Sie sind auf Seite 1von 20

doi: 10.1111/ffe.

12254

Low-cycle fatigue delamination initiation and propagation in fibre


metal laminates

F. MAZAHERI1 and H. HOSSEINI-TOUDESHKY2


1
Department of Aerospace Engineering, Amirkabir University of Technology, Tehran, Iran, 2Fatigue and Fracture Lab., Department of Aerospace
Engineering, Amirkabir University of Technology, Tehran, Iran
Received Date: 27 April 2014; Accepted Date: 24 September 2014; Published Online: 2014

A B S T R A C T The aim of this paper is to develop an elastic–plastic-damage constitutive law and a tool
for simulation of delamination initiation and propagation in fibre metal laminates
(FMLs) under low-cycle fatigue loading regime. In the previous studies, the significance
of plasticity in delamination growth and modelling of FMLs was not considered. Hence,
cohesive zone law that combines the damage evolution with plasticity is developed. The
new fatigue damage model is implemented as user-written subroutines that links with
ANSYS based on the cohesive finite element method. The cohesive zone model constitu-
tive law has been verified by modelling of the delaminated adhesively bonded aluminium
joint under normal and shear loadings and compared with the available results in the
literature. The developed procedure and tool have been used for the analyses of DCB
and ENF specimens under uniform and variable loadings. The obtained results for
progressive damage and delamination and stress–strain curves are discussed in this paper.

Keywords cohesive zone; delamination; fibre metal laminate; low-cycle fatigue; plasticity.

NOMENCLATURE 1,2,3 = components of reference coordinate system


b, c = material parameters related to Coffin–Manson relation
bf, cf = parameters derived from calibration tests of low-cycle fatigue
B = transformation matrix that depends on nodal coordinates and on deriva-
tions of approximation functions
C(d) = inverse of the elastic-damage tensor
d = scalar damage parameter
dmon = monotonic contribution of the damage parameter
dcyc = cyclic contribution of the damage parameter
Ded = elastic-damage tensor that relates the CZM stresses to the elastic strains
Depd = elastic–plastic-damage stiffness matrix
D0ij = undamaged penalty elasticity tensor
E11 = Young’s modulus in the longitudinal direction-1
E22, E33 = Young’s modulus in the transverse direction-2, and through the thick-
ness direction-3
f = generic nonlinear function of dcyc and Vα
G12, G13, G23 = shear moduli
Gc = critical strain energy release rate in the mixed-mode fracture condition
GT = sum of strain energy release rates in various pure modes
GIc = critical strain energy release rate in mode I

Correspondence: H. Hosseini-Toudeshky. E-mail: Hosseini@aut.ac.ir

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20 1
2 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

GIIc = critical strain energy release rate in mode II


h0 = CZM element thickness
H = hardening modulus
K = CZM penalty stiffness
K Te = stiffness matrix of each element
KT = CZM penalty stiffness in plastic part
m = degree of nonlinearity of damage evolution law
N = cycle number
Nf = cycle number at final failure
R = displacement ratio
s, t, n = components of local coordinate system
u, v, w = displacement in the direction of x,y,z, respectively
Vα = causal variables related to the excitation that causes the damage
phenomenon
α = maximum applied effective strain
β = mode mixing ratio
γ0 ; γ0tn ; γ0sn ; ε0n = single-mode shear and normal strains at softening onset
γ_ = plastic multiplier
δij = Kronecker delta function
ε = total strain tensor
εe = elastic strain tensor
εp = plastic strain tensor
ε__p = plastic strain increment
εp = effective plastic strain
ε_
p
= increment of effective plastic strain
εm = effective strain of cohesive zone model element under the mixed-mode
condition at integration point
ε0m = effective strains at delamination onset and complete de-cohesion
f
εm = effective strains at complete de-cohesion
εxz = shear strain
εz = normal strain
η = material parameter obtained from experiments
v12, v13, v23 = Poisson’s ratios
ρ = reference mass density
σ = cohesive zone model stress tensor
σ 0n ; σ 0 = normal cohesive zone model strength
σy = yield stress
δmin, δmax = minimum and maximum displacement values in a cycle
σz = normal stress
τ 0, τ 0 = shear cohesive zone model strength
τ xz = shear stress
ϕ = yield function
ψ = Helmholtz free energy per unit mass
ψ ed = elastic-damage contribution to the free energy
ψp = plastic contribution to the free energy respectively
ψ0 = a convex function
Ωe = represents the sub-region allocated by element

INTRODUCTION composite layers. Currently, FMLs such as GLARE


(glass fibre/aluminium), ARALL (aramid fibre/aluminium)
Fibre metal laminates (FMLs) consist of alternating and CARAL (carbon fibre/aluminium) attract a number of
layers of aluminium alloy and fibre-reinforced composite, aircraft manufacturers.1,2 FMLs have been successfully in-
which are adhesively bonded using the matrix material of troduced into the Airbus A380.3–5 For example, ARALL

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 3

has been developed for the lower wing skin panels of zone model (CZM) was first suggested by Dugdale,23
the former Fokker 27 aircraft and the cargo door of and later, Barenblatt24 and Hillerborg et al.25 added
the Boeing C-17,6,7 and GLARE was selected for the significant contributions. Since then, many authors have
Boeing 777 impact resistant bulk cargo floor. It has published papers on the cohesive zone approach or devel-
been well documented that these hybrid materials display opment of CZM elements for progressive damage
improved characteristics in terms of weight saving and analyses of components. Cohesive zone modelling is a
impressive fatigue and damage properties of fibre- powerful method to consider the material nonlinearity
reinforced composite materials in addition to fire resis- of bonded joints in both monotonic and cyclic loading;
tance and durability of metals.8 The fibres have bridging hence, it is capable to model low-cycle fatigue condition.
effect; that is, they are less sensitive to fatigue loading Hosseini-Toudeshky et al.26 investigated concurrent
and transfer the loads over a crack, so lower the stress transverse cracking and induced delamination propaga-
intensity at the crack front and reduce the crack growth tion using a generalized micro-meso approach and exper-
rate. As a result, the crack growth lives of the FMLs have imental validation. A layer-wised finite element model
been significantly extended comparing to the monolithic was developed to predict the delamination growth of
aluminium panels.9,10 post-buckled composite laminates under quasi-static
Nowadays, it is often stated that fatigue accounts for at compression loading using cohesive elements.27 Further-
least 90% of all service failures due to mechanical more, cohesive zone modelling was employed in Ref. [28]
reasons. Fatigue failure is specifically insidious because to study the effects of delamination size and stacking
it occurs without any obvious warning. Therefore, meth- sequence of laminates on delamination growth under
odology for fatigue failure prediction is of immense quasi-static compression loading. Hosseini-Toudeshky
importance in practice.11 Damage evolution mechanism et al.29 also proposed a CZM finite element model to
is one of the important focuses on fatigue life prediction simulate the progressive debonding between composite
of components made of composites, aluminiums and patch and cracked aluminium panels concurrent with
other hybrid materials.12,13 For low-cycle fatigue condi- the crack growth in the base panel under high-cycle
tions, the stress is large enough to create plastic strain; fatigue loading. Hosseini-Toudeshky et al.30 also
hence, fatigue failure results from cyclic strain under developed de-cohesive law to simulate delamination
several thousand cycles. Low-cycle fatigue is usually growth of post-buckled laminate under high-cycle
characterized by the Coffin–Manson relation, best fatigue loading. The proposed finite element model
described by the material relation between plastic strain considered the geometry nonlinearity of post-buckling
amplitude and life, which is known as the strain-life and material nonlinearity of CZM simultaneously.
curves.14,15 The novelty of this work is considering elastic–
In FMLs, the fatigue phenomenon involves several plastic-damage constitutive law for cohesive zone
damage mechanisms that result in the degradation of modelling that is valid for low-cycle fatigue regime.
mechanical properties. One of the most important Previously developed models are valid for high-cycle
fatigue damage mechanisms is the interlaminar delam- fatigue regime considering the Paris law for damage
inations.16–18 Guo et al.19 derived an accurate expres- growth. The origin of considered irreversibility in
sion and test method for the delamination growth some of the previously performed works was not plas-
under high-cycle fatigue loading with different stress ticity that occurs in low-cycle fatigue regime. In this
ratio in hybrid bonded laminates. Lin20 performed an paper, the delamination growth of FMLs under low-
experimental study to examine the fatigue delamina- cycle fatigue loading is analysed. For this purpose,
tion growth in CARAL under cyclic loading. Guo 20-node three-dimensional CZM elements with small
and Wu21 developed an analytical model for fatigue thickness have been developed as a user element in
crack growth (FCG) in FMLs considering bridging the ANSYS finite element software. The mixed-mode
effect under various cyclic stress levels and stress ratio bilinear elastic–plastic-damage constitutive law with a
in high-cycle fatigue regime. The predicted crack damage evolution derived from the Coffin–Manson
growth rates were in good agreement with the experi- relation has been used in order to simulate the
mental results. mixed-mode stress field at the delamination front.
Using fracture mechanics approach in mixed-mode Therefore, the new proposed CZM links plasticity
progressive delamination analysis requires complex parameters to damage mechanics. The damage param-
moving mesh techniques to advance the initial crack eter is a function of the loading conditions as well as
front. One of the most appealing continuum damage the experimentally determined material parameters
mechanical techniques to predict the initiation and for low-cycle fatigue. Various loading conditions for
growth of delamination instantaneously without the need GLARE3 are considered for the analyses, and the
of remeshing is the cohesive zone approach.22 Cohesive obtained results are discussed in this paper.

