Department of Mechanical Engineering

Govind Ballabh Pant Institute of Engineering and Technology

Pauri
Report on Summer Internship and Training
(MEP 020)
On the topic of

Study on Transverse Mechanical Behaviour of

Unidirectional Composites using RVE approach
Under the Supervision of
Dr Indra Vir Singh,
Professor
Department of Mechanical and Industrial Engineering

Indian Institute of Technology

Roorkee

Submitted to Submitted by
Dr Sunil Chamoli Anmol Dimri (2200408)
Assistant Professor 4th Year
1
CERTIFICATE OF COMPLETION

2
ACKNOWLEDGEMENT

3
 TABLE OF CONTENT
[Link]. Title Page Number
1. LITERATURE REVIEW

2. INTRODUCTION

1.1. UNIDIRECTIONAL COMPOSITES

1.2. RANDOM SEQUENTIAL EXPANSION
ALGORITHM (RSE)
1.3. COHESIVE ZONE MODELLING (CZM)

3. NUMERICAL MODELLING AND SIMULATION

2.1. RVE GENERATION

2.2. GEOMETRY AND MATERIAL DEFINITION
2.3. TRANSVERSE SIMULATION USING COHESIVE
ZONE MODELLING

4. RESULTS AND DISCUSSIONS

5. CONCLUSION
6. REFERENCES

4
LIST OF TABLES

5
LIST OF FIGURES

6
CHAPTER 1
INTRODUCTION
Composite materials have become increasingly important across a wide range of engineering
applications due to their impressive mechanical properties. Their ability to deliver high
strength and stiffness relative to weight makes them especially attractive when compared to
traditional materials like metals or ceramics. These benefits, combined with the flexibility
they offer in terms of tailoring performance to specific needs, make composites an ideal
choice for fields such as aerospace, automotive design, and civil engineering—areas where
both mechanical efficiency and weight reduction are critical.
At their core, most fibre-reinforced composites are built from three main ingredients: (a) the
fibres, which do most of the heavy lifting in terms of carrying loads and providing structural
rigidity; (b) the matrix, which surrounds the fibres and helps distribute stress while also
shielding them from environmental damage; and (c) the interface, a less visible but equally
important region that controls how well the fibres and matrix work together.
One of the most influential factors in this interaction is the fibre volume fraction, which refers
to the proportion of the composite's volume taken up by fibres. This parameter has a direct
impact on mechanical properties such as stiffness, tensile strength, and how much the
material can deform before it fails. To better understand this relationship, the present study
focuses on evaluating how changes in fibre volume fraction affect these properties in
unidirectional composites, especially when the material is loaded transversely—that is,
perpendicular to the direction of the fibres, where the matrix and the interface play more
dominant roles.
Rather than relying solely on physical experiments, a series of virtual tensile tests were
carried out using Representative Volume Elements (RVEs), which are small-scale models
designed to capture the microstructural details of the material. These simulations were
conducted in ABAQUS, a widely used finite element software package known for its
capability to model complex materials. Each RVE was constructed with a different fibre
volume fraction and a randomly generated fibre layout to reflect the kinds of variations and
imperfections typically found in real composite microstructures.
This simulation-based approach offered a clear view into how the composite responds to
transverse loading as the fibre content changes. The results provide a deeper understanding of
the roles played by each phase in the material and highlight how microstructural features
influence macroscopic mechanical behaviour. Such insights are especially valuable for
engineers and designers working to optimize composite structures for high-performance
applications, where getting the right balance between stiffness, strength, and weight is
essential.
1.1. Unidirectional (UD) Composites
In unidirectional fibre-reinforced composites, the reinforcing fibres are aligned uniformly
along a single direction throughout the entire length of the material. This highly ordered fibre
orientation results in superior tensile strength and stiffness along the fibre axis, making the
composite exceptionally effective in carrying loads in that direction. However, this
directional reinforcement introduces significant anisotropy; the transverse properties of the

