You are on page 1of 12

Acta Mechanica Solida Sinica, Vol. 24, No.

5, October, 2011 ISSN 0894-9166


Published by AMSS Press, Wuhan, China

A DAMAGE MECHANICS MODEL FOR FATIGUE LIFE


PREDICTION OF FIBER REINFORCED POLYMER
COMPOSITE LAMINA

Wenjing Shi Weiping Hu Miao Zhang Qingchun Meng


(Institute of Solid Mechanics, School of Aeronautics Science and Engineering, BeiHang University, Beijing
100191, China)

Received 14 April 2011, revision received 27 June 2011

ABSTRACT A damage mechanics fatigue life prediction model for the fiber reinforced polymer
lamina is established. The stiffness matrix of the lamina is derived by elastic constants of fiber
and matrix. Two independent damage degrees of fiber and matrix are introduced to establish
constitutive relations with damage. The damage driving forces and damage evolution equations
for fiber and matrix are derived respectively. Fatigue tests on 0◦ and 90◦ unidirectional laminates
are conducted respectively to identify parameters in damage evolution equations of fiber and
matrix. The failure criterion of the lamina is presented. Finally, the life prediction model for
lamina is proposed.

KEY WORDS fiber reinforced polymer composite lamina, continuum damage mechanics, fatigue
life prediction, fiber breakage, matrix cracking

I. INTRODUCTION
Composites have many engineering applications owing to their excellent properties, and some of
these applications involve components subjected to cyclic loading which always causes damage and
deterioration in material. To ensure the high reliability of the structures during lifetime, it is important to
evaluate the degraded properties and fatigue remaining life accurately. Since 1970s, the issue of composite
fatigue has been extensively explored and many achievements have been obtained[1–4] . However, due
to the complexity of the fatigue damage mechanism for composites, many problems have not been well
resolved yet.
Existing approaches to the fatigue problems of composites can be divided into two classes: the
S-N curve method and the damage accumulation theory. The S-N curve method[1–5] has been widely
employed in engineering to deal with the fatigue issue of composites. However, only under the conditions
of low stress and simple stress state, is the method suitable. The damage accumulation theory, which
can be applied under complex loading conditions, is a hotspot in researches of the fatigue of composites.
Multitudinous approaches, such as the residual strength model[6–8] , residual stiffness model[9, 10] , energy
model[11–13] , and Markov chain model[14, 15] , have been conducted with respect to different damage
variables and different research methodologies.

 Corresponding author. E-mail: daizi8429@163.com


 Project supported by the FanZhou Science and Research Foundation for Young Scholars.
· 400 · ACTA MECHANICA SOLIDA SINICA 2011

Continuum damage mechanics and meso-damage mechanics are two branches of damage mechanics.
Meso-damage mechanics[16] employs a representative volume element to provide an insight into the
influence of microstructures on the failure behaviors and the damage mechanisms of materials. However,
due to the complexity of the micromechanisms and a number of factors having impact on damage
behaviors, a micromechanical model generally contains many microstructural parameters which are
difficult to be identified, and is based on a series of assumptions for simplicity yielding the results
quantitatively unsatisfactory. Hence, it is not readily applicable for engineering applications. On the
other hand, continuum damage mechanics[17, 18] applies the phenomenological method to study the
mechanical behavior of damage on the basis of continuum thermodynamics and continuum mechanics.
As the physical mechanisms are not deeply considered, the phenomenological model usually has an
unsubstantial physical foundation and can not explain some complicated phenomena with respect to
microstructure evolution. However, without considering the damage mechanisms, a phenomenological
one characterizes the damage effects in terms of an internal variable in such a fashion as the scalar,
vector, or tensor, which leads the framework relatively simple and the engineering application more
convenient. For composite fatigue damage, even though the phenomenological one is more applicable,
the parameters, which are associated with the independent variables of the stiffness matrix with damage,
also are too numerous to identify easily by fitting the results of experiments.
In this article, we aim to establish a continuum damage mechanics model to predict the fatigue life of
fiber reinforced polymer composite lamina. Firstly, the stiffness matrix of the fiber reinforced polymer
composite lamina is expressed in terms of the elastic constants of the fiber and the matrix. Secondly, two
independent damage variables, the damage degree of the fiber and the damage degree of the matrix, are
introduced to derive the constitutive relation taking account of damage coupling effects for fiber reinforced
polymer composite lamina. Thus the damage state of lamina can be characterized by the damage of the
fiber and matrix. Thirdly, the damage driving forces are derived based on the damage thermodynamics,
and the damage evolution equations of the fiber and the matrix are presented respectively. Then, the
tension-tension fatigue experiments on the standard specimens of the unidirectional laminate with 90◦
and 0◦ ply angles are performed to determine the parameters in the damage evolution equations of the
matrix and fiber. Fourthly, the failure criterion for life prediction of the lamina is proposed through
studying two independent damage evolution processes in lamina, which are related to the fiber breakage
and matrix cracking. Finally, the model for fatigue life prediction of the angle-ply lamina is constructed,
and examples of fatigue life prediction of the lamina with different ply angles are given. The proposed
model provides a new theoretical method for fatigue life prediction of fiber reinforced polymer composite
lamina quantitatively, which transforms the complex issue of composite lamina fatigue into the relative
simple damage analysis of the fiber and matrix. Another obvious advantage of the above model is that,
among others, the number of parameters in the damage evolution equation is significantly reduced and
only a few fatigue tests on standard specimens are needed. It overcomes the difficulty of parameter
identification in previous damage mechanics models and is convenient for practical applications.