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
4 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

THEORETICAL FORMULATIONS determine the consistent elastic–plastic-damage stiffness


matrix with integration scheme used for the determina-
Solutions and analyses of complex mechanical phenom- tion of the stresses in the integration points in part 2.3
ena without assumptions and simplifications can be a very (category (a)). Finally, the total damage is assumed to be
cumbersome task or impossible in many cases. As the the sum of the damage caused by the monotonic loading
issue in this paper is the analysis of low-cycle fatigue and the damage resulted from the cyclic loading (assump-
delamination initiation and propagation analyses in tion of category (a)). To determine the damage parame-
FMLs, we are faced with several complexities in terms ter in low-cycle fatigue, the differential equation of
of loading, plasticity, damage and anisotropic multi- damage accumulation developed by Lemaitre and
material component. All assumptions are briefly Chaboche in 199632 is used. For this purpose, a general
explained here, and they can be also followed in the nonlinear damage evolution law relating cyclic damage
following sections of the paper through the equations, parameter and causal variables is considered. Then this
results presentations and discussions. The main assump- relation is simplified by the assumption of decoupling be-
tions invoked in different sections can be divided into tween the cyclic damage and casual variables (assumption
two categories: of category (b)). The plastic strain increment in each
cycle is considered as the only causal variable of this non-
a. Those related to specific material behaviour.
linear damage evolution. This increment of plastic strain
b. Those related to neglecting specific effects or simpli-
in the low-cycle fatigue regime is related to the number
fying relations.
of cycles to failure and the material parameters of the
For developing the elastic–plastic cohesive zone Coffin–Manson rule (assumption of category (a)). In
constitutive law, the Helmholtz free energy is calculated Lemaitre and Chaboche model,32 the contribution of
with the hypothesis of decoupling between elastic- the elastic part, that is, high cycle part of strain in the
damage and plastic hardening parts (category (a)). In Coffin–Manson rule is negligible with respect to the total
the elastic-damage part of Helmholtz free energy, we part of damage (assumption of category (b)). A linear
neglect the compressive normal stress effect because of accumulation rule for damage evolution law is also
preventing the penetration of layers (category (b)). For considered for determining of the cyclic damage parameter
the three-dimensional CZM element with small thick- (assumption of category (b)).
ness, we assume the penalty stiffness to be equal to the
slope of the stress–strain curve (assumption of category
(a)). We derived stress–strain relation using the second Elastic–plastic CZM constitutive law
principle of thermodynamics in the form of the The constitutive equation for the elastic–plastic-damage
Clausius–Duhem inequality (assumption of category theory associated with cohesive zone modelling in low-
(a)). A bilinear cohesive constitutive equation has been cycle fatigue regime is described in this section. In this
used in this paper. This assumption is in the category theory, it is assumed that the Helmholtz free energy per
(a). It is assumed that the onset of damage can be unit mass, ψ, is a function of state variables:33
predicted using the quadratic failure criterion (assump-
tion of category (a)), and compressive normal stresses  p 
do not affect the delamination onset (assumption of cate- ψ ¼ ψ εei ; εi ; d ; i ¼ 1; 3 (1)
gory (b)). It has been assumed that the critical strain
energy release rate per unit thickness of the cohesive p
zone is equal to the area under the effective stress–strain where εei is the elastic strain tensor, εi is the plastic strain
relation (assumption of category (a)). By this assumption, tensor, and d is the scalar damage parameter. Under the
the effective strain at the complete de-cohesion is hypothesis of decoupling between elastic-damage and
derived. Also, the B-K (Benzeggagh and Kenane31) crite- plastic hardening, the specific free energy is assumed to
rion has been used to determine the critical strain energy be given as34
release rate (assumption of category (a)). To describe the
return mapping algorithm for the integration of the    p
ψ ¼ ψ ed εei ; d þ ψ p εi ; i ¼ 1; 3 (2)
elastic–plastic cohesive zone modelling, we assume the
von Mises plasticity model with isotropic hardening
(assumption of category (a)). For the derivation of the where ψ ed and ψ p are the elastic-damage and plastic con-
integration algorithm, it is more convenient to use the tribution to the free energy, respectively. In the present
squared form of the yield function (assumption of formulation of solid-like CZM element35 shown in Fig. 1,
category (b)). To guarantee the quadratic convergence the following form is postulated for the elastic-damage
of the Newton–Raphson method, it is necessary to potential with crack closure effect:36

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 5

Fig. 1 Reference and local coordinate systems in solid-like 20 node cohesive zone model element.

    function of the damage parameter by the following


ρψ ed εei ; d ¼ ð1  d Þψ 0 εei (3) equations:
  8
d ψ 0 δ3i < εe3 > ; i ¼ 1; 3 >
< KI εm ≤ ε0m
Ded ¼ ð1  d ÞK I þ dK I c ε0m < εm < εmf ; Ic
>
: f
In this equation, the second term prevents the pene- KIc εm ≥ εm
2 3
tration of layers due to compressive normal stresses. 0 0 0
Where the < > operator indicates the zero value for neg- 60 0 0 7
¼6 4
7
ative inside value, δij is the Kronecker delta function, ρ is a < εe3 > 5
0 0
reference mass density and ψ 0 is a convex function < εe3 > (8)
defined as
  1 where εm is the effective strain of CZM element under
ψ 0 εei ¼ εei D0ij εej ; i ¼ 1; 3 ; j ¼ 1; 3 (4) the mixed-mode condition at the integration point,
2
which is determined by the following equations. ε0m and
where D0ij is undamaged penalty elasticity tensor, which f
εm are the effective strains at the delamination onset
can be determined by the following equation: and complete de-cohesion, respectively, and will be in-
troduced later.
D0ij ¼ δij K ; i ¼ 1; 3 ; j ¼ 1; 3 (5)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
εm ¼ < εn >2 þ γ2shear (9)
where K is the penalty stiffness. For zero-thickness CZM
elements, the penalty stiffness parameter is equal to the
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
slope of the stress-relative displacement curve. On the
γshear ¼ γ2sn þ γ2tn ; γsn ¼ 2ε23 ; γtn
other hand, for the solid-like CZM, this parameter is
the slope of the stress–strain curve.27 Using the second ¼ 2ε13 ; εn ¼ ε33 (10)
principle of thermodynamics in the form of the
Clausius–Duhem inequality, the state laws are obtained
Under pure mode I, II or III loading, the onset of
as follows:37
damage can be determined simply by comparing the
∂ψ stress components with their respective allowable values.
σi ¼ ρ ¼ ð1  d ÞD0ij εej (6) However, under mixed-mode loading, the damage onset
∂εei
  and the corresponding softening behaviour may occur
d D0ij δ3i < εe3 > ; i ¼ 1; 3
before any of the involved stress components reaching
their respective allowable values.38 It is assumed that
A bilinear cohesive constitutive equation has been the onset of damage can be predicted using the quadratic
used in this paper. A conveniently compact representa- failure criterion, and compressive normal stresses do not
tion of elastic-damage constitutive law can be obtained affect the delamination onset:39
by using the matrix notations:  
< σ n > 2 τ sn 2 τ tn 2
þ 0 þ 0 1¼0 (11)
σ ¼ Ded εe (7) σ 0n τ τ

where εe and σ are the elastic strain and CZM stress ten-
sors and Ded is the elastic-damage tensor, which relates In the aforementioned equation, σ 0n and τ 0 are the nor-
the stresses to the elastic strains. It can be found as a mal and shear CZM strengths. In solid-like CZM, model

