Composite Structures 351 (2025) 118613

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Fatigue delamination damage analysis in composite materials through a rule

of mixtures approach
Alireza Taherzadeh-Fard a,b ,∗, Sergio Jiménez a,b , Alejandro Cornejo a,b , Eugenio Oñate a,b ,
Lucia Gratiela Barbu a,b
a
Polytechnic University of Catalonia (UPC), Campus Nord, 08034 Barcelona, Spain
b International Center for Numerical Methods in Engineering (CIMNE), Campus Nord UPC, 08034 Barcelona, Spain

ARTICLE INFO ABSTRACT

Keywords: The present study investigates delamination damage initiation and propagation within a homogenization theory
Rule of mixtures of mixtures, using the concept of virtual layers and virtual interfaces. It eliminates spatial discretization of
Homogenization layers, introducing a resultant damage variable to capture structure’s bulk response under both monotonic
Fatigue
and cyclic loads. Fatigue-induced deterioration is classified into sub-critical, critical, and over-critical stages
Delamination
based on interfacial stresses. Calibration is conducted employing the widely-available Wöhler curves for each
Composites
Finite element method
loading mode independently. An advance-in-time strategy is included in the model to enhance the simulation
speed. The reliability of the approach is assessed for crack initiation and propagation separately through
standard test coupons, showing good correlation with experimental data in mode I, mode II, and mixed-
mode loading conditions. Depending on the calibration procedure adopted, the model is applicable to a wide
range of stress ratios. In addition, it could be integrated into any standard finite element framework using
the desired number of elements through the thickness regardless of the physical amount of layers. This allows
easy modification of stacking sequences or the number of layers within the constitutive law without mesh
structure changes, facilitating simulation of large-scale composite laminates with minimal accuracy loss and
reduced computational costs.

1. Introduction its major effect on the strength and stiffness of the composite [12–
15]. Indeed, laminated composites are fabricated through building up
Over the past recent years, composites have attracted much atten- the in-plane fiber-reinforced laminae and are mainly expected to bear
tion as the potential materials to be utilized in different industries loading in the fiber direction. However, general loading conditions
including aerospace, automotive, marine, etc. [1,2], due to their su- usually impose some through-the-thickness stresses to the composite,
perior mechanical properties such as corrosion and fatigue resistance, and can lead to failure of the matrix and initiation of the delamination
light weight, high strength and ease of transportation [3–6]. The multi- phenomenon [16,17]. Debonding in the adjacent layers may not be
material nature of composite materials gives more freedom to engineers catastrophic within the in-plane loading, nonetheless, for composites
for designing purposes, however, this would lead to more damage under out-of-plane loading or bending conditions, it could be fatal [18,
mechanisms being engaged once these materials are under structural 19]. Prediction of this initiation and consequent propagation of the
loading as well as more uncertainty in their bulk response due to manu- delamination crack has been associated with experimental full-scale
facturing complexities [7–9]. Therefore, more advanced methodologies testing procedures owing to the lack of enough knowledge and accurate
are required in studying the behavior of composites. numerical models, which, in turn, has made the design procedures to
Damage mechanisms in composites are mainly comprised of intra- be highly expensive in terms of time and cost [20].
laminar and inter-laminar modes [10]. According to the complex nature Fatigue delamination damage is mainly composed of three distinct
of these damages and their interactions within the structure, simu- stages of initiation, onset and propagation, and each of them has been
lation schemes have encountered challenges to compromise between characterized by different numerical approaches [21]. The initiation
accuracy and computational cost [11]. Among all degradation modes, phase refers to the formation of the crack from an intact material
interface failure or delamination has been of significant importance for and commonly originates from the imperfections formed during the

∗ Corresponding author at: Polytechnic University of Catalonia (UPC), Campus Nord, 08034 Barcelona, Spain.
E-mail addresses: alireza.taherzadeh.fard@upc.edu, atfard@cimne.upc.edu (A. Taherzadeh-Fard).

https://doi.org/10.1016/j.compstruct.2024.118613
Received 11 June 2024; Received in revised form 8 August 2024; Accepted 23 September 2024
Available online 27 September 2024
0263-8223/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

manufacturing process [11]. A common method to account for this of fatigue scenarios while being capable of capturing both damage
stage is defining an initiation zone where damage could start appearing initiation and propagation over different amount of stress ratios.
according to the models based on S-N curves [22,23]. In this way, S- In the current study, a homogenization theory of mixtures has
N curves are formulated using experimentally-defined parameters and been presented to model fatigue delamination damage response in
each mode is calibrated independently. Delamination onset, however, laminated composites. The work is based on a previous research by the
is defined as a visible increment in the crack through the medium authors [51] where loading conditions were limited to the monotonic
from a crack starter. Although there is an unstable delamination crack regime. Rather than considering cohesive elements at each interface,
propagation after crack initiation, crack onset is followed by a stable the proposed model employs a homogenized medium where there is no
condition of crack extention [24]. Crack onset could be analyzed in need to explicitly discretize composite layers, which could significantly
a similar manner as initiation by considering fitting parameters to facilitate the pre-processing stage as well as the optimization of the
the experimental data [25], however this method lacks the ability to design process, since no re-meshing is required when the internal
consider the mixed-mode and stress ratios. configuration of the composite changes. This is even more pronounced
Linear elastic fracture mechanics (LEFM) is a widely-used approach when modeling composites with large number of layers by accounting
in simulating the fatigue crack in the propagation phase [26,27]. the number of elements needed through the thickness irrespective of
These models mainly employ any forms of the Paris law to make the actual number of layers. The cyclic loading effect is introduced to
a connection between the crack growth and the energy release rate the formulation by defining a fatigue reduction function which could
(ERR) [11]. Therefore, other methods should also be incorporated be calibrated according to the common S-N curves for mode I and
within this formulation to calculate the energy release rate or the stress mode II, independently. Delamination damage effect, in both contexts
intensity factor such as Virtual Crack Closure Technique (VCCT) [8,28– of crack initiation and propagation, is reflected to the adjacent virtual
30], virtual crack extension [31,32], or crack surface displacement bulk layers so that the bulk response of the structure is accurately
extrapolation methods [33]. Despite this, LEFM based models suffer reproduced, while not imposing any limitations on the constitutive law
from several problems such as the inability to be used in hybrid used for each layer. Therefore, in addition to the inter-layer damages,
there is the possibility of intera-layer mechanical phenomena such
composites [34] or failure to initiate the fatigue crack within a pristine
as damage and plasticity. Since the ad-hoc interface penalty stiffness
material [24].
concept is eliminated from the formulation, structural response will no
Cyclic Cohesive Zone Method (CCZM) can also be used to simulate
longer be manipulated in the elastic regime. The proposed model is
fatigue crack in composites which is based on defining a traction-
compatible with any finite element code and is considered to be well-
separation law for damage evolution within the interface [23,35–39].
suited for simulating fatigue delamination in large-scale composites
Based on how the cyclic load effects are taken into account, two
while maintaining the accuracy within an acceptable range.
different methods of hysteresis loop and envelope load damage models
have been considered [40]. Hysteresis loop damage models degrade the
2. Mathematical formulation
strength and the stiffness within a cycle-by-cycle algorithm and are well
suited for low-cycle fatigue regime [41]. Nevertheless, calibration of
2.1. Preliminary concepts
the parameters is considered to be a highly challenging task [40]. Enve-
lope load damage models, on the other side, utilize jumping strategies
In this section, some introductory information regarding the under-
combined with the Paris law to evolve delamination damage within the lying rule of mixtures method and interfacial stresses is provided. The
interface. This formulation could be beneficial in the high-cycle fatigue reader is referred to the previous work [51] for a complete description
regime and is capable of considering different factors such as mode mix and fundamental hypotheses.
and stress ratios. However, implementing Paris law within the cohesive Composite laminates are mainly composed of a specific number of
framework can be complex [42], and linking the damage variable to fiber reinforced laminae oriented at different directions. Trusdell and
the traction-separation formulations is not straightforward [20]. In ad- Toupin [52] firstly presented a method to unify all layers’ behavior
dition, it is not clear if the Paris law should be considered as a material through a homogenization theory. Later on, the method has been
property or rather a problem dependent variable. Consequently, fatigue developed by other researchers as well [53,54]. It is mainly based on
delamination damage models that are not based on the Paris law are the assumption that strains experienced by all laminae are the same
highly valuable. whilst the stress of the composite is obtained by taking a super-position
Moreover, cohesive zone methods have several shortcomings. of the stress in each laminae according to its volumetric participation,
Mainly since they are dependent on the definition of an interface i.e.:
medium, crack path should be known a priori [43]. Additionally, incor-
porating interface elements within all interfaces would be a challenging 𝜺1 = 𝜺2 = ⋯ = 𝜺𝑛 (1)
task especially in composites with large number of layers. Since CCZMs
∑
𝑛
are dependent on the definition of a traction-separation law with an 𝝈= (𝑘𝑖 ⋅ 𝝈 𝑖 ) (2)
initial linear part, they could manipulate the bulk response of the 𝑖=1
structure in its elastic region when there is not damage evolved within where 𝜺𝑖 , 𝑘𝑖 and 𝝈 𝑖 are the strain tensor, volumetric participation and
the interface [44]. Selection of the slope in CCZM linear part is also the stress tensor of the 𝑖th layer, respectively. From the homogenization
ad-hoc and requires a tuning process [45,46]. point of view, the composite is made up of virtual bulk layers and
Recently, a continuum mechanics based fatigue model has been virtual interfaces, as presented in Fig. 1, in a way that the response
proposed by unifying different phenomena such as damage, plasticity, of the structure is reproduced correctly in the macro scale.
viscosity and temperature effects [47], and has been improved in other The first step in formulating the delamination is to prepare a foun-
applications as well [48,49]. Indeed, the formulation triggers strength dation to calculate interfacial stresses. Since free edge effects are not
and stiffness deterioration once a specific stress criterion is satisfied. accounted for in the homogenized medium, a straightforward averaging
The degradation is accounted for by a damage-type constitutive law, scheme is employed to obtain stresses at the interfaces, as follows:
while being sensitive to the cyclic loads through a fatigue reduction
𝝈 𝑏𝑢𝑙𝑘
𝑖 + 𝝈 𝑏𝑢𝑙𝑘
factor [50]. The factor also affects the fracture energy so that an energy 𝝈 𝑖𝑛𝑡𝑒𝑟𝑓
𝑖
𝑎𝑐𝑒
= 𝑖−1
(3)
dissipation is introduced to the model as a result of the cyclic loading. 2
Model parameters are calibrated by using curve fitting to the common where 𝝈 𝑖𝑛𝑡𝑒𝑟𝑓
𝑖
𝑎𝑐𝑒
is the interfacial stress tensor at the 𝑖th interface while
S-N curves. The model is expected to be applicable in a wide spectrum 𝑏𝑢𝑙𝑘
𝝈𝑖 is the bulk layer stress tensor at the 𝑖th layer.

