You are on page 1of 15

Composites Science and Technology 245 (2024) 110316

Contents lists available at ScienceDirect

Composites Science and Technology


journal homepage: www.elsevier.com/locate/compscitech

A new integrated modeling method for predicting low-velocity impact


behavior and residual tensile failure of Z-pinned T-joints
Jianwu Zhou a, Zhibin Zhao b, Liyong Jia c, **, Chao Zhang a, d, e, *
a
School of Aeronautics, Northwestern Polytechnical University, Xi’an, Shaanxi, 710072, China
b
Xi’an Research Institute of High-tech, Xi’an, Shaanxi, 710025, China
c
First Aircraft Institute of Aviation Industry Corporation, Xi’an, Shaanxi, 710089, China
d
School of Civil Aviation, Northwestern Polytechnical University, Xi’an, Shaanxi, 710072, China
e
Joint International Research Laboratory of Impact Dynamics and Its Engineering Application, Xi’an, Shaanxi, 710072, China

A R T I C L E I N F O A B S T R A C T

Keywords: Composite T-joints are highly susceptible to low-velocity impact, which can significantly affect their residual
A. Adhesive joints performance due to the primary working condition of bearing out-of-plane tensile loads. Currently, most
B. Impact behavior methods that employ multiple models or analytical steps to sequentially assess the mechanical properties of
C. Residual stress
composites generally exhibit certain limitations, leaving room for improvement. This study has developed a finite
Bridging effect
element (FE) model to simulate the low-velocity impact and post-impact tensile behaviors of carbon fiber
reinforced polymers (CFRP) T-joints using an integrated analysis method. The model is based on stress failure
criteria and continuous stiffness degradation theory and incorporates corrections to the damage variables. Both
the low-velocity impact and quasi-static tensile portions of the model are implemented using an explicit solver
with the VUMAT subroutine for calculations in Abaqus. The element damage states are transferred between the
two models via a Python script, mitigating the inefficiencies and uncontrollable errors associated with the
traditional method of transferring element information between multiple models or analytical steps. Finally, the
numerical results of mechanical response and damage states are compared with experimental findings from
various perspectives, and the bridging mechanism of Z-pins is thoroughly investigated. The results show that the
model exhibits a maximum error of 10.41 % in the main key parameters during low-velocity impact and a
maximum error of 10.30 % in the ultimate load during post-impact tension. The model’s delamination damage
state and final tensile failure mode closely align with the experimental results. Furthermore, a comprehensive
analysis of the FE model indicates that the pull-out force of the Z-pin is unrelated to its implantation position or
pull-out rate, and the reinforcing effect of the Z-pin becomes significantly apparent only after the CFRP T-joint
reaches a certain degree of initial damage.

1. Introduction Considering that T-joints commonly operate under tensile loading con­
ditions [4,5], conducting a study on the response to low-velocity impacts
The integrated design and manufacturing of composite material and the ensuing residual tensile performance holds significant engi­
structures is a crucial trend in advancing the lightweight development of neering value.
future aircraft. Composite T-joints, being a quintessential form of inte­ At present, research on the structural performance of composite T-
grated structure in aircraft, offer notable advantages including the joints predominantly centers around aspects like tension, bending, and
reduction of fastener count, decreased structural weight, and lowered fatigue, etc. [6–8]. In these investigations, T-joint failures often stem
assembly costs [1–3]. Nevertheless, during actual service operations, from delamination damage. As a result, various reinforcement tech­
T-joints inevitably suffer from external low-velocity impact, such as niques have been employed to address the interlaminar toughening
tools accidently falling during aircraft maintenance, generating impact challenge in composites, including Z-pin, tufting, 3D weaving, and more
energies typically ranging from several joules to tens of joules. [9]. Among these methods, Z-pin reinforcement has garnered

* Corresponding author. School of Aeronautics, Northwestern Polytechnical University, Xi’an, Shaanxi, 710072, China.
** Corresponding author.
E-mail addresses: taishanbuzuo@163.com (L. Jia), chaozhang@nwpu.edu.cn (C. Zhang).

https://doi.org/10.1016/j.compscitech.2023.110316
Received 16 July 2023; Received in revised form 11 October 2023; Accepted 21 October 2023
Available online 29 October 2023
0266-3538/© 2023 Elsevier Ltd. All rights reserved.
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

