3230 PDF

MICROMECHANICAL MATERIAL MODELS FOR POLYMER
COMPOSITES THROUGH ADVANCED NUMERICAL

SIMULATION TECHNIQUES

Von der Fakultt fr Maschinenwesen der Rheinisch-Westflischen Technischen
Hochschule Aachen zur Erlangung des akademischen Grades eines Doktors
der Ingenieurwissenschaften genehmigte Dissertation

vorgelegt von
Ghayath Al Kassem

Berichter:

Univ.-Prof. Dr.-Ing. Dieter Weichert
Univ.-Prof. Dr.-Ing. Christoph Broeckmann

Tag der mndlichen Prfung: 12.11.2009

Diese Dissertation ist auf den Internetseiten der Hochschulbibliothek online verfgbar.

2005 - 2009

Kurzfassung

Anstatt intensive experimentelle Prfttigkeiten zur Materialcharakterisier-
ung durchzufhren, knnen durch den Einsatz moderner numerischer
Simulationsmethoden das Verhalten von neuen mehrphasigen Werkstoffen auf
Basis der Mikrostruktur vorausberechnet werden. Eine adquate Abbildung der
werkstofflichen Zusammensetzung ist hierbei notwendig, um reale Phnomene auf
mikroskopischer Ebene zu beschreiben, die nicht mit Hilfe von analytischen
Modellen abgebildet werden knnen. Durch die beeindruckende Entwicklung der
zur Verfgung stehenden Computerleistung in den letzten Jahren, sind die
Mglichkeiten und Anwendungsgebiete der numerischen Simulation in Bezug auf
heterogene Werkstoff-Mikrostrukturen signifikant erweitert worden. In der
vorliegenden Arbeit wird nun eine Methode vorgestellt, Mikrostrukturen mit Hilfe
der Finiten Elemente Methode abzubilden und akkurat zu beschreiben. Dies
beinhaltet sowohl eine Beschreibung der numerischen Homogenisierung, als auch
die Anwendung der Methode der reprsentativen Volumenelemente (RVE).

Die makroskopischen Eigenschaften von Kunststoff - Verbundwerkstoffen,
gefllt mit kugelfrmigen Partikeln und mit Glasfasern, werden durch den Einsatz
von RVE und numerischer Simulation berechnet. Ein Schwerpunkt bildete dabei
die korrekte Anwendung von geeigneten 3D-Randbedingungen in der numerischen
Simulation. Die Ergebnisse werden mit existierenden empirischen und semi-
analytischen Modellen, wie Mori-Tanaka und Double Inclusion, verglichen und
diskutiert. Experimente werden an partikel-gefllten Polymer-Komposites und an
der ungefllten Matrix durchgefhrt, um die bentigten Materialeigenschaften fr
die Simulation zu generieren und die Ergebnisse zu validieren. 3D-periodische und
homogene Randbedingungen werden umfassend untersucht, weiter entwickelt und
auf die erstellten RVEs angewendet. Ein neuer Ansatz fr die Anwendung von
periodischen Randbedingungen wird vorgestellt. Verschiedene numerische
Homogenisierungsmethoden werden untersucht; isotrop mit partikel-gefllten

Polymer-Komposites und transvers isotrop/orthotrop mit unidirektionalen und
beliebig orientierten Kurzfaserkomposites.

Haupt Softwares und Tools: Abaqus (numerical simulation), DigiMat MF (mean-
field homogenization), DigiMat FE (microstructure generator),
NetgenSinterStrict (microstructure generator), Origin (data assessment),
Python Modules (programmed especially for this work),

Keywords: Micromechanics, FEM material modeling, fiber and particle reinforced
polymers, Representative Volume Element, Numerical Homogenization

ABSTRACT

In order to reduce laboratory and experiments expenses, one would try to
make predictions of a new materials behavior and response by numerical
simulations, with the chief goal being to speed up the trial and error experimental
testing and to be able to simulate real phenomena that occur at the micro level of
the composites that cannot be accurately implemented in the existing analytical
models. The recent dramatic increase in computational power available for
mathematical modeling and simulation raises the possibilities that modern
numerical methods can play a significant role in the analysis of heterogeneous
microstructures. This fact has motivated the work that will be presented in this
work, which focuses on the methodology of building up an appropriate finite
element material model describing the microstructure of the composite. It contains
numerical homogenization practice and theory, as well as micro structural material
modeling by using numerical simulation techniques on representative volume
elements (RVEs).

This work deals with the determination of macroscopic material properties of
polymer composites by meso-mechanical numerical modeling. Focus is laid on the
methodology how to build up appropriate representative volume elements (RVE) to
describe the microstructure of spherical-particles and fibers reinforced composites
and how to apply appropriate 3D boundary conditions. This work includes the
comparison of the micro structural simulated FE-models with existing empirical
and semi analytical formulations like Mori-Tanaka and the interpolative double
inclusion (Lielens Model) that are used extensively in material modeling. Material
characterization experiments are done on a particle reinforced polymer composite
and its unfilled matrix to extract the material properties then compared with
numerical homogenization applied on our micro material models. Various

conclusions and results are discussed for the know how in building the
appropriate or preeminent representative material model based on the
microstructure of the composite. 3D periodic and homogeneous boundary
conditions are comprehensively studied, developed and applied to our RVEs. A new
approach and technique is established for the 3D periodic boundary conditions.
Different cases of numerical homogenization are examined, the isotropic case
assumed for the particle filled composites (spherical inclusions) and the transverse
isotropic/Orthotropic cases assumed for the fully-aligned/General-Orientation
short-fiber reinforced composites (sphero-cylindrical and cylindrical inclusions).

Major Softwares and Tools: Abaqus (numerical simulation), DigiMat MF (mean-
field homogenization), DigiMat FE (microstructure generator),
NetgenSinterStrict (microstructure generator), Origin (data assessment),
Python Modules (programmed especially for this work),

Keywords: Micromechanics, FEM material modeling, fiber and particle reinforced
polymers, Representative Volume Element, Numerical Homogenization

Content
CONTENT ..................................................................................................................... 3
NOTATIONS ................................................................................................................. 6
OPERATORS ................................................................................................................ 6
SYMBOLS ..................................................................................................................... 6
ABBREVIATIONS .......................................................................................................... 7
CHAPTER ................................................................................................................... 8
1. INTRODUCTION .................................................................................................... 8
1.1. MOTIVATION ...................................................................................................................... 9
1.2. THEORETICAL INTRODUCTION ........................................................................................ 11
1.3. MICRO-MACRO CONCEPTUAL MODELING...................................................................... 14
1.4. HISTORICAL THEORETICAL BACKGROUND .................................................................... 15
1.5. SCOPE, FRAMEWORK AND OBJ ECTIVES ......................................................................... 17
CHAPTER I ................................................................................................................ 19
2. HOMOGENIZATION THEORIES AND METHODS .................................................. 19
2.1. CONCEPTUAL BACKGROUND.......................................................................................... 20
2.2. PRELIMINARIES AND BASIC EQUATIONS ........................................................................ 20
2.3. SEMI-ANALYTICAL METHODS .......................................................................................... 24
2.3.1. The Theoretical Bounds of HILL REUSS VOIGT ....................................................... 24
2.3.2. The asymptotic HASHIN SHTRIKMAN bounds ............................................................. 25
2.3.3. The ESHELBY Model ............................................................................................................ 26
2.3.4. The MORI-TANAKA Model .................................................................................................. 27
2.3.5. The Self-Consistent Method ................................................................................................. 29
2.3.6. Interpolative Double Inclusion Model (LIELENS Model) ................................................. 29
2.3.7. The HALPIN-TSAI Model ..................................................................................................... 31
2.4. COMPARISON OF SEMI-ANALYTICAL METHODS ............................................................. 33
2.5. DIRECT NUMERICAL METHODS (RVE) .......................................................................... 37
2.6. AVERAGE STRAIN THEOREM .......................................................................................... 38
2.7. AVERAGE STRESS THEOREM ......................................................................................... 38
2.8. HILLS THEOREM ............................................................................................................. 39

2.9. THE EFFECTIVE MATERIAL LAW ..................................................................................... 39
2.9.1. Small Elastic Deformation .................................................................................................... 39
2.10. MAIN STREAM AVERAGING HOMOGENIZATION ............................................................. 42
2.10.1. Formulation ........................................................................................................................ 43
2.10.1.1. Isotropic Case .................................................................................................................... 44
2.10.1.2. Transverse Isotropic/Orthotropic Case .......................................................................... 45
CHAPTER II ............................................................................................................... 48
3. A COMPREHENSIVE STUDY, 3D BOUNDARY CONDITIONS .............................. 48
3.1. INTRODUCTION ................................................................................................................ 49
3.2. IMPLEMENTED 3D HOMOGENEOUS BOUNDARY CONDITIONS FOR RVE ..................... 50
3.2.1. Theory ..................................................................................................................................... 50
3.2.2. Development and Construction ........................................................................................... 51
3.2.3. Formulation ............................................................................................................................. 53
3.3. IMPLEMENTED 3D PERIODIC BOUNDARY CONDITIONS FOR RVE ............................... 55
3.3.1. Theory ..................................................................................................................................... 56
3.3.2. Development and Construction ........................................................................................... 59
3.3.3. Formulation ............................................................................................................................. 62
3.3.4. Examination and Verification................................................................................................ 67
3.3.4.1. Different Windowing Approach ....................................................................................... 74
3.3.4.2. Numerical Accuracy .......................................................................................................... 77
3.4. FEM MODELING AUTOMATED PROCESS ANATOMY .................................................. 78
CHAPTER IV .............................................................................................................. 80
4. RVES WITH SPHERICAL INCLUSIONS ................................................................ 80
4.1. TYPES OF MICROSTRUCTURES (RVE) .......................................................................... 81
4.2. RVE WITH EMBEDDED SPHERICAL INCLUSIONS ........................................................... 82
4.2.1. FE-model Mesh Sensitivity ................................................................................................... 83
4.2.2. Influence of Number of Inclusions ....................................................................................... 86
4.2.2.1. Using Periodic Boundary Conditions ............................................................................. 87
4.2.2.2. Using Homogeneous Boundary Conditions .................................................................. 88
4.2.2.3. Comparison between Homogeneous and Periodic Boundary Conditions ............... 89
4.2.3. FE-modeling with Different Volume Fraction ..................................................................... 90
4.2.3.1. Using Periodic Boundary Conditions ............................................................................. 91
4.2.3.2. Using Homogeneous Boundary Conditions .................................................................. 92

4.3. RVES WITH NON-FULLY-EMBEDDED INCLUSIONS ........................................................ 93
4.3.1. Method ..................................................................................................................................... 93
4.3.2. RVEs with Periodic Geometry (edge-trimmed inclusions) ............................................... 96
4.3.3. RVEs with non-Periodic Geometry (edge-trimmed inclusions) ....................................... 97
4.4. FE-MODEL, ANALYTICAL MODELS AND MEASUREMENTS ............................................. 99
CHAPTER V ............................................................................................................. 106
5. RVES FOR FIBER REINFORCED COMPOSITES ................................................ 106
5.1. TYPES OF MICROSTRUCTURES (RVE) ........................................................................ 107
5.2. RVES WITH FULLY ALIGNED FIBER INCLUSIONS (EDGE-TRIMMED) ........................... 108
5.2.1. Influence of Number of Inclusions ..................................................................................... 109
5.2.2. FE-modeling with Different Volume Fraction ................................................................... 114
5.3. FE-MODELING OF A PARTIALLY ORIENTED FIBER REINFORCED COMPOSITE ............ 119
5.3.1. Statistical study .................................................................................................................... 122
5.4. FE-MODELING OF FIBER REINFORCED COMPOSITES WITH A SINGLE AND MULTI LAYER
ORIENTATION .............................................................................................................................. 123
CHAPTER VI ............................................................................................................ 127
6. SUMMARY AND OUTLOOK ............................................................................... 127
6.1. SUMMARY ................................................................................................................... 128
6.1.1. 3D Boundary Conditions ..................................................................................................... 128
6.1.2. FE-Modeling of Particle Reinforced Composites ............................................................ 129
6.1.3. FE-Modeling of Fiber Reinforced Composites ................................................................ 131
6.1.3.1. Fully aligned fiber reinforced composite ...................................................................... 131
6.1.3.2. Partially oriented fiber reinforced composite ............................................................... 132
6.1.3.3. Multi-layer fiber orientation FE-model .......................................................................... 133
6.2. FUTURE-WORKS....................................................................................................... 134
REFERENCES .............................................................................................. 137

Notations

Operators

x A scalar
xi Components of a vector x

xij Components of a second- order tensor x
[ ] x , X Tensor
x Time derivative of x ( dt dx x = )
ij
Kronecker delta; 1 =
ij
if j i = , 0 =
ij
if j i
D
ij
x Components of the deviatoric tensor
ij kk ij
D
ij
x x x
3
1
=
x Absolute value of x
x Averaged value over volume such as,
= d
def
.
1

x Mean value of x
, , d Differential operators
H Single inclusion strain concentration tensor

Ln Napierian logarithm

Symbols

ijkl
C Stiffness tensor components
E Youngs Modulus
n
Unit normal vector

p
n
Unit vector - orientation of the maximal principal stress

t Time
T Temperature
Stress tensor
ij
Components of a stress tensor

eff
Effective stress tensor
eff
ij
Components of the effective stress tensor
Strain tensor
ij
Components of a strain tensor
eff
Effective strain tensor
eff
ij
Components of the effective strain tensor

D
Stored energy threshold
e
Elastic strain energy density
Density
12
Engineering strain,
12 12
2 =

Abbreviations

RVE Representative Volume Element
MT Mori Tanaka Model
FEM Finite Element Method
FE Finite Element
DI Interpolative double inclusion Lielens interpolative Model
DoF Degree of freedom
MSA Main Stream Averaging (Homogenization)
MP Material Properties
BC Boundary Conditions
PBC Periodic Boundary Conditions
HBC Homogeneous Boundary Conditions
SUBC Static uniform boundary conditions
KUBC Kinematic uniform boundary conditions
RP Reference Point

CHAPTER

1. Introduction

Max IS
mises
I =111.3 MPa
Max IS
22
I =143.7 MPa
Max IS
mises
I =34.1 MPa
Max IS
22
I =45.7 MPa
Max IS
mises
I =110.1 MPa
Max IS
22
I =141.5 MPa

Max IS
mises
I =111.3 MPa
Max IS
22
I =143.7 MPa
Max IS
mises
I =34.1 MPa
Max IS
22
I =45.7 MPa
Max IS
mises
I =110.1 MPa
Max IS
22
I =141.5 MPa

In this chapter, the main concept of this work is introduced. It describes the
efficiency and benefits in understanding the mechanical behavior of composite
materials while using a modern engineering approach in simulating the
composite at the micro-structural level

MICROMECHANICAL MATERIAL MODELS FOR POLYMER COMPOSITES THROUGH ADVANCED NUMERICAL SIMULATION TECHNIQUES
CHAPTER - Introduction - 9 -
1.1. Motivation

Contemporary engineering tools and applications have been significantly improved
by the use of composite materials specially designed to provide the preferred
mechanical behavior. Some desired properties for instance are light weight, high
stiffness or high flexibility, good thermal and mechanical durability, high yield
strength under static or dynamic loading and good surface hardness. Usually
homogeneous materials satisfy only some of the desired properties. That is why
more and more interest in composite materials exists in engineering, which
combine the specific properties of its constituents in a well application-oriented
approach.

As illustrated in the following example, some concrete structures used in traffic
systems need to be protected from penetrating water and frost. Usually this is done
by a polymer coating. The polymer has the desired property of great flexibility under
mechanical and thermal loads. It can sustain large deformations under tension and
compression without initiating cracks. However, for traffic systems pure polymer
coatings would have a much too weak surface hardness and stiffness, and for these
purposes its great flexibility under shear loads is a disadvantage, too. For this
reason such polymer coatings in general are stiffened with sand. The sand
increases the hardness and resistance to shear loads, while the necessary flexibility
under thermal loads is kept. Other widely used composite materials are e.g.
concrete, fiber reinforced materials, ceramics and metal composites used in high
tech applications in industry. A drawback to the use of microheterogeneous
materials could be that inhomogeneities within the microstructure cause local
stress concentration, which is often responsible for inelastic material behavior,
damage and debonding of the inclusions from the matrix material. It is essential to
know about these phenomena and to evaluate their influence on the macroscopic
behavior correctly.

What is common to all composite materials is that their macroscopic properties are
strongly influenced and determined by the properties of the microconstituents and
phenomena on the microscale. The description of the microstructural phenomena
leads to a better understanding of the macroscopic behavior. However, most often
the exact microstructure is not known, so in general some statistical assumption
has to be made. The macroscopic properties are determined by a homogenization
process which yields the effective stresses and strains acting on the effective,
homogenized sample of material. This sample of material is often called statistically
representative volume element (RVE).

The goal of the homogenization process is to provide data which can be used to find
a material model for the effective material, and to identify the parameters
introduced in this material model. The effective material is supposed to represent
all macroscopic properties of the microheterogeneous material. In general, one
cannot assume the effective material model to be of the same type as the model
used for the microconstituents, which significantly complicates the search for an
effective material model. Here, an exception is linear elastic material at small
strains, since the superposition principle holds for this material.

Until some years ago, homogenization and the determination of effective material
parameters could only be done by either performing experiments or tests with the
existing material sample or by applying semi-analytical methods making rather
strong assumptions on the mechanical field variables or on the microstructure of
the material. Quite often, those semi-analytical methods do not lead to sufficiently
accurate results. Especially for microconstituents with extreme properties like near
incompressibility, the determination of effective material parameters with the
commonly used semi-analytical methods leads to considerable deviation in results
from reality.

Recently it is commonly accepted that numerical simulations of the microstructural
behavior in 3D are necessary to get more accurate a priori results for the effective
properties of the material. These numerical simulations can significantly reduce the
number of time consuming and expensive experiments with laboriously
manufactured material samples. This clearly improves the development and design
of new materials for modern engineering applications.
1.2. Theoretical Introduction

One of the foremost progresses in contemporary structural components is the
enhancement done on the materials to obtain the optimum behavior relevant to its
application. This is done through the exploitation of the materials microstructure.
Composite materials have their macroscopic characteristics based on the mixture of
two or more pure components like particles, platelets or fibers suspended in a
binding matrix as it is shown in Fig. 1.2.1. This mixing is used in many materials
like metal, concrete,..., polymer matrix composites, etc. In the construction of
composite materials, the basic philosophy is to select material combinations to
produce desired cumulative responses. For example, in aeronautic engineering
applications the basic choice is a harder particulate phase that acts as a stiffening
factor that adds to the metal or polymer matrix enhanced properties against
abrasion and extreme temperature-fluctuation.
Binding matrix
material
Base material
additives
Mixing
propeller
New
material
Macroscopic
structure
Binding matrix
material
Base material
additives
Mixing
propeller
New
material
Macroscopic
structure

Fig. 1.2.1 Mixing a base matrix material with spherical particles

This suggests to carry out direct numerical simulation of microstructures and to try
to establish a realistic representation of the heterogeneous structure that appends
and contains all the microscale details. Doing so and in order to capture all the
details would lead to an extremely fine spatial discretization with a very large
meshes of finite elements to carry the microscale information. Such problems are
beyond the capacity of the computational power available in the foreseeable future.
That is why the approach of taking a small micro sample that contains a finite part
of inclusions to demonstrate a representative volume element in combination with
proper boundary conditions to represent as close as possible the real composite
material macro-behavior would provide us with a tool to enhance and practicle
understanding of the composites material behavior based on its micro-
constituents.

Effective
properties
determined
E
(1)
E
(2)
:
*
IE =
*
IE
Actual structure
Structure with
effective properties
HOMOGENIZATION
Effective
properties
determined
E
(1)
E
(2)
:
*
IE =
*
IE
Actual structure
Structure with
Effective
properties
determined
E
(1)
E
(2)
:
*
IE =
*
IE
Actual structure
Structure with
HOMOGENIZATION

Fig. 1.2.2. Effective material properties by homogenizing the heterogeneous
microstructure

This macro response is calculated from the micro response through a variety of
methods known as numerical homogenization. Because of these essentials the use
of homogenized material models are of common place in practically all branches of
the physical sciences.

The usual approach is to compare a constitutive relation between averages, relating
volume averaged field variables. Consequently, the so called effective properties can
be used in macroscopic analysis as illustrated in Fig. 1.2.2.

L
(1)
L
(2)
L
(3)
L
(1)
>> L
(2)
>> L
(3)
L
(1)
L
(2)
L
(3)
L
(1)
>> L
(2)
>> L
(3)

Fig. 1.2.3. The size requirements of a representative volume element

The volume averaging takes place over a statistically representative volume element
(RVE). The internal fields to be volumetrically averaged must be computed by
solving a series of boundary value problems with test loadings [6]. Such
homogenization processes are referred to as Numerical Homogenization, mean
field theories, theories of effective properties, etc. For details, see Jikov et al. [5]
for mathematical aspects see Aboudi [1], Hashin [2], Mura [3] and Nemat-Nasser
and Hori [4] for more in-depth studies into this subject. For a sample to be
statistically representative it must usually contain a sufficient number of inclusions
and should have a larger size relative to the size of each inclusion as described in
Fig. 1.2.3. The calculations for a RVE are still large, but are much inferior in
comparison with the simulation of the real structure. Historically most classical
analytical or semi analytical methods for estimating the macroscopic response of
such engineering materials have strongly phenomenological basis, and are in reality
non-predictive of material responses that are unidentified a-priori. This is true even
in the linearly elastic, infinitesimal strain range. In plain words such models require
extensive experimental data to tune parameters that have little or no physical
significance. The arguments about this issue have led to the computational
approaches which require relatively simple description on the microscale,
containing parameters that are physically meaningful or realistic. In other words,
the phenomenological aspects of the material modeling are reduced, with the
burden of the work being shifted to high performance computational methods.
Stated clearly, the aim of computational micro-macro mechanics is to develop
relationships between the microstructure and the macroscopic response of a
composite material, using representative models on the microscale that are as
simple as possible and provide an acceptable presentation for the composite
material in investigation [6].

The use of the FEM for the micromechanical analyses of random composites,
which represent most of the real composites, is very expensive from a point of view
of processing time and use of computer memory. In fact, the FEM discretization of a
representative volume element (RVE) with many heterogeneities involves a problem
with a large number of degrees of freedom (the RVE contains the heterogeneities
characterizing the microstructure of the composite). Such problems have been
analysed by Ghosh et al. [74], who have developed a plane finite element model
based on a polygonal Voronoi cell [75]. Inconveniences due to the use of random
distributions of inclusions and defects can be avoided by assuming a periodic
distribution of such heterogeneities. In fact, in this case it is possible to adopt an
RVE containing a small number of heterogeneities and equipped with suitable
periodic boundary conditions.

1.3. Micro-macro Conceptual Modeling

Throughout this work, the case of linear elasticity is considered. In this perspective,
the mechanical properties of micro heterogeneous material are characterized by a
spatially variable elasticity tensor C. Generally, in order to demonstrate the
homogenized effective macroscopic response of such materials, the relation between
averages turns to be

C
eff
= 1.3-1
where
= d
def
.
1

1.3-2
and where
and
are the volume average stress and strain tensor fields

within a statistically representative volume element (RVE) of volume . The
quantity
eff
C , is known as the effective property, and is the elasticity tensor used in
usual structural scale analysis. Similarly, one can describe other effective
quantities such as conductivity or diffusivity, in virtually the same manner, relating
other volumetrically averaged field variables. However, in this work, we describe
exclusively the mechanical linear elastostatic problems. It is emphasized that
effective quantities such as
eff
C are relations between averages.

1.4. Historical Theoretical Background

Composite materials analysis is active for a considerable period of time; estimates
on effective responses have been made under a range of assumptions on the
internal fields within the microstructure. Works dating back at least to Maxwell [9]
(1867), [10] (1873) and Lord Rayleigh [11] (1892) have dealt with determining overall
macroscopic transport phenomena of materials consisting of a matrix reinforced
with spherical particles. As for Voigt [13] (1889), he is usually credited with the
early study of the effective mechanical properties of micro heterogeneous materials.
A complementary contribution was given later by Reuss [12] (1929). Voigt assumed
that the strain field within an aggregate sample of heterogeneous material was
uniform, leading to
C as an expression of the effective property, while Reuss

approximated the stress fields within the aggregate of polycrystalline material as
uniform. If the Reuss field is assumed within the RVE, then an expression for the
effective property is

1
1
C . A fundamental result (Hill [15], 1952) will be discussed

later on in the next chapter.

Within the last decades improved estimates have been pursued. For example, the
dilute-methods family assumes that there is no particle interaction. With this
assumption one requires only the solution to a single ellipsoidal particle in an
unbounded domain. This is the primary use of the elegant Eshelby [14] (1957)
formalism, based on eigen strain concepts, which is used to determine the solution
to the problem of a single inclusion embedded in an infinite matrix of material
under uniform exterior loading. By itself, this result is of little practical interest;
however the solution is relatively compact and easy to use, and thus has been a
basis for the development of many approximation methods based on non-
interacting and weakly interacting (particle) assumptions. According to Wriggers
[60], non-interaction of particulates is an unrealistic assumption for materials with
randomly dispersed particulate microstructure, at even a few percent volume fraction
(Vf).

An effort to improve the Dilute Approximations is through ideas of self consistency
[6]. For example, in the standard Self-Consistent method, the idea is simply that a
particle sees the effective medium instead of the matrix in the calculations. In
other words the matrix material in Eshelby analysis is simply the effective
medium. Unfortunately, the self-consistent method can produce negative effective
bulk and shear responses, for voids, for volume fractions of 50% and higher [60].
For rigid inclusions it produces infinite effective bulk responses for any volume
fraction and infinite effective shear responses above 40%.

Attempts have also been made to improve these approaches. For example the
Generalized Self-Consistent method encases the particles in a shell matrix material
surrounded by the effective medium (see Christensen [18]). However, such methods
also exhibit problems, primarily due to mixing scales of information in a
phenomenological manner, which are critically discussed at length in Hashin [2].
For a relatively recent and thorough analysis of a variety of classical approaches,
such as the ones briefly mentioned here see Torquato [19]-[20]. In addition to ad-
hoc assumptions and estimates on the interaction between microscale constituents,
many classical methods of analysis treat the microstructure as being a regular,
periodic infinite array of identical cells [1].

1.5. Scope, framework and objectives

This work comprises a build-up of a fully automated process for FE-models starting
from micro-structure generation to the calculation of the composite material
effective properties. It investigates, verifies and compares different types of RVEs for
sphere and sphero-cylinder inclusions representing particle and fiber reinforced
composites to existing semi-analytical models like Mori-Tanaka model, see Section
2.3.4, and the interpolative double inclusion or Lielens Model, see Section 2.3.6.
Micro-structures (RVEs) are generated as geometry with the help of the DigiMat
software package then meshed with a python script through Abaqus to obtain an
orphan mesh file. 3D boundary conditions (B.C.) are developed, tested and verified
to meet the intended numerical homogenization requirements. A new approach or
method is used to develop the 3D periodic boundary conditions that enabled us to
use it on any micro-structure independent of its geometry and does not require a
periodic mesh on the boundaries as other methods. It also enables us to control our
structure through a single independent dummy node or element; more details are
discussed in CHAPTER II. A boundary condition module is programmed with
python language to prepare and append to the input orphan mesh file the
appropriate B.C. code. After this stage, a targeted perturbation is applied and the
job is sent to Abaqus solver. A homogenization module is developed to handle the
post processing stage over the outcome of the simulation. The types of numerical
homogenization are prompted for: Isotropic and transverse isotropic/orthotropic
cases described thoroughly in CHAPTER I through Sections (2.9.1) and (2.10). This
module is responsible for calculating the composites effective material properties in
terms of the volume averaged stress and strain tensors of the RVE.

The framework is applied for various types of sphere, sphero-cylindrical and
cylindrical inclusion-filled RVEs. A variety of statistical testing is done to identify
the appropriate criteria for the most reliable representative volume element. A small
experimental part (single data point) is added and compared to FE-models, MT, and
Lielens model [59]. The main objectives can be summarized through the following
points:

Build up of a micromechanical model (representative volume element) for
spherical-particles and fiber reinforced composite materials to be used as
a useful tool to verify and evaluate existing analytical and semi-analytical
material models, and to have a reliable FE-model to be used in numerical
simulation experiments that replaces real experiments
Development of suitable 3D boundary conditions that adhere to all the
special requirements for the intended simulations and numerical
homogenization.
Establishing an automated homogenization process acting as a tool to
determine the effective material properties of an isotropic, transverse
isotropic or orthotropic composite material based on the matrix/filler
properties and their realistic micro-geometry or structure.
Verification and comparison between the FE-models and existing semi-
analytical models like Mori-Tanaka and the Interpolative Double
Inclusion Model (Lielens Model).
A sample comparison between experimental data, semi-analytical models,
and the FE-model for spherical inclusions.