2
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 1. Virtual bulk layers and virtual interfaces through the homogenized medium [51].

According to Balzani et al. [55], delamination is driven by a normal

stress component acting perpendicular to the delamination plane and
two shear stress components acting on the delamination plane. Consid-
ering the coordinate system presented in Fig. 1, two equivalent stresses
are calculated to be fed into the delamination damage model for mode
I and mode II loadings:
⟨ ⟩
𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒
𝜎𝑛,𝑖 = 𝜎𝑧𝑧,𝑖 (4)
√
𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 2 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 2
𝜎𝑠,𝑖 = (𝜎𝑥𝑧,𝑖 ) + (𝜎𝑦𝑧,𝑖 ) (5)
𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒
where 𝜎𝑛,𝑖 and 𝜎𝑠,𝑖 are the interfacial equivalent normal and
shear stresses at the 𝑖th interface. ⟨∙⟩ are the Macaulay brackets to
prevent damage evolution in normal compression loads [51].

2.2. Fatigue delamination damage evolution

In order to accumulate the fatigue effect into the current delam-

ination formulation, a similar scheme to that described in [49] is
adopted here. To this end, the failure indicators 𝐹𝑛,𝑖 and 𝐹𝑠,𝑖 previously
developed at each interface 𝑖 in [51] are modified by a fatigue reduction
function (𝑓 𝑟𝑒𝑑 ) as below:
𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒
𝜎𝑛,𝑖 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒
𝐹𝑛,𝑖 = 𝑟𝑒𝑑 𝑐 𝑚𝑎𝑥 )
− 𝜎𝑛,𝑡ℎ,𝑖 >0 (6)
𝑓𝑛,𝑖 (𝑁𝑛,𝑖 , 𝑅𝑛,𝑖 , 𝑆𝑛,𝑖
𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒
𝜎𝑠,𝑖 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒
𝐹𝑠,𝑖 = 𝑟𝑒𝑑 (𝑁 𝑐 , 𝑅 , 𝑆 𝑚𝑎𝑥 )
− 𝜎𝑠,𝑡ℎ,𝑖 >0 (7)
𝑓𝑠,𝑖 𝑠,𝑖 𝑠,𝑖 𝑠,𝑖

where subscriptions n and s refer to normal mode I and shear mode

𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒
II loadings, respectively. 𝜎𝑛,𝑡ℎ,𝑖 and 𝜎𝑠,𝑡ℎ,𝑖 are historical variables
of the interface threshold stress in mode I and mode II, and their
initial value would consider to be normal and shear interface strengths,
respectively. These values will be updated at each step to the maximum
𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 Fig. 2. Different stages experienced by a band of elements in (a) Sub-critical (blue),
historical normal (𝜎𝑛,𝑖 ) and shear (𝜎𝑠,𝑖 ) uni-axial stresses (b) Critical (gray), and (c) Over-critical (red) regions as fatigue crack propagates in the
obtained previously. Once inequalities (6) and (7) are satisfied, the direction illustrated by arrows.
crack is permitted to grow.
𝑐 , 𝑅 , 𝑆 𝑚𝑎𝑥 ) is introduced to the failure indicators to take
𝑓 𝑟𝑒𝑑 (𝑁𝑚,𝑖 𝑚,𝑖 𝑚,𝑖
the cyclic load effects into account by amplifying the stress state. Its 𝑆 𝑚𝑖𝑛 ∕𝑆 𝑚𝑎𝑥 , matching the blue area in Fig. 2(a). An element located in
value ranges from 1, indicating no cyclic load effect, down to 0 at this area (black dashed strip) should not experience any cyclic load
the asymptote. According to the fatigue crack propagation through the effects, and hence, the fatigue related quantities will be maintained in
interface, which is shown in Fig. 2, three different stages could be their initial values, as below:
considered at each integration point, whilst each stage requires specific
𝑟𝑒𝑑
formulation for 𝑓 𝑟𝑒𝑑 , as described in the following. 𝑓𝑚,𝑖 =1 (8)

2.2.1. Sub-critical fatigue stage 𝑐

𝑁𝑚,𝑖 =0 (9)
Once fatigue crack initiates and starts propagating, there are some
where 𝑐 is the number of cycles at the 𝑖th interface and 𝑚 stands
areas in the far-field where the maximum cyclic stress at the interface 𝑁𝑚,𝑖
(𝑆 𝑚𝑎𝑥 ) is below the fatigue limit, 𝑆𝑡ℎ , at a specific stress ratio, 𝑅 = for 𝑛 or 𝑠 in normal and shear loadings, respectively. As the crack

