Received: 30 June 2019 Revised: 9 October 2019 Accepted: 15 October 2019

DOI: 10.1111/ffe.13152

ORIGINAL CONTRIBUTION

Damage behaviour of laminated composites during fatigue

loading

Ali Mahmoudi1 | Bijan Mohammadi1 | Hossein Hosseini‐Toudeshky2

1
School of Mechanical Engineering, Iran
Abstract
University of Science and Technology,
Tehran, Iran This study deals with the modelling of damage evolution in the carbon/epoxy
2
School of Aerospace Engineering, laminated composites under static and fatigue loading. A cumulative damage
Amirkabir University of Technology,
model is developed on the basis of damage evolution due to static and fatigue
Tehran, Iran
during cyclic loading. A continuum damage mechanics (CDM)‐based damage
Correspondence model coupling with the micromechanics has been utilized to predict the
Bijan Mohammadi, School of Mechanical
fatigue behaviour of laminate composites. A multicriterion approach has been
Engineering, Iran University of Science
and Technology, Tehran, Iran. introduced to predict the damage behaviour in the longitudinal, transverse,
Email: bijan_mohammadi@iust.ac.ir and shear direction at the ply scale. Extensive experimental results on
T300/EPL1012 carbon/epoxy laminates are prepared to characterize under
static and fatigue loading and to evaluate the proposed model in different con-
ditions. The obtained results show that at the beginning of the cyclic loading,
the damage grows suddenly and increases until final failure, which justifies
the proposed method is able to predict the evolution of the damage due to static
and fatigue loading separately during cyclic loading. The obtained results show
that considering damage due to static loading leads to more accurate results,
particularly in low‐cycle fatigue.

K E YWO R D S
damage, fatigue life, fatigue loading, fibre‐reinforced composite, static loading

1 | I N T RO D U C T I O N A survey in the literature reveals that there are many

researches developed to quantify the fatigue process in
Failure of engineering components undergoing fatigue order to predict the number of cycles to fracture of com-
loading is one of the most dominant modes of failures posite materials. Despite the different behaviours of
in industrial systems. Composite materials are widely composite materials under fatigue loading from metals,
used in aerospace industry because of their high fatigue the first models were developed following the models
strength; however, they are not exempt from fatigue that had been developed previously for metals. Hashin
deteriorations. and Rotem1 proposed one of the first models based

NOMENCLATURE: A12,B12,C12,Af,Bf,Cf,Af,Bf,Cf, the material parameters of fatigue‐damage model; b, the constant of static‐damage model; E, the
stiffness of damaged state; E0, the stiffness of undamaged state; G, shear modulus; N, number of cycles; XT, static strength of longitudinal direction;
Y1,Y2,Y12,Yf,Ym, the conjugated thermodynamic forces of damage parameters; Y 1C ; Y 2C ; Y 12C , critical values of static‐damage model; Y O ; Y O ; Y O ,
1 2 12

threshold values of static‐damage model; YT, static strength of transverse direction; εe, elastic strain; ν, Poisson ratio; ρ, density; σ, stress; ψ, strain
energy; ω1,ω2,ω12, damage parameters of material directions; ωf,ωm, damage parameters of fibre and matrix; ωs,ωf, damage parameters due to static
and fatigue loading

Fatigue Fract Eng Mater Struct. 2019;1–13. wileyonlinelibrary.com/journal/ffe © 2019 Wiley Publishing Ltd. 1
2 MAHMOUDI ET AL.