II. CONSTITUTIVE RELATION WITH DAMAGE


Figure 1 sketches a typical fiber reinforced poly-
mer composite lamina, in which the hatchings
stand for the fiber. Refer to the material coordinate
system (1,2) and the natural coordinate system
(x, y) originating at the lamina center, as shown
in Fig.1, where the 1-axis parallels to the fiber, 2-
axis is perpendicular to the fiber, and θ is defined
as the off-axis angle, and it is supposed that the
counterclockwise turning from x axis to 1 axis is Fig. 1 A fiber reinforced polymer composite lamina.
positive.

2.1. Constitutive Relation for Fiber Reinforced Polymer Composite Lamina without Damage
The constitutive relation for the angle-ply lamina is
σ xy = Q̄εxy (1)
Vol. 24, No. 5 Wenjing Shi et al.: A Damage Mechanics Model for Fatigue Life Prediction · 401 ·

where the subscript xy refers to the natural coordinate system (x, y), as defined in Fig.1, σ xy is the
stress component, εxy is the strain component, and Q̄ is the off-axis stiffness matrix with respect to
the xy-axes, which can be calculated by transforming the stiffness matrix Q, referred to the 12-axes in
the following form:
Q̄ = W Q (2)
where W denotes the transformation matrix[17] , which only depends on the off-axis angle θ.
A unidirectional fiber reinforced polymer composite lamina is anisotropic, and, in the state of plane
stress, the constitutive relation for the lamina with respect to the material coordinate system (1,2) can
be written as
σ 12 = Qε12 (3)
where the subscript 12 corresponds to the material coordinate system (1,2), as defined in Fig.1, σ 12 and
ε12 denote the stress component and the strain component, respectively. The principal axis stiffness
matrix Q can be composed in terms of only four independent elastic constants: the elastic modulus in
the fiber direction E1 , the elastic modulus in the transverse direction E2 , the value of Poisson’s ratio
ν12 , and the in-plane shear modulus G12 .
According to the micromechanics, we can obtain the aforementioned elastic constants of the lamina
in terms of the elastic constants of the fiber and the matrix as[19]
E1 = E1 (Ef , Vf , Em ) = Ef Vf + Em (1 − Vf )
Ef Em [Vf + η2 (1 − Vf )]
E2 = E2 (Ef , Vf , Em ) =
Ef η2 (1 − Vf ) + Em Vf
(4)
ν12 = ν12 (νf , νm , Vf ) = νf Vf + νm (1 − Vf )
Gm Gf [Vf + η12 (1 − Vf )]
G12 = G12 (Gm , Gf , Vf ) =
Gf η12 (1 − Vf ) + Gm Vf
where the subscript f stands for the fiber, m stands for the matrix (the same below), Em and Ef are the
elastic moduli of fiber and matrix, νf and νm are values of the major Poisson’s ratio of fiber and matrix,
respectively, Gf and Gm are the shear moduli of fiber and matrix, Vf is the fiber volume fraction, and
η2 and η12 are the correction factors determined by the experiment.
By substituting Eq.(4) into Eqs.(1)-(3), the elastic constants and constitutive relation of the lamina
are expressed by the elastic constants of the fiber and matrix.