3
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

propagates, the black dashed area becomes closer to the crack tip and Once S-N surface has been defined and formulated, an expression
the maximum cyclic stress levels will increase, which necessitates a could be proposed for fatigue reduction function in the form
different description for the 𝑓 𝑟𝑒𝑑 .
⎡ ( )
( )
⎤
⎢ ln 𝑆𝑚,𝑖
𝑚𝑎𝑥 ∕𝑆
𝑚,𝑢,𝑖 ( ) 𝛽𝑓 2 ⎥
) = exp ⎢ ⎥.
𝑟𝑒𝑑 𝑐 𝑚𝑎𝑥 𝑐 𝑚,𝑖
𝑓𝑚,𝑖 (𝑁𝑚,𝑖 , 𝑅𝑚,𝑖 , 𝑆𝑚,𝑖 ( ) log 𝑁𝑚,𝑖 (13)
2.2.2. Critical fatigue stage ⎢( ) 𝛽𝑓 2 ⎥
⎢ log 𝑁 𝑓 𝑚,𝑖 ⎥
When a specific element experiences stresses greater than the fa- ⎣ 𝑚,𝑖 ⎦
tigue limit and below the static strength, the cyclic load effects should 𝑚𝑎𝑥 is applied
According to Fig. 3, when a cyclic load at a specific 𝑆𝑚,𝑖
be considered through 𝑓 𝑟𝑒𝑑 . This situation is observed by the elements 𝑟𝑒𝑑
within the critical fatigue stage, a unique 𝑓𝑚,𝑖 function could be found
within the shaded area in the gray region in Fig. 2(b), where variation 𝑓 𝑓
which intersects the normalized S-N curve at 𝑁𝑚,𝑖 . 𝑁𝑚,𝑖 is a critical
of 𝑓 𝑟𝑒𝑑 could trigger two kinds of non-linearity. The first one is the
quantity which will be used in the next sessions for the jumping
gradual amplifying of the interfacial stresses (or literally, the gradual 𝑐 = 𝑁 𝑓 in Eq. (13),
strategy. Its value could be obtained by setting 𝑁𝑚,𝑖
decrease of the strength) as a result of the 𝑓 𝑟𝑒𝑑 evolution when 𝑁 𝑐 is 𝑚,𝑖
then equating the outcome to the result from Eq. (10):
growing. The second one, which is a consequence of the first one, is
the rapid propagation of the crack once 𝑁 𝑐 reaches the critical number
⎡ 1 ⎤
of cycles (𝑁 𝑓 ) and inequalities (6) and (7) are met. ⎢⎡ ⎛ 𝑆 𝑚𝑎𝑥 − 𝑆 𝑡ℎ (𝑅 ) ⎞⎤ 𝑓 ⎥
⎢⎢ 1 ⎜ 𝑚,𝑖 𝑚,𝑖 𝛽
⎟⎥ 𝑚,𝑖 ⎥
The definition of the 𝑓 𝑟𝑒𝑑 function can be done based on either G- ⎢⎢−
𝑚,𝑖
⎥
⎢⎢ 𝛼 𝑡 (𝑅 ) ⎜⎜ 𝑆 𝑢 − 𝑆 𝑡ℎ (𝑅 ) ⎟⎟⎥⎥
⋅ln
⎥
N (energy release rate vs. number of cycles) or S-N (Interfacial stress ⎢⎣ 𝑚,𝑖 𝑚,𝑖 ⎝ 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 ⎠⎦ ⎥
𝑓 ⎢ ⎥
vs. number of cycles) diagrams. Indeed, the model in this study has 𝑁𝑚,𝑖 𝑚𝑎𝑥
(𝑅𝑚,𝑖 , 𝑆𝑚,𝑖 )= 10⎣ ⎦ (14)
originated from the high-cycle fatigue (HCF) formulation presented by
Generally, as long as 𝑁𝑚,𝑖𝑐 < 𝑁 𝑓 , the non-linearities would be strength
Oller et al. [47], which has been developed for metals based on S- 𝑚,𝑖
deterioration and fracture energy dissipation, according to the 𝑓𝑚,𝑖 𝑟𝑒𝑑
N curves. Successful implementation and satisfactory results of this
evolution. Once 𝑁𝑚,𝑖 𝑐 ≥ 𝑁 𝑓 , in addition to the strength, the stiffness
formulation in HCF regime [49,56] triggered the idea of applying 𝑚,𝑖
the same approach in composite materials [57]. Therefore, although will be degrading as well, which corresponds to a rapid failure process
calibrating the model based on G-N curves seems to be more applicable in the integration point level according to inequalities (6) and (7).
as for more availability of these curves for composite materials, present Matlab software [60] is employed in the fitting process and to
formulation has been established based on the stress curves. Supporting find proper values for the engaged parameters. To this end, the more
this approach, there are recently-published papers where S-N curves experimental data is provided, the wider the reliable region will be in
have been employed, and a calibration process has been conducted terms of stress ratio, 𝑅, and maximum stress, 𝑆 𝑚𝑎𝑥 . In other words, the
accordingly to model the delamination propagation in laminated com- model should be well-prepared prior to simulations to be able to treat
posites [58,59]. This may indicate the potential of the S-N curves in this as close as possible to the provided experimental Wöhler surface. In this
context. Moreover, according to [42], some inter-connections could be way, the formulation could be applicable in different loading scenarios.
established between the common Paris law (da/dN vs. G) and the S-N
curve in composites. This would be an asset for the current approach in 2.2.3. Over-critical fatigue stage
this study, since there is the potential to feed the Paris law to the model Over-critical fatigue stage refers to the region where predictive
and make calibrations accordingly. This issue has not been addressed stresses are greater than the static strength of the material. This sit-
in current study, and would be a matter of future investigations. uation happens as the crack propagates through the interface and the
Definition of 𝑓 𝑟𝑒𝑑 for the interface is conducted through a fitting crack opening increases cycle-by-cycle, as is the case for the elements
procedure with respect to the common S-N surfaces of the type [49]: located within the red region in Fig. 2(c). The irreversible degradation
( ) 𝑟𝑒𝑑 quantity. How-
process of the interface has to be reflected in the 𝑓𝑚,𝑖
𝑐 𝑡ℎ 𝑢 𝑡ℎ 𝑡
𝑆𝑚,𝑖 (𝑅𝑚,𝑖 , 𝑁𝑚,𝑖 ) =𝑆𝑚,𝑖 (𝑅𝑚,𝑖 ) + 𝑆𝑚,𝑖 − 𝑆𝑚,𝑖 (𝑅𝑚,𝑖 ) ⋅ exp{−𝛼𝑚,𝑖 (𝑅𝑚,𝑖 ) ever, a different formulation than the one for the critical region should
( )𝛽 𝑓 be employed. Consequently, a similar approach to [61] is adopted by
𝑐 𝑚,𝑖
⋅ log 𝑁𝑚,𝑖 } (10) considering a fatigue-induced degradation as:
𝑢 being the stress ratio and static strength for mode 𝑚 at
𝑅𝑚,𝑖 and 𝑆𝑚,𝑖 𝑟𝑒𝑑
d𝑓𝑚,𝑖 ( )−𝜂𝑚,𝑖
1 𝑐
𝑡ℎ (𝑅 ) is the fatigue limit at the 𝑖th =− 𝑁𝑚,𝑖 (15)
the 𝑖th interface, respectively. 𝑆𝑚,𝑖 𝑚,𝑖
𝑐
d𝑁𝑚,𝑖 𝛾𝑚,𝑖
interface:
where 𝛾𝑚,𝑖 and 𝜂𝑚,𝑖 are interface parameters describing the rate of degra-
⎧ ( ) ( )𝑆𝑚,𝑖
𝑅1
dation, and their value could be obtained according to the experimental
⎪ if |𝑅| ≤ 1 ⇒ 𝑆 𝑡ℎ (𝑅 ) = 𝑆 𝑒 + 𝑆 𝑢 − 𝑆 𝑒 ⋅ 1 + 𝑅𝑚,𝑖
⎪ 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 2 campaign.
⎨ (11)
( ) ( 1 + 𝑅 )𝑆𝑚,𝑖𝑅2
⎪ 𝑚,𝑖
⎪ if |𝑅| > 1 ⇒ 𝑆 𝑡ℎ (𝑅 𝑚,𝑖 ) = 𝑆 𝑒 + 𝑆 𝑢 − 𝑆 𝑒 ⋅
𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 2𝑅𝑚,𝑖 2.3. Fatigue delamination damage effect
⎩
𝑒 is the fatigue limit at 𝑅
where 𝑆𝑚,𝑖 𝑡
𝑚,𝑖 = −1. 𝛼𝑚,𝑖 (𝑅𝑚,𝑖 ) is a parameter Once the failure criteria in Eqs. (6) and/or (7) are satisfied – either
which depends on the state of loading as described below: by evolution of 𝑓𝑚,𝑖𝑟𝑒𝑑 or increasing the 𝜎 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 – the damage variable
𝑚,𝑖
( )
⎧ if |𝑅| ≤ 1 ⇒ 𝛼 𝑡 (𝑅 ) = 𝛼 𝑓 + 1 + 𝑅𝑚,𝑖 ⋅ 𝐴𝑈 𝑋 𝑅1 for each mode 𝑚 and interface 𝑖 is calculated independently according
⎪ 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 2 𝑚,𝑖 to an exponential softening law [51,62]:
⎨ ( ) (12)
1 + 𝑅𝑚,𝑖 [ 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 ]
⎪ if |𝑅| > 1 ⇒ 𝛼𝑚,𝑖 (𝑅𝑚,𝑖 ) = 𝛼𝑚,𝑖 −
𝑡 𝑓
⋅ 𝐴𝑈 𝑋 𝑅2 0,𝑡ℎ
𝜎𝑚,𝑖 ( 𝜎𝑚,𝑖 )
⎩ 2𝑅𝑚,𝑖 𝑚,𝑖
𝑑𝑚,𝑖 = 1 − 𝑖𝑛𝑡𝑒𝑟𝑓 𝑎𝑐𝑒 exp 𝐴𝑚,𝑖 1 − (16)
0,𝑡ℎ
𝜎𝑚,𝑖 𝜎𝑚,𝑖
𝑅1 , 𝑆 𝑅2 , 𝛼 𝑓 , 𝛽 𝑓 , 𝐴𝑈 𝑋 𝑅1 and 𝐴𝑈 𝑋 𝑅2 are material properties and
𝑆𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖 𝑚,𝑖
could be calibrated at each interface 𝑖 for each mode of loading 𝑚. The where
more Wöhler diagrams in different stress ratios are considered in the 1
𝐴𝑚,𝑖 = (17)
calibration, the more accurate description of the S-N surface will be 𝐶𝑚,𝑖 𝐺𝑚,𝑖 1
− 2
obtained in a wide spectrum of stress ratios. 0,𝑡ℎ 2
(𝜎𝑚,𝑖 ) 𝑙𝑐

4
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 3. Normalized Wöhler and fatigue reduction function curves in a general loading state.

