Accepted Manuscript

Analytical model for predicting the tensile strength of unidirectional composites based
on the density of fiber breaks

J.D. Vanegas-Jaramillo, A. Turon, J. Costa, L.J. Cruz, J.A. Mayugo

PII: S1359-8368(17)30696-0
DOI: 10.1016/j.compositesb.2017.12.012
Reference: JCOMB 5441

To appear in: Composites Part B

Received Date: 24 February 2017

Revised Date: 6 November 2017
Accepted Date: 15 December 2017

Please cite this article as: Vanegas-Jaramillo JD, Turon A, Costa J, Cruz LJ, Mayugo JA, Analytical
model for predicting the tensile strength of unidirectional composites based on the density of fiber
breaks, Composites Part B (2018), doi: 10.1016/j.compositesb.2017.12.012.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Analytical model for predicting the tensile strength of unidirectional composites

based on the density of fiber breaks

J. D. Vanegas-Jaramillo1-2-3*, A. Turon1, J. Costa1, L. J. Cruz2, J.A.Mayugo1

1
AMADE - Polytechnic School, University of Girona, Girona, Spain.
2
Grupo de Investigación Sobre Nuevos Materiales, GINUMA, Universidad Pontificia Bolivariana,

PT
Medellín, Colombia
3
Grupo de Investigación Materiales y Energía, MATyER, Instituto Tecnológico Metropolitano, Medellín,
Colombia.

RI
*corresponding author email: juan.vanegas@udg.edu, Tel: (+57) 312 722 7380

SC
Abstract

While analytical fiber fragmentation models following the global load-sharing (GLS) assumption

U
efficiently reproduce the stress strain curves of unidirectional composites loaded in the direction of the
AN
reinforcement when the number of breaks is moderate, they completely fail to predict tensile strength. In

this paper, we propose that failure takes place when a critical density of breaks, which depends entirely on
M

the constituent properties, is reached. Therefore, we rewrite classic GLS fragmentation models in terms of

the linear density of breaks. The critical number of breaks for a set of glass and carbon reinforced
D

polymer composites is extracted from published experimental data and fitted to an empirical law, with
TE

good predictive capability. Our approach complements GLS fragmentation models because it identifies

the ultimate stress, from which the stress-strain curve given by the model becomes unrealistic.
EP

Keywords

A: Polymeric-matrix composites (PMCs); B: Fragmentation; C: Analytical modeling; D: Strength.

C
AC

1 Introduction

Physically-based analytical approaches to anticipate the mechanical properties of composite

laminates are fast tools for structural design and permit their constituents to be optimized for a better

mechanical performance. Analytical models for the stress-strain behavior of unidirectional plies in the

direction of the reinforcement rely on the representation of the stochastic occurrence of breaks along the

fibers and, therefore, are referred to as fragmentation models [1,2].

1
 ACCEPTED MANUSCRIPT

Fragmentation models assume a stochastic distribution of defects along the fibers [3], (the Weibull

distribution is most commonly used), as well as a certain rule of load sharing from a broken fiber to the

rest of the composite. A useful baseline of results has been obtained by assuming that stress re-

distribution around fiber breaks satisfies the global load-sharing (GLS) assumption (i.e. that the load is

spread evenly among all the fibers in the composite) [4–6]. Under this condition, there is no interaction

PT
between fiber breaks of different fibers. These models aim to relate the remote uniaxial stress, , to the

strain, . Due to the stochastic nature of the phenomenon and the large number of fibers in the composite,

RI
it is assumed that the average stress built up along the fragmented fiber matches the average stress in a

cross-section of the composite [5,7].

SC
For instance, Curtin [8] deduced an analytical stress-strain solution by expanding the Weibull

distribution of the fiber strength with a Taylor series. Curtin’s model was later extended by Neumeister

U
to consider the overlap of influence zones adjacent to fiber breaks [9,10]. Later, Hui and collaborators
AN
developed an exact solution for the fiber fragmentation problem [11], which needs to be solved

iteratively. Furthermore, Turon et al. [12] developed a progressive damage model for unidirectional
M

composites laminates based on fiber fragmentation. These models show very similar behavior in the

stress-strain curves up to the maximum of the curve. From this point on, the curves deviate because of the
D

different micro-mechanisms considered (fiber-matrix pullout and sliding, etc.) and the mathematical
TE

complexity involved. The model developed by Neumeister [10] is of practical interest because it

reproduces the tensile behavior of brittle-fiber composites without the need to solve differential equations
EP

or to evaluate auxiliary functions numerically.

While those models reasonably agree with the experimental stress-strain curve for a low
C

density of breaks (before the maximum of the curve), there is considerable discrepancy between them

when predicting strength. First, in a GLS context, and in spite of the lack of physical grounds for this
AC

hypothesis, it is often assumed that strength corresponds to the maximum of the curve and so GLS models

largely overpredict the experimental strength. In fact, composite failure will take place when the number

of fiber breaks reaches such a high level that an avalanche of fiber failure events take place. Indeed, the

load released by fiber-break(s) is redistributed among the intact adjacent fibers to reestablish local load

equilibrium. The stress is intensified in the adjacent fibers, increasing the probability of neighboring

fibers fracturing and, hence, controlling the strength. Consequently, the way this concentration factor is

2
 ACCEPTED MANUSCRIPT

modeled determines the accuracy of the predicted strength (i.e. the global load-sharing assumption is

clearly not obeyed [4]).