considerable attention due to its cost-effectiveness, ease of imple­ steps. Post-impact simulation is achieved by resetting boundary condi­
mentation, and other advantages, and it has demonstrated substantial tions and conducting software restart calculations. By this means, Liu
improvements in damage tolerance [10–12]. Nevertheless, some studies et al. [22], Tuo et al. [23,24], and Jia et al. [25] successfully simulated
have also revealed instances where the damage tolerance of Z-pin the evolution of impact damage in composite laminates and forecasted
reinforced composite structures not only failed to achieve the antici­ their residual compressive strength after impact. Similarly, using the
pated enhancements, but experienced some degree of deterioration. For same method, Cheng et al. [26–28] have gone a step further, achieving
instance, Francesconi et al. [13] examined the impact of Z-pin on the progressive damage simulation in composite laminates and even heli­
residual performance of composite laminates following low-energy im­ copter tail structures across various load scenarios, including tension,
pacts (below 5 J), using a comprehensive array of mechanical tests and compression, and fatigue following an impact. They have achieved high
non-destructive testing methods. They observed that the Z-pin rein­ levels of prediction accuracy for performance metrics. However, the
forced laminates exhibited lower residual compressive strength restart analysis method has also exposed persistent challenges in the
compared to regular laminates under such conditions. Zhang et al. [14] research process. Owing to the intricate failure behavior of composites,
analyzed the effect of Z-pin implantation angles on the post-impact certain elements may experience substantial mesh distortion during
compression performance of composite laminates. Their findings indi­ impact without meeting the deletion threshold in the impact model. This
cated that laminates with Z-pins implanted at a 60◦ angle displayed poses significant hurdles for mesh convergence in subsequent
weaker compressive strength after high-energy impacts (40–60 J) than quasi-static simulations, often necessitating manual searches and de­
ordinary laminates. Koh et al. [15] discovered that more Z-pin implan­ letions prior to implementing subsequent simulations. These laborious
tations did not necessarily yield better results. Excessive Z-pin implan­ manual operations further impede the computational efficiency of the
tation in certain regions led to earlier T-joint failures during quasi-static model. In summary, the primary issue with the aforementioned two
tension tests. Additionally, Hoffmann et al. [16] delved into the adverse methods lies in the ineffective transfer of element information, espe­
effects of Z-pin usage on the tensile strength and fatigue life of laminates cially element damage states, between adjacent models or analytical
in quasi-static tensile and dynamic fatigue tests. It is evident that the steps. Furthermore, the existing research mainly focuses on predicting
application of Z-pin reinforcement technology in aerospace composite the residual performance of simple composite structures, especially the
structures continues to encounter challenges, especially in the context of plate. The structural complexity of composite T-joints and the influence
composite T-joints. of Z-pin reinforcement on these structures further heighten the chal­
Over the past few decades, researchers have made concerted efforts lenges associated with analyzing their damage tolerance. If these two
to address the challenge of damage tolerance in composites by inte­ methods were applied to analyze the low-velocity impact and residual
grating experimental and simulation methods. However, due to the performance of CFRP T-joints, it would inevitably escalate modeling
intricate and dispersed nature of composite damage, simulating their complexities and compromise calculation accuracy. Therefore, it is
dynamic mechanical behavior and predicting overall structural perfor­ imperative to develop an efficient FE integrated analysis method
mance remains an arduous task, when it comes to forecasting the re­ capable of facilitating seamless and minimal loss transfer of element
sidual performance of composites after impact—an issue of paramount information, particularly element damage states, between multiple
importance that has garnered substantial attention [17]. Currently, the models.
numerical analysis methods commonly employed to predict the per­ Building upon prior experimental research [29], this study pre­
formance of composites after low-velocity impact can be categorized dominantly relies on numerical methods in tandem with the VUMAT
into two primary types: the equivalent damage method and restart subroutine to simulate the mechanical behavior and damage states in
analysis method. The former method operates on a relatively straight­ CFRP T-joints subjected to low-velocity impact condition (model 1) and
forward principle. It predominantly relies on non-destructive testing subsequent post-impact quasi-static tension (model 2). Furthermore, a
results to ascertain the internal damage state of composites post-impact. Python script serves as a conduit between these two models, facilitating
By simplifying the damage region or equivalently weakening material the extraction of final element damage state information from the pre­
parameters, it seeks to establish an equivalent substitution between the ceding model and its direct or indirect incorporation into corresponding
initial state of the residual performance analysis model and the final elements of the subsequent model. This efficient and low loss trans­
state of the low-velocity impact model. Rozylo et al. [18] and Baluch mission of damage states between models during continuous analysis
et al. [19] have utilized this approach to predict post-impact residual avoids issues commonly encountered in the restart analysis method,
strength by simplifying the impact damage characteristics of composites such as reduced calculation efficiency or failures due to mesh distortion
and mapping the damage region onto a compression FE model dis­ in the preceding model. Additionally, it eliminates the need for an
cretized with continuous shell elements. Ouyang et al. [20] based on the equivalent processing of actual damage caused by low-velocity impact,
measured impact damage dimensions, simplified the damage region into thereby circumventing discrepancies arising from different approaches.
concentric circles with varying stiffness and strength. They employed Finally, through comparisons with corresponding experimental results
the virtual crack closure technique to swiftly predict the residual from various angles, this method and model’s validity is confirmed.
compressive strength of CFRP laminates following impact. Zou et al. Furthermore, a comprehensive analysis of the failure mechanism of
[21] used non-destructive detection technology, partitioned damage in CFRP T-joints and the reinforcement mechanism of Z-pins is undertaken.
CFRP laminates post-impact into distinct regions. They then simplified This study introduces an integrated modeling method that paves the way
these regions based on size and shape, subsequently constructing a for addressing multi-model continuous analysis challenges.
phenomenological equivalent impact damage finite element (FE) model,
thus enabling simulation of the initiation, propagation, and failure 2. The damage model
behavior of compressive damage post-impact. However, due to the
excessive idealization of equivalent damage, particularly in the presence Generally speaking, composite damage can be roughly categorized
of highly complex composite failure mechanisms, it may underestimate into two main types: (a) intralaminar damage, which occurs within a
or even neglect the influence of critical damage modes, resulting in single ply and includes damage to both the fibers and the matrix; and (b)
significant deviations in prediction results. As a result, the restart interlaminar damage, which occurs between neighboring plies and in­
analysis method has garnered increased attention from scholars in volves delamination damage. Both of these damage process can be
recent years. effectively simulated using progressive damage models, which consist of
The restart analysis method entails incorporating setting multiple failure initiation criteria and damage evolution model. To enhance
analysis steps within a single model, where low-velocity impact simu­ simulation accuracy, the damage state of materials is modified to avoid
lation and post-impact simulation are positioned in successive analysis the delay of damage initiation and the advance of ultimate failure.

