changes, the principal stretches are modified [11] such as:
1
= 3
( 2.13 )
1 2 3 = 1
( 2.14 )
Which yields:
1 = 3 1 = 3 3 1
( 2.15 )
2 = 3 2 = 3 3 2
( 2.16 )
3 = 1
( 2.17 )
The remaining task is to add the hydrostatic part to the strain energy. A simple form of a hydrostatic
term can be expressed by the following:
1
2
12
( 2.18 )
Accordingly, ABAQUS uses in its code the following expressions for the neo-Hookean and MooneyRivlins forms respectively:
= 10 1 3 +
1
1
1
= 10 1 3 + 01 2 3 +
1
1
1
( 2.19 )
( 2.20 )
0 =
2
1
( 2.21 )
2.1.2. Viscoelasticity
The classical theory of elasticity mentioned in 2.1.1 states that, for sufficiently small strains, the
stress in an elastic solid is proportional to the instantaneous strain and is independent of the strain
rate. On the opposite end, in a viscous fluid, according to the theory of hydrodynamics, the stress is
proportional to the instantaneous strain rate and is independent of the strain. Viscoelastic materials,
such as rubbers, exhibit both solid and fluid behaviour, with their deformation response being
determined by both current and past stress states, and conversely, the current stress state being
determined by both current and past deformation states. It may be said that viscoelastic materials
have memory; this characteristic constitutes the foundation on which their mathematical modelling is
based [12].
In the limit of infinitesimal strains, the responses of viscoelastic materials to mechanical perturbations
are well described by the theory of linear viscoelasticity, which may be defined by using either
differential or integral constitutive equations. The differential form of the constitutive law for a one
dimensional linear viscoelastic solid is [13]:
=0
=0
( 2.22 )
Where and are constants and and are the stress and strain respectively. The number of time
derivates retained are denoted by and .
13
The simplest models for one-dimensional viscoelastic materials are the Maxwell model, which consists
of a linear spring and a linear dashpot connected in series, and the Kelvin-Voigt model, which consists
of the same elements connected in parallel. In order to better approximate material behaviour over a
frequency range, more terms may be retained in equation ( 2.22 ).
Other way of defining linear viscoelastic stress-strain laws is by using hereditary integrals, i.e. by using
the fact that, if the applied load consists of infinitesimal steps, the response to each one of these steps
is summed to give the total response.
( 2.23 )
Where is the relaxation modulus function. Equations ( 2.22 ) and ( 2.23 ) are equivalent.
Let us now assume that a harmonic shear strain = 0 sin is imposed on a linear viscoelastic
solid, where 0 and are, respectively, the amplitude and frequency of the perturbing strain. By
employing ( 2.22 ), one observes that the shear stress (response) is an angle out of phase with
respect to the harmonic strain, i.e.,
= 0 sin +
= 0 cos sin + 0 sin cos
( 2.24 )
Equation ( 2.24 ) indicates that the stress has one component in phase with the perturbation (0 cos )
and another 90 out of phase (0 sin ) with the perturbation. Moreover, it can be written:
= 0 sin + cos
( 2.25 )
Where
=
0
cos
0
( 2.26 )
0
sin
0
( 2.27 )
From ( 2.26 ) and ( 2.27 ) is possible to define a parameter, , called loss factor which is a measure of
the damping in the material:
= tan =
( 2.28 )
Thus, one can define a complex shear modulus that characterizes the relationship between the
shear stress and the deformation.
= + = 1 +
14
( 2.29 )
The mechanical work performed per cycle and per unit of volume by a material that undergoes
sinusoidal shear deformation = 0 sin is given by:
=
( 2.30 )
=
= 0 +
0 +
( 2.31 )
02
In equation ( 2.31 ) it can be seen that, in the integral containing , the work done during part of
the cycle is recovered during the other part of the cycle, and the integral of the cycle is zero. Thus,
is called the storage relaxation modulus since it is related to the stored energy. The term
containing is associated to the viscous dissipation, and hence this viscoelastic function is called
loss relaxation modulus. Equation ( 2.31 ) indicates also that the dissipated energy is proportional to
both the square of the amplitude of the deformation and the loss relaxation modulus (which is
frequency dependent).
Similar to the general theory of elasticity one can determine viscoelastic properties for type of
deformation using information available from the behaviour in other kinds of deformation. For instance,
one may determine viscoelastic properties in shear deformation using information available from
volumetric and tensile deformation:
=
3
9
( 2.32 )
Where and are the complex uniaxial and bulk moduli respectively.
The dynamic properties of general polymeric damping materials usually vary quite markedly with
frequency and temperature. On the other hand, they usually vary only a little with strain amplitude up
to quite large strains and generally depend only to a minor extent on static preload and on exposure to
long-duration cyclic loading. This dependence with frequency and temperature under steady-state
dynamic loading is directly related to the deformation mechanisms associated with the long chain
nature of polymer molecules. It depends not only on its basic chemical composition, i.e. on the
monomers that make up the material, but also on several other factors such as the materials
molecular weight spectrum, the degree of cross-linking, and the amounts and types of included
plasticizers and fillers, among other factors [13]. Thats why experience has shown that even samples
taken from the same production batch of a given material may exhibit considerably different modulus
15
and loss factor values, making it difficult to define with accuracy the values for the dynamic properties
of rubber-like materials.
The real parts (storage components) of the complex moduli are found experimentally to increase in
value when frequency increases or when temperature decreases. In these experiments it is possible
to notice that, for instance, if one varies the frequency keeping constant the temperature, the storage
moduli varies drastically, starting from small values at low frequencies and progressing to high values
at high frequencies via a range of intermediate frequencies in which the rate of change is greatest. At
low frequencies the material is said to be in its rubber-like state and at high frequencies in its glasslike state. The region in which rapid changes occur is called the rubber-to-glass transition region. The
highest loss factor values occur at or near this frequency.
Depending on the frequency at which the rubber-to-glass transition occurs, one may divide the
behaviour of rubber-like material in two groups [14]: low-damping rubber-like materials and highdamping rubber-like materials. The former has its transition frequency, at room temperature, at very
high frequencies (thousands of Hz) so that through the range of frequencies normally of concern in
vibration problems the loss factor is small and varies slowly with frequency (the behaviour of the
storage moduli is similar). The latter group has its transition frequencies occurring at or near
frequencies that are normally of interest in vibration problems at room temperature. For these
materials the storage moduli increases very rapidly with frequency, while the loss factor assumes large
values. A qualitative plot for the properties associated with these two groups is shown in Fig. 2.1,
where the curves shaded portion indicates the typical range of frequencies of interest in vibration
problems at room temperature.
b)
16
Vibrating systems can be divided into two categories: linear systems, whose motion is described by
linear differential equations and nonlinear systems whose motion is governed by nonlinear differential
equations. For linear systems, many well-developed methods of vibration analysis are known.
Furthermore, in the study of small oscillations, a nonlinear system can be treated as a linear one using
proper linearization techniques. Nevertheless, nonlinearities must be considered when dealing with
large oscillations in highly nonlinear systems.
The essential physical properties of any linearly elastic structural or mechanical system subjected to
an external source of excitation or dynamic loading are its mass, elastic properties (flexibility or
stiffness), and energy-loss mechanism or damping. In the simplest model of a single degree of
freedom (SDOF) system, each of these properties is assumed to be concentrated in a single physical
element - Fig. 2.2.
The entire mass of this system is included in the rigid block which is constrained to a vertical
movement so that it can move only in simple translation; thus, the single displacement coordinate
completely defines its position. The elastic resistance to displacement is provided by the weightless
linear spring of stiffness , while the energy-loss mechanism is represented by the damper . The
external dynamic loading applied to this system is the time-varying force .
17
( 2.33 )
Let us assume the case where there is no force applied to the system3, i.e. = 0. Consequently,
the free-vibration response of the undamped system is obtained by solving the linear second order
differential equation, whose solution is of the form:
=
( 2.34 )
2 + = 0
( 2.35 )
Therefore:
=
( 2.36 )
The constant is the so called natural frequency of the system. Another way of writing ( 2.34 ) is:
= +
( 2.37 )
( 2.38 )
Considering the case where damping exists, the equation of movement, for the free-vibration case is:
+ + = 0
( 2.39 )
Equation ( 2.39 ) is also a second order linear differential equation with constant coefficients, whose
solution takes the form of ( 2.34 ). Substituting ( 2.34 ) into ( 2.39 ):
2 + + = 0
Oscillatory motions taking place with no applied force are called free vibrations.
18
( 2.40 )
1,2 =
1
2 4
2
( 2.41 )
that, depending on the value of one may have three different cases:
1. Critically damped system = :
= +
( 2.42 )
Critical damping represents the limit of periodic motion, hence the displaced body is restored to
equilibrium in the shortest possible time, and without oscillation or overshoot.
2. Overdamped system > 1:
= 1 1 + 2 2
( 2.43 )
Since both values of are negative, the motion is the sum of two exponential decays.
3. Underdamped system < 1:
= 1
1 2
+ 2
1 2
( 2.44 )
The motion of the body is therefore an exponentially decaying harmonic oscillation with frequency
= 1 2 .
The free vibration of an undamped multi-degree of freedom (MDOF) system is described by
M x
+ K x
= 0
( 2.45 )
Where M and K are matrices with being the number of degrees of freedom of the system.
Extending the concept of harmonic motion, let us assume that x
= u . Hence ( 2.45 )
becomes
2 M u = K u
( 2.46 )
2
This is a classical eigenvalue problem with eigenvalues 12 , 22 ,
which are called the roots of
the characteristic equation. The are called the natural frequencies of the system. Corresponding to
, the eigenvector u describes a specific distribution of displacements and is called a mode shape.
The time dependent response of the mode of vibration of the system is given by
= +
19
( 2.47 )
Where and are determined by the initial conditions. The free vibration response of a MDOF
system is given by a linear combination of the response of each vibration mode, i.e.
( 2.48 )
=1
For a MDOF system with viscous damping, its motion can be described by the matrix differential
equation:
M x
+ C x
+ K x
= 0
( 2.49 )
Where C is the damping matrix consisting of the coefficients of dampers. The concepts of free
vibrations and modes of vibration presented previously can be extended here.