Other types of models account for the stress concentrations around broken fibers and are generically

named Local Load Sharing (LLS) models [13–16]. They make use of complex mathematical formulations

[17,18] or numerical Monte-Carlo Simulations [19–22], although unlike the analytical GLS models, the

PT
computational cost of such models is very high and they do not lead to analytical expressions. In addition,

LLS models tend to underpredict the observed strength [23].

RI
Given the advantage GLS models have of being able to deliver fast results (allowing their use, for

example, in optimization iterative processes), there is interest in predicting composite strength within

SC
their framework. To achieve this, the composite strength must be related to a parameter delivered by the

GLS model. Koyanagi et al. [24] followed this approach and presented the Simultaneous Fiber-Failure

U
(SFF) model, which assumes that the UD composite fails when a certain number (n) of simultaneous fiber
AN
breaks occur. This critical number of fiber breaks was deduced from the experimental strength of a set of

different composite materials and the fragmentation model. Then, for the composite materials considered,
M

a phenomenological relationship was established between n and the strength ratio , where is

the interfacial strength between fiber and matrix, is the characteristic stress of the Weibull distribution
D

of the fiber strength and is the strength of the matrix. The empirical relationship between and the
TE

strength ratio reads:

ln = + ln + (1)
EP

where and are constants.

C

Then, for an arbitrary system of known material parameters , and , equation 1 is used to estimate
AC

, which determines the ultimate tensile strength through the SFF model as follows:
%
&'%
+1 2 1 !" + − 1
= ∙ $
+2 +2 !"
(2)

where is the critical stress, is the Weibull coefficient and !" is the fiber volume fraction. The critical

stress is the theoretical stress that causes uniform fragment spacing in a single filament, in the moment

in which the fiber fragmentation process ceases. In practice, the composite would fail before reaching

3
 ACCEPTED MANUSCRIPT

this stress. The critical stress controls several major composite properties such as tensile strength, fiber

pullout and work of fracture [25]:

%
2 ∙ *+ ∙ &'%
= ) -
(
,∙
(3)

with * being the characteristic length of the Weibull distribution and , the fiber diameter.

PT
Koyanagi et al. mixed, in a unique relationship (equation 1), different material systems such as glass and

carbon fiber-reinforced polymeric composites (GFRP, CFRP), ceramic matrix composites (CMC), and

RI
carbon/carbon composites (C/C). Although their predictions were reasonably accurate for C/C and CMC,

in the case of GFRP and CFRP it was much lower. While Koyanagi’s GLS-based model may have

SC
improved strength estimation accuracy, when compared to the predictions based on the maximum of the

curve, they are not particularly suitable for CFRP and GFRP.

U
In this work, we propose an alternative approach to predict the strength of a unidirectional composite
AN
in a GLS fragmentation model framework. Our approach relies on assuming that failure will take place at

a critical crack density (i.e. number of breaks per unit length), which will be specific to each material
M

system. To determine critical density, two different fragmentation models (Neumeister’s and Turon’s)

formerly formulated to relate the remote stress, , with the far-field fiber stress, ", were rewritten in

terms of the number of breaks per unit length, Λ , to yield the Λ curve. The critical number of
D

breaks per unit length Λ

TE

for several CFRP and GFRP composites was evaluated using the current

model and data from the literature concerning their experimental strength. We observed clear trends in the

dependence of the critical number of breaks on the parameters of the system (fiber, matrix and interface
EP

properties) and which were distinct between CFRP and GFRP.

Polymer matrices reinforced with structural fibers as carbon, glass or aramid (fiber-reinforced
C

polymers or FRPs) possess excellent specific mechanical properties as strength and stiffness. As a result,
AC

structural composites are commonly used in applications driven by weight reduction in aerospace,

although they are continuously expanding to other industrial sectors, such as automotive, energy, sports or

civil engineering. Excellent examples of carbon composite applications in aerospace are found in the last

two civil airplanes developed by Airbus and Boeing, the A350 and 787 Dreamliner, respectively, in

which composites made up to 50 % in weight of structural parts ranging from fuselage barrels or wings to

stabilizers. However, despite the increasing number of engineering applications of structural composites,

4
 ACCEPTED MANUSCRIPT

the accurate prediction of their mechanical behavior still remains an arduous task because of the

complexity of the failure mechanisms involved, specially at the microscopic level [26]. This kind of

studies can be faced using numerical methods and computational tools. However, they have a high

computational cost and time consuming. The present work offers a practical analytical tool of design that

allows accurately predicting the mechanical response of unidirectional reinforced polymers used on the

PT
above-mentioned applications.