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
6 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

K is the slope of the stress–strain curve. Because of the is the sum of the strain energy release rates in various
lack of material behaviour data in 1, 2 and 3 directions, pure modes.
we used the simplification assumption and considered
the same penalty stiffness in all directions. Using the Return mapping algorithm
same penalty stiffness in modes I, II and III, the stress
To describe the implicit elastic predictor/return mapping
components before softening onset are
scheme for the integration of elastic–plastic cohesive
σ i ¼ K εi ; i ¼ 1; 3 (12) zone modelling, the elastic–plastic strain split is defined
as

The single-mode normal and shear strains at soften- εe ¼ ε  εp (19)


ing onset are
where ε is the total strain tensor. Assuming isotropic
σ0 hardening von Mises model, the yield function, ϕ, is
ε0n ¼ n (13) defined as
K
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  3 T
τ0 ϕ σ ðd; ε Þ; σ y ðε Þ ¼
e p
σ Pσ (20)
γ0 ¼ γ0tn ¼ γ0sn ¼ (14) 2
K σ y ðε Þp
;
2 3
2 0 0
The mode mixing ratio, β, is also defined as 60 2 07
P¼64
7
γshear 25
β¼ (15) 0 0
εn 3
where σ y is the yield stress and εp is the effective plastic
strain. The plastic flow rule is also defined as
The effective strain at delamination onset, ε0m , is ob-
tained by the substitution of (9), (10), (12), (13), (14) rffiffiffi
∂ϕ 3 Pσ
and (15) into (11) and solving for εm: _ε p ¼ γ_ ¼γ_ pffiffiffiffiffiffiffiffiffiffiffiffi (21)
∂σ 2 σ T Pσ
8 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
>
> 1 þ β2
< ε0 γ0 where ε__p is the plastic strain increment and γ_ is the plastic
n   2 εn > 0
εm ¼
0 2
ðγ0 Þ þ βε0n (16) multiplier. The hardening variable evolution is defined as
>
>
:
γ 0
εn ≤0
ε_ ¼ γ_
p
(22)

where ε_ is the increment of the effective plastic strain.


p
It has been assumed that the critical strain energy
The loading and unloading conditions are41
release rate of the cohesive zone is equal to the area under
effective stress–strain relation. By this assumption, the γ_ ≥0 ; ϕ≤0 ; γ_ ϕ ¼ 0 (23)
effective strain at the complete de-cohesion is given by
8
>
< 2Gc For the derivation of an integration algorithm, it is
εn > 0
εfm ¼ K h0 εm
0
(17) more convenient to use the squared form of the yield
>
: f
γsn εn ≤0 function, which, in the present case, can be handled more
easily. Then, Eqs 21–23 are replaced with
where h0 is the CZM element thickness. The B-K 8
(Benzeggagh and Kenane31) criterion has been used to > 1 1
>
> ϕ ¼ σ T Pσ  σ 2y ðεp Þ
>
> 2 3
determine the critical strain energy release rate, Gc, in >
<
the mixed-mode fracture condition: ∂ϕ
ε_ p ¼ γ_ ¼ γ_ Pσ (24)
>
> ∂σ
  >
> rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
GT  GI η >
>
Gc ¼ GIc þ ðGIIc  GIc Þ (18) : ε_ p ¼ γ_ 2σ T Pσ
GT 3
where η is a material parameter in mixed-mode crack
growth conditions. It can be obtained by curve fitting The implicit elastic predictor/return mapping algo-
of the obtained experimental critical strain energy rithm starts with computing the elastic-damaged predic-
release rate versus mode mixing ratio curve,40 and GT tor state in time intervals [tn, tn + 1]:

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 7

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
8 e trial 1 1 2
> ε ¼ εen þ Δε ; Δε ¼ εenþ1  εen ðσ nþ1 ÞT Pσ nþ1  σ 2y εpn þΔγ ðσ nþ1 ÞT Pσ nþ1 ¼ 0
>
< nþ1 2 3 3
σ enþ1
trial
¼ Ded εenþ1
trial
(25) (33)
>
>
: p trial p
εnþ1 ¼ εn
where the unknowns now are the stress array, σ n + 1, and
the plastic multiplier, Δγ, and C(d) is the inverse of the
elastic-damage tensor:
The next step is to check for plastic admissibility of 8
the elastic trial state by the following computation: > ðK I Þ1 εm ≤ε0m
 1 > <
1
C ðd Þ≡ D ed
¼ ðð1  d ÞK I þ dK I c Þ εm < εm < εmf
0
1  trial T trial 1 2  p trial  >
>
:
ϕ trial ¼ σ Pσ nþ1  σ y εnþ1 (26) ðK I c Þ1 εm ≥εm
f
2 nþ1 3
(34)

If the elastic trial state is admissible, that is, if


Finally, by substituting (32) into the consistency con-
dition (33), the return mapping for the elasto-plastic
ϕ trial
≤0 (27)
CZM element is reduced to the following scalar nonlin-
ear equation having Δγ as the only unknown:
Then the process is in an elastic condition within [tn, rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
tn + 1], and we update the values by e ðΔγÞ≡ ξ ðΔγÞ  σ ε þΔγ 2 ξ ðΔγÞ
ϕ
1 1 2 p
(35)
2 3 y n 3

ð:Þnþ1 ¼ ð:Þtrial
nþ1 (28)
Here, we have conveniently defined
 T T
Otherwise, the return mapping algorithm needs to be ξ ðΔγÞ≡ σ trial
nþ1 A ðΔγÞ P AðΔγÞσ trial
nþ1 (36)
applied. The implicit return mapping in the present case
consists of solving the following system of algebraic
equations: With

εenþ1 ¼ εenþ1
trial
 ΔγPσ nþ1 (29) AðΔγÞ≡½C ðd Þ þ ΔγP 1 C ðd Þ (37)

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi From Eqs 35–37, the damage parameter is incorpo-


p 2
εnþ1 ¼ εpn þΔγ ðσ nþ1 ÞT Pσ nþ1 (30) rated with the yield function. Thus, the return mapping
3 for the elastic–plastic CZM element is carried out as fol-
lows. Firstly, the consistency Eq. (35) is solved using the
1 1 p  Newton–Raphson algorithm. Then the other variables
ðσ nþ1 ÞT Pσ nþ1  σ 2y εnþ1 ¼ 0 (31) are updated as
2 3
8
>
> σ nþ1 ¼ AðΔγÞσ trial
nþ1
p
for εenþ1, εnþ1
and Δγ where σ n + 1 is a function of >
>
Solved >
>
< εnþ1 ¼ C ðd Þσ nþ1
> e
εenþ1defined by the constitutive law (6). It is also possible
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (38)
to reduce the number of unknowns of return mapping >
> p p 2
>
> εnþ1 ¼ εn þ Δγ ξ ðΔγÞ
equations for the von Mises model. Having the incre- >
> 3
>
: P
mental plastic multiplier as the unknown, the five- εnþ1 ¼ εPn þ ΔγPσ nþ1
variable return mapping system (29)–(31) can be reduced
to a single scalar nonlinear equation. To this end, we
substitute (30) into (31) and rearrange (29) using the
inverse constitutive law (6). Then the original return Elastic–plastic-damage consistent tangent operator
mapping system reduces to
To guarantee the convergence of the Newton–Raphson
method, it is necessary to determine properly the tangent
stiffness in every iteration step. It can be obtained in a
σ nþ1 ¼ ½C ðd Þ þ ΔγP 1 C ðd Þσ trial
nþ1 (32) classical way from the stiffness matrices of all elements.