( 2.50 )
0
2
20
( 2.51 )
Where
= tan1
( 2.52 )
The complete solution is the sum of the complementary solution to the steady-state solution.
For these systems, one can define a parameter called transmissibility which quantifies the amount of
force which is transmitted to the foundation:
1 + 2 2
1 2 2 + 2
( 2.53 )
( 2.54 )
The effect of and in the transmissibility is shown in Fig. 2.3. It can be seen that for existing
isolation, the frequency ratio must be greater than
Transmissibility
= 2
The response MDOF systems to harmonic excitations can be similarly determined. To this end,
consider
M x
+ C x
+ K x
= F0
( 2.55 )
By complex analysis, the excitation is the imaginary part of the exponential form F0 . Thus the
steady-state response can be written as
= X 0
( 2.56 )
F0
( 2.57 )
Now let us consider the case when, instead of a harmonic load, a step-like load is applied to a SDOF
system defined as:
F=
0 for t = 0
F0 for t > 0
( 2.58 )
0
1
1
sin +
1 2
( 2.59 )
With
= sin1
1 2
( 2.60 )
Fig. 2.4 shows the influence of the damping factor, , in the response of the SDOF system to a steplike load. As it can be seen, for a rapid suppression of the oscillation movement, a high value of is
recommended.
Fig. 2.4 Effect of in the transient response of a SDOF system subjected to a step-like load
Since the engine of an automotive vehicle creates loads (with both low and high frequencies
components) which must be isolated from the vehicles body, it is natural that rubber-like materials
have been used for a long time as a vibration isolating material.
22
Similar to what was discussed earlier about the theory of vibration isolation using springs and
dampers, analogous expressions can be found for rubber-like materials. Detailed information about
the formulations of the expressions for the transmissibility of SDOF systems composed of elastomeric
materials is presented in [14]. The general equation for the transmissibility of this kind of SDOF
systems when subjected to harmonic loading is:
1 + 2
=
1
+ 2
( 2.61 )
= tan1
( 2.62 )
( 2.63 )
Where is a constant4 that is determined solely by the geometry of the rubber-like material.
possesses the dimensions of length: e.g. for a bar loaded at its ends, = 3 where and are the crosssectional area and length respectively (with the assumption that the material is incompressible, resulting in
= 3 ).
23
Structural
analogue
substitution
Hrenikoff 1941
McHenry 1943
Newmark 1949
Argyris 1955
Turner et al.
Direct
continuum 1956
elements
Trial
functions
Piecewise
continous
trial
functions
Variational methots
Rayleigh 1870
Ritz 1909
Weighted residuals
Gauss 1795
Galerkin 1915
Biezeno-Koch 1923
Richardson
1910
Liebman 1918
Finite
Southwell
differences 1940
Courant 1943
Prager-Synge 1947
Modern FEM
Fig. 2.5 FEM historical development process (from [16])
Finite element method as it is know today can be traced back to the work by Courant [17] and Prager
and Synge [18] with the former using an assemblage of triangular elements and the principle of
minimum potential energy to study the Saint Venant torsion problem and the latter leading to the
development of the hypercircle method. The term finite element was mentioned for the first time in a
report by Clough from 1960 [19], where the technique was presented for the special case of plain
stress. In the years since 1960 the FE method has received widespread acceptance in engineering.
Thousands of papers, hundreds of conferences, and many books appeared on the subject since then.
A 1991 bibliography [20] lists nearly 400 finite element books in English and other languages.
+ = 0
( 2.64 )
where is the system kinetic energy, is the work performed by internal and external forces,
represents time, and is the variational operator. In the case of statics ( = 0), Hamiltons Principle
reduces to the Principle of Virtual Work, stated mathematically as:
24
= 0
( 2.65 )
For most mechanical systems of interest, may be expressed in terms of a conservative interior
elastic potential energy , dissipative interior work , and the work associated with externally applied
forces . Thus Hamiltons Principle is stated as
1
0
+ + = 0
( 2.66 )
The kinematics of a mechanical system composed of a single element of volume are described in
terms of the displacement field:
=
( 2.67 )
where is the displacement array at any point in , is an array of discrete nodal displacements.
The transformation matrix describes assumed shape functions for the particular type of FE.
In matrix notation, the strain field within the element volume is related to the assumed displacements
by the differential operator matrix as:
= =
, , ,
( 2.68 )
= =
( 2.69 )
Where is the elasticity matrix containing the appropriate material constants. In the case of mixed
(hybrid) finite element formulations, for which there is an assumed element stress field other than
simply , the situation is more involved (see 2.3.3.1 for more details).
Using the above general expressions, the kinetic and strain energies associated with a FE are:
=
1
2
1
2
( 2.70 )
( 2.71 )
where is the material density matrix, and and are the individual element mass and
stiffness matrices, respectively. The superscript shown as
and
In the case of viscous damping (which is a common yet not necessarily realistic assumption), the
element dissipative work is:
=
25
( 2.72 )
( 2.73 )
( 2.74 )
where and are the systems mass and stiffness matrices. For the case of viscous damping, the
complete system virtual dissipative work is:
=
( 2.75 )
Finally, the work associated with externally applied forces on the complete system is defined as:
=
( 2.76 )
where represents the vector containing the externally applied loads. Substitution of the above
expressions for the complete system energies and virtual work into Hamiltons Principle, followed by
key manipulations, results in the FE system differential equations
+ + =
( 2.77 )
The task of defining a FE model is not yet complete at this point. Constraints and boundary conditions,
as required, must be imposed to fully complete the formulation.
26
( 2.78 )
For a linear problem, the solution of ( 2.78 ) can be found explicitly by inverting the stiffness matrix and
multiply it by the force vector, i.e.:
=
( 2.79 )
However, some engineering problems have characteristics that are incompatible with this linear
formulation. Generally, three types of nonlinearities may exist in a solid mechanics or structural
problem: material nonlinearity due to a nonlinear stress-strain relationship; geometric nonlinearity
arising from significant changes in the geometry of the structure so that the equilibrium equations must
be formulated with respect to the deformed geometry; and boundary nonlinearity usually caused by
contact interactions between surfaces in the model.
Let us consider a simple case when only material nonlinearity is present in the FE model (this a
particular case since generally the three types of nonlinearities mentioned earlier may be present
simultaneously in a FE model). The nonlinear equilibrium equations can be written as ( 2.78 ). In this
case nonlinearity occurs in the stiffness matrix due to the nonlinear stress-strain relationship:
=
( 2.80 )
Since the material properties are not constant, the equilibrium equations can be expressed as:
,
( 2.81 )
To solve this problem the solution step has to be divided into increments where a new stiffness matrix
has to be computed at every increment, and therefore the matrix must be determined at each
current displacement. Several iterative methods have been proposed to address this problem (e.g.
Newton-Raphson method). In [22] a detailed discussion about these methods is present.
27
The main problem arising from the application of a standard displacement formulation to
incompressible or nearly incompressible materials lies in the determination of the mean stress or
pressure which, in the case of isotropic materials, is related to the volumetric part of the strain. To
overcome this problem it is convenient to separate this from the total stress field and treat it as an
independent variable. A detailed presentation about this kind of formulation can be found in [16].
If no damping and forcing terms (free vibration) exist in the dynamic problem, ( 2.77 ) reduces to:
+ = 0
( 2.82 )
( 2.83 )
= 0
( 2.84 )
This is a general linear eigenvalue or characteristic value problem and for non-zero solutions the
determinant of the above coefficient matrix must be zero:
2 = 0
( 2.85 )
number of unknowns in the system), which is sometimes called a normal mode of the system (it can
be proved these eigenvectors are and -orthogonal and linear independent to each other [23]).
The lowest nonzero is called the fundamental vibration frequency.
While the solution of ( 2.85 ) cannot determine the actual values of we can find vectors
proportion of the various terms.
28
Let us consider the full form of ( 2.77 ) for free response conditions:
+ + = 0
( 2.86 )
= exp
( 2.87 )
Substituting
= 0
( 2.88 )
Where and will generally be complex. The real part of the solution represents a decaying
vibration.
If the load applied to the system is periodic, or, in a more general way, it can be expressed as:
= exp
( 2.89 )
where is a complex variable, the general solution can be written in the same way as ( 2.87 ).
Substituting ( 2.87 ) into ( 2.77 ) gives:
2 + +
( 2.90 )
Which is no longer an eigenvalue problem but can be solved by inverting matrix , i.e.
=
( 2.91 )
This is similar to the solution of static problems but, however, has to be determined in terms of
complex quantities.
29
( 2.92 )
Where is an arbitrary function of time. It has been seen that the general solution for the free
response if of the form:
exp
( 2.93 )
=1
forced response let us assume that the solution can be written as a linear combination of modes, i.e.
( 2.94 )
=1
If ( 2.94 ) is substituted into ( 2.92 ) and the result is premultiplied by the complex conjugate
transposed , then the result is simply a set of scalar, independent, equations:
+ + =
( 2.95 )
( 2.96 )
Each scalar equation of ( 2.95 ) can then be solved by elementary procedures independently and the
total vector of response can further be obtained by superposition using ( 2.96 ).
30
Chapter 3
Static behaviour of the mount
Overview
The static behaviour of the mount is discussed in this chapter.
First, the experimental tests performed in order to obtain the
mounts static load-displacement curve are presented and its
results are shown. Then, the characteristics of the FE model
created to simulate the experimental tests and therefore determine
the mounts rubber approximate hyperelastic parameters are
mentioned. To conclude, the numerical load-displacement curves
for three loading directions are presented. The results obtained
made possible the determination of the mounts rubber
approximate hyperelastic parameters.
The key aspect that sets finite analysis of elastomers (FEA) apart from FEA of metal components is
the specification of its material properties. Metals can, generally, be considered as Hookean materials,
i.e. materials with a linear stress-strain relationship over its useable stress-strain range. Values for
Youngs modulus and Poissons ratio for various metals are readily available in the literature and have
generally well-known values. In the particular case when the applied loads are sufficient to cause
yielding, a non-linear analysis of the metal component can generally be made using a bilinear stressstrain curve or a nonlinear strain hardening rule. In the case of elastomers, each formulation is
different, yielding a large spectrum of behaviours, where, generally, there is only a small linear region
in the stress-strain curve. Hence, the specification of nonlinear material properties for elastomers is
more difficult.