on S‐N curves to determine the fibre and matrix strain energy. Mohammadi et al23 proposed a CDM model
failure mode. based on effective average local stresses that consider
Some models were developed to predict the residual three damage parameters for fibre, matrix, and shear
mechanical properties such as stiffness and strength on direction. In most of the CDM‐based models, the effect
the basis of experimental procedures.2-8 These models of load magnitude on the damage evolution, particularly
correlate between the residual stiffness/strength of com- in the first cycle, is not considered, which results in differ-
posites and the number of cycles to failure. The requiring ent critical damage parameters at different stress levels
of extensive experimental data and lack of specific failure and higher life predictions in low‐cycle fatigues.
criteria are the main problems of developing residual The recently developed entropy‐based models consider
mechanical properties models. a maximum value of released entropy during cyclic load-
The classical progressive damage models determine ing until final failure called fracture fatigue entropy (FFE)
the specific criteria for failure of laminated composites. as the failure criteria.24-27 In these models, the dissipated
These models degrade the stiffness on the basis of the energy through different mechanisms is utilized to calcu-
degradation law and using the applied stress in each late the generated entropy and temperature during cyclic
cycle. Shokrieh and Lessard9,10 proposed a model to pre- loading.
dict the damage state during cyclic loading and also pre- The damage evolution in the fatigue process is an accu-
dict the residual stiffness and strength and fatigue life. mulation of different damage mechanisms including
They considered seven damage modes, and the model damage due to static loading in the first cycle. In this
can predict the stage of these modes up to final failure. paper, a CDM‐based damage model coupled with
Despite the experimental observation of gradual degrada- the micromechanical bridging method is proposed to
tion of mechanical properties during the fatigue process, predict the stiffness degradation and fatigue life of com-
in the classical progressive damage models, the mechani- posite laminates under static and fatigue loading in
cal properties drop suddenly when a situation meets the carbon/epoxy laminates. Three different damage vari-
failure criteria. Besides, the classical progressive damage ables are considered to predict the fatigue behaviour of
models are not able to consider the thickness scale effects longitudinal, transverse, and shear directions. A lot of
and the stacking sequence. experiments have been carried out for the characteriza-
Some models are developed on the basis of fracture tion and verification of the proposed model. A numerical
mechanics, which predict the fatigue behaviour using algorithm is developed and implemented to evaluate the
micromechanical approaches and considering the physics capability of the model to predict the fatigue life of com-
of damage. These models utilize a closed‐form solution of posite laminates.
the stress‐strain field of representative unit cells to derive
the energy release rate of damage modes on the basis of
damage evolution laws.11-16 One of the advantages of
these models is simple characterization methods. How-
2 | MODELLING OF DAMAGE
ever, the predictions of these models for single‐layer com-
BEHAVIOUR UNDER STATIC AND
posites are not reliable. And also, the fracture mechanics–
F A T I G U E LO A D I N G
based models have complex mathematical formulation
Describing the damage evolution requires formulating the
and high computational costs.
damage behaviour of composite laminates under static
The continuum damage mechanics (CDM) models,
and cyclic loading.
firstly applied for static loading,17 are introduced to pre-
dict the fatigue behaviour and the damage parameters
according to changes in thermodynamic forces. Talreja18
proposed a CDM‐based model that considers the damage 2.1 | Damage mechanics of anisotropic
variables as tensors. Payan and Hochard19 developed a materials
CDM model to predict the damage under static and cyclic
loading. Kawai and Honda20 proposed a mesoscale model In this ply‐scale model, the internal variables are used to
based on in situ strength. Movaghghar and Lvov21 pro- describe the loss of stiffness. Three scalar variables
posed a method considering a single scalar damage are introduced to characterize the loss of rigidity in
parameter and a damage evolution law based on the longitudinal tension, transverse tension, and shear load-
strain energy. On the basis of Movaghar's Model, Salimi ing conditions.
et al22 considered three separate damage parameters for In a thermodynamic framework, the strain energy den-
three material directions. They correlated between the sity of damaged layer using the plane‐stress condition can
evolution rate of each damage parameter and related be written in terms of mean stress as follows17:
MAHMOUDI ET AL. 3

shear directions. The damage evolution law for ωs1 , ωs2 ,

1 σ 11 2þ σ 11 2 ν0
ρψ ¼ þ 0− − 2 120 σ 11 σ 22 and ωs12 are defined as
2 E1 ð1 − ω1 Þ
0
E1 E1
 (1) pffiffiffiffiffiffi qffiffiffiffiffiffiffi
σ 22 þ
2
σ 22 −
2
σ 12 2
þ 0 þ 0 þ 0 ; Y 1 − Y 1O
E2 ð1 − ω2 Þ E2 G12 ð1 − ω12 Þ ωs1 ¼ qffiffiffiffiffiffi ; (4)
Y 1C
where subscripts 1, 2, and 12 denote the fibre, transverse, qffiffiffiffi qffiffiffiffiffiffiffi
and shear directions, respectively. E 01 , E 02 , G012 , and ν012 are Yb − Y2
O
the elastic coefficient of the undamaged state. ω1, ω2, and ω2 ¼
s
qffiffiffiffiffiffi ;
2
ω12 are scalar variables describing the damage state under YC
longitudinal, transverse, and shear loading conditions, qffiffiffiffi qffiffiffiffiffiffiffi
respectively. The bracket notation in Equation (1) is b − Y 12
Y O
ω12 ¼
s
qffiffiffiffiffiffiffi ;
defined as 12
YC

σ þ ¼ σ if σ > 0; otherwise σ þ ¼ 0;
where Y 1O , Y 2O , and Y 12
O are threshold values that indicate
σ − ¼ σ if σ < 0; otherwise σ − ¼ 0:
the damage initiation. Y 1C , Y 2C , and Y 12
C are critical values
ν12 ν012 that are utilized to determine the slope of the damage
During damage evolution, ¼ 0 was assumed to
E1 E1 evolution. Y 1 is the maximum values of Y1 in previous
remain constant. From this strain energy, the constitutive times, and Y b is the maximum value of Y b , which is a linear
equations for elastic deformations are combination of transverse and shear thermodynamic
forces, for all previous times, defined as
σ 11þ σ 11− ν012
εe11 ¼ þ 0 − 2 0 σ 22 ;
E01 ð1 − ω1 Þ E1 E1 b ¼ Y 12 þ bY 2 :
Y (5)
σ 22þ σ 22− ν0
εe22 ¼ 0 þ 0 − 2 120 σ 11 ; (2) Typical damage evolution of longitudinal, transverse,
E2 ð1 − ω2 Þ E2 E1
σ 12 and shear directions are shown in Figure 1.
ε12 ¼ 0
e
: Y 1O , Y 2O , Y 12 1 2 12
2G12 ð1 − ω12 Þ O , Y C , Y C , Y C , and b are material constants,
which should be determined from an experimental proce-
And the conjugated thermodynamic forces associated dure described in Section 3.1.
with the damage variables are defined as
 2.3 | The evolution of damage due to