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
8 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

The stiffness matrix of each element is then defined as42  T


ξ ¼ εtrial
nþ1 E P Eεtrial
nþ1 (46)
h i h i
K Te ¼ ∫Ωe ½BT Depd ½BdΩ (39)
Using the standard relation for the differential of the
inverse of a matrix in (42) and the chain rule yields the
where [Depd] is so called elastic–plastic-damage stiffness identity as
matrix, [B] is the transformation matrix that depends on
dE ¼ E P E dΔγ (47)
the nodal coordinates and on derived of approximation
functions and Ωe represents the sub-region allocated by
the element. The quadratic convergence of the Straightforward differentiation of the aforementioned
Newton–Raphson method is ensured by the use of a con- expression (46) together with (47) gives
sistent elastic–plastic-damage stiffness matrix with an  
integration scheme used for the determination of stresses dξ ¼ 2 σ Tnþ1 P E dεtrial
nþ1  σ nþ1 PE Pσ nþ1 dΔγ
T
(48)
at the integration points, so called consistent tangent
operator.43 The elastic–plastic-damage consistent tangent
operator is defined as Substitution of the aforementioned formula into (45)
followed by the substitution of the resulting expression
dσ nþ1 dσ nþ1 into (41) and straightforward algebraic manipulations
Depd ≡ ¼ trial (40)
dεnþ1 dεnþ1 gives

where σ n + 1 is the outcome of the return mapping algo- dσ nþ1 ¼ ½E  μðE Pσ nþ1 Þ ⊗ðE Pσ nþ1 Þdεtrial
nþ1 (49)
rithm. We start by differentiating (29) that, together with
where the scalar μ is defined as
the elastic-damage law, gives

1
dσ nþ1 ¼ E dεtrial μ¼ (50)
nþ1  dΔγ Pσ nþ1 (41) σ Tnþ1 PE Pσ nþ1 þ 32HΔγ
2ξH

where
The explicit expression for the matrix form of the
E ¼ ½C ðd Þ þ ΔγP 1 (42) elastic–plastic-damage tangent operator consistent with
the CZM return mapping is then given by
Differentiation of the plastic consistency (35) and Depd ¼ E  μðE Pσ nþ1 Þ ⊗ðE Pσ nþ1 Þ (51)
qffiffiffiffiffi
using the identity, σ y ¼ 32 ξ , which holds under plastic
flow, yield the equation
  Damage evolution law in low-cycle fatigue
e ¼ 1 dξ  2 dσ y ξ dΔγ þ 1 Δγ dξ ¼ 0
dϕ (43)
2 3 dεp 2 In this section, the general constitutive law of the bilinear
CZM element in the low-cycle fatigue regime is pro-
posed as shown in Fig. 2. In this figure, plastic deforma-
The hardening modulus is defined as tion is large enough to produce degradation in the
material property of the cohesive zone. It is worth to note
dσ y KKT that in this modelling, loading and unloading must be
H¼ ¼ (44)
dεp K  K T accounted in each cycle of elastic–plastic-damage analysis
of the CZM element.
where KT is the penalty stiffness in plastic region. Then, Generally, it is defined that d = 0 at undamaged condi-
from the aforementioned equations, we have tion and d = 1 at final failure. Under general cyclic load-
  ing, the total damage, d, is the sum of the damage
3 2
dΔγ ¼ 1  HΔγ dξ (45) caused by monotonic loading and the damage resulted
4Hξ 3
from cyclic loading.

d ¼ d mon þ d cyc (52)


Using the elastic-damage law and the aforementioned
definition for E, the scalar ξ defined by (36) can be equiv- where dmon is the monotonic and dcyc is the cyclic contri-
alently written as bution of the damage parameter. For bilinear constitutive

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 9

Fig. 2 Low-cycle fatigue bilinear cohesive zone model element constitutive law.

equation, the monotonic damage parameter can be deter-


mined by the following equation:
d cyc ðN Þ
δd cyc   N   
∫   ¼ G d cyc ðN Þ ¼ ∫ h V α N ’ δN ’ (56)
0 g d cyc 0
f 
εm α  ε0m
d mon ¼   (53)
f
α εm  ε0m The integral in the right hand side of (56) is a linear
operator. Hence, we have a nonlinear damage evolution,
the nonlinearity being related to the G(dcyc), associated
where α is the maximum applied effective strain, εm, in all with a linear accumulation. Let us consider particular
previous iterations. In 1996, Lemaitre and Chaboche32 excitation of Vα(N) = Vα = constant, with explicit integra-
developed a common mathematical basis for the damage tion of (56) leads to
parameter in low-cycle fatigue in terms of a differential
equation of damage accumulation described as follows.   N
If the general nonlinear damage evolution law is consid- G d cyc ðN Þ ¼ ∫ hðV α Þ δN ’ ¼ hðV α Þ N (57)
0
ered,

If Nf is the cycle number at final failure, hence dcyc(Nf)


∂d cyc  
¼ f d cyc ; V α ðN Þ α (54) = 1. Then, we can rewrite (57) with assuming N = Nf as
∂N
¼ 1; 2; …; nα Gð1Þ
Nf ¼ (58)
hðV α Þ
where N is the cycle number and an independent variable
where G(1) = 1 can be considered without loss of general-
in the damage evolution. The causal variables Vα are re-
ity and hence
lated to the excitation that causes the damage phenome-
non. For instance, they may be the cyclic stress/strain  
amplitude or mean stress amongst other causes. The 1
V α ¼ h1 (59)
function f is generic nonlinear. By the assumption of Nf
decoupling between the cyclic damage and casual vari-
ables in (54), we have
Considering Eq. (54) with Eq. (59), the general ex-
pression yields to
∂d cyc  
¼ g d cyc hðV α ðN ÞÞ α (55)
∂N   
∂d cyc 1 1
¼ 1; 2; …; nα ¼ f d cyc ; h (60)
∂N Nf

As a result, the explicit integration of (55) is possible Then the simple form of the aforementioned equation
and leading to assuming no relation with damage evolution and cyclic

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
10 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

damage parameter can be derived. So, the simple nonlin-


ear relation for damage evolution is

 m
∂d cyc 1
¼ (61)
∂N Nf

where m is the degree of nonlinearity of damage evo-


lution law. In the analysis of the low-cycle fatigue,
when the plastic deformation effects become dominant
compared to the elastic deformation effects, a suitable
choice for causal variable is the plastic strain incre-
ment. Hence

V α ðN Þ ¼ Δεp ðN Þ (62)

The Coffin–Manson rule postulates the existence of a


power law that relates the number of cycles to the final
failure and the constant amplitude of the effective plastic Fig. 3 Typical nonlinear damage evolution.
strain applied in each cycle:44
At this point, we have d < 0.5:
c
N f ¼ N f ðΔε p Þ ¼ b½Δεp  (63)
N f =2 m
X 1
N =N f ¼0:5
d cyc ¼ < 0:5 (68)
where b and c are material parameters with the following i¼1
N f ði Þ
relations:
 c1
1 1 f Assuming a small increment, Nf (i) ≈ constant, and
c¼ ; b¼ (64)
cf bf Eq. (68) can be rewritten as