RI
2 Fiber Fragmentation models

SC
When a tensile load in the direction of the reinforcement is applied monotonically to a composite,

fibers fail randomly at several flaws along the composite in question [27]. Around a fiber break, the load

U
is transferred by the matrix and the fiber may slide (depending on the fiber/matrix interface). The stress

level on the fiber is fully recovered over a load transfer length *0 , due to the presence of the interfacial
AN
shear stress, [9,10]. The reinforcement is, thus, subjected to significant fragmentation prior to failure

[28] (Figure 1). Since any component made of such a composite contains a very large number of fibers,
M

when the density of fiber breaks is low, the overall behavior of the material is described by the averaged
D

value of the individual constituent’s behavior. Furthermore, by considering the stochastic response of the

reinforcement, the net stress over any cross-section of the composite containing enough contributing
TE

fibers can be obtained from the mean stress in a long (fragmented) fiber subjected to the same strain

[9,10].
EP

The probability of fracture for a fiber of length * under stress " is described by a Weibull distribution:

4 78 :
1 ", * =1− 2 5 75
3 6 9
4 (4)
C

where * is a characteristic length for which a representative strength (

AC

is known, and is the Weibull

coefficient. Because of the loss of ability to carry the load involved in the fragmentation process, the

apparent stiffness of the system decreases with the number of breaks. The apparent axial stiffness of the

composite can be obtained from the number of fiber breaks, ; [43], which is assumed to follow a

Poisson Law:

* " &
= ) -
;
*
(5)

5
 ACCEPTED MANUSCRIPT

Once the mean number of breaks is known, the distance between two breaks, i.e., the length of the

different fragments into which the fiber has split, can be computed as [29]

< = = Λ ∙ 2 3> (6)

where Λ is the number of breaks per unit length within a fiber:

1 " &
Λ= = ) -
;
* *
(7)

PT
Considering the importance of the critical stress (3), the number of breaks per unit length (7) can be

RI
rewritten as [29]

1 &'%
Λ = ) -
"
2*0 (
(8)

SC
where the load transfer length or ineffective length, *0 is the interval required to build up the full fiber

stress (or stress recovery), which is equal to

*0 =
U ,∙
4∙
"
AN
(9)

Small ineffective lengths (small fiber diameters or high interfacial shear stress) create large stress
M

concentrations in the neighboring fibers, thus raising the possibility of cumulative flaws leading to

fracture [30]. Large ineffective lengths increase the size of the flaw within the composite. Then, the
D

interfacial shear stress, , plays a crucial role in the ultimate tensile strength of unidirectional composites
TE

[31]. All of these parameters (equations 5 to 9) are the base for defining the fragmentation models.

In Neumeister’s model [10], the remote tensile stress reads

1 1 E
= !" ∙ D + 6 9 G
"
E+1 2∙F E+1 E+1
EP

@ABC
(10)

where E is a damage variable that accounts for the relationship between the load of the fiber, " , and the
C

critical stress (3):

&'%
AC

E=) -
"
(11)
(

In the same way, Turon et al. [27] proposed an approximate analytic relationship for stress-strain

curve:

1 − 2 3M 1 4 78
:
3 6 9
= !" ∙ L + E ∙ 2 NO 7P Q
HBIJK "
E 2
(12)

where R equals twice the stress recovery region at the critical stress (.

6
 ACCEPTED MANUSCRIPT

Furthermore, Hui et al. [11] obtained an equivalent expression for the approximation of the remote

tensile stress as

1 %
= !" ∙ D)1 + E + Θ ∙ E - 2 3M6%3VWM9 G
SBT "
2
(13)

where

X=
+1

PT
(14)

7 + 12
Θ=
24 2 + 3
(15)

RI
To compare the three above-mentioned fragmentation models, the predicted stress-strain curves for

SC
typical carbon and glass reinforced composites are depicted in Figure 2 by using (10), (12) and (13).

Table 1 contains the constituent properties of each material.

U
AN
Table 1. Weibull strength parameters for typical carbon-epoxy and glass-epoxy composites [4]

Material
Parameter
Carbon Fiber Glass Fiber
Characteristic strength,
Characteristic length, *
M

5000 MPa 2500 MPa

25.0 mm 25.0 mm
Weibull modulus, 7.0 7.0
Longitudinal fiber moduli, ["
Fiber diameter, ,
D

230 GPa 70 GPa

7.0 µm 12.0 µm
Interfacial shear stress, \]^_`/bcdef
TE

50 MPa 50 MPa

The three models provide practically the same stress-strain behavior for low density of breaks before
EP

peak stress is reached. If the composite is assumed to fail when the stress-strain curve reaches peak stress,

the predicted strength largely exceeds the experimental values. This is because these models rely on the
C

GLS assumption and so do not take into account local effects that will be predominant as the number of
AC

breaks increases. The following sections develop the idea that failure occurs when the composite reaches

a critical density of breaks.