∂ψ  σ 11 2 fatigue loading
Y1 ¼ ρ  ¼ 0 þ 2;
∂ω1  2E 1 ð1−ω1 Þ
 σ; ω ; ω In this paper, to predict the stiffness reduction in longitu-
2 12
 dinal and transverse directions due to fatigue loading, two


∂ψ  σ 22 2 damage parameters in fibre and matrix due to fatigue
Y2 ¼ ρ  ¼ 0 þ 2; (3) loading are defined as
∂ω2  2E 2 ð1−ω2 Þ
 σ; ω ; ω
1 12
 E 0k − E k
 ωfk ¼ ; (6)
 E0k
∂ψ  σ 12 2
Y 12 ¼ ρ  ¼ 0 :
∂ω12  2G12 ð1−ω12 Þ2 k = m, f , represents the matrix and fibre, E 0k is the stiff-
 σ; ω ; ω
1 2 ness of undamaged state, and Ek represents the stiffness
of damaged state.
The damage in shear direction due to fatigue loading is
2.2 | The evolution of damage due to static defined independently as
loading
f G012 − G12
ω12 ¼ ; (7)
In case of unidirectional (UD) plies under static loading, G012
on the basis of the Ladeveze and Le Dantec17 model for
highly orthotropic materials such as carbon/epoxy com- where G012 is the shear modulus of undamaged material
posites, three scalar damage variable are introduced to and G12 is the shear modulus of damaged material. It is
determine the loss of rigidity in fibre, transverse, and noted that the shear damage is influenced by the
4 MAHMOUDI ET AL.

(A) Longitudinal damage evolution

FIGURE 1 Typical damage evolution:

A, longitudinal damage evolution; B,
transverse damage evolution; and C, shear
damage evolution [Colour figure can be
(B) Transverse damage evolution (C) Shear damage evolution viewed at wileyonlinelibrary.com]

interphase properties, which have anisotropic properties.

Therefore, it is better to consider a different damage 1 Em
parameter for shear direction. a22 ¼ a33 ¼ a44 ¼ þ
2 2Ef
It is needed to calculate average stress distribution in
fibres and matrix to evaluate the damage evolution.
According to bridging micromechanical theory,28 the 1 Gm
a55 ¼ a66 ¼ þ
averaged stresses in the fibres and matrix of UD compos- 2 2Gf
ites can be correlated using a bridging matrix:

 m   h f i

σ i ¼ Aij σ j ; (8) νm νf 1 1
a12 ¼ a13 ¼ ða11 − a22 Þ − = − (10)
Em Ef Ef Em
where the bridging matrix [Aij] represents the load share
capacity of the matrix with respect to fibres and defined and Ef is the elastic modulus of fibre and Em is the elastic
through28 modulus of the matrix. Vf and Vm are the volume frac-
tions of fibre and matrix, respectively. Now, the load
2 3
a11 a12 a13 0 0 0 share of fibre and matrix is defined as28
6 7
6 0 a22 0 0 0 0 7  f    
6 7
6 7 σ ¼ Bij σ j ;
  6 6 0 0 a33 0 0 0 7 7  m  i    
Aij ¼ 6 7; (9) σ i ¼ Aij Bij σ j ; (11)
6 0 0 0 a44 0 0 7
6 7
6 7
6 0 0 0 0 a55 0 7
4 5
0 0 0 0 0 a66 where [σi] = [σ11, σ22, σ33, σ23, σ12, σ12]T is the stress of com-
posite, [Bij] = (Vf[I]+Vm[Aij])−1, and [I] is a unit matrix.
where Em The rigidity of undamaged lamina in longitudinal and
a11 ¼ transverse directions can be defined in terms of elastic
Ef
constants of fibre and matrix as
MAHMOUDI ET AL. 5

ω1 ¼ ωs1 þ ωf1 ;
E 01 ¼ V f E 0f þ V m E0m ; ω2 ¼ ωs2 þ ωf2 ; (17)
ðV f þ V m a11 ÞðV f þ V m a22 Þ ω12 ¼ ωs12 þ ωf12 :
E 02 ¼
Vf Vm Vf Vm
ðV f þ V m a11 Þ 0 þ a22 0 þ V f V m 0 − 0 a12
Ef Em Ef Em
(12) 3 | MATERIAL
Now, the elastic constant of damaged material can be CH ARA C TERIZATION
defined as
In the following, the characterization of damage due
E1 ¼ V f 1 − ωff E0f þ Vm 1 − ωfm E0m ; to static and fatigue loading of T300/EPL1012
carbon/epoxy composites is described. The mechanical
properties obtained from the static test are shown in

ðV f þ V m a11 ÞðV f þ V m a22 Þ

E2 ¼ ! ! ;
Vf Vm νf νm
ðV f þ V m a11 Þ þ a22 þ VfVm − a12
1 − ωff E 0f 1 − ωfm E0m 1 − ωff E 0f 1 − ωfm E 0m