N =N f ¼0:5
where bf and cf derived from calibration tests of low-cycle N f ð1mÞ ¼ 2d cyc <1 (69)
fatigue. In Lemaitre and Chaboche model, the contribu-
tion of elastic part, that is, high cycle part of strain in the
Coffin–Manson rule is negligible with respect to the total By making use of Eq. (63), Eq. (69) gives
damage. Equation (63) corresponds to the definition of 2   3
the failure domain. Therefore, assuming a linear accumu- N =N f ¼0:5
log 2d cyc
lation rule, one has a damage evolution law as m¼14 5 (70)
log ðbÞ  clogðΔεp Þ
 m  m
∂d cyc 1 1 pc
¼ ¼ ½Δε  (65)
∂N N f ðΔεp Þ b With attention to the order of the material parameter
in (70) from a calibration of the low-cycle fatigue in
and damage accumulation after N cycle is Coffin–Manson rule and the amplitude of effective plas-
tic strain, the degree of nonlinearity of damage evolution
N  m N  m law is always 0 < m < 1. It must be determined in every
cycN
X 1 X 1
¼ ¼ ½Δεp ðiÞc (66) analysis with its particular loading condition and material
d
i¼1
N f ðΔεp ðiÞÞ i¼1
b selection to give the damage parameter, 0 < d < 1. For ex-
ample, we estimate m = 0.2 for high-density polyethylene
adhesive and m = 0.25 for epoxy adhesive.
To determine the range of the degree of nonlinearity Figure 4 illustrates the flowchart of progressive
for the damage evolution law, we may focus on a typical elastic–plastic-damage analyses. In these analyses, the
point of N/Nf = 0.5 in Fig. 3: yield stress, CZM strengths, fracture energies and pen-
alty stiffness in the elastic and plastic regions should be
N =N f ¼0:5
d ¼ d mon þ d cyc (67) updated.45 This figure also shows the location of the user

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 11

node brick element with small thickness (Fig. 1). The el-
ement thickness has been recommended to be about 0.01
of the layer thickness in elastic-damage analyses. But,
using of such small thickness values produces numerical
and convergence difficulties in the elastic–plastic-damage
analyses. Therefore, the CZM element thicknesses of 0.1
and 0.2 mm were used for elastic–plastic-damage analyses
in different cases. These values are almost equal to the
cohesive thickness between the layers. The elastic–
plastic-damage law for the description of material deg-
radation under low-cycle fatigue loading is investigated
with respect to its prediction for three fundamental
tests as follows:
1 Validation of the developed model with the analysis of
the damage growth in adhesively bonded aluminium
joints with different loading conditions.
2 Investigation of delamination growth and cycle num-
ber to final failure in FML under mode I and mode
II with a displacement ratio R ¼ δδmax
min
¼ 0.
3 The influence of the variable amplitude loading (VAL)
Fig. 4 Flowchart of progressive elastic–plastic-damage analysis. on damage growth rate in FML under mode-I and
mode-II fracture and the displacement ratio R =  1.

subroutines written in FORTRAN to perform these


analyses and linked to the master solver. Damage growth in adhesively bonded aluminium
plates
To validate the proposed model of elastic–plastic-damage
RESULTS AND DISCUSSION constitutive law for CZM element, the obtained results
from the developed procedure are compared with the
The presented CZM element and user material have available numerical results in Ref. [46]. In Ref. [46],
been linked to the ANSYS software by user programmable FCG along the CZM elements is studied. Instead of
features. The CZM element is a three-dimensional 20- using the Paris equation, the actual process of material

Table 1 Material properties for aluminiums bonded with adhesive46–49

Aluminium

E(GPa) v
70 0.33
Cohesive zone model element

GIc(N/mm) GIIc(N/mm) η bf cf m
0.422 1.279 2 0.06514562 0.116 0.2
K(GPa) KT (GPa) σ y(MPa) σ 0(MPa) τ 0(MPa)
0.3 0.1 3.3 6.66 6.66

E: Young’s modulus.
GIc: critical strain energy release rate in mode I.
η: material parameter obtained from experiments.
m: degree of nonlinearity of damage evolution law.
KT: cohesive zone model penalty stiffness in plastic part.
σ 0: normal cohesive zone model strength.
v: Poisson’s ratio.
GIIc: critical strain energy release rate in mode II.
bf, cf: parameters derived from calibration tests of low-cycle fatigue.
K: cohesive zone model penalty stiffness.
σ y: yield stress.
τ 0: shear cohesive zone model strength.

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
12 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

separation during FCG along the CZM elements is de- In the first test, a shear loading was applied, and in the
scribed by the use of an irreversible constitutive equation second test, the model response was examined in two level
for the cyclic traction-separation behaviour within the of loading sequences with different load directions and
CZM approach. This reference considered two- load levels. Besides, we developed gap element50 with the
dimensional constitutive law for stress–strain relation user material subroutine to model the initial delamination
and therefore used four-node cohesive elements for in order to prevent penetration of each layer to another
modelling. Also, an exponential constitutive law was one under compressive cyclic loading (Fig. 5).
considered in Ref. [46], but we used a three-dimensional To investigate the effect of mesh size on the results,
bilinear constitutive law that is compromising between the model has been analysed with two different mesh
the numerical cost and approximation. sizes. The coarse mesh contained 352 elements including
This model is representative of initially delaminated 60 CZM, 60 gap and 232 solid elements and the fine
adhesively bonded plates consisting of aluminium sub- mesh contained 532 elements including 120 CZM, 60
strates bonded by a modified high-density polyethylene gap and 352 solid elements were analysed. Figure 6 indi-
adhesive. The material data used in this investigation cates the damage accumulation in the middle layer of the
are presented in Table 1.46–49 initial delamination front (point A is located within the
The study investigated an adhesively bonded joint, solid adhesive and 0.1 mm far from the two adhesive-
with a thickness of 9.4 mm, a width of 25.4 mm, a length aluminium joints) due to the first loading sequence, that
of 216.0 mm and an initial delamination length of is, shear loading. Figure 6(b) also shows the damage con-
127.0 mm for normal and shear loading using the principles tour of the CZM elements and damage accumulation at
of normal superposition. When we consider the elastic- point A versus the number of cycles. This figure shows
damage model, the problem will be irreversible, and we a reasonable agreement between the predicted damage
can’t use the superposition to derive the accurate result results from this study and in Ref [46]. It should be noted
for the damage accumulation. Similarly in the elastic– that the resultant damage accumulation values in the first
plastic-damage analysis, we have the two irreversibility loading sequences are smaller than those predicted in
sources, and using the superposition for damage accu- Ref. [46]. This difference is due to the consideration of
mulation leads to inaccurate results. But to prevent the compressive stresses at the front of the delaminated area
problem complexity, we assumed that the irreversibility because of plastic strain accumulation. This compressive
of two loading block is not coupled, and the superposition stress reduces the rate of damage evolution in a low-cycle
of the damage accumulation is acceptable. The thickness fatigue regime.
of CZM elements is 0.2 mm. Computations in the example In the second loading sequence shown in Fig. 7, the
are performed under load-control condition. Both the model predicts a reduction in the fatigue strength with
normal and shear loadings followed a sinusoidal function. ongoing damage accumulation during the first loading

Fig. 5 Geometry and loading conditions for adhesively bonded joint: (a) shear loading and (b) normal loading.

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 13

Fig. 6 One level loading case study: (a) cyclic shear loading and (b) comparison of obtained damage evolution with those predicted by Roe and
Siegmund.46

block (15 cycles), such that damage will continue to accu- law. Also, it should be noted that the plastic strain accumu-
mulate during the second loading block (45 cycles). lation in the coarse meshed model is smaller than the
Specifically, the results presented in Fig. 7 demonstrate fine mesh; hence, the damage growth rate is smaller
that using the present model, the damage accumulation than fine mesh.
is coupled between the normal and shear direction of
loading in the elastic–plastic-damage cohesive zone.
Delamination growth in FMLs under mode I and
It should be noted that the resultant damage accu-
mode II
mulation values in the second loading sequences are
greater than those predicted in Ref. [46]. This differ- The composite layers in FMLs may consist of 2, 3 or 4
ence is due to the consideration of three-dimensional unidirectional or cross-ply layers. The alloy for the metal
elastic–plastic CZM constitutive law. The plastic strain layers can be selected from a wide range of different
accumulation increases the rate of damage evolution in metal alloys such as aluminium alloys. A similar freedom
a low-cycle fatigue regime. Furthermore, Figs 6 and 7 of choice is available for the composite materials. Of
show good agreement between the obtained results course, compatibility is important, and the hybrid lami-
with coarse and fine mesh models and available results nates should offer special features when compared with
in both loading sequence with nonlinear damage evolution individual metal alloys and full composites.