2
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

2.1. Intralaminar and interlaminar damage undergoes complete stiffness degradation. In other words, when
ε0 ≤ ε ≤ εf , the material is in the stage of stiffness degradation. How­
Hashin’s failure criterion [30] is a widely recognized and frequently ever, in actual numerical calculations, it is unlikely for f = 1 even at the
employed method for predicting damage initiation in single layer of onset of stiffness degradation. When f > 1 and the equivalent initial
composite laminates, both in academic and engineering fields. In this damage strain at this time is marked as ε′0 , then ε′0 > ε0 , meaning it is
study, the three-dimensional Hashin criterion is implemented, which overestimated. Since the fracture toughness (Gc ) of the material is
incorporates the through-thickness normal stress component in matrix constant and the element size remains unchanged, the corresponding
failure. The details of the four distinct failure modes can be found in equivalent failure strain (ε′f ) will be underestimated, viz., ε′f < εf , as
Appendix A. shown in Fig. 1(b).
Then the intralaminar damage evolution employs the mainstream Based on the aforementioned rules, when the VUMAT subroutine is
energy-based linear continuous degradation model, as shown in Fig. 1 executed, the stiffness degradation stage of the material may be mis­
(a). Specifically, prior to any damage occurring within the element, its
judged as ε′0 ≤ ε ≤ ε′f . As a result, when ε0 ≤ ε ≤ ε′0 occurs, the program
stress-strain relationship is represented by σ = C0 ε. Once damage oc­
is unable to reduce the stiffness of the material even though it has
curs, the stress-strain relationship changes to σ = Cd ε. C0 represents the
stiffness matrix of the undamaged material, while Cd represents the already undergone stiffness degradation. Conversely, when ε′f ≤ ε ≤ εf
stiffness matrix of the damaged material, and the specific expression is occurs, the program completely reduces the stiffness of the material
shown in Appendix B. It should be noted that in the compression mode, even though it has not completely failed. In this case, the damage var­
the material still has some residual strength (marked as σres iable does not match the actual situation, as expressed in Eq. (1), leading
i ) after
to premature or delayed damage.
complete failure, Faggiani and Falzon et al. [31] has shown that σres i was
equivalent to the transverse ultimate strength (marked as Yc ) of the ε′f ( ε′ ) εf ( ε0 )
material itself, so σ res 1− 0 ∕ (1)
i = Yc is taken.
d= = 1−
ε′
f − ε′
0
ε εf − ε0 ε
Since the failure strain of the element is influenced by itself, it is not
appropriate to treat the failure strain of the material as a fixed material To tackle this problem, this study incorporates f into the calculation
parameter. Based on Hillerborg’s theory of material fracture energy of damage variables to correct the values of ε0 and εf for the material.
[32], the characteristic length of the element is incorporated into the The correction principle is to record the equivalent strain, equivalent
stiffness degradation criterion of the material to mitigate the influence stress, and corresponding failure factor of the material at each moment
of the element size on the failure behavior of the material, and the during the entire loading process. When f exceeds 1 for the first time, the
fracture toughness is considered as a fixed material parameter and uses equivalent strain is scaled to the ideal state where f = 1 (viz., ε0 =
√̅̅̅
it to solve for failure strains under different modes, as expressed in Eqs. ε′0 / f ), thus obtaining the corrected value for ε0 . Simultaneously, the
(C1) of Appendix C. However, the characteristic length of the element equivalent stress is also scaled to the ideal state where f = 1 (viz., σ0 =
√̅̅̅
may become excessively large in tensile or compressive, resulting in σ′0 / f ) and used as the equivalent strength of the material under the
negative damage variables. To address this issue, while refining the current stress state. Then εf is calculated through Gc (viz., εf = 2Gc /σ 0 l).
mesh of the core region, its characteristic length is constrained by using Since this study adopts the three-dimensional Hashin’s failure criterion,
Eqs. (C2) of Appendix C. In other region, a coarser mesh is utilized. this method is used to correct the equivalent initial damage strain and
On the other hand, the built-in cohesive element in Abaqus based on strength under different failure modes. The results are shown in Ap­
a bilinear traction-separation relationship is adopted to simulate pendix E.
delamination at the interface. The quadratic stress criterion is chosen as
the initial criterion for delamination damage, which is expressed as Eq. 3. Integrated modeling method
(D1) of Appendix D. Then the damage evolution of interlaminar mate­
rials adopts the Benzeggagh-Kenane failure criterion [33,34], which is Sequential analysis between two or more FE models can be accom­
expressed as Eq. (D3) of Appendix D, and its damage variable under plished through two common methods: the restart method and the
mixed-mode is expressed in Eq. (D2) of Appendix D. equivalent damage method. The restart method utilizes the computation
results from the previous model as initial conditions for the subsequent
2.2. Correction of damage variables model by defining a predefined field. This approach is logically sound
and avoids the introduction of new errors during the transfer of damage
As stated in Section 2.1 and Appendix B, the accuracy of calculating state information between models. However, its implementation can be
Cd depends on obtaining precise values for the equivalent damage initial challenging when the previous model exhibits severe damage, such as
strain (ε0 ) and equivalent failure strain (εf ) of the element. Theoreti­ significant material stiffness degradation or element deletion. On the
cally, ε0 represents the equivalent strain when the failure factor (f) other hand, the equivalent damage method assigns initial damage, such
equals to 1. εf represents the equivalent strain when the material

Fig. 1. (a) Schematic diagram of energy-based linear continuous degradation model and (b) schematic diagram of the principle of damage initiation and failure
strain correction.

3
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

as pre-existing holes or cracks, to approximate the final mechanical with preset conditions to guide the calculation progression.
performance of the previous model with the initial mechanical perfor­ The data connection module forms the heart of the entire process. It
mance of the subsequent model. While this method has broad applica­ utilizes a Python script to parse the “.odb” file generated from the final
bility, it can be operationally complex and may introduce significant impact module calculation results and extracts intralaminar and inter­
errors, thus hindering an accurate representation of the true damage laminar element information (including element numbers) into two
state of the previous model. To address these challenges, this paper separate “.dat” files. For intralaminar elements, impact damage is
successfully transfers element damage state information between two transferred directly following this principle: using the VUMAT subrou­
models in a continuous analysis using Python scripts. tine to map the damage state variables obtained from the impact model
As shown in Fig. 2, this is the complete technical flowchart for FE to the corresponding intralaminar elements in the tensile model. For
simulation of low-velocity impact and post-impact tensile testing of a interlaminar elements, Cohesive elements provided by Abaqus are
composite T-joint. It can be roughly divided into four modules: pre- employed since it’s not feasible to directly map the damage state vari­
processing module, impact module, tensile module, and data connec­ ables using VUMAT. Instead, impact damage is transferred indirectly
tion module. following this principle: based on the damage state variables of different
In the pre-processing module, the T-joint and impactor are created, elements in the impact model, a degraded material model is established
and they are assigned appropriate material properties (including intra­ for each element (in this study, the damage state variables are divided
laminar and interlaminar properties), contact properties (including into 100 segments, and the material parameters of the intralaminar el­
surface and internal contact), and boundary conditions (only those ements are adjusted accordingly). These degraded models are then
required for low-velocity impact) to ensure the model faithfully repre­ assigned to the corresponding interlaminar elements in the tensile model
sents the experimental setup. Both the impact and tensile modules are in batches. This allows for low loss transmission of all element state
similar in that they employ Abaqus’ explicit solver, in conjunction with a information between two adjacent calculation models, as shown in
VUMAT subroutine, to carry out calculations. The key distinction is that Fig. 3. Subsequently, by defining new boundary conditions for the ten­
the latter includes an additional function that writes element informa­ sile test, the calculations can be seamlessly continued.
tion from a specified “.dat” file into the corresponding elements. During
the model solving, loads are incrementally applied to obtain material 4. Experiments and FE models
stiffness matrices, strains, stresses, and damage variables for each
analysis increment. These results are subsequently compared against To validate the effectiveness of the integrated model, a comparative

Fig. 2. Overall flowchart of the coordination of various modules.