2.2. Constitutive Relation with Damage


The constitutive relation for linear elastic materials without damage is[17]

σij = δij λδkl εkl + 2μεij (5)

where σij and εij are the stress


 components and the strain components, respectively, δij and δkl are
1, i = j
the Kronecker symbols: δij = , λ and μ are the Lame constants
0, i = j
Eν E
λ= , μ=G= (6)
(1 + ν)(1 − 2ν) 2(1 + ν)
where E refers to the Young’s modulus without damage, G is the shear modulus without damage, and
ν is the Poisson’s ratio without damage.
The damage degree is introduced to characterize the deterioration of materials under the cyclic
loading:
E − ED
D= (7)
E
where ED is the Young’s modulus with damage. As ED ranges from E to 0, D varies between 0 and 1.
Substituting Eq.(7) into Eq.(6), the Lame constants with damage are rewritten as

λD = (1 − D)λ, μD = (1 − D)μ (8)


· 402 · ACTA MECHANICA SOLIDA SINICA 2011

The constitutive relation with damage is derived by substituting Eq.(8) into Eq.(5), that is

σij = (1 − D)δij λδkl εkl + 2(1 − D)μεij (9)

For both the fiber and matrix can be considered as linear elastic materials, the damage degrees of
fiber and matrix are defined as
Ek − ED,k
Dk = (10)
Ek
where the subscript k = f, m, similarly to Eq.(4), and f and m stand for the fiber and matrix, respectively.
Substituting Eq.(10) into Eq.(9), the constitutive relations for the fiber and matrix taking account of
damage coupling effects are formulated respectively in the following:

σij,f = (1 − Df )δij λf δkl εkl,f + 2(1 − Df )μf εij,f (11)


σij,m = (1 − Dm )δij λm δkl εkl,m + 2(1 − Dm )μm εij,m (12)

For the isotropic material, there is

ED E
GD = μD = = (1 − D) (13)
2(1 + ν) 2(1 + ν)

where GD is the shear modulus with damage.


From Eqs.(4), (10) and (13), the elastic constants with damage of the lamina can readily be written
as
E1,D = ED,f Vf + ED,m (1 − Vf ) = (1 − Df )Ef Vf + (1 − Dm )Em (1 − Vf )
ED,f ED,m [Vf + η2 (1 − Vf )] (1 − Df )Ef (1 − Dm )Em [Vf + η2 (1 − Vf )]
E2,D = =
ED,f η2 (1 − Vf ) + ED,m Vf (1 − Df )Ef η2 (1 − Vf ) + (1 − Dm )Em Vf
(14)
ν12,D = νf Vf + νm (1 − Vf )
GD,m GD,f [Vf + η12 (1 − Vf )] (1 − Dm )Gm (1 − Df )Gf [Vf + η12 (1 − Vf )]
G12,D = =
GD,f η12 (1 − Vf ) + GD,m Vf (1 − Df )Gf η12 (1 − Vf ) + (1 − Dm )Gm Vf
Thus, the components of the principal axis stiffness matrix taking account of damage coupling effects,
QD , can be obtained as
2
E1,D
Q11,D = 2
E1,D − ν12,D E2,D
E1,D E2,D
Q22,D = 2 (15)
E1,D − ν12,D E2,D
Q66,D = G12,D
ν12,D E2,D E1,D
Q12,D = Q21,D = 2
E1,D − ν12,D E2,D
Note that θ keeps constant during damage evolution. Substituting Eq.(14) into Eqs.(15) and (2),
the off-axis stiffness matrix of lamina with damage, Q̄D , can be calculated by

Q̄D = W QD (16)

Then, the constitutive relation with damage for the angle-ply lamina can be obtained as

σ xy,D = Q̄D εxy,D (17)

In summary, the issue of fatigue damage for the lamina is transformed into the fatigue damages of
the fiber and matrix. Hence, the key problem becomes how to analyze the damage evolutions of fiber
and matrix.
Vol. 24, No. 5 Wenjing Shi et al.: A Damage Mechanics Model for Fatigue Life Prediction · 403 ·

III. DAMAGE DRIVING FORCES AND DAMAGE EVOLUTION EQUATIONS


3.1. The Damage Driving Forces and the Damage Evolution Equations of the Fiber and Matrix
Because fatigue failure is an irreversible thermodynamic process, according to the law of thermo-
dynamics, the damage driving force can be presented in the form[17]
∂f
Y = −ρ (18)
∂D
where f is the free energy per unit mass, and ρ is the medium mass density. During the isothermal
process, there is 
ρf = W = σij dεij (19)

where ρf refers to the free energy per unit volume, and W is the strain energy density.
Employing the constitutive relations in terms of stress components:
 