0,𝑡ℎ
𝜎𝑚,𝑖 , 𝐶𝑚,𝑖 , 𝐺𝑚,𝑖 , and 𝑙𝑐 being either mode I or mode II interfacial within the simulation. This could be problematic, specifically in high-
strength, modulus, fracture toughness and characteristic length, re- cycle fatigue situations due to the high computational cost associated.
spectively. As explained in previous work [51], transverse Young’s Therefore, a procedure should be considered to skip a number of cycles
and transverse shear moduli of the bulk layers are considered as the while compensating for its effects within the formulation to reduce the
interfacial normal and shear moduli, whilst half of the element length computational costs.
is regarded as the characteristic length. To this end, an advance-in-time (AIT) strategy is considered once
Since there is not dedicated medium for the interface layer, de- stable conditions are obtained in terms of the cyclic loading exerted.
lamination damage effects should be properly reflected in the virtual The method is based on the jump strategy proposed in [49,50,57,63],
bulk layers so that the behavior of the structure at macro-scale is well- however it is enhanced to account for two different loading mode con-
captured. To this end, each bulk layer is affected by two maximum ditions. Indeed, previous version of the AIT method has been presented
normal (𝑑𝑛 ) and shear (𝑑𝑛 ) damages as below [51]: for isotropic materials where the distinction between mode I or mode
𝑏𝑢𝑙𝑘, 𝑑 𝑏𝑢𝑙𝑘
II loadings was not necessary. In the case of the numerical framework
𝜎𝑧𝑧,𝑖 = (1 − max[𝑑𝑛,𝑖 , 𝑑𝑛,𝑖−1 ])𝜎𝑧𝑧,𝑖 (18) proposed in this study, that is no longer the case, a method is developed
to identify to most critical mode of loading, and the jump is set up
𝑏𝑢𝑙𝑘, 𝑑 𝑏𝑢𝑙𝑘
𝜎𝑥𝑧,𝑖 = (1 − max[𝑑𝑠,𝑖 , 𝑑𝑠,𝑖−1 ])(1 − max[𝑑𝑛,𝑖 , 𝑑𝑛,𝑖−1 ])𝜎𝑥𝑧,𝑖 (19) accordingly.
Let us consider that mode I and mode II loadings are activated in
𝑏𝑢𝑙𝑘, 𝑑 𝑏𝑢𝑙𝑘 the critical fatigue region, as shown schematically in Fig. 5. To maintain
𝜎𝑦𝑧,𝑖 = (1 − max[𝑑𝑠,𝑖 , 𝑑𝑠,𝑖−1 ])(1 − max[𝑑𝑛,𝑖 , 𝑑𝑛,𝑖−1 ])𝜎𝑦𝑧,𝑖 (20)
generality, loading modes are referred to as modes 𝑘 and 𝑙. Depending
where 𝑏𝑢𝑙𝑘 , 𝜎 𝑏𝑢𝑙𝑘 , 𝜎 𝑏𝑢𝑙𝑘 )
(𝜎𝑧𝑧,𝑖 are predictive stresses acting on the delamina-
𝑥𝑧,𝑖 𝑦𝑧,𝑖 on the time when either mode has reached the critical fatigue region,
𝑏𝑢𝑙𝑘, 𝑑
tion plane. They would be considered as the integrated stresses (𝜎𝑧𝑧,𝑖 , the number of cycles before jump (𝑁𝑖𝑐 ) could be different, since in the
𝑏𝑢𝑙𝑘, 𝑑 𝑏𝑢𝑙𝑘, 𝑑
𝜎𝑥𝑧,𝑖 and 𝜎𝑦𝑧,𝑖 ) once affected by the delamination damage. All sub-critical fatigue stage cycles are not counted according to Eq. (9).
other stress components remain unaffected in terms of the delamination On the other hand, the required number of cycles to reach the damage
phenomenon [55]. initiation conditions could be obtained using Eq. (14) for each loading
To illustrate the performance of the presented formulation in dif- mode, independently. It would be reasonable to employ the most
ferent scenarios, two types of loading conditions are schematically critical loading mode in terms of the distance to the damage initiation
presented in Fig. 4. In Fig. 4(a), a cyclic load is applied within the conditions, so that the other mode remains in the allowable S-N area
critical fatigue stage followed by a monotonic loading part. The model after the jump. As a result, the loading mode with the minimum amount
𝑟𝑒𝑑
is expected to activate the first non-linearity as the result of the 𝑓𝑚,𝑖 of 𝑁 𝑓 − 𝑁 𝑐 will be fed into the AIT strategy.
evolution and decrease the interface strength accordingly. However, in Other combinations of loading modes are still possible, where the
Fig. 4(b), a monotonic load is applied followed by a cyclic load within selection of the mode for the AIT strategy is quite straightforward.
the over-critical fatigue region. As it is obvious, the response follows Within all other possibilities, loading mode in the over-critical fatigue
the same path as in the monotonic loading regime in the first cycle, stage would be selected for AIT strategy compared to the one in the
whilst a progressive degradation in both strength and stiffness happens critical fatigue stage, while critical fatigue stage will be prioritized over
once the load is switched to a cyclic one. the sub-critical fatigue region.
The jump is expected to occur once stable conditions are reached
2.4. Advance-in-time strategy in the stress state. While the crack is propagating, the maximum cyclic
stress is continuously evolving at each integration point. In this way,
According to Eqs. (13) and (15), the fatigue delamination method stability condition means a plateau in the damage and stress evolution
developed in the present study employs three independent quantities which could be monitored employing two indicators:
𝑐 , 𝑅
of 𝑁𝑚,𝑖 𝑚𝑎𝑥 𝑐
𝑚,𝑖 and 𝑆𝑚,𝑖 in the critical regime, and one quantity of 𝑁𝑚,𝑖 | 𝑆𝑚,𝑖,𝑗+1
𝑚𝑎𝑥 𝑚𝑎𝑥 |
− 𝑆𝑚,𝑖,𝑗
| |
in the over-critical region to identify the state of the sample within the 𝜑𝑚,𝑖 = | | < 𝑡𝑜𝑙𝑒𝑟𝑎𝑛𝑐𝑒 (21)
| 𝑚𝑎𝑥
𝑆𝑚,𝑖,𝑗+1 |
cyclic loading. Although 𝑅𝑚,𝑖 and 𝑆𝑚,𝑖 𝑚𝑎𝑥 are computed at each step and | |
are completely dependent on the exerted external loads, 𝑁𝑚,𝑖 𝑐 is a con- | 𝑅𝑚,𝑖,𝑗+1 − 𝑅𝑚,𝑖,𝑗 |
| |
𝜓𝑚,𝑖 = | | < 𝑡𝑜𝑙𝑒𝑟𝑎𝑛𝑐𝑒 (22)
tinuous variable in the time domain and requires progressive growth | 𝑅𝑚,𝑖,𝑗+1 |
| |

5
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 4. Model performance in two different loading scenarios: (a) cyclic load followed by a monotonic tension within the critical area and (b) monotonic tension followed by a
cyclic loading within the over-critical region.

Fig. 5. Criticality level of two different loading modes in terms of the number of remaining cycles to failure.

𝜑𝑚,𝑖 and 𝜓𝑚,𝑖 being maximum stress and stress ratio stabilization norms, in the absence of the fiber bridging effects, the results still fall within
𝑚𝑎𝑥 and 𝑅
while 𝑆𝑚,𝑖,𝑗 𝑚,𝑖,𝑗 being the maximum stress and the stress ratio an acceptable range of the experimental data.
of mode 𝑚, interface 𝑖 and time step 𝑗, respectively. Once stability The presented fatigue formulation is implemented through the
conditions in Eqs. (21) and (22) are satisfied, the process time as well open-source code Kratos Multi-physics [64,65]. Geometry and mesh
as the number of cycles at each integration point will be updated, and generation has been conducted utilizing pre and post processing soft-
𝑟𝑒𝑑 is calculated based on the newly-obtained number of cycles.
the 𝑓𝑚,𝑖 ware GiD [66].

3. Numerical examples and results

3.1. Constitutive law parameters calibration

The potential of the proposed model is studied in this section

As it has been mentioned in Section 2.2.2, a Matlab fitting algorithm
through different loading scenarios in both damage initiation and dam-
age propagation regimes. Unidirectional carbon fiber/epoxy prepreg is employed to calibrate the fatigue parameters of the model. To
(IM7/8552) properties have been employed, and the validations are this end, Wöhler diagrams of the interface in each loading mode is
conducted on this composite in all fatigue crack evolution assessments. required to effectively characterize the fatigue delamination behavior.
It should be noted that, in the current study, the effect of fiber Transverse tension [18,67] and shear [22,68] fatigue tests have been
bridging has not been considered in the simulations. Indeed, the pre- conducted on the IM7/8552 carbon/epoxy composite at different stress
𝑚𝑎𝑥 ) until the crack is initiated within
ratios (𝑅𝑚,𝑖 ) and stress levels (𝑆𝑚,𝑖
sented fatigue model acts in an integration-point-wise approach, where
all the calculations are conducted at each Gauss-point within the finite the interface. The resulted experimental S-N diagrams are utilized to
element model. This implies that the basis of the fatigue model in fit a curve in the form of Eq. (10) for each mode I and mode II. The
current assessment is a local framework. However, taking into account experimental Wöhler curves along with the calibrated model responses
the bridging effect necessitates additional information regarding the are presented in Fig. 6.
extent of delamination damage crack, since the R-curve phenomenon Tables 1 and 2 present the material properties as well as the
depends upon the length of the crack propagated. To this end, non-local obtained values for the calibrated interfacial parameters, respectively.
methods should be taken into account to efficiently consider the fiber These properties will be used within the simulations in the following
bridging effects. Ignorance of the R-curve effect would not impose any sections for both fatigue damage initiation and propagation regimes.
problems for the damage initiation stage, however, damage propaga- Since all the simulations are conducted at 𝑅𝑚,𝑖 < 1, there is no need to
tion would be affected by this assumption. It will be demonstrated that 𝑅2 and 𝐴𝑈 𝑋 𝑅2 parameters.
calibrate 𝑆𝑚,𝑖 𝑚,𝑖

6
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 6. Fatigue material model calibration in (a) transverse normal (mode I) and (b) shear (mode II) configurations compared to the experimental data [18,68].