7
composite are primarily governed by the matrix material, which is typically much weaker
than the fibres. As a result, failure mechanisms such as matrix cracking, fibre–matrix
debonding, and delamination become more pronounced under transverse or off-axis loading
conditions.
1.2. Representative Volumetric Element (RVE)
The Representative Volume Element (RVE) is a fundamental concept in multiscale
modelling, defined as the smallest volume of a heterogeneous material that can be considered
statistically representative of the entire material's microstructure. It serves as a
computationally efficient yet physically meaningful model to analyse and predict the
effective behaviour of complex materials under various physical loading conditions. In the
field of micromechanics, RVEs are extensively employed to simulate and understand the
response of composite materials, porous media such as concrete, and, in some cases, the
crystalline microstructure of metals and ceramics. By capturing essential microstructural
features—such as fibre distributions, voids, or grain orientations—the RVE facilitates
accurate estimation of macroscopic properties through numerical homogenization techniques.
1.3. Periodic Boundary Condition (PBC)
Periodic Boundary Conditions (PBCs) are widely applied in computational modelling of
heterogeneous materials to simulate a small, representative segment known as the
Representative Volume Element (RVE). Instead of analysing the entire material domain, only
this limited section is considered. PBCs ensure that the deformation on one face of the RVE
matches that on the opposite face, maintaining continuity and reflecting the repeating
structure of the material’s microstructure. This approach minimizes artificial edge effects and
enables accurate estimation of the material’s overall response.
By imposing compatibility of displacement and other mechanical variables across opposite
boundaries, PBCs allow the RVE to behave as part of an effectively infinite periodic medium.
Consequently, PBCs play a crucial role in multiscale modelling and homogenization methods
aimed at predicting macroscopic material properties from detailed microstructural
information.
1.4. Cohesive Zone Modelling (CZM)
Cohesive Zone Modelling (CZM) is an advanced computational approach employed to
simulate fracture processes in materials by representing crack initiation and propagation
through a defined cohesive zone ahead of the crack tip. Unlike traditional fracture mechanics
that treat cracks as sharp discontinuities, CZM captures the gradual degradation of material
integrity by introducing traction–separation laws that govern the relationship between stresses
transmitted across the interface and the relative displacement between adjacent surfaces.
The cohesive zone characterizes the progressive damage evolution, beginning with an elastic
response under increasing load, followed by damage initiation and softening, ultimately
culminating in complete separation once the interface strength is exceeded. This methodology
enables the accurate representation of complex failure mechanisms such as delamination in
composites, interfacial debonding in layered materials, and crack growth in metals and
ceramics without the need for explicit crack tracking or remeshing.

8
CZM has become an indispensable tool in multiscale modelling and structural analysis,
offering a robust and physically grounded framework for predicting fracture behaviour and
enhancing the design of materials and structures with improved damage tolerance.

CHAPTER 2
LITERATURE REVIEW
Previous studies have explored a range of micromechanical phenomena including interfacial
debonding, matrix cracking, and the influence of microstructural features such as fiber
geometry and volume fraction. These works serve as the foundation for the current study.
Gonzalez and Llorca (2007) presented a detailed micromechanical analysis of UD
fiber-reinforced polymers under transverse compression. Using finite element simulations,
they identified key damage mechanisms such as matrix yielding and interfacial debonding.
One of their important findings was that a representative volume containing as few as 30
fibers was sufficient to capture the essential mechanical response of the composite. This
finding directly informed the present study’s RVE design, enabling reduced computational
cost without sacrificing accuracy.
Zhao et al. (2018) combined experimental testing and micromechanical modeling to
study the strength of T300/7901 composites. Their work demonstrated that damage in these
composites initiates with interfacial debonding, followed by matrix cracking, with fibers
contributing the least to transverse failure. This sequential failure pattern closely aligns with
the results observed in the present simulations, validating the relevance of cohesive interface
modeling in transverse tensile analysis.
Jeong et al. (2020) investigated how variations in fiber cross-sectional shape—
specifically noncircular geometries—affect the transverse damage behavior of UD
composites. Through virtual testing, they showed that geometric irregularities intensify local
stress concentrations and significantly influence failure initiation. While the current study
assumes ideal circular fiber geometries, their findings support the importance of capturing
geometric realism in micromechanical models and motivate future extensions incorporating
such effects.
Yang et al. (2013) introduced an algorithm to generate random fiber distributions for
composites, ensuring non-overlapping placement while achieving desired volume fractions.
Their method, inspired by random sequential addition principles, mirrors the present study’s
implementation of a custom Python-based Random Sequential Expansion (RSE) algorithm.
This approach was used to create RVEs with physically realistic, stochastic fiber layouts—an
essential feature for simulating real-world composite behavior under transverse loading.
The ABAQUS User Guide (2020) was instrumental in defining simulation protocols,
particularly regarding unit consistency. It emphasizes that ABAQUS is a dimensionally
consistent software package, meaning users are responsible for ensuring consistent units
throughout the model. This understanding enabled accurate scaling of material properties for
micrometer-scale RVEs used in the present work.
Collectively, these studies provide a robust framework for modeling and simulating the
transverse mechanical response of UD composites. They reinforce the significance of fiber