1 (1 + ν) ν
εij = σij − δij δkl σkl (20)
(1 − D) E E
the strain energy density with damage can be rewritten as
   
1 1 1 2
W = (1 − D) λ(δij εij )2 + μεij εij = (1 + ν) σij σij − ν (δij σij ) (21)
2 2E 1 − D
Thus, we can get the damage driving forces of fiber and matrix respectively as
∂Wk Wk
Yk = − = (k = f, m) (22)
∂Dk 1 − Dk
The following damage evolution equations are introduced to characterize the damage evolution of
the fiber and matrix, that is[17] :
Bk
dDk Ak Ymax,k
= (k = f, m) (23)
dN (1 − Dk )Ck
where Ak , Bk and Ck are the material parameters, Ymax,f and Ymax,m are the maximal damage driving
forces for the fiber and matrix, respectively.
In the case of uniaxial tension, the damage evolution equations expressed in terms of stress can be
derived as
2Bk
dDk Ak σmax,k
= B
· (k = f, m) (24)
dN (2Ek ) k (1 − Dk )2Bk +Ck
where σmax,k denotes the maximal tensile stress and remains the same in the stress controlled experiment.
Integrating Eq.(24) from D = D0,k to D = 1, we can get
B
2Bk (2Ek ) k
σmax,k · Nk = (1 − D0,k )2Bk +Ck +1 (k = f, m) (25)
Ak (2Bk + Ck + 1)
where D0,k refers to the initial damage degree of the material. Taking logarithm of Eq. (25), the relations
between the logarithm life and the maximal stress can be obtained as
B
(2Ek ) k
log Nk = log (1 − D0,k )2Bk +Ck +1 − 2Bk log σmax,k (k = f, m) (26)
Ak (2Bk + Ck + 1)

3.2. Identification of the Parameters in Matrix Damage Evolution Equations


Fatigue experiments on the matrix material are necessary for the identification of the material
parameters in the damage evolution equation of matrix. However, there are few data available from
the fatigue test on the single matrix material. As the damage of fiber evolves rarely when the 90◦
unidirectional laminate is subjected to uniaxial tension, the crucial failure mode of this laminate is
matrix cracking. In other words, the fatigue life of the 90◦ unidirectional laminate depends on the
· 404 · ACTA MECHANICA SOLIDA SINICA 2011

fatigue properties of matrix entirely. Hence, in this article, specimens of the 90◦ unidirectional laminate
replacing specimens of single matrix material are tested for identifying the parameters in matrix damage
evolution equations.
Specimens of 90◦ unidirectional laminates are laid by the E-glass fiber reinforced unsaturated
polyester resins GT200(GFRP), and some material constants are listed in Table 1. The fatigue tests
are carried out under cyclic tensile stress with ratio R = 0.1 and in the stress control mode. The
experimental data are listed in Table 2, where ‘Force’ refers to the maximal test force, σmax stands
for the maximal stress calculated by force and section areas, X̄ is the mean logarithm fatigue life, S is
standard deviation of the logarithm fatigue life, and Cv denotes the coefficient of variation. The fatigue
limit is tested by means of the up and down method, while the others are tested by the grouping test
method. (The same below).
Based on the least square method, the parameters in the damage evolution equation of matrix are
identified, as shown in Table 3.

Table 1. Mechanical properties of fiber and matrix of GFRP

Ef (GPa) νf Em (GPa) νm Vf (%)


76 0.25 3.4 0.35 52.2

Table 2. Fatigue test data of GFRP unidirectional laminate with 90◦ ply angle

Force (kN) σmax (MPa) X̄ N50 S Cv


1.50 17.1 4.4302 26926 0.06162 0.0139
1.35 15.4 4.7655 58277 0.02227 0.0047
1.20 13.7 5.2931 196370 0.24222 0.0458
1.05 12.0 5.7472 558733 0.20624 0.0359
1.01 11.6 1000000

Table 3. Parameters of the matrix damage evolution equation of GFRP

Parameters Am Bm Cm Dm,0
Values 7.11558×10−11 4.50905 277.514 0.07280

Substituting the parameters in Table 3 into Eq.(26), we can obtain the relationship curve of the
maximal stress versus fatigue life for matrix GT200. The comparison of the fitting curve and the
experimental point is sketched in Fig.2.