Fig. 7 Two level loading case study: (a) shear and normal loading amplitudes and sequence and (b) comparison of the obtained damage evo-
lution with those predicted by Roe and Siegmund.46

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
14 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

We consider double cantilever beam specimens The explained components have been analysed for the
(DCB) for mode I and end notch flexure specimens delamination growth rate under various loading condi-
(ENF) for mode-II condition. Computations in this part tions. For this purpose, firstly, the mode-I delamination
are performed under displacement-controlled conditions growth in the DCB specimens with different applied
with sinusoidal function and the ratio of R = 0. An impor- displacements is investigated. Figure 9(a) shows varia-
tant difference between the two material types is the tions of the damage parameter versus a number of load
occurrence of plastic deformation of the thin FML in cycles at the middle of the initial delamination front
mode-I and mode-II tests. As the plastic deformation (point A) for different applied displacements (from 0.75
consumes energy and cannot be distinguished from the to 1.05 mm) and the ratio of R = 0. Also, Fig. 9(b) shows
required energy for delamination, special considerations the normal stress–strain variations at point A for different
have to be employed to obtain reliable values for the frac- applied displacements. Figure 9(c) shows the contour of
ture toughness. Several modifications were made to the the damage parameters for applied displacement 0.75
original test specifications. The original test specimens and 1.05 mm. Figure 9(a) indicates that with increasing
were designed for composites instead of FML that show the magnitude of applied displacement, the number of
elastic deformations during the experimental procedure. cycles to create delamination decreases.
One modification is using additional layers of 7075 For each applied displacement, by increasing the
aluminium alloy with the thickness of 4.1 mm that pre- number of cycles the damage accumulation and there-
vents plasticity occurrence during both mode-I and fore, delamination growth increases (Fig. 9(a)) and the
mode-II experiments.51 stress amplitude in the stress–strain curve decreases
The final specimens used for the simulations (Fig. 9(b)). Figure 9(b) implies that increasing the applied
contained a set of Aluminium 7075-T6 doublers on both displacement causes more plastic strain accumulation.
sides with the thickness of 4.2 mm. In between, it was The contours of the damage parameters at the CZM
chosen a GLARE3 lay-up in which the aluminium layers elements show the delamination growth as depicted in
and composite laminate [0/90] had a thickness of 0.3 mm Fig. 9(c) after 5 cycles for applied displacement of
and 0.5 mm, respectively. The thickness of the CZM 1.05 mm and 120 cycles for applied displacement of
element was 0.1 mm, and resin adhesive layer between 0.75 mm. The damage parameter value of 1 at an integration
the aluminiums and doublers was 0.2 mm. The final spec- point of an element means delamination at that integration
imen contains the thickness of 10 mm, width of 25 mm, point. This figure also shows that at an applied displacement
and length of 150 mm. The initial delamination length of 0.75 mm, the domain of delaminated area (black area with
was 50 mm. Aluminium substrates and composite lami- damage factor of about unity) is smaller than that for the
nates bonded to each other with epoxy as an adhesive. applied displacement of 1.05 mm, but the domain of the
The geometry and loading condition of the specimen in damaged area is much greater than that for the applied
both modes are illustrated in Fig. 8(a) and (b). Totally, displacement of 1.05 mm, as it passed more cycles to reach
each model contains of 1048 elements where 80 elements the final failure.
of them are CZM elements. To evaluate the delamination In the second stage, the mode-II delamination growth
growth in this FML containing the initial delamination in the ENF specimens with different applied displace-
under mode I and mode II in low-cycle fatigue loading ments is investigated. Figure 10(a) shows variations of
using the developed approach, the presented material the damage parameter versus a number of load cycles at
properties in Table 2 were used.47–53 the middle of the initial delamination front (point A)

Fig. 8 Double cantilever beam and end notch flexure geometries and loading conditions for of GLARE3 specimens: (a) mode I and (b) mode
II.54

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 15

Table 2 Material properties for GLARE3 with aluminium doublers47–53

Aluminiums

E(GPa) v
70 0.33
Glass/epoxy

E11(GPa) E22 = E33(GPa) G12(GPa) G13(GPa) G23(GPa) v12 = v13 = v23


53.98 9.412 5.548 3 5.548 0.33
Cohesive zone model element

GIc(N/mm) GIIc(N/mm) η bf cf m
4 2 2 0.06514562 0.116 0.25
K(GPa) KT(GPa) σ y(MPa) σ 0(MPa) τ 0(MPa)
1.8 0.5 22 40 40
E: Young’s modulus.
E11: Young’s modulus in the longitudinal direction 1.
G12, G13, G23: shear moduli.
GIc: critical strain energy release rate in mode I.
η: material parameter obtained from experiments.
m: degree of nonlinearity of damage evolution law.
KT: cohesive zone model penalty stiffness in plastic part.
σ 0: normal cohesive zone model strength.
v: Poisson’s ratio.
E22, E33: Young’s modulus in the transverse direction 2, and through the thickness direction 3.
v12, v13, v23: Poisson’s ratios.
GIIc: critical strain energy release rate in mode II.
bf, cf: parameters derived from calibration tests of low-cycle fatigue.
K: cohesive zone model penalty stiffness.
σ y: yield stress.
τ 0: shear cohesive zone model strength.

Fig. 9 Double cantilever beam (DCB) results for various applied displacements and R = 0: (a) damage parameter versus number of cycles, (b)
normal stress–strain curves and (c) damage contours for displacements of 0.75 and 1.05 mm.

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
16 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

Fig. 10 End notch flexure results for various applied displacements and R = 0: (a) damage parameter versus number of cycles, (b) shear stress–
strain curves and (c) damage contours for displacement 0.3 and 0.55 mm.

Fig. 11 Comparison of log–log relation of plastic strain amplitude versus number of cycles to failure between end notch flexure (ENF) and
double cantilever beam (DCB) specimens for R = 0.

for different applied displacements (from 0.3 to 0.55 mm) are shown for the applied displacements of 0.3 and
and the ratio of R = 0. Also, Fig. 10(b) shows the shear 0.55 mm. For the applied displacement of 0.3 mm, the
stress–strain variations at point A for different applied domain of the delaminated and damaged area is much
displacements. Similar to the previous DCB tests, it is greater than that for the applied displacement of
found that with increasing the magnitude of the applied 0.55 mm. In both Figs 9 and 10, greater plastic strain in-
displacement, the required number of cycles to create crements are observed for larger applied displacements.
delamination at point A decreases. In Fig. 10(c), the Finally, Fig. 11 shows the plastic strain amplitude Δεp
contours of the damage parameters at the cohesive area against the number of cycles to failure, Nf, in the ENF

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 17

Fig. 12 Double cantilever beam (DCB) specimens with different displacement cycle sequences and R = 1: (a) damage parameter versus num-
ber of cycles, (b) normal stress–strain curves and (c) contours of damage for high to low and low to high blocks.