Table 1. The mechanical properties that are obtained

G12 ¼ G012 1− ωf12 : (13) from the static stress‐strain data consists of at least five
tests on separate specimens.
Therefore, the damage parameters of fatigue loading in
longitudinal and transverse directions are defined using
3.1 | Characterization of static‐damage
the above equations as
model
E01 − E 1
ωf1 ¼ ; Three cyclic loading on 0° UD plies, [±45]s, and [±67.5]s
E 01
angle plies are needed to obtain Y 1O , Y 2O , Y 12 1 2 12
O , Y C, Y C, Y C ,
and b.
E 02 − E2
ωf2 ¼ : (14)
E 02
3.1.1 | [04] tensile test
Mohammadi et al23 proposed the power‐law relation-
The stresses of principal material coordinates for 0° UD
ships to predict to stiffness degradation with respect to
plies under an axial load can be expressed in terms of
the number of cycles as
applied average stress σ x as
dωfk Ak Y Bk k
¼ C
; (15)
dN 1−ωfk k
TABLE 1 Material properties of T300/EPL1012 carbon/epoxy
where k represents the fibre, matrix, and shear direction. Parameters T300/EPL1012
The thermodynamic forces associated with the damage
E11 79.2 GPa
parameters are defined as
E22 4.5 GPa
σ 2k G12 4.47 GPa
Yk ¼ 2: (16)
2Ek 1−ωfk υ12 0.28

2.4 | Framework of accumulated damage XT 1086 MPa

YT 14.6 MPa
Development of the damage variables depends on the S12 56 MPa
stress level and loading conditions. In this study, the total
Ply thickness 0.3 mm
damage ω is the summation of damage due to static load-
Fibre volume fraction 0.51
ing ωs and damage due to fatigue loading ωf:
6 MAHMOUDI ET AL.

σ 11 ¼ σ x ;

σ 22 ¼ 0;

σ 12 ¼ 0: (18)

Figure 2 shows the stress‐strain response of longitudinal

cyclic loading on [04] UD plies. The tensile modulus
degrades with each higher maximum stress in a new cycle.
The values of longitudinal damage evolution curve are
determined from the unloading/reloading modulus of
each cycle from the results of Figure 2. The values of ωs1
are then used to calculate the corresponding thermody- FIGURE 3 Longitudinal damage evolution [Colour figure can be
namic associated force Y1 and obtain Y b . Figure 3 shows viewed at wileyonlinelibrary.com]
p ffiffiffiffiffi
ffi
the diagram of ωs1 ‐ Y 1 , and the fitted curve is used to
determine the form of evolution law.
The parameters of evolution law Y 1O and Y 1C can be σ 22 ¼ 0;
obtained using the slope and the intercept of the equation
of the fitted curve as σx
σ 12 ¼ − : (20)
qffiffiffiffiffiffiffi 2
0:0152
Y 1O ¼ ;
0:0344
Regarding the small values of σ x relating to the ulti-
mate strength of the fibre direction of carbon/epoxy com-
1
qffiffiffiffiffiffi ¼ 0:0344: (19) posite, the shear stress σ12 dominate the damage
Y 1C evolution in [±45]s carbon/epoxy laminate.
Representative results for the stress/strain response of
By solving these equations, parameters of the longitu- carbon/epoxy obtained from [±45]s laminates are shown
dinal damage evolution are calculated and presented in in Figure 4. The shear stress is obtained using Equa-
Table 2. tion (20), and the shear strain is obtained using strain
transformation equations. The modulus of each
unloading/reloading is determined through the slope of
3.1.2 | [±45]s angle ply tensile test
the straight lines in Figure 4.
The stresses of principal material coordinates for a highly
orthotropic [±45]s angle ply composite can be expressed TABLE 2 Parameters of modelling damage due to static loading
for T300/EPL1012 carbon/epoxy
in terms of applied average stress σ x as
Damage Model Parameters T300/EPL1012
σ 11 ¼ σ x ;
Static‐damage model Y 1O 0.19448 MPa
Y 1C 845.05 MPa
Y 2O 0.00146 MPa
Y 2C 3.38408 MPa
Y 12
O
0.00547 MPa
Y 12
C
1.1627 MPa
b 0.35
Fatigue‐damage model Am 3.756302807 * 1020
Bm 13.52
Cm −31.02
Af 2.301944291 * 10−14
Bf 14.06
Cf −33.07
A12 4.251776777 * 108
B12 10.3335
FIGURE 2 Cyclic longitudinal stress‐strain response of [04]
C12 −25.0067
laminate [Colour figure can be viewed at wileyonlinelibrary.com]
MAHMOUDI ET AL. 7

fitted curve in Figure 5 as

qffiffiffiffiffiffiffi
0:0686
O ¼
Y 12 ;
0:9278

1
qffiffiffiffiffiffiffi ¼ 0:9278: (21)
Y 12
C

Y 12 12
O and Y C are calculated using these equations and
presented in Table 2.