3. Critical number of breaks (CNB) model

3.1 Fiber fragmentation models in terms of the number of breaks per unit length

Fiber fragmentation can be understood as a progressive degradation process caused by the increase of

the number of fiber breaks as a load is monotonically applied. Then, the remote tensile stress given in

7
 ACCEPTED MANUSCRIPT

equation (10) or (12) can be rewritten as a function of the number of breaks per unit length instead of the

far-field fiber stress ", given that from equation (7)

%
" = Λ * & ∙ (16)

Replacing (16) in (11) the damage variable, E, reads

E = 2*0 ∙ Λ (17)

PT
And substituting (16) and (17) in (10), the Neumeister-CNB model becomes
%
= !" ∙ Λ * & ∙

RI
ghi 3 gj

1 1 2*0 ∙ Λ
(18)
∙L + Q
2*0 ∙ Λ + 1 2 ∙ F k 2*0 ∙ Λ + 1l 2*0 ∙ Λ + 1

SC
Similarly, the Turon-CNB model (12) results in

% 1 − 2 3 4q∙>
= !" ∙ Λ * ∙ ∙p + Λ*0 2 34J ∙> r

U
&
min+o3 gj 2*0 ∙ Λ
(19)
AN
The corresponding expression for Hui’s model is not included here because it cannot be solved

explicitly and because, before maximum stress, the three models are similar. The models expressed in
M

equations (18) and (19) relate composite stress with crack density (Figure 3). The overall stress evolution

of each composite is practically equal for both models until the peak of the curves is reached (this result is
D

expected in view of the similarity of the models (Figure 2), and the fact that only a change of variables
TE

has been introduced).

EP

3.2 Prediction of the critical density of fiber breaks

The approach proposed in this work relies on assuming that the ultimate stress of the composite
C

corresponds to a critical density of fiber breaks characteristic of the material in question (volume fraction
AC

of fibers in the composite, fiber diameter, Weibull parameters of the fiber and interfacial shear strength).

A necessary additional assumption is that the Λ curves (equations 18 and 19) apply when the

critical density of fiber breaks is reached. This last assumption is somehow contradictory. On the one

hand, the fragmentation models (equations 18 and 19) follow a global load-sharing rule, so the interaction

between fiber breaks is not considered. On the other hand, the fracture of the composite occurs when a

critical crack density is reached, in which the interaction between fiber breaks is important enough to

create avalanches of broken fibers and eventually the collapse of the component. When a fiber breaks, the

8
 ACCEPTED MANUSCRIPT

load on that fiber in locally relaxed and the fiber springs back. This creates a stress wave travelling along

each fiber, causing a temporary increase in the stress concentration [32]. Some authors have stressed the

importance of dynamic stress concentrations in the failure of unidirectional composites [33–37]. In fact,

the assumption is that as soon as the interaction between fiber breaks becomes important, the critical

stress (or the critical density of breaks) is reached.

PT
By admitting these hypotheses, the critical stress measured in an experimental test must correspond to

the stress predicted by the fragmentation models when the crack density reaches the critical crack density.

RI
Or, in other words, the critical crack density can be deduced from the measured ultimate stress by using

the fragmentation model.

SC
We followed this approach to compute the critical number of breaks for seven CFRP and three GFRP

composites with different constituent properties and fiber volume fractions (Table 2). Composites 1 to 6

U
in Table 2 are composites with different types of carbon reinforcements, all of which are embedded in
AN
epoxy (EP) resins. Item 7 corresponds to a carbon fiber and vinyl ester (VE) composite system.

Furthermore, composites 8 to 10, are materials reinforced with glass fibers in EP or VE matrices.

(
M

On the other hand, we have extended the concept of the critical stress for a single fiber, (eq. 3, the
s
stress at which fiber fragmentation saturates), to the overall fragmentation limit stress, (, of the
D

composite to account for the fiber volume fraction

s
=
TE

!" (20)

s
(, which only depends on the constituent properties and the volume fraction, corresponds to the
EP

remote stress that, on average, causes a fragmentation saturation in all the fibers of the composite.

That is, the composite is not able to develop further fragmentation, so other damage mechanisms, such as
C

fiber sliding and fiber pullout, among others, start to appear. This sequence of events is already
AC

evidenced in, for example, the Neumeister model (eq. 10). The equation has two softening terms: the first

corresponds to the contribution the intact fibers make to the overall stress within the composite (dashed

line in Figure 4), whereas the second accounts for the contribution the sliding and separation of the fibers

makes (dashed-dotted line in Figure 4). When the first term curve is close to reaching the peak,

fragmentation ceases, the sliding and separation of fibers becomes predominant and the failure of the

composite is imminent.

9
 ACCEPTED MANUSCRIPT

Table 2 contains the constituent properties, the derived s

(, the measured ultimate stress, tu , and the

associated critical number of breaks, Λ .

Table 2. Constituent properties, experimental tensile strength, critical number of breaks and overall
fragmentation limit stress of several GFRP and CFRP composites systems.