31
The most simple hyperelastic constitutive relationship known is the neo-Hookean model that has a first
coefficient equal to one-half of the shear modulus and a second coefficient equal to zero. This material
model exhibits a constant shear modulus and, according to [24], gives good correlation with
experimental data up to 40% strain in uni-axial tension and up to 90% in simple shear. However, this
model is a particular case of a more advanced strain energy function the Mooney-Rivlin one (more
details are present in 2.1.1). Gent [24] also mentions that the Mooney-Rivlin model is currently the
most widely used strain energy function in FEA due to its simplicity and robustness. For the purpose of
this work, these two material models will be tested in the FE model.
The standard approach to determine the behaviour of rubber-like materials its to perform a set of
experimental tests in order to obtain the materials stress-strain curves for a certain group of possible
modes of deformation (uniaxial, equibiaxial, planar, and, if the material is compressible, a volumetric
compression test) and then calibrate the material coefficients in the strain energy functions from this
experimental information using a least-squares-fit procedure, which minimizes the relative error in
stress [25]. Depending on the accuracy requested, one must perform one or more of these tests in
order to fully describe the behaviour of the elastomeric materials when subjected to various loading
conditions.
For the case of the work presented, it wasnt possible do perform such tests due to the inexistence of
samples made out of the same material as the one used in the mount. The alternative, and adopted,
approach was to perform a static test of the engine mount in order to obtain its load-displacement
curve. Afterwards, using the FE model of the mount this experimental test was numerically reproduced
and its resulting load-displacement curve was used to determine the material coefficients that best
described the behaviour of the elastomeric material. In the next paragraphs the work done is
presented.
For the vertical direction two tests were performed to determine the behaviour of the mount when an
upward or downward displacement was applied. In the case of the transversal direction the mount was
tested in only one direction (due to the mounts geometrical symmetry in this load direction), although
32
three experimental curves were determined corresponding to three different rotation angles in order to
discard the possibility of existing some misalignment in the experimental set-up.
a) Vertical test
b) Transversal test
Fig. 3.1 Experimental set-up to obtain the static load-displacement curves
For each test the load resulting from the applied displacement was recorded and the results obtained
are plotted in the load-displacement curves shown in Fig. 3.2 to Fig. 3.4.
Fig. 3.2 Experimental load-displacement curve for the vertical direction downward displacement
33
Fig. 3.3 Experimental load-displacement curve for the vertical direction upward displacement
In the vertical direction, it is easy to see that the load-displacement curves obtained reveal two distinct
regions: for small displacements, the load-displacement curve has a almost linear shape with a slope
of approximately 140 N/mm; when the imposed displacement takes larger values, contact between the
V shaped rubber block and the overload and rebound rubber blocks take place resulting in a sudden
increase in the mount static stiffness.
For the transversal direction the results show that the transversal stiffness is approximately constant
(showing only a minor decrease with the increasing displacement) for the range of displacements
applied and takes an approximate value of 125 N/mm for displacements up to 2 mm, while there is
almost no difference between the results obtained for the different rotations angles tested.
34
Material Property
Steel
Aluminium
Natural Rubber
Youngs modulus,
70 GPa
1 100 MPa6
Poisons ratio,
0,25 0,30
0,33
0,45 0,49
As one can see, the Youngs modulus for steel and aluminium is several orders of magnitude above
the one for natural rubber. In terms of physical sense, one may say that steel and aluminium have a
much higher stiffness than natural rubber. When this information is used in a FE model, one may
proceed with a simplification of the model and assume that the metallic components in the mount can
be modelled as rigid since their stiffness is several orders of magnitude higher than natural rubbers
stiffness. This assumption translates in a much simpler FE model because these metallic components
can be modelled solely by the use of a pair of surfaces with an imposed rigid behaviour.
35
To improve mesh quality over the isolator, this component was divided in three individual regions
defining a core component and two other ones placed in opposite sides of the core respectively as
depicted in Fig. 3.5.
In terms of element shape, ABAQUS/Standard offers three major element shapes to model a threedimensional solid, namely tetrahedral, wedge and hexahedral elements. While tetrahedral and wedge
elements are very useful in creating complex geometry meshes with little or no effort, ABAQUS
documentation [25] discourage their use, especially first-order fully integrated tetrahedral elements, as
they are usually overly stiff and exhibit a slow convergence with mesh refinement, requiring extremely
fine meshes to obtain accurate results. Therefore, where it was possible, hexahedral elements were
used to build the mesh (still requiring a high level of geometry partition as shown in Fig. 3.5a), while in
36
certain regions the complexity of the geometry made impossible the use of this element shape, being
opted the use of tetrahedral shapes.
a) Mount assembly
ABAQUS/Standard finite element library contains a vast set of element types, differing from linear and
quadratic shape functions to special formulations developed to address certain type of problems. Their
use must therefore be carefully chosen as a function of the problem at hand.
One of the difficulties associated with this analysis is the almost incompressible behaviour of the
material. In this kind of problems, fully integrated elements often suffer from a phenomenon called
volumetric locking. This problem consists of spurious pressure stresses arising at the integration
points causing an element to behave too stiffly for deformations that should cause no volume
changes. When suspected, the presence of this phenomenon can be determined by looking at the
pressure stress at the integration points: if it exists, severe and frequent discontinuities in the pressure
37
stresses should be visible across the mesh (initial analyses performed using these elements revealed
the presence of such phenomenon and yielded non-realistic results). To address this problem
ABAQUS/Standard offers a class of elements called hybrid. The fact that, in these cases, a very
small change in displacement produces extremely large changes in pressure over the element is
resolved by treating the pressure stress as an independently interpolated solution variable, coupled to
the displacement solution through the constitutive theory and the compatibility condition (further
information about this formulation can be found in [27]). Therefore, and following the
recommendations made by ABAQUS documentation manuals, in the case of the hexahedral
elements, an 8-node linear interpolation with hybrid formulation element (C3D8H) was used.
As mentioned in 3.2.2.1, the complex geometry of the mount made impossible the use of a
hexahedral element shape in certain regions. Alternatively, a 4-node linear interpolated tetrahedral
element with hybrid formulation (C3D4H) was used in these regions, with compatibility between
different meshed regions being imposed using tie constraints. Either one of the element types used
have 3 translational degrees-of-freedom for each node.
Fig. 3.7 Nodal and face numbering of the element types used (from [25])
Tab. 2 Finite element mesh properties
No. of elements
Part instance
No. of dofs
Tetrahedral
Hexahedral
12128
2790
20211
19039
11910
3870
14592
805*
2520
1260*
4032
Total
39892
53265
38
In this type of analysis, the solution cannot be obtained by solving a single system of linear equations,
as it happens in a linear problem. Instead, the loads are applied in increments, where in each
increment an approximate equilibrium state is searched and the displaced shape of each element
taken into account when formulating each successive load increment. Because the solution is history
dependent, the time increments must be kept small in order to ensure that the strains within each
increment are sufficiently small so that history dependent effects are modelled correctly.
ABAQUS/Standard uses by default Newtons method (see [27]) as a numerical technique for solving
the nonlinear equilibrium equations. The motivation for this choice is primarily the convergence rate
obtained by using this method when compared to the convergence rate of alternative methods
(modified Newton, quasi-Newton, etc.) for the types of nonlinear problems often studied with
ABAQUS.
3.2.4. Contact
One of the characteristics of this FE model is the presence of contact between surfaces of the mount
when the applied loads/displacements become relatively high. Hence, it is important to understand
how the finite element software handles these constraints.
According to ABAQUS documentation [25], ABAQUS/Standard contact algorithm is built around the
Newton/Raphson technique. During the analysis, ABAQUS/Standard examines the state of all contact
interactions at the beginning of each increment and verifies if whether the slave nodes are open or
closed, applies a contact constraint for each closed node and removes constraints from any node
where contact state has changed from closed to open. A slave node is said having an open or closed
state if the clearance, i.e. the distance separating two surfaces that interact with each other, is positive
or zero respectively, being, therefore, the contact constraint applied when the clearance becomes
zero. In the case of the present work, ABAQUS/Standards default option for contact constraints was
used named hard contact. This formulation assumes that there is no limit on the magnitude of
contact pressure that can be transmitted between the surfaces involved.
39
For the definition of contact interactions ABAQUS requires the user to point out which surfaces are
likely to interact with each other (and with itself) during the analysis. The contact surfaces selected
pairs are depicted in Fig. 3.9.
a)
c)
The use of rigid surfaces allows transferring all surface belonging DOFs to a single node (the reference node).
Hence, in the case of the external and internal sleeve rigid surfaces, it was only necessary to specify boundary
conditions on a single node.
40
a)
b)
The choice to use a Mooney-Rivlin hyperelastic material model was made because, while it is still a relatively
simple model, it is the most widely used constitutive relationship for the nonlinear stress analysis of elastomers
[24].
41
The best fit that could be found for each material model is presented in Fig. 3.11. As one can see, the
two simplest models, i.e. the linear elastic and the neo-Hookean ones, cannot describe with sufficient
accuracy the static curve obtained in the experimental tests, namely for a negative vertical load. In
fact, if one tried to adjust the material parameters to fit the negative portion of the load displacement
curve, the positive portion of the numerical curve would fall far away from the experimental one. On
the other hand, the FE model that used a Mooney-Rivlin material model was able to reproduce with
more accuracy the experimental load-displacement curves. As is possible to see in Fig. 3.11 and Fig.
3.12, this model provided a good fit for both experimental loading directions (vertical and transversal).
Although no experimental tests were performed in the horizontal direction, the numerical loaddisplacement curve in this direction, for a material with the same Mooney-Rivlin parameters, is shown
in Fig. 3.13.