FIGURE 4 Cyclic shear stress‐strain response of [±45]s laminate

[Colour figure can be viewed at wileyonlinelibrary.com] 3.1.3 | [±67.5]s angle ply tensile test

The values of ωs12 , which are obtained from the results The approximated stresses of principal material coordi-
of Figure 4, are used to calculate the corresponding ther- nates for a highly orthotropic [±67.5]s angle ply compos-
modynamic forces Y12 and Y b from Equation (5) with ite can be expressed in terms of applied average stress
qffiffiffiffi
σ x as29
Y2 = 0. The values of ωs12 and Y b and the best fitted curve
to results are plotted in Figure 5 to determine the form of σ 11 ¼ 0:209σ x ;
the shear evolution law.
The parameters of shear damage evolution law Y 12 O and
σ 22 ¼ 0:791σ x ;
Y 12
C are calculated using the slope and the intercept of the

σ 12 ¼ −0:415σ x : (22)

The results of shear and transverse response of [±67.5]s

laminate under a biaxial state of stress are shown in
Figure 6.
By using the method described in previous sections,
qffiffiffiffi
the values of ωs2 as a function of b are plotted in
Y
Figure 7.
The parameters of transverse damage evolution law Y 2O
and Y 2C are calculated using the slope and the intercept of
the fitted curve in Figure 7 as
qffiffiffiffiffiffiffi
0:0208
FIGURE 5 Shear damage evolution [Colour figure can be viewed Y 2O ¼ ;
0:5436
at wileyonlinelibrary.com]

(A) Shear stress-strain response (B) Transverse stress-strain response

FIGURE 6 Cyclic stress‐strain response of [±67.5]s laminate. A, Shear stress‐strain response; B, transverse stress‐strain response [Colour
figure can be viewed at wileyonlinelibrary.com]
8 MAHMOUDI ET AL.

2E 0k Bk
max;k N f;k ¼
σ 2B k
(25)
Ak ð1 − R2 Þð2Bk þ C k þ 1Þ

2Bk þCk þ1
1 − 1−ωfc;k ;

where ωfc;k is the critical value of damage parameter,

which is determined using the maximum value of the
damage parameter at different loading conditions.
Taking the logarithm of Equation (25) yields
FIGURE 7 Transverse damage evolution [Colour figure can be
viewed at wileyonlinelibrary.com]
logN f;k ¼ alogσ max;k þ b;

1 a ¼ −2Bk ; b
qffiffiffiffiffiffi ¼ 0:5436: (23)

Y 2C 2E 0k Bk
¼ log
Ak ð1 − R2 Þð2Bk þ C k þ 1Þ (26)
2Bk þCk þ1 
Y 2O and Y 2C are calculated using these equations and 1 − 1−ωc;k f
:
presented in Table 2.
As the ply angle does not change the material parame- Now, Bk can be determined through the slope of the
ters of the evolution law, the parameter b can be deter- diagram logNf,k versus loglogσmax,k.
mined using the results of ωs as a function of Y b under Parameters Ak and Ck can be determined using the
12
b
biaxial stress state and results of ωs12 as a function of Y dωf −1
slope of the diagram log k versus log 1−ωfk ; there-
under pure shear as 29 dN
fore, taking a logarithm of both sides of Equation (24) is
needed:
qffiffiffiffiffiffiffi qffiffiffiffiffiffiffi2
ωs12 " #
C þ
Y 12 Y 12 − Y 12
b¼
O
: dωfk Ak ð1 − R2 Þσ 2B k
max;k
Y2 log ¼ log
dN 2E 0k Bk
−1
þ ð2Bk þ C k Þlog 1−ωfk : (27)
The parameters of modelling of the damage due to
static loading are summarized in Table 2. It is noted that to characterize the damage due to
fatigue loading, the fatigue damage values are obtained
using the Equation (17) through subtracting the static
3.2 | Characterization of fatigue‐damage damage from total damage. The total damage is obtained
model with measuring the portion of stiffness reduction to stiff-
ness of the pristine specimen using the experimental
Tensile, transverse, and shear loadings are needed to results of stiffness degradation.
characterize the fatigue model. Under different loading
conditions, with a nonnegative load ratio, R, and consid- 3.2.1 | Characterization of matrix damage
ering Equations (15) and (16), the corresponding evolu-
tion law can be considered as For characterization of the fatigue‐damage model in the
matrix material, tension‐tension fatigue tests on 90° plies
are required. Figure 8 shows the Log NF,m versus Log
dωfk Ak ð1 − R2 Þ σ 2B k

¼
max;k
2B þC
; (24) σmax,m at a load ratio of 0.1, the frequency of 6 Hz, and
dN 2E0k Bk 1−ωfk k k two stress levels (70% and 80% of static strength). Now,
Bm can be determined through the slope of the diagram
as shown in Figure 8A.
where σmax,k is the magnitude of maximum applied
As depicted in Figure 8B, the parameters Am and Cm
stress. Integrating Equation (24) from ωfk ¼ 0 to
dωf −1
ωfk ¼ ωfc;k yields are determined through log m − log 1−ωfm diagram
dN
MAHMOUDI ET AL. 9