No. 1 2 3 4 5 6 7 8 9 10

PT
CARBON
Fiber GLASS
T700SC AS400 T700 M40 AS-4 AS-4 M40
VINYL VINYL

RI
Matrix EPOXY EPOXY
ESTER ESTER

vw
0.300 0.300
0.590 0.700 0.600 0.677 0.677 0.300 0.567 0.100
0.400 0.400
xy
z1{

SC
230 294 220 392 234 234 392 76 72 72

|
μ~
6.9 7.1 7.0 6.0 7.1 7.1 6.0 13.0 14.9 16.0

•€
•1{
2700 4275 5470 4500 4275 4493 4500 1150 1649 2500

‚€
~~
100.0 12.5 20.0 25.0

U
12.7 10.0 25.0 24.0 5.0 25.0
AN
ƒ 9.03 10.30 5.60 16.0 10.7 8.33 16.0 6.34 3.09 13.0

„
•1{
23 40 40 50 40 40 20 42 23 40
M

Ref. [38–40] [41] [41] [24] [42] [42] [24] [22] [38] [24]

•s…
12306,0 9871,9 13756,3 9955,0 8524,4 9374,0 18550,8 3956,3 7497,6 33059
D

MPa
tu 1055 1890 3409 2310 1890 1890 1390 940 719 252
MPa
TE

†…
131.9E-4 41.2E-4 269.4E-4 43.5E-4 8.2E-4 19.0E-4 1520.9E-4 10283.5E-4 5168.0E-4 529.6E-4
1/mm

Consequently, the critical number of breaks per unit length Λ‡ and the overall fragmentation limit
EP

stress s
, point to the same stage in the loading of the composite: the transition from fragmentation to

the final failure of the composite. Their mutual dependence shown for the materials included in Table 2,
C

demonstrates distinct behaviors for CFRP and GFRP composites (Figure 5). For CFRP (dashed-dotted
AC

line) Λ‡ increases with s

while for GFRP composites (dashed line) this decreases. In the present study,

we fitted the relationship between them by using the following empirical equation:
‰
Λ‡ = 2 ∙k7Ô l 'Š∙k7Ô l'( (21)

where the parameters {, ‹ and Œ for each material (Table 3) were determined by a second order

polynomial fitting for ln Λ‡ vs s

.

10
 ACCEPTED MANUSCRIPT

Table 3. Estimation of the parameters of the critical number of breaks model (CNB) for carbon
and glass fiber reinforced composites

{ ‹ Œ
Coefficient
Fiber

Carbon Fiber -3.21E-8 1.35E-4 -16.06

Glass Fiber 3.613E-9 -2.36E-5 0.903

PT
As can be observed in Figure 5, the curves for CFRP and GFRP present different trends, with the first one

monotonically increasing and the other one, monotonically decreasing. A possible way to give an

RI
explanation to the different trends observed is to analyze independently the influence of the different

parameters ([" , + , *+ …) on the critical stress and the associated critical density of breaks. However,

SC
using the experimental data available, it is difficult to perform this analysis and provide a conclusive

explanation. This analysis can be done, for example, using detailed micro-mechanical models and will be

U
addressed by the authors in the near future.
AN
4 Discussion

Applying the CNB model to (Table 2) material 2, AS400 carbon fiber embedded in an epoxy matrix,
M

is used to discuss the proposed procedure and to compare the predicted composite strength using two
D

alternative approaches: the maximum of the stress-strain curve, MAX, and the Simultaneous Fiber

Failure, SFF, from Koyanagi [24]

TE

Both Neumeister’s (10) and Turon’s (11) fragmentation models provide the same stress-strain curve

before its maximum, so they can be used equivalently to predict the strength (Figure 6). The measured
EP

tensile strength of AS400/Epoxy is Su = 1890 MPa. The material has a critical stress of σ( =

5824.5 MPa which, considering the fiber volume content !" = 0.590, leads to an overall fragmentation
C

limit stress of s
= 9872.11 MPa. Then, the corresponding critical number of breaks according to the
AC

empirical relationship of the equatio1n (21) is Λ‡ = 4.12[ − 3 ~~3% . Using (18) or (19), this critical

level of breaks leads to a predicted ultimate tensile stress of tu gj

= 1836.1 MPa.

Then, the error in the CNB value, with regards the experimental one, is ˜ gj = 2.85%. The error is

obviously small as this material is one of the seven CFRP materials used to fit Equation (21).

On the other hand, using Koyanagi’s SFF model with a matrix strength of = 80 MPa (normal

epoxy resin strength [24]), the number of fibers that fail simultaneously and the predicted ultimate

11
 ACCEPTED MANUSCRIPT

tensile strength k š››

l are, respectively, = 3.79 and tu š›› = 2293.57 MPa. Then, the error of the

SFF value, with respect to the experimental one, is ˜š›› = 21.35%

Finally, the maximum of the curve, (MAX, Figure 6) using the Neumeister model (10) gives a tensile

strength of tu = 2709.44 MPa, (˜ = 43.35%).