42
43
1)
F=-1500 N
2)
F=-1200 N
3)
F=-450 N
4)
F=450 N
5)
F=750 N
Fig. 3.11 Load-displacement curves comparison (vertical direction) with deformed shapes
6)
F=1500 N
1) F=-1000 N
2) F=0 N
3) F=1000 N
Fig. 3.12 Load-displacement curves comparison (transversal direction) with deformed shapes
3
1)
F=-2000 N
2) F=0 N
3) F=2000 N
Fig. 3.13 Load-displacement curve (horizontal direction) with deformed shapes
44
a) Full geometry
b) Cut-out view
Fig. 3.14 Von-Mises stresses for a load of -1500N in the vertical direction (values in MPa)
c) Full geometry
d) Cut-out view
Fig. 3.15 - Von-Mises stresses for a load of 1500N in the vertical direction (values in MPa)
45
3.5. Discussion
Despite the relative simplicity of the Mooney-Rivlin strain energy function, results indicate that, without
any additional experimental data concerning the static behaviour of the rubber at hand, it can be used
as a first choice in the modelling of the hyperelastic behaviour of this kind of natural rubber. Although
Gent [24] mentions that a two-coefficient Mooney-Rivlin model shows good agreement with tensile test
data up to 100% strain but has been found inadequate in describing the compression mode of
deformation, it can be seen that, in this particular application, the use of this model has resulted in a
quite macroscopically satisfactory mounts static behaviour. However, the material parameters found
may only be used as a guideline in describing the static behaviour of this mount and should not be
used to determine the exact characteristics of the rubber used in the mount. In fact, is possible that,
choosing other material model with completely different (and not equivalent) material parameters one
could obtain the same static load-displacement curve. The material parameters found to fit with better
accuracy the static behaviour of the mount using a Mooney-Rivlin material model are, according to (
2.20 ), shown in Tab. 3.
Tab. 3 Best fitting Mooney-Rivlin parameters
1
1
1
[MPa]
[MPa]
[MPa-1]
0.382
0.096
0.01
As is can be seen, the material is almost incompressible since 1 , which represents the inverse of the
initial bulk modulus, is almost zero, i.e. the initial bulk modulus 0 is almost infinity. For reference only,
according to [28] the properties shown in Tab. 3 resemble the ones possessed by rubbers with a
shore A hardness9 of 55.
In terms of the macroscopic behaviour of the mount, results show that the vertical and the transversal
loading directions have similar stiffnesses, namely 140 N/mm and 125 N/mm respectively, whilst the
horizontal loading direction exhibit a much greater stiffness approximately 470 N/mm for small
displacement values. Since the vertical direction is expected to be the one which will be more affected
by the engines running disturbances, the corresponding stiffness value will result in a systems natural
frequency (considering that the engines weight applied on each mount will be 50 kg) that falls
between the range 6 12 Hz recommended by some authors [29]. This shows that the mounts design
took into account not only its static performance but also its dynamic capabilities to absorb engine
induced vibration.
The analysis of the stress distribution in the mount shows that von-Mises equivalent stresses are far
below typical values for the natural rubbers tensile strength (Ashby [26] indicates that natural rubber
9
Rubber hardness is defined as its resistance to indentation. Hardness measurements in rubber are expressed in
Shore A or Shore D units according to the ASTM D2240 test procedures.
46
has a tensile strength about 30 MPa). However most rubber mounts fails in service not due to the
static loads applied but mostly due to the fatigue failure caused by dynamic loading associated with
prolonged unfavourable environment conditions, namely relatively high temperatures and exposure to
chemical agents such as lubrication oils. Hence, during the mounts development stage, mechanical
failure by fatigue should also be analysed. Some authors (e.g. [30]) have been working on the subject,
proposing techniques for fatigue life estimation of rubber mounts, subject that does not fall into the
scope of this thesis.
47
Chapter 4
Frequency domain dynamic behaviour
Overview
In this chapter frequency dependent dynamic behaviour of the
mount is analysed. First, a one-dimensional model of the mount is
analysed to determine its frequency response function curve. Next,
the FE model created in Chapter 3 is adapted in order to perform a
linear steady-state dynamic analysis of the mount using the
dynamic properties of four types of rubbers. The results for two
loading states (engine idling and acceleration) that result in
different operating stiffnesses are then shown and analysed.
Fig. 4.1 Experimental results from an ON/OFF engine acceleration test (from 750 rpm to 3000 rpm)
10
These tests and the consequent results did not make part of the scope of this thesis and were performed by an
automotive constructor.
49
The results indicate that, at an engine idling speed of 750 rpm, these small disturbances could be
idealized as:
2
sin
( 4.1 )
=1
with
= 1
1 = 25
2 = 50
( 4.2 )
If, for instance, the engine was subjected to an acceleration until 3000 rpm the results shown that the
disturbances would not be the same as ( 4.3 ). Instead
3
= 0 +
sin
( 4.4 )
=1
with
0 = 1200
= 1
1 = 100
2 = 200
2 = 250
( 4.5 )
I.e. besides the small engine disturbances loads, the mount was subjected to a static load cause by
the changes in torque that result from vehicle acceleration.
From this data is possible to assume that the small vibratory disturbances at which the mount will be
subjected in normal operating conditions (discarding the engines start-up and shut-down phases)
have a frequency not lower than 25 Hz (or approximately 157 rad/s). This information will be of great
importance in the analysis of the mounts frequency response curves presented in this chapter.
50
rubber usually vary with frequency making it difficult to use the same approach. Alternatively, the
approach taken was to use material properties available in the bibliography for several rubbers and
see which one results in a dynamic behaviour of the mount similar to the one observed when one uses
the hysteretic damping model with a loss factor of approximately 0.1 in the FE model.
As mentioned in 2.1.2.2 rubber-like materials can be divided in two categories according to their
dynamic properties: low-damping and high-damping rubbers. For this analysis, rubbers from the two
categories were picked, namely an unfilled natural rubber and a natural rubber filled with 50 parts by
weight of HAF carbon black13 per 100 parts of rubber (from the low-damping category) and a butyl
rubber filled with 40 parts by weight of MPC carbon black per 100 parts of rubber (from the highdamping category. Dynamic properties for these rubbers (for a temperature of 20 C) are shown in Fig.
4.2.
a)
c)
b)
d)
13
Carbon black has a major influence on the dynamic properties of compounded rubber since it is a source of
hysteresis or damping. According to [35], the amount of damping increases with the surface area of the carbon
black and the level used in the compound.
51
1 + 2
=
1
+ 2
( 4.6 )
= tan1
( 4.7 )
Applying this type of dynamic analysis to the engine mount studied in this report, one sees that,
considering the one-dimensional direction to be the one corresponding to the vertical direction of the
physical mount, the equivalent SDOF system must have a static stiffness = 140 / ,
resulting in:
=
( 4.8 )
In ( 4.8 ), the term 0 accounts for the changes with frequency in the dynamic storage modulus.
F(t)
X(t)
K*()=K()[1+i()]
Considering equations ( 4.6 ) and ( 4.7 ) and using the rubber properties shown in 4.1, for each
rubber the plot of the transmissibility as a function of the frequency of the applied load is made in Fig.
4.4.
Fig. 4.4 Equivalent SDOF system transmissibility plot for the vertical direction
52
Fig. 4.5 Equivalent SDOF system phase angle plot for the vertical direction
As expected, similar to what was mentioned in 2.2.2, the transmissibility plot for a SDOF system
composed of elastomeric materials has three distinct regions: for low frequencies, all of the applied
force magnitude is transmitted to the foundation with the same phase angle; at frequencies in the
neighbourhood of the systems natural frequency, the transmitted force is higher than the exiting force,
whereas the phase angle changes more quickly than anywhere else; and at higher frequencies the
transmitted force has essentially a phase which is opposite to the one of the applied force, while the
magnitude is smaller than the one of the applied force (for frequency values much greater than the
systems natural frequency, the transmissibility becomes approximately inversely proportional to the
square of the exciting frequency for most low damping rubbers).
In the vicinity of the systems natural frequency, the benefits of a high damping rubber are evident
since its dynamic properties result in smaller transmissibility values. On the opposite end, an unfilled
natural rubber has the worst performance of the lot. In the high frequency region, the benefits of the
filled butyl rubbers dynamic properties cease to exist since the transmissibility decreases at a much
slower rate than the one observed with an unfilled natural rubber for example.
It is interesting to notice that the decrease in the transmissibility with the square of the exciting
frequency observed in the low damping rubbers is not solely due to low values of damping (loss
factor). In fact, the major contributor to a slow decrease rate when using filled butyl rubbers is not the
associated high values of damping but instead the relatively high increase of the dynamic properties
(storage modulus and loss factor) with frequency observed in these kinds of rubbers. Snowdon in [14]
discusses this behaviour and, based on this, mentions some guidelines in designing antivibration
mounts: first it should possess a relatively high loss factor and then it should possess storage modulus
and damping factor that remain constant or increase only slowly with frequency.
53
Fig. 4.6 Relative difference in transmissibility between the hysteretic material and the other rubbers
Fig. 4.6 shows the plot of the relative difference in transmissibility between the hysteretic material
model a loss factor of 0.1 and the other rubbers. This plot shows that the filled natural rubber yields a
dynamic response which resembles the most the one yielded by the simple material model mentioned.
Hence it is reasonable to assume that the rubber used in this mount should have dynamic properties
that resemble the ones possessed by a natural rubber filled with 50 parts by weight of HAF carbon
black per 100 parts of rubber.
former was simulated introducing a concentrated mass14 of 50 kg in the internal rigid surfaces
reference point. Since this mass causes the mount to suffer a displacement, it was necessary to
perform a static non-linear analysis to calculate the displaced shape of the mount and therefore
update its stiffness prior to the frequency domain analysis. The small disturbances loads were
simulated using a vertical sinusoidal time varying concentrated load (with an amplitude of 1 N) in the
internal rigid surfaces reference point whose frequency was changing throughout the analysis and the
resulting reactions in the external rigid surfaces reference point were measured (see Fig. 4.7 for a
boundary conditions and measured reactions diagram).
a)
b)
14
55
Before analysing the frequency response of the mount it is important to determine its free vibration
behaviour. Vibration is said to be free if the loads applied to a structure are either zero or constant.