CHAPTER I

2. Homogenization Theories and Methods

M
i
c
r
o
M
e
s
o
Homogenization Methods
M
a
c
r
o
RVE
M
i
c
r
o
M
e
s
o
Homogenization Methods
M
a
c
r
o
RVE

In this chapter, micro-macro scaling or homogenization theory is presented.
The most commonly known homogenization methods are described in details.

CHAPTER I - Homogenization Theories and Methods - 20 -
2.1. Conceptual Background

Dealing with composite materials properties is an early interest in the science of
materials. Homogenisation tools and methods to determine the effective material
properties of composites have been developed and dealt with since late 19
th
century
and until our times. Recently numerical homogenisation approaches based on FE
methods are being developed and enhanced to meet more realistic and precise
results based on the microstructure of the composite.

Starting from probing the inhomogeneous material, for example in linear elastic
materials to measure the Youngs modulus and Poisson ratio, a sample is tested
under tension and compression load tests. The stresses and strains measured are
the averages for the specimen. Assuming that the material is homogeneous and the
results are the effective properties of this material while in reality there is no such a
homogeneous material. Even at the micro scale when two phases are analysed, one
should always assume that each phase is homogenous by itself which is not the
case in reality since even pure material components are inhomogeneous at a certain
scale. So the assumption of homogenisation is always taken when using any type of
material properties predictions, whether using empirical, analytical, semi-analytical
or numerical FE-based methods.

2.2. Preliminaries and Basic Equations

In this section, some basic notations are defined. There are some formulations and
results which are needed later when we discuss the formulation of the Mori-Tanaka
model (section 2.3.4) and the Interpolative Double Inclusion or Lielens Model
(section 2.3.6). Boldface symbols denote tensors, the order of which is indicated by
the context. Einsteins summation convention over repeated indices is used unless
otherwise indicated:
kj ij
k
kj ij
b a b a

=
3
1

The inner product is generally defined as follows,
[ ] [ ] ( ) B A

= = tr B A
def
ij ij
def
:
2.2-1
The symbols 1 and I designate the second- and fourth-order symmetric identity
tensors, respectively,
ij ij
= 1 , and ( )
jk il jl ik ijkl
+ =
2
1
I 2.2-2
Where
ij
is the Kroneckers symbol
1 =
ij
, if j i = and 0 =
ij
if j i
2.2-3
Hooks elasticity operator is designated by C. In the isotropic case, it is given by:
dev vol
I I C 2 3 + = 2.2-4
where and are the elastic bulk and shear moduli, respectively [21].

The notation D C comparing two fourth-order tensors means stiffer than and is
defined as follows [21]:
, : : : : a D a a C a D C
a second order symmetric tensor ) (
ji ij
a a =
2.2-5
In a micro-macro-approach, at each macro-point x , we know the macro-strain ) (x
and need to compute the macro-stress ) (x , or vice-versa. At micro-level, the
macro-point is viewed as the center of an RVE with domain and boundary . It
can be shown that if linear boundary conditions (BCs) are applied to the RVE, then
the problem of relating macro-strains and stresses and can be transformed
onto that of relating average strains and stresses and over the RVE [21].

We consider two-phase composites: a number of inclusions in a matrix. The matrix-
phase (m) has volume fraction
m
V . The inclusions-phase has a volume fraction
m I
V V =1 .

As mentioned, the sub- and superscripts I and m refer to the inclusions and the
matrix respectively. Hence the following are the constitutive equations for the
composite, inclusion and matrix material
C
eff
= ,
I I I
C = ,
m m m
C = .
2.2-6
The volume average stress and strain tensors for inclusion and matrix materials are
written as follows,
=
I
V I
I
dV
V

1
,
=
m
V m
m
dV
V

1
,.
2.2-7
=
I
V I
I
dV
V

1
,
=
m
V m
m
dV
V

1
.
2.2-8
The relationships between the total averaged stress and strain tensors with the
inclusion and matrix averages are defined by the rule of mixtures as follows,
m
m
I
I
V V + = , 2.2-9
m
m
I
I
V V + = . 2.2-10
Where
I
V and
m
V are the volume fractions of the inclusion and matrix respectively.

Average properties and the strain concentrations are going to be portrayed for a
later use. An important concept first introduced by Hill [62] is the stress and strain
concentration tensors A and B. They are mainly correlating the average inclusion
strain/stress (
I
,
I
) with the corresponding average strain/stress ( , ) of
the whole composite material. They are fourth order tensors. They are used as
follows,
A =
I
, 2.2-11
B =
I
. 2.2-12
We will introduce another two tensors, the first is called alternate strain
concentration tensor and denoted by A
, the second is called alternate stress

concentration tensor and denoted by B
. Both of these tensors relates the averages of

the inclusions and the matrix strains and stresses accordingly as shown below [48],
m I
A

= ,
m I
B

= . 2.2-13
From eqns. 2.2-10, 2.2-11, and 2.2-13 we obtain a relation between A and the A

as follows,
( ) [ ]
1

+ = A I A A
I I
V V .
2.2-14
Similarly from eqns. 2.2-9, 2.2-12, and 2.2-13 using the alternate stress
concentration tensors B
by analogy we will obtain the following relationship,

( ) [ ]
1

+ = B I B B
I I
V V .
2.2-15
With eqns. 2.2-6, one can relate the average composite stiffness and compliance in
terms of the strain/stress concentration tensors A , B and the inclusion/matrix
properties,
( )A C C C C
m I
I
m eff
V + = , 2.2-16
( )B S S S S
m I
I
m eff
V + = . 2.2-17
The effective or average compliance is related to the average stiffness by,
( )
1
=
eff eff
C S .
2.2-18

2.3. Semi-analytical Methods

2.3.1. The Theoretical Bounds of HILL REUSS VOIGT

As discussed in Chapter I, early studies were carried on the concept of
homogenization of heterogeneous microstructures. Some were done more than a
century ago by VOIGT (1889)[13] and REUSS (1929)[12] who both proposed different
simple approximations for the effective material data of heterogeneous linear elastic
materials, which have been the basis for a primary result by HILL (1952)[15].

The VOIGT and REUSS assumptions have a physical interpretation as being displayed
in (Fig. 2.3.1). Looking at a rod under a tension load the VOIGT approach would be
exact for different materials being connected in parallel relative to the applied load.
The REUSS model would be exact for different materials being connected in series.

Materials in series
connection
u
Reuss
Materials in parallel
connection
u
Voigt
Materials in series
connection
u
Reuss
Materials in parallel
connection
u
Voigt

Fig. 2.3.1 Geometrical Interpretation of VOIGT and REUSS theoretical bounds

Taking into consideration the basic theory behind the Voigt and Reuss bounds we
can illustrate the use of the strain concentrations mention in the previous section.
Regarding Voigt bound, it corresponds to the assumption that the inclusion and the
matrix both experience the same uniform strain. Then the average strain tensors of
the inclusion and matrix are equal to the composite average strain =
I
, so
using eqn. ( 2.2-11) we get I A = , therefore we conclude from the stiffness tensor
eqn. ( 2.2-16) the following expression [48],
( )
m
m
I
I
m I
I
m Voigt
V V V C C C C C C + = + = . 2.3-1
This leads to the representation of the Voigt upper bound (or the rule of mixtures) of
the effective stiffness of the composite. On the other hand, Reuss assumes that the
inclusion and matrix experience same uniform stress. This means that the stress
concentration I B = and by analogy the compliance tensor eqn. ( 2.2-17) turns to be
[48],
( )
m
m
I
I
m I
I
m
V V V S S S S S S + = + =
Reuss
. 2.3-2
This leads to the representation of the Reuss lower bound of the effective stiffness of
the composite.
2.3.2. The asymptotic HASHIN SHTRIKMAN bounds

More tight bounds for linear elasticity were proposed by HASHIN and SHTRIKMAN
(1962), (1963). HASHIN and SHTRIKMAN bounds are based on variational principles.
They engage the principle of minimum potential energy and the principle of
minimum complementary potential energy.

The bounds for the material parameters of an isotropic linear elastic effective
material consisting of two phases with volume fractions
1
V and
2
V and material
parameters
1
K ,
2
K ,
1
and
2
, respectively are (see [60] and [61])
2 2
2
2 1
2
2
1 1
2
2 1
2
1
4 3
3 1
1
4 3
) 1 ( 3 1
+
+
+
+
+
K
V
K K
V
K K
K
V
K K
V
K
eff

2.3-3
) 4 3 ( 5
) 2 ( 6 1
1
) 4 3 ( 5
) 2 )( 1 ( 6 1
2 2 2
2 2 2
2 1
2
2
1 1 1
1 1 2
2 1
2
1

+
+
+
+
+
+
K
K V
V
K
K V
V
eff

2.3-4

The HASHIN and SHTRIKMAN bounds are only asymptotic bounds and strictly valid only
for a theoretically infinite size of the representative volume element they are used
for. But they are the tightest possible bounds for general isotropic materials without
restrictions on the geometry of the microstructure [61].

2.3.3. The ESHELBY Model

The introduction of the Eshelby model middle of the 20
th
century is one of the major
achievements in the analytical approach for predicting the effective material
properties of heterogeneous microstructures besides the previously mentioned
bounds. Many models are based on this analysis going back to the work of ESHELBY
(1957) who found a general solution for one ellipsoidal particle embedded in an
infinite matrix in linear elasticity. ESHELBY (1957) found that for a homogeneous
isotropic infinite body with an ellipsoidal inclusion subjected to a uniform
eigenstrain
*
, the resulting strain field within the inclusion is uniform and can be
described by,
*
= , 2.3-5
where is the called the fourth rank ESHELBY tensor. It only depends on the
geometry of the ellipsoidal inclusion and POISSONs ratio. Eshelbys model can be
used to predict the effective stiffness of a composite with ellipsoidal inclusions at
dilute concentrations. That is why it is sometimes called the dilute Eshelbys model.
Note that the average strain is identical to the applied strain
A
[48]
A
= . 2.3-6
Therefore the relation between the average composite strain and the average
inclusions strain is
( ) [ ]
m I m I
C C S I + = , 2.3-7
For a representative formulation of the above equation, see Tucker (1999)[27].
Using the principle of strain concentration tensors introduced in section 2.2 then
eqns. ( 2.2-11) and ( 2.3-7) give us,

Eshelby
I
A = . 2.3-8
Therefore,
( ) [ ]
1
+ =
m I m Eshelby
C C S I A .
2.3-9
This form will be later used for the derivation of Mori-Tanakas and Lielens models,
which are based on Eshelbys model.

2.3.4. The MORI-TANAKA Model

The Mori-Tanaka (MT) model was proposed by Mori and Tanaka (1973) and it is
suited for composites with moderate inclusion volume fraction. For details on
formulation and numerical implementation see Doghri and Ouaar [21].

Mori-Tanakas assumption was that when many identical particles are introduced
to the composite microstructure, the average inclusion strain is given by
m
Eshelby
I
A = . 2.3-10
This means, within a concentrated composite each single inclusion sees a far field
strain that is equal to the average strain of the matrix rather than the composite as
in Eshelbys case (See eqn. 2.3-8). Here we will use the alternate strain concentration
tensor described in eqn. ( 2.2-13) to correlate MT to Eshelbys strain concentration
tensor,
Eshelby MT
A A =
.
2.3-11
Using the above result with eqn. ( 2.2-14) one obains the Mori-Tanaka strain
concentration tensor,

( ) [ ]
1
1

+ =
Eshelby
I I
Eshelby MT
v v A I A A .
2.3-12
This can be used in eqn. (2.2-16) to calculate the overall effective stiffness tensor of
the composite.

I
c
m
c
I
c
m
c

Fig. 2.3.2 Mori-Tanakas proposition

2.3.5. The Self-Consistent Method

Another approach to account for higher inclusions volume fractions is the self
consistent method which assumes the surrounding material of one inclusion to be
the effective material of the composite itself. Therefore, eqn. ( 2.3-9) leads to the
following form by substituting the matrix properties with the composites properties
( ) [ ]
1
+ = C C S I A
I SC

2.3-13
This self consistent approach is generally credited to Hill [17][16] who originally
focused on spherical particles and fully aligned fibers.

2.3.6. Interpolative Double Inclusion Model (LIELENS Model)

The double inclusion model (DI) was proposed by Nemat-Nasser and Hori [4] see
also Lielens [43], and supposes that each spheroidal inclusion ) (I of stiffness
) (
I
C is wrapped with a matrix material of stiffness ) (
m
C . The outer reference
material has a stiffness ) (
R
C . The composite has an average or effective stiffness
) (
eff
C . For details on formulation and numerical implementation see Doghri and
Ouaar [21].

By choosing the surrounding reference stiffness ) (
R
C to be either the matrix,
inclusions or the effective composites stiffness; one can retrieve many
homogenization models. The choice ) (
eff R
C C = , this means that the inclusion is
surrounded by a material having the effective stiffness of the whole composite, gives
the generalized self consistent model. A second choice is ) (
m R
C C = , the stiffness of
the real matrix material gives the Mori-Tanaka model, thus describing a lower
bound for the alternate concentration tensor on the presumption that the inclusion
is stiffer than the matrix, as it is shown below,
( ) [ ]
1

+ = = =
m I m Eshelby MT lower
C C S I A A A .
2.3-14
A third choice of ) (
I R
C C = , the stiffness of the real inclusion ) (I . This means that
the matrix is stiffer and engulfing a softer material of the inclusion. In this case, it
is found that
[ ] [ ] ( ) [ ]
I m I I Eshelby MT upper
C C S I A A A + = = =
1 1

.
2.3-15
This case can be called the inverse MT model, as it corresponds to MT for a
composite where the material properties of the inclusion and the matrix are
permuted. This describes the upper bound since the reference is taken to be the
stiffer material of the inclusion. Note that
I
is now calculated for inclusions of
matrix material and surrounded by the inclusion material, i.e. inversed.

lower
A
and
upper
A
correspond to lower and upper stiffness estimates

respectively, which are very much related to Hashin-Shtrikman bounds (Hashin
and Shtrikman, 1963). Consequently, Lielens (Chapter 9 in (Lielens, 1999))[59]
proposed a homogenization model that assumes for higher volume fractions the
effective properties are more drifted towards the upper bound (the stiffer phase), in
this case the inclusion. While at relatively lower volume fractions the model is
drifting results towards the lower bound (the softer phase), in this case the binding
matrix, which is based on the following interpolation between these bounds based
on the inverse of the alternate concentration tensors
( )[ ] [ ] { }
1
1 1

1

+ =
upper lower Lielens
A A A
2.3-16
where ) (
I
V is a function of the inclusions volume fraction and is a smooth
interpolation function which satisfies:
, 0 ) ( >
I
V , 0 ) ( >
I
I
V
dv
d

, 0 ) (
lim
0
=
I
v
V
I

, 1 ) (
lim
1
=
I
v
V
I

2.3-17
Lielens proposed the following expression for ) (
I
V :
( )
I I I
V V V + = 1
2
1
) ( 2.3-18

2.3.7. The HALPIN-TSAI Model

There exist numerous micromechanics-based models which were developed to
predict a complete set of elastic constants for aligned short-fiber composites. One of
the most popular ones is the Halpin-Tsai model [71] which was initially developed
for continuous fiber composites and which was derived from the self-consistent
models of Hermans [22] and Hill [23].

The Halpin-Tsai equations can be expressed in a short and effortlessly usable form
which might be one of the reasons why they have found a broad usage:
I
I
m
V
V
M
M
+
=
1
1
with
=
I
I
M
M 1

2.3-19
I
V is the fiber volume fraction and M stands for any of the moduli listed in Table:
2.3-1. E11 and E22 are the longitudinal and the transverse Youngs modulus, G12
and G23 the in-plane and out-plane shear modulus, respectively. K23 is the plane-
strain bulk modulus and
21
the longitudinal Poissons ratio of the unidirectional
transversely isotropic short fiber composite. The corresponding values of the
empirical parameter are also listed in Table: 2.3-1.

M MI Mm

E11 EI Em 2a
E22 EI Em 2
G12 GI Gm 1
G23 GI Gm
=
I
I
M
M 1

Table: 2.3-1 is listed for the different substitutions of MI and Mm

is correlated with the geometry of the reinforcement and it was found empirically
that predictions for E11, the Youngs modulus in fiber direction, are best if a 2 = ,
where a is the fiber aspect ratio, defined as:
d
L
a = , 2.3-20
L is the fiber length and d the fiber diameter. It can be shown that for the
Halpin-Tsai equations become the rule of mixtures (Voigt bound) where fiber and
matrix experience the same, uniform strain:
m m I I
M V M V M + = , 2.3-21
The rule of mixtures is also applied to calculate the longitudinal Poissons ratio
21

although predictions are not accurate when matrix and fibers have considerably
different Poissons ratios.
m m I I
v V v V v + =
21
, 2.3-22
The Halpin-Tsai model can deal both with isotropic and transversely isotropic fibers
e.g carbon fibers because the underlying self-consistent theories of Hermans [22]
and Hill [23] apply also to transversely isotropic fibers.

2.4. Comparison of semi-analytical methods

The following comparison is done between different homogenization methods for
three different cases. The tested composite materials are isotropic with the following
properties listed below (shear and bulk moduli):

CASES Material Matrix (MPa) Inclusion (MPa)
Case -I- Low Contrast
G = 3
K = 7
G = 30
K = 70
Case -II- High Contrast
G = 3
K = 7
G = 3000
K = 7000
Case -III- Voids
G = 3
K = 7
G = 0
K = 0

As shown in (Fig. 2.4.1, Fig. 2.4.2 and Fig. 2.4.3), the effective material properties
are plotted as a function of the volume fraction of the inclusion material. One can
clearly see that the direct methods have severe difficulties handling the extreme
cases of rigid or void inclusions, and the more the material data of the matrix and
the inclusion material differ, the less reliable are the analytical approximations even
for the linear elastic case. A general overview of the various analytical
approximation methods for linear elastic materials is given for example in Aboudi
(1992) [29].

0
10
20
30
40
50
60
70
80
90
0 0.2 0.4 0.6 0.8 1
Volume fraction (inclusion) [-]
E
-
m
o
d
u
l
u
s

[
M
P
a
]
Voigt
Reuss
HSlow
HSup
Halpin
Lielens
Mori-Tanaka
Self consistent
K
matrix
=7 MPa
G
matrix
=3 MPa
K
inclusion
=70 MPa
G
inclusion
=30 MPa

0
5
10
15
20
25
30
35
0.000 0.200 0.400 0.600 0.800 1.000
G
-
m
o
d
u
l
u
s

[
M
P
a
]
Voigt
Reuss
HSlow
HSup
Halpin
Lielens
Mori-Tanaka
Self Consistent
K
matrix
=7 MPa
G
matrix
=3 MPa
K
inclusion
=70 MPa
G
inclusion
=30 MPa

Fig. 2.4.1 CASE I Low contrast

0
1000
2000
3000
4000
5000
6000
7000
0 0.2 0.4 0.6 0.8 1
E
-
m
o
d
u
l
u
s

[
M
P
a
]
Voigt
Reuss
HSlow
HSup
Halpin
Lielens
Mori-Tanaka
Self Consistent
K
matrix
=7 MPa
G
matrix
=3 MPa
K
inclusion
=7000 MPa
G
inclusion
=3000 MPa

0
500
1000
1500
2000
2500
3000
0.000 0.200 0.400 0.600 0.800 1.000
G
-
m
o
d
u
l
u
s

[
M
P
a
]
Voigt
Reuss
HSlow
HSup
Halpin
Lielens
Mori-Tanaka
Self Consistent
K
matrix
=7 MPa
G
matrix
=3 MPa
K
inclusion
=7000 MPa
G
inclusion
=3000 MPa

Fig. 2.4.2 CASE II High contrast

0
1
2
3
4
5
6
7
8
9
10
0 0.2 0.4 0.6 0.8 1
Voluma fraction (inclusion) [-]
E
-
m
o
d
u
l
u
s

[
M
P
a
]
Voigt
HSlow
Mori-Tanaka
Self Consistent
K
matrix
=7 MPa
G
matrix
=3 MPa
K
inclusion
=0 MPa
G
inclusion
=0 MPa
-Hashin-Shtrikman
upper bound, HS
up
-Reuss
-Halpin-Tsai
-Lielens
(are not applicable)

0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
0.000 0.200 0.400 0.600 0.800 1.000
Volume fraction (inclusion) [MPa]
G
-
m
o
d
u
l
u
s

[
M
P
a
]
Voigt
HSlow
Halpin
Mori-Tanaka
Self Consistent
K
matrix
=7 MPa
G
matrix
=3 MPa
K
inclusion
=0 MPa
G
inclusion
=0 MPa
-Hashin-Shtrikman
upper bound, HS
up
-Reuss
-Halpin-Tsai
-Lielens
(are not applicable)

Fig. 2.4.3 CASE III Voids

2.5. Direct Numerical Methods (RVE)

The limitations faced when using analytical approximation methods make direct
numerical simulations necessary. Therefore, during the last years numerical
methods to directly compute the effective material data gained more and more
interest and importance. Most of those methods are developed only for linear
material laws and small deformations. Just recently and due to the increasing
computational power available, a couple of methods for non-linear elasticity and
general non-linear material behavior have been developed [61]. Theoretical work has
been done by HUET (1982)[30], HUET (1990)[31] and TORQUATO (1991)[32].
ZOHDI & WRIGGERS (2001a)[33] and ZOHDI & WRIGGERS (2001b) [34] have
worked on computational homogenization of geometrically linear and possibly
materially non-linear microstructures. For fiber reinforced materials including
delamination at small strains see e.g. DBERT (2001)[35]. For general 2D problems
with DIRICHLET boundary conditions fulfilled only in the integral over the
representative volume element work has been done by LARSSON & RUNESSON
(2004)[36]. Homogenization at finite strains and possibly inelastic material behavior
has been done by SCHRDER (2000)[37], MIEHE ET AL. (2002)[39] and MIEHE
(2003)[38]. This approach yields a load dependent effective material tangent
stiffness and this way is applicable to multi-scale methods. Multi-scale methods
have been investigated for example by ORTIZ (2003)[40].

1
0
2
0
tion Homogeniza
1
0
2
0
tion Homogeniza

Fig. 2.5.1 Representative Volume Element (RVE)

The representative volume element (RVE) is a sample of the material microstructure
that the macroscopic body consists of as shown in (Fig. 2.5.1). In order to be
representative, the size of the RVE is of great importance. As will be discussed in
details later in this monograph, HILLs condition can be used as a criterion for the
quality of the choice of the size of the RVE. In the initial configuration the RVE
assumes the domain
0
. Assuming the RVE consists of a matrix material and an
inclusion material the domain
0
can be divided into the two sub-domains
1
0
and
2
0
representing the matrix and the particle material.

2.6. Average Strain Theorem

The average strain theorem states that for any perfectly bonded material within the
RVE and for an exterior homogeneous displacement given on the entire boundary of
the RVE, the volume average of the strain is the applied displacement on the
boundary. For details on the theory and derivation see Lhnert [61].

2.7. Average Stress Theorem

The average stress theorem states that for a given uniform external load on an RVE
the volume average of the forces within the RVE is identical to the given force on the
boundary in case of absence of body forces. Again, this can easily be proved for the
small deformation theory. For more details on the theory and derivation, see
Lhnert [61].

2.8. Hills Theorem

As for the practical applications, the choice of the size of the RVE is very important.
It has to be chosen in a way such that on the one hand the RVE is small relative to
the macroscopic body of interest and on the other hand it is large enough such that
the boundary field fluctuations are relatively small. It is widely accepted that HILLs
condition is an appropriate criterion for the right choice of the size of the RVE. It
states that the energy on the micro-level has to be the same as the effective energy
for the homogenized material over the volume if uniform stresses or strains exist on
the boundary of the RVE, and it is formulated as follows,

: : = , 2.8-1

2.9. The Effective Material Law

The aim of the homogenization procedure is to find constitutive equations for the
effective material and to fit the parameters introduced in this constitutive model to
the results of the numerical homogenization. In order to find a material law and
parameters for the effective material some further assumptions and requirements
are necessary. In case the constitutive model for the effective material is assumed to
be isotropic, it is necessary that the microstructure is statistically able to reflect
isotropic behavior. For non-spherical inclusions it is important that the distribution
as well as the orientation of the inclusions is random.

2.9.1. Small Elastic Deformation

The small elastic deformation conjecture is applied in our calculations of the
effective material properties of sphere-filled RVEs (material-microstructures) in the
case of isotropy with linear elastic assumption. The numerical homogenization is
part of a whole automated process in creating our material model that will be
discussed in details in the following chapters. The simplest case to get the effective
material model and the corresponding parameters is given for microstructures
which only consist of linear elastic materials which are perfectly bonded. Due to the
linearity on the micro-level, the effective material model is linear elastic as well
since the principle of superposition holds. As a result the material parameters will
definitely not be path dependent.

In general, even if the materials on the micro-level are isotropic, the effective
material can show anisotropic behavior. A general anisotropic linear elastic material
may have twenty one independent material parameters. With one type of test,
meaning one set of boundary conditions it is possible to calculate at most six of
those parameters, since one obtains only six independent effective stress or strain
components for the six independent strain or stress components prescribed on the
boundary of the RVE [61].
0
0
0
0
0
0
0
0
0
0
0
0
31
23
12
33
22
11
3131 2331 1231 3331 2231 1131
2331 2323 1223 3323 2223 1123
1231 1223 1212 3312 2212 1112
3331 3323 3312 3333 2233 1133
2231 2223 2212 2233 2222 1122
1131 1123 1112 1133 1122 1111
31
23
12
33
22
11
2
2
2
eff eff eff eff eff eff

C C C C C C
C C C C C C
C C C C C C
C C C C C C
C C C C C C
C C C C C C

2.9-1

For the isotropic case it is reduced to the following,
[ ] [ ] [ ]
[ ] [ ] [ ]
[ ] [ ] [ ]
[ ]
[ ]
[ ]

+
+
+
=
0
0
0
0
0
0
0
0
0
0
0
0
31
23
12
33
22
11
31
23
12
33
22
11
2
2
2
3 0 0 0 0 0
0 3 0 0 0 0
0 0 3 0 0 0
0 0 0 4 3 2 3 2 3
0 0 0 2 3 4 3 2 3
0 0 0 2 3 2 3 4 3
3
1
k k k
k k k
k k k

2.9-2

So for a general anisotropic linear elastic effective material at least 4 different
linearly independent types of tests have to be performed in order to be able to
calculate all the effective material parameters. In case of isotropy the number of
parameters reduces to two, so one type of test with an appropriate choice of
boundary conditions will be enough to obtain both material parameters. However, if
the effective material is assumed to be nearly incompressible, to get more accurate
effective material data it might be better to perform two tests, one isochoric test to
obtain the shear modulus
eff
, and one volumetric test to obtain the bulk modulus
eff
K [61]. Then the effective material parameters can be computed from eqn. 2.9-2
as follows,
0
0
3
3
3
1
tr
tr
K
eff

2.9-3
and
0 0
0 0
:
:
2
1

=
d d
d d
eff

2.9-4
where 1

3
tr
d
= and 1

3
tr
d
= are the deviatoric parts of the stress and strain
tensor, respectively.
The elastic modulus and Poisson ratio are calculated according to the classical
relations of isotropy,
eff eff
eff eff
eff
K
K
E
+
=
3
9
and
eff eff
eff eff
eff
K
K
2 6
2 3
+
= 2.9-5

In the homogenization process for isotropic material, we followed a statistical
procedure in order to dilute the effect of the non-structural geometries of our
randomly generated microstructures which exist in reality. Taking for example an
RVE made up of randomly distributed spheres, we apply a tensile load in the X, Y
and Z directions and we run a simulation job for each case. Afterward, we compute
the effective MP (
eff eff
and E ) for each case then we calculate the average of the
three results as follows:

( ) 3
eff
CaseIII
eff
CaseII
eff
CaseI
eff
E E E E + + =
2.9-6

( ) 3
eff
CaseIII
eff
CaseII
eff
CaseI
eff
+ + =
2.9-7

In CHAPTER IV, the statistical implementation of the above homogenization will be
thoroughly discussed and compared to Mori-Tanaka and Lielens Models.

2.10. Main Stream Averaging Homogenization

The aim of the main stream averaging (MSA) homogenization procedure is to find
constitutive equations for the effective material properties using a simple direct
volume averaging equations. Many researchers like Charles L. Tucker III and Erwin
Liang [48] used this method in transverse isotropic homogenization to calculate the
longitudinal and transverse independent material properties. Also this method
bypasses the time and computation efforts needed by the optimization method to
calculate the five independent material properties as done by Gusev, see [49],[50].