Fig. 2. Comparison of the fitting curve and the experiment points for the 90◦ GFRP laminate.
Vol. 24, No. 5 Wenjing Shi et al.: A Damage Mechanics Model for Fatigue Life Prediction · 405 ·

3.3. Identification of the Parameters in Fiber Damage Evolution Equations


In principle, the fatigue test on fiber bundles is the most direct approach to identifying the parameters
in the damage evolution equation of fiber. However, it is not practical to manufacture fiber bundles for
experiment.
While the tensile fatigue loading is applied on the 0◦ lamina in the direction of fiber, the following
relations
σf σm
= (27)
Ef Em
and
σf Af + σm Am = F (28)
can be easily derived, in which σm is the stress associated with matrix, σf is the stress of fiber, F
denotes the tension of the lamina, and Af and Am are the cross section areas of the fiber and the matrix,
respectively. Thus, the fiber volume fraction of the lamina, Vf , can be expressed as
Af
Vf = (29)
Af + Am
Obviously, we can calculate the stress of the lamina, σl , by the following means:
F
σl = (30)
Af + Am
From Eqs.(27)-(30), we can determine the stresses of the fiber and matrix in terms of the stress of
lamina as
σl
σf = (31)
Vf + Em (1 − Vf )/Ef
σl
σm = (32)
Ef Vf /Em + (1 − Vf )
For the fiber elastic modulus is usually far greater than the matrix, according to Eqs.(31) and (32),
the stress of fiber is generally much greater than that of matrix. We carry out the fatigue experiments
on GFRP unidirectional laminates with 0◦ off-axis angle and list the test data in Table 4.

Table 4. Fatigue test data of GFRP unidirectional laminates with 0◦ ply angle

Force (kN) σmax (MPa) X̄ N50 S Cv


24.00 274.3 3.9745 9430 0.08532 0.0215
21.00 240.0 4.3532 22553 0.31327 0.0720
18.00 205.7 4.9477 88652 0.12528 0.0253
15.00 171.4 5.4998 316022 0.21725 0.0395
12.63 144.3 1000000

The matrix stress σm corresponding to a given σmax can be calculated by using material constants
in Table 1 and Eq.(32), and then the life of matrix will be evaluated by previous parameters and
equations. It is easily confirmed that the fatigue life of matrix is much less than the experimental life
of the laminate. On the other hand, according to the fatigue fractures of the 0◦ GFRP unidirectional
laminate, it can be obviously found that the definitive failure mode for the laminate is fiber breakage.
Therefore, when the 0◦ lamina is subjected to the tensile fatigue loading, we can state that the fatigue
life of the 0◦ GFRP laminate depends on the fiber life.
Based on the analysis above, we use the tension-tension fatigue tests on the standard specimens of
the 0◦ unidirectional laminates to determine the parameters in the damage evolution equation of fiber.
For the sake of simplicity, the stress redistribution caused by the different evolution rates of fiber and
matrix is neglected. Table 5 shows the parameters in the damage evolution equation of glass fiber.
By the aforementioned means, using parameters in Table 5, the stress-life relation of glass fiber can
be figured out. We present the comparison between the fitting curve and the test points in Fig.3.
· 406 · ACTA MECHANICA SOLIDA SINICA 2011

Table 5. Parameters in the damage evolution equation of glass fiber

Parameters Af Bf Cf D0,f
Values 1.89969×10−7 3.70585 13.7888 0.04529

Fig. 3. Comparison of the fitting curve and experiment


points for the 0◦ GFRP laminates.
Fig. 4. Stress transformation for the lamina.

IV. FATIGUE LIFE PREDICTION


Fiber breakage and matrix cracking are two major failure modes of the fiber reinforced lamina in
case of the tensile fatigue loading. To establish a model for predicting the fatigue life of the lamina
incorporating the two failure modes, the fiber is assumed to bear the tensile stress along the fiber
direction only, while the matrix bears the rest in this article. In the establishment, the key step is to
transform the original tensile stress of lamina, σx , into the stresses, σ1 , σ2 , τ12 , in the material coordinate
system (1,2) in terms of Eq.(33):
⎧ ⎫ ⎧ ⎫
⎨ σ1 ⎬ ⎨ M 2 ⎬
σ2 = N2 σx (33)
⎩ ⎭ ⎩ ⎭
τ12 −M N
where M = cos θ, and N = sin θ. The transformation is sketched in Fig.4.
Considering the evolutions of the damages for both fiber and matrix at the same time, the stresses
generated from Eq.(33) are discussed as follows:
1) σ1 refers to the stress along the fiber direction. According to Eq.(31), we can obtain the stress
of fiber, σ1,f , which influences the increment of damage degree of fiber and induces the fiber breakage
failure.
2) σ2 is the stress perpendicular to the fiber. τ12 is the shear stress. Based on Eq.(32), the matrix
stress, σ1,m , is derived from σ1 . All the three kinds of stresses are all bore by matrix, and it means the
matrix is under the multi-axial stress state. In this case, the strain energy density can be calculated in
terms of Eq.(22). Thus, the corresponding damage driving force in the multi-axial stress state can be
written as  
2
1 σ1,m + σ22 − 2νm σ1,m σ2 2
τ12
Ym = + (34)
2(1 − Dm )2 Em Gm
which is the foundation of the analysis on the damage degree of the matrix, and induces the matrix
cracking.