Furthermore, constant loads affect natural frequencies only if the loads create significant prestress,
hence modifying the structures stiffness or if they are associated with a certain mass. Both of these
situations existed in this model: the load originated by the point mass that simulated the engines
weight affected the mounts configuration and eventually its stiffness and, hence, had to be considered
in the free vibration analyses of the mount.
A structure vibrating at only one of its natural frequencies does it with a characteristic pattern of
amplitude distribution called a normal mode of vibration. Usually, unless there is shock loading, only
the modes of lowest frequency are important in the structural response. Accordingly, despite the fact
that the model had a great number of DOFs, and consequently modes of vibration, in the present
analysis only the six lowest frequencies and modes were requested.
Tab. 4 Natural frequencies for the idling state [Hz]
Natural Frequency
1st
2nd
3rd
4th
5th
6th
Unfilled NR
6.86
8.28
15.98
239.46
357.64
399.02
Filled NR
7.60
9.06
17.47
262.71
393.23
438.68
Filled BR
10.67
12.41
23.82
361.27
543.94
606.89
Tab. 4 shows that the three lowest natural frequencies obtained for each rubber type are very similar
between rubbers in opposition to what happens for higher frequencies especially in the case when a
filled butyl rubber is used. This is explained by the fact that this rubbers dynamic properties increase
significantly as frequency increases, which results in a higher dynamic stiffness and hence higher
values of the natural frequencies. In Fig. 4.8, the normal modes of vibration corresponding to the
mounts first six natural frequencies are shown.
56
a)
1st mode
b)
2nd mode
c)
3rd mode
d)
4th mode
e)
5th mode
f)
6th mode
Analysing Fig. 4.8 it is possible to see that the three lowest frequency modes correspond to
translations in the coordinate directions. In addition, the three next modes correspond to rotations
about the same axes.
57
The information obtained in this type of analysis will complement the frequency response analysis of
the mount presented in the next paragraph.
To determine the frequency response of the mount, a direct-solution steady-state dynamic analysis
was performed [25]. This kind of analysis provides the steady-state amplitude and phase angle of the
response of the system due to harmonic excitation at a given frequency. This procedure is a
perturbation procedure, where the perturbed solution is obtained by linearization about the current
base state (for the calculation of the base state the structure may exhibit material and geometrical
nonlinear behaviour as well as contact nonlinearities). The fact that it is a linear perturbation
procedure, in this case, does not yield major problems since the loads applied are extremely small and
no nonlinear effects are expected to be significant. Two base states were considered in this analysis:
the first corresponding to the load caused by the engines weight; and the second with an additional
load of -1200 N in the vertical direction, which simulates the conditions of vehicle acceleration. The
frequency of the applied force (of unitary amplitude) was swept over the range 1 300 Hz
(approximately 0.16 - 1885 radians per second). The measured reactions (forces and moments) about
the external rigid surfaces reference point (see Fig. 4.7c for an illustration) are plotted in Fig. 4.9 to
Fig. 4.16.
a) Reaction force
b) Phase angle
Fig. 4.9 Frequency dependence (in idling state) of the reaction forces - constant G' and material
RF1;
RF2;
RF3;
RF2 SDOF
a) Reaction Moment
b) Phase angle
Fig. 4.10 Frequency dependence (in idling state) of the reaction moments - constant G' and material
RM1;
RM2;
RM3
58
a) Reaction force
b) Phase angle
Fig. 4.11 Frequency dependence (in idling state) of the reaction forces unfilled NR material
RF1;
RF2;
RF3;
RF2 SDOF
a) Reaction Moment
b) Phase angle
Fig. 4.12 Frequency dependence (in idling state) of the reaction moments unfilled NR material
RM1;
RM2;
RM3
a) Reaction force
b) Phase angle
Fig. 4.13 Frequency dependence (in idling state) of the reaction forces filled NR material
RF1;
RF2;
RF3;
RF2 SDOF
a) Reaction Moment
b) Phase angle
Fig. 4.14 Frequency dependence (in idling state) of the reaction moments filled NR material
RM1;
RM2;
RM3
59
a) Reaction force
b) Phase angle
Fig. 4.15 Frequency dependence (in idling state) of the reaction forces filled BR material
RF1;
RF2;
RF3;
RF2 SDOF
a) Reaction Moment
b) Phase angle
Fig. 4.16 Frequency dependence (in idling state) of the reaction moments filled BR material
RM1;
RM2;
RM3
From Fig. 4.9 to Fig. 4.16 it can be seen that, opposing to what was mentioned in 4.2, the mount
cannot be considered as a SDOF system since, when one applies a load in the vertical direction, there
are reaction components in all coordinate directions. Furthermore, despite the good agreement
between the reaction in the direction of the applied load and the equivalent SDOF model for low
frequencies, the fact is that for higher frequencies multidimensional effects are visible, corresponding
to the influence of other vibration modes shapes. To be able to assess the effectiveness of the
isolation it is necessary to determine the resulting reaction from all of these components. This is a
difficult concept to visualize since all of the reaction components vary over time with different
amplitudes and phase angles. The resulting reaction can be determined by using the following
expression:
= 11 + 22 + 33
( 4.9 )
Where are the three directions of an orthonormal coordinate system. The norm of can be written
as:
1 sin 1
+ 2 sin 2
+ 3 sin 3
1
2 2
( 4.10 )
If one takes the maximum of ( 4.10 ) it is possible to determine an estimate for the vibration amplitude
of the response and, hence, since the amplitude of the perturbation was unitary, the amplitude of the
60
response corresponds to the transmissibility of the system. A plot of the force transmissibility over the
working frequency range is shown in Fig. 4.17.
It is possible to notice in Fig. 4.17 that there is an additional resonance peak in the transmissibility
when the exciting frequency is about 100 rad/s (approximately 15,9 Hz). Considering that, in the idling
state, the perturbations can be treated as a sum of two harmonic excitations with frequencies of 25
and 50 Hz respectively the mount only transmits to the body structure the load percentages indicated
in Tab. 5.
Tab. 5 Percentage of the applied loads transmitted to the vehicles body in idling conditions
Exciting Frequency
25 Hz
50Hz
2.9%
Rubber type
Unfilled NR
Filled NR
14.1%
16.1%
2.9%
3.3%
Filled BR
23.4%
5.9%
61
Similar to what was mentioned in 4.3.2.1, in Tab. 6 the six lowest natural frequencies considering the
case when, in addition to the engines weight, a static load of -1200 N is applied to the mount are
presented.
Tab. 6 Natural frequencies for the acceleration state [Hz]
Natural Frequency
1st
2nd
3rd
4th
5th
6th
Unfilled NR
6.53
10.88
16.88
252.80
385.65
407.69
Filled NR
7.56
12.09
18.65
285.80
433.52
454.74
Filled BR
11.30
16.93
25.71
409.38
616.23
645.12
Comparing Tab. 6 to Tab. 4 is possible to see that, as expected, the application of a nominal load and
consequent increase in stiffness results, in general, in an increase of the mounts resonance
frequencies. This increase will consequently be seen in the steady-state response of the mount,
analysed in the next paragraph.
Again, considering the base state where a static load of -1200 N is applied to the mount, a harmonic
load with unitary amplitude and a frequency that was swept over the range 1 300 Hz (approximately
0.16 - 1885 radians per second) as applied to the mount. The measured reactions (forces and
moments) about the external rigid surfaces reference point are plotted in Fig. 4.18 to Fig. 4.16.
c) Reaction force
d) Phase angle
Fig. 4.18 Frequency dependence (in acceleration state) of the reaction forces - constant G' and material
RF1;
RF2;
RF3;
RF2 SDOF (idle)
62
c) Reaction Moment
d) Phase angle
Fig. 4.19 Frequency dependence (in acceleration state) of the reaction moments - constant G' and material
RM1;
RM2;
RM3
c) Reaction force
d) Phase angle
Fig. 4.20 Frequency dependence (in acceleration state) of the reaction forces unfilled NR material
RF1;
RF2;
RF3;
RF2 SDOF (idle)
c) Reaction Moment
d) Phase angle
Fig. 4.21 Frequency dependence (in acceleration state) of the reaction moments unfilled NR material
RM1;
RM2;
RM3
c) Reaction force
d) Phase angle
Fig. 4.22 Frequency dependence (in acceleration state) of the reaction forces filled NR material
RF1;
RF2;
RF3;
RF2 SDOF (idle)
63
c) Reaction Moment
d) Phase angle
Fig. 4.23 Frequency dependence (in acceleration state) of the reaction moments filled NR material
RM1;
RM2;
RM3
c) Reaction force
d) Phase angle
Fig. 4.24 Frequency dependence (in acceleration state) of the reaction forces filled BR material
RF1;
RF2;
RF3;
RF2 SDOF (idle)
c) Reaction Moment
d) Phase angle
Fig. 4.25 Frequency dependence (in acceleration state) of the reaction moments filled BR material
RM1;
RM2;
RM3
As expected, the increase in the mounts natural frequencies due to the changes in its stiffness has
reflected itself in the mounts frequency response curves. However, in the frequency range of interest
the reaction forces still possess values lower that unity, indicating that the mount it is still
accomplishing its mission of reducing vibration transmission. An interesting fact to be noticed is the
frequency decrease of the anti-resonance valleys for the RF1 and RF2 reactions which, in opposition
to the effect of an increase of the mounts natural frequencies, helps to reduce the overall systems
transmissibility in the frequency range of interest (determined using ( 4.10 )) as it can be seen in Fig.
4.26 and Tab. 7.
64
Exciting Frequency
Rubber type
Unfilled NR
Filled NR
0.41%
0.39%
100Hz
Filled BR
0.69%
200Hz
1.24%
1.25%
1.47%
3.40%
250Hz
1.25%
2.15%
1.57%
4.04%
4.4. Discussion
This chapter dealt with the steady-state response of the mount when subjected to harmonic loading of
small amplitude. Frequency domain analyses have proven to be very useful in describing the
effectiveness of vibration isolator devices, and, in this case, of an automotive engine mount.