In this work, we actually suggested the naming of this method to be Main Stream
Averaging Homognisation, because it calculates the effective material property in
the main direction of stresses relative to the perturbation used on the model. A
prerequisite for this method to properly function is to have a uni-axial stress state
in the direction of the perturbation when applying compressive or tensile loadings
on the micro-structure. I.e., we apply a definite perturbation to calculate a definite
material property. For example, if applying a tensile perturbation in the x-direction
along fiber alignment in a transverse isotropic case, then the effective longitudinal
E-modulus is a function of the volume average stress and strain tensors in the
direction of the perturbation,
0
0
=
xx
xx
eff
xx
E

2.10-1
In this case we neglect the other stress tensor components (with a negligible error).
Actually the study of different types of RVEs that will be discussed in the coming
chapters, shows less than 0.001% error if compared to an exact solution which
exists in the isotropic case, see Section 2.9.1.

Fig. 2.10.1 Same size, non-periodic, edge trimmed sphere-filled RVE

This is valid only if we used proper boundary conditions that retain a uni-axial
stress state, which is given by the 3D homogeneous boundary conditions and our
3D periodic boundary conditions that are discussed comprehensively in CHAPTER
II.

2.10.1. Formulation

The formulations described in the following sections are arranged according to the
respective perturbation applied to the model and divided into two procedures, one
for the isotropic case with two independent parameters ( , E ) and the second for the
transverse isotropic case with five independent parameters (
12 23 22 12 11
, , , , G E E ).

2.10.1.1. Isotropic Case

Main stream averaging (MSA) homogenization method is not mainly used for the
isotropic case in view of the fact that the exact solution exists for the two
independent parameters as described in Section 2.9.1. But it is later on used in the
verification and validation of the MSA homogenization in comparison to the exact
formulation in the case of isotropy; consequently to be adopted for our transverse
isotropic case of fiber filled RVEs.

Case I: MSA Formulation for tensile perturbation in the x-direction (Fig. 2.10.1) is:

0
0
11
11
11
= =
eff eff
E E
2.10-2

( ) 2
;
13 12
11
33
13
11
22
12
0
0
0
0
eff eff eff
eff eff

+ =
= =

2.10-3

Case II: MSA Formulation for tensile perturbation in the y-direction (Fig. 2.10.1) is:

0
0
22
22
22
= =
eff eff
E E
2.10-4

( ) 2
;
23 21
22
33
23
22
11
21
0
0
0
0
eff eff eff
eff eff

+ =
= =

2.10-5

Case III: MSA Formulation for tensile perturbation in the z-direction (Fig. 2.10.1) is:

0
0
33
33
33
= =
eff eff
E E
2.10-6

( ) 2
;
32 31
33
22
32
33
11
31
0
0
0
0
eff eff eff
eff eff

+ =
= =

2.10-7

In this case of isotropic homogenization and for statistical purposes, if wished, the
above equations can be used in conjunction with each other, i.e. the three
perturbations are performed and the material properties from each case are
calculated and then averaged,

( ) 3
eff
CaseIII
eff
CaseII
eff
CaseI
eff
E E E E + + =
2.10-8

( ) 3
eff
CaseIII
eff
CaseII
eff
CaseI
eff
+ + =
2.10-9
In our statistical studies on sphere-filled RVEs, we will use the exact solution in
Section 2.9.1 to calculate the two independent material properties.

2.10.1.2. Transverse Isotropic/Orthotropic Case

In our study, MSA homogenization method is used for the transverse isotropic case
of fiber reinforced composite material. The representative volume elements are filled
with sphero-cylindrical inclusions (Fig. 2.10.2).

2
3
1
2
3
1
2
3
1

Fig. 2.10.2 Same size, non-periodic, edge trimmed
fully aligned fiber-filled RVE

In order to calculate the five independent effective material parameters, the
microstructure undergoes three types of specific perturbations where each type
enables the calculation of the effective properties in its direction in terms of the
volume averaged stress and strain tensors as shown below:

Case I: MSA Formulation for tensile perturbation in the 1-direction, (
eff eff
E
12 11
; ) are:

0
0
11
11
11
eff
E
2.10-10

0
0
11
22
12
eff

2.10-11

Case II: MSA Formulation for tensile perturbation in the 2-direction, (
eff eff
E
23 22
; ) are:

0
0
22
22
22
eff
E
2.10-12

0
0
22
33
23
eff

2.10-13

Case III: MSA Formulation for tensile perturbation in the 3-direction, (
eff eff
E
31 33
; ) are:

0
0
33
33
33
eff
E
2.10-14

0
0
33
11
31
eff

2.10-15

Case IV: MSA Formulation for shear perturbation in the 12-direction, (
eff
G
12
) is:
0
0
12
12
12
eff
G ,
0
0
32
32
32
eff
G where
12 12
2 = and
32 32
2 =
2.10-16
In our statistical studies on fiber-filled RVEs, we will use the MSA homogenisation
for the transverse isotropic case to calculate the five independent material
properties and compare them to the existing semi-analytical models. This can be
surely expanded by adding other perturbations to calculate the other independent
parameters in a model with fiber orientation, where orthotropic behaviour governs
the composite material model. In our orthotropic models we will calculate up to 7
independent parameters enough for reasonable comparison and evaluation of the
semi-analytical models like Mori-Tanakas and Lielens.

CHAPTER II

3. A Comprehensive Study, 3D Boundary Conditions

X- Tensile
Perturbation
X- Tensile
Perturbation
Y
X
PRESCRIBED
3D PERIODIC
X
Y
Z
X
Y
Z
X- Tensile
Perturbation
X- Tensile
Perturbation
X- Tensile
Perturbation
YY
XX
PRESCRIBED
3D PERIODIC
X
Y
Z
X
Y
Z
X
Y
Z
XX
YY
ZZZ

This chapter includes a detailed study of the construction of three dimensional
periodic and homogeneous boundary conditions describing the targeted
advantages they endow with later usage in the RVEs micro-simulations

CHAPTER II - A Comprehensive Study, 3D Boundary Conditions - 49 -
3.1. Introduction

In a numerical simulation, it is unnecessary and not efficient to simulate everything
we encounter, especially when there are reliable analytical models that sufficiently
describe and solve the task. Generally we choose a region of interest in which we
conduct a numerical simulation where limitations of analytical models exist. The
interesting region has a certain boundary with the surrounding environment.
Numerical simulations therefore have to consider the physical processes in the
boundary region by appropriately chosen boundary conditions (BC). Different
boundary conditions may cause quite different simulation results. Improper sets of
boundary conditions may introduce nonphysical influences on the simulation
system. So arranging the boundary conditions for different problems becomes very
important. While at the same time, different variables in the environment may have
different boundary conditions according to certain physical problems.

Starting by the concept of micro-scale heterogeneous materials reveals the
need to build up FE models that describe their realistic heterogeneous
microstructure. So it is not only important to define or build a representative
volume element (RVE) but also to have the proper choice of boundary conditions
that enables us to correctly predict the simulation response from which we will
calculate the effective material properties using direct numerical homogenization
methods as described in CHAPTER I in Sections (2.9) and (2.10).

Boundary conditions mainly portray the macro behavior of the microstructure in
the simulation process. For instance, the two types of boundary conditions we are
using in this work depict a couple of aspects. If we pick the choice of an RVE with
3D periodic boundary conditions then the simulation results being generated would
represent a macro structure consisting of periodically-repeated cells. While the
choice of 3D homogeneous boundary conditions, the results would consider the RVE
as the macro structure itself with its micro-constituents.

Generally speaking, boundary conditions represent the type and value of
constraints that control the simulation response of our microstructure (RVE) and
their usage is dependant on the type of the RVE used due to the various geometric
possibilities in generating them. The types of RVEs will be thoroughly discussed in
CHAPTER IV and CHAPTER (V).

In this chapter, the index notation is used for the detailed formulation of the
boundary conditions.

3.2. Implemented 3D Homogeneous Boundary Conditions
for RVE

The concept of the 3D homogeneous boundary conditions (HBC) is also used in our
microscopic approach in modeling and simulating the composite material
(inclusion/matrix). HBC are less time consuming than periodic boundary
conditions, hence they are more suitable for sufficiently large RVEs, i.e. containing
a high number of inclusions to be represented as close as possible to reality. In our
study they are intensively used in the modeling of fiber reinforced composite
material, where cylindrical and sphero-cylindrical inclusions are used. Keeping in
mind that PBC is logically useful with periodic microstructures (RVEs with periodic
boundary-trimmed inclusions, or fully embedded inclusions). In this work the RVEs
representing fully aligned fiber reinforced composites are not geometrically periodic
microstructures. Consequently the studies on those RVEs that are laid down in
CHAPTER V used homogeneous boundary conditions.

3.2.1. Theory

Following (Aboudi 1991) [54], homogeneous boundary conditions are obtained in
the following forms,
i. Kinematic uniform boundary conditions (KUBC): The displacement
i
u is
imposed at point
i
x belonging to the boundary such that:
( )
j ij i
x u
0
=
j
x 3.2-1
Where
0
ij
is a constant tensor representing the global strains that does not
depend on
i
x

The KUBC are the adopted in this work representing the applied homogenous
boundary conditions on our FE-models.

ii. Static uniform boundary conditions (SUBC): The traction vector
i
t is
prescribed at the boundary such that:
( )
j ij i
n t
0
=
j
x 3.2-2
Where
0
ij
is a constant tensor representing the global stresses that does not depend
on
i
x , and
i
n the vector normal to at
i
x

3.2.2. Development and Construction

Homogeneous boundary conditions (HBC) are used on all our material models
especially on RVEs with relatively large number of finite elements, because HBC
provide less computational effort in simulation than the periodic boundary
conditions we mentioned above, due to the large number of constraints or
equations used in the periodicity analysis, see Section 3.3.3. During the
development stage, the first condition was to prevent rigid body movement. The idea
was to restrain certain degrees of freedom (DOF) of three adjacent faces of the cube
and set free the others and apply a certain perturbation. Two cases are developed
for the microstructure. They are homogeneous boundary conditions for:

Uni-axial perturbations (Tension/Compression)
Shear Loadings

In order to have a flexible control over the boundaries of the microstructure, six
reference points for each of the faces of the cube have been introduced. Six node
sets are established in reference to Fig. 3.3.3 and Fig. 3.2.1 as follows:

Node Set - Description (Cube faces) Ref. Point

o Xinf
o Xsup
o Yinf
o Ysup
o Zinf
o Zsup

- All Rear Nodes (AA1D1D)
- All Front Nodes (BB1C1C)
- All Bottom Nodes (AB1C1D1)
- All Top Nodes (ABCD)
- All Right Nodes (CC1D1D)
- All Left Nodes (AA1B1B)

o RP 1
o RP 4
o RP 2
o RP 5
o RP 3
o RP 6
Table 3.2-1 Boundary sets and their corresponding reference points

In the above table, each node set represents a group nodes that belongs to one of
the RVE boundary surfaces (cube faces). For example, Xinf is the set of nodes
belonging to inferior cube-face normal to the x-axis while Xsup is the set of nodes
belonging to the superior cube-face normal to the x-axis. The nomenclature of the
RVE boundary surfaces are described in Fig. 3.3.3.

In Fig. 3.2.1, a fully meshed RVE is presented along with the relative reference
points for each face (boundary). These projections show each reference point with
its relative surface. Each RP is tied to the set of nodes of the corresponding cube
face using the *TIE Abaqus command. In the following section we will establish the
two types of homogeneous boundary conditions implemented on our RVEs for:

Tensile Perturbation
Shear Perturbation


Fig. 3.2.1 Reference points analogy in homogeneous boundary conditions

3.2.3. Formulation

The HBC for tensile perturbation, as its shown in the schematic of Fig. 3.2.2, are
implemented by fixing only the normal directions of three adjacent faces of the cube
and applying a tensile perturbation on one or a combination of the other three free
faces, in one of the three directions X, Y, Z or a combination for a multi-axial
loading.

X- Tensile
Perturbation
Y
X
PRESCRIBED
TENSILE
X
Y
Z
X
Y
Z
X- Tensile
Perturbation
Y
X
PRESCRIBED
TENSILE
X- Tensile
Perturbation
X- Tensile
Perturbation
Y
XX
PRESCRIBED
TENSILE
X
Y
Z
X
Y
Z
X
Y
Z
XX
YY
ZZZ

Fig. 3.2.2 Homogeneous Boundary Conditions Tensile Perturbation
(only the normal directions are fixed at the boundaries)

The loads implemented in our statistical testing of RVEs are uni-axial. The
superscripts represent the node sets described in Table 3.2-1. These displacement
conditions are formulated as follows:

General conditions
0
inf
1
=
X
u
0
inf
2
=
Y
u
0
inf
3
=
Z
u
X-direction
0
sup
1
=
Load X
u u
Y-direction
0
sup
2
=
Load Y
u u
Z-direction
0
sup
3
=
Load Z
u u
Where the tensile
load . const u
Load
=
Table 3.2-2 Homogeneous tensile perturbation boundary conditions

The HBC for shear perturbation, as its shown in Fig. 3.2.3, are implemented by
fixing the respective in-plane directions of three adjacent faces of the cube and
applying a shear perturbation on one of the other three free faces, in one of the
three in-plane directions X, Y or Z.

X
Y
Z
XY- Shear
Perturbation
X
Z
PRESCRIBED
SHEAR
Y
X
Z
U
Laod
X
Y
Z
X
Y
Z
XY- Shear
Perturbation
X
Z
PRESCRIBED
SHEAR
XY- Shear
Perturbation
X
Z
PRESCRIBED
SHEAR
Y
X
Z
Y
X
Z
U
Laod

Fig. 3.2.3 Homogeneous Boundary Conditions Shear Perturbation
(only the in-plane directions are fixed at the boundaries)

The superscripts represent the node sets described in Table 3.2-1. These shear
displacement conditions are formulated as follows:

General conditions
0
inf
1
=
Y
u
0
inf
2
=
Z
u
0
inf
3
=
X
u
XY-direction
0
sup
1
=
Load Y
u u
YZ-direction
0
sup
2
=
Load Z
u u
ZX-direction
0
sup
3
=
Load X
u u
Where the shear
load . const u
Load
=
Table: 3.2-3 Homogeneous shear perturbation boundary conditions

3.3. Implemented 3D Periodic Boundary Conditions for RVE

The periodic boundary conditions on the RVE surfaces described in Cartesian co-
ordinates are given in an early study by Suquet [46]. The concept of the 3D periodic
boundary conditions (PBC) is assumed in our microscopic approach in modeling
and simulating the composite material (inclusion/matrix). In order to homogenize
data from a 3D microstructure to a 3D meso or macrostructure with the periodicity
assumption, proper 3D periodic boundary conditions (PBC) must be used. The
assumptions needed for using PBC engulf two major conditions, first the geometric
periodicity of the microstructure, second a very small and limited number of
inclusions compared to the targeted macrostructure.

In this work, the implementation for the 3D periodicity on micro-structures,
several properties and advantages are obtained that will be discussed thoroughly in
the coming sections.

3.3.1. Theory

Consider a periodic structure consisting of a periodic array of repeated unit cells
under a macroscopic strain. The displacement field for the periodic structure can be
expressed as:
( ) ( )
3 2 1
* 0
3 2 1
, , , , x x x u x x x x u
i j ij i
+ = .
3.3-1
In the above,
0
ij
is the macroscopic strain tensor of the periodic structure and the
first term on the right side of the equation (
j ij
x u
0 0
= ) represents a linear
displacement field. The second term on the right side of the equation ( )
3 2 1
*
, , x x x u
i
is
the periodic function. It represents a modification to the linear displacement field
due to the heterogeneous structure of the composites. Since the periodic array of
the repeated unit cells represents a continuous physical body, two continuities
must be satisfied at the boundaries of the neighboring unit cells. One is that the
displacements must be continuous, i.e., the adjacent unit cells cannot be separated
or penetrate each other at the boundaries after the deformation. The second
condition implies that the traction distributions at the opposite parallel boundaries
of a unit cell must be the same. In this manner, the individual unit cell can be
assembled as a physically continuous body. The assumption of the displacement
field in the form of eqn. 3.3-1 can not be directly applied to the boundaries since
the periodic part, ( )
3 2 1
*
, , x x x u
i
is generally unknown. For any unit cell, its boundary
surfaces must always appear in parallel pairs, the displacements on a pair of
parallel opposite boundary surfaces can be written as

* 0
i
k
j ij
k
i
u x u + =
+ +
, 3.3-2
* 0
i
k
j ij
k
i
u x u + =

.
3.3-3

Where indices
+
k and

k identify the
th
k pair of two opposite parallel boundary
surfaces of a repeated unit cell. Note that ( )
3 2 1
*
, , x x x u
i
is the same at the two
parallel boundaries (periodicity), therefore, the difference between the above two
equations is

k
j ij
k
j
k
j ij
k
i
k
i
x x x u u = =
+ + 0 0
) ( . 3.3-4
Since
k
j
x are constants for each pair of the parallel boundary surfaces, with a
specified
0
ij
, the right side becomes constant and such equations can easily be
applied in the finite element analysis as nodal displacement constraint equations.

A schematic graphical representation of a 2D unit cell under macro strain with
periodic boundary conditions is displayed in Fig. 3.3.1.
Periodicity with
Macrostrain
2D Representation
Meso-scale periodic array
PBC
+
Macrostrain
k k
x u
1
0
11 1
+
+
k k
x u
2
0
22 2

k
u
2
k
u
1
0
ij
Periodicity with
Macrostrain
2D Representation
Meso-scale periodic array
PBC
+
Macrostrain
k k
x u
1
0
11 1
+
+
k k
x u
2
0
22 2

k
u
2
k
u
1
0
ij

Fig. 3.3.1 Schematic representing a 2D unit cell under macro strain with
periodic boundary conditions

Eqn. 3.3-4 is a special type of displacement boundary condition. Instead of giving
known values of boundary displacements, it specifies the displacement-differences
between two opposite boundaries. Obviously, the application of it will guarantee the
continuity of the displacement field. However, in general, such displacement-
difference boundary conditions, eqn. 3.3-4, may not be complete or may not
guarantee the traction continuity conditions. The traction continuity conditions can
be written as the following:
0 =
+
t t , where n t = , 3.3-5
where
t and
+
t are the tractions at the corresponding parallel boundary surfaces
respectively. Fig. 3.3.2 displays a graphical representation of the traction continuity
condition of a unit cell.

-t
x
k
-
t
y
k
+
t
yx
k
+
-t
yx
k
-
-t
xy
k
-
-t
y
k
-
t
x
k
+
t
xy
k
+
BOUNDARY
TRACTIONS
(2D Schematic)
INFINITESIMAL
STRESSES
-t
x
k
-
t
y
k
+
t
yx
k
+
-t
yx
k
-
-t
xy
k
-
-t
y
k
-
t
x
k
+
t
xy
k
+
BOUNDARY
TRACTIONS
(2D Schematic)
INFINITESIMAL
STRESSES

Fig. 3.3.2 Traction continuity condition on the boundary of a unit cell

For the general periodic boundary value problems, eqn. 3.3-4 and eqn. 3.3-5
constitute a complete set of boundary conditions apart from rigid body motion. In
the following illustrative examples in the verification section, however, if a unit cell
is analyzed by using a displacement-based finite element method as in our case, the
application of only eqn. 3.3-4 can guarantee the uniqueness of the solution and
thus eqn. 3.3-5 is automatically satisfied. In other words, the latter boundary
condition is not necessary to be explicitly applied in the analysis.

3.3.2. Development and Construction

The idea of properly implementing 3D periodicity is to attain the desired periodic
mechanical response (strains and stresses) on the boundaries of the simulated
structure. Our developed method delivers several features in dealing with 3D
periodic analysis. These aspects can be summarized as follows:

Requirements
No gap or overlap between the neighboring RVEs after deformation of a
periodic array representing the composite material, i.e. boundary
displacement continuity is preserved.
Traction continuity is preserved between periodic surfaces.
Sustaining a uni-axial stress state in case of unidirectional perturbations.

Advantages
The whole simulation process of the micro-structure is managed through
a single control-node referred to as the dummy-node.
Independent of the geometry of the micro-structure.
All types of perturbations are applicable (uni-axial, Multi-axial)
Periodic mesh on the boundaries is not a necessity, which allows easier
and more flexible geometry built-up of the micro-structures.

In order to fulfill all the requirements, a dummy node is introduced. As the name
shows, it has nothing to do with the micro-structure; it is not linked to any element
within the model but simply to provide it with the necessary degrees of freedom
through which the RVE mechanical response is controlled.

C
X
Y
Z
A
D
B
C
1
A
1
D
1
B
1
C
X
Y
Z
X
Y
Z
A
D
B
C
1
A
1
D
1
B
1

Fig. 3.3.3 Representative Volume Element planar notation

The first step is to choose and group the regions on the RVE boundaries that have
to be constrained. The planar notation and numbering of the cubic representative
volume element is shown in Fig. 3.3.3. Nodes on the boundary faces of the RVE are
divided into the following groups:
Group I: Inner face nodes are the nodes that belong to one face excluding
the nodes on the common edges and corners.

Group II: Inner edge nodes are the nodes that belong to an edge excluding
the end nodes which resembles the corners.

Group III: Corner or Master nodes are the eight corner nodes of the cubic
RVE.

These three groups contain separate sets of nodes using the following notation as
described in Table 3.3-1,

Group I: Group IIa/b/c: Group III:
o FaceBottom
o FaceTop
o FaceFront
o FaceRear
o FaceLeft
o FaceRight

o EdgeCC1
o EdgeBB1
o EdgeAA1
o EdgeDD1
---
o EdgeCD
o EdgeAB
o EdgeA1B1
o EdgeC1D1
---
o EdgeAD
o EdgeBC
o EdgeB1C1
o EdgeA1D1
o A
o B
o C
o D
o A1
o B1
o C1
o D1
Table 3.3-1 Notation of the node sets used in the PBC analysis
3.3.3. Formulation

Setting up the system of equations between the relative pairs is done through a
linear multi-point constraint. It requires that a linear combination of nodal
variables is equal to zero; that is,
0 ...
2 1
= + + +
R
k N
Q
j
P
i
u A u A u A , 3.3-6
where
P
i
u is a nodal variable at node P , degree of freedom i ; and the
N
A are the
coefficients that define the relative motion of the nodes. After this setup, the
perturbation is introduced to the system of equation through a dummy node which
only acts as a carrier for the load and is not attached to any element in the model.
The general form of the complete set of the system of equations used in our PBC
formulation is represented in eqn. 3.3-7, where
+
k and k are superscripts
representing the relative opposite periodic pairs or node sets noted in Table 3.3-1,
and
dummy
i
u is the perturbation carried on the dummy or control node, so eqn.
3.3-6 which is an interpretation of eqn. 3.3-4 becomes,
0 = +
+ dummy
i
k
i
k
i
u u u . 3.3-7
For example, to apply a tensile displacement load in the y-direction, see Fig. 3.3.5,
while preserving periodicity, we translate this displacement through a linear
constraint between the dummy node, FaceBottom and FaceTop node sets in the y-
direction and fully equate the other two degrees of freedom, eqn. 3.3-7 becomes,

0
1 1
=
FaceTop FaceBottom
u u
0
2 2 2
=
dummy FaceTop FaceBottom
u u u
0
3 3
=
FaceTop FaceBottom
u u
3.3-8
This form of equation is applied to all other pairs in relevance to each other and the
applied load as it is shown further ahead in this section. By applying the dummy
node approach, the model attains no fixation since no restriction of the rigid body
movement is described in the above set of equations, and also enables us to control
the mechanical response of the micro-structure through a single node denoted as
the dummy node.

Another challenge is faced in order to have or maintain a uni-axial stress state in
the model. For the isotropic case, it is non-relevant whether to have a General
Stress State or a Uni-axial Stress State, since we can calculate the exact solution of
the effective material properties of the two independent parameters as a function of
the volume averaged stress and strain tensors as described in Section 2.9.1. While
for the Main Stream Averaging (MSA) Homogenization case, our model must
maintain the uni-axial stress state since the formulation of the material parameters
is a function of the main stresses and strains in the direction of the perturbation
(see Section 2.10.).

The detailed sets of equations satisfying this condition and used in the periodic
boundary condition code
1)
are the following,

Faces of the RVE
Group I:

0
1 1
=
FaceTop FaceBottom
u u
0
2 2 2
=
dummy FaceTop FaceBottom
u u u
0
3 3
=
FaceTop FaceBottom
u u

0
1
Re
1 1
=
dummy ar Face FaceFront
u u u
0
Re
2 2
=
ar Face FaceFront
u u
0
Re
3 3
=
ar Face FaceFront
u u

0
1 1
=
FaceRight FaceLestt
u u
0
2 2
=
FaceRight FaceLestt
u u
0
3 3 3
=
dummy FaceRight FaceLeft
u u u

1)
Boundary Conditions Code: is written in a python script as part of the automated process
done through this work to calculate the effective material properties of the composite material
based on numerical simulations done on its microstructure (RVE).
Edges of the RVE
Group IIa: Group IIb: Group IIc:

0
1
1
1
1
=
EdgeBB EdgeCC
u u
0
1
2
1
2
=
EdgeBB EdgeCC
u u
0
3
1
3
1
3
= +
dummy EdgeBB EdgeCC
u u u

0
1 1
=
EdgeAB EdgeCD
u u
0
2 1
=
EdgeAB EdgeCD
u u
0
3 3 3
= +
dummy EdgeAB EdgeCD
u u u

0
1 1 1
= +
dummy EdgeBC EdgeAD
u u u
0
2 2
= +
EdgeBC EdgeAD
u u
0
3 3
=
EdgeBC EdgeAD
u u

0
1
1
1
1
1
=
dummy EdgeAA EdgeBB
u u u
0
1
2
1
2
=
EdgeAA EdgeBB
u u
0
1
3
1
3
=
EdgeAA EdgeBB
u u

0
1 1
1 1
= +
B EdgeA EdgeAB
u u
0
2
1 1
2 2
= +
dummy B EdgeA EdgeAB
u u u
0
1 1
3 3
= +
B EdgeA EdgeAB
u u

0
1 1
1 1
= +
C EdgeB EdgeBC
u u
0
2
1 1
2 2
= +
dummy C EdgeB EdgeBC
u u u
0
1 1
3 3
=
C EdgeB EdgeBC
u u

0
1
1
1
1
=
EdgeDD EdgeAA
u u
0
1
2
1
2
=
EdgeDD EdgeAA
u u
0
3
1
3
1
3
=
dummy EdgeDD EdgeAA
u u u

0
1 1
1
1 1
1
=
D EdgeC B EdgeA
u u
0
1 1
2
1 1
2
=
D EdgeC B EdgeA
u u
0
3
1 1
3
1 1
3
=
dummy D EdgeC B EdgeA
u u u

0
1
1 1
1
1 1
1
=
dummy D EdgeA C EdgeB
u u u
0
1 1
2
1 1
2
=
D EdgeA C EdgeB
u u
0
1 1
3
1 1
3
=
D EdgeA C EdgeB
u u

Corners of the RVE (Master Nodes)
Group III:

0
1 1
=
C B
u u
0
2 2
=
C B
u u
0
3 3 3
=
dummy C B
u u u

0
1
1 1
=
C C
u u
0
2
1
2 2
= +
dummy C C
u u u
0
1
3 3
=
C C
u u

0
1
1
1
1
=
B C
u u
0
1
2
1
2
=
B C
u u
0
3
1
3
1
3
= +
dummy B C
u u u

0
1 1
=
D A
u u
0
2 2
= +
D A
u u
0
3 3 3
=
dummy D A
u u u

0
1
1 1
=
D D
u u
0
2
1
2 2
= +
dummy D D
u u u
0
1
3 3
=
D D
u u

0
1
1
1
1
=
A D
u u
0
1
2
1
2
=
A D
u u
0
3
1
3
1
3
= +
dummy A D
u u u

0
1
1
1
1
1
= +
dummy B A
u u u
0
1
2
1
2
= +
B A
u u
0
1
3
1
3
=
B A
u u
A simple single centered particle RVE is shown in Fig. 3.3.4. A single dummy node
is used to carry the perturbation necessary to execute the intended simulation. This
process is done through establishing equations between the relative periodic sets of
nodes and inserting the perturbation of the dummy node into these equations. See
Fig. 3.3.5.

Fig. 3.3.4 Single centered-particle RVE

What is actually done, is that in order to equate a certain node set to its
counterpart, the slave and master node-based surfaces must coincide and in a
peer-to-peer nodal sequence, this means same number and position of nodes must
be preserved.

