materials
Article
Thermal Study of Carbon-Fiber-Reinforced Polymer
Composites Using Multiscale Modeling
Wiem Nasri 1, * , Zied Driss 1 , Ridha Djebali 2 , Kyu-Yeon Lee 3 , Hyung-Ho Park 3 , Abderazak Bezazi 4
and Paulo N. B. Reis 5, *
Citation: Nasri, W.; Driss, Z.; Djebali,
R.; Lee, K.-Y.; Park, H.-H.; Bezazi, A.;
Reis, P.N.B. Thermal Study of
Carbon-Fiber-Reinforced Polymer
Composites Using Multiscale
Modeling. Materials 2023, 16, 7233.
https://doi.org/10.3390/
ma16227233
Academic Editor: Karim Benzarti
Received: 29 October 2023
Revised: 16 November 2023
Accepted: 17 November 2023
Published: 19 November 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Materials 2023, 16, 7233. https://doi.org/10.3390/ma16227233
1 Laboratory of Electro-Mechanic Systems (LASEM), National School of Engineers of Sfax (ENIS),
University of Sfax (US), B.P. 1173, Road Soukra km 3.5, Sfax 3038, Tunisia; zied.driss@enis.tn
2 UR 22ES12: Modeling, Optimization and Augmented Engineering, ISLAI Béja, University of Jendouba,
Béja 9000, Tunisia; ridha.djebali@islaib.u-jendouba.tn
3 Department of Materials Science and Engineering, Yonsei University, Seoul 03722, Republic of Korea;
dwrky@yonsei.ac.kr (K.-Y.L.); hhpark@yonsei.ac.kr (H.-H.P.)
4 Laboratory of Applied Mechanics of New Materials (LMANM), Université 8 Mai 1945—Guelma, B.P. 401,
Guelma 24000, Algeria; ar_bezazi@yahoo.com
5 University of Coimbra, CEMMPRE, ARISE, Department of Mechanical Engineering,
3030-788 Coimbra, Portugal
* Correspondence: wiem.nasri@enis.tn (W.N.); paulo.reis@dem.uc.pt (P.N.B.R.)
Abstract: The layered fibers of carbon-fiber-reinforced polymer (CFRP) composites exhibit low
thermal conductivity (TC) throughout their thickness due to the poor TC of the polymeric resin.
Improved heat transmission inside the hydrogen storage tank during the filling process can reduce
further compression work, and improved heat insulation can minimize energy loss. Therefore, it is
crucial to understand the thermal properties of composites. This paper reports the thermal behavior
of plain-woven CFRP composite using simulation at the micro-, meso-, and macro-scales. The TC
was predicted numerically and compared to experimental findings and analytical models. Good
results were found. Using the approach of multi-scale modeling, a parametric study was carried out
to analyze in depth the influence of certain variables on thermal properties. The study revealed that
both fiber volume fraction and temperature significantly influenced the TC of the composite, with
the interphase fiber/matrix thickness following closely in terms of impact. The matrix porosity was
found to have a relatively slighter impact, particularly within the porosity range of 5 to 15%.
Keywords: carbon-fiber-reinforced polymer; multi-scale modeling; micro-scale; woven fabric;
thermal conductivity
1. Introduction
Composite materials reinforced with carbon fiber are relatively widespread, and their
application is increasing. These high-performance composites are lightweight, robust, and
resistant to corrosion. Beyond these inherent qualities, carbon-fiber-reinforced polymer
(CFRP) distinguishes itself through exceptional fatigue and creep resistances, particularly
when subjected to harsh environmental conditions and varying loads [1–6]. These unique
attributes position CFRP as a superior choice among fiber-reinforced polymers (FRPs). In
the realm of applications, CFRP is at the forefront of diverse industries, playing a pivotal
role in aerospace, automotive, construction, and sports equipment. Its prevalence in these
sectors underscores its versatility and underscores the material’s significant impact on
advancing technology and engineering solutions. Most composites developed in previous
years were designed and manufactured to enhance mechanical features such as toughness,
stiffness, and strength. However, high thermal capacity, especially the ability to endure high
temperatures and remove heat quickly from thermal sensorial equipment or heat exchang-
ers, is becoming extremely relevant. In the literature, there is a wealth of studies on the
https://www.mdpi.com/journal/materials
Materials 2023, 16, 7233 2 of 24

mechanical characteristics of composites. Thermal assets, on the other hand, are addressed
in fewer accounts. Because the heat-conduction qualities of carbon may be modified to fit
the demands of a specific engineering application, it is among the most adaptable materials
for thermal control. Thermal conductivity (TC) is increasingly being given more thought in
order to broaden the applications of composites. When reinforced polymer composites are
employed, they face several challenges. In fact, they are frequently subjected to a quickly
changing temperature setting. As a result, thermal stress is created, which has considerable
influence on composite stiffness and strength [7]. Considering that polymer composites are
frequently thin, the simplest way to eliminate heat from a composite structure is through
its thickness direction. Nonetheless, Shim et al. [8] reported that the ratio of transverse to
in-plane thermal conductivities may be adjusted from 30 to 130, restricting heat transport
in the thickness direction. The thermal stress generated by temperature changes has been
shown to have a significant influence on material failure [9–12]. As a result, it is critical and
necessary to explore the heat conductivity of composites.
Many researchers have considered various techniques for predicting the heat transfer
and thermal properties of such composites. Most anisotropic composite material research
has concentrated on mechanical properties, with just a few models developed to estimate
heat conductivity characteristics in several directions. Mechanical parameters that have
been experimentally established can be validated using recently reported analytical models.
However, there are just a few models for TC that are available for casual verification. Most
studies predict effective thermal conductivity (ETC) across a material’s thickness. Using
the law of mixtures, however, heat conductivity along a composite’s fiber orientation may
be easily anticipated. Models that predict transverse conductivity are mostly dependent on
the geometrical configuration of the reinforcement in the investigation. The unidirectional
reinforcement models of Charles and Wilson [13] and Maxwell [14] were employed. A
crimp angle was included with these models to account for the undulation phenomena of
the analyzed woven fabric. Ning and Chou’s model [15] was employed to estimate the
transversal TC of braided composites. However, these models are valid if the effective TC
is independent of temperature. This is not the case with this study. TC is believed to be
constant in ordinary problems. Nonetheless, when the temperature difference is significant
and coupled with a change in characteristics, the influence of temperature on TC must be
considered while developing a practical problem. Currently, the mathematical formulation
of heat conduction related to temperature-dependent TC is described as nonlinear. The
strong superposition concepts of linear theory cannot be employed to develop analytical
solutions. Some researchers were able to provide analytical solutions for the non-linearity
of heat conduction problems, but they had to assume material homogeneity. Chang and
Payne [16], for instance, provided an analytical expression for heat conduction in a two-
layered block with linear temperature-dependent TC using the Kirchhoff transformation.
Many additional studies concentrated on constructing predictive models and performing
tests to assess the heat conductivity of unidirectional and particle composites.
So far, all analytical solutions have either concentrated on geometrical variables while
ignoring the role of temperature on thermal properties, or they assume material homogene-
ity and ignore the influence of geometrical factors. With the advancement of computer
computational capabilities, finite element analyses (FEA) have become frequently em-
ployed in the numerical computation of heat conduction. Li et al. [17] determined both the
in- and out-of-plane thermal conductivities of woven composites using the representative
volume element (RVE) technique with two-unit cells created at varied scales and periodic
boundary conditions. Ran et al. [18] created a numerical model to assess combined heat and
mass transmission in woven fibrous composites while taking geometrical properties into
account. Matusiak [19] established a model of woven fabric thermal resistance based on a
square cross-section of yarn. Rafiee and Salehi [20,21] and Yazdanparast and Rafiee [22]
presented two-scale and multiscale studies using “bottom-up” and “top-down” approaches
to predict the burst pressure of a wound composite vessel while taking into consideration
the influence of numerous factors on composite failure. Also, different types of RVE shapes
Materials 2023, 16, 7233 3 of 24

were discussed. It was discovered that fiber packing has little impact on the mechanical
characteristics of micro-scale RVE and that fiber contiguity and spacing have a significant
impact. Even though these approaches can accurately compute the effective properties of
woven composites, there is insufficient FEA research dealing with the influence of micro
voids, cracks, and internal defects on the thermal properties of CFRP at the microscale level.
During the early preparation of composites, flawed interfaces are frequently produced
because of chemical reactions, heat treatment, and interfacial debonding, among other
factors. This has a significant effect on the thermomechanical properties of the compos-
ite. It has been shown that the tensile stresses and potential cracks in woven composite
engineering structures typically start in the microstructure when subjected to mechan-
ical loads, cyclic internal pressure, or harsh environments. Progressive failure begins
with the introduction of voids in CFRP. Neglecting these facts would be harmful to the
regular review of industrial applications, especially when using CFRP for storing hydro-
gen in gaseous high-pressure vessels or applications where the composite is facing cyclic
load change.
While there have been numerous studies on the mechanical behavior of CFRPs, there
has been relatively less research on their thermal behavior. Most existing thermal studies
are empirical in nature and the numerical studies that do exist often overlook important
variables that affect thermal behavior. One key novelty of the present work is the incor-
poration of temperature effects, which is essential due to the nonlinear nature of CFRP
composites under varying thermal conditions. By considering temperature-dependent be-
havior, it provides a more comprehensive understanding of how thermal factors influence
the composite’s overall performance. Furthermore, various factors that significantly impact
its thermal behavior were incorporated. These factors include temperature, fiber volume
fraction, matrix porosity (voids), and interphase thickness. Moreover, two different coating
materials were used for the interphase layer, acknowledging the influence of material choice
on the thermal properties. Overall, our proposed extended finite element analysis not
only advances the understanding of CFRP’s thermal behavior but also has the potential to
enhance design considerations and practical applications in real-world industrial scenarios.
By incorporating multiple influential factors and overcoming the limitations of previous
studies, it contributes to a more comprehensive understanding of CFRP composites and
their thermal behavior.
Regarding this issue, this paper proposes to conduct an extended finite element analysis
of heat transfer of CFRP woven fabric composites to determine the effective thermal conduc-
tivities of composites using both unit cells with various length scales. The content of this paper
can be outlined as follows. First, unit cells are built in two sizes and periodic boundary condi-
tions are applied to estimate the thermal conductivities in the in-plane and through-thickness
directions of the unit cells. The model is then validated. Those results could also be applied
to unidirectional (UD) composites with fiber reinforcement. The calculated output material
characteristics of the unit cell design are then employed as input for the macro-scale model.
Consequently, the thermal properties of the whole composite can be computed. Following the
establishment of the fundamental framework of the multi-scale model for CFRP, a parametric
study is performed to evaluate the influence of various factors such as temperature, fiber
volume fraction, matrix porosity, and interphase fiber/matrix thickness on the microscale
level. The final part presents and summarizes the study’s findings.