9
volume fraction, interfacial integrity, and realistic microstructural representation in predicting
composite behavior. The present work builds upon these foundations by using cohesive zone
modeling and randomized fiber placement to explore how changes in fiber content influence
the stiffness, strength, and ductility of composites under transverse tension.

10
CHAPTER 3
NUMERICAL MODELLING AND SIMULATION
3.1. RVE Generation
In real-world applications, unidirectional (UD) composites typically exhibit random fibre
distributions due to bulk manufacturing processes such as spray-up, compression moulding,
and injection moulding. These processes do not allow for precise control over fibre
positioning, resulting in microstructural variability that significantly influences mechanical
behaviour. To replicate this randomness in simulation, custom Python scripts were developed
to automate pre-processing using a Random Sequential Expansion (RSE) algorithm. This
approach ensures non-overlapping, randomly distributed fibres within the Representative
Volume Elements (RVEs), improving model realism and reducing manual setup time.
The RSE algorithm [6] is a geometry-based method used to sequentially place random
inclusions—such as fibres—within a domain while ensuring that they do not intersect or
overlap. The process begins with an empty RVE domain. Fibres are added one at a time at
random locations. For each candidate fibre, the algorithm checks whether the new fibre
overlaps with any previously placed fibres. If there is no overlap, the fibre is accepted and
permanently added to the domain; otherwise, it is discarded and a new location is generated.
This process continues until the desired number of fibres or the target volume fraction is
achieved.
In the present work, the RSE algorithm was implemented using Python to generate fibre
geometries for different fibre volume fractions. The algorithm ensures that all fibres are
placed without overlaps while preserving randomness in location.
Three RVEs were generated of the volumetric fraction of 30%, 40% and 50% respectively.
The script was written in such a way that it would generate a different RVE each time it is
used to maintain randomness in the structure.

Fig.3.1. Geometric Models for RVEs at different volumetric fractions: 30%, 40% and 50%

3.2. Geometry and Material Definition

The RVE size was taken to be of length 10 micron, as previously proved by Gonzalez and
LLorca [3], that only 30 fibres are enough to simulate the behaviour of a composite, 30 fibres
of radii 0.56, 0.64, and 0.73 for RVEs with VF 30%, 40%, and 50%, respectively. The

11
interface thickness was assumed to be very less, nearly about the 1/70 th time of the fibre
radius [1].

Fig.2.2. Representation of Fibres (T300) and Matrix (7901)

Fig.3.3. Interface between the Fibres and Matrix

For the materials used in modelling of the fibre and matrix, we have our used mechanical
properties mentioned in the table below:
Table no.3.1. Material Properties for T300 and 7901 [4].
Material Property Fibre (T300) Matrix (7901)
E11 (GPa) 230 3.17
E22 (GPa) 15 3.17
G12 (GPa) 15 1.17
G23 (GPa) 7 1.17
V12 0.2 0.35
Xt (MPa) - 85.1
Xc (MPa) - 106.4
0
α1 (ppm/ C) -0.7 55
0
α2 (ppm/ C) 12 55
Xt = Tensile Strength, and Xc = Compressive Strength

As mentioned in the ABAQUS User Guide [6], ABAQUS is a dimension consistent software
that works on the order of units rather than having fixed units. To get the most accurate
results at the 10 microns RVE, we scaled down the properties to be in a uniform system as
mentioned in the table below:

12
Table no.3.2. Conversion of Units within Different Systems.
Material Units used for Units used for
Property Model in meters Model in microns
Length m μm
Area m2 μm2
Modulus KN/m2 KN/μm2
Stress N/m2 N/μm2
Fracture Energy J/m2 J/μm2
Stiffness KN/m3 KN/μm3

3.3. Transverse Simulation using Cohesive Zone Modelling

To simulate the behaviour of the T300/7901, two-dimensional linear plane strain elements
with reduced integration (CPE4R) were used to decrease the computational time along with
preserving accuracy of the simulation in the desired range.
In the Interface Section, ABAQUS standard 2-Dimensional 4 node Cohesive Elements
(COH2D4) were used. In the Cohesive elements the value of viscosity was set as 0.01 to
remove any possible convergence errors during the simulation.