The ultimate stress prediction, according to the CNB, SFF and MAX models, for all the materials in

PT
Table 2 is illustrated in Figure 7. These results show the suitability of the number of breaks parameter in

identifying where in the fragmentation curve composite failure will take place. Fragmentation models

RI
based on the GLS rule extend continuously up to practically unattainable strains. The ability to locate the

point in the GLS curve where the material fails complements and enhances the usefulness of the

SC
fragmentation models to support structural design and/or optimize the constituent properties for an

improved mechanical design.

U
A GLS fragmentation model has clear advantages in terms of readiness of implementation and
AN
delivery of outputs, although it does have intrinsic limitations. For instance, the fact that GLS models do

not account for stress concentrations around fiber breaks rules out their ability to capture size scaling
M

effects; something which LLS models do [16].

The predictive accuracy for composites other than those explored in Table 2, remains an open
D

question, especially if more ductile matrices are considered, where the stress redistribution near a fiber
TE

break can be very different. We would have liked to have included other CFRP or GFRP composites

(different fibers and matrices) in this study to challenge the validity of the phenomenological CNB model.
EP

However, the mechanical properties of the constituents required to calculate the overall fragmentation
s
limit stress, (, are difficult to determine experimentally with confidence [43] and are scarcely found in
C

the literature. Investigations based on computational micromechanics in a Finite Element Modeling

AC

framework are a more controlled scenario where the effect of the constituent properties on the ultimate

strength can be investigated. In spite of the high computational cost associated to analyzing the

fragmentation of a relevant number of fibers, this remains an ongoing research activity carried out in the

author’s laboratory.

5 Conclusions

Neumeister’s and Turon’s global load-sharing (GLS) fragmentation models have been reformulated to

express composite remote stress in terms of the number of breaks per unit length. We propose that

12
 ACCEPTED MANUSCRIPT

composite failure, which cannot be directly captured by GLS models, takes place at a critical number of

breaks per unit length which, in turn, depends on the properties of the constituents and the volume

fraction of the fibers (CNB model).

A set of CFRP and GFRP materials found in the literature, including the data for constituent

properties and ultimate stress, has been used to deduce an empirical law for the critical number of breaks

PT
as a function of the overall fragmentation limit stress, which depends solely on the material properties and

fiber volume fraction. While the empirical law shows differing behavior for CFRP and GFRP, there are

RI
clear tendencies for each of them. For the materials considered, the CNB model better predicts the

strength when compared to other approaches previously published in the literature. However, the model’s

SC
accuracy for other materials requires further investigation.

The CNB approach complements GLS fragmentation models by adding the capability to identify the

U
ultimate stress in the stress strain curve.
AN
Acknowledgements
M

This work was supported by Administrative Department of Science, Technology and Innovation,

Colciencias (Colombia) [grant number 1210-669-46014]; Spanish Ministerio de Ciencia e Innovación

D

with the grant number MAT2015-69491-C3-1-R; Polymers Laboratory of Instituto Tecnológico

TE

Metropolitano, Medellín (Colombia); Enlazamundos, Agencia Superior de Medellín, (Colombia) [grant

year 2015].
EP

References
[1] Lamon J. Stochastic models of fragmentation of brittle fibers or matrix in composites. Compos
Sci Technol 2010;70:743–51. doi:10.1016/j.compscitech.2010.01.005.
[2] Naresh K, Shankar K, Velmurugan R. Reliability analysis of tensile strengths using Weibull
C

distribution in glass/epoxy and carbon/epoxy composites. Compos Part B 2017;133:129–44.

doi:10.1016/j.compositesb.2017.09.002.
AC

[3] Zhuang L, Pupurs A, Varna J, Ayadi Z. Fiber/matrix debond growth from fiber break in
unidirectional composite with local hexagonal fiber clustering. Compos Part B 2016;101:124–31.
doi:10.1016/j.compositesb.2016.07.005.
[4] Swolfs Y, McMeeking RM, Rajan VP, Zok FW, Verpoest I, Gorbatikh L. Global load-sharing
model for unidirectional hybrid fibre-reinforced composites. J Mech Phys Solids
2015:JMPSD1400292. doi:10.1016/j.jmps.2015.08.009.
[5] Yu Y, Zhang B, Tang Z, Qi G. Stress transfer analysis of unidirectional composites with
randomly distributed fibers using finite element method. Compos Part B Eng 2015;69:278–85.
doi:10.1016/j.compositesb.2014.09.035.
[6] Ma Y, Yang Y, Sugahara T, Hamada H. A study on the failure behavior and mechanical
properties of unidirectional fiber reinforced thermosetting and thermoplastic composites. Compos
Part B 2016;99:162–72. doi:10.1016/j.compositesb.2016.06.005.
[7] Majumdar BS, Matikas TE, Miracle DB. Experiments and analysis of fiber fragmentation in