2. Methodology
It is necessary to measure the material characteristics of each individual lamina in
order to assess components constructed from laminated composites. Micromodeling anal-
ysis involves employing unit cells to compute the homogenized material properties of a
lamina. The homogenized material properties can be calculated if the material properties
and distribution of the fibers and matrix are known. Composite materials are commonly
periodic micro-structured, such as carbon fibers. In order to reduce computational costs,
simulations of a composite material with a homogeneous domain with equivalent prop-
Materials 2023,
Materials 16,16,
2023, 7233
x FOR PEER REVIEW 4 of 25 4 of 24

and distribution
erties are used. ofThetheUD
fibers and matrix
laminas mayare
beknown.
employed Composite materials
to forecast are commonly
the thermal properties of
periodic micro-structured, such as carbon fibers. In order to reduce
complicated fabric structural composite materials using the reduced unit-cell computational costs,
model. The
simulations of a composite material with a homogeneous domain with equivalent prop-
smallest volume element of a lamina, known as RVE, is the tiniest volume element that
erties are used. The UD laminas may be employed to forecast the thermal properties of
may be repeated to define the whole lamina. A lamina is made up of fibers and matrix
complicated fabric structural composite materials using the reduced unit-cell model. The
in general. The matrix material is isotropic, while the fibers are frequently orthotropic.
smallest volume element of a lamina, known as RVE, is the tiniest volume element that
In
maythis
be work,
repeatedthe
to following material
define the whole properties
lamina. A laminaare computed
is made fromand
up of fibers a selected
matrix in unit cell
shape.
general.The
The material properties
matrix material at room
is isotropic, temperature
while are frequently
the fibers are mentioned in Table 1In[23]. The
orthotropic.
meso-microscopic scalematerial
this work, the following is represented byare
properties unit cell 1. The
computed fromlatter was built
a selected within
unit cell COMSOL
shape.
Multiphysics (version 5.2a)
The material properties andtemperature
at room includes one arefiber and four
mentioned 1/4 fibers.
in Table 1 [23]. The
The dimensions
meso- of
the unit cellscale
microscopic are 0.38 × 0.22 × by
is represented 0.8unit
mmcell
as 1.
sketched
The la er inwas
Figure
built1.within COMSOL Mul-
tiphysics (version 5.2a) and includes one fiber and four 1/4 fibers. The dimensions of the
unit cell
Table are 0.38 ×properties
1. Thermal 0.22 × 0.8 of
mm as sketched
materials in Figure
at room 1.
temperature (T = 25 ◦ C) [23].

Table 1. Thermal properties

Densityof materials at room
Thermal temperature (T
Conductivity = 25 −
(W.m 1 .K
°C) −1 )
[23].
Specific Heat Capacity
Materials (g.cm −3 ) −1 −1 )
Density ThermalλConductivity
k (W.m−1λ
.K⊥−1) Specific Heat(J.kg .K
Capacity
Materials
(g.cm−3) ∥ (J.kg−1.K−1)
Carbon fiber
Carbon fiber 1.8 10.2 1.256 750
(T700S-12K) 1.8 10.2 1.256 750
(T700S-12K)
Epoxy resin
Epoxy resin 1.13 0.18 0.18 1200
(JC-02A) 1.13 0.18 0.18 1200
(JC-02A)

Figure 1. Micro-scale hexagonal unit cell model: unit cell 1.

Figure 1. Micro-scale hexagonal unit cell model: unit cell 1.

Thefiber
The fibersize
size
in in
thethe
towtow is determined
is determined so thatsothethat the fraction
volume volumeoffraction
the fiberof andthethefiber and
the external
external diameter
diameter of the
of the tow are tow are proportional
proportional to those of to those
a real tow.ofAccording
a real tow. According
to the lit- to
the literature,
erature, a real genuine
a real genuine carbon fiber carbon fiber reinforced
reinforced composite iscomposite
transversely is isotropic
transversely
whichisotropic
which is maintained
is maintained by hexagonal
by hexagonal packing
packing because of because
the random of the randomofdistribution
distribution fibers along of fibers
along the cross-section
the cross-section of the yarnof the
[17].yarn [17]. Therefore,
Therefore, the inculcatedthe fibers
inculcated
insidefibers inside
the yarns are the
sup-yarns are
posed to beto
supposed distributed in hexagonal
be distributed arrays. There
in hexagonal arrays.are There
hundreds areof fibers in aof
hundreds single
fibers yarnin a single
with awith
yarn random distribution.
a random The shape of
distribution. Thetheshape
fibers isofcircular with aisreasonably
the fibers consistent
circular with a reasonably
diameter when
consistent viewed
diameter in cross-section.
when It is reasonableIt to
viewed in cross-section. is imagine
reasonable the to
yarns in a woven
imagine the yarns in a
composite as a UD fiber-reinforced composite, since the scale
woven composite as a UD fiber-reinforced composite, since the scale of yarn of yarn undulations is often
undulations
considerably larger than that of carbon fiber diameter. For the macro–mesoscopic scale,
is often considerably larger than that of carbon fiber diameter. For the macro–mesoscopic
unit cell 2 is shown in Figure 2 and the impact of waviness is considered. Fabric-reinforced
scale, unit cell 2 is shown in Figure 2 and the impact of waviness is considered. Fabric-
composites can be made by weaving together fluctuated warp and weft threads and after-
reinforced composites
ward impregnating them can be epoxy
with made by weaving
resin. To ensuretogether
that thefluctuated
fiber volumewarpportion
and weft re- threads
and
mained relatively stable, and the characteristics were unaffected by position, the cross- volume
afterward impregnating them with epoxy resin. To ensure that the fiber
portion remained
section form of yarnsrelatively
is intended stable,
to be and the and
elliptical characteristics
fixed. Figurewere unaffected
3 presents the macro- by position,
the
scalecross-section
model of the form of yarns
laminated is intended
composite which wasto be elliptical
created usingand fixed. Figure
SolidWorks software 3 presents
the macro-scale
(version 16.0). Both model of the
unit cells laminated
depend composite
on translational which was
symmetry. created
All plies can using
be consid-SolidWorks
ered similar
software and can16.0).
(version be replicated
Both unit by another with a simple
cells depend translation.symmetry.
on translational To be er explain All plies can
be considered similar and can be replicated by another with a simple translation. To
better explain the design of the composites, the unit cell approach was used. For the
1-layered composite, the dimensions were obtained by replicating the unit cell in the x
and y directions, four times. This approach ensured that the composite had consistent
dimensions and allowed for a representative analysis of the composite structure. Similarly,
 Materials 2023, 16, x FOR PEER REVIEW 5 of 25

Materials 2023, 16, x FOR PEER REVIEW 5 of 25

Materials 2023, 16, 7233 the design of the composites, the unit cell approach was used. For the 1-layered composite, 5 of 24
the dimensionsthe were obtained
design by replicating
of the composites, the cell
the unit unitapproach
cell in the
wasx used.
and yFordirections, fourcomposite,
the 1-layered
times. This approach ensured were
the dimensions that the composite
obtained had consistent
by replicating the unitdimensions
cell in the and
x andallowed
y directions, four
for a representative
times.analysis of the ensured
This approach composite thatstructure. Similarly,
the composite the 12-layered
had consistent compo-
dimensions and allowed
the 12-layered composite replicated the 1-layered composite 12 times in the z direction or
site replicated for
the a1-layered composite
representative 12 times
analysis of the in the z direction
composite or through
structure. Similarly,the thickness
the 12-layered compo-
through the thickness
direction. Thissite
stacking
direction.
replicated
of multiple
This
the 1-layered
identical
stacking
composite 12of multiple
times
layers allowed in
usthe zidentical
direction
to create
layers
or through
a composite
allowed us to
the thickness
with
create a composite
enhanced properties with
direction.
while enhanced
This properties
stacking of multiple
maintaining a consistent while
identical maintaining
design.layers allowed
Hence, a consistent
us to create
the design design.
of theacom- Hence,
composite with
the design
posites ofenhanced
involved the composites
properties
replicating involved
unit cells replicating
whiletomaintaining
establish the unit cells
a consistent
dimensions design.
andto Hence,
establish
stacking the dimensions
themultiple
design of the com-
and stacking multiple layers to create the desired composite architecture.
layers to create the desired composite architecture. The assumption of similar and repli- multiple
posites involved replicating unit cells to establish the dimensions and The assumption
stacking
of similar
cable andlayers
layers allowed to create
replicable
for the desired
layers
a focused composite
allowed
analysis on a for aarchitecture.
single focused
layer, while The
analysisassumption
accurately of similar
on representing
a single andwhile
layer, repli-
cable layers allowed for a focused analysis
accurately representing the overall behavior of the composites.
the overall behavior of the composites. on a single layer, while accurately representing
the overall behavior of the composites.

Figure The
Figure2.2.The mesoscale
Figure Theunit
2. unit
mesoscale cell cell
mesoscale ofunit
plain-woven composite:
cell ofcomposite:
of plain-woven plain-woven cellunit
2. cell
composite:
unit unit2.
cell 2.

(a) (b)
(a) (b)

(c) (d)
(c) (d)
Figure 3. Macro-scale composite model: (a) woven fabric; (b) matrix; (c) 1-layered composite; (d) 12-
Figure 3. Macro-scale composite model: (a) woven fabric; (b) matrix; (c) 1-layered composite;
Figure 3. layered
Macro-scale composite.
composite model: (a) woven fabric; (b) matrix; (c) 1-layered composite; (d) 12-
(d) 12-layered composite.
layered composite.
The fiber volume fraction Vy in unit cell 1 is calculated as:

4πabl
Vy = (1)
h d2
where a and b are the elliptical cross section’s long and short diameters, respectively, and l
denotes the yarn’s length. The mesoscale model’s side length and height are d and h. Table 2
summarizes the fundamental parameters. The fiber packing ratio in the yarn Vfy represents
the dispersed composition of fibers in an individual yarn. It is a crucial determinant when
Materials 2023, 16, 7233 6 of 24

calculating the dimensions of the micro-scale model. The volume fraction of the whole
composite Vf is expressed as follows:

Vf = Vfy Vy (2)

Table 2. Parameters of the meso-scale unit cell model [23].