and DCB specimens with zero displacement ratios. The amplitude blocks and low to high amplitude blocks are
major resource for a comparison is the Coffin–Manson considered. The displacement amplitudes are 1 and
rule, which postulates a power law relation between the 0.8 mm for high and low loading blocks, respectively. Each
amplitude of the plastic strain increment and the number displacement sequence contains a high and a low ampli-
of cycles to failure, that is, a straight line in a log–log plot. tude loading block 5 cycles each. These blocks are firstly
This figure shows that for the same number of cycles to applied to the DCB specimen. Figure 12(a) shows the
failure, in a low-cycle regime, the plastic strain amplitude variations of damage accumulations versus cycle numbers
in ENF specimen is greater than that for DCB specimen. at point A due to different blocks. Figure 12(b) shows
According to the damage evolution relation (Eq. (65)), by the normal stress–strain curves at point A for different
increasing the plastic strain increment, the rate of cyclic blocks. Figure 12(a) and (b) indicates that in the high to
damage growth also increases. Hence, the cyclic damage low block case, accumulation of plastic strains is larger
evolution in mode II is greater than in mode I in a low- than that obtained for the low to high block case; hence,
cycle regime. damage accumulation and damage growth rates is greater
than the low to high block case. Damage parameter
contours at the cohesive area have been shown for both
Damage growth in FMLs under variable loadings
loading blocks in Fig. 12(c). These contours clearly show
In this section, the influence of VAL and influence of the that the damaged area in the high to low applied displace-
displacement ratio on the damage growth rate in FMLs ment sequence is larger than that predicted for the low to
under mode-I and mode-II conditions are studied. Similar high sequence.
to the previous section, DCB and ENF specimens are The described loading blocks are also applied to ENF
analysed under displacement-controlled condition with specimens with the displacement ratio of R =  1. The
sinusoidal function and the ratio of R =  1. For this displacement amplitudes are 0.5 and 0.3 mm for high
purpose, two displacement sequences of high to low and low loading blocks, respectively. Figure 13(a) shows

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
18 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

Fig. 13 End notch flexure (ENF) specimens with different displacement sequence and R = 1: (a) damage parameter versus number of cycles,
(b) shear stress–strain curves and (c) contours of damage for high to low and low to high blocks.

Fig. 14 Variable amplitude loading results for double cantilever beam (DCB) and end notch flexure (ENF) specimens with R = 1: (a) damage
parameter versus number of cycles and (b) stress–strain curves.

the variation of damage accumulation versus cycle num- of plastic strains is larger than that predicted for the low
bers at point A due to different loading blocks. Figure 13 to high loading block; hence, damage accumulation and
(b) shows the obtained shear stress–strain curves at point the damage growth rates are greater than those obtained
A for different loading blocks. Fig. 13(a) and (b) indicates for the low to high loading blocks. Similar to the ob-
that for the high to low amplitude blocks, accumulation tained results for zero displacement ratio, below

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
LOW-CYCLE FATIGUE DELAMINATION IN FIBRE METAL LAMINATES 19

10 cycles, the rate of cyclic damage growth in ENF spec- developed formulations and tool can be used for progres-
imens is greater than those obtained for DCB specimens. sive damage analyses of laminates under the low-cycle
Hence, for the high to low amplitude blocks, the damage fatigue loadings. Moreover, the delamination growths
growth rate is large enough to create delamination, that versus the load cycle number in FMLs under mode-I
is, damage parameter reaches to about unity (delamination and mode-II fracture conditions with the displacement
initiation) at about 8 cycles for ENF specimens and 10 - ratio of R = 0 were investigated. It was shown that in a
cycles for DCB specimens. Damage parameter contours low-cycle fatigue regime, the delamination growth rate
at the cohesive area for two loading blocks are shown in of FML in mode II is greater than mode-I condition. It
Fig. 13(c). These contours clearly show that delaminated was also shown that the damage growth rate and domain
and damaged area in the high to low applied displacement of the delaminated area in FMLs under VAL in mode-I
sequence is larger than the low to high sequence. and mode-II conditions with the displacement ratio of
Finally, for a general sense, the differences between R =  1 in the high to low applied displacement sequence
the predicted damage growth rates for VAL with is greater than the low to high sequence.
R =  1 under mode-I and mode-II conditions are
discussed. Figure 14(a) shows the damage accumulation
versus load cycle numbers. Dimensionless stress–strain
curve associated with this damage accumulation has been REFERENCES
also presented in Fig. 14(b). Obviously, the obtained
results for DCB and ENF specimens are not under 1 Fan, J. Y., Guan, Z. W., Cantwell, W. J. (2011) Structural
similar displacement amplitudes. behaviour of fibre metal laminates subjected to a low velocity
impact. Sci China - Phys Mech Astron, 54, 1168–1177.
It can be observed from Fig. 14 that the stress level in 2 Sinmazçelik, T., Avcu, E., Bora, M. Ö., Çoban, O. (2011) A
ENF specimen is lower than the DCB specimen, but the review: fibre metal laminates, background, bonding types and
damage growth rate in mode II is larger than mode I in applied test methods. Mater. Des., 32, 3671–3685.
both high to low and low to high loading blocks. The 3 Beumler, T., Pellenkoft, F., Tillich, A., Wohlers, W., Smart, C.
reason for this phenomenon is the larger plastic strain (2006) Airbus customer benefit from fiber metal laminates.
Airbus Deutschland GmbH, p. 1–18.
accumulation in mode II under variable loading that 4 Botelho, E. C., Silva, R. A., Pardini, L. C., Rezende, M. C.
causes unstable damage growth. Hence, it is predicted (2006) A review on the development and properties of continuous
that the final failure in mode II in the high to low loading fiber/epoxy/aluminum hybrid composites for aircraft structures.
sequence occurred in lower stress level and smaller Mater. Res, 9, 247–256.
applied displacement values. Two facts are also found 5 Yehia, A., Bahei-El-Din, A., Amany, G., Botrous, B. (2003)
Analysis of progressive fiber debonding in elastic laminates. Int
from Fig. 14(a). Firstly, as the plasticity is an irreversible
J Solids Struct, 40, 7035–53.
phenomenon, the predicted damages at the end of high 6 Vogelesang, L. B., Vlot, A. (2000) Development of fibre metal
to low and low to high applied displacement sequence laminates for advance. J. Mater. Process. Technol., 103, 1–5.
are quite different in both ENF and DCB tests. Secondly, 7 Alderliesten, R. (2009) On the development of hybrid material
the rate of damage growth of high loading block in high concepts for aircraft structures. Recent Patents Eng, 3, 25–38.
8 Vlot, A., Gunnink, J. W. (2001) Fiber Metal Laminates – An
to low applied displacement sequence is greater than the
Introduction. Kluwer Academic Publishers, Dordrecht, Netherland.
damage growth rate of high loading block in low to high 9 Khan, S. U., Alderliesten, R. C., Benedictus, R. (2008) Fatigue
applied displacement sequence. This fact can be due to crack growth prediction of fibre reinforced metal laminate under
the compressive stress resultant from plastic strain accu- variable amplitude loading. In: 26th International congress of the
mulation after low applied displacements in both ENF aeronautical sciences.
10 Shima, D. J., Alderliesten, R. C., Spearing, S. M., Burianek, D.
and DCB tests.
A. (2000) Fatigue crack growth prediction in GLARE hybrid
laminates. Compos Sci Technol, 63, 1759–1767.
11 Joadder, B., Shit, J., Acharyya, S., Dhar, S. (2011) Fatigue failure
CONCLUSION of notched specimen-Astrain life approach. Mater. Sci. Appl., 2,
1730–1740.
12 Yahya, M. M., Mallik, N. (2012) Low-cycle fatigue failure of
A three-dimensional CZM element with mixed-mode
composite materials/aluminium alloys at different heat
bilinear elastic–plastic-damage constitutive law has been treatments/process-A review. Int. J. Sci. Eng. Res., 3, 1–6.
presented in this paper for the prediction of delamination 13 Fuqiang, W., WeiXing, Y. (2010) A fatigue damage model of
growth of FMLs under low-cycle fatigue condition. To composite materials. Int. J. Fatigue, 32, 134–138.
verify the developed procedure, the predicted damage 14 ASTM Standard E606-8 (1969) Manual on low-cycle fatigue
testing. Constant amplitude low-cycle fatigue testing. in ASTM
parameter versus the load cycle number for adhesively
Specical Technical Publication 465.
bonded aluminium joints with initial delamination and 15 Manson, S. S., Halford, G. R. (1981) Practical implementation
different loading conditions was compared with the avail- of the double linear damage rule and damage curve approach
able results from literature. It was shown that the for treating cumulative fatigue damage. Int. J. Fracture, 17, 169–192.