To achieve this task while having a non conform mesh on the periodic regions, a
copy of the selected surface is translated to the opposite surface and an equation
(Abaqus command *Equation) is set between the surface and its own copy, since a
copy guarantees the above mentioned criteria to equate the corresponding sets.
Afterwards, the copy is tied to the opposite surface (Abaqus command *Tie). The
choice of these node sets that are mentioned above is very crucial since the opposite


Fig. 3.3.5 Schematic representation of the dummy element method

faces of the cube have common edges and corners which make a conflict by losing
the degree of freedom if used more than once. This scheme is graphically presented

Fig. 3.3.6 Setting periodic constraint for the top and bottom node-based surfaces

in details in Fig. 3.3.6 for only one pair of node sets FaceBottom and FaceTop.
Similarly, all other corresponding pairs of node sets are equated.

Tying the two node-based surfaces exhibits the following functionalities [47] as it
ties two surfaces together for the duration of a simulation and constrains each of
the nodes on the slave surface to have the same motion and the same value of any
simulation response used on the master surface to which it is closest, it eliminates
the degrees of freedom of the slave surface nodes that are constrained, where
possible.

3.3.4. Examination and Verification

Three types of examinations are done on simple and structured Representative
Volume Elements to verify the above mentioned 3D periodic boundary conditions in
different aspects. Checks of the PBC approach have been done with respect to:
o The periodicity conditions.
o The numerical homogenization approach.

In the verification process, the periodic boundary conditions are applied on a cubic
representative volume element of a unit volume. The RVE has the following
specifications as described in Table 3.3-2,

<< 1 >> Matrix
Isotropic Elastic behavior,
(Young modulus, Poisson's ratio)
(2800 MPa, 0.35)
<< 2 >> Sphere Fillers
(min. 191 Inclusions)
(72000 MPa, 0.20)
o inclusion_shape = ellipsoid
o aspect_ratio = 1
o phase_definition = by_size
o size_distribution = random
o min_inclusion_size = 0.1
o size_rve_x = 1
o size_rve_y = 1
o size_rve_z = 1
o allow_interpenetration = no
o minimum_relative_distance = 0
o max_inclusion_size = 0.4
o clustering = no
o Incl. volume_fraction = 25%
o Periodic boundary conditions
o Number of elements: 111268
o (Embedded)
minimum_relative_vol =1
o (Inclusions allocation)
max_number_of_tests = 400
o tensile perturbation in 3
rd
direction (Z)
Table 3.3-2 Representative Volume Element Specifications

As shown in Fig. 3.3.7, the PBC satisfy both conditions of continuity on the
boundary of the micro-structure. FaceRight and FaceLeft are two opposite faces or
parallel pairs where stress and strain continuity should be satisfied. The color
gradient stands for the von Mises stresses while the peaks and valleys represent the
deformation on the parallel faces.

Fig. 3.3.7 Periodic B.C. Graphical representation of the traction and
displacement continuity on a parallel pair of faces of the RVE
As it follows from the formulation of the PBC, the simulations have predefined
displacement constraints satisfying the displacement continuity condition that
yields the traction continuity condition as shown in Fig. 3.3.7 in the form of von-
Mises stresses calculated at the centroids of the finite elements.

Stress and strain continuity is also quantified by the calculation of the extrapolated
maximum principle stresses and strains on the parallel opposite mapped node-
pairs that belong to the FaceLeft and FaceRight respectively. The values of the
stresses and strains are extrapolated from the respective integration points using
the shape functions of the element or elements sharing the nodes.

The following figures show a statistical distribution of the relative difference of the
nodal extrapolated stresses and strains on the opposite parallel mapped node-pairs
of the FaceLeft and FaceRight respective node sets.

Mises Stresses Continuity - Relative Difference Statistical
Distribution
Diff. < 1%
728 nodes
Diff. < 3%
555 nodes
Diff. < 5
130 nodes
Diff. < 7%
50 nodes
Diff. > 7%
39 nodes
0
100
200
300
400
500
600
700
800
1
N
u
m
b
e
r

o
f

N
o
d
e
s

[
-
]
Difference <1%
Difference <3%
Difference <5%
Difference <7%
Difference >7%
Average Difference
1.642 %

Fig. 3.3.8 Extrapolated Von-Mises Stresses - Relative difference between
opposite parallel node pairs of FaceLeft and FaceRight of the RVE

Stresses Continuity - Relative Difference Statistical Distribution
Diff. < 1%
995 nodes
Diff. < 3%
415 nodes
Diff. < 5%
65 nodes
Diff. < 7%
20 nodes
Diff. > 7%
7 nodes
0
110
220
330
440
550
660
770
880
990
1100
1
N
u
m
b
e
r

o
f

N
o
d
e
s

[
-
]
Difference <1%
Difference <3%
Difference <5%
Difference <7%
Difference >7%
Average Difference
1.025 %

Fig. 3.3.9 Extrapolated max. Principle Stresses - Relative difference between
opposite parallel node pairs on FaceLeft and FaceRight of the RVE

Strains Continuity - Relative Difference Statistical Distribution
Diff. < 1%
824 nodes
Diff. < 3%
502 nodes
Diff. < 5%
114 nodes
Diff. < 7%
39 nodes
Diff. > 7%
23 nodes
0
100
200
300
400
500
600
700
800
900
1
N
u
m
b
e
r

o
f

N
o
d
e
s

[
-
]
Difference <1%
Difference <3%
Difference <5%
Difference <7%
Difference >7%
Average Difference
1.375 %

Fig. 3.3.10 Extrapolated max. Principle Strains - Relative difference between
opposite parallel node pairs on FaceLeft and FaceRight of the RVE
The relative difference is calculated by subtracting the peer-to-peer fields of the
respective opposite parallel entities and dividing them by the minimum of both. The
percentage relative difference is given by

100 *
) , (
+
+

i i
i i
Min
, 3.3-9
where
+

i i
and represents the respective field output (example, von Mises
stresses, max. principle strains,,etc) of the corresponding parallel mapped entities,
in our case the nodes of the LeftFace and RightFace sets.

Fig. 3.3.8 shows the extrapolated von Mises stresses on the opposite nodes with
peer-to-peer comparison. 85.4% of the node-pairs have an difference less than 3%
yielding an overall average relative difference of 1.642%. Fig. 3.3.9 is the same
study but over the extrapolated maximum principle stresses, where 93.9% of the
node-pairs shows a consistency with a difference less than 1% yielding an overall
average relative difference of 1.025%. Similarly, Fig. 3.3.10 relates to the
extrapolated maximum principle strains, where 88.3% of the node-pairs showing an
difference less than 1% yielding an overall average relative difference of 1.375%.
Nevertheless, one should keep in mind that the numerical error due to the
extrapolation from the integration points to the nodes is a major contributor to the
above results. These results show the continuity conditions of stresses and strains
on the boundaries are acceptably fulfilled by the implementation of our 3D periodic
boundary conditions with negligible numerical errors.

Another sample test is made to check whether the PBC code provides the uni-axial
stress state in case of uniaxial perturbation. In Fig. 3.3.11, the deformed and the
undeformed micro-structure is presented, a tensile perturbation is applied, and as
we can see the deformation occurs periodically in the direction of the perturbation.
Undeformed
RVE DeformedRVE
Uni-axial Stress State
No fixations
Undeformed
RVE DeformedRVE
Uni-axial Stress State
No fixations

Fig. 3.3.11 Periodic B.C. Tensile perturbation Uniaxial Stress State

This phenomenon is quantified by measuring the summation of the volume
averaged stresses of all the elements in the x, y and z directions, and keeping in
mind that the representative volume elements have unit volume. The RVE used has
the following specifications:

<< 1 >> Matrix
(2890 MPa, 0.38)
(72000 MPa, 0.20)
o aspect_ratio = 1
o size_distribution = fixed
o clustering = no
o Homogeneous boundary conditions
o Number of elements = 181253
o size_rve_x = 1
o size_rve_y = 1
o size_rve_z = 1
o (Embedded)
o tensile perturbation in 3
rd
direction (Z)
Table: 3.3-3 Representative Volume Element Specifications

In Table: 3.3-4, HBC and PBC are compared to each other on the same RVE.
Theoretically homogeneous boundary conditions yield no stresses other than in the
direction of the perturbation; therefore they result in a uni-axial stress state under
uni-axial loading. The results show that all the principal stresses of the
microstructure are accumulated in the z-direction or the direction of the
perturbation while using the periodic boundary conditions in comparison to the
homogeneous boundary conditions. This property is very much useful when
applying the Main Stream Averaging homogenization (MSA) to our microstructures.
Fig. 3.3.12 is a graphical representation in two dimensions showing an example of
uniaxial stress state. For more details of the constraints applied in the PBC please
go back to the PBC code formulation in Section 3.3.3.

Periodi city with
General Stress
State
Periodi city with
Uniaxi al Stress State
Compression Perturbation
Periodi city with
Meso-scale
Periodi c Array
Periodi city with
General Stress
State
Periodi city with
Compression Perturbation
Periodi city with
Meso-scale
Periodi c Array

Fig. 3.3.12 Periodicity with uni-axial stress state - Graphical representation

Boundary
Conditions
xx

yy

zz

Homogeneous 0.00 MPa. 0.00 MPa. 250.11 MPa.
Periodic 2.20 MPa. 2.18 MPa. 249.50 MPa.
Table: 3.3-4 Uniaxial Stress State Periodic and Homogeneous
Boundary Condition - Summation of stresses in the three
principle directions x, y and z

3.3.4.1. Different Windowing Approach

In this example, it is intended to verify the periodicity condition of a microstructure
distributed in a meso-scale periodic array as shown in (Fig. 3.3.13),

Fig. 3.3.13 Windowing approach, meso-scale periodic array

Consider the above meso-scale periodic array of sphere particles, by taking two
different windows we define two single celled micro-structures, the first is called
cornered single-particle RVE and the other is centered single-particle RVE. The two
micro-structures are generated and computed with the FE software package
Abaqus.

They both have structured geometry and mesh as shown by the cross sections in
(Fig. 3.3.14). The analysis is geometrically linear with a tensile strain load in the
second direction or y-direction, having a value of 0.6% of the cube edge. A three
dimensional solid continuum element is used; it is an 8-node linear brick, reduced
integration with hourglass control. The two microstructures represent the same
composite, i.e. the same volume fraction and material properties of their individual
material components.

Both representative volume elements show traction and displacement continuity on
the boundaries, the periodic boundary conditions are successful in showing the
influence of the inclusions with respect to each other within the meso-scale periodic
array.

MATRIX
E =1.5 GPa
=0.22
INCLUSION
V
f
=44.4%
E =75 GPa
=0.17
Centeredparticle Cornered particle
Max lS
22
l =130.4 MPa
Mises lSl =99.98 MPa
Max lS
22
l =129.1 MPa
Different Windowing
Comparison
MATRIX
E =1.5 GPa
=0.22
INCLUSION
V
f
=44.4%
E =75 GPa
=0.17
Centeredparticle Cornered particle
Max lS
22
l =130.4 MPa
Max lS
22
l =129.1 MPa
Different Windowing
Comparison

Fig. 3.3.14 3D periodicity verification - different windowing approach

In Fig. 3.3.14 and Table: 3.3-5, the maximum local stresses are compared as they
are found in the area of interaction between the repeated inclusions whether in the
centered or cornered RVE, they show a very minor numerical difference due to
different structure of the microstructures and subsequently a minor difference of
the mesh on the boundaries. The 3D periodic boundary condition in debate has put
forward the intended targets to move ahead in applying it to more geometrically
complicated types of microstructures as they will be described in CHAPTER IV and
CHAPTER V.

Maximum Stresses
CORNERED
Single Particle RVE
CENTERED
Single Particle RVE
Difference
(MPa Units) (MPa Units) (%)
*Maximum Stress
22
S =
129.1 130.4 1.01%
*Max. von Mises Stress
Mises
22
S =
99.12 99.98 0.87%

Table: 3.3-5 Max. Stresses Comparison - different windowing approach

In this sample test we also intend to check and compare the results after applying
numerical homogenization for an isotropic case over the two RVEs in Fig. 3.3.14.
The effective material properties (
eff eff eff eff
K E ; ; ; ) are calculated using the
formulation of the small elastic deformation theory for isotropic material described
in Section 2.9.1.

Effective Material Properties
CORNERED
Single Particle RVE
CENTERED
Single Particle RVE
Difference
(MPa Units) (MPa Units) (%)
*Effective Shear Modulus
eff
=
2232.894 2253.333 0.9071%
*Effective Bulk Modulus
eff
K =
2294.737 2300.851 0.2657%
*Effective Elastic Modulus
eff
E =
5058.090 5096.312 0.7500%
*Effective Poisson Ration
eff
=
0.132631 0.130839 1.3700%

Table: 3.3-6 PBC Numerical Stability Comparison

The comparison in Table: 3.3-6 shows the difference in the final effective material
properties of the composite, given that with the application of periodic boundary
conditions to these RVEs should theoretically provide the same results. But as we
mentioned that due to different geometries the meshing have a minor difference
that cannot be avoided especially on the boundaries, this explains the small relative
difference in the homogenized material properties. Nevertheless, it suggests the PBC
method is correctly implemented.

3.3.4.2. Numerical Accuracy

In this sample test we intend to check the numerical accuracy of the whole process
of applying the periodic boundary conditions and homogenization. In order to
achieve this we assumed both phases (Matrix, filler) to have the same material
properties, i.e. we expect to have the MP of the resulting homogenized composite to
be exactly the same as the inputs of its constituents except for a minor error that
enables us to check the numerical accuracy of the method.

Table: 3.3-7 shows a comparison between the resulting effective Elastic modulus of
the structured micro-structures, the centered and the cornered single-particle
RVEs.

MP
CORNERED
Single Particle RVE
CENTERED
Single Particle RVE
Difference
RVE1 & RVE2
O/P (MPa) I/P (MPa) Difference O/P (MPa) I/P (MPa) Difference (MPa Units)
eff
E

75066.4 75000.0 0.0885% 75067.6 75000.0 0.0902% 0.0017%

Table: 3.3-7 PBC Numerical Accuracy Comparison

Similarly, the analysis applied is linear with tensile perturbation of 0.6% strain load
with periodic boundary conditions and both micro structures representing same
material. The input material properties of the matrix and the filler are the same; in
this case both have same elastic modulus of 75000 MPa. The difference between the
output (O/P) effective E-modulus of the homogenized material and the input (I/P)
E-modulus is very small, indicating that the homogenization process is numerically
highly accurate. The difference between the cornered and the centered RVEs shows
an insignificant effect of the different windowing approach used proving the
periodicity is well implemented.

3.4. FEM Modeling Automated Process Anatomy

The flow chart in Fig. 3.4.1 shows in details the path through which the FE-model
passes through. The major steps like Boundary Conditions and the Implemented
Homogenization Methods modules are discussed comprehensively in CHAPTER II
and CHAPTER I Sections (2.9.1 and 2.10). The other steps and levels are discussed
in the statistical tests and results done throughout CHAPTER IV and CHAPTER (V)

This process is used for the statistical studies done in the next coming two chapters
on composite materials with different types of microstructures.

Fig. 3.4.1 Flow chart of FE material modeling

The effective material properties are computed through the numerical
homogenisation implemented in the above flow chart (Fig. 3.4.1), and then they are
compared to Mori-Tanaka and Lielens analytical models. Some of the studies
include verifications and tests for the influences of different important parameter
that are predominant and essential for building up the best representative material
model or microstructure (RVE). It is employed throughout CHAPTER IV and (V).

CHAPTER IV

4. RVEs with spherical inclusions

This chapter includes the modeling of sphere shaped particle reinforced
composite materials and a comprehensive statistical testing in comparison
with semi-analytical models (Mori-Tanakas and Lielens models).

CHAPTER IV - RVEs with spherical inclusions - 81 -
4.1. Types of Microstructures (RVE)

Using the finite element analysis of a periodic/non-periodic cubic RVE of volume
3
L
with randomly distributed spherical particles, we calculate the homogenized effective
elastic constants of the composite. This process and the developed tools will be
presented in the coming sections.

The representative volume element used is taken as a cube with a volume,
. . 1
3
v u L V
edge RVE
= = (unit volume) and an edge length, . . 1 l u L
edge
= (unit length). This
standardization is required because of the wish to implement an automated process
starting from the generation of the RVE up to the calculation of the effective
material properties.

The RVE is generated by the DigiMat [72], a software package that allows the
construction of the geometry before it is exported to Abaqus software for meshing
the microstructure.

Periodic distribution
of identical particles
Non-periodic distribution
Randomdistribution
Randomdistribution
of non-identical
particles
Microstructure Generator
TYPE (3)
TYPE (1) TYPE (2)
TYPE (4)
Embedded
Inclusions
Trimmed
Inclusions
Periodic distribution
Non-periodic distribution
Randomdistribution
Randomdistribution
of non-identical
particles
Microstructure Generator
TYPE (3)
TYPE (1) TYPE (2)
TYPE (4)
Embedded
Inclusions
Trimmed
Inclusions

Fig. 4.1.1 displays the different types of sphere-filled RVEs that are going to be used
in several studies against a series of variable-parameters and boundary conditions
influences. These types are categorized into the following groups:

Fully Embedded spheres (no trimming of inclusions on RVE boundaries)
o Same inclusions size (random distribution in space) - (type-2)
o Different inclusions size in a given range (random distribution in space) -
(type-4)

Edge Trimmed spheres (allow trimming of inclusions on RVE boundaries)
o Periodic Geometry on the boundaries - (type-3)
Same inclusions size (random distribution in space)
o Non-periodic Geometry on the boundaries - (type-1)
Same inclusions size (random distribution in space)*

The meshing of the RVEs is done through Abaqus meshing tools. The boundary
conditions and the numerical homogenization are written in python programing
language [73] as scripts part of the the overall process described in Section 3.4.

4.2. RVE with Embedded Spherical Inclusions

Using this type of RVE with different criteria and boundary conditions, the following
main points of the statistical studies, verifications and comparisons are presented:
Microstructure mesh sensitivity
Influence of the number of fillers on the homogenized material properties
FE-modeling and homogenization with different volume fraction of inclusions
Periodic and homogeneous boundary conditions are compared and studied on the
different types of RVEs.

4.2.1. FE-model Mesh Sensitivity

The RVEs that are used in the analysis are generated with random distribution of
inclusions, thus having different geometries each time the same microstructure is
created. This would make it impossible to generate identical meshes. Identical
meshes for different RVEs are needed in order to properly compare results obtained
from simulations. Every RVE generated is made up of randomly distributed
inclusions; this means if we generate the same RVE twice we would get a different
distribution of inclusions, thus having different meshes. The only variable we
control that provides the mesh size is the seeding size. Seedsize is the size of the
segments that divide a geometrical entity that yield either a finer or a coarser mesh.
For example, a Seedsize of 0.05 means that the geometric entity is divided into an
equal number of segments with the size of 5% of the overall entity size. This means,
as Seedsize decreases the resulting mesh is finer and vice versa.

So the criterion we used for having similar rather than identical meshes for different
RVEs, is the Seedsize. In order to measure the influence of Seedsize on the effective
material properties a default minimum Seedsize value is chosen to be 0.05. A
relative variation or difference is calculated with respect to this value as follows

100 *
) , (
05 . 0
05 . 0
X X Min
X X
s
s
, 4.2-1
where X represents the homogenized or effective material property at a given
Seedsize, in this case, it is the Youngs modulus and Poisson ratio. The superscript
s is the value of the Seedsize.

Periodic BC
50 same size
Embedded
spherical
inclusions
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
0.30%
0.35%
0.40%
0.45%
0.50%
0.04 0.06 0.08 0.1 0.12 0.14 0.16
" Mesh Seedsize" [-]
R
e
l
.

V
a
r
i
a
t
i
o
n

[
%
]
10000
17500
25000
32500
40000
47500
55000
62500
70000
77500
85000
92500
E
l
e
m
e
n
t
s

N
u
m
.

[
-
]
Rel. variation Poisson's Ratio
Rel. variation E-Modulus
Elements Number
Poly. (Elements Number)

Fig. 4.2.1 Effect of mesh refinement (seedSize) on: a. the effective elastic
modulus and Poisson ratio of the composite. b. number of finite elements

Seedsize is inversely proportional to the number of elements in the model; this
means lower seed size results of a finer mesh. In Fig. 4.2.1 the relative variation of
the homogenized E-modulus and Poisson ratio is plotted against the Seedsize used
on an RVE consisting of 50 embedded randomly distributed identical spheres with
periodic boundary conditions. The analysis is linear elastic and the composite is
assumed isotropic. Linear perturbation on the RVE is applied in the X, Y, and Z
directions and the resulting material parameters from the homogenization of the
three simulations are averaged. This procedure is followed throughout the
homogenization process in the case of isotropy for all types of RVEs. It is used in
order to dilute the non-structural geometry of the microstructures, for details and
formulations see Section 2.9.1.

It is noticeable that at a Seedsize of 0.08 the relative variation of the E-modulus is
0.02% and the variation slope is minimal and constant below this value, while for
Seedsize greater than 0.08, the variation slope accelerates, thus making the
influence of the mesh more critical. Upon this study it is concluded that an
optimum Seedsize should be less than 0.08 and a value of 0.05 is chosen as a
default value to be used for all our FE-models unless otherwise mentioned.

Mesh Sensitivity (element size)
0.E+00
2.E-06
4.E-06
6.E-06
8.E-06
1.E-05
1.E-05
0 10 20 30 40 50 60
Number of Particles [-]
A
v
e
r
a
g
e

E
l
e
m
e
n
t

V
o
l
u
m
e
(
u
n
i
t

v
o
l
u
m
e
)
Average Element size (80%)

Fig. 4.2.2 Mesh sensitivity criterion for different geometries with respect to
the number of inclusions, taken as the average volume of the largest
80% of all the elements of the microstructure (Same Vf and seedSize)

After the choice of 0.05 mesh seed side, another test is done to check the meshing
sensitivity of the FE-models. We tried to find a criterion to compare the NON-
IDENTICAL meshes we obtain for the same RVE being generated multiple times,
since the randomly distributed inclusions create different geometries i.e. different
meshes. This criterion is based on distinguishing the refinement of the meshes of
the analyzed FE-models. It states that we chose the largest 80% (by volume) of the
finite elements of each RVE and we calculate their average volume. This means that
as the mesh is finer this average volume should decrease and vice versa. Fig. 4.2.2
35
Negligible Effect
shows a study done on a group of 15 RVEs of the same material and inclusion
volume fraction = 15% and a fixed Seedsize = 0.05 with randomly distributed
spherical inclusions. They are statistically distributed as 3 RVEs on each number of
particles interval from 10 to 50 particles per microstructure. The data show that
above 35 particles the effect of mesh refinement becomes asymptotically smaller
and will be neglected in the sequel. Consequently, with a fixed Seedsize and a high
number of inclusions (>35) the microstructures become less mesh sensitive even
with their dissimilar geometries resulting from the random distribution of
inclusions.

4.2.2. Influence of Number of Inclusions

The RVEs used in this study are generated with random distribution of totally
embedded same-size inclusions of the same material and having the same volume
fraction but different numbers of spheres. The plots presented in this section are
described in a way that for each data point plotted on the resulting graphs
corresponds to 9 simulations (or 9 FE-models), this means at each point 3
statistically different RVEs are generated having same input data except for the
random distribution of their inclusions. Each RVE is simulated with 3 tensile
perturbations in X, Y, and Z directions. Given that the microstructure represents an
isotropic material, an average of the resulting effective MP is taken to the plot.
Repeating this homogenization process on the other remaining two RVEs will make
3 average data points, afterwards we average these points to get the final
homogenized data point called FE-Average. Statistically speaking, each data point
represents therefore an averaging of 9 models. The particles volume fraction used is
15% and the input material properties for the two phases are the following,

Matrix: 35 . 0 2800 = =
m m
and MPa E
Inclusion (Quartz): 2 . 0 72000 = =
I I
and MPa E

4.2.2.1. Using Periodic Boundary Conditions

In Fig. 4.2.3 the Mori-Tanaka and Lielens material models are compared with the
FE-model having periodic boundary conditions, for details see Section 3.3.

3700
3775
3850
3925
4000
4075
4150
0 10 20 30 40 50 60 70
E
-
M
o
d
u
l
u
s

[
M
P
a
]
Number of Spheri cal Incl usi ons [-]
FE Average Mori-Tanaka
Lielens Average 1
Average 2 Average 3
FE Average (trend line)
Periodic Boundary Conditions
Em=2800MPa
m=0,35
Ep =72000MPa
p =0,2
Same size
Embeddedspheres
Seedsize=0.05
Vf=15%

1350
1380
1410
1440
1470
1500
1530
1560
1590
1620
0 10 20 30 40 50 60 70
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Lielens Average 1
Average 2 Average 3
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2
Same size
Embedded spheres
Seedsize =0.05
Vf =15%

3770
3790
3810
3830
3850
3870
3890
3910
3930
0 10 20 30 40 50 60 70
B
u
l
k

M
o
d
u
l
u
s

[
M
P
a
]
Mori-Tanaka Lielens
Average 1 Average 2
Average 3 FE Average
Same size
Embedded spheres
Seedsize =0.05
Vf =15%
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2

0,3230
0,3280
0,3330
0,3380
0,3430
0,3480
0,3530
0 20 40 60 80
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Lielens Average 1
Average 2 Average 3
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2
Same size
Embedded spheres
Seedsize =0.05
Vf =15%

Fig. 4.2.3 Influence of the number of inclusions on an FE-model.
Statistical results and comparison of the effective MP of isotropic
composite (
eff eff eff eff
and K E , , , ) with respect to Mori-Tanaka and
Lielens material models using periodic boundary conditions.

The isotropic material properties are statistically calculated using a numerical
homogenization for the respective RVEs with different number of spheres with a
range from 10 to 60 particles, for details on the numerical homogenization and the
whole process see Section 2.9.1 and Section 3.4. As it is noticed, the FE-model is
stiffer than both analytical ones.

As the number of particles increases the FE-model becomes more statistically stable
and as it contains more than 35 inclusions the influence of the inclusions number
becomes negligible.

4.2.2.2. Using Homogeneous Boundary Conditions


3700
3800
3900
4000
4100
4200
4300
0 10 20 30 40 50 60 70
E
-
M
o
d
u
l
u
s

[
M
P
a
]
Lielens Average 1
Average 2 Average 3
Homogeneous Boundary Conditions
Same size
Embedded spheres
Seedsize =0.05
Vf =15%
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2

1350
1380
1410
1440
1470
1500
1530
1560
1590
1620
0 10 20 30 40 50 60 70
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Lielens Average 1
Average 2 Average 3
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2
Same size
Embedded spheres
Seedsize =0.05
Vf =15%

1350
1380
1410
1440
1470
1500
1530
1560
1590
1620
0 10 20 30 40 50 60 70
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Lielens Average 1
Average 2 Average 3
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2
Same size
Embedded spheres
Seedsize =0.05
Vf =15%

0,3230
0,3280
0,3330
0,3380
0,3430
0,3480
0,3530
0 20 40 60 80
P
o
i
s
s
o
n

R
a
t
i
o
n

[
M
P
a
]
Lielens Average 1
Average 2 Average 3
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2
Same size
Embedded spheres
Seedsize =0.05
Vf =15%

composite (
eff eff eff eff
Lielens material models using homogeneous boundary conditions.

FE-model having homogeneous periodic boundary conditions, for details see Section
3.2. The homogenization process is applied to calculate the isotropic effective
material properties of the FE-models. Similar trend of results is obtained in
comparison to the periodic boundary conditions.
So as a conclusion, with higher number of inclusions (>35) the statistical scattering
of results is minimized. Also, no matter what boundary conditions are used, the FE-
models are stiffer than MT and Lielens models and the minimum number of
inclusions to have a reliable and representative volume element is taken through
out this study to be 35 inclusions. Lielens model shows, in both B.C. cases, a better
overall result than the MT model in comparison to the FE-model. Regarding the
Bulk modulus, it shows that the FE-model lies between both but closer to Mori-
Tanaka. It is still not a persuasive result since the bounding range for all three
models is very small nearly 1.4%, this leaves us with a conclusion that the Bulk
modulus graph has no significance in regards to the models comparison. The
Poisson ratio for the two analytical models is nearly the same.