4.1. Failure Criterion


In the presented model, the evolutionary damages of fiber and matrix are coexisting and independent.
Incorporating the two damages, the fatigue failure criterion of fiber reinforced composite lamina is
proposed as the stiffness matrix of lamina singularity under the boundary condition, which, in other
words, means that the lamina fatigue failure should occur as long as the determinant of coefficient of
Vol. 24, No. 5 Wenjing Shi et al.: A Damage Mechanics Model for Fatigue Life Prediction · 407 ·

stiffness matrix with damage for the lamina is zero:


 
Q̄ij  = 0 (35)
D

which is equivalent to

⎨ Df = 1 and Dm = 1 (θ = 0◦ )
Dm = 1 (θ = 90◦ ) (36)

Dm = 1 or Df = 1 and Dm = 1 (θ = 0◦ and θ = 90◦ )

where Df = 1 means the fiber breakage , and Dm = 1 denotes the matrix cracking.
4.2. Closed Solution of the Fatigue Life for the Lamina
Substituting Eq.(34) into Eq.(23), the life of the matrix subjected to both tension and shear loadings
is determined:
 1
2Bm (1 − Dm )2Bm +Cm
Nm,θ =    Bm dDm
D0,m Am σx 2 cos4 θ/H 2 + sin4 θ − 2ν cos2 θ sin2 θ/H /E + σ 2 cos2 θ sin2 θ/G
m m x m
(37)
in which
Ef (1 − Df )Vf
H= + (1 − Vf ) (38)
Em (1 − Dm )
The life of fiber can be derived in terms of the fiber damage evolution equation by the following
means:
 1 B
(2Ef ) f (1 − Df )2Bf +Cf {Vf + Em (1 − Dm )(1 − Vf )/[Ef (1 − Df )]}2Bf
Nf = dDf (39)
D0,f Af σx2Bf
Then, the closed solution of fatigue life for the lamina can be obtained according to Eqs.(37) and
(39), as long as the proposed fatigue failure criterion is satisfied. The damage coupling effect is reflected
in terms of the redistribution of the stress which is caused by the different and coexisting evolutions of
the Dm and the Df .

4.3. Numerical Method for Fatigue Life Prediction of the Lamina


In an attempt to obtain the quantificational fatigue life containing the damage coupling effect of
the damage field and the stress field, a numerical method is used to predict the life of the lamina as
follows:
1) Transform the stress in the natural coordinate system (x, y) into the stresses with respect to the
material coordinate system (1,2) by coordinate transformation.
2) Divide the stresses with respect to 12-axes into two parts. One is σ1 , the other is σ2 and τ12 .
Substituting σ1 into Eqs.(31) and (32), the corresponding fiber stress σ1,f and the matrix stress σ1,m
are obtained. The fiber damage evolution is determined by σ1,f , while the matrix damage evolution is
dependent on σ2 , τ12 , σ1,m .
3) Give the damage degree increment Dm , based on the initial damage degree of the matrix Dm,0 .
Then the cycle increment of matrix, Nm , can be calculated in terms of Eq.(40), as the damage degree
of the matrix, Dm , increases by Dm :
2Bm (1 − Dp,m )2Bm +Cm Dm
Nm = 2 (40)
Am [(σ1,m + σ22 − 2νm σ1,m σ2 )/Em + τ12
2 /G ]
m

and the increment of the fiber damage degree, Df , after Nm cycles can be obtained as
2Bf
Af σ1,f Nm
Df = (41)
(2Ef )Bf (1 − Dp,f )2Bf +Cf
where Dp,f and Dp,m are the initial values of damage degrees for the fiber and the matrix, respectively,
in the beginning of every increment.
· 408 · ACTA MECHANICA SOLIDA SINICA 2011