4
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

Fig. 3. Low loss transmission of intralaminar element damage states (taking the matrix tensile as an example) and interlaminar element damage states (taking the
bonding interface as an example) between two models.

study on the low-velocity impact behavior and residual tensile proper­ direction to prevent the formation of a weak resin-rich zone.
ties of CFRP T-joints was conducted based on our previous experimental The Z-pins were crafted from S35/YH69 (PL/epoxy) prepreg fiber
work [29]. As shown in Fig. 4(a), two types of T-joint specimens with a 5 bundles with a diameter of 0.3 mm. They were inserted into the bonding
mm fillet radius were manufactured. The key difference between them region between the skin and flanges at 4 × 4 mm intervals. A total of 39
lies in the utilization of Z-pins for local interlaminar reinforcement. Z-pins were incorporated on each side of the specimen. The specimen
An IM7/M91 (carbon/epoxy) unidirectional prepreg with a nominal with or without Z-pins is referred to as the “Z-pinned T-joint” and
cured ply thickness of 0.175 mm was used to manufacture the skin and “Unpinned T-joint” in this study, respectively.
L-pieces with corresponding stacking sequences of [45/0/-45/90]3S and As depicted in Fig. 4(b), the T-joint specimen was fixed upside down
[45/0/-45/90]S, respectively. Additionally, the deltoid (highlighted in using a specialized fixture. The Instron Dynatup 9340 impact testing
Fig. 4(a)) was filled with the same material oriented along the 90o machine with a 10 mm hemispherical impactor and a 5.421 kg

Fig. 4. (a) Two types of T-joint specimens, test platforms of (b) low-velocity impact and (c) quasi-static tension, (d) 3D FE model of Z-pinned T-joint.

5
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

impacting part was used to strike at the center of the skin. The anti- mesh refinement for the deltoid and Z-pin regions is automatically
rebound device was activated to catch the impacting part and prevent determined based on the aforementioned settings. Consequently, the T-
a second impact. Subsequently, post-impact interlaminar damage was joint mesh comprises 338,304 C3D8R elements, 10,000 C3D6 elements,
meticulously documented for each individual specimen with the aid of and 252,904 COH3D8 elements. The impactor is modeled as an
ultrasonic C-scan equipment. Following the impact evaluations, quasi- analytical rigid body, with its initial velocity calculated based on the
static tensile tests were conducted on both intact specimens and those impactor mass and impact energy, which is then applied through a
that had incurred damaged during impact. These tests persisted until predetermined field. A dynamic penalty function is employed to address
complete failure was observed, all while maintaining a controlled tensile the contact problem within the current model, with a friction coefficient
rate of 0.5 mm/min, facilitated by a CRIMS DDL100 tensile testing set to 0.3.
machine, as shown in Fig. 4(c). The specific test arrangements are shown In alignment with the clamping conditions during the experiment,
in Appendix F. for the low-velocity impact simulation, a 30 mm section of the skin at
As per the experimental findings, it is evident that damage occur­ both ends is immobilized (including both upper and lower sides), while
rences are infrequent in both the terminal sections of the skin and the the other regions remain unconstrained, as shown in Fig. 4(b). For the
upper portion of the L-pieces in the T-joint during both low-velocity quasi-static tensile simulation, in addition to the same constraints at
impact and quasi-static tensile tests. Consequently, to enhance compu­ both ends of the skin as in the low-velocity impact simulation, a 30 mm
tational efficiency, the progressive damage and failure behavior of the segment of the L-pieces end is controlled using point-to-surface kine­
core region is considered, while only elastic behaviors are considered for matic coupling (including both left and right sides), with other regions
other regions. The 3D FE models, exemplified by the Z-pinned T-joint, is remaining free, as shown in Fig. 4(c).
based on the aforementioned experiments, as shown in Fig. 4(d). In Global and local coordinate systems are defined to account for ply
these models, C3D8R elements are used for the skin, L-pieces and Z-pin, orientations and material behaviors. The detailed material parameters
while the C3D6 elements are used for the deltoid region. Zero-thickness of the IM7/M91 and S35/YH69 unidirectional laminate used in this
COH3D8 elements are employed to simulate interface delamination in study are listed in Appendix G.
the core zone. It is important to note that, to ensure the smooth pro­
gression of the calculations, an element-based cohesive model is 5. Results and discussion
implemented between layers, while a surface-based cohesive model is
utilized in the Z-pin regions. Additionally, one element is designated for The numerical results of low-velocity impact responses and residual
each layer in the thickness direction. To mitigate the occurrence of zero- tensile properties for the two types of CFRP T-joints are compared
energy modes during Abaqus/Explicit simulations, the relaxed stiffness against the experimental data in Ref. [29]. This comparison serves the
hourglass control method was adopted. dual purpose of validating the efficacy of the method for transferring
To validate the mesh sensitivity of the FE models, a comparison is element damage states between two sequential FE models and assessing
made among coarse, middle, and fine mesh models for the unpinned T- their accuracy. It is worth emphasizing that all the detailed repeated test
joint. These models feature element sizes of 1.2 × 2.0 mm, 0.6 × 1.0 results can be found in the appendix of our previous work [29]. In line
mm, and 0.4 × 0.8 mm, respectively, at the central region, as illustrated with the experimental findings, the impact simulation process involves
in Fig. 5(a). The coarse mesh model exhibits a more substantial devia­ recording impact force-time, transferred energy-time, and impact
tion from the experimental curve than the other two models under an 8 J force-central displacement response curves. Additionally, the regions
low-velocity impact, as shown in Fig. 5(b). The maximum ration of the exhibiting delamination damage are delineated. The residual tensile
artificial strain energy to the total impact energy for coarse mesh model, load-displacement response curves are collected and the final failure
middle mesh model and fine mesh model is 13.14 %, 6.62 % and 5.31 %, modes of T-joint and Z-pin are laid out. Furthermore, the bridging effect
respectively, as shown in Fig. 5(c). of Z-pin is investigated thoroughly.
In consideration of computational efficiency, accuracy, and the
geometric characteristics of the model, the element size within the core 5.1. Low-velocity impact response prediction of T-joints
region is primarily set at 0.6 × 1.0 mm, complemented by 1.0 × 1.0 mm
and 1.0 × 1.4 mm. Coarser meshing is applied to other regions. The Previous studies have indicated that Z-pin reinforcement in the non-

Fig. 5. Mesh convergence analysis for unpinned T-joint: (a) coarse mesh, middle mesh and fine mesh models at central region, (b) the impact force-time and (c) the
artificial strain energy-time responses using three models.