Fig.3.4. Meshed model of the Composite with Plane Strain Elements in the fibre and matrix,
(red and green) and Cohesive Elements for the interface (blue).
In the actual transverse tensile test, the failure in the material takes place in such a way that
firstly the debonding of the fibre and matrix takes place, where the interface is damages,
followed up by damage in the matrix and the fibres are most likely the last to be affected.
Hence, they were modelled based on elasticity, and no damage law was set. This leads to
change in mechanical properties as the displacement occurs in the specimen. For the plane
strain assumption, displacement changes through the thickness were taken to be linear.
To model the debonding phenomenon in an exact manner, the interface zone between the
fibre and matrix was modelled using the Cohesive Zone Model following the Traction-
Saperation Laws, using the Quadratic Stress Damage Law, using Power law (with value of
constant = 1.2) to define the damage evolution in Interface. The parameters used for the
Quadratic Stress Damage Criterion and Damage Evolution are mentioned in table 2.3.

13
Table no.3.3. Cohesive Zone Properties of the Interface [1]
Material Property Value
KI=KII (GPa/m) 108
T0I (MPa) 44
T0II (MPa) 82
GCI (J/m2) 1.0
GCII (J/m2) 2.0

A thermal field [1] was applied to simulate the effect of the thermal stresses due to the
temperature change in the material from curing temperature, which was 120 Celsius to 20
Celsius (Room Temperature). In the Static Step I, static stabilization [5] was set as 0.001 to
ensure there are no convergence error when damage is initiated in the Interface.

Fig.2.5. Temperature Field Boundary Condition during the initial step

To simulate the transverse tensile test, periodic boundary conditions [2] were applied to the
RVE as following:
upr+ − upr- = εpqΔxqr, (p,q,r=1,2) (1)
u1r+ − u1r- = ε11X1, u12+ − u12− = 0 (2)
Where the superscripts r+ and r- represent the positive and negative sides of the RVE in a
direction, up is the displacement for the periodic pair using the constraint equation (1). Here
Δx denotes the changes in the coordinates of the nodes and ε is the strain imposed on the
RVE.
In this script, an output was printed out to show whether the periodicity is maintained along
the number of elements in each boundary of the RVE, then Boundary Conditions as
represented in fig (6) were imposed.

14
 Fig.3.6. Number of nodes on each boundary (edge) of the RVE

CHAPTER 4
RESULT AND DISCUSSSION
To ensure the reliability and accuracy of the numerical model, a convergence study was
performed by systematically varying the RVE size and fibre radius, while keeping the fibre
volume fraction constant. Throughout this process, the number of fibres within the RVE was
maintained at 30 or more, in accordance with earlier findings indicating that this number is
sufficient to capture the essential mechanical response of unidirectional composites [3]. The
objective of this study was to evaluate the sensitivity of the results to changes in geometric
parameters and to confirm that the chosen RVE dimensions are representative and
sufficiently large to eliminate size-dependency in the simulated response.

Fig.4.1. Convergence Results for model at different RVE sizes

The convergence study showed that simulation results were largely consistent when the RVE
size was varied between 5 μm and 10 μm. At 5 μm, the predicted stiffness and ultimate
strength closely matched those from the larger 10 μm model, with only minor deviations
observed at intermediate sizes like 7 μm. These results indicate that the RVE size can be
reduced by half without significantly affecting accuracy, offering a more computationally
efficient approach for micromechanical simulations.
During the transverse tensile simulations, failure was observed to initiate with interfacial
debonding, followed by matrix cracking, while the fibres remained largely unaffected
throughout the process. This indicates that the interface and matrix play dominant roles in
governing transverse failure behaviour.