Starting from a simple one-dimensional model, results showed that the static properties of the mount
were carefully chosen since they result in a low natural frequency (between the range 6 12 Hz
recommended by [29]) and, hence, in the excitation frequency of interest, indicated that the mount
should performed according to what is expected of it, i.e. it is capable of isolating the engine induced
disturbances from the vehicles body. Furthermore, the comparison between the results obtained from
several rubber types and the ones yielded by an hysteretic damping model showed that the rubber
used in the mount should be some sort of filled natural rubber with dynamic properties similar to those
showed in 4.1.
The frequency domain analysis conducted on the FE model showed that, despite common onedimensional treatment of these kinds of problems, the mount should be analysed as a threedimensional structure since, even when dynamic loads are applied in just one direction, three65
dimensional effects are evident in the analysis of the mounts transmitted loads. However, the overall
conclusions that resulted from the one-dimensional approach were proved to be valid, with the mount
being able to isolate from the vehicles body the disturbances produced by the engine in either idling
or acceleration states, with transmissibility values that are very reasonable.
66
Chapter 5
Time domain dynamic behaviour
Overview
This chapter presents the mounts transient dynamic behaviour
when is subjected to high amplitude low frequency loads. Starting
with a presentation of the boundary conditions imposed upon the
FE model, a discussion about the validity of the analysis type
chosen is then made. Afterwards, the mounts response to a high
amplitude step-like load is computed using four distinct dynamic
rubber properties sets and the results obtained are then analysed.
In this chapter the known numerical FEA types are presented and the choice used in the transient
analysis of this mount is justified and validated. Subsequently, the response of the mount to
instantaneously applied loads is analysed and a discussion about the optimal mounts material
properties is presented.
Although the fact that most of the loads applied in this manner to the mount have a magnitude high
enough to result in temporary contact between the mounts V shaped rubber core and its overload
67
and rebound rubber stoppers, this constraint could not be treated with the current FE model. It is of
common knowledge that high speed contact between surfaces is very difficult or maybe impossible to
treat in some finite element models. This was the case of the mounts FE model. Earlier analyses
showed that the inclusion of contact interactions was out of reach and could not be performed.
However, for loads with a smaller magnitude which didnt yielded in contact between surfaces several
conclusions could also be made. Hence, to analyse the mounts transient response to instantaneously
applied loads, a load of 250 N applied in the mounts downward vertical direction was considered as
shown in Fig. 5.1. Similar to what was mentioned in 4.3.1 and 4.3.3, the mount was fixed in its outer
rigid sleeve surface and a static nonlinear analysis was performed to determine the initial conditions
caused by the presence of the engines weight.
Fig. 5.1 Boundary conditions for the time domain analysis of the mounts dynamic behaviour
decomposition analysis. This type of analysis has some advantages and disadvantages that should be
mentioned:
i.
It can be performed using the frequency dependent dynamic material data available;
ii.
In this type of analysis most of the computational effort occurs in solving for frequencies and
modes of the system, which, in overall, results in less computational costs when compared to
a direct method making it suitable for large problems;
iii.
Transient modal decomposition analysis available in ABAQUS is only a linear method which,
at first sight, can be a problem considering the nonlinear nature of the mount analysed in this
work;
iv.
It requires the analyst to choose the number of modes to be considered in the analysis, having
in mind the fact that more modes result in more computational expenses.
It is important to consider how many natural frequencies of the structure may be excited by the loading
and how many of these modes may be needed to calculate the desired response. [31] refers that in
structural dynamics problems, such as loading by an earthquake, only the lower frequencies and
modes are significantly excited (in some cases only the ten lowest modes may be important); in wave
propagation problems, such as shock or blast loading, higher modes and frequencies are also excited
(the number of significant modes may be large reaching sometimes two-thirds the number of DOFs).
To assess the number of modes required to obtain a fairly accurate solution and to evaluate the
validity of this type of linear analysis applied to a nonlinear structure, a comparison between solutions
obtained using a direct integration analysis and transient modal decomposition analysis was perform.
To be able to evaluate the same material properties in both analyses, a material modelled using a
standard linear solid model with arbitrary properties was used15. Its dynamic properties are shown in
Fig. 5.2.
a)
Frequency domain
b)
Fig. 5.2 SLS material model dynamic properties
Time domain
To perform this analysis, boundary conditions similar to those presented in 5.1.1 were used, but
instead of a load of -1200 N was applied (please note that, because a transient modal decomposition
15
Dynamic material properties of a standard linear solid model can be converted between frequency and time
domains without error using ABAQUS conversion algorithm.
69
analysis is linear, contact that would exist due to this load magnitude had to be ignored). This load
magnitude was chosen instead of a load of 250 N with the objective to show in a more evident way the
eventual differences that may exist between results obtained for both analyses types. Considering the
assumptions mentioned previously, the measured time dependent displacements in the mount relative
to the displaced shape induced by the engines weight load are presented in Fig. 5.3 and Fig. 5.4.
Fig. 5.3 Transient modal decomposition analysis validation relative horizontal displacement (U1)
Fig. 5.4 Transient modal decomposition analysis validation relative vertical displacement (U2)
As could be expected there are some differences in the final stationary displacements mainly due to
the fact that the mounts stiffness, even without the presence of contact, is not truly constant in this
range of displacements. In fact it can be seen that using a nonlinear direct implicit integration analysis
the stationary displacements are smaller than the ones obtained with a linear modal decomposition
analysis indicating that the mount exhibits some softening in this load range. As a consequence, this
difference also influences the frequency at which the mount oscillates.
70
There is, however, a general good agreement between results from both analyses indicating that it is
possible to analyse with a reasonable degree of accuracy the transient dynamic behaviour of this
mount using a linear modal decomposition analysis.
Fig. 5.5 Transient response to an instantaneously applied load of -250 N relative horizontal displacement (U1)
Fig. 5.6 - Transient response to an instantaneously applied load of -250 N relative vertical displacement (U2)
71
5.2. Discussion
From the analysis of the dynamic results several conclusions can be taken. First it is possible to see a
good agreement between the results obtained for the mount with a filled natural rubber and the ideal
material with a constant storage modulus and loss factor, reinforcing the conclusions taken from the
results in the frequency domain. Next, as expected, is possible to verify that an unfilled natural rubber,
which possessed the best vibration isolation capabilities to low amplitude-high frequency loads, now
has the worst performance of the lot of rubbers analysed, mainly due to its low damping capabilities. In
contrast a filled butyl rubber has the best performance with the engine bounce being suppressed in
about 0.5 seconds. With intermediate performance is the filled natural rubber being able to most of the
engine bouncing movement in approximately 1 second.
Joining these conclusions to the ones obtained in the frequency domain analyses it is possible to
conclude that special care was taken in selecting the appropriate rubber for use in this mount, with its
static and dynamic properties being able to offer a good compromise between vibration isolation of the
engines low amplitude-high frequency disturbances and engine bounce control when the mount is
subjected to high amplitude-low frequency loads arising from sudden vehicle acceleration and/or
driving in rough terrain.
72
Chapter 6
Suggestions for improving the dynamic performance of
anti-vibration rubber mounts
Overview
In this chapter, some possible approaches for improving the
performance of engine mount components are discussed. The
advantages of parallel placement of low and high damping rubbers
are discussed and the usage of nonlinear material models as a
mean to tune the dynamic behaviour of mounts is introduced.
Snowdon in [14] and [32] mentions an approach for overcoming these limitations of rubbers as
vibration isolating materials. He suggests that pairs of high and low damping rubbers should be
mounted in a parallel configuration of suitable cross-sectional areas in a way that the overall storage
modulus increases relatively slowly with frequency while the loss factor still maintains reasonable
values. A schematic diagram of this approach is depicted in Fig. 6.1.
73
According to the system depicted in Fig. 6.1, the parallel rubbers experience the same strain when the
system is subjected to a load (). Hence, the dynamic properties of the equivalent system are readily
derived when the rubbers have uniform cross-sectional areas. Let us assume that 1 and 2 are the
cross-sectional areas of the low and high damping rubbers respectively. If 1 and 2 are the
storage moduli of the same rubbers, the resulting systems storage modulus is given by:
1
2
1 +
1 + 2
1 + 2 2
( 6.1 )
( 6.2 )
1 1 + 2 2
1 + 2
( 6.3 )
Let us consider a test case where an unfilled natural rubber is mounted in parallel with a butyl rubber
filled with 40 parts by weight of MPC carbon black per 100 parts of rubber whose dynamic properties
are shown in Fig. 4.2. According to equations ( 6.2 ) and ( 6.3 ), the equivalent systems dynamic
properties as a function of frequency and the ratio of cross-sectional areas, , are shown in Fig. 6.2.
As it can be seen, as the cross-sectional area of the high damping rubber increases relatively to the
cross-sectional area of the low damping rubber, so does the equivalent systems dynamic properties,
so that, when goes to infinity, the dynamic properties of the equivalent system are equal to those of
the high damping rubber.
However, this combination of rubbers is not the most optimal, since, for low frequency values, both
rubbers have relatively low loss factors and, hence, the equivalent system will always have low
damping values in this frequency range. To take advantage of the parallel placement of rubbers, a
rubber with high damping at low frequencies should be chosen instead. The alternative rubber chosen
was a commercially available unfilled synthetic rubber of brand name Thiokol RD developed by
Thiokol Chemical Corporation whose dynamic properties are available in [14]. This rubber is
characterized for having a considerably high loss factor throughout the usable frequency range of
common structural applications, namely due to the fact that its transition frequency, as it can be seen
in Fig. 6.3b, occurs at about 900 Hz. On the other hand, its storage modulus increases severely with
frequency making it unusable for direct application in the vibration isolation application studied in this
thesis.
74
The association of this high damping rubber with an unfilled natural rubber in parallel configuration
results in the dynamic properties depicted in Fig. 6.3. From the analysis of Fig. 6.3, one may already
see that even for low values of the equivalent loss factor takes relatively large values throughout the
frequency range of interest in the isolation of engine induced disturbances (0 1000 Hz).