FIGURE 8 Material characterization of

matrix using [9010] laminates under
uniaxial fatigue loading in the transverse
direction. A, Log NF,m versus Log σmax,m;
B, log dωfm /dN versus log (1‐ωfm )−1 [Colour
figure can be viewed at wileyonlinelibrary.
com]
(A) (B)

of tension‐tension fatigue tests on 90° plies at a load ratio laminates are required. Parameter B12 is determined
of 0.1 at the stress level of 70% of static strength. through the slope of the Log NF,12 versus Log σmax,12 at
The parameters of damage evolution of matrix due to a load ratio of 0.1 and different stress levels as shown in
fatigue loading are calculated and presented in Table 2. Figure 10A.
Parameters A12 and C12 are determined through
dωf −1
3.2.2 | Characterization of fibre damage log 12 − log 1−ωf12 diagram of tension‐tension
dN
fatigue tests on [±45]s laminates at a load ratio of 0.1 at
For characterization of the fatigue‐damage model in fibre the stress level of 80% of static strength as depicted in
material, tension‐tension fatigue tests on 0° plies are Figure 10B.
required. Parameter Bf is determined through the slope The parameters of damage evolution of shear direction
of the Log NF,f versus Log σmax,f at a load ratio of 0.1 due to fatigue loading are calculated and presented in
and different stress levels as shown in Figure 9A. Table 2.
Parameters Af and Cf are determined through
dωf −1
log f − log 1−ωff diagram of tension‐tension fatigue
dN
tests on 0° plies at a load ratio of 0.1 at the stress level
of 80% of static strength as depicted in Figure 9B. 4 | EXPERIMENTAL EVALUATION
The parameters of damage evolution of fibre due to O F T H E PR O P O S E D M O D E L
fatigue loading are calculated and presented in Table 2.
Carbon/epoxy laminates are fabricated using the hand
lay‐up technique. All specimens are cured at the room
3.2.3 | Characterization of shear damage temperature and under vacuum bag pressure. Specimens
were cut from carbon/epoxy sheets with water‐cooled dia-
For characterization of the fatigue‐damage model in shear mond saws. An epoxy‐based adhesive was used to glue
direction, tension‐tension fatigue tests on [±45]s the glass/epoxy tabs to the specimens.

(A) (B)

FIGURE 9 Material characterization of fibre direction using [04] laminates under uniaxial fatigue loading. A, Log NF,f versus Log σmax,f; B,
log dωff /dN versus log (1‐ωff )−1 [Colour figure can be viewed at wileyonlinelibrary.com]
10 MAHMOUDI ET AL.

FIGURE 10 Material characterization of shear direction using [±45]s laminates. A, Log NF,12 versus Log σmax,12; B, log dωf12 /dN versus log
(1 − ωf12 )−1 [Colour figure can be viewed at wileyonlinelibrary.com]

A series of tension‐tension fatigue test were performed of static strength. On the basis of the simulations, because
under load control mode. Referring to Figure 11, of higher stress values rather than the damage threshold,
T300/EPL1012 carbon/epoxy specimens were loaded the damage grows suddenly to the value of 0.06 because
using a 5‐ton DARTEC LTD servo‐hydraulic fatigue test- of static loading at the end of the first cycle. The damage
ing machine at a load ratio of 0.1 and frequency of 6 due to fatigue loading increases gradually until the final
Hz. The fatigue test were carried out up to final failure failure. Because of stiffness degradation during fatigue
of specimens. The test data were extracted on the specific loading, the thermodynamic force associated with the
cycles during fatigue test. static loading increases, which leads to damage evolution
For prediction of fatigue life of laminated composite, due to static loading. The damage parameter due to static
a numerical solution was developed on the basis of the loading increases from 0.06 at the end of the first cycle to
theories described in previous sections and implanted a value of 0.073 at the failure moment. The damage due to
as a USERMAT subroutine in Ansys commercial fatigue loading reaches the value of 0.15 at the failure
software. moment. The total damage, which is obtained from the
summation of damage parameters due to static and
fatigue loading, is in good agreement with the experimen-
4.1 | Stiffness degradation tal results. The critical damage of longitudinal direction at
which final failure occurs is 0.22.
The experimental results and the simulated ones for lon- As depicted in Figure 12B, at the stress level of 80% of
gitudinal, transverse, and shear directions are compared static strength in the transverse direction, the damage
in Figure 12A‐C using the material parameters presented parameter has a sudden increase in the first cycle of about
in Tables 2. 0.05 due to static loading. The damage parameter due to
Figure 12A shows the evolution of the damage param- fatigue loading increase gradually until final failure and
eter in the longitudinal direction at the stress level of 80% reaches the value about 0.45 before final failure. After a

FIGURE 11 Carbon epoxy specimens and experimental set up for fatigue test [Colour figure can be viewed at wileyonlinelibrary.com]
MAHMOUDI ET AL. 11

Figure 12C shows the stiffness degradation of shear

direction at the stress level of 80% of static shear strength.
The results obtained from simulations indicate that the
damage parameter has a sudden increase about 0.15 at
the end of the first cycle because of static loading. In most
part of fatigue loading, the shear damage parameter due
to static loading is dominant to the damage parameter
due to fatigue loading. The damage due to fatigue loading
increases gradually until final failure and reaches the
value of 0.23. The damage due to static loading grows
steadily after the first cycle and reaches the value of
(A)
0.22. The critical damage of shear direction, which is the
summation of static and fatigue damage at the final fail-
ure, is 0.45.