Table 1 Table 2
Mechanical properties of the IM7/8552 carbon/epoxy prepreg [22]. Calibrated interfacial fatigue parameters for modes I and II [68].
Mechanical property Value Interface property Normal load Shear load
(Mode I) (Mode II)
𝐸1 [GPa] 161
𝐸2 = 𝐸3 [GPa] 11.4 Energy release rate (𝐺) [J/m2 ] 212 600
0
𝐺12 = 𝐺13 [GPa] 5.17 Interfacial strength (𝜎𝑡ℎ ) [MPa] 111 85
𝐺23 [GPa] 3.98 Interfacial modulus (𝐶) [GPa] 11.4 5.17
𝜈12 = 𝜈13 0.32 𝑆 𝑒 [MPa] 87.7 39.1
𝜈23 0.43 𝑆 𝑅1 [MPa] 48.88 55.81
𝛼𝑓 2.14e−3 1.69e−4
𝛽𝑓 4.49 6.71
𝐴𝑈 𝑋 𝑅1 −2.80e−3 −2.58e−4
3.2. Fatigue delamination damage initiation assessment 𝛾 25 30
𝜂 0.72 0.72

In order to investigate the model capability to initiate fatigue crack

in an undamaged medium, two standard tests have been employed in
each loading mode, as illustrated in Fig. 7. Three-point bending test is in 90◦ angle with respect to the longitudinal direction. Hexahedron
utilized to initiate the fatigue crack in mode I loading. To this end, a elements with the size of 0.1 mm are employed in the central region,
central displacement is cycled with a frequency of 1 Hz while the two where the crack is expected to initiate, and coarse meshes are opted for
ends of the cantilever beam are set to be fixed through the whole width the far-field area. The cycle is defined to induce 0.1 stress ratio states
(𝑤 = 6.35 mm), as shown in Fig. 7(a). according to the experimental procedure, and the results are presented
Cyclic displacement domains are chosen in a way that they produce in Fig. 8 compared against the experimental data. Within the experi-
maximum normal stresses of 90, 93, 97, 101, and 105.5 MPa within the mental tests, specimens in both standard tests fail catastrophically once
central region. Carbon fibers are aligned in the loading direction and damage has initiated, which allows simple detection of the damage

7
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 7. Schematic view of (a) three-point bending and (b) double-notched shear tests under cyclic load (dimensions in millimeters).

Fig. 8. Initiation of the fatigue crack in the three-point bending configuration under mode I loading compared with the experimental data at 𝑅 = 0.1 [18,23].

onset. In the numerical framework, however, the necessary number of pure mode II simulations [68]. The configuration of the test is shown in
cycles to initiate the delamination damage is calculated according to Fig. 7(b) where carbon fibers are aligned in the longitudinal direction.
Eq. (14). This value has been considered as the crack initiation point A cyclic axial displacement is applied to one end through the whole
within the simulations. width (𝑤 = 12.7 mm) so that it produces shear stresses at the central
As it can be observed, the model captures the crack initiation part. From the numerical point of view, considering solely the central
cycle fairly well at each stress level, and the numerical output lies part would be enough to run the simulations. Two stress ratios of 0.1
within the experimental scatter region. In the simulations, fatigue crack and 0.3 are used along with a cyclic load of 1 Hz. The element size and
propagates catastrophically once it has been initiated, which aligns
type are the same as in the three-point bending test. Fig. 10 summarizes
with the experimental findings [18,23]. In Fig. 9, the crack initiation
the numerical and experimental results.
step and its complete propagation is presented at the stress level of
97 MPa. Through the simulation, a deactivation algorithm is employed The model output for mode II loading shows a good agreement with
to erase fully damaged elements so that the model converges more the experimental results. This indicates that the proposed formulation
easily. would be able to deal with the crack initiation phenomenon in both
Validating the performance of the model in mode I loading, it would mode I and mode II loading conditions in an accurate manner. The
be valuable to check its applicability in shear mode as well. Therefore, next stage is assessing the performance of the model in the propagation
experimental double-notched shear test results are utilized for running regime, which will be discussed in the following section.

8
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 9. Three-point bending fatigue simulations in the stress level of 97 MPa in (a) initiation stage at 𝑁 𝑐 = 218 964 and (b) complete failure at 𝑁 𝑐 = 218 986.

Fig. 10. Initiation of the fatigue crack in double-notch shear configuration in (a) 𝑅 = 0.1 and (b) 𝑅 = 0.3 compared with the experimental data [68].

9
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 11. Double Cantilever Beam (DCB) configuration within the crack propagation regime in pure mode I loading.

Table 3 could also be calculated as a function of the reaction force (𝑃 ) as well

Dimensions of the DCB, ENF and MMB specimens. as the crack length as:
Test configuration a [mm] w [mm] L [mm] 2h [mm] C [mm]
𝑃 2 (𝑎 + 𝜒ℎ)2
DCB 50.8 25.4 110.8 4.5 – 𝐺𝐼 = . (27)
ENF 25.4 25.1 101.6 4.5 – 𝑤𝐸𝐼
MMB (50%) 28.8 25.4 101.6 4.5 41.3 According to the experimental procedure, the simulations are con-
ducted in a displacement control mode where the maximum amounts
of the displacement applied are 𝛿 = 1.48 mm, 1.7 mm, 1.92 mm, 𝑎𝑛𝑑
2.25 mm, and the stress ratio is 𝑅 = 0.1. Crack propagation rate
3.3. Fatigue delamination damage propagation assessment
(da∕dN) is calculated utilizing the finite difference method accord-
ing to Eq. (23). Fatigue crack extension rates versus the correspond-
In order to investigate the crack propagation through the homoge-
ing energy release rates are compared to the experimental values as
nization theory proposed herein, three standard test methods, i.e., Dou- illustrated in Fig. 12.
ble Cantilever Beam (DCB), End Notch Flexure (ENF), and Mixed Mode Arrows in Fig. 12 indicate the direction of the diagram as crack is
Bending (MMB) are employed. The same material properties as in propagated. Initially, fatigue crack propagates at high rates, which is
Section 3.2 are utilized, and an orthotropic material model has been attributed to the formation of a process zone as the result of a mono-
used for describing the mechanical behavior of laminae within the tonic damage within the first cycle. The crack then reaches a stable
classical mixing theory. condition in propagation which considers as the Paris law region. To
DCB test is simulated according to the extensive experimental pro- reach a unified Paris law description among all simulations conducted,
cedure presented in [69]. The configuration is illustrated in Fig. 11 the point where 5 percent increase in compliance is detected would be
where two cyclic loads with the same characteristics are exerted on considered in each load level, and a Paris law equation of the form
both flanges of the sample in opposite directions. In this way, the pre-
da
cracked area is expected to grow in a pure tension mode. Table 3 = 𝑚(𝐺𝑚𝑎𝑥 )𝑛 (28)
dN
summarizes the dimensions of the sample as well as the initial crack
is fitted through these points. In case of the DCB test, Paris law
within the mid-plane of the specimen.
description is also included in Fig. 12 by considering 1.79 and 5.03
Crack propagation assessment requires the amount of the crack
for 𝑚 and 𝑛 parameters, respectively. As it can be observed, the present
extension within the sample at regular time intervals. However, mea-
model has the ability to predict mode I fatigue crack behavior in the
suring the crack extent directly from the finite element model in the propagation stage with an acceptable accuracy.
post-process could be time consuming and inefficient. To address this On the other hand, however, it seems to be necessary to investi-
issue, an experimentally-measurable quantity could be employed as a gate the fatigue crack propagation in mode two shear loading, since
reference, and the crack length will be obtained accordingly after a the model treats different modes in independent ways. Validating the
calibration process. For the DCB test, an analytical formulation based numerical model in the shear mode would pave the way in applying the
on the corrected beam theory has been presented to calculate the crack current formulation in more complex loading conditions and structures.
length, 𝑎, as a function of the specimen compliance, 𝐶, [42]: ENF is considered as the standard test method to evaluate the current
( ) 13 implementation. A schematic view of this test is presented in Fig. 13,
3
𝑎= 𝐶 𝐸𝐼 − 𝜒ℎ (23) while the dimensions are mentioned in Table 3. A cyclic displacement
2
is applied at the center of the beam, and the two ends are modeled to
in which
be simply-supported.
𝑏ℎ3 As in the DCB test, ENF configuration is also accompanied by closed
𝐸𝐼 = 𝐸11 (24)
12 form solutions to calculate the crack extension and the energy release
√ [
𝐸11 ( )2 ] rate. Crack length could be calculated according to the instantaneous
𝛤 specimen compliance (𝐶) as below [71]:
𝜒= 3−2 (25)
11𝐺13 1+𝛤
( )1
96𝐶 𝐸𝐼 − 2𝐿3 3
where 𝑎= − 0.42𝜒ℎ (29)
√ 3
𝐸11 𝐸22
𝛤 = 1.18 (26) where 𝐸𝐼 and 𝜒 can be calculated from Eqs. (24) and (25), respectively.
𝐺13 Mode II energy release rate then would be:
being 𝜒 and 𝛤 the crack length and transverse modulus correction
3(𝑎 + 0.42𝜒ℎ)2 𝑃 2
parameters, respectively. Energy release rate in mode I loading (𝐺𝐼 ) 𝐺𝐼𝐼 = (30)
64𝑤 𝐸𝐼

10
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 12. Numerical results of the crack propagation within the DCB configuration compared to the experimental data [42,69,70].

Fig. 13. End Notch Flexure (ENF) configuration within the crack propagation regime in pure mode II loading.