RVE (mm) Yarn (mm)

d h Vfy a b e Vf Vy
8 1 75.3% 1.8 0.2 0.4 32.56% 43.25%

To consider the correlation between micro-scale, meso-scale, and macro-scale models

in simulations, a multiscale approach is employed. This approach aims to capture the
behavior of composite materials at different length scales and integrate them into a cohesive
simulation framework. At the micro-scale, the behavior of individual constituents, such as
fibers, matrix, and interphase, is analyzed. This can be done through detailed finite element
analysis (FEA). By examining the micro-scale behavior, important information regarding
thermal conductivity, thermal diffusivity, and other properties can be obtained. Moving
to the meso-scale, the focus shifts to the organization and arrangement of the constituents
at a larger length scale, such as the woven fabric structure or laminate configuration. The
effective properties derived from the micro-scale analysis are incorporated into an RVE or
unit cell. This meso-scale model plays a crucial role in capturing the macroscopic behavior
of the composite, including thermal transport and mechanical response. Finally, at the
macro-scale, the behavior of the composite material at the system or component level is
considered. This involves taking into account the overall geometry, loading conditions,
and boundary conditions. By integrating the information from the meso-scale model,
the macro-scale model provides predictions of the composite’s global behavior, such as
temperature distribution, thermal conductivity, and other macroscopic properties. To
establish the correlation between these models, information flows from the micro-scale
to the meso-scale and then to the macro-scale. The effective properties obtained from
the micro-scale analysis, such as thermal conductivity and diffusivity, serve as inputs for
the meso-scale model. The meso-scale model, in turn, provides the effective properties
of the composite, which are then used as inputs for the macro-scale model. Maintaining
consistency and accuracy in transferring information between the scales is crucial. Proper
homogenization techniques can be employed to determine effective properties at each
scale. Additionally, the validation and calibration of the models against experimental data
at different scales are essential for establishing correlation and ensuring the predictive
capability of the simulation framework. In summary, the correlation between micro-scale,
meso-scale, and macro-scale models in simulations is achieved by integrating information
and properties obtained at each scale. This multiscale approach allows for a comprehensive
understanding of the material behavior and enables accurate predictions of the composite’s
performance. Before carrying out heat transfer analysis, the following hypotheses are
presumed. First, the fibers inside the yarn are distributed in a regular pattern with no
overlap. The yarn is made of untwisted UD fibers joined together by epoxy resin. Second,
the yarns in the woven fabric are arranged in a periodic pattern. The interwoven yarns
are assembled and bonded with resin to constitute a layer. Finally, in order to validate
our models primordially, we assumed that the interface fiber/matrix is properly bonded
with the strength of the epoxy resin and that there are no micro voids in the matrix. In
other words, the simulation assumed an ideal scenario where the fiber/resin interface
was perfectly bonded, implying an interphase thickness of 0. This assumption allowed
to establish a baseline for the interfacial bonding performance in order to validate the
model with the work of Dong et al. [23]. Following this, variations were systematically
introduced by incrementally increasing the interphase thickness. This step was taken
to simulate the presence of an additional layer between the matrix and the fiber. The
Materials 2023, 16, 7233 7 of 24

purpose of this increment was to explore how changes in the interphase thickness influence
the overall interfacial bonding characteristics. By introducing this layered approach, we
aimed to capture the nuanced effects of interfacial interactions. As the interphase thickness
increased, it represented a more realistic depiction of the actual physical conditions within
fiber-reinforced composites. This iterative process allowed us to observe and analyze the
evolving behavior of the interface under different interphase thicknesses. The impact of
matrix porosity and interphase thickness at the microscopic level, with two carbon fiber
coating materials, will be developed in Section 5 along with the effect of temperature and
fiber volume fraction.

3. Numerical Models
3.1. Governing Equations
Heat conduction FEA were performed using the commercially available finite element
software COMSOL Multiphysics (version 5.2a). The finite element computations of micro-scale
and meso-scale models were carried out in a steady state. The governing equation is described as:

∂2 T ∂2 T ∂2 T ∂2 T ∂2 T ∂2 T
λxx 2
+ λyy 2
+ λzz 2
+ (λxy + λyx ) + (λxz + λzx ) + (λyz + λzy ) =0 (3)
∂x ∂y ∂z ∂x∂y ∂x∂z ∂y∂z
while the FEA of the full-scale model was run for transient thermal conditions as described
in the following equation:

∂2 T ∂2 T ∂2 T ∂2 T ∂2 T ∂2 T ∂T
λxx + λ yy + λzz + ( λ xy + λ yx ) + ( λxz + λ zx ) + ( λyz + λ zy ) = Cp ρ (4)
∂ x2 ∂ y2 ∂ z2 ∂x∂y ∂x∂z ∂y∂z ∂t
where T is the temperature, λxi , λyi, and λzi are the three lines of the TC tensor, and i is a
direction among x, y, and z. The carbon fiber is considered to be transversely isotropic, and
the resin matrix is isotropic. The specific heat and density, correspondingly, are Cp and q.
The constitutive relations used in FEA are identical to those found in Fourier’s law of heat
conduction and are listed below:
q = −λ ∇T (5)

3.2. Boundary Conditions

To preserve the consistency of the temperatures of the relevant nodes on the parallel
sides, periodic temperature boundaries were employed. At the opposing boundaries of the
three directions x, y, and z, a temperature offset is applied. The temperature profiles at these
opposing borders are not necessarily homogeneous, but they are pointwise equal up to a
constant offset of 1 K. The fibers and nearby unit cells should not be separated or overlapped.
Periodic boundary conditions (PBCs) were designed by Xia et al. [24] and improved by Li
Materials 2023, 16, x FOR PEER REVIEW 8 of 25
et al. [25]. Figure 4a shows the PBCs applied for the micro-scale unit cell model. The ETC for
the macro-scale model is estimated by implementing a thermal gradient along the thickness
direction and supposing the remaining surfaces are adiabatic (Figure 4b).

(a) (b)
Figure 4. Boundary conditions: (a) Periodic boundary conditions; (b) macro boundary conditions.
Figure 4. Boundary conditions: (a) Periodic boundary conditions; (b) macro boundary conditions.

3.3. Meshing
COMSOL Multiphysics was used to generate a triangular mesh. Following pre-test-
ing with various mesh sizes, the most appropriate mesh is chosen as “Finer”. For the mi-
cro-scale unit cell model, after confirming that additional refining would not change the
 (a) (b)
Materials 2023, 16, 7233 8 of 24
Figure 4. Boundary conditions: (a) Periodic boundary conditions; (b) macro boundary conditions.

3.3. Meshing
3.3. Meshing COMSOL Multiphysics was used to generate a triangular mesh. Following pre-test-
ing with various mesh sizes, the most appropriate mesh is chosen as “Finer”. For the mi-
COMSOL Multiphysics was used to generate a triangular mesh. Following pre-testing
cro-scale unit cell model, after confirming that additional refining would not change the
with varioussolution,
mesh sizes, the most
an optimal appropriate
mesh density ofmesh
15,492iselements
chosen asis“Finer”.
obtained For the micro-scale
as shown in Table 3. As
unit cell model,
for meso- and macro-scale models, the optimal number of elementssolution,
after confirming that additional refining would not change the anand
is 475,222
optimal mesh densityrespectively.
1,590,741, of 15,492 elements
The mesh is distribution
obtained asfor shown in Table 3.unit
the micro-scale Ascell
formodel
meso-isand
shown
macro-scale in
models,
Figure the
5. optimal number of elements is 475,222 and 1,590,741, respectively.
The mesh distribution for the micro-scale unit cell model is shown in Figure 5.
Table 3. Meshing.
Table 3. Meshing.Mesh Type Extra Coarse Normal Finer Extra Fine Extremely Fine
Number of Ele-
Mesh Type Extra Coarse963 Normal4903 Finer 15,492 Extra Fine 47,473Extremely 259,950
Fine
ments
Number of Elements λxx 963 0.513 4903 0.506 15,492 0.505 47,473 0.505 259,950 0.505
λxx λyy 0.513 0.503 0.506 0.505 0.505 0.504 0.505 0.504 0.505 0.504
λyy λ zz
0.503 6.439 0.505 6.507 0.504 6.508 0.504 6.508 0.504 6.508
Time of computa-
λzz 6.4391 min 15 s 6.507
1 min 22 s6.508 3 min 30 s6.508 11 min 34 s 6.50814 min 45 s
tion
Time of computation 1 min 15 s 1 min 22 s 3 min 30 s 11 min 34 s 14 min 45 s

Materials 2023, 16, x FOR PEER REVIEW 9 of 25

(a) (b) (c)

(d) (e)
Figurefor
Figure 5. Meshing 5. Meshing for micro-scale
micro-scale unit cell
unit cell model: (a)model:
extra (a) extra (b)
coarse; coarse; (b) normal;
normal; (c) (d)
(c) finer; finer; (d) extra
extra fine;
fine; (e) extremely fine.
(e) extremely fine.
4. Models Validation
4. Models Validation
4.1. Micro-Scale
4.1. Micro-Scale Model Validation
Model Validation
To confirmTo theconfirm the of
accuracy accuracy of the micro-scale
the micro-scale TC computation,
TC computation, mathematical
mathematical models
models
tailored to UD fiber-reinforced composites were used with the basic thermal
tailored to UD fiber-reinforced composites were used with the basic thermal characteristics characteris-
of the fiber tics
andofmatrix
the fiber and matrixi.e.,
employed, employed,
carbon i.e.,
fibercarbon fiber and
and epoxy epoxy
resin. resin.
The The thermal
thermal conduc-con-
ductivities of the yarns at an ambient temperature equal◦to T = 25 °C are then calculated.
tivities of the yarns at an ambient temperature equal to T = 25 C are then calculated. The
The expression of the yarn’s axial TC λya for the parallel model [23] is wri en as:
expression of the yarn’s axial TC λya for the parallel model [23] is written as:
ya  f  Vfy  m (1 Vfy ) (6)
λya = λf k Vfy + λm (1 − Vfy ) (6)
The yarn’s transversal TC λyt for the Pilling model [26] is expressed as follows:
2
 2 
 (1  Vfy ) 2   f   1   4  f   (1  Vfy )   f   1  
  m  m  m 
Materials 2023, 16, 7233 9 of 24

The yarn’s transversal TC λyt for the Pilling model [26] is expressed as follows:
r  2   2
 (1 − Vfy )2 λf⊥
λm −1 +4 λf ⊥
λm − (1 − Vfy ) λf ⊥
λm −1 
λyt = λm   (7)
 2 

The expressions for the yarn’s axial and transverse thermal conductivities of the yarn
for the Kulkarni and Brady model [27] are defined as follows:

λya = λfk Vfy + λm (1 − Vfy ) (8)

" #
λf⊥ (1 + Vfy ) + λm (1 − Vfy )
λyt = λm (9)
λf⊥ (1 − Vfy ) + λm (1 + Vfy )

where λfk , λf⊥ , λm are respectively the TC along the axial direction of the fiber, the TC
normal to the axial direction of the fiber, and the TC of the matrix. Vfy is the volume
fraction of the fiber in the yarn. Table 4 shows the values of TCs computed with COMSOL
Multiphysics compared to the values calculated using analytical models and referring to the
experimental findings of Dong et al. [23] for a volume fraction in the yarns equal to 75.3%.
A good agreement is observed with an error of less than 9%. The model of Kulkarni and
Brady [27] is the closest to the actual experimental and numerical findings, with an error
of less than 2.5%. As a result, this model can compute the yarns’ thermal conductivities
made from any material at any temperature. Figure 6a shows a comparison of our findings
with those of Dong et al. [23] for the axial and transverse thermal conductivities of yarns
as a function of temperature for a fiber volume fraction of 75.3%. The TC is found to be a
linear function of temperature, and the heat conductivity of the material is greater in the
axial direction than in the radial direction. The distribution of the temperature within the
micro-scale unit cell in three local directions is shown in Figure 7.