4.2.2.3. Comparison between Homogeneous and

In Fig. 4.2.5 a comparison between the HBC and PBC for a totally embedded
particles RVEs shows stiffer results when using the HBC. This is due to the higher
constraint model when using the homogeneous boundary conditions which shows
that the type of boundary conditions along with the type of RVE, plays a crucial role
in determining the mechanical response of the microstructure. Further studies will
be conducted in the following sections to oversee the compatibility of boundary
conditions and other types of RVEs

3700
3750
3800
3850
3900
3950
4000
4050
4100
4150
0 10 20 30 40 50 60 70
E
-
M
o
d
u
l
u
s

[
M
P
a
]
PBC (FE-average) HBC (FE-average)
Lielens Mori-Tanaka
Same size
Embedded spheres
Seedsize =0.05
Vf =15%
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2

1350
1400
1450
1500
1550
1600
0 10 20 30 40 50 60 70
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
PBC (FE-average) PresBC (FE-average)
Lielens Mori-Tanaka
Same size
Embedded spheres
Seedsize =0.05
Vf =15%
Em=2800 MPa
m=0,35
Ep =72000 MPa
p =0,2

3770
3790
3810
3830
3850
3870
3890
3910
3930
0 10 20 30 40 50 60 70
B
u
l
k

M
o
d
u
l
u
s

[
M
P
a
]
PBC (FE-average) PresBC (FE-average)
Lielens Mori-Tanaka

0,3230
0,3280
0,3330
0,3380
0,3430
0,3480
0 20 40 60 80
P
o
i
s
s
o
n

R
a
t
i
o
n

[
M
P
a
]
PBC (FE-average) HBC (FE-average)
Lielens Mori-Tanaka

composite (
eff eff eff eff
and K E , , , ) between Homogeneous and Periodic
boundary conditions.

4.2.3. FE-modeling with Different Volume Fraction

The RVEs used in this study are generated with random distribution of totally
embedded different size spherical inclusions ranging in diameter size (0.1<d<0.4) or
between 10% and 40% of the RVE side length. Multi data point graphs are
generated with different volume fractions and compared to the FE-model. All RVEs
contain a minimum of 50 up to 200 particles and a seed size = 0.05. The same
material properties and statistical approach is considered as described in Section
4.2.2 but in this case four different RVEs are generated at every volume fraction
with X, Y, and Z perturbations making up to 12 simulations per data point. Periodic
and homogeneous boundary conditions are applied and results are discussed as
shown in Fig. 4.2.6 and Fig. 4.2.7.
4.2.3.1. Using Periodic Boundary Conditions

FE-model having periodic boundary conditions, for details see Section 3.3. The
isotropic material properties are statistically calculated as discussed in Section
2.9.1 using a numerical homogenization for the respective RVEs with different
volume fractions. At lower volume fractions a very small difference is observed
between the three models, while as the volume fraction increases a larger difference
is seen especially with Mori-Tanakas model. Overall, the FE-model shows a slightly
stiffer result than the two analytical models. Lielens model shows the closest
results to our FE-model in the main three parameters
eff eff eff
and E , , . For example

3200
3700
4200
4700
5200
5700
5 10 15 20 25 30
Inclusion's Volume Fraction [%]
E
-

M
o
d
u
l
u
s

[
M
P
a
]
Average 1
Average 2
Average 3
Average 4
FE Average
Lielens
Mori-Tanaka
Poly. (FE Average)
Inclusions =spheres
B.C. =Periodic
Em =2800 MPa
m =0,35
Ep =72000 MPa
p =0,2
Embedded RVE
0.1 <Incl. Size <0.4
Periodic B.C.
Minimumof
50 upto 200 Inclusions
Seedsize =0.05

1100
1300
1500
1700
1900
2100
2300
5 10 15 20 25 30
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Average 1
Average 2
Average 3
Average 4
FE Average
Lielens
Mori-Tanaka
Poly. (FE Average)
Inclusions =spheres
B.C. =Periodic
Em =2800 MPa
m =0,35
Ep =72000 MPa
p =0,2
Embedded RVE
Periodic B.C.
Minimumof
Seedsize =0.05

3400
3600
3800
4000
4200
4400
4600
4800
5000
5 10 15 20 25 30
B
u
l
k

M
o
d
u
l
u
s

[
M
P
a
]
Average 1
Average 2
Average 3
Average 4
FE Average
Lielens
Mori-Tanaka
Poly. (FE Average) Inclusions =spheres
B.C. =Periodic
Em =2800 MPa
m =0,35
Ep =72000 MPa
p =0,2
Embedded RVE
Periodic B.C.
Minimumof
Seedsize =0.05

0.305
0.310
0.315
0.320
0.325
0.330
0.335
0.340
0.345
5 10 15 20 25 30
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Average 1
Average 2
Average 3
Average 4
FE Average
Lielens
Mori-Tanaka
Poly. (FE Average)
Inclusions =spheres
B.C. =Periodic
Em =2800 MPa
m =0,35
Ep =72000 MPa
p =0,2
Embedded RVE
Periodic B.C.
Minimumof
Seedsize =0.05

Fig. 4.2.6 FE-modeling with different inclusions volume fraction.
composite (
eff eff eff eff
Lielens material models using periodic boundary conditions

at 23.75% volume fraction the E-modulus is, 4761 MPa. (FE-Model), 4715 MPa.
(Lielens), and 4443 MPa. (MT). For the Bulk modulus, as discussed previously in
Section 4.2.2.2, it shows similar results, where Mori-Tanaka is closer to the FE-
model, but the difference between the three models is rather small.

4.2.3.2. Using Homogeneous Boundary Conditions

3200
3700
4200
4700
5200
5700
5 10 15 20 25 30
E
-

M
o
d
u
l
u
s

[
M
P
a
]
Average 1
Average 2
Average 3
Average 4
FE-Average
Lielens
Mori-Tanaka
Poly. (FE-Average)
Inclusion =spheres
Em =2800 MPa
m =0,35
Ef =72000
f =0,2
Embedded RVE
Homogeneous BC
Minimumof
Seedsize =0.05

1100
1300
1500
1700
1900
2100
2300
5 10 15 20 25 30
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Average 1
Average 2
Average 3
Average 4
FE-Average
Lielens
Mori-Tanaka
Poly. (FE-Average)
Inclusion =spheres
Em =2800 MPa
m =0,35
Ef =72000
f =0,2
Embedded RVE
Homogeneous BC
Minimumof
Seedsize =0.05

3400
3600
3800
4000
4200
4400
4600
4800
5000
5 10 15 20 25 30
B
u
l
k

M
o
d
u
l
u
s

[
M
P
a
]
Lielens
Average 1
Average 2
Average 3
Average 4
Mori-Tanaka
FE-Average
Poly. (FE-Average) Embedded RVE
Homogeneous BC
Minimumof
Seedsize =0.05
Inclusion =spheres
Em =2800 MPa
m =0,35
Ef =72000
f =0,2

0.300
0.305
0.310
0.315
0.320
0.325
0.330
0.335
0.340
0.345
0.350
5 10 15 20 25 30 35
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Average 1
Average 2
Average 3
Average 4
FE-Average
Lielens
Mori-Tanaka
Poly. (FE-Average)
Embedded RVE
Prescribed BC
Minimumof
Seedsize =0.05
Inclusion =spheres
Em =2800 MPa
m =0,35
Ef =72000, f =0,2

composite (
eff eff eff eff
Lielens material models using homogeneous boundary conditions

In Fig. 4.2.7 results show that in general also HBC gives stiffer results than the
PBC. This is plausible due to the more constraint conditions. As discussed in the
previous results with periodic boundary conditions, the three models yield
approximately the same results at low volume fractions (<10%,15%). At higher
volume fractions they slightly differ, but overall, Lielens model shows better results
in comparison to the FE-model, giving the way to conclude that regardless of the
boundary conditions, Lielens model is the most suitable among the considered
analytical models.

4.3. RVEs with Non-fully-embedded Inclusions

4.3.1. Method

Other types of RVE are those ones generated with edge-trimmed inclusions as
described in Section 4.1. In this study we used only one RVE at each volume
fraction but by applying the X, Y, and Z perturbations making three simulation
models for every data point. This will allow us to see how much statistical
dispersion one gets for an FE-model of an isotropic composite with minimum of 50
particles per RVE if one used the X, Y, and Z perturbations. In this study we
compared which type of boundary conditions (HBC or PBC) suits which type of RVE
whether its periodic geometry or non-periodic geometry on the boundaries (RVE
type-1 and 3). It includes the following points:

Periodic RVE with Periodic and Homogeneous Boundary Conditions
Non-periodic RVE with Periodic and Homogeneous Boundary Conditions

A 40% inclusion volume fraction is reached. The matrix/filler properties are:
Matrix: 22 . 0 1500 = =
m m
and MPa E
Inclusion: 17 . 0 75000 = =
I I
and MPa E

P.S.: In the following graphs of Youngs and shear moduli, Voigt upper bound is not
visible and out of range because the composite used has a very high stiffness
contrast between its matrix and the fillers.

PBC - periodic RVE
1600
2000
2400
2800
3200
3600
4000
4400
10 15 20 25 30 35 40 45
Volume fraction [%]
E
-
M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
FE Model (Average X,Y,Z)
Poly. (FE Model (Average X,Y,Z))
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

PBC non-periodic RVE
1600
2000
2400
2800
3200
3600
4000
4400
10 15 20 25 30 35 40 45
Volume fraction [%]
E
-
M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

PBC - periodic RVE
600
800
1000
1200
1400
1600
1800
10 15 20 25 30 35 40 45
Volume fraction [%]
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

600
800
1000
1200
1400
1600
1800
2000
10 15 20 25 30 35 40 45
Volume fraction [%]
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

PBC - periodic RVE
0.160
0.178
0.195
0.213
0.230
10 20 30 40
Volume fraction [%]
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

0.160
0.178
0.195
0.213
0.230
10 20 30 40
Volume fraction [%]
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

composite (
eff eff eff
and E , , ) with respect to Mori-Tanaka, Lielens
material models and the upper and lower bounds of Reuss and Voigt
respectively using Periodic boundary conditions


HBC - periodic RVE
1600
2000
2400
2800
3200
3600
4000
4400
10 15 20 25 30 35 40 45
Volume fraction [%]
E
-
M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Expon. (FE Model (Average X,Y,Z))
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

HBC non-periodic RVE
1600
2000
2400
2800
3200
3600
4000
4400
10 15 20 25 30 35 40 45
Volume fraction [%]
E
-
M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
nm=0,22
Ef=75000MPa
nf=0,17

HBC - periodic RVE
600
800
1000
1200
1400
1600
1800
10 15 20 25 30 35 40 45
Volume fraction [%]
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Expon. (FE Model (Average X,Y,Z))
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

600
800
1000
1200
1400
1600
1800
10 15 20 25 30 35 40 45
Volume fraction [%]
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

HBC - periodic RVE
0.160
0.178
0.195
0.213
0.230
10 20 30 40
Volume fraction [%]
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Double Inclusion
Mori Tanaka
Voigt
Reuss

Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

0.160
0.178
0.195
0.213
0.230
10 20 30 40
Volume fraction [%]
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Double Inclusion
Mori Tanaka
Voigt
Reuss
Em=1500MPa
m=0,22
Ef=75000MPa
f=0,17

composite (
eff eff eff
and E , , ) with respect to Mori-Tanaka, Lielens
material models and the upper and lower bounds of Reuss and Voigt
respectively using Homogeneous boundary conditions

Periodic boundary conditions are used on two types of RVEs, the first set with
periodic geometry and the second set with non-periodic geometry. Knowing a priore
that PBC is suitable only to periodic microstructures, either the inclusions are
edge-trimmed or fully embedded. So the results of the other set with non-periodic
microstructures are shown for emphasis. The effective material properties of the
FE-model are plotted with the Voigt-Reuss bounds, Mori-Tanakas and Lielens
model. It is clear that the FE-model lies within the bounds and that Lielens model
shows the closest fit at higher volume fraction, and at low volume fractions both
Lielens and MT are suitable. Similarly, in Fig. 4.3.2 results show that Lielens
model coincides with the FE-model of HBC with non-periodic microstructure with
minimal data dispersion.

It is concluded that the use of PBC is suitable on periodic microstructures and HBC
on non-periodic microstructures, while the former generally yields softer results
than the latter. More statistical studies are done to confirm the results and
comparisons with the analytical models by developing more RVEs at each data
point as it will be shown in the next sections.

4.3.2. RVEs with Periodic Geometry (edge-trimmed
inclusions)

We will now show the relevant comparison results of the RVEs with periodic
geometry and periodic boundary conditions with Mori-Tanaka and Lielens analytical
models. The data represent 3 different RVEs generated at each volume fraction with
3 different perturbations for each microstructure. The legends RVE1, RVE2, and
RVE3 refer to the average of the results of 3 models with different tensile
perturbations in different directions. The legend FE-average is the total average
value of the homogenization results of 9 simulations at each data point.

Results show that at higher volume fractions (up to 30% in this case), DI or Lielens
model yields the best results for the Poisson ratio, Youngs and shear moduli while
both MT and DI are suitable at lower volume fractions.

The dispersion or statistical scatter is minimal (maximum of 3%) in comparison
with the initial study in Fig. 4.3.1 that is done on a single RVE with three results at
each data point.

2750
3250
3750
4250
4750
5250
5750
0 5 10 15 20 25 30 35
E
-

M
o
d
u
l
u
s

[
M
P
a
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average)
Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2
Edge-trimmed
same-size inclusions
Periodic geometry
Periodic BC
Seedsize =0.05

1000
1200
1400
1600
1800
2000
2200
0 5 10 15 20 25 30 35
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average)
Edge-trimmed
Periodic geometry
Periodic BC
Seedsize =0.05
Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2

3000
3500
4000
4500
5000
5500
0 5 10 15 20 25 30 35
B
u
l
k

M
o
d
u
l
u
s

[
M
P
a
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average)
Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2
Edge-trimmed
Periodic geometry
Periodic BC
Seedsize =0.05

0.305
0.310
0.315
0.320
0.325
0.330
0.335
0.340
0.345
0.350
0 5 10 15 20 25 30 35
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average)
Edge-trimmed
Periodic geometry
Periodic BC
Seedsize =0.05
Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2

composite (
eff eff eff eff
and K E , , ) with respect to Mori-Tanaka, Lielens
material models using periodic geometry with periodic boundary
conditions

4.3.3. RVEs with non-Periodic Geometry (edge-trimmed
inclusions)

The comparison results of the RVEs with non-periodic geometry and homogeneous
boundary conditions are displayed in Fig. 4.3.4. the same statistical procedure is
followed with 9 simulations on each volume fraction data point, the homogenization
results of the Poisson ratio, Youngs and shear moduli show that DI or Lielens
model is the best fit model for all volume fractions (up to 30% in this study), while
MT is for volume fractions <10%.

2750
3250
3750
4250
4750
5250
5750
0 5 10 15 20 25 30 35
E
-

M
o
d
u
l
u
s

[
M
P
a
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average) Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2
Edge-trimmed
non-periodic geometry
Homogeneous BC
Seedsize =0.05

1000
1200
1400
1600
1800
2000
2200
0 5 10 15 20 25 30 35
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average)
Edge-trimmed
Homogeneous BC
Seedsize =0.05
Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2

3000
3500
4000
4500
5000
5500
0 5 10 15 20 25 30 35
B
u
l
k

M
o
d
u
l
u
s

[
M
P
a
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average)
Edge-trimmed
Homogeneous BC
Seedsize =0.05
Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2

0.305
0.310
0.315
0.320
0.325
0.330
0.335
0.340
0.345
0.350
0 5 10 15 20 25 30 35
P
o
i
s
s
o
n

R
a
t
i
o

M
o
d
u
l
u
s

[
-
]
RVE1
RVE2
RVE3
FE-average
Lielens
Mori-Tanaka
Poly. (FE-average)
Edge-trimmed
Homogeneous BC
Seedsize =0.05
Inclusion =spheres
Em =2800 MPa
m=0,35
Ef =72000 MPa
f =0,2

composite (
eff eff eff eff
and K E , , ) with respect to Mori-Tanaka, Lielens
material models using non-periodic geometry with homogeneous
boundary conditions

These results show the same trend as for periodic microstructures with periodic
boundary conditions in the previous section but with stiffer results as it is already
discussed in Section 4.2.2.3 and shown in Fig. 4.2.5 . If looking at both studies the
periodic microstructures with PBC and non-periodic microstructures with HBC, we
find that DI or Lielens model fits better the latest.

4.4. FE-model, Analytical models and Measurements

In this section an FE model with spherical inclusions is generated and analyzed to
calculate the homogenized effective material properties and compared to the
analytical models of Mori-Tanaka and Lielens and with experimental measurements
of an epoxy resin composite material. The MP measurements of the individual
components of the composite are listed in Table: 4.4-1, while the FE-model
specifications are described in Table: 4.4-2.

Quartz E-Modul [MPa]
[]
Filler - - - 72000 - - - 0.20
Duroplasts
epoxy resin
T = 23C, v =1mm/min
E-Modul [MPa]
[]
Test 1 Test 2 Test 3 Average Test 1 Test 2 Test 3 Average
Unfilled - 2910 2870 2890 - 0.38 (0,38) 0.38
Duroplasts
epoxy resin
T = 23C, v =1mm/min
E-Modul [MPa]
[]
Test 1 Test 2 Test 3 Average Test 1 Test 2 Test 3 Average
Filled
(17% Incl.) 4310 4230 4430 4323 0.355 0.36 0.355 0.357
Table: 4.4-1 Material properties measurements of the unfilled and sphere
filled composite material, epoxy-resin.

<< 1 >> Matrix
(2890 MPa, 0.38)
(72000 MPa, 0.20)
o aspect_ratio = 1
o clustering = no
o 179000<Number of elements: <182000
o size_rve_x = 1
o size_rve_y = 1
o size_rve_z = 1
o (Embedded Inclusions)
o tensile perturbation in three

directions
Table: 4.4-2 Specifications of the RVEs used in the FE modeling in
comparison to Mori-Tanaka and Lielens models and the
experimental measurements of a particle-filled composite
This study is performed on three RVEs with similar specifications but the only
difference is the random distribution of the inclusions. The spherical inclusions are
totally embedded and of same size. Tensile perturbations in the three directions are
applied to each RVE with two types of boundary conditions, the periodic and the
homogeneous ones. Table: 4.4-3 to Table: 4.4-9 display the statistical results of
the numerical homogenization of the FE model with periodic and homogeneous
boundary conditions. The analysis is linear and the composite material has an
isotropic behavior. As it is mentioned in previous sections, the statistical approach
in this study is achieved by applying tensile perturbations in the three directions
and calculating the effective material properties for each case. Three RVEs with
random distribution of inclusions are generated, so this means the average of 9 FE-
models of the isotropic composite is calculated.

The tables below show the average deviation percentage between individual cases.
Average deviation is the average of the absolute deviations of data points from their
mean value. It is a measure of the variability in a data set. The average deviation
percentage is calculated using the following expression,

100 *
1
M
M M
n
i

= deviation Average ,
and
=
i
M
n
M
1
,
4.4-1
i
M represents the value of the material parameter of data point n i ,..., 2 , 1 = , where
n is the number of data points that are used to calculate their average deviation;
M is the material parameter average or mean value of all the data points in study.

From Table: 4.4-3 until Table: 4.4-6, periodic boundary conditions are used for the
FE-model. In Table: 4.4-3 the Youngs modulus results show a maximum average
deviation value of 0.15% between the different perturbations in X, Y, and Z
directions of the same RVE, a maximum average deviation value of 0.22% between
different RVEs, and the overall average deviation did not exceed the 0.06%.
Similarly, for the other material parameters, the overall average deviation did not
exceed 0.23%. Such results show that the FE -model with periodic boundary
conditions is statistically reliable.

Elasti c Modulus [MPa] - PBC
X Y Z Average
RVE1 4150 4140 4132 4141 0.14%
RVE2 4148 4152 4162 4154 0.13%
RVE3 4167 4159 4177 4168 0.15%
FE-average 4155 4150 4157 4154 0.06%
0.20% 0.17% 0.40% 0.22%
Average
Deviation
Table: 4.4-3 Finite element modeling with PBC - Statistical results
for the Youngs modulus

Shear Modulus [MPa] - PBC
X Y Z Average
RVE1 1526 1522 1518 1522 0.17%
RVE2 1525 1527 1530 1527 0.12%
RVE3 1533 1529 1536 1533 0.14%
FE-Average 1528 1526 1528 1527 0.05%
0.22% 0.18% 0.43% 0.24%
Average
Deviation
for the Shears modulus

Bulk Modulus [MPa] - PBC
X Y Z Average
RVE1 4938 4935 4960 4944 0.22%
RVE2 4940 4935 4965 4946 0.24%
RVE3 4942 4942 4967 4950 0.22%
FE-average 4940 4937 4964 4947 0.23%
0.03% 0.06% 0.05% 0.04%
Average
Deviation
for the Bulks modulus

Poisson Ratio - PBC
X Y Z Average
RVE1 0.3599 0.3602 0.3612 0.3604 0.14%
RVE2 0.3600 0.3598 0.3603 0.3600 0.05%
RVE3 0.3594 0.3597 0.3598 0.3597 0.04%
FE-average 0.3598 0.3599 0.3604 0.3600 0.07%
0.07% 0.05% 0.14% 0.07%
Average
Deviation
for the Poisson ratio number

From Table: 4.4-7 until Table: 4.4-10, homogeneous boundary conditions are used
for the FE-model. In Table: 4.4-7 the Youngs modulus results show a maximum
average deviation value of 0.25% between the different perturbations in X, Y, and Z
directions of the same RVE, a maximum average deviation value of 0.28% between
different RVEs, and the overall average deviation did not exceed the 0.07%.
Similarly, for the other material parameters, the overall average deviation is
negligible and did not exceed 0.27%. Such results show that the FE -model with
homogeneous boundary conditions is also statistically reliable.

Elasti c Modulus [MPa] - HBC
X Y Z Average
RVE1 4192 4182 4169 4181 0.20%
RVE2 4194 4196 4211 4201 0.16%
RVE3 4205 4217 4234 4215 0.25%
FE-average 4197 4199 4205 4200 0.07%
0.12% 0.29% 0.57% 0.28%
Average
Deviation
Table: 4.4-7 Finite element modeling with HBC - Statistical results
for the Youngs modulus

Shear Modulus [MPa] - HBC
X Y Z Average
RVE1 1543 1539 1533 1538 0.23%
RVE2 1544 1545 1549 1546 0.15%
RVE3 1548 1552 1559 1551 0.23%
FE-average 1545 1545 1547 1546 0.05%
0.15% 0.30% 0.61% 0.30%
Average
Deviation
for the Shears modulus

Bulk Modulus [MPa] - HBC
X Y Z Average
RVE1 4945 4941 4959 4948 0.15%
RVE2 4948 4933 4974 4952 0.30%
RVE3 4926 4959 4982 4962 0.40%
FE-average 4940 4944 4972 4952 0.27%
0.19% 0.20% 0.17% 0.11%
Average
Deviation
for the Bulks modulus

Poisson Ratio - HBC
X Y Z Average
RVE1 0.3587 0.3589 0.3599 0.3592 0.13%
RVE2 0.3587 0.3582 0.3589 0.3586 0.07%
RVE3 0.3577 0.3583 0.3584 0.3584 0.07%
FE-average 0.3584 0.3585 0.3590 0.3586 0.08%
0.12% 0.08% 0.16% 0.08%
Average
Deviation
for the Poisson ratio number

The homogenized effective material properties, as shown in details in the previous
statistical data tables, are calculated and compared to Mori-Tanakas model, the
interpolative double inclusion model (Lielens model) and the experimental
measurements as shown in Fig. 4.4.1.

Exp.
4323
FE - HBC
4200
FE - PBC
4154
DI
4142
MT
4038
3500
3600
3700
3800
3900
4000
4100
4200
4300
4400
4500
1
E
-

M
o
d
u
l
u
s

[
M
P
a
]
Exp.
FE-HBC
FE - PBC
DI
MT

Exp.
1593
FE - HBC
1546
FE - PBC
1527
DI
1520
MT
1480
1300
1350
1400
1450
1500
1550
1600
1650
1
S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Exp.
FE-HBC
FE - PBC
DI
MT

Exp.
5038
FE - HBC
4952
FE - PBC
4947
DI
5023
MT
4953
4700
4750
4800
4850
4900
4950
5000
5050
5100
1
B
u
l
k

M
o
d
u
l
u
s

[
M
P
a
]
Exp.
FE-HBC
FE - PBC
DI
MT

Exp.
0.3570
FE - HBC
0.3586
FE - PBC
0.3600
DI
0.3626
MT
0.3641
0.3460
0.3480
0.3500
0.3520
0.3540
0.3560
0.3580
0.3600
0.3620
0.3640
0.3660
1
P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Exp.
FE-HBC
FE - PBC
DI
MT

Fig. 4.4.1 Comparison between measurements, analytical models (Mori-
Tanaka and Lielens/interpolative Double Inclusion models) and the
FE-model.

The graphs in Fig. 4.4.1 show that the FE-model with homogeneous boundary
conditions gives the closest result although the FE-model with PBC and Lielens
model are of not much difference for this material. The experimental measurements
are the stiffest as shown in Youngs modulus graph.

These results shows the real need to have a micromechanical material model that
helps us identify the best suitable analytical model for a range of data, especially
the inclusions volume fraction. Other benefits can be extracted from such FE-
models, now we have a representative microstructure of our material that enables
us to include any phenomenon at the micro level within the composite, and to do
relevant analysis and simulations in order to predict material behavior and
response. For instance, micro cracking can be included, or micromechanical
damage models, in order to observe the effect of those phenomena on the
composites material properties. Also studies on the composites life time can be
carried out knowing that damage and creep experiments are usually expensive.

These facts are the basic motivation of this work, so that further built-up can be
done in many other aspects based on the micromechanical FE-model and the
methodology that is being developed and verified throughout this work.

CHAPTER V

5. RVEs for fiber reinforced composites

This chapter includes the modeling of fiber reinforced composite materials and
a comprehensive statistical testing in comparison with the existing semi-
analytical models (Mori-Tanakas and Lielens models).

CHAPTER V - RVEs for fiber reinforced composites - 107 -
5.1. Types of Microstructures (RVE)

The types of RVEs used in this chapter represent the microstructure of three types
of fiber reinforced composite material. As it is known that the overall stiffness of the
composite is strongly influenced by the orientation of the fiber inclusions
themselves, RVEs with different fiber orientation will be analyzed. The fiber
orientation defines each RVE type as shown in Fig. 5.1.1,
Fully-aligned (Type 1)
Partially-aligned (Type 2)
General orientation (single and multi-layered) (Type 3)

Fig. 5.1.1 Types of RVEs representing a fiber-reinforced composite material

This chapter focuses on the effect of number of inclusions on the homogenized
material properties and what is the minimum number of inclusions that provides a
volume element representative of the composite. It will also deal with the effect of
fiber orientation on the effective material properties. The results of the numerical
homogenization will be compared to the analytical models of Mori-Tanaka and
Lielen.

5.2. RVEs with Fully Aligned Fiber Inclusions (edge-
trimmed)

In this section, RVE Type (1) as displayed in Fig. 5.1.1, a fully aligned fiber
reinforced FE-model, is generated and statistically examined and compared versus
the MT (Mori-Tanaka) and DI (Interpolative Double Inclusion, Lielens) analytical
models. The global axes are X, Y, and Z and represented by 1,2, and 3 indices or
subscripts. The fibers in the following studies are aligned in the X-direction. First,
the influence of the number of inclusions within the RVE on the effective material
properties is studied. Second, inclusions volume fraction in comparison with
analytical models is investigated. Both studies are carried out on a statistical basis
of three different RVEs at each data point. A transverse isotropic behavior is
assumed for the composite. Main stream averaging (MSA) homogenization of
Section 2.10 is applied through the homogenization module (see Section 3.4) that is
programmed with Python language to calculate the five independent effective
material parameters of the homogenized composite. In this case, the longitudinal
material parameters are
12 12 11
, , G and E (XY longitudinal plane) and the transverse
material parameters are
23 22
and E (YZ transverse plane of symmetry).

Sphero-cylindrical inclusions are considered in these studies. The RVEs have a
non-periodic geometry and trimming of the inclusions on the boundaries is allowed.
The RVE geometry is generated by the DigiMat-FE software and meshed using
Abaqus software meshing tool. Homogeneous boundary conditions are applied on
the microstructures as described in Section 3.2 using the boundary conditions
module (see Section 3.4).
5.2.1. Influence of Number of Inclusions

In this study, a fully aligned fiber reinforced FE-model is statistically studied for the
influence of the number of inclusions on the effective material properties. A
sufficient or minimum number of inclusions should be reached in order to have a
minimum dispersion in the results.

RVEs with the same inclusions volume fraction but different number of inclusions
are generated. A range of 15 to 60 inclusions is taken and the effective material
properties are calculated for each RVE. Statistically, three RVEs are generated with
the same inclusions and volume fraction but with random distribution. The
material behavior of the composite is assumed to be transverse isotropic and five
independent parameters of the homogenized composite are calculated by applying
the corresponding perturbations (see Section 2.10).