As opposed to the DCB configuration, ENF tests are conducted under theory has been employed. Indeed, the theory utilizes a combination
load-control conditions at load levels of 𝑃 = 660.8 N, 809.5 N, 931.7 N, of stresses of all layers according to their volumetric participation at
𝑎𝑛𝑑 1037.7 N with a stress ratio of 𝑅 = 0.1. Crack propagation and the integration point level. This could lead to the loss of some data
energy release rates are calculated and compared to the experimental regarding the position of the layers, as it has been commented in [51].
data [72] at each load level as shown in Fig. 14. The issue could be more pronounced in cases that there is a neutral axis
Comparing experimental and numerical values, a good agreement within the sample, as in case of the ENF configuration, which can lead
could be observed in terms of the energy release rate and crack exten- to some deviations from the experimental results.
sion characteristics. There are two drops in the crack propagation rates. A Mixed Mode Bending (MMB) test is considered to assess the
The first one, which is observed in all loading levels at the beginning present model within a combination of mode I and mode II loadings.
The configuration is presented in Fig. 16(a), where a rigid lever is
of the propagation stage, is due to the formation of the process zone
employed to exert certain amount of loading to the central and end
whilst the second one, which could be noticed at Figs. 14(c) and 14(d)
part of the sample. The value of 𝐶 parameter would specify the mode
at the end of the propagation, is mainly due to the crack arrival to the
mix ratios within the fatigue tests. Dimensions of the MMB sample are
center of the specimen, where stress conditions are changed according
mentioned in Table 3.
to the exerted central load at the crack tip vicinity. In order to have
According to Fig. 16(b), the loading lever is eliminated in the
a unified response composed of all loading levels, points of 5 percent simulations and two equivalent loads are considered. The proportion
increment in specimen’s compliance are obtained, and a Paris law in the of these two loads determines the mode mix ratio of the problem as
form of Eq. (28) is fitted through these points, as illustrated in Fig. 15. follows [12]:
Values of 0.30 and 2.93 are used for 𝑚 and 𝑛 parameters, respectively. 𝐿
𝑃1 𝑐+
As it can be seen, although the numerically-obtained Paris law is = 2
. (31)
within an acceptable range of the experimental values, there are some 𝑃2 𝑐
discrepancies, specifically at high crack propagation rates. This could In current simulation, a value of 41.3 mm is chosen, which corresponds
be attributed to the main assumptions made once the homogenization to a mode mixity of 50%. To obtain the crack length, a quadratic

11
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 14. Fatigue crack propagation in shear mode II at (a) 𝑃 = 1037.7 N, (b) 𝑃 = 931.7 N, (c) 𝑃 = 809.5 N, and (d) 𝑃 = 660.8 N compared to the experimental data [72].

Fig. 15. Paris law fitted curve on the overall response in all the loading levels compared with the experimental results in loading mode II conditions [72].

polynominal has been presented in [42], which is used in the current Load (𝑃 ), and Load-point deflection (𝛿) are obtained as [12,51]:
study. For a specimen with 25.4 mm width, the crack length, 𝑎 [mm], 𝐿∕2
𝑃 = 𝑃 (33)
based on compliance, 𝛿∕𝑃 [mm∕N], is: 𝑐 + 𝐿∕2 1

( )2 ( )
𝛿 𝛿 2𝑐 + 𝐿 2𝑐
𝑎 = −3.37 ⋅ 105 + 8.55 ⋅ 103 − 1.572 (32) 𝛿= △𝐶 − △𝐸 (34)
𝑃 𝑃 𝐿 𝐿

12
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 16. Mixed Mode Bending (MMB) configuration within the crack propagation area in (a) real test conditions and (b) equivalent numerical configuration.

△𝐶 and △𝐸 being the central deflection and end opening as illustrated by incorporating the R-curve effects, in which the critical energy release
in Fig. 16(b). Mixed-mode energy release rate could be obtained using rate is increased as a function of the crack extension.
the following relation [71]:
[ ( ) ] 4. Conclusion
𝐿 2 𝑃2
𝐺 = 4 3𝑐 − (𝑎 + 𝜒ℎ)2 + 3 (𝑐 + 𝐿)2 (𝑎 + 0.42𝜒ℎ)2 ⋅ ( )2
2 𝐿
64𝑤 𝐸𝐼𝑏 In the present study, a novel methodology was developed to sim-
2
ulate fatigue crack initiation and propagation within a homogenized
(35) medium. To this end, explicit representation of the bulk layers and in-
in which terfaces were eliminated and their mechanical response was considered
( ) ( ) [ ( )3 ] inside a rule of mixtures approach. Interfacial stresses were calculated
𝐿 2 𝐿 2 𝐿
4 3𝑐 − (𝑎 + 𝜒ℎ)3 + 𝑐 + 3 (𝑎 + 0.42𝜒ℎ)3 + 2 based on the state of stress in the adjacent bulk layers. In addition
2 2 2
𝐸𝐼𝑏 = to the damage under monotonic loads, the possibility of mechanical
( )2
𝐿 properties deterioration under cyclic loads was added to the homoge-
96 𝐶
2 nized formulation via a fatigue reduction function for each mode I and
(36) mode II loadings, independently. Conceptualization of this function was
conducted in three distinct fatigue regions, i.e., sub-critical, critical,
Simulations are conducted in a load-control mode at three load levels of
and over-critical stages, based on the stress level at each integration
𝑃 = 222 N, 𝑃 = 193 N, and 𝑃 = 170 N, while maintaining the stress ratio
point. An advance-in-time strategy was added to the calculations to skip
to be 𝑅 = 0.1. Simulation results are compared with the experimental
values reported in [42] at the same mode mixity, as it is presented in some cycles once a relative stability had been obtained in terms of the
Fig. 17. damage evolution.
Although a good correlation could be found between numerical Calibration of the model was conducted according to the Wöhler
and experimental results in high load levels, there is a deviation in diagrams of mode I and mode II independently for unidirectional car-
smaller load domains, as it can bee observed in the case of 𝑃 = 170 N, bon/epoxy (IM7/8552) composite material. Three-point bending and
however, the simulation results still fall within an acceptable range of double-notched shear tests were employed to assess the delamination
the experimental counterparts in terms of the rate of diagram evolution. initiation at different stress ratios. The results highlighted the capability
The error could be attributed to the homogenization assumption, as of the present model in capturing the number of cycles to initiate
it has been outlined for the ENF loading case. A Paris law curve of a catastrophic crack formation within the model. On the other side,
the form of Eq. (28) is fitted to the points of 5 percent increase in crack propagation was investigated through Double Cantilever beam
compliance with 0.11 and 2.47 for 𝑚 and 𝑛 parameters, respectively. A (DCB), End Notch Flexure (ENF), and Mixed Mode Bending (MMB)
close examination of the obtained results would uncover the fact that tests for normal, shear, and mixed mode loading scenarios. Closed form
there are some shifts in the experimental data to the right or left side of equations were utilized in combination with the finite element solution
the fitted Paris law. As it is mentioned in [42], this could be improved to calculate fatigue crack extension on the basis of the specimen’s

13
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

Fig. 17. Simulation results of the MMB configuration compared with the experimental data [42].

compliance. A good agreement was observed in the crack propagation Acknowledgement

behavior with respect to the experimental data, however, some discrep-
ancies were detected specifically at low loading levels, which might The authors Lucia Gratiela Barbu and Alejandro Cornejo are Serra
be ascribed to the homogenization theory and the elimination of the Húnter Fellows.
explicit representation of the layers.
Utilizing virtual layers and interfaces makes the present model a References
highly favorable option for simulating large-scale structures with large
[1] Tsokanas P, Loutas T, Kotsinis G, Kostopoulos V, van den Brink WM, Martin
number of layers. This also facilitates the pre-process stage as there
de la Escalera F. On the fracture toughness of metal-composite adhesive joints
will be no need to consider a layer-by-layer approach. In addition, with bending-extension coupling and residual thermal stresses effect. Composites
calibrating the fatigue parameters using only one interface, the model B 2020;185:107694. http://dx.doi.org/10.1016/j.compositesb.2019.107694, URL
would be able to take different orientations and stacking sequences into https://www.sciencedirect.com/science/article/pii/S1359836819317548.
account. Although there would be some loss of accuracy according to [2] Eliasson S, Hultgren G, Wennhage P, Barsoum Z. Numerical fatigue assessment
of a cross-ply carbon fiber laminate using a probabilistic framework. Composites
the homogenization procedure, the present study demonstrated that
B 2024;281:111514. http://dx.doi.org/10.1016/j.compositesb.2024.111514, URL
this issue is not significant at the coupon level, which will pave the https://www.sciencedirect.com/science/article/pii/S1359836824003251.
way of applying the method at the component level such as naval, [3] Zhou X-Y, Qian S-Y, Wang N-W, Xiong W, Wu W-Q. A review on
automotive, or aerospace structures. stochastic multiscale analysis for FRP composite structures. Compos Struct
2022;284:115132. http://dx.doi.org/10.1016/j.compstruct.2021.115132, URL
The model is compatible with any standard finite element code, and
https://www.sciencedirect.com/science/article/pii/S026382232101549X.
is currently implemented in the Kratos Multi-physics framework, which [4] Granados JJ, Martinez X, Barbu LG, Di Capua D. Fatigue prediction
is accessible in an open-source platform. of composite materials, based on serial/parallel mixing theory. Compos
Struct 2022;291:115516. http://dx.doi.org/10.1016/j.compstruct.2022.115516,
CRediT authorship contribution statement URL https://www.sciencedirect.com/science/article/pii/S0263822322003063.
[5] Nishi Y, Hosoi A, Kawada H. Evaluation of matrix crack growth in interlam-
inar toughened quasi-isotropic carbon-fiber reinforced plastic laminates up to
Alireza Taherzadeh-Fard: Writing – original draft, Software, the very-high cycle regime by ultrasonic fatigue testing. Compos Sci Technol
Methodology, Investigation, Conceptualization. Sergio Jiménez: Writ- 2024;253:110623. http://dx.doi.org/10.1016/j.compscitech.2024.110623, URL
ing – review & editing, Software. Alejandro Cornejo: Writing – review https://www.sciencedirect.com/science/article/pii/S0266353824001933.
[6] Kotrotsos A, Geitona A, Kostopoulos V. On the mode I and mode II fa-
& editing, Software. Eugenio Oñate: Writing – review & editing, Super-
tigue delamination growth of CFRPs modified by electrospun bis-maleimide
vision, Funding acquisition. Lucia Gratiela Barbu: Writing – review & resin. Compos Sci Technol 2023;237:110000. http://dx.doi.org/10.1016/j.
editing, Supervision, Project administration, Funding acquisition. compscitech.2023.110000, URL https://www.sciencedirect.com/science/article/
pii/S0266353823000933.
Declaration of competing interest [7] Stoll MM, Weidenmann KA. Fatigue of fiber-metal-laminates with aluminum
core, CFRP face sheets and elastomer interlayers (FMEL). Int J Fatigue
2018;107:110–8.
The authors declare that they have no known competing finan- [8] Heidari-Rarani M, Sayedain M. Finite element modeling strategies for 2D and
cial interests or personal relationships that could have appeared to 3D delamination propagation in composite DCB specimens using VCCT, CZM
influence the work reported in this paper. and XFEM approaches. Theor Appl Fract Mech 2019;103:102246.
[9] Abdul Hamid ZM, Findeisen C, Laveuve DM, Rohrmüller B, Richter H, De
Monte M, Spancken D, Hohe J. A continuum damage mechanics model for fatigue
Data availability degradation and failure of short fiber reinforced composites. Compos Sci Technol
2024;110679. http://dx.doi.org/10.1016/j.compscitech.2024.110679, URL https:
Data will be made available on request. //www.sciencedirect.com/science/article/pii/S0266353824002495.