Table 4. Thermal conductivities of the yarns at room temperature (T = 25 ◦ C) at Vfy = 75.3% compared
to the analytical models and experimental results.

Models
Experimental [23] This Study
Parallel Pilling Kulkarni and Brady
λya 7.725 - 7.725 - 7.715
λyt - 0.723 0.646 0.664 0.662
λya 0.129 - 0.129 -
Error (%)
λyt - 8.43 2.416 0.346

4.2. Full-Scale Validation

The thermal characteristics of the yarn obtained from unit cell 1 at different tempera-
tures are included as input into unit cell 2. The FEA results are compared with the FEA
values of Dong et al. [23] and with the experimental values at temperature T = 240 ◦ C. The
values are consistent with the FEA and experimental results of Dong et al. [23], as shown
in Figure 6b. A comparison of the TC of woven composite and UD lamina can simply
determine that the woven composite has better TC than 90◦ UD lamina and lower than
0◦ UD lamina when the heat path is known. Obviously, carbon fiber tows have superior TC
compared to epoxy resin, especially through the axial direction. Therefore, heat will follow
the path of the carbon fiber. The 0◦ and 90◦ lamina resemble a plain woven composite
when they are combined. Thus, a simplification of plain-woven composites can be done by
combining UD laminas. The decomposition method is further explained by Dong et al. [23].
 Table 4. Thermal conductivities of the yarns at room temperature (T = 25 °C) at Vfy = 75.3% com-
pared to the analytical models and experimental results.

Models
Experimental [23] This Study
Parallel Pilling Kulkarni and Brady
λya 7.725 - 7.725 - 7.715
Materials 2023, 16, 7233 λyt - 0.723 0.646 0.664 0.662
10 of 24
λya 0.129 - 0.129 -
Error (%)
λyt - 8.43 2.416 0.346

λt FEA λa FEA
λt experimental [Dong et al. 2016] λa experimental [Dong et al. 2016]

(a)
λ UD 90° FEA λ UD 90° FEA [Dong et al. 2016] λ UD 90° experimental Dong et al. 2016]
λ UD 0° FEA λ UD 0° FEA [Dong et al. 2016] λ UD 0° experimental [Dong et al. 2016]
λxx FEA λxx FEA [Dong et al. 2016] λxx experimental [Dong et al. 2016]
λyy FEA λyy FEA [Dong et al. 2016] λyy experimental [Dong et al. 2016]
λzz FEA λzz FEA [Dong et al. 2016] λzz experimental [Dong et al. 2016]

(b)
Materials 2023, 16, x FOR PEER REVIEW Figure 6. Thermal 11 of 25
Figure 6. Thermal conductivity versusconductivity versus the
the temperature intemperature in the local directions
the local directions [23]: (a)[23]: (a) micro-scale
micro-scale
(carbon fiber tow); (b) macro-scale: comparison of the UD lamina with the plain woven composite.
(carbon fiber tow); (b) macro-scale: comparison of the UD lamina with the plain woven composite.

°C

(a) (b) (c)

Temperaturedistributions
Figure7.7.Temperature
Figure distributionsofofthe
the yarn
yarn in
in the
the three
three local
local coordinate
coordinatedirections:
directions:(a)
(a)TxT;x(b)
; (b)Ty ;
Ty(c)
; (c)TzT. z.

4.2. Full-Scale Validation

The thermal characteristics of the yarn obtained from unit cell 1 at different temper-
atures are included as input into unit cell 2. The FEA results are compared with the FEA
values of Dong et al. [23] and with the experimental values at temperature T = 240 °C. The
 Materials 2023, 16, 7233 11 of 24

5. Results and Discussion

5.1. Local Characteristics
In the macro-scale model, only heat transfer by conduction is considered. The latter
depends
Materials 2023, 16, x FOR PEER REVIEW basically on the TC of specific materials. Effective properties of the composite
12 of 25
can be estimated based on the results concerning simulated temperature distribution or
Materials 2023, 16, x FOR PEER REVIEW heat flux. The distribution of temperature for 1-layered and 12-layered
12 ofcomposites
25 for a
different set temperatures is shown sequentially in Figures 8 and 9.
°C °C °C

°C °C °C

(a) (b) (c)

(a) (b) (c)
°C °C °C
°C °C °C

(d) (e) (f)

(d) (e) (f)
Figure 8. Distribution of the temperature for 1-layered composite for different set temperatures and
Figure 8. Distribution ofof the temperature for 1-layered composite for different
different setset temperatures
temperaturesand
andat
at Vf = Figure
32.56%:8.
(a)Distribution
T = 40 °C; (b) ◦T the
= 100temperature for
°C; (c) T =◦ 140 1-layered
°C; composite
(d) T = ◦180 °C; (e) T =for
200
◦ °C; (f) T = 240 °C.
◦ ◦
V =f32.56%:
atf V (a)(a)
= 32.56%: TT= =4040 C;
°C;(b)
(b)TT==100 C; (c)
100 °C; (c)TT==140 C; (d)
140 °C; (d) T
T = 180 °C;
C; (e)
(e) TT ==200
200°C;
C;(f)
(f)TT==240
240°C.C.
°C °C °C
°C °C °C

(a) (b) (c)

(a) (b) (c)

Figure 9. Cont.
Materials 2023, 16, x FOR PEER REVIEW 13 of 25
Materials 2023,16,
Materials2023, 16, 7233
x FOR PEER REVIEW 13 12
of of
2524

°C °C °C
°C °C °C

(d) (e) (f)

(d) (e) (f)
Figure 9. Distribution of the temperature for 12-layered composite for different set temperatures
Figure
and
Figureat V9.
9.f =Distribution
32.56%: (a) of
Distribution Tof=the
40 temperature
the °C;
temperature for
(b) T = 100 12-layered
°C;
for composite
(c) T = 140
12-layered °C; (d) Tfor
composite different
=for
180 °C; (e)set
different =temperatures
Tset 200 °C; (f) T =
temperatures
and
240 at
°C. V f = 32.56%: (a) T = 40 °C; ◦(b) T = 100 °C; (c)
◦ T = 140 °C; (d)
◦ T = 180 °C; (e) T
◦ = 200 °C; (f) T =◦
and at V = 32.56%: (a) T = 40 C; (b) T = 100 C; (c) T = 140 C; (d) T = 180 C; (e) T = 200 C;
240 °C. f
(f) T = 240 ◦ C.
In the 1-layered composite, the temperature distribution is typically more uniform
In the 1-layered composite, the temperature distribution is typically more uniform
acrossInthe
theentire
1-layered composite,
composite and the thetemperature
temperaturegradient
distribution
withinis typically moreisuniform
the composite gener-
across the entire composite and the temperature gradient within the composite is gener-
across
ally the entire
gradual. Thiscomposite
is becauseand the the
heattemperature
flows more gradient
easily in awithin
singlethe composite
layer withoutissignificant
generally
ally gradual. This is because the heat flows more easily in a single layer without significant
gradual. Thisorisinterfaces.
interruptions because the In heat flows
contrast, more
the easily in
12-layered a single layer
composite without
introduces significant
additional in-
interruptions or interfaces. In contrast, the 12-layered composite introduces additional in-
interruptions
terfaces or interfaces.
between the layers. In Thecontrast,
presence theof12-layered composite
these interfaces introduces
can affect additional
the temperature
terfaces between the layers. The presence of these interfaces can affect the temperature
interfaces between
distribution within the layers.
composite. The Depending
presence of on these
theinterfaces can affect themismatch
thermal conductivity temperature be-
distribution within the composite. Depending on the thermal conductivity mismatch be-
tween the layers, there may be temperature variations at the interfaces due tomismatch
distribution within the composite. Depending on the thermal conductivity thermal
tween the layers, there may be temperature variations at the interfaces due to thermal
between the layers,
resistance. there may be temperature variations at the interfaces due tothe
thermal
resistance. This can result
This can result in localized
in localized temperature
temperature gradients
gradients ororhot
hotspots
spotsnear
nearthe inter-
inter-
resistance.
faces. This
Figures can
10–12 result
serve in
to localized
illustrate temperature
the temporal gradients
evolution or hot spots
profiles ofnear the interfaces.
temperature, TC,
faces. Figures 10–12 serve to illustrate the temporal evolution profiles of temperature, TC,
Figures
and 10–12
thermal serve to illustrate
diffusivity across thethe temporal
entirety of evolution
the material’sprofiles
layers,ofwith
temperature,
a specific TC, and
scenario
and thermal diffusivity across the entirety of the material’s layers, with a specific scenario
thermal diffusivity
in across the aentirety of the material’s layers, with a specific scenario in
in focus.
focus. This scenario
This scenario entails
entails prescribed
a prescribed elevated
elevated temperature,
temperature, denoted
denoted asasTT= =100100°C,
°C,
focus. This scenario entails a prescribed elevated temperature, denoted as T = 100 ◦ C, and
and
and aa fabric
fabric volume
volume fraction of
fraction of Vff == 32.56%.
32.56%.
a fabric volume fraction of Vf =V32.56%.

100
100 Layer1212
Layer
Layer1111
Layer
90
90 Layer1010
Layer
80 Layer9 9
Layer
80
Layer8 8
Layer
70
70 Layer7 7
Layer
Layer6 6
Layer
(°C)
TT (°C)

60
60 Layer5 5
Layer
Layer4 4
Layer
50
50
Layer3 3
Layer
40
40 Layer2 2
Layer
Layer1 1
Layer
30
30
20
20
00 200 400 600
600 800
800 1000
1000
t(s)
t(s)

Temporal evolution
evolutionof
ofthe
thetemperature
temperaturethrough
throughall
alllayers
layersatat
atTT
T =100
100 ◦ C.
Figure 10. Temporal
Figure evolution of the temperature through all layers hh==100
h °C.
°C.

Notably, layer 12, positioned as the exterior interface exposed to the elevated tempera-
ture, showcases distinct characteristics within this context. Anticipatedly, the curves for
the specimens’ temperature progression reflect a swift initial ascent, followed by a gradual
and stabilizing phase where the temperature attains a consistent value aligning with the
volume-averaged representation of the applied elevated temperature. An observation of
significance emerges as the temporal patterns of TC and thermal diffusivity mirror that of
Materials 2023, 16, 7233 13 of 24

the temperature curve model. This concurrence underscores the intricate interplay between
temperature, TC, and diffusivity. A discernible pattern surfaces in the transmission of heat
from one layer to the next, leading to temperature differentials that oscillate and conse-
quentially impact the TC within each layer. This dynamic interlayer interaction bestows
unique thermal behaviors upon each layer. Consequently, the sequential layers each exhibit
a distinctive TC value. Considering the composite’s homogeneity, wherein uniform layers
comprise the composition, an inherent predictability arises regarding the effective thermal
properties. In this context, the directional orientation of heat flux ceases to exert any influ-
Materials
Materials2023,
2023,16,
16,xxFOR
FORPEER
PEERREVIEW
ence on these properties. The inherent similarity in the composition and arrangement
REVIEW 14
14 of
of 25
of
25
layers thus bolsters the foreseeability of the composite’s thermal characteristics, even in the
presence of varying heat flux directions.
0.18
0.18 Layer
Layer12
12
Layer
Layer11
11
0.16 Layer
Layer10
10
0.16
Layer
Layer99
Layer
Layer88
0.14
0.14 Layer
Layer77
-1.K-1-1)
K )

Layer
Layer66
(W.m-1.