Fig. 6.4 shows the transmissibility curves for a SDOF system with static properties equal to those
treated in 4.2 ( = 140 and = 50 ) whose dynamic properties are those depicted
in Fig. 6.3. An analysis of the results obtained reveal that there are no obvious advantages in using a
parallel mount composed of an association of an unfilled natural rubber with an unfilled Thiokol RD
rubber in parallel when compared with the results using a filled butyl rubber (presented in 4.2). In
fact, if one wants to improve the mounts performance near the resonance frequency (hence improving
its response to instantaneously applied loads) by increasing the relative proportion of high damping
rubber in the mount this will lead to poor performance in the isolation of small high frequency engine
induced disturbances.
75
This conclusion may not, however, be generalized to all possible combinations of known rubbers.
Each case should be analysed individually as a function of each rubbers dynamic properties, since,
as mentioned throughout this thesis, the static and dynamic properties of rubbers can be very distinct
between any two sets of rubbers.
Fig. 6.4 Transmissibility curves of a SDOF system with an unfilled natural rubber mounted in parallel to an
unfilled synthetic rubber Thiokol RD for various cross-sectional area ratios
76
Tab. 8 SDOF system transmissibility comparison between a filled butyl rubber mount and a parallel antivibration mount working at distinct frequencies
Dynamic properties
Resonance peak
30Hz
300 Hz
Filled BR16
3.4134
0.1506
0.0031
3.9007
0.1281
0.0026
3.4789
0.1369
0.0030
2.5483
0.1742
0.0048
A schematic drawing of one of such nonlinear models is presented in Fig. 6.5. Consider a pair of low
and high damping rubbers placed in parallel configuration with the latter being shorter that the former,
i.e. leading to the existence of a gap.
Fig. 6.5 Nonlinear SDOF system with parallel mounted rubbers (nonlinearity caused by the introduction of a
gap)
Now assume that a sinusoidal load = 0 sin is applied to the system which then produces a
sinusoidal strain in the rubber, = 0 sin . If the amplitude of this strain is lower than the strain
required to eliminate the existing gap, , the energy dissipated per loading cycle is given, according
to ( 2.31 ), by:
= 1 02
16
The transmissibility curve for a SDOF system with these dynamic properties is shown in Fig. 4.4.
77
( 6.4 )
Where 1 is the loss modulus of the low damping rubber. If, however, the amplitude of the strain is
higher than , the energy dissipated per loading cycle is
2
= 1
+
0
( 6.5 )
Where
is the loss modulus resulting from the association of the low and high damping rubbers in
parallel. A plot of the dissipated energy per loading cycle as a function of the strain amplitude is
depicted in Fig. 6.6. As it can be seen, for low strain amplitudes, increases relatively slowly
with the strain amplitude due to a low value of 1 (associated with the low damping rubber). When the
strain amplitudes become greater, the gap ceases to exist and the mount starts to dissipate energy at
Fig. 6.6 Dissipated energy in a loading cycle as a function of the strain amplitude for a nonlinear SDOF system
with parallel mounted rubbers
To understand the advantages of this model, one need to remember the characteristics of the loads
usually applied to the mount: first, it is required that the mount isolates low amplitude high frequency
loads this is accomplished by this model since, when the mounts strains are relatively low, the
dissipated energy is also low favouring low transmissibility values; second, the mount should prevent
the engine to oscillate severely when high amplitude low frequency loads are applied since the
dissipated energy is very high for high strain values, the engine bounce is therefore suppressed
quickly.
Nevertheless, despite the obvious advantages of this kind of approach, the fact is that the numerical
modelling of these types of nonlinear models is difficult to perform. In the particular case of the finite
element software ABAQUS, only linear viscoelastic models (whose behaviour is only frequency and
not amplitude dependent) are numerically implemented. In order to use such nonlinear models,
ABAQUS have the ability to allow the user to program a subroutine called UMAT which can be used
as part of the material definition, and, hence, to define material models that are not implemented in
ABAQUS native code. However, due to the complexity and difficulty of this kind of programming, the
advantages of using nonlinear dynamic material models were not tested using the FE model
developed in this thesis. These tests are then suggested to be performed in future works.
78
Chapter 7
Conclusions and recommendations
7. Conclusions and recommendations
The main goal of this thesis was to develop a FE model of an available automotive engine mount
component so that its current static and dynamic behaviour could be analysed and that, in the future,
several materials could be tested in this model and, perhaps, improvements to the current
performance specifications can be achieved. In this final chapter some of the most important
conclusions will be summarized so that it is possible to evaluate if the objectives proposed at the
beginning of the work were overcome or not. In addition, some recommendations are made regarding
possible future studies that can be made based on the work developed in this thesis.
7.1. Conclusions
In this thesis, computational modelling and simulation of an available engine mount was performed to
gain an understanding of the dynamics of automotive elastomeric engine mounts and to evaluate the
effectiveness of current market available solutions. Special attention was given to the correct
modelling of nonlinear effects on the static behaviour of the mount. The experimental static analysis of
the mount revealed the presence of two distinct regions in the load-displacement curve: one (with
displacements up to 10 and 6.5 mm in the downward and upward direction respectively) in which the
mount had an almost constant stiffness and another where the existence of contact between surfaces
caused an abrupt increase in the mounts stiffness. The numerical simulation of the experimental tests
revealed that the mounts rubber nonlinear static material behaviour could be modelled fairly
accurately using a simple two parameter Mooney-Rivlin hyperelastic model.
Since the mount is mainly subjected to two kinds of dynamic loads, the mounts dynamic response
was simulated using the FE model created. First, the response to harmonic loads of small amplitude
was determined. A linear steady-state dynamic analysis was performed using dynamic properties for
three types of commercially available rubbers and the results compared to the ones obtained using a
hysteretic damping material model suggested by the automotive constructor. The results indicated that
the rubber used in the mount had dynamic properties similar to those of a particular type of filled
natural rubber (a natural rubber filled with 50 parts by weight of HAF carbon black per 100 parts of
rubber). Furthermore, the results showed that the mount performs quite satisfactory in both engine
idling and accelerating operating conditions with transmissibility values lower that 15% even when the
operating conditions led to an increase in the mounts static stiffness and, hence, a shift in the
transmissibility curve to higher frequencies. An important fact that was verified in this work was the
multi-dimensional nature of the mounts dynamic response (despite the common treatment of these
problems as one-dimensional ones). Nevertheless the analyses proved that the mount is well
79
developed since its task to isolate, from the vehicles body, engine induced vibrations is successfully
performed.
The other dynamic loads applied to the mount belong to the high amplitude low frequency type. The
analysis of the mounts response to these kinds of loads required the computation of the mounts time
dependent transient response. Problems with the specification of the rubbers dynamic properties in
the time domain arose making impossible the use of a nonlinear direct integration time domain
dynamic analysis. The alternative was to use a linear mode decomposition analysis which proved to
yield acceptable results for moderate amplitude loads. The results indicated again a good agreement
between the use of the dynamic properties of a filled natural rubber and those of a hysteretic damping
model proposed by the automotive constructor. Furthermore, the time dependent response using this
latter model proved to be acceptable, with the mount being able to suppress in about 1 second the
engine shake caused by a step-like load of 250 N.
The analysis of the mounts response to these two kinds of loads confirmed the existing knowledge
that the rubber which yields better results in isolating engine vibration from the vehicles structure (a
rubber with a low dynamic stiffness) has the worst performance in suppressing engine shake arising
from high amplitude transient loads and vice-versa.
Finally, the approach of using rubbers placed in parallel as a mean to improve the mounts
performance was tested. The results indicate that this approach should be considered in a case by
case analysis since the conclusions taken for a pair of rubbers cannot be extrapolated for every
possible rubber combination.
7.2. Recommendations
As mentioned, the work developed in this thesis was performed using material properties available in
some references. As future work it would be interesting to evaluate the accuracy of these properties
through experimental testing of various rubber samples. Some static tests as well as dynamic
mechanical analyses (DMA) can be performed so that the difficulties associated with the specification
of material properties (specially the dynamic ones) can be suppressed.
Again, the main objective of this thesis was to develop a FE model of an engine mount so that, in the
future, several material models could be tested. Hence, in the future the FE model created can be
used, for instance, to assess the effectiveness of the nonlinear material models mentioned in 6.2
and/or other models developed to simulate the behaviour of cellular/composite materials.
Finally, changes can be made to the mounts geometry so that its performance can be improved.
These changes can be achieved, for example, using a geometrical optimization algorithm associated
to the FE model created.
80
Bibliography
[1]
[2]
C. Olsson, "Active automotive engine vibration isolation using feedback control," Journal of
Sound and Vibration, vol. 294, pp. 162-176, 2006.
[3]
[4]
H. C. Lord, "Vibration Dampening Mounting," Patent US Patent 1,778,503, Oct. 14, 1930.
[5]
M.
Hibi.
(2006,
Oct.)
Rubber
vibration
isolator
for
automobile.
[Online].
http://isotc.iso.org/livelink/livelink/fetch/2000/2122/138117/860973/3831013/6030200/Seminar_1
_Rubber_vibration_isolator_for_automobile.pdf?nodeid=6032192&vernum=0
[6]
[7]
[8]
R. S. Rivlin, "Large elastic deformations of isotropic materials. IV. Further developments of the
general theory," Philosophical Transactions of the Royal Society of London, vol. A 241, pp. 379397, 1948.
[9]
L. R. TREOLAR, The Physics of Rubber Elasticity, 3rd ed. Oxford University Press, 1975.
[10] A. N. Gent and A. G. Thomas, "Forms for the stored strain energy function for vulcanized rubber,"
Journal of Polymer Science, vol. 28, pp. 625-628, 1958.
[11] R. W. Ogden, Non-linear Elastic Deformations. New York: Dover Publications Inc. , 1997.
[12] R. M. Christensen, Theory of Viscoelasticity, An Introduction. New York: Academic Press, 1982.
[13] H. F. Brinson and L. C. Brinson, Polymer Engineering Science and Viscoelasticity - An
Introduction. Springer, 2008.
[14] J. C. Snowdon, Vibration and Shock in Damped Mechanical Systems. Wiley, 1968.
[15] J. M. Silva and N. M. Maia, Vibraes e Rudo - Teoria. Lisboa: AEIST, 1986.