6
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

impact region had a limited influence on the impact-related mechanical specimen preparation). However, following impacts of 15 J or 25 J, both
responses of the T-joint [29], and more test results are shown in Fig. 6. In show conspicuous delamination damage, characterized by a distinctive
simpler terms, the mechanical responses of both types of T-joints under "bow tie" pattern. Higher impact energy caused more serious delami­
low-velocity impact conditions exhibit remarkable similarity. To facili­ nation damage. Nonetheless, some disparities between the model and
tate concise presentation, typical curves were selected for a comparative experimental results exist, primarily concerning the extent of the
analysis of mechanical responses. It is essential to highlight that the damaged region. In the experiments, after a 25 J impact, the damage has
impact force underwent a smoothing process through low-pass fast already spread to a portion of the Z-pin reinforcement region, whereas
Fourier transform, with a cut-off frequency of 15 Hz. This specific fre­ this is not replicated in the numerical results. This divergence is likely
quency was chosen to effectively filter out force oscillations introduced attributed to initial defects arising during specimen preparation, which
by the natural frequency inherent in the experimental apparatus setup. represents the most significant distinction between the experiment and
When examining the mechanical response to low-velocity impact, a simulation at present.
commonly employed research approach involves a quantitative analysis
of key parameters, as these parameters offer effective insights into the 5.2. Residual tensile failure prediction of T-joints
specimen’s impact resistance from various perspectives. Specifically, the
Hertzian failure load is often considered the pivotal threshold value for Displayed in Fig. 9 are the experimental outcomes alongside corre­
the initiation of extensive delamination damage within the laminate sponding numerical results, depicting typical residual tensile load-
[35]. The maximum load typically exhibits a positive correlation with displacement response curves for T-joints subjected to varying energy
impact energy [36], the time duration captures the entire contact pro­ impacts. The experimental data reveal that the tensile stiffness of the
cess between the impactor and the specimen, the peak energy moment intact specimen, i.e., following a 0 J impact, remains virtually un­
provides a direct reflection of the structural stiffness of the specimen changed until nearing the ultimate load (Fig. 9(a)(e)). After an 8 J
[29], the final energy accounts for the total energy dissipated by the impact, the specimen only experiences minor tensile stiffness degrada­
specimen [37], the maximum displacement offers insights into the tion as the load approaches the ultimate value (Fig. 9(b)(f)). However,
elastic characteristics of the structure, and the final displacement char­ specimens subjected to 15 J and 25 J impacts exhibit pronounced
acterizes the extent of damage incurred by the specimen [38]. In addi­ stiffness degradation during the later stages of loading (Fig. 9(c)(d)(g)
tion, fluctuations in load serve as indicators of the intensity of internal (h)). These phenomena can primarily be attributed to disparities in the
damage evolution within the specimen. Fig. 7(a)~(c), (d)~(f), (g)~(i) extent of internal damage incurred by the specimens following low-
present typical experimental data alongside corresponding numerical velocity impact.
results for impact force-time response curves, transfer energy-time Notably, all Z-pinned T-joints subjected to 15 J and 25 J impacts
response curves, and impact force-central displacement response exhibit two discernible load recovery stages following the attainment of
curves for T-joints subjected to low-velocity impact at energy levels of 8 maximum tensile load. Subsequently, the load stabilizes for a period
J, 15 J, and 25 J, respectively. Both sets of data show high degree of after reaching near the regional ultimate load (Fig. 9(g)(h)). This
consistency in overall fluctuation patterns and key parameters. The behavior arises from the fact that the tensile load is insufficient to induce
differences between the average experimental results and corresponding instantaneous failure of the remaining region, given the significant and
numerical results for Hertzian failure load (Fh), maximum load (Fmax), progressive damage within the specimen. The propagation of cracks is
time duration (dur), peak energy moment (PEM), final energy (Efin), temporarily impeded when they reach the Z-pin reinforcement region,
maximum displacement (Dmax) and final displacement (Dfin) are as resulting in the load recovery stage observed in the curve. Meanwhile,
follow: 47.28 % (Fh), 4.19 % (Fmax), 3.70 % (dur), 0.50 % (PEM), 5.18 % stress concentration at the crack tip continues to intensify. This stress
(Efin), 1.69 % (Dmax), and 10.41 % (Dfin). Notably, the majority of these concentration induces ongoing matrix damage before Z-pin fracture or
differences fall within a reasonable range, with the exception of a pro­ pull-out, owing to the relatively weaker strength of the resin. This
nounced difference in Hertzian failure load, which can be attributed to phenomenon is reflected in the load stabilization stage of the curve.
its relatively small value. Ultimately, once the stress reaches the limit necessary for Z-pin fracture
Conversely, the bonding interface stands out as the weakest segment or pull-out, Z-pins near the deltoid region are the first to fail, leading to
within the entirety of the T-joint, experiencing the most severe delam­ local stress redistribution. Consequently, multiple stages of load recov­
ination damage following low-velocity impact. In prior investigations, ery and stabilization in tensile load occur.
non-destructive testing was performed on this region using ultrasonic C- The simulation results align closely with the experimental observa­
scan equipment. Therefore, the damage outcomes pertaining to the tions mentioned earlier. Specifically, the differences between the
bonding interface within the FE model are extracted and juxtaposed average experimental results and their corresponding numerical coun­
against the experimental findings, as shown in Fig. 8. The comparative terparts regarding the residual ultimate tensile load of both unpinned T-
analysis reveals a high degree of similarity between the damage states joint (un) and Z-pinned T-joint (Z) following 0 J, 8 J, 15 J, and 25 J
observed in both the model and the experiments. Notably, after 8 J impacts are as follows: 8.92 % (un), 6.73 % (un), 9.77 % (un), 8.08 %
impact, both exhibit minimal damage (with the slight damage in the (un), 3.14 % (Z), 6.04 % (Z), 9.98 % (Z), and 10.30 % (Z), respectively. It
experimental results attributed to unavoidable initial defects during is worth noting that the simulation results also exhibit a notable load

Fig. 6. Test results of unpinned and Z-pinned T-joints under low-velocity impact of (a) 8 J, (b) 15 J, and (c) 25 J.

7
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

Fig. 7. Typical experimental and numerical results for (a)~(c) impact force-time response curves, (d)~(f) transfer energy-time response curves, and (g)~(i) impact
force-central displacement response curves for T-joints during low-velocity impact at 8 J, 15 J, and 25 J energy levels, respectively.