Layer
Layer55
λλ(W.m

0.12
0.12 Layer
Layer44
Layer
Layer33
0.1
0.1 Layer
Layer22
Layer
Layer11

0.08
0.08
00 200
200 400
400 600
600 800
800 1000
1000
tt(s)
(s)
Figure
Figure11. ◦ C.
Figure 11.Temporal
Temporalevolution
Temporal evolutionof
evolution ofthe
thethermal
thermalconductivity
conductivitythrough
throughall
alllayers
layersat
atTThh===100
100°C.
100 °C.

0.13
0.13 Layer
Layer12
12
Layer
Layer11
11
0.12
0.12 Layer
Layer10
10
Layer
Layer99
0.11
0.11 Layer
Layer88
Layer
Layer77
(mm22..ss-1-1))

0.1
0.1 Layer
Layer66
αα(mm

0.09 Layer
Layer55
0.09
Layer
Layer44
0.08
0.08 Layer
Layer33
Layer
Layer22
0.07
0.07 Layer
Layer11

0.06
0.06
00 200
200 400
400 600
600 800
800 1000
1000
t(s)
t(s)
Figure
Figure12.
Figure 12.Temporal
Temporalevolution
Temporal evolutionof
evolution ofthe
thethermal
thermaldiffusivity
diffusivitythrough
diffusivity throughall
alllayers
layersat
atTThhh===100
100°C.
100 ◦ C.
°C.

The rate layer

Notably,
Notably, of change
layer 12, and the ultimate
12, positioned
positioned as
as the equilibrium
the exterior
exterior values
interface
interface of TC can
exposed
exposed to provide
to the
the elevated
elevated valuable
tem-
tem-
information
perature, about
showcases the material’s
distinct ability
characteristics to conduct
within thisand distribute
context. heat
Anticipatedly,
perature, showcases distinct characteristics within this context. Anticipatedly, the curves over time.
the This
curves
information
for can be used to assess the composite’s transient thermal
for the specimens’ temperature progression reflect a swift initial ascent, followed by aa
the specimens’ temperature progression reflect a swift initial behavior,
ascent, followedheat dissi-
by
pation characteristics,
gradual
gradual and stabilizingand
and stabilizing phasesuitability
phase where
where the fortemperature
the applications awith
temperature a ains
ainsvarying temperature
aa consistent
consistent value profiles.
value aligning
aligning
As
with evident
withthe from the
thevolume-averaged graph, the
volume-averagedrepresentation TC exhibits
representationof a
ofthedistinctive
theapplied trend
appliedelevated along the
elevatedtemperature. thickness
temperature.An of the
Anobser-
obser-
composite.
vation
vationof Starting
ofsignificance from
significanceemerges the
emergesasoutermost
asthe layer,
thetemporal
temporalpa which is
pa erns
ernsoftypically
ofTC
TCand dominated
andthermal by the
thermaldiffusivity polymer
diffusivitymir- mir-
matrix,
ror that we
of observe
the a gradual
temperature decline
curve in
model. TC.
ThisThis reduction
concurrence can be associated
underscores
ror that of the temperature curve model. This concurrence underscores the intricate inter- the with
intricate the lower
inter-
TC of
play
play the matrix
between
between comparedTC,
temperature,
temperature, to and
TC, the
andcarbon fibers.
diffusivity.
diffusivity. A As we move
Adiscernible
discernible padeeper
pa ern into the
ernsurfaces
surfaces incomposite,
in the
thetrans-
trans-
mission
mission of heat from one layer to the next, leading to temperature differentials thatoscil-
of heat from one layer to the next, leading to temperature differentials that oscil-
late
late and
and consequentially
consequentially impactimpact thethe TC
TC within
within each
each layer.
layer. This
This dynamic
dynamic interlayer
interlayer inter-
inter-
action
action bestows
bestows unique
unique thermal
thermal behaviors
behaviors upon upon each
each layer.
layer. Consequently,
Consequently, the the sequential
sequential
layers each exhibit a distinctive TC value. Considering the composite’s
layers each exhibit a distinctive TC value. Considering the composite’s homogeneity, homogeneity,
 formation can be used to assess the composite’s transient thermal behavior, heat dissipa-
tion characteristics, and suitability for applications with varying temperature profiles. As
evident from the graph, the TC exhibits a distinctive trend along the thickness of the com-
posite. Starting from the outermost layer, which is typically dominated by the polymer
Materials 2023, 16, 7233 matrix, we observe a gradual decline in TC. This reduction can be associated with14 the of 24
lower TC of the matrix compared to the carbon fibers. As we move deeper into the com-
posite, closer to the central layers, where the carbon fibers are densely packed, a significant
increase in TC is observed. This sudden change suggests that carbon fibers significantly
closer to the central layers, where the carbon fibers are densely packed, a significant increase
contribute to improving heat conduction.
in TC is observed. This sudden change suggests that carbon fibers significantly contribute
Thermal diffusivity is a parameter that reflects how quickly heat is conducted and
to improving heat conduction.
diffused within a material, and in the context of CFRP composites, it plays a significant
Thermal diffusivity is a parameter that reflects how quickly heat is conducted and
role in understanding heat transfer characteristics. Obviously, the thermal diffusivity in-
diffused within a material, and in the context of CFRP composites, it plays a significant
creases as the temperature increases. The mean effective thermal diffusivity represents the
role in understanding heat transfer characteristics. Obviously, the thermal diffusivity
composite’s ability to transport heat and is typically an average value. Variations in ther-
increases as the temperature increases. The mean effective thermal diffusivity represents
mal
the diffusivity
composite’satability
different temperatures
to transport heat can
andreveal how the
is typically material’s
an average heat Variations
value. conductionin
properties change with
thermal diffusivity thermal temperatures
at different stress. Understanding
can revealthis
howdistribution is essential
the material’s for pre-
heat conduction
dicting how the composite will respond to temperature variations in real
properties change with thermal stress. Understanding this distribution is essential forapplications.
Examining
predicting this
howdistribution
the composite canwill
provide insights
respond into how heat
to temperature is conducted
variations in realthrough dif-
applications.
ferent layers of the composite at different temperatures. It illustrates the
Examining this distribution can provide insights into how heat is conducted throughanisotropic char-
acteristics of the material.
different layers of the composite at different temperatures. It illustrates the anisotropic
characteristics of the material.
5.2. Effect of the Fiber Volume Fraction
5.2. The
Effectmicro-scale
of the Fiber RVE
Volume Fraction
based primarily on the fiber content in the yarn Vfy, which in
The micro-scale
this study is 75.3%, can RVEbebased primarily
regarded on the fiber content
as UD-fiber-reinforced in the yarnThis
composites. Vfy ,scaled
which
in this study is 75.3%, can be regarded as UD-fiber-reinforced composites.
model’s thermal conductive behavior is investigated as a function of the thermal conduc- This scaled
model’s
tive thermalofconductive
parameters the fibers behavior is investigated
and the resin. Figure 13asshows
a function of the thermal
the relationship conductive
between the
parameters of the fibers and the resin. Figure 13 shows the relationship
TC and the volume fraction of fibers at room temperature equal to T = 25 °C for the micro- between the TC
and the
scale volume
model. Basedfraction of fibers
on these at room
findings, it istemperature
clear that theequal
TC to
rises 25 ◦ Cthe
T =with forincreasing
the micro-scale
vol-
model.
ume Based
fraction on these
of the fibers findings,
inside theityarn.
is clear
Thisthat
canthe TC rises with
be explained by thetheincrease
increasing volume
in the fiber
fractionand
content of the fibersininside
decrease the yarn.
the resin content This can beinexplained
resulting by in
the increase thethe
increase in the
TC of the fiber
fabric.
content and decrease in the resin content resulting in the increase in the TC of the fabric.

10
λxx λyy λzz
8
λ (W. m-1 .K-1)

0
0.4 0.5 0.6 0.7 0.8 0.9
Vfy

Figure13.
Figure Thermal conductivity
13.Thermal conductivity as
as aa function
functionof
ofthe
thevolume
volumefraction ofof
fraction fibers inside
fibers thethe
inside yarn for for
yarn unit
cell cell
unit 1. 1.

InInFigure
Figure 14,
14, aa visual
visual representation
representationisisprovided,
provided, elucidating thethe
elucidating variability in the
variability in ETC
the
exhibited by the 12-layered composite. This variation is assessed with respect to
ETC exhibited by the 12-layered composite. This variation is assessed with respect to al- alterations
in the fabric
terations volume
in the fabricfraction
volumewhile accounting
fraction for differentfor
while accounting prescribed
differentsetting temperatures.
prescribed se ing
Notably, a consistent trend becomes evident across the depicted scenarios. As foreseen,
the observed TC exhibits a clear dependency on the fabric volume fraction, aligning with
expectations surrounding this relationship. At lower temperatures (40 ◦ C and 80 ◦ C), the
curves demonstrate a gradual rise in ETC as the proportion of fabric layers increases. This
suggests an enhanced heat conduction through the fabric layers under these conditions.
During intermediate temperatures (100 ◦ C and 140 ◦ C), the curves exhibit a steeper incline,
indicating a more pronounced enhancement in ETC as the fabric volume fraction grows.
This implies that fabric layers likely play a heightened role in facilitating heat conduction
at these intermediary temperatures. Conversely, at higher temperatures (180 ◦ C and
 aligning with expectations surrounding this relationship. At lower temperatures (40 °C
and 80 °C), the curves demonstrate a gradual rise in ETC as the proportion of fabric layers
increases. This suggests an enhanced heat conduction through the fabric layers under
these conditions. During intermediate temperatures (100 °C and 140 °C), the curves ex-
Materials 2023, 16, 7233 hibit a steeper incline, indicating a more pronounced enhancement in ETC as the fabric 15 of 24
volume fraction grows. This implies that fabric layers likely play a heightened role in fa-
cilitating heat conduction at these intermediary temperatures. Conversely, at higher tem-
peratures
200 ◦ C),(180
the °C and 200
curves °C),tothe
begin curves
level off, begin to levelthat
suggesting off, the
suggesting
influencethat
ofthe influence
fabric volume
of fraction
fabric volume
on ETC diminishes as temperatures rise. This reduction in influenceincould
fraction on ETC diminishes as temperatures rise. This reduction influ-be
ence could betoa the
attributed ributed to thedominance
increasing increasing dominance of heat
of alternative alternative heat
transfer transfer mecha-
mechanisms at these
nisms at these
elevated elevated temperatures.
temperatures.