[16] O. C. Zienkiewicz and R. L. Taylor, The Finite Element Method - Volume 1 - The Basis, 5th ed.
Butterworth-Heinemann, 2000.
[17] R. L. Courant, "Variational Methods for the Solution of Problems of Equilibrium and Vibration,"
Bulletin of the American Mathematical Society, vol. 49, pp. 1-23, 1943.
[18] W. Prager and J. L. Synge, "Approximation in Elasticity Based on the Concept of Function
Space," Q. Appl. Math., vol. 5, pp. 241-269, 1947.
81
[19] R. W. Clough, "The Finite Element Method in Plane Stress Analysis," in Proceedings of 2nd
ASCE Conference on Electronic Computation, Pittsburgh, 1960, pp. 8-9.
[20] A. K. Noor, "Bibliography of Books and Monographs on Finite Element Technology," Appl. Mech.
Rev, vol. 44, no. 8, pp. 307-317, Jun. 1991.
[21] C. Lanczos, The Variational Principles of Mechanics, 4th ed. New York: Dover Publications, Inc.,
1986.
[22] O. C. Zienkiewicz and R. L. Taylor, The Finite Element Method - Volume 2 - Solid Mechanics, 5th
ed. Butterworth-Heinemann, 2000.
[23] R. D. Cook, D. S. Malkus, and M. E. Plesha, Concepts and Applications of Finite Element
Analysis, 3rd ed. 1989: Wiley.
[24] A. N. Gent, Engineering With Rubber - How to Design Rubber Components, 2nd ed. Hanser,
2001.
[25] Hibbit, Karlsson, and Sorensen, ABAQUS Documentation Manuals. Version 6.7, 2007.
[26] M. F. Ashby and D. H. Jones, Engineering Materials, 2nd ed. Butterworth Heinemann, 1996.
[27] O. C. Zienkiewicz and R. L. Taylor, The Finite Element Method for Solid and Structural
Mechanics, 6th ed. Elsevier, 2005.
[28] E. F. Gbel, Rubber Springs Design. Newnes-Butterworths, 1974.
[29] T. K. Garrett, K. Newton, and W. Steeds, The Motor Vehicle, 13th ed. Butterworth Heinemann,
2001.
[30] W. D. Kim, H. J. Lee, J. Y. Kim, and S. K. Koh, "Fatigue life estimation of an engine rubber
mount," International journal of fatigue, vol. 26, no. 5, pp. 553-560, 2004.
[31] R. D. Cook, Finite Element Modeling for Stress Analysis. Wiley, 1995.
[32] J. C. Snowdon, "Reduction of the Response to Vibration of Structures Possessing Finite
Mechanical Impedance - Part II ," University of Michigan - Willow Run Laboratories , Michigan,
Technical Report 2892-6-T, 1960.
[33] E. Riande, R. Diaz-Calleja, M. Prolongo, R. Masegosa, and C. Salom, Polymer Viscoelasticity,
1st ed. Marcel Dekker, 2000.
[34] G. Lechner and H. Naunheimer, Automotive Transmissions. Springer, 1999.
[35] C. M. Harris and A. G. Piersol, Harris' Shock and Vibration Handbook, 5th ed. McGraw-Hill, 2002.
[36] M. R. Roque, "Anlise Terica e Experimental de Elastmeros," MSc. Thesis, Universidade
Tcnica de Lisboa, Portugal, 2007.
[37] P. G. Zavala, M. G. Pinto, R. Pavanello, and J. Vaqueiro, "Experimental and computational
simulation approaches for engine mounting development and certification," SAE technical paper
series 2000-01-3239E, 2000.
82
[38] A. Renault, "Modelling of Pneumatic Engine Mount," MSc. Thesis, Chalmers University of
Technology, Gteborg, 2005.
[39] A. Geisberger, A. Khajepour, and M. L. Golnaraghi, "Nonlinear Modeling and Experimental
Verification of a MDOF Hydraulic Engine Mount,," in Proceedings of the Symposium on Control
of Vibration and Noise - New Millennium, Orlando, 2000.
[40] F. Jansson, "A Study of Active Engine Mounts," MSc. Thesis, Linkpings Universitet, 2003.
[41] C. Olsson, "Disturbance Observer-Based Automotive Engine Vibration Isolation Dealing With
Non-linear Dynamics and Transient Excitation," Uppsala University, 2005.
[42] D. W. Fotsch, "Development of Valid FE Models for Structural Dynamic Design," PhD. Thesis,
Imperial College of Science, 2001.
[43] S. Huang, "Dynamic Analysis of Assembled Structures With Nonlinearity," PhD. Thesis, Imperial
College of Science, 2007.
[44] F. Wahl, G. Schmidt, and L. Forrai, "On the Significance of Antiresonance Frequencies in
Experimental Structural Analysis," Journal of Sound and Vibration, vol. 219, 1999.
[45] C. Woo, W. Kim, and J. Kwon, "A Study on the Material Properties and Fatigue Life Prediction of
Natural Rubber Component," International Journal of Fatigue, vol. 27, no. 3, pp. 263-272, 2005.
[46] A. Chaikittiratana and N. Moonpa, "Modelling of Visco-hyperelastic behaviour of Carbon Black
Filled Rubber - Finite Element Analysis Approach," King Mongkut's University of Technology North Bangkok, 2003.
[47] S. O. Oyadiji, How to Analyse the Static and Dynamic Response of Viscoelastic Components.
NAFEMS, 2004.
[48] M. Sjberg, "On Dynamic Properties of Rubber Isolators," PhD. Thesis, Royal Institute of
Technology , Stockholm, 2002.
[49] W. D. Kim, W. S. Kim, C. Woo, and H. Lee, "Some Considerations on Mechanical Testing
Methods of Rubbery Materials Using Nonlinear Finite Element Analysis," Polymer international,
vol. 53, pp. 850-856, 2004.
83
84
Appendix A
ABAQUS Viscoelastic Model Verification
In order to ensure that the dynamic analyses of the mount were conducted using a well implemented
linear viscoelastic formulation, some validation tests were performed in order to verify the accuracy of
the results provided by ABAQUSs linear viscoelastic theory. To do so, a set of examples were
analysed and the results compared with the ones predicted by classic viscoelastic models.
Since these elements are connected in parallel, they will experience the same strain when the model
is subjected to a stress . For each element (spring and dashpot respectively) the stress-strain
relationship may be written as:
1 =
2 =
=
( A.1 )
( A.2 )
A-1
( A.3 )
where is the frequency, the stress response will take the form:
= 0
( A.4 )
Inserting the previous equations into ( A.2 ), the following takes place:
= = +
0
( A.5 )
Equation ( A.5 ) defines the so called complex stiffness . The storage stiffness (the real part of )
and the loss factor (the quotient between the imaginary and real parts of ) may be written
respectively as:
=
= tan =
( A.6 )
If a force = 0 sin is applied to the system, the general transmissibility equation may be
written as:
=
1 + 2
1 2 2 +
( A.7 )
Where is defined as the ratio of the applied frequency to the natural frequency of the system, i.e.
A-2
( A.8 )
systems storage stiffness and loss factor are depicted in Appendix Fig. A.3.
As it can be seen, the storage stiffness of the Kelvin model is constant over the frequency range while
the loss factor is directly proportional to the frequency. This behaviour yields that, for small
frequencies, the magnitude of the complex stiffness approximately equals the spring constant , while
at high frequencies the dynamic stiffness increases very rapidly, i.e. the material becomes more and
more stiff.
In Appendix Fig. A.4 a plot of the transmissibility curves obtained using either equation ( A.7 ) and an
equivalent SDOF system modelled in ABAQUS is shown. As it can be seen, both curves match each
other.
Appendix Fig. A.4 Transmissibility curves comparison for a test Kelvin-Voigt model
A-3
=
2
( A.9 )
( A.10 )
= 1 2
( A.11 )
sin + cos
<1
( A.12 )
Considering the same conditions of the test case presented in A.1.1, Appendix Fig. A.5 shows the
time dependent displacement when the mass is released from an initial position with = 0 =
0.001 using either equation ( A.12 ) and an equivalent SDOF system modelled in ABAQUS. Again,
the results show an excellent agreement between both curves.
Appendix Fig. A.5 Time dependent mass displacement curves comparison for a test Kelvin-Voigt model
A-4
For the three-element spring and dashpot combination shown in Appendix Fig. A.6, the following
differential equation can be written:
+
1 + 2
= 1 +
2
2
( A.13 )
Based on the previous equation, one may determine the corresponding complex stiffness when a
sinusoidal strain is applied to the model:
= 1 +
1
1 2 + 1
( A.14 )
1 22 + 2 2 1 + 2
22 + 2 2
( A.15 )
22
1 22 + 2 2 1 + 2
A-5
( A.16 )
( A.17 )
( A.18 )
And
=
( A.19 )
The transmissibility associated with the application of a sinusoidal load = 0 sin is therefore
defined as:
=
2 + 2
2 + 2 2 2 + 2
+ 2
1 + 2
1 + 2
2
2
+ 2
( A.20 )
Let us consider a test case with 1 = 2.5 105 , 2 = 2.5 105 , = 10000 and
= 100 . The systems storage stiffness and loss factor are depicted in Appendix Fig. A.3.
A-6
Appendix Fig. A.8 Dynamic properties of a test standard linear solid model
As it can be seen, this model provides a qualitative correct behaviour of the storage modulus and loss
factor of rubber-like materials. Its storage modulus increases with increasing frequency and the loss
factor reaches a maximum where the increase in the dynamic modulus has a maximum.
Considering a load = 0 sin to be applied to the system depicted in Appendix Fig. A.7, the
transmissibility curves obtained using either equation ( A.20 ) and an equivalent system modelled in
ABAQUS are shown in Appendix Fig. A.9. Again, there is a good agreement between the results
obtained with each approach.
Appendix Fig. A.9 Transmissibility curves comparison for a test standard linear solid model
The results obtained in this validation stage indicate that the viscoelastic formulation implemented in
ABAQUS code can be used with confidence in the dynamic analyses performed in this thesis.
A-7
A-8
Appendix B
Technical drawings
B-1
B-2
!
"
%
&'
#
$
'
!
(