recovery phenomenon. Although the extent of load fluctuation during Consequently, during the post-impact tensile process, these delaminated
this stage is considerably greater than what is observed in the experi­ regions, in addition to the inherently fragile bonding interface, undergo
mental data (and lacks a distinct load stabilization stage), and the further extension of delamination damage. The certain differences do
starting positions of multiple load recovery stages are not well distin­ exist between the two types of failure phenomena observed in all
guished, the phenomenon within the numerical results strongly supports experiments.
the effectiveness of Z-pin reinforcement in T-joints that have undergone In both types, the deltoid is completely separated from the skin, a
substantial damage. result consistent with experimental findings. However, simulation yields
On the other hand, all test specimens, encompassing both intact two distinct final positions for the deltoid. In the first scenario, the
specimens and those that have sustained damage, manifest two distinct simulation aligns with the experimental results, while in the second
types of failure occurrences. The first type entails the observation of scenario, following the complete pull-off failure of the T-joint, the del­
damage extension solely within specific bonding interfaces, more pre­ toid remains adhered to the skin. Additionally, most experiments
cisely, between the deltoid and one of the fillet regions, and between the generate cracks near a specific tip of the deltoid (Fig. 10(c)), whereas the
skin and both the flange region and the deltoid, as shown in Fig. 10(a). simulation does not produce similar outcomes. It is highly likely that
The second type of failure involves the extension of damage within the these two distinct phenomena share similar underlying causes. This
fillet regions in addition to the same regions as the first type, as shown in stems from the fact that during the actual preparation process, the del­
Fig. 10(c). The FE simulation also replicates these two distinct failure toid is formed by gradually filling unidirectional prepreg material,
phenomena, as illustrated in Fig. 10(b)(d). inevitably resulting in the formation of voids, particularly in regions
Upon comparison, it becomes evident that the failure mode of the T- adjacent to the skin. The so-called "interlaminar" performance in this
joint obtained through simulation closely aligns with the failure mode region is considerably weaker than the genuine interlaminar perfor­
observed experimentally, exhibiting only minor discrepancies. Both mance between layers. However, the FE model has yet considered these
types of failure phenomena primarily occur during the pull-off process of discrepancies.
intact specimens and specimens subjected to an 8 J impact. Meanwhile, Lastly, the FE model faithfully reproduces the pull-off failure phe­
the second type of failure phenomenon primarily emerges during the nomenon of Z-pins in Z-pinned T-joints, as shown in Fig. 10(h) (cross-
pull-off process of specimens subjected to 15 J or 25 J impacts. An ex­ sectional view). The outcomes reveal that nearly all Z-pins experience
amination of the cross-sectional damage state within the FE model re­ failure through pull-out mechanisms throughout the entire process.
veals that specimens subjected to 15 J and 25 J impacts exhibit more Only a minority of Z-pins incur fiber and matrix damage at the end
pronounced delamination damage across multiple layers, including where being pulled out, and the locations of the pull-out events
within the fillet regions (Fig. 10(f)(g)), whereas specimens subjected to consistently occur within the flange regions. This phenomenon arises
an 8 J impact display only limited damage in these regions (Fig. 10(e)). from the fact that the skin’s thickness is four times that of the L-pieces,

8
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

Fig. 8. Comparison of experimental and simulation results of delamination damage region of T-joints after low-velocity impact at different energy levels.

endowing the skin with greater capacity to exert constraint force during After 0 J and 8 J energy impacts, the fluctuations are nearly synchro­
the load-bearing process of the Z-pin. nous, whereas after 15 J and 25 J energy impacts, they occur in clusters.
The former exhibits a notably swifter fluctuation process, characterized
by markedly shortened durations of Stages 2, 3, and 4.
5.3. Bridging effect and failure mechanism of Z-pin This distinction arises primarily from the impact of 8 J energy on the
T-joint, which incurs minimal damage, akin to that of an intact T-joint.
Given that Z-pin exhibits distinct strengthening effects on CFRP T- In this scenario, progressive damage is scarcely observable as the tensile
joints depending on the extent of damage, delving deeper into its un­ load nears its limit, thus accumulating a significant internal stress. The
derlying mechanism necessitates an analysis from the perspective of the damage propagates rapidly once the stress reaches the critical failure
Z-pin’s constraining influence on the layers. Sections 5.1 and 5.2 notably threshold, and the separation strength between the layers in the region
validate the element damage state transmission method employed in reinforced by Z-pins far surpasses the bonding and bridging strength.
this study and affirm the precision of the FE model. Consequently, this Consequently, the failure of the Z-pin reinforced region occurs nearly
model serves as a valuable tool for conducting the pertinent analysis. simultaneously with the failure of other regions, resulting in the
Research findings [39] underscore that the bridging effect, specif­ simultaneous pull-out of these three rows of Z-pins.
ically bridging tractions, stands as the pivotal factor by which Z-pins Conversely, the impacts of 15 J and 25 J energy levels inflict sub­
exert constraint on the layers. These bridging tractions predominantly stantial damage on the T-joint, with internal stress levels insufficient to
arise from the fractional stress interactions between the Z-pin and the reach the point of complete T-joint failure. Instead, this stress is pro­
layers. Therefore, this paper analyzes the variation process of the total gressively released in tandem with the ongoing damage. Consequently,
force due to frictional stress on all Z-pin surfaces throughout the com­ the damage extends along the bonding interface, sequentially traversing
plete pull-off process of the T-joint. The aims to elucidate the impact these three rows of Z-pins, leading to their batch pull-out. Even after the
characteristics of Z-pins at various positions on the residual tensile Z-pins have been pulled out, the equivalent bridging tractions continue
performance within T-joints that exhibit varying degrees of damage. to exhibit some fluctuations (Fig. 11(c)(d)). This phenomenon occurs
As shown in Fig. 11, the Z-pins in the Z-pinned T-joint are catego­ because the subject of measurement is the total force due to frictional
rized into three groups according to their positions: Row 1, Row 2 and stress across the entire Z-pin surface. Even after Z-pins have separated
Row 3. Subsequently, the total force due to frictional stress of all Z-pins from the L-pieces, they maintain contact with the skin, inducing
(26 each) in each group is calculated to determine the equivalent continued oscillation.
bridging tractions for each group. It is important to highlight that, in the case of T-joints subjected to
The results unveil that the fluctuation process of all equivalent 15 J and 25 J energy impacts, a more pronounced drop in equivalent
bridging tractions encompasses four distinct stages: a stable stage (Stage bridging tractions occurs during the pull-out process of the third row of
1), a gradual ascent stage (Stage 2), a rapid ascent stage (Stage 3), and a Z-pins before final failure, as shown in Fig. 11(c)(d). This drop signifies a
swift descent stage (Stage 4). Moreover, the fluctuations in equivalent notable slippage occurring between the Z-pin and the contact surface at
bridging tractions among different groups during the T-joint pull-off this particular moment. The primary reason for this phenomenon is that
process, at various levels of damage, fall into two primary categories.