15
 Fig.3.2. Equivalent Stress Contour for (30% VF) Composite
Additionally, a study was carried out to investigate the effect of increasing fibre volume
fraction on the mechanical behaviour of the composite, with particular focus on changes in
stiffness, strength, and failure strain under transverse loading.

Fig.4.3. Effect of Volumetric Fraction on the mechanical properties of UD Composite

The primary observation from this study was a clear increase in composite stiffness with
rising fibre volume fraction, with the highest stiffness recorded at 50% and the lowest at
30%. This trend confirms that in applications where high stiffness is a critical requirement,
composites with a higher fibre volume fraction are preferable due to their enhanced load-
bearing capability. Also, it was observation that, as the fibre volume fraction (VF) increased,
the failure strain of the composite decreased, indicating a more brittle response. This suggests

16
that composites with higher VF may be less suitable for applications requiring high ductility
or significant elastic deformation.

CHAPTER 5
CONCLUSION
This report presented a comprehensive investigation into the transverse mechanical behaviour
of unidirectional (UD) fibre-reinforced polymer (FRP) composites, focusing on how
variations in fibre volume fraction (VF) affect stiffness, ultimate tensile strength, and failure
strain. Recognizing that manufacturing processes such as spray-up, compression moulding,
and injection moulding result in inherently random fibre distributions, a custom Python-based
Random Sequential Expansion (RSE) algorithm was developed to generate Representative
Volume Elements (RVEs) that realistically capture this randomness.
The RSE approach enabled automated, non-overlapping fibre placement for RVEs with 30%,
40%, and 50% VF, producing statistically unique microstructures. Using 30 fibres per RVE,
in line with prior findings by Gonzalez and Llorca (2007), the models effectively simulated
localized stress concentration, matrix cracking, and interfacial degradation without requiring
large domains.
A convergence study confirmed that reducing the RVE size to 5 μm preserved accuracy, with
stiffness and strength differing by less than 2% from the 10 μm model. This validated the use
of smaller RVEs for efficient and reliable micromechanical simulation.
To model interfacial behaviour, a Cohesive Zone Model (CZM) was implemented in
ABAQUS. Simulations revealed that under transverse tensile loading, interfacial debonding
initiated failure, followed by matrix cracking, while the fibres remained structurally intact—
highlighting the dominant role of the interface and matrix in transverse failure.
Key observations include:
1. Stiffness increased by ~55% from 30% to 50% VF, affirming the fibres’ load-bearing
role.
2. Ultimate strength also improved with VF, though less sharply due to interface-limited
failure.
3. Failure strain decreased by ~30%, indicating greater brittleness at higher VF.
These results have important implications for composite design. For high-stiffness
applications, composites with higher VF are preferred. Conversely, where ductility or energy
absorption is critical, lower VF composites offer more suitable performance.

17
REFERENCES
[1] Kim D-W, Lim JH and Lee S. Prediction and validation of the transverse mechanical
behaviour of unidirectional composites considering interfacial debonding through
convolutional neural networks. Compos Part B Eng 2021; 225: 109314.
[2] Tian W, Qi L, Chao X, Liang J and Fu M. Periodic boundary condition and its
numerical implementation algorithm for the evaluation of effective mechanical
properties of the composites with complicated micro-structures. Compos Struct 2020;
245: 112370.
[3] Gonzalez C and LLorca J. Mechanical behaviour of uni directional fibre-reinforced
polymers under transverse compression: microscopic mechanisms and modelling.
Compos Sci Technol 2007; 67: 2795–2806.
[4] Zhao YQ, Zhou Y, Huang ZM and Batra RC. Experimental and micromechanical
investigation of T300/7901 unidirectional composite strength. J Reinf Plast Compos
2018; 24: 1739–1755.
[5] Jeong G, Lim JH, Choi C and Kim S-W. A virtual experimental approach to evaluate
transverse damage behaviour of a unidirectional composite considering noncircular
fibre cross-sections. Compos Struct 2020; 251: 112638.
[6] Yang L, Yan Y, Ran Z and Liu Y. A new method for generating random fibre
distributions for fibre reinforced composites. Compos Sci Technol 2013; 76: 14–20.
[7] Dassault Systèmes. Abaqus Analysis User’s Guide, Version 2020. Section 1.8.2:
Units. Providence, RI: Dassault Systèmes Simulia Corp., 2020

18