4) Calculate the new damage fields of the matrix and fiber after a cycle increment as follows:

Dp,m = Dp,m + Dm (42)

Dp,f = Dp,f + Df (43)
 
where Dp,f and Dp,m refer to the new damage degrees of the fiber and matrix respectively.
5) Judge whether
 the new damage fields satisfy the failure criterion. If so, the fatigue life of the
lamina is N = N; if not, repeat the steps from the second to the fifth until the failure criteria is
satisfied.
It must be emphasized that the value of increment Dm should be small enough to ensure that a
convergent result of Nf can be reached.

V. RESULTS AND DISCUSSION


5.1. Precision of the Parameter Identification Method for Fiber Damage Evolution Equations
As illustrated in §3.3, the parameters in the damage evolution equation of fiber are determined from
the fatigue test on 0◦ GFRP laminates, and the damage evolution of matrix is ignored for simplicity.
In other words, the stress redistribution in fiber and matrix caused by the ununiformity of damages
is not considered in the aforementioned parameter identification method. To make sure the method is
appropriate, the fatigue lives of lamina are calculated accurately, taking account of damage evolution
of both fiber and matrix, by the numerical means mentioned in §4.3 with parameters in Table 5 and
Table 3. The comparison between calculated lives and experimental data is reported in Table 6, where
X̄ refers to the mean logarithmic fatigue lives of the experiment, log Ni is the numerical solution of the
logarithmic fatigue life, errori is the relative error defined by
 2
log Ni − X̄
errori = (44)

and Δ is the sum of errori .

Table 6. Comparison between the test data and the numerical fatigue lives of the 0◦ GFRP laminates

i σmax,i (MPa) X̄ log Ni errori


1 274.3 3.9745 3.9994 3.91×10−5
2 240.0 4.3532 4.4358 3.6 ×10−4
3 205.7 4.9476 4.9409 1.83×10−6
4 171.4 5.4997 5.5409 5.62×10−5
5 144.3 6 6.1106 3.4 ×10−4
Δ 7.97 ×10−4

Table 6 shows that when the damage evolution of matrix is taken into account, the lives which are
obtained with the parameters in Table 5 agree with the experimental results. So, the fiber parameters
obtained without considering the matrix are sufficiently precise and the feasibility for the fitting method
is illustrated.
From both the experimental phenomenon and the calculated results, we find that the damage
evolution process can be divided into two stages. In the early stage, the damages for fiber and matrix
should evolve at different rates until the matrix cracks, which results in the redistribution of stress.
However, the modulus of the matrix is far less than that of fiber, which leads the stress redistribution
unremarkable enough to be neglected. Then, in the latter stage, there is no stress redistribution, since
the matrix failures and the residual life of lamina is decided only by the fiber. The features of the two
stages indicate that the stress redistribution of fiber and matrix has little influence on the fatigue life
of the lamina.
In conclusion, not only is it feasible and accurate to obtain the parameters in the damage evolution
equation of fiber in terms of the presented means without considering damage evolution of matrix, but
the method also is simple for application compared to the method of parameter identification with
stress redistribution between fiber and matrix.
Vol. 24, No. 5 Wenjing Shi et al.: A Damage Mechanics Model for Fatigue Life Prediction · 409 ·

5.2. Life Prediction for the Lamina with Different Ply Angles
Based on the proposed model, the fatigue lives of the angle-ply lamina are calculated and plotted
in Table 7.

Table 7. The fatigue lives of the GFRP lamina with different ply angles

Stress (MPa)
log N
30.0 24.0 20.0 15.4
0◦ >6 >6 >6 >6
30◦ 3.345 4.219 4.933 5.950
θ
45◦ 2.394 3.269 3.983 5.000
90◦ 2.215 3.090 3.804 4.929

In Table 7, the fatigue lives of lamina with four common off-axis angles under four different tensile
fatigue loadings are given. As the given load is much lower than the fiber limit force, the fatigue lives
of the 0◦ lamina can be found to be infinite.
Under the condition of the same cyclic load, the fatigue lives of the lamina associated with the ply
angles 0◦ , 30◦ , 45◦ and 90◦ decrease successively. The failure behaviors of the lamina with different
off-axis angles come out to be in similar modes, namely, the matrix cracks firstly, and, the fiber failure
supervenes immediately, for the fiber can not bear the tension not along the fiber direction.