14
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

[10] Zhao L, Zhi J, Zhang J, Liu Z, Hu N. XFEM simulation of delamination in [30] Bender J, Bak B, Carreras L, Lindgaard E. Intralaminar crack growth rates of a
composite laminates. Composites A 2016;80:61–71. glass fibre multiaxial laminate subjected to variable amplitude loading. Compos-
[11] Bak BL, Sarrado C, Turon A, Costa J. Delamination under fatigue loads in com- ites B 2023;252:110510. http://dx.doi.org/10.1016/j.compositesb.2023.110510,
posite laminates: a review on the observed phenomenology and computational URL https://www.sciencedirect.com/science/article/pii/S1359836823000136.
methods. Appl Mech Rev 2014;66(6):060803. [31] Zhang H, Qiao P. On the computation of energy release rates by a peri-
[12] Garg N, Prusty BG, Ooi ET, Song C, Pearce G, Phillips AW. Application of scaled dynamic virtual crack extension method. Comput Methods Appl Mech Engrg
boundary finite element method for delamination analysis of composite lami- 2020;363:112883. http://dx.doi.org/10.1016/j.cma.2020.112883, URL https://
nates using cohesive zone modelling. Compos Struct 2020;253:112773. http:// www.sciencedirect.com/science/article/pii/S0045782520300657.
dx.doi.org/10.1016/j.compstruct.2020.112773, URL https://www.sciencedirect. [32] Davis B, Wawrzynek P, Hwang C, Ingraffea A. Decomposition of 3-D mixed-
com/science/article/pii/S0263822320326994. mode energy release rates using the virtual crack extension method. Eng Fract
[13] Hu P, Li X, Lubineau G. Prediction of a complex delamination front using a Mech 2014;131:382–405. http://dx.doi.org/10.1016/j.engfracmech.2014.08.014,
general cohesive model. Compos Sci Technol 2023;233:109911. http://dx.doi. URL https://www.sciencedirect.com/science/article/pii/S0013794414002768.
org/10.1016/j.compscitech.2023.109911, URL https://www.sciencedirect.com/ [33] Farshidi A, Berggreen C, Schäuble R. Numerical fracture analysis and model
science/article/pii/S0266353823000040. validation for disbonded honeycomb core sandwich composites. Compos Struct
[14] Zhu M, Gorbatikh L, Lomov SV. An incremental-onset model for fa- 2019;210:231–8. http://dx.doi.org/10.1016/j.compstruct.2018.11.052, URL
tigue delamination propagation in composite laminates. Compos Sci Technol https://www.sciencedirect.com/science/article/pii/S0263822318322347.
2020;200:108394. http://dx.doi.org/10.1016/j.compscitech.2020.108394, URL [34] Zhou H, Li S, Xie K, Lu X, Zhao Y, Tay T. Mode II interlaminar fracture of hybrid
https://www.sciencedirect.com/science/article/pii/S0266353820321849. woven carbon-dyneema composites. Composites A 2020;131:105785. http://
[15] Carreras L, Bak B, Jensen S, Lequesne C, Xiong H, Lindgaard E. Benchmark dx.doi.org/10.1016/j.compositesa.2020.105785, URL https://www.sciencedirect.
test for mode I fatigue-driven delamination in GFRP composite laminates: com/science/article/pii/S1359835X20300233.
Experimental results and simulation with the inter-laminar damage model [35] Turon A, Costa J, Camanho P, Dávila C. Simulation of delamination in com-
implemented in SAMCEF. Composites B 2023;253:110529. http://dx.doi.org/10. posites under high-cycle fatigue. Composites A 2007;38(11):2270–82. http://
1016/j.compositesb.2023.110529, URL https://www.sciencedirect.com/science/ dx.doi.org/10.1016/j.compositesa.2006.11.009, URL https://www.sciencedirect.
article/pii/S135983682300032X. com/science/article/pii/S1359835X0600337X, CompTest 2006.
[16] Knopp A, Funck E, Holtz A, Scharr G. Delamination and compression-after-impact [36] May M, Hallett SR. An advanced model for initiation and propagation of
properties of z-pinned composite laminates reinforced with circumferentially damage under fatigue loading–part I: Model formulation. Compos Struct
notched z-pins. Compos Struct 2022;285:115188. http://dx.doi.org/10.1016/ 2011;93(9):2340–9.
j.compstruct.2022.115188, URL https://www.sciencedirect.com/science/article/
[37] May M, Pullin R, Eaton M, Featherston C, Hallett SR. An advanced model
pii/S0263822322000058.
for initiation and propagation of damage under fatigue loading–part II: Matrix
[17] Ye Q, Chen P. Prediction of the cohesive strength for numerically simulating cracking validation cases. Compos Struct 2011;93(9):2350–7.
composite delamination via CZM-based FEM. Composites B 2011;42(5):1076–
[38] Zhao M, Zhao Y, Wang A, Chang Z, Zhang J, Wang Z. Investigation of the
83. http://dx.doi.org/10.1016/j.compositesb.2011.03.021, URL https://www.
mode-I delamination behavior of double-double laminate carbon fiber reinforced
sciencedirect.com/science/article/pii/S1359836811001181.
composite. Compos Sci Technol 2024;248:110463. http://dx.doi.org/10.1016/j.
[18] O’Brien TK, Chawan AD, Krueger R, Paris IL. Transverse tension fatigue life
compscitech.2024.110463, URL https://www.sciencedirect.com/science/article/
characterization through flexure testing of composite materials. Int J Fatigue
pii/S0266353824000332.
2002;24(2–4):127–45.
[39] Tao C, Qiu J, Yao W, Ji H. A novel method for fatigue delamination simu-
[19] Brod M, Dean A, Scheffler S, Gerendt C, Rolfes R. Numerical modeling and
lation in composite laminates. Compos Sci Technol 2016;128:104–15. http://
experimental validation of fatigue damage in cross-ply CFRP composites under in-
dx.doi.org/10.1016/j.compscitech.2016.03.016, URL https://www.sciencedirect.
homogeneous stress states. Composites B 2020;200:108050. http://dx.doi.org/10.
com/science/article/pii/S026635381630104X.
1016/j.compositesb.2020.108050, URL https://www.sciencedirect.com/science/
[40] Springer M, Turon A, Pettermann H. A thermo–mechanical cyclic cohesive zone
article/pii/S1359836820306624.
model for variable amplitude loading and mixed–mode behavior. Int J Solids
[20] Bak B, Turon A, Lindgaard E, Lund E. A benchmark study of simulation methods
Struct 2019;159:257–71. http://dx.doi.org/10.1016/j.ijsolstr.2018.10.004, URL
for high-cycle fatigue-driven delamination based on cohesive zone models.
https://www.sciencedirect.com/science/article/pii/S0020768318303986.
Compos Struct 2017;164:198–206.
[41] Roth S, Kuna M. Prediction of size-dependent fatigue failure modes by means of
[21] Maragoni L, Carraro PA, Simonetto M, Quaresimin M. A novel method to
a cyclic cohesive zone model. Int J Fatigue 2017;100:58–67. http://dx.doi.org/
include crack-induced delamination in a fatigue damage predictive procedure
10.1016/j.ijfatigue.2017.01.044, URL https://www.sciencedirect.com/science/
for composite laminates. Compos Sci Technol 2023;238:110011. http://dx.doi.
article/pii/S0142112317300531.
org/10.1016/j.compscitech.2023.110011, URL https://www.sciencedirect.com/
science/article/pii/S0266353823001045. [42] Dávila C. From S-N to the Paris law with a new mixed-mode cohe-
[22] May M, Hallett SR. An advanced model for initiation and propagation of sive fatigue model for delamination in composites. Theor Appl Fract Mech
damage under fatigue loading – part I: Model formulation. Compos Struct 2020;106:102499. http://dx.doi.org/10.1016/j.tafmec.2020.102499, URL https:
2011;93(9):2340–9. http://dx.doi.org/10.1016/j.compstruct.2011.03.022, URL //www.sciencedirect.com/science/article/pii/S0167844219305579.
https://www.sciencedirect.com/science/article/pii/S0263822311001061. [43] Falcó O, Ávila R, Tijs B, Lopes C. Modelling and simulation methodology for
[23] May M, Hallett SR. A combined model for initiation and propagation of unidirectional composite laminates in a virtual test lab framework. Compos
damage under fatigue loading for cohesive interface elements. Composites Struct 2018;190:137–59. http://dx.doi.org/10.1016/j.compstruct.2018.02.016,
A 2010;41(12):1787–96. http://dx.doi.org/10.1016/j.compositesa.2010.08.015, URL https://www.sciencedirect.com/science/article/pii/S0263822317329768.
URL https://www.sciencedirect.com/science/article/pii/S1359835X10002356. [44] Settgast RR, Rashid MM. Continuum coupled cohesive zone elements for analysis
[24] Rafiee R, Sotoudeh S. A hysteresis cohesive approach for predicting mixed-mode of fracture in solid bodies. Eng Fract Mech 2009;76(11):1614–35.
delamination onset of composite laminates under cyclic loading: Part I, model [45] Roy P, Deepu S, Pathrikar A, Roy D, Reddy J. Phase field based peridynam-
development. Compos Struct 2021;277:114667. ics damage model for delamination of composite structures. Compos Struct
[25] Krueger R. Development of a benchmark example for delamination fatigue 2017;180:972–93.
growth prediction. Fatigue Compos Mater 2010;3:54–73. [46] Reinoso J, Blázquez A, Távara L, París F, Arellano C. Damage tolerance of com-
[26] Carloni C, Subramaniam KV. Investigation of sub-critical fatigue crack growth posite runout panels under tensile loading. Composites B 2016;96:79–93. http://
in FRP/concrete cohesive interface using digital image analysis. Composites dx.doi.org/10.1016/j.compositesb.2016.03.083, URL https://www.sciencedirect.
B 2013;51:35–43. http://dx.doi.org/10.1016/j.compositesb.2013.02.015, URL com/science/article/pii/S1359836816301743.
https://www.sciencedirect.com/science/article/pii/S135983681300070X. [47] Oller S, Salomón O, Oñate E. A continuum mechanics model for mechani-
[27] Tang S, Lemanski S, Zhang X, Ayre D. Fatigue life prediction of z-fibre pinned cal fatigue analysis. Comput Mater Sci 2005;32(2):175–95. http://dx.doi.org/
composite laminate under mode I loading. Compos Sci Technol 2019;174:221– 10.1016/j.commatsci.2004.08.001, URL https://www.sciencedirect.com/science/
31. http://dx.doi.org/10.1016/j.compscitech.2019.02.010, URL https://www. article/pii/S0927025604001958.
sciencedirect.com/science/article/pii/S0266353818321249. [48] Barbu LG, Oller S, Martinez X, Barbat AH. High-cycle fatigue consti-
[28] Feng W, Arouche MM, Pavlovic M. Fatigue crack growth characterization tutive model and a load-advance strategy for the analysis of unidirec-
of composite-to-steel bonded interface using ENF and 4ENF tests. Compos tional fiber reinforced composites subjected to longitudinal loads. Compos
Struct 2024;334:117963. http://dx.doi.org/10.1016/j.compstruct.2024.117963, Struct 2019;220:622–41. http://dx.doi.org/10.1016/j.compstruct.2019.04.015,
URL https://www.sciencedirect.com/science/article/pii/S0263822324000916. URL https://www.sciencedirect.com/science/article/pii/S0263822318336481.
[29] Tanulia V, Wang J, Pearce GM, Baker A, Chang P, Gangadhara Prusty B. [49] Jiménez S, Barbu LG, Oller S, Cornejo A. On the numerical study of fatigue
Experimental and computational assessment of disbond growth and fatigue life process in rail heads by means of an isotropic damage based high-cycle fatigue
of bonded joints and patch repairs for primary airframe structures. Int J Fatigue constitutive law. Eng Fail Anal 2022;131:105915. http://dx.doi.org/10.1016/
2022;159:106776. http://dx.doi.org/10.1016/j.ijfatigue.2022.106776, URL https: j.engfailanal.2021.105915, URL https://www.sciencedirect.com/science/article/
//www.sciencedirect.com/science/article/pii/S0142112322000561. pii/S1350630721007767.