The specifications of the RVEs are presented in details in Table 5.2-1 below,
<< 1 >> Matrix
(2800 MPa, 0.35)
<< 2 >> Fiber Fillers
(72000 MPa, 0.20)
o inclusion_shape = sphero-cylinder
o aspect_ratio = 10
o phase_definition = by_number
o number_of_inclusions = 15
o clustering = no

o 44982 <Number of elements: < 118116
o size_rve_x = 1
o size_rve_y = 1
o size_rve_z = 1
o (Edge-trimmed Inclusions)
o tensile and shear perturbations
Table 5.2-1 Specifications of the RVEs used in the FE modeling for the
influence of the number of inclusions in comparison to Mori-Tanaka
and Lielens models of fully aligned fiber-reinforced composite

In order to have same inclusions volume fraction in all the RVEs but considering a
different number of inclusions, the inclusions length and diameter differ from RVE
to RVE but preserving a constant aspect ratio of the fibers (L/D = constant,
Inclusions length divided by its diameter).

Fig. 5.2.1 to Fig. 5.2.5 present the results of the five independent effective material
parameters
12 23 22 12 11
, , , G and E E respectively. Inclusions volume fraction of 10%
and an aspect ratio equal to 10 are purposely chosen at these relatively low levels,
since at this range usually both analytical models (MT and DI) provide a good
approximation as it is experienced in the particle reinforced composites examined in
CHAPTER IV. In Fig. 5.2.1 it is observed that for increasing number of inclusions
the difference between the effective longitudinal Youngs modulus and the analytical
models decreases.

Fig. 5.2.6 shows the statistical dispersion or the percentage difference between
maximum and minimum values of the homogenized effective longitudinal Youngs
modulus E11 of the three RVEs generated with the same number of inclusions,
keeping in mind that all the RVEs in this study have constant fiber aspect ratio and
volume fraction.

The statistical dispersion in RVEs with up to 20 inclusions is approximately 9% and
is much bigger than that of RVEs with more than 20 inclusions which have an
acceptable range between 0.88% and 4.8%. Comparing the graphs of the material
parameters, we see that most of the influence of the inclusions number is on the
E11 which is the stiffness in the direction of the alignment of the fibers and generally
on the longitudinal properties, while the number of inclusions has a minimal
influence on the transverse properties in a fully aligned fiber reinforced composite.

6000
6200
6400
6600
6800
7000
7200
7400
7600
7800
8000
8200
8400
10 20 30 40 50 60 70
Number of Inclusions [-]
E
1
1

-

L
o
n
g
i
t
u
d
i
n
a
l

E
-
m
o
d
u
l
u
s

[
M
P
a
]
Mori-Tanaka
Lielens
RVE1
RVE2
RVE3
FE-average
Poly. (FE-average)
Sphero-cylindrical fibers
Aspect ratio =10
Inclusions' V
f
=10%
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
Edge-trimmed
Incl. Size =0.8
Homogeneous BC
Seedsize =0.05

Fig. 5.2.1 E11 The effective longitudinal Youngs modulus

0.320
0.324
0.327
0.331
0.334
0.338
0.341
0.345
10 20 30 40 50 60 70
1
2

-

L
o
n
g
i
t
u
d
i
n
a
l

P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Mori-Tanaka
Lielens
RVE1
RVE2
RVE3
FE-average
Poly. (FE-average)
Aspect ratio =10
Inclusions' V
f
=10%
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
Edge-trimmed
Incl. Size =0.8
Homogeneous BC
Seedsize =0.05

Fig. 5.2.2 12 The effective longitudinal Poisson ratio
3440
3490
3540
3590
3640
3690
10 20 30 40 50 60 70
E
2
2

-

T
r
a
n
s
v
e
r
s
e

E
-
m
o
d
u
l
u
s

[
M
P
a
]
Mori-Tanaka
Lielens
RVE1
RVE2
RVE3
FE-average
Poly. (FE-average)
Aspect ratio =10
Inclusions' V
f
=10%
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
Edge-trimmed
Incl. Size =0.8
Homogeneous BC
Seedsize =0.05

Fig. 5.2.3 E22 The effective transverse Youngs modulus

0.420
0.425
0.430
0.435
0.440
0.445
0.450
0.455
10 20 30 40 50 60 70
2
3

-

T
r
a
n
s
v
e
r
s
e

P
o
i
s
s
o
n

R
a
t
i
o

[
-
]
Mori-Tanaka
Lielens
RVE1
RVE2
RVE3
FE-average
Poly. (FE-average)
Edge-trimmed
Incl. Size =0.8
Homogeneous BC
Seedsize =0.05
Aspect ratio =10
Inclusions' V
f
=10%
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2

1120
1170
1220
1270
1320
1370
1420
10 20 30 40 50 60 70
G
1
2

-

l
o
n
g
.

S
h
e
a
r

M
o
d
u
l
u
s

[
M
P
a
]
Mori-Tanaka
Lielens
RVE1
RVE2
RVE3
FE-average
Poly. (FE-average)
Aspect ratio =10
Inclusions' V
f
=10%
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
Edge-trimmed
Incl. Size =0.8
Homogeneous BC
Seedsize =0.05

Fig. 5.2.5 G12 The effective longitudinal shear modulus

Statistical Dispersion for E
11
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
10%
10 15 20 25 30 35 40 45 50 55 60 65
P
e
r
c
e
n
t
a
g
e

d
i
f
f
e
r
e
n
c
e

[
-
]
Difference between Max. and Min.
Range = 3.94 %

Fig. 5.2.6 Statistical Dispersion for the effective longitudinal Youngs
modulus E11 for RVEs with different number of inclusions
As for the appropriate number of inclusions, we have used a minimum number of
approximately 100 inclusions per RVE in the sequel.

5.2.2. FE-modeling with Different Volume Fraction

A fully aligned fiber reinforced FE-model is statistically studied for different
inclusions volume fraction. The homogenized effective material properties for the
FE-model are calculated using the same process mentioned in Section 3.4. The two
analytical models Mori-Tanaka and the interpolative double inclusions known as
Lielens model (MT and DI) are compared with the FE-model for an inclusions
volume fraction that ranges from 5% to 20%, with an inclusions aspect ratio equal
to 20.

The specifications of the RVEs are presented in details in Table 5.2-2 below,
<< 1 >> Matrix
(2800 MPa, 0.35)
(72000 MPa, 0.20)
o inclusion_shape = sphero-cylinder
o aspect_ratio = 20
o inclusion_size = 0.8
o 100<number_of_inclusions <451
o clustering = no
o Incl. volume_fraction = variable
o 327494 <Number of elements: < 686482
o size_rve_x = 1
o size_rve_y = 1
o size_rve_z = 1
o (Edge-trimmed Inclusions)
Table 5.2-2 Specifications of the RVEs used in the FE modeling for the
influence of inclusions volume fraction in comparison to Mori-Tanaka
and Lielens models of fully aligned fiber-reinforced composite

The RVEs have non-periodic geometry and edge-trimmed inclusions. Homogeneous
boundary conditions are applied and main stream averaging (MSA) homogenization
method is used to compute the five independent effective material parameters of the
transverse isotropic composite.

In Fig. 5.2.7 to Fig. 5.2.11 the independent homogenized effective material
parameters of a transverse isotropic composite polymer are plotted with MT and DI
analytical models against the inclusions volume fraction. It is known that the
interpolative double inclusion model is developed to give a better result at relatively
higher volume fractions than the Mori-Tanakas model, for details see Section 2.3.6.
In this study, the maximum inclusions volume fraction reached is 20%. The
complexity or limitations of generating RVEs with high inclusions volume fractions
and relatively high aspect ratios lies in the big amount of fibers needed to fill up the
RVE results in a complex geometry that requires very fine meshing which
consequently leads to a huge number of finite elements that exceeds our
computational capacity for statistical studies. For this reason we reached a 20%
fibers volume fraction and an aspect ratio of 20 in fiber reinforced models. The
statistical approach we used requires generating three RVEs for each data point.
The large the number of inclusions generated under these specifications of
relatively high fibers aspect ratio of 20, each RVE contained between 100 to 450
inclusions depending upon the inclusions volume fraction. This yielded a very large
number of finite elements (between 327494 and 686482), thus requiring a very high
computational effort (CPU power and mainly memory) to perform the simulations
and numerical homogenization. If the inclusions volume fraction is increased at a
constant aspect ratio, then the number of inclusions will increase dramatically,
thus developing a more complicated geometry that results in a finer mesh and
consequently a larger number of finite elements.

Fig. 5.2.7 and Fig. 5.2.11 show that at low inclusionss volume fractions (<10%), the
analytical and the numerical models give nearly overlapping results, while up to the
20% volume fraction both analytical models diverge but still in an acceptable
distance from our FE-model. This shows that for fully aligned fibers up to 20%
inclusions volume fraction, both analytical models and the FE-model yield similar
results that deviate from each other to a maximum of 3% at higher volume
fractions.

It is noticeable, that the statistical dispersion of data points of the FE-model is
minimal at all volume fractions and did not exceed 0.9%. The different RVEs
generated at each data point differ only in the fibers distribution, but there is no
clustering of inclusions and all have same orientation. This means, the
rearrangement and distribution of fibers in the RVE would not have any significant
influence on the overall effective material properties.

4500
5500
6500
7500
8500
9500
10500
11500
12500
13500
14500
0 5 10 15 20 25
E
1
1

-

L
o
n
g
i
t
u
d
i
n
a
l

E
-
m
o
d
u
l
u
s

[
M
P
a
]
Mori-Tanaka
Lielens
RVE 1
RVE 2
RVE 3
FE-Average
Poly. (FE-Average)
Edge-trimmed RVE
Incl. Size =0.8
Aspect ratio =20
Homogeneous BC
Minimum of
Seedsize =0.05
Sphero-cylindrical
Fibers
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
E11

Fig. 5.2.7 E11 The effective longitudinal Youngs modulus

0.315
0.320
0.325
0.330
0.335
0.340
0.345
0 5 10 15 20 25
1
2

-

L
o
n
g
i
t
u
d
i
n
a
l

P
o
i
s
s
o
n

r
a
t
i
o

[
-
]
Mori-Tanaka
RVE 1
RVE 2
RVE 3
Lielens
FE-Average
Poly. (FE-Average)
Sphero-cylindrical
Fibers
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
(
12
)
Edge-trimmed RVE
Incl. Size =0.8
Aspect ratio =20
Homogeneous BC
Minimum of
Seedsize =0.05


3100
3300
3500
3700
3900
4100
4300
4500
0 5 10 15 20 25
E
2
2

-

T
r
a
n
s
v
e
r
s
e

E
-
m
o
d
u
l
u
s

[
M
P
a
]
Mori-Tanaka
Lielens
RVE 1
RVE 2
RVE 3
FE-Average
Poly. (FE-Average)
Sphero-cylindrical
Fibers
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
E
22
Edge-trimmed RVE
Incl. Size =0.8
Aspect ratio =20
Homogeneous BC
Minimum of
Seedsize =0.05

Fig. 5.2.9 E22 The effective transverse Youngs modulus
0.415
0.425
0.435
0.445
0.455
0.465
0.475
0 5 10 15 20 25
2
3

-

T
r
a
n
s
v
e
r
s
e

P
o
i
s
s
o
n

r
a
t
i
o

[
-
]
Mori-Tanaka
Lielens
RVE 1
RVE 2
RVE 3
FE-Average
Poly. (FE-Average)
Sphero-cylindrical
Fibers
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
(
23
)
Edge-trimmed RVE
Incl. Size =0.8
Aspect ratio =20
Homogeneous BC
Minimum of
Seedsize =0.05


1100
1150
1200
1250
1300
1350
1400
1450
1500
1550
1600
0 5 10 15 20 25
G
1
2

-

L
o
n
g
i
t
'
a
l

S
h
e
a
r

m
o
d
u
l
u
s

[
M
P
a
]
Mori-Tanaka
Lielens
1
2
3
Average
Poly. (Average)
Sphero-cylindrical
Fibers
E
m
=2800 MPa
m
=0,35
E
f
=72000
f
=0,2
Edge-trimmed RVE
Incl. Size =0.8
Aspect ratio =20
Homogeneous BC
Minimum of
Seedsize =0.05

Fig. 5.2.11 G12 The effective longitudinal shear modulus
5.3. FE-Modeling of a partially oriented fiber reinforced
composite

In this section, RVE Type (2) as displayed in Fig. 5.1.1, a partially oriented fiber
reinforced FE-model is generated and analyzed. The inclusions volume fraction is
7.75% and their aspect ratio is 25. Partially oriented fibers means that
approximately 90% of the fibers are aligned in a single direction.

The RVE specifications are listed in details in Table 5.3-1 below,

<< 1 >> Matrix
(3294 MPa, 0.345)
(72000 MPa, 0.20)
Orientation tensor components
o orientation_11 = 0.90

o inclusion_shape = cylinder
o aspect_ratio = 25
o phase_definition = by_size_range
o max_inclusion_size = 0.99
o size_distribution = random
o number_of_inclusions = 144
o Incl. volume_fraction = 7.75%

o Number of elements: = 357913
o size_rve_x = 1
o size_rve_y = 1
o size_rve_z = 1
o (Fully embedded Inclusions)
o clustering = no
Table 5.3-1 Specifications of the RVE used in the FE modeling for a partially
oriented fiber reinforced composite in comparison to Mori-Tanakas
analytical model

3956 3985
1636
4005 3989
1576
6483
7054
0
1000
2000
3000
4000
5000
6000
7000
8000
E11 E22 E33 G12
Y
o
u
n
g
'
s

a
n
d

S
h
e
a
r

m
o
d
u
l
i

[
M
P
a
]
FE Model
Analytical Model

Fig. 5.3.1 FE-model homogenized effective Youngs moduli in the three
principle directions and the shear modulus in XY plane in comparison
to DigiMat-MF (Mean Field), based on Mori-Tanakas analytical model.

To calculate the respective material parameters, tensile perturbations in x,y, and z
directions and a shear perturbation in the xy-plane are applied to the RVE with
each time to determine the relevant homogenized material properties as described
in Section 2.10.1.2.

Results in Fig. 5.3.1 show a comparison graph for the Youngs moduli in all
directions but a difference in the x-direction (E11) is noted between the MF and FE
models where most of the fibers are aligned. The results calculated using DigiMat-
MF (Mean Field) based on Mori-Tanakas analytical model is approximately 7.1%
stiffer than our FE-model, while Youngs moduli in the other directions (E22, E33)
yielded similar results with a minor difference not exceeding 1.2%. The difference in
shear modulus G12 is 3.6%.
0.360
0.398
0.201
0.356
0.415
0.183
0.000
0.040
0.080
0.120
0.160
0.200
0.240
0.280
0.320
0.360
0.400
0.440
0.480
12 23 31
P
o
i
s
s
o
n

r
a
t
i
o

i
n

t
h
e

p
r
i
n
c
i
p
l
e

p
l
a
n
e
s

[
-
]
FE Model
Analytical Model

Fig. 5.3.2 FE-model homogenized effective Poisson ratio in the three
principle planes in comparison to to DigiMat-MF (Mean Field), based
on Mori-Tanakas analytical model.

These results for partially oriented fiber reinforced composites show that the major
difference between the analytical model and the RVE is in the stiffness in the
direction where most of the fibers are aligned (x-direction). These results also
demonstrate that in a partially oriented fiber reinforced composites, the overall
material behavior can be considered as transverse isotropic. Orthotropic behavior is
not much evident in this model. If we compared the Youngs moduli E22 and E33 of
each model alone, for example in the FE-model they are 3956 MPa and 4005 MPa
respectively; they give very close results proving that plane Y-Z is a plane of
symmetry of a transverse isotropic material. This is true, since orthotropy increases
as the fibers attain higher planar orientation which is not the case here.

5.3.1. Statistical study

A statistical study is presented for four representative volume elements modeling a
partially oriented fiber reinforced composite. All have the same specifications
mentioned in Table 5.3-2, but with different random fibers distribution for
statistical purposes.

<<1 >>Matrix inclusion_shape =cylinder orientation_11 =0.9
Isotropic Elastic behavior aspect_ratio =25 orientation_22 =0.08
(Young modulus, Poisson's ratio) phase_definition =by_size orientation_33 =0.02
(3294, 0.345) inclusion_size =0.99 orientation_12 =0.04
<<2 >>Cylinder Fillers size_distribution =fixed orientation_23 =0.004
Incl_volume_fraction =5% coated =no orientation_13 =0.017
Isotropic Elastic behavior
(Youngs mod., Poisson's
ratio)
Minimum number of.
Elements =171430

(72000, 0.20)
Maximum number of.
Elements =177288

Table 5.3-2 Representative Volume elements specifications

E11 E22 E33 G12 V12 V23 V31
RVE1 5231 3715 3747 1503 0.3692 0.3803 0.2382
RVE2 5286 3708 3749 1520 0.3697 0.3817 0.2354
RVE3 4824 3709 3768 1603 0.3889 0.3615 0.2495
RVE4 5393 3716 3749 1483 0.3629 0.3860 0.2332
FE-Average 5184 3712 3753 1527 0.3727 0.3774 0.2391
Average
Deviation 2.78% 0.07% 0.16% 1.99% 1.74% 1.69% 1.75%
Table 5.3-3 Homogenized material parameters for partially oriented fiber reinforced
composite

Table 5.3-3 shows the results of numerical homogenization of four RVEs extracting
the effective orthotropic material parameters (
12 31 33 23 22 12 11
, , , , , G and E E E ).
Average deviation percentage is calculated as per eqn. 4.4-1, showing no major
statistical significance and recorded a maximum of 2.78% for the longitudinal
Youngs modulus
11
E .

5.4. FE-Modeling of fiber reinforced composites with a
single and multi layer orientation

In this section, RVE Type (3) as displayed in Fig. 5.1.1, a generally oriented fiber
reinforced FE-model is generated of inclusions volume fraction of 5% and an aspect
ratio of 20. This FE-model is generated with three layers of orientation where the
fibers of the outmost layers have the same orientation which differs from the one in
the middle as shown in Fig. 5.4.1.

Fig. 5.4.1 Multi-layered RVE for a fiber reinforced composite with general
fiber orientation

In reality, fibers usually have different orientation at different sections of the
composite due the flow profile during the injection molding process. Usually when
the fibers orientations are experimentally measured, they are calculated over the
thickness of the specimen in an averaging scheme. In this chapter a RVE with three
layers are built. The two outmost layers are with the same orientation while the
orientation of the middle layer differs. Further a RVE with a single layer of
orientation which is the average of the three layers as shown in Table 5.4-1 is
analyzed.

3-layer RVE

1-layer RVE
Layer-1 Layer-2 Layer-3 Layer-1
orientation_11 0.7875 0.3 0.7875
average
0.625
orientation_22 0.1675 0.665 0.1675
average
0.332
orientation_33 0.045 0.035 0.045
average
0.043
orientation_12 0.04 -0.05 0.04
average
0.01
orientation_13 0.025 0.01 0.025
average
0.02
orientation_23 0.02 0.025 0.02
average
0.022
Table 5.4-1 Orientation tensor components of a 3-layered and a
single layered RVE for a fiber reinforced composite

The specifications of the RVEs generated for the 3-layered and single-layered
models are described in Table: 5.4-2.

<< 1 >> Matrix
(3294 MPa, 0.345)
(72000 MPa, 0.20)
Orientation tensor components
o As in Table 5.4-1

o inclusion_shape = cylinder
o aspect_ratio = 20
o size_rve_x = 1
o size_rve_y = 1
o size_rve_z = 1
o inclusion_size 1-layer RVE
0.7
o inclusion_size 3-layer RVE
0.33
o number_of_inclusions 1-layer RVE
74
o number_of_inclusions 3-layer RVE
266+229+266 = 761
o Number of elements 3-layer RVE:
398441
o Number of elements 1-layer RVE:
216837
o (Fully embedded Inclusions)
o clustering = no
o tensile perturbations
Table: 5.4-2 three-layer and single-layer RVEs Specifications

4030
3820
4266
3975
3816
4697
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
E11 E22 E33
Y
o
u
n
g
'
s

m
o
d
u
l
i

i
n

t
h
r
e
e

d
i
r
e
c
t
i
o
n
s

[
M
P
a
]
3-layer FE-Model
1-layer FE-Model

Fig. 5.4.2 FE-model comparison of the homogenized effective Youngs
moduli in the three principle directions of a single and multi layer
orientation of a fiber reinforced composite

To calculate the respective material parameters, tensile perturbations in x, y, and z
directions are applied to the RVE with each time calculating the relevant
homogenized material properties as described in Section 2.10.1.2.

Results in Fig. 5.4.2 and Fig. 5.4.3 show that the E11 modulus of the multi-layer
RVE differs from the value predicted for the RVE with an averaged single layer
orientation. The difference in the stiffness or Youngs modulus (E11) of the 3-layer
RVE is approximately 10%.

0.346
0.358
0.271
0.343
0.346
0.301
0.000
0.040
0.080
0.120
0.160
0.200
0.240
0.280
0.320
0.360
0.400
0.440
12 23 31
P
o
i
s
s
o
n

r
a
t
i
o

i
n

t
h
r
e
e

d
i
r
e
c
t
i
o
n
s

[
-
]3-layer FE-Model
1-layer FE-Model

Fig. 5.4.3 FE-model comparison of the homogenized effective poisson ratio
in the three principle planes of a single and multi layer orientation of a
fiber reinforced composite

CHAPTER VI

6. Summary and Outlook

This chapter includes a summary of major conclusions and the future works
which can be based on this monograph

CHAPTER VI - Summary and Outlook - 128 -
6.1. SUMMARY

6.1.1. 3D Boundary Conditions

Periodic and homogeneous boundary conditions developed in this work intended to
optimize their usage for the numerical simulation of microstructures. Regarding the
3D periodic boundary conditions, previous attempts were used for the same
purpose of homogenization but many obstacles faced the developers. For instance,
Charles L. Tucker III and Erwin Liang [27] in their publication [27] could not apply
multi-axial perturbations on the microstructure using their approach of periodic
boundary conditions. In addition, they had to have periodic geometry of the
microstructure and symmetrical mesh on the boundaries as it was the case of Xia
et al. [67]. These obstacles are bypassed by applying our approach in using the
dummy or control node concept in conjunction with an appropriate set of
equations and constraints on the boundaries. We were able to apply the load and
translate it to the microstructure while preserving the periodicity conditions.

The above combination of obstacles is solved in our 3D periodic boundary
conditions approach, i.e. any type of loading can be applied (uni-axial, multi-axial)
and on any microstructure (periodic, non periodic, symmetrical mesh on boundaries,
non symmetrical mesh, .. etc).

Comprehensive studies in CHAPTER II are done on the development and
verification of the homogeneous and periodic boundary conditions (HBC and PBC),
where studies showed that the numerical stability and the accuracy in satisfying
the periodicity conditions of the PBC are very much acceptable with no significant
errors on the overall homogenized effective material properties of the FE-model
(RVE). The PBC requirements and advantages are as follows,

Properties
No separation or overlap between the neighboring RVEs after deformation
of a periodic array representing the composite material, i.e. boundary
displacement continuity is preserved.
Traction continuity is preserved between periodic surfaces.
Sustaining a uni-axial stress state in case of unidirectional perturbations.

Advantages
The whole simulation process of the micro-structure is managed through
a single control-node referred to as the dummy-node.
Independent of the geometry of the micro-structure.
All types of perturbations or loadings are applicable (uni-axial, Multi-
axial, thermal, thermo-mechanical,)
Periodic mesh on the boundaries is not a necessity, which allows easier
and more flexible built-up of the microstructures.

6.1.2. FE-Modeling of Particle Reinforced Composites

The FE-model or the representative volume element is generally seen as a volume of
a heterogeneous material that is sufficiently large to engulf all major heterogeneities
of the composites microstructure, and to be statistically representative of that
composite material.

Several major aspects had to be studied to affirm the compatibility of our
model with respect to the targeted material behavior of the macro-structure.
Boundary condition studies revealed that periodic boundary conditions are more
relevant and appropriate to use with the two cases of periodic geometries whether
edge-trimmed particles or fully-embedded. RVEs with PBC yielded softer results
than homogenous boundary conditions (HBC), and this is plausible due to the over
constrained conditions the HBC provide.
In addition, another study on the influence of the number of particles which has
been carried out can also be considered a study versus the size of the particles. The
volume fraction is fixed and all particles have same size, thus as number of
particles increases the size of particles decreases proportionally. The results showed
that as number of inclusions increases, more stable homogenized material
properties are obtained or minimal statistical dispersion (<<< 1%). This is plausible,
because as the number of inclusions in the RVE tends to infinity the heterogeneous
microstructure tends to homogeneity and the effect of the inclusions randomness in
the microstructure tends to zero. As the results show, a minimum of 40 spherical
inclusions is a good number to avoid undesired statistical errors. Previously, some
research was done on the influence of the size of the RVE on output results; for
example, (Zeman,2003) [68] reported that the transverse elastic behavior of
continuously reinforced composites can be satisfactorily described by unit cells
containing arrangements of 10 to 20 fibers. For the case of elastic statistically
isotropic composites with matrix and sphere-like particles, (Drugan and
Willis,1996)[69] estimated that for approximating the overall moduli with errors of
less than 5% or less than 1%, respectively, a non-periodic RVE with edge sizes of
approximately two or five inclusion diameters are sufficient for any volume fraction;
see also (Drugan,2000) [70]. Those researchers used the ratio between the cube
edge length L and the same size spheres diameter D and came up with a range
of 2 to 5. In our case the range of 10 to 60 same size particles is equivalent to range
of L/D between 3.3 and 6.0, but the less than 1% relative deviation was found
near and above the L/D = 5.2 which is equivalent to approximately 40 same size
spherical particles inside a unit volume cubic RVE.

A statistical study is done for different inclusions volume fraction calculating
the overall homogenized effective material properties and comparing the FE-model
with the Mori-Tanakas and Lielens models. This study is also done on same-size,
fully embedded spherical particles RVEs Up to 30% inclusions volume fraction.
Both homogeneous and periodic boundary conditions (HBC and PBC) are used.
Results of RVEs with periodic and homogenous boundary conditions show that at
low volume fractions (<10%) the three models overlap in their representation of the
effective properties of the composite. While at higher volume fractions (>15%) they
differ rather little, but overall, Lielens model shows better results in comparison to
the FE-model. This allows concluding that regardless of the boundary conditions,
Lielens model is closer to the FE-model especially at higher volume fractions. A
similar study is carried out with the HBC and PBC but on non-fully embedded
inclusions, i.e. they are trimmed on the boundaries of the cubic RVE. It is
concluded that the use of PBC is suitable on periodic microstructures and HBC on
non-periodic microstructures, while the former generally yields softer results than
the latter.

Last but not least the comparison between experimental measurements of a
particle reinforced composite material with 17% inclusions volume fraction,
analytical models, and the FE-model has been made. The material properties of the
filler and the unfilled material (matrix) were measured and used as inputs for the
analytical models and the RVEs. Though knowing in advance that usually in such
composites, the material properties (MPs) measured for the unfilled-material may
not be accurately compared to the MPs of the same material in the reinforced state.
This means, experimentally it is not possible to separate the filler from the matrix
and characterize the matrix disconnectedly to identify the exact properties at the
given reinforced state. So it has to be prepared from scratch but with no fillers, and
this would yield some differences from the properties when it is in the reinforced
state. The results show that the FE-model gives the closest result to the
measurements, although the FE-model and Lielens model are of not much
difference for this material.

6.1.3. FE-Modeling of Fiber Reinforced Composites

6.1.3.1. Fully aligned fiber reinforced composite

Fully aligned fiber reinforced FE-model is statistically studied for the influence of
the number of inclusions on the effective material properties, and compared to
Mori-Tanaka and Lielens analytical models. The RVEs are consistent and have
constant inclusions volume fraction and aspect ratio, while the only variable is the
number of inclusions and their random distribution within the RVE. A sufficient or
minimum number of inclusions is reached in order to have a minimum dispersion
in the results. As it is shown in Fig. 5.2.6, the statistical dispersion in RVEs with up
to 20 inclusions is approximately 9% and is much bigger than that of RVEs with
more than 20 inclusions which have an acceptable range of dispersion of 3.94%. So
for a minimal influence of inclusions number on the effective MP, a minimum of 25
inclusions should be considered. Results show that most of the influence of the
inclusions number is on the E11 which is the stiffness in the direction of the
alignment of the fibers and generally on the longitudinal properties, while the
number of inclusions has a minimal influence on the transverse properties in a
fully aligned fiber reinforced composite.