13
 ACCEPTED MANUSCRIPT

single and multiple-fiber SiC / Ti-6AI-4V metal matrix composites. Compos Part B
1998;8368:131–45.
[8] Curtin WA. Exact theory of fibre fragmentation in a single-filament composite. J Mater Sci
Technol 1991;26:5239–53.
[9] Neumeister JM. Bundle pullout—a failure mechanism limiting the tensile strength of continuous
fiber reinforced brittle matrix composites— and its implications for strength dependence on
volume and type of loading. J Mech Phys Solids 1993;41:1405–24.
[10] Neumeister JM. A constitutive law for continuous fiber reinforced brittle matrix composites with
fiber fragmentation and stress recovery. J Mech Phys Solids 1993;41:1383–404.
[11] Hui CY, Phoenix SL, Ibnabdeljalilt M, Smiths RL. An exact closed form solution for

PT
fragmentation of Weibull fibers in a single filament composite with applications to fiber-
reinforced ceramics. J Mech Phys Solids 1995;43:1551–85.
[12] Turon A, Costa J, Maimí P, Trias D, Mayugo JA. A progressive damage model for unidirectional
fibre-reinforced composites based on fibre fragmentation. Part I: Formulation. Compos Sci

RI
Technol 2005;65:2039–48. doi:10.1016/j.compscitech.2005.04.012.
[13] Ibnabdeljalil M, Curtin WA. Strength and reliability of fiber-reinforced composites: Localized
load-sharing and associated size effects. Int J Solids Struct 1997;34:2649–68. doi:10.1016/S0020-

SC
7683(96)00179-5.
[14] Beyerlein IJ, Phoenix SL. Stress concentrations around multiple fiber breaks in an elastic matrix
with local yielding or debonding using quadratic influence superposition. J Mech Phys Solids
1996;44:1997–2036. doi:10.1016/S0022-5096(96)00068-3.

U
[15] Sastry a. M, Phoenix SL. Load redistribution near non-aligned fibre breaks in a two-dimensional
unidirectional composite using break-influence superposition. J Mater Sci Lett 1993;12:1596–9.
doi:10.1007/BF00627024.
AN
[16] Zhou SJ, Curtin WA. Failure of fiber composites: a Lattice Green function model. Acta Metall
Mater 1995;43:3093–104.
[17] Harlow DG, Phoenix SL. The Chain-of-Bundles Probability Model For the Strength of Fibrous
Materials I: Analysis and Conjectures. J Compos Mater 1978;12:195–214.
M

doi:10.1177/002199837801200207.
[18] Harlow DG, Phoenix SL. The Chain-of-Bundles Probability Model For the Strength of Fibrous
Materials II: A Numerical Sudy of Converfence. J Compos Mater 1978;12:195–214.
doi:10.1177/002199837801200207.
D

[19] Curtin WA, Takeda N. Tensile Strength of Fiber-Reinforced Composites: I. Mode and Effects of
Local Fiber Geometry. J Compos Mater 1998;32:2042–59.
TE

[20] Curtin WA, Takeda N. Tensile Strength of Fiber-Reinforced Composites: II. Application to
Polymer Matrix Composites. J Compos Mater 1998;32:2060–81.
[21] Landis CM, Beyerlein IJ, McMeeking RM. Micromechanical simulation of the failure of fiber
reinforced composites. J Mech Phys Solids 2000;48:621–48.
EP

[22] Okabe T, Takeda N, Kamoshida Y, Shimizu M, Curtin WA. A 3D shear-lag model considering
micro-damage and statistical strength prediction of unidirectional fiber-reinforced composites.
Compos Sci Technol 2001;61:1773–87. doi:10.1016/S0266-3538(01)00079-3.
[23] Behzadi S, Curtis PT, Jones FR. Improving the prediction of tensile failure in unidirectional fibre
composites by introducing matrix shear yielding. Compos Sci Technol 2009;69:2421–7.
C

doi:10.1016/j.compscitech.2009.06.010.
[24] Koyanagi J, Hatta H, Kotani M, Kawada H. A Comprehensive Model for Determining Tensile
AC

Strengths of Various Unidirectional Composites. J Compos Mater 2009;43:1901–14.

doi:10.1177/0021998309341847.
[25] Curtin WA. Stochastic Damage Evolution and Failure in Fiber-Reinforced Composites. Adv Appl
Mech 1999;36:163–253.
[26] Herráez M, Naya F, González C, Monclús M, Molina J, Lopes CS, et al. Microscale
Characterization Techniques of Fibre-Reinforced Polymers. Struct. Integr. Carbon Fiber
Compos., 2016, p. 283–99.
[27] Turon A, Costa J, Trias D, Blanco N, Mayugo J a. A progressive damage model for unidirectional
fibre-reinforced composites based on fibre fragmentation. Part II: Stiffness reduction in
environment sensitive fibres under fatigue. Compos Sci Technol 2005;65:2269–75.
doi:10.1016/j.compscitech.2005.05.011.
[28] Jalalvand M, Wisnom MR, Czig T. Design and characterisation of high performance , pseudo-
ductile all-carbon / epoxy unidirectional hybrid composites. Compos Part B Eng 2017;111:348–