T= 40 °C T= 80 °C T= 100 °C
T= 140 °C T= 180 °C T= 200 °C
5

3
λeff (W.m-1.K-1)

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Vf
Figure 14.14.
Figure Variation in in
Variation thethe
effective thermal
effective conductivity
thermal of of
conductivity thethe
12-layered composite
12-layered versus
composite fabric
versus fabric
volume fraction for different set temperatures.
volume fraction for different set temperatures.

Some
Somecorrelations
correlationsofofthetheTC
TC versus the
the fabric
fabricvolume
volumefraction
fraction can
can be be wri as
written enaas a
linear
linear function with a correlation coefficient equal 2
to
function with a correlation coefficient equal to R = 0.98. R 2= 0.98.

ForFor 40 ◦itC,isitpossible
T=T40= °C, to obtain
is possible thethe
to obtain following expression:
following expression:
 eff
λeff =
0 .7 4 3 7 VVf +00.7921
0.7437 .7 9 2 1 (10)
(10)
f
while for T= 80 °C,◦it is:
while for T = 80 C, it is:
 eff
λeff =
0 .9 7 0 3 VVf f +11.1136
0.9703 .1 1 3 6 (11)
(11)
forfor
T=T100 °C:◦ C:
= 100
λeff = 1.1481 Vf + 1.3688 (12)
 eff  1.1481 V f  1.3688 (12)
for T = 140 ◦ C:
for T= 140 °C: λeff = 1.6226 Vf + 2.0383 (13)
for T = 180 ◦ C:  eff  1 .6 2 2 6 V f  2 .0 3 8 3 (13)
λeff = 2.2929 Vf + 2.9476 (14)
for T= 180 °C:
and for T = 200 ◦ C:
 eff  2 .2 9 2 9 V f  2 .9 4 7 6 (14)
λeff = 2.724 Vf + 3.5197 (15)
and for T= 200 °C:
5.3. Effect of Temperature
 e ff  2 .7 2 4 V f  3 .5 1 9 7 (15)
The temperature dependency of the TCs of the yarns has been empirically proven.
Figure 15 depicts the effective thermal conductivities of a yarn estimated using a micro-
scale unit cell. At low fiber volume fraction, the ETC demonstrates a relatively modest
trend across temperatures, suggesting that the material’s heat conduction efficiency is
hindered due to the scarcity of thermally conductive fibers. As the fiber content rises within
the range of 19.625% to 50.24%, there is a consistent pattern of the ETC increasing with
rising temperatures, indicating that a higher proportion of carbon fibers contributes to
improved heat conduction capabilities. At the highest fiber volume fraction of 63.59%, the
ETC reaches its peak value and maintains a steady trend across the temperature range,
possibly indicating that the material’s heat conduction capacity has saturated due to the
 Materials 2023, 16, 7233 16 of 24

high concentration of thermally conductive fibers. The equation of the ETC as a function of
the temperature and the volume fraction for carbon fiber may be written for the micro-scale
model as well as the UD carbon fiber reinforced composite as follows:

λeff = 0.0195 T + 0.216 Vfy − 0.53 (16)

where T is the temperature and Vfy is the volume fraction of fibers inside the yarn. This
equation can be used as an input for the macroscale model. It will facilitate the computation
for this kind of material. A parametric sweep combining the two variables could be used
directly by choosing for each variable an interval of values, which would give much
more reasonable and effective results. It also helps to do optimization at the level of
meso-microscopic scale, knowing that failure may develop gradually due to thermal
Materials 2023, 16, x FOR PEER REVIEW
loads without interference from mechanical stresses. In fact, thermal stress generated
18 of 25
by
temperature change can affect the stiffness and the strength of the material which can lead
to material failure.

Vfy = 12.56% Vfy = 19.625% Vfy = 28.26%

Vfy = 38.47% Vfy = 50.24% Vfy = 63.59%
6

4
λeff (W.m-1.K-1)

0
0 50 100 150 200 250 300
T (°C)

Figure 15. Variation in the effective thermal conductivity inside carbon tow as a function of tempera-
Figure 15. Variation
ture for in thevolume
different fiber effective thermal conductivity inside carbon tow as a function of temper-
fractions.
ature for different fiber volume fractions.
Figures 16 and 17 show the variation in the ETC of the 1-layered composite and the
12-layeredVcomposite,
f= 32.56% respectively, V versus temperature at different
f= 28.75% volume fractions of
Vf= 23.97%
fabric. ForVaf=volume
19.17% fraction corresponding
V1f= 15.98%to 32.56% and for the 1-layered
V−f=1 10.72% composite,
− − 1 ◦ −
the ETC increases from 0.6176 W.m .K at 40 C to 2.912 W.m .K at 200 C. For the 1 ◦

12-layeredVcomposite,
f= 4.76%
it ranges from 1.0427 W.m−1 .K−1 at 40 ◦ C to 4.445 W.m−1 .K−1 at
◦ 3
200 C. For the single-layered carbon/epoxy composite, the correlation between the ETC
and temperature at Vf = 32.56% can be written as:
2.5
λeff = 7 10−5 T2 − 0.002 T + 0.61 (17)
λeff (W.m-1. K-1)

2
and for the 12-layered carbon/epoxy composite, it can be written:
1.5
λeff = 9 10−5 T2 − 0.001 T + 0.95 (18)
1
Those two correlations are second-degree polynomials, and they give an improved TC
a correlation coefficient equal to R2 = 0.999.
with 0.5
As expected, TC increases as the volume fraction of the fibers rises. Also, it has been
observed0 that it rises as the temperature increases. As a result, the TC varies proportionally
0
to the temperature 50 volume fraction.
and fiber 100 The same pattern
150 can be noticed 200 for thermal
T (°C)
diffusivity and heat flux. Figures 18 and 19 depict the distribution of the TC and the

Figure 16. Variation in the effective thermal conductivity of 1-layered composite versus temperature
for different fabric volume fractions.
 λeff (W.m
2

1
Materials 2023, 16, 7233 17 of 24
0
0 50 100 150 200 250 300
T (°C)
thermal diffusivity over thickness for various temperatures and Vf = 32.56%. The red zones
represent the highest values, and the blue ones represent the lowest. The degradation
between
Figure 15. them shows
Variation thateffective
in the the TC thermal
and thermal diffusivity
conductivity are carbon
inside dependent
tow on
as atemperature.
function of temper-
Indeed,
ature it propagates
for different fiber progressively as the temperature increases.
volume fractions.

Vf= 32.56% Vf= 28.75% Vf= 23.97%

Vf= 19.17% Vf= 15.98% Vf= 10.72%
Vf= 4.76%
3

2.5
λeff (W.m-1. K-1)

1.5

0.5

0
0 50 100 150 200
T (°C)
Materials 2023, 16, x FOR PEER REVIEW 19 of 25
Figure 16. Variation in the effective thermal conductivity of 1-layered composite versus temperature
Figure 16. Variation
for different in the fractions.
fabric volume effective thermal conductivity of 1-layered composite versus temperature
for different fabric volume fractions.
Vf= 32.56% Vf= 28.75% Vf= 23.97%
Vf= 19.17% Vf= 15.98% Vf= 10.72%
Vf= 4.76%

3
(W.m-1.K-1)

1
λeff

0
0 50 100 150 200
T (°C)
Figure
Figure17.
17.Variation inthe
Variation in theeffective
effective thermal
thermal conductivity
conductivity of 12 layered
of 12 layered composite
composite versus tempera-
versus temperature
ture
for for different
different fabric
fabric volume
volume fractions.
fractions.

As expected, TC increases as the volume fraction of the fibers rises. Also, it has been
observed that it rises as the temperature increases. As a result, the TC varies proportion-
ally to the temperature and fiber volume fraction. The same pa ern can be noticed for
thermal diffusivity and heat flux. Figures 18 and 19 depict the distribution of the TC and
the thermal diffusivity over thickness for various temperatures and Vf = 32.56%. The red
zones represent the highest values, and the blue ones represent the lowest. The degrada-
tion between them shows that the TC and thermal diffusivity are dependent on tempera-
 observed that it rises as the temperature increases. As a result, the TC varies proportion-
ally to the temperature and fiber volume fraction. The same pa ern can be noticed for
thermal diffusivity and heat flux. Figures 18 and 19 depict the distribution of the TC and
the thermal diffusivity over thickness for various temperatures and Vf = 32.56%. The red
Materials 2023, 16, 7233 zones represent the highest values, and the blue ones represent the lowest. The degrada-18 of 24
tion between them shows that the TC and thermal diffusivity are dependent on tempera-
ture. Indeed, it propagates progressively as the temperature increases.

W.m−1.K−1 W.m−1.K−1 W.m−1.K−1

(a) (b) (c)

W.m .K
−1 −1 W.m .K
−1 −1 W.m−1.K−1

(d) (e) (f)

Figure 18. Distribution of the thermal conductivity through thickness for different temperatures
Vf = 32.56%: (a) T = 40 ◦ C; (b) T = 100 ◦ C; (c) T = 140 ◦ C; (d) T = 180 ◦ C; (e) T = 200 ◦ C; (f) T = 240 ◦ C.