9
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

Fig. 9. Typical experimental and numerical results of tensile load-displacement response curves of (a)~(d) unpinned T-joints and (e)~(h) Z-pinned T-joints after
different energy impact.

the T-joint, at this juncture, is approaching the state of complete pull-off, (1) Through the utilization of Python scripts to extract the ultimate
leading to a significant reduction in the mutual constraint between the element information from the preceding model and subsequently
skin and L-pieces. Consequently, it becomes susceptible to generating incorporating it into the corresponding elements of the subse­
substantial instantaneous fluctuations during the progressive damage quent model, the model’s damage state can be efficiently trans­
process along the bonding interface. ferred with low loss. Building upon this foundation, the
Furthermore, by examining the initial moments of Stage 2, it can be integrated model, employing VUMAT subroutines for calcula­
inferred that for T-joints subjected to 15 J and 25 J energy impacts, Z- tions, demonstrates precision in simulating the low-velocity
pins at different positions commence bearing loads at distinct time impact behavior of CFRP T-joints and their residual tensile per­
points. This observation indicates that not all Z-pins initiate the bridging formance after impact. This analytical methodology can be
effect simultaneously. This phenomenon lends robust support to the extrapolated to encompass a wide spectrum of FE simulation
hypothesis concerning the load recovery stages outlined in Section 5.2. challenges that necessitate integrated analysis across two or more
In addition, the statistical findings also reveal that the rate of Z-pin models, encompassing scenarios like post-impact compression,
pull-out and its position exert a negligible influence on the maximum post-impact fatigue, repeated impact, and more.
equivalent bridging tractions. This suggests that bridging tractions (2) While Z-pin reinforcement in the non-impact region exerts little
predominantly governed by frictional stress exhibit minimal depen­ influence on the low-velocity impact response of CFRP T-joints
dence on the rate of change. Z-pins undergoing high-velocity pull-out and minimal effect on the residual tensile performance of intact
still supply significant bridging tractions, although their contributions or slightly damaged CFRP T-joints, it proves highly effective in
become inconsequential in the face of substantial interlaminar separa­ curtailing damage propagation within the bonding region when
tion stress. It is only when the interlaminar separation stress approxi­ significant interlaminar damage is induced by low-velocity
mates the magnitude of the bridging tractions that the constraining impact. Specifically, once the CFRP T-joint reaches its ultimate
effect of Z-pins on the layers becomes fully apparent. load and encounters a sudden load reduction (reduced to
approximately 53.5 % of the ultimate load), the Z-pin intervenes
6. Conclusions to prevent complete failure, allowing the structure to regain some
of its load-bearing capacity within a limited range (approxi­
This study investigates the mechanical response of CFRP T-joints mately 12.3 % of the ultimate load can be restored).
through a new integrated modeling method. Through the study, the (3) The location of interlaminar damage resulting from low-velocity
following conclusions can be drawn: impact on CFRP T-joints plays a pivotal role in determining the
number of regions where damage extends during subsequent

10
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

Fig. 10. (a)~(d) Comparison of different failure modes during tensile process, (e)~(g) difference in distribution of delamination damage regions of T-joints after
low-velocity impact, (h) comparison of final failure modes of Z-pin.

Fig. 11. Comparison of the bridging effect of Z-pins at different positions in the pull-off process of T-joint after (a) 0 J (viz., intact), (b) 8 J, (c) 15 J, and (d) 25 J
energy impact.

quasi-static pull-off process. Likewise, manufacturing process (4) In CFRP T-joints exhibiting extensive damage, Z-pins situated at
imperfections significantly influence the precise failure location various positions do not initiate the bridging effect simulta­
of the deltoid in CFRP T-joints. Collectively, these factors govern neously. Conversely, in CFRP T-joints with minor damage, the
the ultimate failure mode of the CFRP T-joint, whether it involves opposite scenario unfolds, with the latter experiencing a notably
crack propagation at a single location or multiple locations. shorter duration than the former. This disparity is primarily
Furthermore, in the case of almost all Z-pins, their final failure determined by the time span encompassing the failure process of
mode involves complete pull-out within the flange region of the the structure. Furthermore, the bridging tractions primarily
L-pieces. This indicates that increasing the thickness of this re­ governed by frictional stress on the Z-pin surface exhibit insen­
gion can enhance overall interlaminar reinforcement. sitivity to both the pull-out rate and the implantation position of

11
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

the Z-pin. They consistently provide nearly identical equivalent interests or personal relationships that could have appeared to influence
bridging tractions. However, this effect becomes discernible only the work reported in this paper.
when the CFRP T-joint harbors a certain degree of initial damage.
Data availability
Author contributions
Data will be made available on request.
Jianwu Zhou: Investigation, Methodology, Validation, Writing-
original draft, Writing-review draft. Acknowledgments
Zhibin Zhao: Validation, Writing-original draft.
Liyong Jia: Investigation, Methodology, Funding acquisition. This research was supported by the National Natural Science Foun­
Chao Zhang: Investigation, Funding acquisition, Resources, dation of China (No. 12172303), the 111 Project (No. BP0719007) and
Supervision. the Aviation Science Foundation of China (No. 2017ZD03020).

Declaration of competing interest

The authors declare that they have no known competing financial

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.compscitech.2023.110316.

Appendix A. Failure modes of Hashin’s failure criterion

Fiber tensile failure (σ11 ≥ 0)


( )2 ( )2 ( )2
σ11 σ12 σ13
fft = +α +α ≥1 (A1)
Xt S12 S13
Fiber compressive failure (σ11 < 0)
( )2
σ11
ffc = ≥1 (A2)
Xc
Matrix tensile failure (σ22 + σ33 ≥ 0)
( ) ( )2 ( )2
σ 22 + σ33 2 σ12 σ 13 σ 2 − σ 22 σ33
fmt = + + + 23 2 ≥1 (A3)
Yt S12 S13 S23
Matrix compressive failure (σ22 + σ33 < 0)
[( )2 ] ( ) ( )2 ( )2
Yc σ 22 + σ 33 σ22 + σ33 2 σ 12 σ13 σ 2 − σ 22 σ 33
fmc = − 1 + + + + 23 2 ≥1 (A4)
2S23 Yc 2S23 S12 S13 S23

where σij (i, j = 1, 2, 3) is the effective stress tensor; Xt and Xc are the tensile and compressive strengths of the unidirectional composite laminate in the
fiber direction, respectively; Yt and Yc are the tensile and compressive strengths in the transverse direction, respectively; S12 , S13 and S23 are shear
strengths under different modes; α (0 < α ≤ 1) is the shear failure coefficient applied to determine the contribution of shear stresses on the fiber tensile
failure. In the present work, it is set as 1.

Appendix B. Stiffness degradation model.