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20
20 F. MAZAHERI AND H. HOSSEINI-TOUDESHKY

16 Kim, C. W., Oh, D. J. (2008) Progressive delamination with and 35 Wagner, W., Balzani, C. (2008) An interface element for the
without crack propagation in aramid fiber reinforced metal lam- simulation of delamination in unidirectional fiber-reinforced
inates containing a circular notch. Mater. Sci. Eng. A, 483-4, composite laminates. J Eng Fract Mech, 75, 2597–2615.
251–253. 36 Turon, A., Costa, J., Camanho, P., Davila, C. (2007) Simulation
17 Remmers, J. J. C., Borst, R. D. (2001) Delamination buckling of of delamination in composites under high-cycle fatigue. J Comp:
fibre–metal laminates. Compos Sci Technol, 61, 2207–2213. Part A, 38, 2270–2282.
18 de Vries, T. J., Vlot, A., Hashagen, F. (1999) Delamination 37 Ghovanlou, M. K. (2011) Failure analysis of brazed joints using
behavior of spliced fiber metal laminates part 1. Experimental the CZM approach. A thesis submitted for the degree of Doctor
result. Compos Struct, 46, 131–145. of Philosophy in Mechanical Engineering. University of
19 Guo, Y. J., Wu, X. R., Zhang, Z. L. (1997) Characterization of Waterloo.
delamination growth behaviour of hybrid bonded laminates. Fa- 38 Camanho, P., Davila, C., De Moura, M. F. (2003) Numerical
tigue Fract Eng Mater Struct, 20, 1699–708. simulation of mixed-mode progressive delamination in compos-
20 Lin, C. T., Kao, P. W. (1996) Fatigue delamination growth in ite materials. J Compos Mater, 37, 1415–1438.
carbon fibre-reinforced aluminium laminates. Composites Part 39 Ye, L. (1988) Role of matrix resin in delamination onset and
A, 27, 9–15. growth in composite laminates. J Compos Sci Technol, 33, 257–277.
21 Guo, Y. J., Wu, X. R. (1998) A theoretical model for predicting 40 Gözlüklü, B. (2009) Delamination analysis by using cohesive in-
fatigue crack growth rates in fiber-reinforced metal laminates. terface elements in laminated composites. a thesis submitted for
Fatigue Fract Eng Mater Struct, 21, 1133–1145. the degree of Master of Science. Middle East Technical Univer-
22 Naghipour, P. (2011) Numerical simulations and experimental sity, School of Natural and Applied Sciences.
investigations on quasi-static and cyclic mixed mode delamina- 41 Turon, A., Davila, C., Camanho, P., Costa, J. (2005) An engi-
tion of multidirectional CFRP laminates. a thesis submitted for neering solution for using coarse meshes in the simulation of de-
the degree of Doctor of Engineering Sciences. Institute of Air- lamination with cohesive zone models. Langley Research Center
craft Design, University of Stuttgart. Hampton, p. 23681-2199.
23 Dugdale, D. S. (1960) Yielding of steel sheets containing slits. 42 Halama, R., Poruba, Z. (2009) Tangent modulus in numerical
Nonlinear Dyn., 8, 100–104. integration of constitutive relations and its influence on conver-
24 Barenblatt, G. I. (1962). The mathematical theory of equilib- gence of N-R method. Appl. Comput. Mech., 3, 27–38.
43 Simo, J. C., Taylor, R. L. (1985) Consistent tangent operators
rium cracks in brittle fracture. Adv Appl Mech, 7, 55–129.
for rate-independent elastoplasticity. Comput. Meth. Appl. Mech.
25 Hillerborg, A., Modeer, M., Petersson, P. E. (1976) Analysis of
Eng., 48, 101–118.
crack formation and crack growth in concrete by means of frac-
44 Yang, Q. D., Shim, D. J., Spearing, S. M. (2004) A cohesive zone
ture mechanics and finite elements. Cem. Concr. Res., 6, 773–782.
model for low-cycle fatigue life prediction of solder joints.
26 Hosseini-Toudeshky, H., Farrokhabadi, A., Mohammadi, B.
Microelectron. Eng., 75, 85–95.
(2012) Consideration of concurrent transverse cracking and in-
45 Khoramishad, H., Crocombe, A. D., Katnam, K. B., Ashcroft, I.
duced delamination propagation using a generalized micro- A. (2010) A generalised damage model for constant amplitude
meso approach and experimental validation. Fatigue Fract Eng fatigue loading of adhesively bonded joints. Int. J. Adhes. Adhes.,
Mater Struct, 35, 885–901.
30, 513–521.
27 Hosseini-Toudeshky, H., Hosseini, S., Mohammadi, B. (2010) 46 Roe, K. L., Siegmund, T. (2003) An irreversible cohesive zone
Delamination buckling growth in laminated composites using model for interface fatigue crack growth simulation. Eng. Fract.
layer-wise interface element. J Compos Struct, 92, 1846–1856. Mech., 70, 209–232.
28 Hosseini-Toudeshky, H., Hosseini, S., Mohammadi, B. (2010) 47 Gunawardana, S. (2005) Prediction of failure initiation of adhe-
Buckling and delamination growth analysis of composite lami- sively bonded joints using mixed mode fracture data. a thesis
nates containing embedded delaminations. Appl Compos Mater, submitted for the degree of Master of Science. College of engi-
17, 95–109. neering and the faculty of the graduate school of Wichita State
29 Hosseini-Toudeshky, H., Jasemzadeh, A., Mohammadi, B. University.
(2011) Fatigue debonding analysis of repaired aluminium panels 48 Abou-Hamda, M., Mai, Y. W., Lowe, A. (1993) On the low en-
by composite patch using interface elements. Appl Compos Mater, durance fatigue of rubber-toughened adhesives and its implica-
18, 571–584. tion on characterization of damage. Int. J. Damage Mech., 2,
30 Hosseini-Toudeshky, H., Goodarzi, M. S., Mohammadi, B. 177–198.
(2013) Prediction of through the width delamination growth in 49 Ashby, M. F. (2011) Material Selection in Mechanical Design,
post-bukled laminates under fatigue loading using de-cohesive 4th edn. Elsevier, Oxford, UK.
law. Struct. Eng. Mech., 48, 41–56. 50 Lee, S. H. (1992) Handbook for Nonlinear Analysis. Version 67.
31 Benzeggagh, M., Kenane, M. (1996) Measurement of mixed- MaeNeal-Schwendler Corporation, Los Angeles, USA.
mode delamination fracture toughness of unidirectional 51 Sinke, J., de Boer, H., Middendorf, P. (2006) Testing and
glass/epoxy composites with mixed-mode bending apparatus. J modeling of failure behaiviour in fiber metal laminates. In:
Compos Sci Technol, 56, 439–449. 25th international congress of aeronautical sciences.
32 Erlicher, S. (2003) Hysteretic degrading models for the low- 52 Soltani, P., Keikhosravi, M., Oskouei, R. H., Soutis, C. (2010)
cycle fatigue behavior of structural elements: theory, numerical Studying the tensile behaviour of GLARE laminates: a finite el-
aspects and applications. a thesis submitted for the degree of ement modelling approach. Appl Compos Mater, 18, 271–282.
Doctor of philosophy. Department of Civil Engineering, 53 Linde, P., Pleitner, J., de Boer, H., Sinke, J. (2006) Numerical
Università degli Studi di Trento. and experimental simulation of damage behavior of fibre metal
33 Lemaitre, J., Chaboche, J. L. (1990) Mechanics of Solid Mate- laminates. In:25th international congress of aeronautical sciences.
rials. Cambridge University Press, Cambridge, UK. 54 Reinsborg, L. A. (2012) Fatigue properties of adhesive joints in
34 Souza Neto, E. A., Peri’c, D., Owen, D. R. J. (2008) Computa- patch repairs. a thesis submitted for the degree of Master of Sci-
tional Methods for Plasticity: Theory and Applications. John ence. The Norwagian University of Science and Technology,
Wiley and Sons, West Sussex, UK. Department of Engineering Design and Materials.

© 2014 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct 00, 1–20