14
 ACCEPTED MANUSCRIPT

56. doi:10.1016/j.compositesb.2016.11.049.
[29] Gulino R, Phoenix SL. Weibull strength statistics for graphite fibres measured from the break
progression in a model graphite / glass / epoxy microcomposite. J Mater Sci 1991;26:3107–18.
[30] Galiotis C, Paipetis A. Interfacial damage modelling of composites. In: Soutis C, Beaumont
PWR, editors. Multi-Scale Model. Compos. Mater. Syst. Art Predict. Damage Model., New York,
USA: Woodhead Publishing; 2005, p. 33–64.
[31] Drzal LT, Madhukar M. Fibre-matrix adhesion and its relationship to composite mechanical
properties. J Mater Sci 1993;28:569–610.
[32] Swolfs Y, Gorbatikh L, Verpoest I. Fibre hybridisation in polymer composites: A review.
Compos Part A Appl Sci Manuf 2014;67:181–200. doi:10.1016/j.compositesa.2014.08.027.

PT
[33] Hedgepeth JM. Stress Concentrations in filamentary structures. Washington: 1961.
[34] Ji X, Liu X, Chou T. Dynamic Stress Concentration Factors in Unidirectional Composites. J
Compos Mater 1985;19:269–75.
[35] Xing J, Hsiao GC, Chou T-W. A Dynamic Explanation of The Hybrid Effect. J Compos Mater

RI
1981;15:443–61. doi:10.1177/002199838101500504.
[36] Xia Y, Ruiz C. Analysis of damage in stress wave loaded unidirectional composites. Comput
Struct 1991;38:251–8.

SC
[37] Manders PW, Chou T. Enchancement of Strength in Composites Reinforced with Previously
Stressed Fibers Mof. J Compos Mater 1983;17:26–44.
[38] Diao H, Bismarck A, Robinson P, Wisnom MR. Production of continuous intermingled CF/GF
hybrid composite via fibre tow spreading technology. ECCM16 - 16TH Eur. Conf. Compos.

U
Mater., Sevilla: 2014, p. 8.
[39] Zhou Y, Baseer M a., Mahfuz H, Jeelani S. Statistical analysis on the fatigue strength distribution
of T700 carbon fiber. Compos Sci Technol 2006;66:2100–6.
AN
doi:10.1016/j.compscitech.2005.12.020.
[40] Watanabe J, Tanaka F, Okuda H, Okabe T. Tensile strength distribution of carbon fibers at short
gauge lengths. Adv Compos Mater 2014;23:535–50. doi:10.1080/09243046.2014.915120.
[41] Matveev MY, Long a. C, Jones I a. Modelling of textile composites with fibre strength
M

variability. Compos Sci Technol 2014;105:44–50. doi:10.1016/j.compscitech.2014.09.012.

[42] Curtin WA. Tensile Strength of Fiber-Reinforced Composites: III. Beyond the Traditional
Weibull Model for Fiber Strengths. J Compos Mater 2000;34:1301–32.
doi:10.1177/002199830003401503.
D

[43] Swolfs Y, Verpoest I, Gorbatikh L. A review of input data and modelling assumptions in
longitudinal strength models for unidirectional fibre-reinforced composites. Compos Struct
TE

2016;150:153–72. doi:10.1016/j.compstruct.2016.05.002.
C EP
AC

15
 ACCEPTED MANUSCRIPT

PT
RI
Figure 1. Stress profile at a fragment of a broken fiber

U SC
AN
M
D
TE
EP
C
AC

16
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 2. Comparison of fragmentation models for typical carbon/epoxy and glass/epoxy composites
M
D
TE
C EP
AC

17
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 3. Comparison of fragmentation models based on the number of breaks per unit length for typical
M

carbon/epoxy and glass/epoxy composites

D
TE
C EP
AC

18
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 4. Decomposition of the stress-strain curve according to the predominant damage phenomena of a
CFRP composite using Neumeister’s model
M
D
TE
C EP
AC

19
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 5. Relationship between the critical number of breaks per unit length œ( and the overall
fragmentation limit stress ( ′
M
D
TE
C EP
AC

20
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M

Figure 6. Predicted and experimental mechanical response for AS400/Epoxy composite using
Neumeister’s, Turon’s and CNB models
D
TE
C EP
AC

21
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D

Figure 7. Comparison of the predicted strength, using CNB, SFF and MAX models, with the
TE

experimental results.
C EP
AC

22
 ACCEPTED
Tensile strength, MANUSCRIPT
U

1000

2000

3000

4000

5000
0
(1) T700-EP

PT
RI
(2) AS400-EP

U SC
(3) T700-EP

AN
M
(4) M40-EP

D
TE
(5) AS-4-EP
Material

EP

Maximum at the Curve (Ne

SFF Model

CNB Model

Experimental
(6) AS-4-EP
C
AC

(7) M40-VE

(8) Glass-EP