The same model is also statistically studied for an inclusions volume fraction that
ranges from 5% to 20%, with an inclusions aspect ratio of 20. The RVEs have a
non-periodic geometry and edge-trimmed inclusions. Homogeneous boundary
conditions are applied and main stream averaging (MSA) homogenization method is
used to compute the five independent effective material parameters of the
transverse isotropic composite. Fig. 5.2.7 to Fig. 5.2.11, show that at low
inclusionss volume fractions (<10%), the analytical and numerical models are
nearly overlapping, while up to the 20% volume fraction both analytical models
diverge but still in an acceptable range from our FE-model. This shows that for fully
aligned fibers up to 20% inclusions volume fraction, both analytical models and the
FE-model yield similar results that deviate from each other to a maximum of 3% at
higher volume fractions between 10%-20%. So for fully aligned fibers the averaging
approach used in the analytical models is reliable in comparison to the FE-model.

6.1.3.2. Partially oriented fiber reinforced composite

RVE Type (2) as displayed in Fig. 5.1.1, a partially oriented fiber reinforced FE-
model is generated and analyzed. The inclusions volume fraction is 7.75% and their
aspect ratio is 25. Partially oriented fibers, means that approximately 90% of the
fibers are aligned in a single direction. Results in Fig. 5.3.1 show a comparison
graph for the Youngs moduli in all directions but a difference in the x-direction
(E11) is noted between the MF and FE models where most of the fibers are aligned.
The results calculated using DigiMat-MF (Mean Field) based on Mori-Tanakas
analytical model is approximately 7.1% stiffer than our FE-model, while Youngs
moduli in the other directions (E22, E33) yielded similar results with a minor
difference not exceeding 1.2%. The difference in shear modulus G12 is 3.6%. These
results for partially oriented fiber reinforced composites show that the major
difference between the analytical model and the RVE is in the stiffness in the
direction where most of the fibers are aligned (x-direction). These results also
demonstrate that in a partially oriented fiber reinforced composites, the overall
material behavior can be considered as transverse isotropic. Orthotropic behavior is
not much evident in this model. If we compared the Youngs moduli E22 and E33 of
each model alone, for example in the FE-model they are 3956 MPa. and 4005 MPa
respectively, they give very close results proving that plane Y-Z is a plane of
symmetry of a transverse isotropic material. This is true, since orthotropy increases
as the fibers attain higher planar orientation which is not the case here.

6.1.3.3. Multi-layer fiber orientation FE-model

RVE Type (3) as displayed in Fig. 5.1.1, a generally oriented fiber reinforced FE-
model is generated of inclusions volume fraction of 5% and an aspect ratio of 20. A
three layer FE-model is compared to a single layer FE-model, see Fig. 5.4.1. The
fiber orientation tensor of the single layer RVE is calculated by taking the average of
the orientation tensors of the three layered RVE. This approach helps to check
whether it is convenient to model a muli-layer fiber orientation FE-model rather
than an averaged single layer fiber orientation model. In experiments, fibers are
measured through the thickness of the test sample and results for fiber orientation
are collected at different positions (layers). Most injection molded fiber reinforced
composites have several major zones of fiber orientation due to the flow profile, one
at the peripheries and another at the center of the flow. The Fe-model comparison
between 3-layered and single layered results showed a considerable difference of
10% in the value of E11 (longitudinal Youngs modulus), see Fig. 5.4.2. A multi-layer
fiber orientation modeling yields more realistic results and adds a significant
enhancement in predicting the effective material properties.

6.2. FUTURE-WORKS

Now that the micromechanical material FE-model is established, the next step
should be thorough investigations on the interface between the matrix and fillers,
such as debonding and crack initiation phenomena.

The search for suitable failure criteria is to be pursued through specific
experiments to be implemented in the FE-model. Furthermore, micromechanical
damage models ought to be investigated and realized. All the future experiments
and simulations will be focusing on the post yield phase of the composite material
behavior under loadings.

The suggested process is depicted in Fig. 6.2.1 and Fig. 6.2.2. The major goal
is to achieve a macro-scale model based on a micro-scale model that includes the
most prominent micromechanical influences on the quality and life time of the
composite which is the ultimate goal for various industries especially the industries
that deal with polymer composites, as the usage spectrum of sophisticated polymer
composites in industry is on the rise. This could be used with benefit for the design
of composites and the relative manufactured products.


Fig. 6.2.1 Orientation flow chart - Future and ongoing works (road-map)
F
u
t
u
r
e

W
o
r
k
s

I
n
t
e
r
a
c
t
i
v
e

P
l
a
n
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f
t
h
e

M
a
t
r
i
x

U
n
d
a
m
a
g
e
d

S
t
a
t
e

PHASE-I-
EXPERIMENTS AND BASICS
FOR MODELLING
-
-
1
)
P
r
o
p
e
r
t
i
e
s

u
s
e
d

t
h
r
o
u
g
h
o
u
t

t
h
e

s
i
m
u
l
a
t
i
o
n

2
)
T
o

b
e

C
o
m
p
a
r
e
d

w
i
t
h

t
h
e

f
i
n
a
l

r
e
s
u
l
t
s

(
D
a
m
a
g
e

m
o
d
e
l
)
G
h
a
y
a
t
h

A
.

A
l

K
a
s
s
e
m
G
h
a
y
a
t
h

A
.

A
l

K
a
s
s
e
m
O
R
I
E
N
T
A
T
I
O
N
F
L
O
W

C
H
A
R
T
F
r
a
c
t
u
r
e

E
n
e
r
g
y
F
o
r

t
h
e

p
u
r
e

M
a
t
r
i
x
E
X
P
E
R
I
M
E
N
T
A
L
S
I
M
U
L
A
T
I
O
N
PHASE-II-
MICROMECHANICS
SIMULATIONS
PHASE-IV-
MACRO SCALE
SIMULATIONS
PHASE-III-
MACRO SCALE
MODELS
PHASE-V-
ASSESMENT OF
MODELS
F
r
a
c
t
u
r
e

E
n
e
r
g
y
o
f

t
h
e

I
n
t
e
r
f
a
c
e
F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h
i
n

t
h
e

p
u
r
e

M
a
t
r
i
x
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n
o
f

C
o
m
p
o
s
i
t
e

M
a
t
e
r
i
a
l
1
)
F
o
r
m
u
l
a
t
i
o
n

a
n
d

I
m
p
l
e
m
e
n
t
a
t
i
o
n

o
f

V
o
i
d

E
l
e
m
e
n
t
s
C
o
h
e
s
i
v
e

E
l
e
m
e
n
t
s

B
e
t
w
e
e
n

S
o
l
i
d

E
l
e
m
e
n
t
s
M
o
d
e
l
:

F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

i
n

t
h
e

B
u
l
k
M
o
d
e
l
:

F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

a
l
o
n
g

t
h
e

I
n
t
e
r
f
a
c
e
F
a
t
i
g
u
e

A
n
a
l
y
s
i
s
f
o
r

t
h
e

p
u
r
e

M
a
t
r
i
x
F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

a
l
o
n
g

t
h
e

I
n
t
e
r
f
a
c
e
F
a
t
i
g
u
e

A
n
a
l
y
s
i
s
I
n

C
o
m
p
o
s
i
t
e

M
a
t
e
r
i
a
l
2
)
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

w
i
t
h

u
n
d
a
m
a
g
e
d
M
a
t
r
i
x

b
e
h
a
v
i
o
r
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

f
o
r

c
o
m
p
o
s
i
t
e

w
i
t
h

M
a
t
r
i
x

r
e
p
r
e
s
e
n
t
e
d

b
y

p
h
e
n
o
m
e
n
o
l
o
g
i
c
a
l

d
a
m
a
g
e

m
e
c
h
a
n
i
c
s

m
o
d
e
l
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

o
f

c
r
a
c
k

p
a
t
h

i
n

c
o
m
p
o
s
i
t
e
s
;

F
r
a
c
t
u
r
e

v
i
a
C
o
h
e
s
i
v
e

E
l
e
m
e
n
t
s
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

o
f

F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

i
n

C
o
m
p
o
s
i
t
e

M
a
t
e
r
i
a
l
C
R
I
T
E
R
I
O
N

F
O
R

D
A
M
A
G
E
E
x
t
r
a
c
t
e
d

f
r
o
m

M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

o
f

c
r
a
c
k

p
a
t
h

i
n

c
o
m
p
o
s
i
t
e
s
;

F
r
a
c
t
u
r
e

v
i
a
V
o
i
d

E
l
e
m
e
n
t
s
L
i
t
e
r
a
t
u
r
e

S
e
a
r
c
h
C
O
N
S
T
I
T
U
T
I
V
E

M
O
D
E
L

F
O
R

C
O
M
P
O
S
I
T
E

I
N

U
N
D
A
M
A
G
E
D

S
T
A
T
E
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

p
h
e
n
o
m
e
n
o
l
o
g
i
c
a
l

D
a
m
a
g
e

M
o
d
e
l

f
o
r

M
a
t
r
i
x
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

C
o
h
e
s
i
v
e

e
l
e
m
e
n
t
s
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

f
a
t
i
g
u
e

c
r
a
c
k

g
r
o
w
t
h

m
o
d
e
l
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

V
o
i
d

E
l
e
m
e
n
t
s
A
P
P
L
I
C
A
T
I
O
N

O
F

M
A
C
R
O
S
C
A
L
E

M
O
D
E
L
S

T
O

S
A
M
P
L
E

G
E
O
M
E
T
R
I
E
S
A
P
P
L
I
C
A
T
I
O
N

O
F

M
A
C
R
O
S
C
A
L
E

M
O
D
E
L
S

T
O

P
A
R
T
S
I
N
V
E
S
T
I
G
A
T
I
O
N

O
F

C
O
R
R
E
L
A
T
I
O
N

B
E
T
W
E
E
N

L
I
F
E

T
I
M
E

D
A
T
A

A
N
D

R
E
S
U
L
T
S

F
R
O
M

M
A
C
R
O
S
C
A
L
E

S
I
M
U
L
A
T
I
O
N
S

P
H
A
S
E
S
A
r
g
u
m
e
n
t

f
o
r

f
a
t
i
g
u
e

s
a
m
p
l
e
s
;

H
o
m
o
g
e
n
e
o
u
s
l
o
a
d
i
n
g

i
n

t
h
e

i
n
t
e
r
i
o
r

o
f

t
h
e

s
a
m
p
l
e

i
s

r
e
p
r
e
s
e
n
t
e
d
C
o
m
b
i
n
a
t
i
o
n

o
f
C
o
m
b
i
n
a
t
i
o
n

o
f
C
o
m
b
i
n
a
t
i
o
n

o
f
P
r
o
v
i
d
e
d
,

l
i
f
e

t
i
m
e

d
a
t
a

a
v
a
i
l
a
b
l
e

i
n
-
h
o
u
s
e
F
u
t
u
r
e

W
o
r
k
s

I
n
t
e
r
a
c
t
i
v
e

P
l
a
n
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f
t
h
e

M
a
t
r
i
x

U
n
d
a
m
a
g
e
d

S
t
a
t
e

PHASE-I-
EXPERIMENTS AND BASICS
FOR MODELLING
-
-
1
)
P
r
o
p
e
r
t
i
e
s

u
s
e
d

t
h
r
o
u
g
h
o
u
t

t
h
e

s
i
m
u
l
a
t
i
o
n

2
)
T
o

b
e

C
o
m
p
a
r
e
d

w
i
t
h

t
h
e

f
i
n
a
l

r
e
s
u
l
t
s

(
D
a
m
a
g
e

m
o
d
e
l
)
G
h
a
y
a
t
h

A
.

A
l

K
a
s
s
e
m
G
h
a
y
a
t
h

A
.

A
l

K
a
s
s
e
m
O
R
I
E
N
T
A
T
I
O
N
F
L
O
W

C
H
A
R
T
F
r
a
c
t
u
r
e

E
n
e
r
g
y
F
o
r

t
h
e

p
u
r
e

M
a
t
r
i
x
E
X
P
E
R
I
M
E
N
T
A
L
S
I
M
U
L
A
T
I
O
N
PHASE-II-
MICROMECHANICS
SIMULATIONS
PHASE-IV-
MACRO SCALE
SIMULATIONS
PHASE-III-
MACRO SCALE
MODELS
PHASE-V-
ASSESMENT OF
MODELS
F
r
a
c
t
u
r
e

E
n
e
r
g
y
o
f

t
h
e

I
n
t
e
r
f
a
c
e
F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h
i
n

t
h
e

p
u
r
e

M
a
t
r
i
x
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n
o
f

C
o
m
p
o
s
i
t
e

M
a
t
e
r
i
a
l
1
)
F
o
r
m
u
l
a
t
i
o
n

a
n
d

I
m
p
l
e
m
e
n
t
a
t
i
o
n

o
f

V
o
i
d

E
l
e
m
e
n
t
s
C
o
h
e
s
i
v
e

E
l
e
m
e
n
t
s

B
e
t
w
e
e
n

S
o
l
i
d

E
l
e
m
e
n
t
s
M
o
d
e
l
:

F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

i
n

t
h
e

B
u
l
k
M
o
d
e
l
:

F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

a
l
o
n
g

t
h
e

I
n
t
e
r
f
a
c
e
F
a
t
i
g
u
e

A
n
a
l
y
s
i
s
f
o
r

t
h
e

p
u
r
e

M
a
t
r
i
x
F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

a
l
o
n
g

t
h
e

I
n
t
e
r
f
a
c
e
F
a
t
i
g
u
e

A
n
a
l
y
s
i
s
I
n

C
o
m
p
o
s
i
t
e

M
a
t
e
r
i
a
l
2
)
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

w
i
t
h

u
n
d
a
m
a
g
e
d
M
a
t
r
i
x

b
e
h
a
v
i
o
r
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

f
o
r

c
o
m
p
o
s
i
t
e

w
i
t
h

M
a
t
r
i
x

r
e
p
r
e
s
e
n
t
e
d

b
y

p
h
e
n
o
m
e
n
o
l
o
g
i
c
a
l

d
a
m
a
g
e

m
e
c
h
a
n
i
c
s

m
o
d
e
l
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

o
f

c
r
a
c
k

p
a
t
h

i
n

c
o
m
p
o
s
i
t
e
s
;

F
r
a
c
t
u
r
e

v
i
a
C
o
h
e
s
i
v
e

E
l
e
m
e
n
t
s
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

o
f

F
a
t
i
g
u
e

C
r
a
c
k

G
r
o
w
t
h

i
n

C
o
m
p
o
s
i
t
e

M
a
t
e
r
i
a
l
C
R
I
T
E
R
I
O
N

F
O
R

D
A
M
A
G
E
E
x
t
r
a
c
t
e
d

f
r
o
m

M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n
M
i
c
r
o
m
e
c
h
a
n
i
c
s

S
i
m
u
l
a
t
i
o
n

o
f

c
r
a
c
k

p
a
t
h

i
n

c
o
m
p
o
s
i
t
e
s
;

F
r
a
c
t
u
r
e

v
i
a
V
o
i
d

E
l
e
m
e
n
t
s
L
i
t
e
r
a
t
u
r
e

S
e
a
r
c
h
C
O
N
S
T
I
T
U
T
I
V
E

M
O
D
E
L

F
O
R

C
O
M
P
O
S
I
T
E

I
N

U
N
D
A
M
A
G
E
D

S
T
A
T
E
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

p
h
e
n
o
m
e
n
o
l
o
g
i
c
a
l

D
a
m
a
g
e

M
o
d
e
l

f
o
r

M
a
t
r
i
x
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

C
o
h
e
s
i
v
e

e
l
e
m
e
n
t
s
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

f
a
t
i
g
u
e

c
r
a
c
k

g
r
o
w
t
h

m
o
d
e
l
M
i
c
r
o
m
e
c
h
a
n
i
c
s

B
a
s
e
d

D
a
m
a
g
e

M
o
d
e
l

f
o
r

C
o
m
p
o
s
i
t
e
,

b
a
s
e
d

o
n

V
o
i
d

E
l
e
m
e
n
t
s
A
P
P
L
I
C
A
T
I
O
N

O
F

M
A
C
R
O
S
C
A
L
E

M
O
D
E
L
S

T
O

S
A
M
P
L
E

G
E
O
M
E
T
R
I
E
S
A
P
P
L
I
C
A
T
I
O
N

O
F

M
A
C
R
O
S
C
A
L
E

M
O
D
E
L
S

T
O

P
A
R
T
S
I
N
V
E
S
T
I
G
A
T
I
O
N

O
F

C
O
R
R
E
L
A
T
I
O
N

B
E
T
W
E
E
N

L
I
F
E

T
I
M
E

D
A
T
A

A
N
D

R
E
S
U
L
T
S

F
R
O
M

M
A
C
R
O
S
C
A
L
E

S
I
M
U
L
A
T
I
O
N
S

P
H
A
S
E
S
A
r
g
u
m
e
n
t

f
o
r

f
a
t
i
g
u
e

s
a
m
p
l
e
s
;

H
o
m
o
g
e
n
e
o
u
s
l
o
a
d
i
n
g

i
n

t
h
e

i
n
t
e
r
i
o
r

o
f

t
h
e

s
a
m
p
l
e

i
s

r
e
p
r
e
s
e
n
t
e
d
C
o
m
b
i
n
a
t
i
o
n

o
f
C
o
m
b
i
n
a
t
i
o
n

o
f
C
o
m
b
i
n
a
t
i
o
n

o
f
P
r
o
v
i
d
e
d
,

l
i
f
e

t
i
m
e

d
a
t
a

a
v
a
i
l
a
b
l
e

i
n
-
h
o
u
s
e

Fig. 6.2.2 General Plan flow chart - Future and ongoing works (road-map)
(
B
)
L
i
t
e
r
a
t
u
r
e

S
e
a
r
c
h

i
n

D
a
m
a
g
e

a
n
d

F
r
a
c
t
u
r
e

M
e
c
h
a
n
i
c
s
1
-
C
r
a
c
k

p
r
o
p
a
g
a
t
i
o
n

c
r
i
t
e
r
i
a
-
T
h
r
o
u
g
h

t
h
e

m
a
t
r
i
x

i
n
t
o

t
h
e

i
n
t
e
r
f
a
c
e

a
r
o
u
n
d

t
h
e

p
a
r
t
i
c
l
e
s
.

-
O
r

o
n
l
y

t
h
r
o
u
g
h

t
h
e

m
a
t
r
i
x
.
-
O
r

o
n
l
y

d
e
l
a
m
i
n
a
t
i
o
n

i
n

t
h
e

i
n
t
e
r
f
a
c
e
.
G
e
n
e
r
a
l

P
l
a
n
(
A
)
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

n
e
w

c
o
m
p
o
s
i
t
e

m
a
t
e
r
i
a
l

1
-
T
h
e

N
e
w
M
a
t
e
r
i
a
l

p
r
o
p
e
r
t
i
e
s
:
-
D
e
t
e
r
m
i
n
a
t
i
o
n

o
f

t
h
e

Y
o
u
n
g
s

m
o
d
u
l
u
s

a
s

f
u
n
c
t
i
o
n

o
f

t
e
m
p
e
r
a
t
u
r
e
,

t
e
n
s
i
l
e

t
e
s
t
s

a
t

C
R
/
A
P
P
3

(
i
n
-
h
o
u
s
e
)
.
-
P
o
i
s
s
o
n
s

r
a
t
i
o
:

D
K
I

f
o
r

o
p
t
i
c
a
l

s
t
r
a
i
n

m
e
a
s
u
r
e
m
e
n
t
s
.
-
V
i
s
c
o
e
l
a
s
t
i
c
i
t
y
:

D
M
A
,

c
h
e
c
k

c
u
r
r
e
n
t

c
a
p
a
c
i
t
i
e
s

o
f

i
n
-
h
o
u
s
e
,

i
n

c
a
s
e

o
f

r
e
s
t
r
i
c
t
e
d

c
a
p
a
c
i
t
y

g
o

f
o
r

-
-
>

a
l
t
e
r
n
a
t
i
v
e
s

(
C
)
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

p
h
a
s
e
s
(
p
a
r
t
i
c
l
e
s
,

m
a
t
r
i
x

a
n
d

i
n
t
e
r
f
a
c
e
)
1
-
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

m
a
t
r
i
x
:
a
.

U
n
d
a
m
a
g
e
d

s
t
a
t
e
.
-
I
n
f
o
r
m
a
t
i
o
n

f
r
o
m

l
i
t
e
r
a
t
u
r
e

a
n
d

i
n
-
h
o
u
s
e

k
n
o
w
l
e
d
g
e

o
n

t
h
e

d
i
s
t
i
n
c
t
i
o
n

b
e
t
w
e
e
n

c
r
e
e
p

(
i
r
r
e
v
e
r
s
i
b
l
e

d
e
f
o
r
m
a
t
i
o
n
)

a
n
d

v
i
s
c
o
e
l
a
s
t
i
c
i
t
y

(
r
e
v
e
r
s
i
b
l
e

d
e
f
o
r
m
a
t
i
o
n
)

f
o
r

p
o
l
y
m
e
r
s
,

e
s
p
e
c
i
a
l
l
y

e
p
o
x
i
e
s
.
-
D
e
t
e
r
m
i
n
a
t
i
o
n

o
f

Y
o
u
n
g
s

m
o
d
u
l
u
s

a
s

a

f
u
n
c
t
i
o
n

o
f

t
e
m
p
e
r
a
t
u
r
e
:

t
e
n
s
i
l
e

t
e
s
t
s

a
t

A
P
P
3
-
P
o
i
s
s
o
n
s

r
a
t
i
o
:

c
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

e
p
o
x
y
-
b
a
s
e
d

a
d
h
e
s
i
v
e
s
.

D
K
I
*
f
o
r

o
p
t
i
c
a
l

s
t
r
a
i
n

m
e
a
s
u
r
e
m
e
n
t
s
.
-
V
i
s
c
o
e
l
a
s
t
i
c
i
t
y
:

D
M
A
*
,

c
h
e
c
k

c
u
r
r
e
n
t

c
a
p
a
c
i
t
i
e
s

o
f

i
n
-
h
o
u
s
e
,

i
n

c
a
s
e

o
f

r
e
s
t
r
i
c
t
e
d

c
a
p
a
c
i
t
y

g
o

f
o
r

a
l
t
e
r
n
a
t
i
v
e
s

-
-
>

a
l
t
e
r
n
a
t
i
v
e
s
.
b
-
S
t
u
d
y
i
n
g

t
h
e

r
e
v
e
r
s
i
b
l
e

a
n
d

i
r
r
e
v
e
r
s
i
b
l
e

d
e
f
o
r
m
a
t
i
o
n

o
f

t
h
e

m
a
t
r
i
x
.
-
P
l
a
n
n
i
n
g

s
o
m
e

b
a
s
i
c

e
x
p
e
r
i
m
e
n
t
s
:

l
o
a
d

c
h
a
n
g
e

e
x
p
e
r
i
m
e
n
t
s

(
s
t
e
p

l
o
a
d
)
,

l
o
a
d
i
n
g

a
n
d

u
n
l
o
a
d
i
n
g

i
n
t
e
r
v
a
l
s
,

f
o
r

v
i
s
c
o
e
l
a
s
t
i
c
i
t
y
:

o
r
i
g
i
n
a
l

s
h
a
p
e

r
e
c
o
v
e
r
e
d
,

f
o
r

c
r
e
e
p
:

p
e
r
m
a
n
e
n
t

d
e
f
o
r
m
a
t
i
o
n
.
c
-
P
e
r
f
o
r
m
i
n
g

c
y
c
l
i
c

f
a
t
i
g
u
e

t
e
s
t
s

o
n

p
u
r
e

m
a
t
r
i
x

s
p
e
c
i
m
e
n
s
.
-
S
t
r
a
i
n
-
c
o
n
t
r
o
l
l
e
d

f
a
t
i
g
u
e

t
e
s
t
s

a
t

L
B
F
,

c
o
n
s
i
d
e
r

p
a
r
a
m
e
t
e
r

s
p
a
c
e

(
t
e
m
p
e
r
a
t
u
r
e
,

f
r
e
q
u
e
n
c
y
,

R
-
v
a
l
u
e
,

a
m
p
l
i
t
u
d
e
s
,

s
a
m
p
l
e

n
u
m
b
e
r
)
-
C
h
a
r
a
c
t
e
r
i
z
e

d
a
m
a
g
e

m
e
c
h
a
n
i
s
m
,

t
h
i
n
k
i
n
g

a
b
o
u
t

c
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

m
e
t
h
o
d
s
,

c
h
e
c
k
i
n
g

T
E
M

c
a
p
a
b
i
l
i
t
i
e
s

a
t

M
P
I

S
t
u
t
t
g
a
r
t
.
2
-
P
e
r
f
o
r
m
i
n
g

c
y
c
l
i
c

f
a
t
i
g
u
e

t
e
s
t
s

o
n

n
e
w

m
a
t
e
r
i
a
l

s
a
m
p
l
e
s

-
S
t
r
a
i
n
-
c
o
n
t
r
o
l
l
e
d

f
a
t
i
g
u
e

t
e
s
t
s

a
t

L
B
F

*
,

c
o
n
s
i
d
e
r
i
n
g

s
p
a
c
e

p
a
r
a
m
e
t
e
r
s

(
t
e
m
p
e
r
a
t
u
r
e
,

f
r
e
q
u
e
n
c
y
,

R
-
v
a
l
u
e
,

a
m
p
l
i
t
u
d
e
,

s
a
m
p
l
e

n
u
m
b
e
r
)
.
-
U
s
e

L
e
m
a
i
t
r
e
-
b
a
s
e
d

d
a
m
a
g
e

m
o
d
e
l
.

C
h
e
c
k
i
n
g

T
E
M
*
c
a
p
a
b
i
l
i
t
i
e
s

a
t

M
P
I
*
S
t
u
t
t
g
a
r
t
.
2
-
D
e
f
i
n
i
n
g

d
a
m
a
g
e

i
n

t
h
e

m
i
c
r
o
m
e
c
h
a
n
i
c
s

a
p
p
r
o
a
c
h2
-
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

i
n
t
e
r
f
a
c
e
:
a
-
I
m
p
l
e
m
e
n
t
i
n
g

f
r
a
c
t
u
r
e

m
e
c
h
a
n
i
c
s

a
t

t
h
e

i
n
t
e
r
f
a
c
e

i
n

m
o
d
e
l
l
i
n
g
.
-
T
e
s
t

c
o
h
e
s
i
v
e

e
l
e
m
e
n
t
s
,

c
h
e
c
k

i
n
f
l
u
e
n
c
e

a
n
d

b
e
h
a
v
i
o
u
r
.
b
-
D
C
B

(
d
o
u
b
l
e

c
a
n
t
i
l
e
v
e
r

b
e
a
m
)

e
x
p
e
r
i
m
e
n
t
s

b
e
t
w
e
e
n

a

p
u
r
e

m
a
t
r
i
x

a
n
d

p
u
r
e

p
a
r
t
i
c
l
e

s
u
r
f
a
c
e
s
.

T
o

e
x
t
r
a
c
t

t
h
e

c
o
h
e
s
i
v
e

p
r
o
p
e
r
t
i
e
s

a
n
d

t
h
e

c
r
a
c
k

c
r
i
t
e
r
i
a

i
n

n
o
r
m
a
l

d
i
r
e
c
t
i
o
n
.
3
-
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

p
a
r
t
i
c
l
e
s
:

p
r
o
p
e
r
t
i
e
s

t
a
k
e
n

f
r
o
m

t
h
e

l
i
t
e
r
a
t
u
r
e
A
c
h
i
e
v
e
d
Y
E
S
N
O
A
n
o
t
h
e
r

A
p
p
r
o
a
c
h
!
!
!
M
i
c
r
o
m
e
c
h
a
n
i
c
s
A
p
p
r
o
a
c
h
(
D
)
E
s
t
a
b
l
i
s
h
i
n
g

a

s
i
m
p
l
e

3
D

R
e
p
r
e
s
e
n
t
a
t
i
v
e

V
o
l
u
m
e

E
l
e
m
e
n
t

(
R
V
E
)
-
O
n
e

p
a
r
t
i
c
l
e

i
n

t
h
e

c
e
n
t
e
r

o
f

a

c
u
b
e
.