5.4. Effect of the Porosity of Matrix

The presence of voids inside the epoxy matrix is one of the internal flaws that might
harm the CFRP composite. In fact, gradual failure occurs when voids are introduced into
the microstructure due to a variety of reasons, including the curing process. In this part, the
voids are supposed to be filled with air that has a TC equal to 0.026 W.m−1 .K−1 . To facilitate
the simulation, homogenization theory is employed to estimate the TC on the micro- and
macro-scale levels. The porosity ranges from 0 to 15%. Table 5 shows the variation in the TC
of yarn for different porosities and fiber volume fractions. It can be observed that the axial
and transversal thermal conductivities decrease as the porosity increases. The transverse
TC decreases for low porosity values, with an estimated rate of 11.36%, 10.7%, and 8%
for fiber volume fractions of 28.26%, 38.47%, and 63.59%, respectively. The proportion of
reduction in TC drops as the fiber volume fraction increases. However, the porosity slightly
affects the axial TC. In fact, it decreases by 0.55%, 0.35%, and 0.12% for fiber contents of
28.26%, 38.47%, and 63.59% respectively. Transverse TC is believed to be more vulnerable
to the effect of porosity than axial TC. This may be explained by the fact that as the fiber
content rises, carbon fiber becomes increasingly dominant in terms of heat conductivity.
The axial TC of carbon fiber is higher and has a stronger impact on the ETC than the
porous epoxy.
On the other hand, the impact of porosity on the ETC of woven composite and UD
laminas as a function of temperature is reported in Table 6. Obviously, the TC decreases
as the porosity increases. Also, it can be noted that the porosity has a growing impact on
TC as the temperature rises. For instance, the rate of decrease accelerates from 6 to 9%
for temperature rising from 30 to 240 ◦ C for a porosity range of 0.05 to 0.15 for woven
composite. Clearly, even with low porosity values, the TC decreases especially in the
transversal direction, which could cause internal damage to the CFRP composite.
 Materials 2023, 16, x FOR PEER REVIEW 20 of 25

Materials 2023, 16, 7233 19 of 24

Figure 18. Distribution of the thermal conductivity through thickness for different temperatures Vf
= 32.56%: (a) T = 40 °C; (b) T = 100 °C; (c) T = 140 °C; (d) T = 180 °C; (e) T = 200 °C; (f) T = 240 °C.

mm2.s−1 mm2.s−1 mm2.s−1

(a) (b) (c)

mm .s
2 −1 mm2.s−1 mm2.s−1

(d) (e) (f)

Figure 19.
Figure 19.Distribution
Distribution of the thermal
of the diffusivity
thermal through
diffusivity thickness
through for different
thickness temperatures
for different Vf =
temperatures
32.56%: (a) T = 40 °C; (b)◦ T = 100 °C; (c) ◦T = 140 °C; (d) T = 180 °C; (e) T◦= 200 °C; (f) T =◦ 240 °C.
Vf = 32.56%: (a) T = 40 C; (b) T = 100 C; (c) T = 140 C; (d) T = 180 C; (e) T = 200 C; (f) T = 240 ◦ C.
◦

5.4. Effect of the Porosity of Matrix

Table 5. Thermal conductivity of yarn for different fiber volume fractions and porosities.
The presence of voids inside the epoxy matrix is one of the internal flaws that might
Vfy 28.26% the CFRP composite. In fact,38.47%
harm gradual failure occurs when voids are introduced into
63.59%
the microstructure due to a variety of reasons, including the curing process. In this part,
ϕ λxx λyy λzz λ λyy λ λxx λ λ
the voids are supposed xx to be filled with air thatzzhas a TC equal to 0.026yyW.m−1.K−1. zz To
0 0.2772 0.2772 3.0126
facilitate the simulation,0.3272 0.3272 theory
homogenization 4.0355 0.5273
is employed 0.5273
to estimate the TC on6.5543
the
0.01 0.2752 0.2752
micro- and 3.0115
macro-scale 0.3250
levels. The0.3249 4.0345
porosity ranges 0.5246
from 0 to 15%. Table0.5246
5 shows the6.5538
vari-
0.015 0.2741 0.2741 3.0109 0.3238 0.3238 4.0340 0.5232 0.5232 6.5535
ation in the TC of yarn for different porosities and fiber volume fractions. It can be ob-
0.03 0.2710 0.2710 3.0093 0.3204 0.3203 4.0326 0.5191 0.5191 6.5526
0.05 0.2668
served that
0.2668
the axial and
3.0070
transversal
0.3157
thermal conductivities
0.3157 4.0307
decrease as
0.5136
the porosity
0.5136
in-
6.5515
0.1 0.2563 creases.
0.2563 The transverse
3.0015 TC
0.3041 decreases
0.3041for low porosity
4.0260 values,
0.4994with an estimated
0.4994 rate of
6.5487
0.15 0.2457 11.36%, 10.7%,
0.2457 2.9960 and 8% for fiber volume
0.2922 0.2922 fractions of 28.26%,
4.0212 38.47%, and
0.4847 63.59%, respec-
0.4847 6.5459
tively. The proportion of reduction in TC drops as the fiber volume fraction increases.
However, the porosity slightly affects the axial TC. In fact, it decreases by 0.55%, 0.35%,
Table 6. Thermal
and 0.12% conductivity
for fiber contents of of UD lamina
28.26%, and plain
38.47%, andwoven
63.59%composite for different
respectively. porosities
Transverse TC is and
temperatures
believed to be at more
Vf = 32.56%.
vulnerable to the effect of porosity than axial TC. This may be ex-
plained by the fact that as the fiber content rises, carbon fiber becomes increasingly dom-
ϕ 0.05
inant in terms of heat conductivity. 0.1The axial TC of carbon fiber is higher 0.15 and has a
Th (◦ C) UD 0◦ UD 90 ◦
stronger impact
λeff on the ETC
UD 0 ◦ than the ◦
porous
UD 90 epoxy.
λeff UD 0 ◦ UD 90◦ λeff
30 2.1831 0.3726 0.5856 2.1831 0.3726 0.5662 2.1831 0.3726 0.5467
40 2.1834 Table 5. Thermal
0.3729 0.6143 conductivity
2.1834 of yarn0.3729
for different fiber volume fractions
0.5936 2.1834 and porosities.
0.3729 0.5729
V80fy 2.5671 0.4496
28.26% 0.8715 2.5664 0.4494
38.47% 0.8386 2.5657 0.4492
63.59% 0.8057
100 2.8883 0.5138 1.0856 2.8875 0.5137 1.0425 2.8865 0.5136 0.9992
120 3.2549 0.5826 1.3539 3.2539 0.5825 1.2977 3.2527 0.5823 1.2413
160 3.9856 0.7193 2.0536 3.9855 0.7194 1.9634 3.9853 0.7194 1.8728
240 5.7295 1.0202 3.0371 5.7287 1.0199 2.8988 5.7277 1.0195 2.7601
Materials 2023, 16, 7233 20 of 24

5.5. Effect of the Interphase Thickness

A crucial component of composites is the fiber-matrix interfacial region since it is
where loads are transferred from the matrix to the fiber simultaneously. In composites, the
fiber-matrix interfaces are made up of a coating layer made of one or more materials that
are coated on the fiber, called interphase. Here, the impact of interphase thickness will only
be applied to the microscale model. For the interphase, two materials were used: pyrolytic
carbon and polypropylene. The thermal conductivities and densities of pyrolytic carbon
are 3.5 W.m−1 .K−1 and 1.4 g.cm−3 , respectively, and 0.13 W.m−1 .K−1 and 0.9 g.cm−3 for
polypropylene at T = 25 ◦ C. Several interphase thicknesses are selected, ranging from
0.1 to 0.4 µm. The influence of the interphase thickness on the TC of yarn for different
volume fractions in the three local directions is presented in Table 7. It can be noted that
the in-plane and out-of-plane thermal conductivities decrease with increasing interphase
thickness. Taking, for example, fiber volume fraction equal to 63.59%, it can be seen that the
axial TC slightly decreases as the interphase thickness increases with 1.15% for pyrolytic
carbon and 0.05% for polypropylene when the thickness is equal to 0.1 µm. However,
the transverse TC registered a massive drop with 65.48% for pyrolytic carbon and 56.81%
for polypropylene. For ti = 0.1 µm, the transverse TC decreases rapidly, and the rate of
decrease rises as the fiber volume fraction increases. For pyrolytic carbon, it decreases by
34.95% for Vfy = 28.26%, 54.98% for Vfy = 50.24%, and 65.48% for Vfy = 63.59%. For the
polypropylene, the transverse TC registers a drop of 32.68% for Vfy = 28.26%, 49.55% for
Vfy = 50.24%, and 56.81% for Vfy = 63.59%. On the other hand, the axial TC declines
gradually and slowly, at a rate of around 0.05% to 0.08% for every 0.05 µm for polypropylene
and less than 2% for pyrolytic carbon. It is possible to deduce that the interphase thickness
affects the composite’s axial TC less than its transverse TC.

Table 7. Yarn thermal conductivity for various fiber volume fractions and interphase thicknesses.

Pyrolytic Carbon Polypropylene

ti (µm) Vfy 28.26% 50.24% 63.59% 28.26% 50.24% 63.59%
λxx 0.2772 0.4018 0.5273 0.2772 0.4018 0.5273
0 λyy 0.2772 0.4018 0.5273 0.2772 0.4018 0.5273
λzz 3.0126 5.2162 6.5543 3.0126 5.2162 6.5543
λxx 0.1803 0.1809 0.1820 0.1866 0.2027 0.2277
0.1 λyy 0.1803 0.1809 0.1820 0.1866 0.2027 0.2277
λzz 2.9591 5.1243 6.4786 3.0101 5.2119 6.5508
λxx 0.1802 0.1806 0.1813 0.1845 0.1956 0.2133
0.15 λyy 0.1802 0.1806 0.1813 0.1845 0.1956 0.2133
λzz 2.9401 5.0911 6.4511 3.0089 5.2098 6.5491
λxx 0.1801 0.1804 0.1810 0.1834 0.1919 0.2056
0.2 λyy 0.1801 0.1804 0.1810 0.1834 0.1919 0.2056
λzz 2.9244 5.0640 6.4281 3.0076 5.2077 6.5474
λxx 0.1801 0.1803 0.1808 0.1828 0.1896 0.2007
0.25 λyy 0.1801 0.1803 0.1808 0.1828 0.1896 0.2007
λzz 2.9112 5.0411 6.4084 3.0064 5.2056 6.5457
λxx 0.1801 0.1803 0.1807 0.1823 0.1881 0.1975
0.3 λyy 0.1801 0.1803 0.1807 0.1823 0.1881 0.1975
λzz 2.9000 5.0215 6.3915 3.0053 5.2036 6.5441
λxx 0.1801 0.1802 0.1806 0.1820 0.1870 0.1951
0.35 λyy 0.1801 0.1802 0.1806 0.1820 0.1870 0.1951
λzz 2.8904 5.0046 6.3768 3.0041 5.2016 6.5424
λxx 0.1801 0.1802 0.1805 0.1817 0.1861 0.1932
0.4 λyy 0.1801 0.1802 0.1805 0.1818 0.1861 0.1932
λzz 2.8820 4.9900 6.3640 3.0029 5.1996 6.5408

Similar to previous patterns, there is a clear anisotropic behavior, with different TCs
along different axes. This indicates that heat conducts differently in various directions
Materials 2023, 16, 7233 21 of 24

within the yarn material. An interesting observation is that the ratio of TCs (λzz /λxx or
λzz /λyy ) tends to be significantly higher for pyrolytic carbon compared to polypropylene.
This implies that pyrolytic carbon exhibits a stronger anisotropic behavior, with a more
pronounced difference between TCs along different axes. These findings are valuable
for material designers and engineers. They can guide decisions on selecting appropriate
materials, fiber volume fractions, and interphase thicknesses to achieve desired TC profiles
for specific applications. For instance, pyrolytic carbon might be preferred in applications
where anisotropic thermal behavior is advantageous.