4.2 | Cross‐ply laminates

In this section, the fatigue failure of T300/EPL1012

carbon/epoxy [02/902]s cross‐ply laminated under
tension‐tension fatigue test with a load R‐ratio of 0.1 at
different stress levels is investigated. The material param-
eters presented in Table 2 are used to simulate the fatigue
behaviour of laminates. The predicted results by the pro-
(B) posed model for the fatigue life of the [02/902]s cross‐ply
laminates are presented in Figure 13 in comparison with
the experimental results. It is noted that the ultimate
static strength for [02/902]s specimens is 691 MPa. As
can be seen in Figure 13, when the static damage is con-
sidered, the results are more accurate than the ones not
considering it. The differences between the fatigue life
predictions of these cases are because in the case of not
considering the static damage, the damage parameters
meet the failure criteria (critical damage parameters) at
the higher fatigue life. At higher stress levels, the effect
of static damage is more important, and not considering
the static damage leads to the predicted fatigue life higher
(C) than experimental results. The model predicts firstly the
FIGURE 12 The evolution of damage parameter at the stress
level of 80% of static strength of (A) longitudinal direction, (B)
transverse direction, and (C) shear direction [Colour figure can be
viewed at wileyonlinelibrary.com]

sudden increase in the first cycle, the damage due to static

loading increases with a slow slope until final failure and
reaches to the value of 0.1 before final failure. The exper-
imental results show that the total damage before final
failure has a value of 0.55, which approves the results
obtained from the proposed model. It is noted that the
damage due to fatigue loading is dominant in most part
of the fatigue process relative to the damage due to static FIGURE 13 Fatigue life versus stress for [02/902]s laminates at
loading. The critical damage of transverse direction at different stress levels [Colour figure can be viewed at
which final failure occurs is 0.56. wileyonlinelibrary.com]
12 MAHMOUDI ET AL.

TABLE 3 Estimated fatigue life with and without considering the damage due to static loading for T300/EPL1012 carbon/epoxy [02/902]s
laminates at different stress levels

Stress Level, % Estimated Fatigue Life Considering Estimated Fatigue Life Without Considering
of Static Strength the Damage Due To Static Loading the Damage Due To Static Loading Difference, %

65 822 500 898 000 9.2

75 17 500 20 500 16.8
85 915 1135 24

ply failure in 90° plies and finally the fibre breakage in 0° and equals to 0.22 before final failure at a stress level of
plies, which are in agreement with the literature experi- 80% of static strength. In the longitudinal direction, the
mental observations.30 damage due to static loading reaches the value of 0.073
Most of the fatigue models consider no contribution for before final failure at the stress level of 80% of static
damage due to static loading. However, the proposed strength, which equals 33% of total damage. In transverse
model considers both damage parameters due to static direction, the damage due to static loading at the stress
and fatigue loading in the fatigue process, which leads level of 80% of static transverse strength is about 0.1
to more accurate estimations about fatigue life. Table 3 before final failure, which is 18% of total damage. The
shows the estimated fatigue life at different stress levels damage parameters due to fatigue loading are growing
for T300/EPL1012 carbon/epoxy [02/902]s cross‐ply lami- continuously until final failure.
nate. The damage due to static loading has a considerable The model is able to predict the final failure of the lam-
effect in low‐cycle fatigue, which can cause even static inated composites using the critical damage parameters.
failure instead of fatigue failure during cyclic loading. In Obtained results show that at a stress level of 85% of
addition, the damage due to static loading could cause static strength, the estimated life for T300/EPL1012
fracture in specimens with stress concentration because carbon/epoxy [02/902]s cross‐ply laminate is 24% higher
of higher stress levels. without considering damage parameter due to static load-
ing relative to the estimation considering this parameter.

5 | CONCLUSION
REFERENCES
In this study, a coupling static‐fatigue damage model has 1. Hashin Z, Rotem A. A fatigue criterion for fibre reinforced com-
been developed to predict the static and fatigue behaviour posite materials. J Compos Mater. 1973;5:448‐464.
of carbon/epoxy laminated composites undergoing cyclic
2. Hwang W, Han KS. Fatigue of composites—fatigue modulus
loading. Three damage parameters in longitudinal, trans-
concept and life prediction. J Compos Mater. 1986;20(2):154‐165.
verse, and shear directions are considered for modelling
the damage behaviour of composite laminates. The evolu- 3. Whitworth H. Modelling stiffness reduction of graphite epoxy
tion of damage parameters due to static loading is based composite laminates. J Compos Mater 1987;21: 362‐372, 4.
on static evolution laws using the conjugated thermody- 4. Sidoroff F, Subagio B. Fatigue damage modelling of composite
namic forces and the damage thresholds. A CDM‐based materials from bending tests. In: Sixth International Conference
damage model coupling with micromechanics is utilized on Composite Materials (ICCM‐VI) & second European Confer-
for modelling the stiffness degradation at the level of ence on Composite Materials (ECCM‐II). 1987.
fibre, matrix, and in‐plane shear. A numerical method 5. Viellevigne S, Jeulin D, Renard J, Sicot N. Modelling of the
has been utilized to predict the evolution of damage fatigue behaviour of a unidirectional glass epoxy composite sub-
parameters due to static and fatigue loading during cyclic mitted to fatigue loadings. In: International Conference of
loading. Fatigue of Composites. Paris, France. 1997.
The sudden growth of damage at the first cycle of the
6. Brondsted P, Andersen SI, Liholt H. Fatigue damage accumula-
fatigue process resulted from applying the maximum load tion and life time prediction of GFRP materials under block
of the first cycle is predicted by the model and verified by loading and stochastic loading. In: Proceedings of the 18th Riso
the test results. During cyclic loading, the damage param- International Symposium on Materials Science. Denmark:
eter grows due to static loading until final failure, since Roskilde:1997.
the conjugated thermodynamic force increases due to 7. Kawai M. Damage mechanics model for off‐axis fatigue behavior
stiffness degradation. The damage due to static loading of unidirectional carbon fiber‐reinforced composites at room
has a considerable contribution to total damage in the temperatures. In: Proceedings of the Twelfth International Confer-
shear direction, which has a share of 49% of total damage ence on Composite Materials (ICCM‐12). Paris: France:1999.
MAHMOUDI ET AL. 13