-
S
i
z
e

c
o
r
r
e
s
p
o
n
d
i
n
g

t
o

t
h
e

v
o
l
u
m
e

f
r
a
c
t
u
r
e

o
f

t
h
e

c
o
n
s
t
i
t
u
e
n
t
s

(
M
a
t
r
i
x
-
P
a
r
t
i
c
l
e
)

i
n

t
h
e

n
e
w

m
a
t
e
r
i
a
l
-
P
e
r
i
o
d
i
c

b
o
u
n
d
a
r
y

c
o
n
d
i
t
i
o
n
s
.
G
O
A
L
-
D
e
r
i
v
e

p
h
e
n
o
m
e
n
o
l
o
g
i
c
a
l

d
a
m
a
g
e

m
o
d
e
l

f
r
o
m

a

m
e
s
o
s
c
a
l
e

m
o
d
e
l
.
-
C
o
r
r
e
l
a
t
e

l
i
f
e

t
i
m
e

d
a
t
a

w
i
t
h

c
a
l
c
u
l
a
t
e
d

a
n
d

m
e
a
s
u
r
e
d

d
a
m
a
g
e
.
*
L
B
F
,

F
r
a
u
n
h
o
f
e
r
-
I
n
s
t
i
t
u
t

f
r

B
e
t
r
i
e
b
s
f
e
s
t
i
g
k
e
i
t

u
n
d

S
y
s
t
e
m
z
u
v
e
r
l
s
s
i
g
k
e
i
t
-
*
T
E
M
,

T
r
a
n
s
m
i
s
s
i
o
n

E
l
e
c
t
r
o
n

M
i
c
r
o
s
c
o
p
y

-
*
M
P
I
,

M
a
x
-
P
l
a
n
c
k

I
n
s
t
i
t
u
t
e

-
*
D
M
A
,

D
y
n
a
m
i
c

M
e
c
h
a
n
i
c
a
l

A
n
a
l
y
s
i
s
*
D
C
B

D
o
u
b
l
e

C
a
n
t
i
l
e
v
e
r

B
e
a
m
G
h
a
y
a
t
h

A
L

K
A
S
S
E
M
G
h
a
y
a
t
h

A
L

K
A
S
S
E
M
G
E
N
E
R
A
L

P
L
A
N
F
L
O
W

C
H
A
R
T
(
B
)
L
i
t
e
r
a
t
u
r
e

S
e
a
r
c
h

i
n

D
a
m
a
g
e

a
n
d

F
r
a
c
t
u
r
e

M
e
c
h
a
n
i
c
s
1
-
C
r
a
c
k

p
r
o
p
a
g
a
t
i
o
n

c
r
i
t
e
r
i
a
-
T
h
r
o
u
g
h

t
h
e

m
a
t
r
i
x

i
n
t
o

t
h
e

i
n
t
e
r
f
a
c
e

a
r
o
u
n
d

t
h
e

p
a
r
t
i
c
l
e
s
.

-
O
r

o
n
l
y

t
h
r
o
u
g
h

t
h
e

m
a
t
r
i
x
.
-
O
r

o
n
l
y

d
e
l
a
m
i
n
a
t
i
o
n

i
n

t
h
e

i
n
t
e
r
f
a
c
e
.
G
e
n
e
r
a
l

P
l
a
n
(
A
)
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

n
e
w

c
o
m
p
o
s
i
t
e

m
a
t
e
r
i
a
l

1
-
T
h
e

N
e
w
M
a
t
e
r
i
a
l

p
r
o
p
e
r
t
i
e
s
:
-
D
e
t
e
r
m
i
n
a
t
i
o
n

o
f

t
h
e

Y
o
u
n
g
s

m
o
d
u
l
u
s

a
s

f
u
n
c
t
i
o
n

o
f

t
e
m
p
e
r
a
t
u
r
e
,

t
e
n
s
i
l
e

t
e
s
t
s

a
t

C
R
/
A
P
P
3

(
i
n
-
h
o
u
s
e
)
.
-
P
o
i
s
s
o
n
s

r
a
t
i
o
:

D
K
I

f
o
r

o
p
t
i
c
a
l

s
t
r
a
i
n

m
e
a
s
u
r
e
m
e
n
t
s
.
-
V
i
s
c
o
e
l
a
s
t
i
c
i
t
y
:

D
M
A
,

c
h
e
c
k

c
u
r
r
e
n
t

c
a
p
a
c
i
t
i
e
s

o
f

i
n
-
h
o
u
s
e
,

i
n

c
a
s
e

o
f

r
e
s
t
r
i
c
t
e
d

c
a
p
a
c
i
t
y

g
o

f
o
r

-
-
>

a
l
t
e
r
n
a
t
i
v
e
s

(
C
)
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

p
h
a
s
e
s
(
p
a
r
t
i
c
l
e
s
,

m
a
t
r
i
x

a
n
d

i
n
t
e
r
f
a
c
e
)
1
-
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

m
a
t
r
i
x
:
a
.

U
n
d
a
m
a
g
e
d

s
t
a
t
e
.
-
I
n
f
o
r
m
a
t
i
o
n

f
r
o
m

l
i
t
e
r
a
t
u
r
e

a
n
d

i
n
-
h
o
u
s
e

k
n
o
w
l
e
d
g
e

o
n

t
h
e

d
i
s
t
i
n
c
t
i
o
n

b
e
t
w
e
e
n

c
r
e
e
p

(
i
r
r
e
v
e
r
s
i
b
l
e

d
e
f
o
r
m
a
t
i
o
n
)

a
n
d

v
i
s
c
o
e
l
a
s
t
i
c
i
t
y

(
r
e
v
e
r
s
i
b
l
e

d
e
f
o
r
m
a
t
i
o
n
)

f
o
r

p
o
l
y
m
e
r
s
,

e
s
p
e
c
i
a
l
l
y

e
p
o
x
i
e
s
.
-
D
e
t
e
r
m
i
n
a
t
i
o
n

o
f

Y
o
u
n
g
s

m
o
d
u
l
u
s

a
s

a

f
u
n
c
t
i
o
n

o
f

t
e
m
p
e
r
a
t
u
r
e
:

t
e
n
s
i
l
e

t
e
s
t
s

a
t

A
P
P
3
-
P
o
i
s
s
o
n
s

r
a
t
i
o
:

c
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

e
p
o
x
y
-
b
a
s
e
d

a
d
h
e
s
i
v
e
s
.

D
K
I
*
f
o
r

o
p
t
i
c
a
l

s
t
r
a
i
n

m
e
a
s
u
r
e
m
e
n
t
s
.
-
V
i
s
c
o
e
l
a
s
t
i
c
i
t
y
:

D
M
A
*
,

c
h
e
c
k

c
u
r
r
e
n
t

c
a
p
a
c
i
t
i
e
s

o
f

i
n
-
h
o
u
s
e
,

i
n

c
a
s
e

o
f

r
e
s
t
r
i
c
t
e
d

c
a
p
a
c
i
t
y

g
o

f
o
r

a
l
t
e
r
n
a
t
i
v
e
s

-
-
>

a
l
t
e
r
n
a
t
i
v
e
s
.
b
-
S
t
u
d
y
i
n
g

t
h
e

r
e
v
e
r
s
i
b
l
e

a
n
d

i
r
r
e
v
e
r
s
i
b
l
e

d
e
f
o
r
m
a
t
i
o
n

o
f

t
h
e

m
a
t
r
i
x
.
-
P
l
a
n
n
i
n
g

s
o
m
e

b
a
s
i
c

e
x
p
e
r
i
m
e
n
t
s
:

l
o
a
d

c
h
a
n
g
e

e
x
p
e
r
i
m
e
n
t
s

(
s
t
e
p

l
o
a
d
)
,

l
o
a
d
i
n
g

a
n
d

u
n
l
o
a
d
i
n
g

i
n
t
e
r
v
a
l
s
,

f
o
r

v
i
s
c
o
e
l
a
s
t
i
c
i
t
y
:

o
r
i
g
i
n
a
l

s
h
a
p
e

r
e
c
o
v
e
r
e
d
,

f
o
r

c
r
e
e
p
:

p
e
r
m
a
n
e
n
t

d
e
f
o
r
m
a
t
i
o
n
.
c
-
P
e
r
f
o
r
m
i
n
g

c
y
c
l
i
c

f
a
t
i
g
u
e

t
e
s
t
s

o
n

p
u
r
e

m
a
t
r
i
x

s
p
e
c
i
m
e
n
s
.
-
S
t
r
a
i
n
-
c
o
n
t
r
o
l
l
e
d

f
a
t
i
g
u
e

t
e
s
t
s

a
t

L
B
F
,

c
o
n
s
i
d
e
r

p
a
r
a
m
e
t
e
r

s
p
a
c
e

(
t
e
m
p
e
r
a
t
u
r
e
,

f
r
e
q
u
e
n
c
y
,

R
-
v
a
l
u
e
,

a
m
p
l
i
t
u
d
e
s
,

s
a
m
p
l
e

n
u
m
b
e
r
)
-
C
h
a
r
a
c
t
e
r
i
z
e

d
a
m
a
g
e

m
e
c
h
a
n
i
s
m
,

t
h
i
n
k
i
n
g

a
b
o
u
t

c
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

m
e
t
h
o
d
s
,

c
h
e
c
k
i
n
g

T
E
M

c
a
p
a
b
i
l
i
t
i
e
s

a
t

M
P
I

S
t
u
t
t
g
a
r
t
.
2
-
P
e
r
f
o
r
m
i
n
g

c
y
c
l
i
c

f
a
t
i
g
u
e

t
e
s
t
s

o
n

n
e
w

m
a
t
e
r
i
a
l

s
a
m
p
l
e
s

-
S
t
r
a
i
n
-
c
o
n
t
r
o
l
l
e
d

f
a
t
i
g
u
e

t
e
s
t
s

a
t

L
B
F

*
,

c
o
n
s
i
d
e
r
i
n
g

s
p
a
c
e

p
a
r
a
m
e
t
e
r
s

(
t
e
m
p
e
r
a
t
u
r
e
,

f
r
e
q
u
e
n
c
y
,

R
-
v
a
l
u
e
,

a
m
p
l
i
t
u
d
e
,

s
a
m
p
l
e

n
u
m
b
e
r
)
.
-
U
s
e

L
e
m
a
i
t
r
e
-
b
a
s
e
d

d
a
m
a
g
e

m
o
d
e
l
.

C
h
e
c
k
i
n
g

T
E
M
*
c
a
p
a
b
i
l
i
t
i
e
s

a
t

M
P
I
*
S
t
u
t
t
g
a
r
t
.
2
-
D
e
f
i
n
i
n
g

d
a
m
a
g
e

i
n

t
h
e

m
i
c
r
o
m
e
c
h
a
n
i
c
s

a
p
p
r
o
a
c
h2
-
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

i
n
t
e
r
f
a
c
e
:
a
-
I
m
p
l
e
m
e
n
t
i
n
g

f
r
a
c
t
u
r
e

m
e
c
h
a
n
i
c
s

a
t

t
h
e

i
n
t
e
r
f
a
c
e

i
n

m
o
d
e
l
l
i
n
g
.
-
T
e
s
t

c
o
h
e
s
i
v
e

e
l
e
m
e
n
t
s
,

c
h
e
c
k

i
n
f
l
u
e
n
c
e

a
n
d

b
e
h
a
v
i
o
u
r
.
b
-
D
C
B

(
d
o
u
b
l
e

c
a
n
t
i
l
e
v
e
r

b
e
a
m
)

e
x
p
e
r
i
m
e
n
t
s

b
e
t
w
e
e
n

a

p
u
r
e

m
a
t
r
i
x

a
n
d

p
u
r
e

p
a
r
t
i
c
l
e

s
u
r
f
a
c
e
s
.

T
o

e
x
t
r
a
c
t

t
h
e

c
o
h
e
s
i
v
e

p
r
o
p
e
r
t
i
e
s

a
n
d

t
h
e

c
r
a
c
k

c
r
i
t
e
r
i
a

i
n

n
o
r
m
a
l

d
i
r
e
c
t
i
o
n
.
3
-
C
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n

o
f

t
h
e

p
a
r
t
i
c
l
e
s
:

p
r
o
p
e
r
t
i
e
s

t
a
k
e
n

f
r
o
m

t
h
e

l
i
t
e
r
a
t
u
r
e
A
c
h
i
e
v
e
d
Y
E
S
N
O
A
n
o
t
h
e
r

A
p
p
r
o
a
c
h
!
!
!
M
i
c
r
o
m
e
c
h
a
n
i
c
s
A
p
p
r
o
a
c
h
(
D
)
E
s
t
a
b
l
i
s
h
i
n
g

a

s
i
m
p
l
e

3
D

R
e
p
r
e
s
e
n
t
a
t
i
v
e

V
o
l
u
m
e

E
l
e
m
e
n
t

(
R
V
E
)
-
O
n
e

p
a
r
t
i
c
l
e

i
n

t
h
e

c
e
n
t
e
r

o
f

a

c
u
b
e
.

-
S
i
z
e

c
o
r
r
e
s
p
o
n
d
i
n
g

t
o

t
h
e

v
o
l
u
m
e

f
r
a
c
t
u
r
e

o
f

t
h
e

c
o
n
s
t
i
t
u
e
n
t
s

(
M
a
t
r
i
x
-
P
a
r
t
i
c
l
e
)

i
n

t
h
e

n
e
w

m
a
t
e
r
i
a
l
-
P
e
r
i
o
d
i
c

b
o
u
n
d
a
r
y

c
o
n
d
i
t
i
o
n
s
.
G
O
A
L
-
D
e
r
i
v
e

p
h
e
n
o
m
e
n
o
l
o
g
i
c
a
l

d
a
m
a
g
e

m
o
d
e
l

f
r
o
m

a

m
e
s
o
s
c
a
l
e

m
o
d
e
l
.
-
C
o
r
r
e
l
a
t
e

l
i
f
e

t
i
m
e

d
a
t
a

w
i
t
h

c
a
l
c
u
l
a
t
e
d

a
n
d

m
e
a
s
u
r
e
d

d
a
m
a
g
e
.
*
L
B
F
,

F
r
a
u
n
h
o
f
e
r
-
I
n
s
t
i
t
u
t

f
r

B
e
t
r
i
e
b
s
f
e
s
t
i
g
k
e
i
t

u
n
d

S
y
s
t
e
m
z
u
v
e
r
l
s
s
i
g
k
e
i
t
-
*
T
E
M
,

T
r
a
n
s
m
i
s
s
i
o
n

E
l
e
c
t
r
o
n

M
i
c
r
o
s
c
o
p
y

-
*
M
P
I
,

M
a
x
-
P
l
a
n
c
k

I
n
s
t
i
t
u
t
e

-
*
D
M
A
,

D
y
n
a
m
i
c

M
e
c
h
a
n
i
c
a
l

A
n
a
l
y
s
i
s
*
D
C
B

D
o
u
b
l
e

C
a
n
t
i
l
e
v
e
r

B
e
a
m
G
h
a
y
a
t
h

A
L

K
A
S
S
E
M
G
h
a
y
a
t
h

A
L

K
A
S
S
E
M
G
E
N
E
R
A
L

P
L
A
N
F
L
O
W

C
H
A
R
T
References - 137 -
References

*Many references are mentioned directly in the text
[1] Aboudi, J. (1991), Mechanics of Composite Materials - A Unified
Micromechanical Approach, Elsevier, Amsterdam, 328 p.
[2] Hashin, Z. (1983), Analysis of composite materials, a survey, J. Appl. Mech.,
vol. 50, pp. 481-505
[3] Mura, T. (1983), Micromechanics of defects in solids, 2
nd
edition. Kluwer
Academic Publishers, 587 p.
[4] Nemat-Nasser, S., and Hori, M. (1999), Micromechanics: overall properties of
heterogeneous solids. 2
nd
edition. Elsevier, Amsterdam, 810 p.
[5] Jikov, V. V., Kazlov, S. M., and Olenik, O. A. (1994) Homogenization of
differential operators and integral functionals. Springer-Verlag, 570 p.
[6] Zohdi, T. I., and Wriggers, P. (2005), Introduction to Computational
Micromechanics. Springer-Verlag, 196 p.
[7] Huet, C., 1982. Universal conditions for assimilation of a heterogeneous
material to an effective medium. Mech. Res. Comm., vol. 9(3), pp. 165170
[8] Huet, C. (1984), On the definition and experimental determination of effective
constitutive equations for heterogeneous materials. Mech. Res. Comm., vol.
11(3), pp. 195-200
[9] Maxwell, J. C. (1867), On the dynamical theory of gases. Philos. Trans. Soc.
London, 157 p.
[10] Maxwell, J. C. (1873), A treatise on electricity and magnetism, 3
rd
Ed.
Clarendon Press, Oxford, 507 p.
[11] Rayleigh, J. W. (1892), On the influence of obstacles arranged in rectangular
order upon properties of a medium. Phil. Mag., vol. 32, pp. 481-491
[12] Reuss, A. (1929), Berechnung der Fliegrenze von Mischkristallen auf Grund
der Plastizittsbedingung fr Einkristalle Z. angew. Math. Mech., vol. 9, pp.
49-58
[13] Voigt, W. (1889), ber die Beziehung zwischen den beiden
Elastizittskonstanten isotroper Krper. Wied. Ann. vol. 38, pp. 573-587.
[14] Eshelby, J. D. (1957), The elastic field of an ellipsoidal inclusion, and related
problems. Proc. Roy. Soc. A241, pp. 376-396
[15] Hill, R. (1952), The elastic behaviour of crystalline aggregate, Proc. Phys. Soc.
London A65, pp. 349-352
[16] Budiansky, B. (1965), On the elastic moduli of some heterogeneous materials,
J. Mech. Phys. Solids, vol. 13, Elsevier, pp. 223-227
[17] Hill, R. (1965), A self consistent mechanics of composite materials, J. Mech.
Phys. Solids, vol. 13, Elsevier, pp. 213-222
References - 138 -
[18] Christensen, R. (1990), A critical evaluation for a class of micromechanics
models. J. Mech. Phys. Solids, vol. 38(3), Elsevier , pp. 379-404
[19] Torquato, S. (1991), Random heterogeneous media: microstructure and
improved bounds on effective properties, Appl. Mech. Rev. 44, pp. 37-76
[20] Torquato, S. (2002), Random Heterogeneous Materials: Microstructure and
Macroscopic Properties, Springer-Verlag, New York, 701 p.
[21] Doghri, I., and Ouaar, A. (2002), Homogenization of two-phase elasto-plastic
composite materials and structures Study of tangent operators, cyclic
plasticity and numerical algorithms, Int. J. Sol. Struct., vol. 40, Elsavier, pp.
16811712
[22] Hermans, J.J. (1967), The elastic properties of fiber reinforced materials when
the fibers are aligned, Proc. Kon. Ned. Akad. v. Wetensch. B., vol. 65, pp. 1-9
[23] Hill, R. J. (1964), Theory of mechanical properties of fiber strengthened
materials: Elastic behaviour, Mech. Phys. Solids, vol. 12, pp. 199
[24] Tandon, G. P., and Weng, G. J., (1984), The effect of aspect ratio of inclusions
on the elastic properties of unidirectionally aligned composites, Polym. Comp.,
vol. 5, pp. 327-333
[25] Eshelby, J. D (1957), The determination of the elastic field of an ellipsoidal
inclusion, and related problems, Proc. Roy. Soc. A, vol. 241, pp. 376-396
[26] Mori, T., Tanaka, K. (1973), Average stress in matrix and average elastic
energy of materials with misfitting inclusions, Acta Metallurgica, vol. 21, pp.
571-574
[27] Tucker C.L., Liang E. (1999), Stiffness predictions for unidirectional short-fiber
composites: Review and evaluation, Comp. Sci. Tech., vol. 59, pp. 655-671
[28] Qiu Y. P., Wenig G. J., J. (1990, On the application of Mori-Tanaka's theory
involving transversely isotropic spheroidal inclusions, Int. J. Engng Sci., vol.
28, pp. 1121-1137
[29] Aboudi J., (1992), Mechanics of composite materials - a unified
micromechanical approach, Elsevier, Amsterdam, 338 p.
[30] Huet C. (1982), Universal conditions for assimilation of a heterogeneous
material to an effective medium, Mech. Res. Comm., vol. 9(3), pp. 165170
[31] Huet, C. (1990), Application of variational concepts to size effects in elastic
heterogeneous bodies, J. Mech. and Phys. of Solids, vol. 38, pp. 813841
[32] Torquato, S. (1991), Random heterogeneous media: microstructure and
improved bounds on effective properties, Appl. Mech. Rev., vol. 44, pp. 3776
[33] Zohdi, T., and WRIGGERS, P., (2001a), Aspects of the computational testing of
properties of microheterogeneous material samples, Inter. J. Meth. Eng., vol.
50, pp. 25732599
[34] Zohdi, T; and Wriggers, P., (2001), Computational micro-macro material
testing, Archives of Computational Methods in Engineering, vol. 8(2), pp. 131-
228
References - 139 -
[35] Dbert, C. (2001), Meso-Makromechanische Modellierung von Faserverbund-
werkstoffen mit Schdigung, Ph.D. thesis, Institut fr Baumechanik und
Numerische Mechanik, Universitt Hannover.
[36] Larsson F. & RUNESSON K. (2004), A mixed least-squares formulation for rve-
computations, To be published in Int. J. Num. Meth. Eng.
[37] Schrder J. (2000), Homogenisierungsmethoden der nichtlinearen
Kontinuumsmechanik unter Beachtung von Stabilittsproblemen. Technical
report, Habilitationsschrift, Bericht Nr.:I-7(2000) des Instituts fr Mechanik
(Bauwesen) Lehrstuhl I, Universitt Stuttgart.
[38] Miehe, C. (2003), Computational micro-to-macro transitions for discretized
micro-structures of heterogeneous materials at finite strains based on the
minimization of averaged incremental energy, Comp. Meth. App. Mech. Eng.,
vol. 192, pp. 559591
[39] Miehe, C., Schrder, J., and Becker, M. (2002). Computational homogenization
analysis in finite elasticity: material and structural instabilities on the micro-
and macro-scales of periodic composites and their interaction, Comp. Meth.
App. Mech. Eng., vol. 191, pp. 49715005
[40] Ortiz, M. (2003), Mixed continuum/atomistic models: The quasi-continuum
method, In Computational Plasticity vol. 7, pp. 12
[41] Ogden R. (1984), Non-linear Elastic Deformations, Dover, 564 p.
[42] Allix, O., and Corigliano, A., (1999), Geometrical and interfacial non-linearities
on the analysisof delamination in composites. Int. J. of Solids and Structures,
vol. 36, pp. 21892216
[43] Lielens, G. (1999), Micro-Macro Modeling of Structured Materials, Phd thesis,
Universite Catholique de Louvain, Belgium.
[44] Lielens, G., Pirotte, P., Couniot, A., Dupret, F., and Keunings, R. (1998),
Prediction of thermo-mechanical properties for compression moulded
composites. Composite part A 29, 1-2, pp. 6370
[45] Sreedhar, K., Berger, H., and Rodriguez-Ramos, R., Gabbert, U., (2007).
Computational evaluation of effective material properties of composites
reinforced by randomly distributed spherical particles. Composite Structures,
vol. 77, pp. 223231
[46] Suquet, P. (1987), Elements of homogenization theory for inelastic solid
mechanics. In: Sanchez-Palencia E., Zaoui A., editors. Homogenization
Techniques in Composite Media, Berlin: Springer-Verlag; pp. 194278
[47] Abaqus Keywords Reference Manual Version 6.5 documentation.
[48] Tucker III, C. L., and Liang, E., (1999), Stiffness predictions for unidirectional
short-fiber composites: Review and evaluation, Comp. Sc. and Tech., vol.
59(5), pp. 655-671
[49] Gusev, AA. (1997), Representative volume element size in elastic composites: a
numerical study. J. Mech. Phys. of Solids, vol. 45, pp. 144959
References - 140 -
[50] Gusev, A.A., Hine, P.J., and Ward, I.M. (2000), Fiber packing and elastic
properties of a transversely random unidirectional glass/epoxy composite.
Composites Science and Technology, vol 60, pp. 53541
[51] Mura, T. (1987), Micromechanics of defects in solids (2nd ed.), The
Netherlands, Martinus Nijhoff publishers, 614 p.
[52] Voigt (1889) Theory: Strains constant in composite, Rule of Mixtures for
stiffness components.
[53] Reuss (1929) Theory: Stresses constant in composite, Rule of Mixtures for
compliance components.
[54] Aboudi, J., (1991), Mechanics of Composite Materials, A Unified
Micromechanical Approach, Elsevier Science Publishers, Amsterdam, 328 p.
[55] Aboudi, J., (1990), Micromechanical prediction of initial and subsequent yield
surfaces of metal matrix composites, Int. J. Plasticity vol. 6, pp. 134141
[56] Benveniste, Y., (1987), A new approach to the application of Mori-Tanaka's
theory in composite materials, Mech. Mater. Vol. 6, pp. 147157
[57] Abaqus (2001), a general-purpose finite element program, Hibbitt, Karlsson &
Sorensen, Inc., Pawtucket, RI, USA.
[58] Gonzalez, C., Llorca, J., (2000), A self-consistent approach to the elasto-plastic
behaviour of two-phase materials including damage, J. Mech. Phys. Solids vol.
48, pp. 675692.
[59] Lielens, G., (1999), Micro-macro modeling of structured materials, Ph.D.
dissertation, UCL/FSA, Louvain-la-Neuve, Belgium.
[60] Wriggers, P., and Zohdi, T. I. (2005), Introduction to Computational
Micromechanics, Springer-Verlag GmbH, 196 p.
[61] Loehnert, S., (2004), Computational Homogenization of Microheterogeneous
Materials at Finite Strains Including Damage, PhD dissertation.
[62] Hill, R. (1963), Elastic properties of reinforced solids: Some theoretical
principles, J. Mech. Phys. Solids vol. 11, pp. 357372
[63] Pierard, O., Friebel, C., Doghri., I., (2004), Mean-field homogenization of multi-
phase thermo-elastic composites: a general framework and its validation.
Composites Science and Technology vol. 64 ; pp.15871603
[64] Kerner, E. H. (1956), The elastic and thermo-elastic properties of composite
media, Proc. Phys. Soc. B, vol. 69, pp. 808813
[65] Hermans, J. J., (1967), The elastic properties of fiber reinforced materials
when the fibers are aligned, Proc. Kon. Ned. Akad. v. Wetensch. B, vol. 65, pp.
19.
[66] Christensen, R. M. and Lo, K. H., (1979), Solutions for effective shear
properties in three phase sphere and cylinder models, J. Mech. Phys. Solids,
vol. 27, pp. 315330, erratum: vol. 34, pp. 639
[67] Xia et al., (2003) Xia Z., Zhang Y. and Ellyin F., A unified periodical boundary
conditions for representative volume elements of composites and applications,
Int. J. Sol. Struc. vol. 40, pp. 19071921
References - 141 -
[68] Zeman, J. (2003), Analysis of Composite Materials with Random
Microstructure, Doctoral Thesis, Czech Technical University, Prague, CZ
[69] Drugan W.J., Willis J.R. (1996), A Micromechanics-Based Nonlocal
Constitutive Equation and Estimates of Representative Volume Element Size
for Elastic Composites; J. Mech. Phys. Sol. vol. 44, pp. 497 {524, 1996}
[70] Drugan W.J. (2000), Micromechanics-Based Variational Estimates for a
Higher-Order Nonlocal Constitutive Equation and Optimal Choice of Effective
Moduli for Elastic Composites; J. Mech. Phys. Sol. vol. 48, pp. 1359 {1387,
2000}
[71] Halpin, J.C. and Kardos, J. L. (1976), Halpin-Tsai equations: A review,
Polymer Engineering and Science, vol. 16, pp. 344-352.
[72] Digimat is commercial software developed by Xstream Engineering. Companys
website: www.e-xtream.com
[73] Python programming language. Website: www.python.org
[74] Ghosh, S., Ling, Y., Majumdar, B. and Kim, R. , (2000), Interfacial debonding
analysis in multiple fiber reinforced composites, Mechanics of Materials, Vol.
32, pp. 561 591.
[75] Voronoi, G. (1907). Nouvelles applications des paramtres continus la
thorie des formes quadratiques. Journal fr die Reine und Angewandte
Mathematik, vol. 133, pp. 97-178.

From 2005 To 2009 www.ghayyath.com

3230 PDF

Hochgeladen von

Copyright:

Verfügbare Formate

Das könnte Ihnen auch gefallen

3230 PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

3230 PDF

Hochgeladen von

Copyright:

Verfügbare Formate

MICROMECHANICAL MATERIAL MODELS FOR POLYMER

COMPOSITES THROUGH ADVANCED NUMERICAL

H Single inclusion strain concentration tensor

Unit normal vector

Unit vector - orientation of the maximal principal stress

are the volume average stress and strain tensor fields

C as an expression of the effective property, while Reuss

C . A fundamental result (Hill [15], 1952) will be discussed

, the second is called alternate stress

. Both of these tensors relates the averages of

by analogy we will obtain the following relationship,

correspond to lower and upper stiffness estimates

eff eff eff eff eff eff

Das könnte Ihnen auch gefallen