5.6. Effect Mechanism of Thermal Performance

In this section, the effect mechanism of thermal performance in the plain-woven
CFRP composites based on the simulation results is explored. In fact, the simulation
results provide valuable insights into the thermal performance of the composite. The
findings reveal that the temperature and the fiber volume fraction exhibit the highest
impact on thermal conductivity. Those two factors have important roles in enhancing the
heat transmission, which can lead to reduced compression work during the filling process,
resulting in substantial energy saving. Furthermore, the interphase thickness plays a crucial
role in minimizing energy loss and improving heat insulation. In fact, the reduction in
thermal conductivity leads to a reduction in heat loss, which contributes to improved
energy efficiency during storage and transportation of hydrogen. It is worth noting that
the porosity or the voids inside the matrix, while having the least impact, should still be
considered in optimizing the thermal performance of the composite. Indeed, reducing the
porosity level to below a certain threshold can help minimize heat loss caused by air gaps
or voids.
The principles of optimization theory can be applied to further enhance heat trans-
mission and insulation. Techniques such as Design of Experiments and response surface
methodology can be utilized to identify the optimal combination of temperature, fiber vol-
ume fraction, interphase thickness, and porosity. These optimization approaches can help
determine the specific parameter values that maximize heat transmission efficiency and
minimize energy loss within the hydrogen storage tank. By implementing these optimiza-
tion strategies and achieving the targeted improvements, significant energy savings and
enhanced thermal performance in hydrogen storage systems can potentially be achieved.
It is important to note that the actual values for heat transfer efficiency and heat loss can
vary significantly depending on the system’s characteristics and the specific optimization
strategies employed. Therefore, conducting detailed simulations or experiments tailored to
the specific system is crucial for obtaining accurate and reliable results.

6. Conclusions
A multiscale approach for evaluating the effective thermal properties of plain weave
CFRP composites is described in this study, emphasizing the interplay of key factors.
The results demonstrate that both fiber volume fraction and temperature exert significant
influence on the TC of the composite. Notably, the interphase thickness emerges as a critical
factor, closely following the impact of fiber volume fraction and temperature. Specifically,
for a fiber volume fraction in the yarn of 63.59%, a significant reduction in transverse
thermal conductivity for pyrolytic carbon (65.7%) and polypropylene (64%) when the
interphase thickness increased to 0.4 µm is observed. Nevertheless, a slight reduction in the
axial direction occurs, with less than 3% for pyrolytic carbon and 0.05% for polypropylene.
These numerical findings highlight the pronounced influence of interphase thickness on
the composite’s thermal properties. Moreover, our investigation quantifies the impact
of matrix porosity on TC, revealing a relatively slighter effect, particularly within the
porosity range of 5 to 15%. This quantitative insight provides a nuanced understanding
of how matrix porosity decreases TC. Knowing these points highlights its importance in
preventing or delaying micro-cracking in the matrix, which can affect material failure and
fatigue performance under thermal loads. In essence, this study offers comprehensive
Materials 2023, 16, 7233 22 of 24

insights into the factors influencing thermal properties, providing valuable considerations
for optimizing thermal performance in industrial applications.

Author Contributions: W.N.: Conceptualization, Data curation, Formal analysis, Investigation,

Methodology, Software, Validation, Visualization, Writing original draft; Z.D.: Supervision, Writing,
review & editing; P.N.B.R.: Software, Writing, review & editing; R.D., A.B., K.-Y.L. and H.-H.P.: Writ-
ing, review & editing. All authors have read and agreed to the published version of the manuscript.
Funding: This research was sponsored by national funds through FCT-Fundação para a Ciência e a
Tecnologia, under the projects UIDB/00285/2020 and LA/P/0112/2020. This work was also sup-
ported by the National Research Foundation of Korea (NRF) grant funded by the Korean government
(MSIT) No. 2020R1A5A1019131.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The authors declare that the data supporting the findings of this study
are available within the paper.
Acknowledgments: This research was sponsored by national funds through FCT-Fundação para a
Ciência e a Tecnologia, under the projects UIDB/00285/2020 and LA/P/0112/2020. This work was
also supported by the National Research Foundation of Korea (NRF) grant funded by the Korean
government (MSIT) No. 2020R1A5A1019131.
Conflicts of Interest: The authors declare that they have no known competing financial interest or
personal relationships that could have appeared to influence the work reported in this paper.

Nomenclature

T Temperature, ◦ C
Tc Cold temperature, ◦ C
Th Hot temperature, ◦ C
a Long axis of the yarn, m
b Short axis of the yarn, m
Width and length of the meso-scale unit cell
d
model, m
e Yarn spacing, m
h Height of the meso-scale unit cell model, m
l Length of the yarn, m
q Density of heat flow, W.m−2
V Volume fraction
t Time, s
ti Interphase thickness, m
Symbols
α Thermal diffusivity, m2 .s−1
λ Thermal conductivity, W.m−1 .K−1
ϕ Porosity
∆T Temperature difference, K
∇T Temperature gradient, K.mm−1
ρ Density, kg m−3
Cp Specific heat capacity, J.kg−1 .K−1
Subscripts
f Carbon fibers in plain woven composite
m Epoxy resin matrix
eff Effective
y Yarn
fy Fiber in yarn
xx Across the heat flow path
yy Normal to the heat flow path
zz Out-of- plane direction
Materials 2023, 16, 7233 23 of 24

Parallel to the axial direction of the fiber

⊥ Normal to the axial direction of the fiber
ya Axial direction of the yarn
yt transverse direction of the yarn
in In-plane direction
out out-of-plane direction (thickness)
Abbreviation
CFRP Carbon-fiber-reinforced polymer
ETC Effective thermal conductivity
FEA Finite Element Analyses
RVE Representative Volume Element
TC Thermal conductivity
UD Unidirectional lamina

References
1. Wu, J.; Zhu, Y.; Li, C. Experimental Investigation of Fatigue Capacity of Bending-Anchored CFRP Cables. Polymers 2023, 15, 2483.
[CrossRef]
2. Naresh, K.; Shankar, K.; Velmurugan, R.; Gupta, N.K. Statistical analysis of the tensile strength of GFRP, CFRP and hybrid
composites. Thin-Walled Struct. 2018, 126, 150–161. [CrossRef]
3. Xian, G.; Guo, R.; Li, C. Combined effects of sustained bending loading, water immersion and fiber hybrid mode on the
mechanical properties of carbon/glass fiber reinforced polymer composite. Compos. Struct. 2022, 281, 115060. [CrossRef]
4. Wysmulski, P. The analysis of buckling and post buckling in the compressed composite columns. Arch. Mater. Sci. 2017, 36, 36.
[CrossRef]
5. Wysmulski, P. The effect of load eccentricity on the compressed CFRP Z-shaped columns in the weak post-critical state. Compos.
Struct. 2022, 301, 116184. [CrossRef]
6. Rozylo, P. Comparison of failure for thin-walled composite columns. Materials 2021, 15, 167. [CrossRef]
7. Dong, K.; Gu, B.; Sun, B. Comparisons of thermal conductive behaviors of epoxy resin in unidirectional composite materials. J.
Therm. Anal. Calorim. 2015, 124, 775–789. [CrossRef]
8. Shim, H.B.; Seo, M.K.; Park, S.J. Thermal conductivity and mechanical properties of various cross-section types carbon fiber-
reinforced composites. J. Mater. Sci. 2002, 37, 1881–1885. [CrossRef]
9. Jeon, J.; Muliana, A.; La Saponara, V. Thermal stress and deformation analyses in fiber reinforced polymer composites undergoing
heat conduction and mechanical loading. Compos. Struct. 2014, 111, 31–44. [CrossRef]
10. Aboudi, J. The temperature field induced by the dynamic stresses of a transverse crack in periodically layered composites. Int. J.
Eng. Sci. 2011, 49, 732–754. [CrossRef]
11. Deierling, P.; Zhupanska, O.I. Experimental study of high electric current effects in carbon/epoxy composites. Compos. Sci.
Technol. 2011, 71, 1659–1664. [CrossRef]
12. Ogasawara, T.; Hirano, Y.; Yoshimura, A. Coupled thermal–electrical analysis for carbon fiber/epoxy composites exposed to
simulated lightning current. Compos. Part A Appl. Sci. Manuf. 2010, 41, 973–981. [CrossRef]
13. Charles, J.A.; Wilson, D.W. Delaware univ newark center for composite materials. In A Model of Passive Thermal Nondestructive
Evaluation of Composite Laminates; Center for Composite Materials: Washington, DC, USA, 1980.
14. Achard, F.; James, C.M. A Treatise on Electricity and Magnetism, 1st ed.; Elsevier eBooks: Amsterdam, The Netherlands, 1873;
Volume 2005, pp. 564–587. [CrossRef]
15. Ning, Q.G.; Chou, T. A Closed-Form solution of the transverse effective thermal conductivity of woven fabric composites. J.
Compos. Mater. 1995, 29, 2280–2294. [CrossRef]
16. Chang, K.; Payne, U.J. Analytical solution for heat conduction in a Two-Material-Layer slab with linearly temperature dependent
conductivity. J. Heat Transf. 1991, 113, 237–239. [CrossRef]
17. Li, H.; Li, S.; Wang, Y. Prediction of effective thermal conductivities of woven fabric composites using unit cells at multiple length
scales. J. Mater. Res. 2011, 26, 384–394. [CrossRef]
18. Ran, X.; Zhu, Q.; Li, Y. Investigation on heat and mass transfer in 3D woven fibrous material. Int. J. Heat Mass Transf. 2011,
54, 3575–3586. [CrossRef]
19. Matusiak, M. Modelling the thermal resistance of woven fabrics. J. Text. Inst. 2013, 104, 426–437. [CrossRef]
20. Rafiee, R.; Salehi, A. A novel recursive multi-scale modeling for predicting the burst pressure of filament wound composite
pressure vessels. Appl. Phys. A 2022, 128, 388. [CrossRef]
21. Rafiee, R.; Salehi, A. Estimating the burst pressure of a filament wound composite pressure vessel using two-scale and multi-scale
analyses. Mech. Adv. Mater. Struct. 2022, 30, 2668–2683. [CrossRef]
22. Yazdanparast, R.; Rafiee, R. Determining in-plane material properties of square core cellular materials using computational
homogenization technique. Eng. Comput. 2022, 39, 373–386. [CrossRef]
23. Dong, K.; Liu, K.; Zhang, Q.; Gu, B.; Sun, B. Experimental and numerical analyses on the thermal conductive behaviors of carbon
fiber/epoxy plain woven composites. Int. J. Heat Mass Transf. 2016, 102, 501–517. [CrossRef]
Materials 2023, 16, 7233 24 of 24

24. Xia, Z.; Zhang, Y.; Ellyin, F. A unified periodical boundary conditions for representative volume elements of composites and
applications. Int. J. Solids Struct. 2003, 40, 1907–1921. [CrossRef]
25. Li, S. Boundary conditions for unit cells from periodic microstructures and their implications. Compos. Sci. Technol. 2008,
68, 1962–1974. [CrossRef]
26. Pilling, M.; Yates, B.; Black, M.; Tattersall, P. The thermal conductivity of carbon fiber-reinforced composites. J. Mater. Sci. 1979,
14, 1326–1338. [CrossRef]
27. Kulkarni, M.; Brady, R.P. A model of global thermal conductivity in laminated carbon/carbon composites. Compos. Sci. Technol.
1997, 57, 277–285. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.