15
A. Taherzadeh-Fard et al. Composite Structures 351 (2025) 118613

[50] Gonçalves LA, Jiménez S, Cornejo A, Barbu L, Parareda S, Casellas D. Numerical [59] Dávila CG. From SN to the Paris law with a new mixed-mode cohesive fatigue
simulation of a rapid fatigue test of high Mn-TWIP steel via a high cycle fatigue model. Tech. rep., 2018.
constitutive law. Int J Fatigue 2023;168:107444. http://dx.doi.org/10.1016/j. [60] The MathWorks, Inc. MATLAB R2013a (8.1.0.604). Natick, Massachusetts, USA;
ijfatigue.2022.107444, URL https://www.sciencedirect.com/science/article/pii/ 2013, Software.
S0142112322006946. [61] Maiti S, Geubelle PH. A cohesive model for fatigue failure of polymers. Eng Fract
[51] Taherzadeh-Fard A, Cornejo A, Jiménez S, Barbu LG. A rule of mixtures approach Mech 2005;72(5):691–708.
for delamination damage analysis in composite materials. Compos Sci Technol [62] Oller S. Nonlinear dynamics of structures. Springer; 2014.
2023;242:110160. http://dx.doi.org/10.1016/j.compscitech.2023.110160, URL [63] Barbu LG, Oller S, Martinez X, Barbat A. High cycle fatigue simulation: A new
https://www.sciencedirect.com/science/article/pii/S0266353823002531. stepwise load-advancing strategy. Eng Struct 2015;97:118–29. http://dx.doi.org/
[52] Truesdell C, Toupin R. The classical field theories. In: Principles of classical 10.1016/j.engstruct.2015.04.012, URL https://www.sciencedirect.com/science/
mechanics and field theory/prinzipien der klassischen mechanik und feldtheorie. article/pii/S0141029615002527.
Springer; 1960, p. 226–858. [64] Ferrándiz VM, Bucher P, Rossi R, Cotela J, Carbonell J, Zorrilla R, Tosi R.
[53] Cornejo Velázquez A. A fully Lagrangian formulation for fluid-structure in- KratosMultiphysics (version 8.0). 2020.
teraction between free-surface flows and multi-fracturing solids and structures [65] Dadvand P, Rossi R, Oñate E. An object-oriented environment for developing
(Thesis), 2020. finite element codes for multi-disciplinary applications. Arch Comput Methods
[54] Car E, Oller S, Oñate E. An anisotropic elastoplastic constitutive model Eng 2010;17:253–97.
for large strain analysis of fiber reinforced composite materials. Comput [66] Ribó R, Pasenau M, Escolano E. GiD user manual. CIMNE; 2007.
Methods Appl Mech Engrg 2000;185(2):245–77. http://dx.doi.org/10.1016/ [67] Dávila CG, Rose CA, Murri GB, Jackson WC, Johnston WM. Evaluation of fatigue
S0045-7825(99)00262-5, URL https://www.sciencedirect.com/science/article/ damage accumulation functions for delamination initiation and propagation.
pii/S0045782599002625. Tech. rep., 2020.
[55] Balzani C, Wagner W. An interface element for the simulation of delamina- [68] May M, Hallett SR. An assessment of through-thickness shear tests for initiation
tion in unidirectional fiber-reinforced composite laminates. Eng Fract Mech of fatigue failure. Composites A 2010;41(11):1570–8.
2008;75(9):2597–615. [69] Murri GB. Evaluation of delamination onset and growth characterization methods
[56] Gonçalves LA, Jiménez S, Cornejo A, Tedesco MM, Barbu LG. A high cycle fatigue under mode I fatigue loading. Tech. rep., 2013.
numerical framework for component-level virtual fatigue testing: Application to [70] Nojavan S, Schesser D, Yang Q. An in situ fatigue-CZM for unified crack
a light-duty vehicle lower control arm. Eng Struct 2024;311:118198. initiation and propagation in composites under cyclic loading. Compos Struct
[57] Alcayde B, Merzkirch M, Cornejo A, Jiménez S, Marklund E, Barbu L. Fa- 2016;146:34–49.
tigue behaviour of glass-fibre-reinforced polymers: Numerical and experimental [71] Reeder JR, Demarco K, Whitley KS. The use of doubler reinforcement in
characterisation. Compos Struct 2024;337:118057. http://dx.doi.org/10.1016/ delamination toughness testing. Composites A 2004;35(11):1337–44.
j.compstruct.2024.118057, URL https://www.sciencedirect.com/science/article/ [72] O’Brien TK, Johnston WM, Toland GJ. Mode II interlaminar fracture toughness
pii/S0263822324001855. and fatigue characterization of a graphite epoxy composite material. Tech. rep.,
[58] Dávila CG, Joosten MW. A cohesive fatigue model for composite delamination 2010.
based on a new material characterization procedure for the Paris law. Eng Fract
Mech 2023;284:109232.

16