VI. CONCLUSIONS
Incorporating the failure modes of fiber breakage and matrix cracking, we have proposed a new
damage mechanics model to predict the fatigue life of fiber reinforced polymer lamina and adopted
the singularity of stiffness matrix as the failure criterion of lamina in this article, which inventively
transformed the complex anisotropic issue of composite lamina fatigue into the analysis of single-variable
isotropic damages for fiber and matrix.
Furthermore, the parameters in the fiber and matrix damage evolution equations are identified
creatively by the normal fatigue tests data of 90◦ and 0◦ unidirectional laminates, which not only
avoids the unusual fatigue tests on the pure matrix material and fiber bundles, but also significantly
reduces the number of fatigue tests required for identifying, yielding applications more convenient.
To conclude, the damage mechanics model proposed above provides a novel and thinkable path to
study the damage accumulation of composite fatigue and supplies a sufficiently theoretical foundation
for the further research on the fatigue life prediction of the composite laminate.

References
[1] Talerja,R., Fatigue of Composite Materials. Lancaster: Technomic Publishing Company, 1987.
[2] Reifsnider,K.L., Fatigue of Composite Materials. New York: Elsevier Science Publishers, 1991.
[3] Hashin,Z. and Rotem,A., Fatigue failure criterion for fiber reinforced materials. Journal of Composite Ma-
terials, 1973, 7(10): 448-464.
[4] Wang,F., Zeng,X.G. and Zhang,J.Q., Predictive approach to failure of composite laminates with equivalent
constraint model. Acta Mechanica Solida Sinica, 2010, 23(3): 240-247.
[5] Gamstedt,E.K. and Talreja,R., Fatigue damage mechanisms in unidirectional carbon-fiber-reinforced plas-
tics. Journal of materials science, 1999, 34: 2535-2546.
[6] Quaresimin,M., Susmel,L. and Talreja,R., Fatigue behaviors and life assessment of composite laminates
under multiaxial loadings. International Journal of Fatigue, 2010, 32: 2-16.
[7] Tang,C.Y., Fan,J.P. and Tsui,C.P., et al., Quantification of shear damage evolution in aluminum alloy
2024T3. Acta Mechanica Solida Sinica, 2007, 20(1): 57-64.
[8] Yao,W.X. and Himmel,N., A new cumulative fatigue damage model for fiber-reinforced plastics. Composites
Science and Technology, 2000, 60: 59-64.
[9] Talerja,R., Stiffness properties of composite laminates with matrix cracking and interior delamination,
Engineering of Fracture Mechanics, 1986, 25: 751-762.
[10] Zhang,J.Q. and Hermann,K.P., Stiffness degradation Induced by multilayer intralaminar cracking in com-
posite laminates. Composites Materials, 1999, 30: 683-706.
· 410 · ACTA MECHANICA SOLIDA SINICA 2011

[11] Bathias,C., An engineering point of view about fatigue of polymer matrix composite materials. International
Journal of Fatigue, 2006, 28: 1094-1099.
[12] E.Kadi,H. and Al-Assaf,Y., Energy-based fatigue life prediction of fiberglass/epoxy composites using mod-
ular neural networks. Composite Structures, 2002, 57(1): 85-89.
[13] Takahara,A.tsushi, et al., Effect of glass fiber-matrix polymer interaction on fatigue characteristic of short
glass fiber reinforced ploy (butylene tereph thalate) based on dynamic viscoelastic measurement during
the fatigue process. Journal of Polymer Science, Part B, 1994, 32: 174-181.
[14] Kozin,F. and Bogdanoff,J.L., Recent thoughts on probabilistic fatigue crack growth. Applied Mechanics
Reviews, 1989, 42(11): 121-127.
[15] Ganesan,R., Astochastic cumulative damage model for the fatigue response of laminated composite. ICCM-
11, 1997: 145-156.
[16] Feng,X.Q. and Yu,S.W., Damage micromechanics for constitutive relations and failure of microcracked
quasi-brittle materials. International Journal of Damage Mechanics, 2010, 19(11): 911-948.
[17] Zhang,X. and Zhao,J., Applied Fatigue Damage Mechanics of Metallic Structural Members. Beijing: Na-
tional Defence Industry Press, 1994.
[18] Lematre,J., A Course on Damage Mechanics. New York: Springer-Verlag, 1992.
[19] Research Institute of Aerospace Industry Ministry, Composites Design Handbook. Beijing: Aviation In-
dustry Press, 1990.

You might also like