8. Van Paepegem W, Degrieck J. Experimental setup for and 21. Movaghghar A, Lvov GI. An energy model for fatigue life predic-
numerical modelling of bending fatigue experiments on plain tion of composite materials using continuum damage
woven glass/epoxy composites. Compos Struct. 2001;5:1‐8. mechanics. Appl Mech Mater. 2012;110:1353‐1360.
9. Shokrieh MM, Lessard LB. Progressive fatigue damage model- 22. Salimi Majd D, Helmi M, Mohammadi B. Damage growth pre-
ling of composite materials, Part I: Modeling. J Compos Mater. diction of unidirectional layered composites under cyclic
2000;34(13):1056‐1080. loading using an energy based model. Modares Mech Eng.
10. Shokrieh MM, Lessard LB. Progressive fatigue damage modeling 2015;15:173‐180.
of composite materials, Part II: Material characterization and 23. Mohammadi B, Fazlali B, Salimi‐Majd D. Development of a con-
model verification. J Compos Mater. 2000;34(13):1081‐1116. tinuum damage model for fatigue life prediction of laminated
11. Bergmann HW, Prinz R. Fatigue life estimation of composites. Compos A: Appl Sci Manuf. 2016;93:163‐176.
graphite/epoxy laminates under consideration of delamination 24. Naderi M, Khonsari MM. Thermodynamic analysis of
growth. Int J Numer Methods Eng. 1989;27(2):323‐341. fatigue failure in a composite laminate. Mech Mater.
12. Dahlen C, Springer GS. Delamination growth in composites 2012;46:113‐122.
under cyclic loads. J Compos Mater. 1994;28(8):732‐781. 25. Naderi M, Khonsari MM. On the role of damage energy in the
13. Bucinell RB. Development of a stochastic free edge delamination fatigue degradation characterization. Composites: Part B.
model for laminated composite materials subjected to constant 2013;45(1):528‐537.
amplitude fatigue loading. J Compos Mater. 1998;32 26. Mohammadi B, Mahmoudi A. Developing a new model to pre-
(12):1138‐1156. dict the fatigue life of cross‐ply laminates using coupled CDM‐
14. Henaff‐Gardin C, Lafarie‐Frenot MC. The use of a characteristic entropy generation approach. Theor Appl Fract Mech.
damage variable in the study of transverse cracking development 2018;95:18‐27.
under fatigue loading in cross‐ply laminates. In: International 27. Mahmoudi A, Mohammadi B. On the evaluation of damage‐
Journal of Fatigue. 2002;24.2–4:389–395. entropy model in cross‐ply laminated composites. Eng Fract
15. Schon J. model of fatigue delamination in composites. Compos Mech. 2019;219:106626.
Sci Technol. 2000;60:553‐558. 28. Huang ZM. Simulation of the mechanical properties of fibrous
16. Hosseini‐Toudeshky H, Farrokhabadi A, Mohammadi B. Imple- composites by the bridging micromechanics model. Compos A:
mentation of a micro‐meso approach for progressive damage Appl Sci Manuf. 2001;32(2):143‐172.
analysis of composite laminates. Struct Eng Mech. 2012;43 29. Herakovic C. T. Mechanics of Fiberous Composites. New york
(5):657‐678. 1998.
17. Ladeveze P, Le Dantec E. Damage modelling of the elementary 30. Taheri‐Behrooz F, Shokrieh MM, Lessard LB. Progressive
ply for laminated composites. Compos Sci Technol. 1992;43 fatigue damage modelling of cross‐ply laminates, II: Experimen-
(3):257‐267. tal Evaluation. J Compos mater. 2010;44(10):1261‐1277.
18. Talreja R. Stiffness properties of composite laminates with
matrix cracking and interior delamination. Eng Fract Mech.
1986;25(5‐6):751‐762.
19. Payan J, Hochard C. Damage modelling of laminated
How to cite this article: Mahmoudi A,
carbon/epoxy composites under static and fatigue loadings. Int
J Fatigue. 2002;24(2‐4):299‐306.
Mohammadi B, Hosseini‐Toudeshky H. Damage
behaviour of laminated composites during fatigue
20. Kawai M, Honda N. Off‐axis fatigue behavior of a carbon/epoxy
cross‐ply laminate and predictions considering inelasticity and loading. Fatigue Fract Eng Mater Struct. 2019;1–13.
in situ strength of embedded plies. Int J Fatigue. 2008;30(10‐ https://doi.org/10.1111/ffe.13152
11):1743‐1755.