⎡ ⎤
(1 − d1 )2 C11 (1 − d1 )(1 − d2 )C12 (1 − d1 )C13 0 0 0
⎢ (1 − d1 )(1 − d2 )C12 (1 − d2 )2 C22 (1 − d2 )C23 0 0 0 ⎥
⎢ ⎥
⎢ (1 − d1 )C13 (1 − d2 )C23 C33 0 0 0 ⎥
Cd = ⎢


⎥ (B1)
⎢ 0 0 0 (1 − d12 )2 C44 0 0 ⎥
⎣ 0 0 0 0 (1 − d23 )2 C55 0 ⎦
2
0 0 0 0 0 (1 − d13 ) C66

12
J. Zhou et al. Composites Science and Technology 245 (2024) 110316


⎪ 1 − ν23 ν32 ν21 + ν23 ν31

⎪ C11 = , C12 =

⎪ E2 E3 Δ E1 E3 Δ





⎪ ν31 + ν21 ν32 ν32 + ν12 ν31

⎪ C13 = , C23 =

⎪ E1 E2 Δ E1 E2 Δ

1 − ν13 ν31 1 − ν12 ν21 (B2)

⎪ C = , C33 =
⎪ 22
⎪ E1 E3 Δ E1 E2 Δ



⎪ C = G , C = G , C = G12

⎪ 44 23 55 31 66



⎪ 1 − ν ν − ν ν − ν13 ν31 − 2ν12 ν23 ν31

⎩Δ = 12 21 23 32
E1 E2 E3


⎪ ( )( )

⎪ d1 = 1 − t
1 − d11 c
1 − d11

⎪ ( )( )

⎪ t c
⎪ d2 = 1 −
⎪ 1 − d22 1 − d22


⎨ d3 = 0
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (B3)

⎪ d12 = 1 − (1 − d1 )(1 − d2 )

⎪ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅

⎪ d13 = 1 − (1 − d1 )(1 − d3 )



⎪ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅


⎩ d23 = 1 − (1 − d2 )(1 − d3 )

where E1 , E2 and E3 represent the Young’s modulus in the three principal elastic directions, respectively; G12 , G23 and G31 represent the shear modulus
in the three planes, respectively; The Poisson’s ratio νij satisfies the relationship νij /Ej = νji /Ei (i, j = 1, 2, 3 and i ∕
= j) with the Young’s modulus; dtii and
dcii (i = 1, 2) represent the damage variables under tensile and compressive modes in the i direction, respectively. The specific definitions are expressed
in Eqs. (B4).
⎧ ( )
⎪ εtf,1 εt0,1

⎪ d t
= 1 −
⎪ 11 εt − εt
⎪ ε11

⎪ f,1 0,1

⎪ (

⎪ c c )
⎪ c
⎪ ε f,1 ε 0,1

⎪ d11 = c c 1−
⎨ εf,1 − ε0,1 ε11
( ) (B4)

⎪ εtf,2 εt0,2

⎪ d t
= 1 −




22
εtf,2 − εt0,2 ε22



⎪ c ( c )

⎪ ε f,2 ε 0,2

⎩ d22 c
= c 1 −
εf,2 − εc0,2 ε22

where εt0,i and εc0,i (i = 1, 2) represent the equivalent damage initial strain of element under tensile and compressive modes in the i direction,
respectively; εtf,i and εcf,i (i = 1, 2) represent the equivalent failure strain of element under tensile and compressive modes in the i direction, respectively.

Appendix C. Solution of failure strains and limitation of characteristic length.



⎪ t 2Gtic
⎪ εf,i = σt l


(C1)
0,i


⎪ 2Gcic
⎪ c
⎩ εf,i = σc l

0,i



⎪ 2Gt 2Gt

⎪ εtf,i = t ict ≥ εt0,i →lti ≤ t ict (i = 1, 2)

⎪ σ 0,i li σ 0,i ε0,i


(C2)
c
2G 2Gc

⎪ εcf,i = c icc ≥ εc0,i →lci ≤ c icc (i = 1, 2)



σ l
0,i i σ 0,i ε0,i

⎪ { t c}

l = min li , li

where Gtic and Gcic (i = 1, 2) represent the fracture toughness under tensile and compressive modes in the i direction, respectively [40–43]; σ t0,i and
σ c0,i (i = 1, 2) represent the equivalent damage initial stress of element under tensile and compressive modes in the i direction, respectively; l represents
the characteristic length of the element, which can be approximately calculated and obtained by transforming the target element through the char
Length function embedded in VUMAT.

Appendix D. Damage model for the interface.

[ ]2 [ ]2 [ ]2
〈σ n 〉 σt σs
fini = + + (D1)
Nmax Tmax Smax

13
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

where σn , σt and σ s are interlaminar normal stress and two kinds of interlaminar shear stresses under different modes, respectively; Nmax , Tmax and Smax
are interlaminar tensile stress and two kinds of interlaminar shear strength under different modes; fini is the damage initiation variable, and the
cohesive element starts to damage when it equals to 1.
( )
δf δm − δ0m
D = m( f ) (D2)
δm δm − δ0m

where D is the damage evolution variable. When D = 0, the material does not undergo stiffness degradation. When D = 1, the material undergoes
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
complete stiffness degradation. δm = 〈δn 〉2 + δ2t + δ2s , which represents the comprehensive opening displacement. δ0m represents the comprehensive
0
displacement when the material starts to damage. δfm = 2Gequivc /Teff , which represents the equivalent displacement when the material completely
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
2 0
fails. Teff = 〈σ n 〉 + σt + σs , which represents the comprehensive stress. Teff
2 2 represents the comprehensive stress when the material starts to damage.
In addition, the equivalent fracture toughness under the mixed-mode is expressed in Eq. (D3).
( )η
GII + GIII
Gequivc = GIc + (GIIc − GIc ) (D3)
GI + GII + GIII

where GIc = 850 J/m2 and GIIc = 1450 J/m2 is the fracture toughness under the mode I and II loading, respectively; GI , GII and GIII is the strain energy
release rate under the mode I, II and III loading, respectively; η is the damage factor, it is set as 1.45.

Appendix E. Correction of equivalent damage initiation strain and strength under different failure modes.

Failure mode Longitudinal tension Longitudinal compression

Failure factor (fiber) fft ffc


√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Equivalent strain (fiber)
ε = 〈ε11 〉2 + αε212 + αε213 ε = 〈− ε11 〉2
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Modified equivalent damage initiation strain (fiber)
〈ε11 〉2 + αε212 + αε213 〈− ε11 〉2
ε0 = √̅̅̅̅̅ ε0 = √̅̅̅̅̅
fft ffc
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Modified equivalent damage initiation strength (fiber)
〈σ11 〉2 + ασ212 + ασ213 〈− σ11 〉2
σ0 = √̅̅̅̅̅ σ0 = √̅̅̅̅̅
fft ffc
Failure factor (matrix) fmt fmc
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Equivalent strain (matrix)
ε = 〈ε22 〉2 + 〈ε33 〉2 + ε212 + ε213 + ε223 ε = 〈− ε22 〉2 + 〈− ε33 〉2 + ε212 + ε213 + ε223
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Modified equivalent damage initiation strain (matrix)
〈ε22 〉2 + 〈ε33 〉2 + ε212 + ε213 + ε223 〈− ε22 〉2 + 〈− ε33 〉2 + ε212 + ε213 + ε223
ε0 = √̅̅̅̅̅̅ ε0 = √̅̅̅̅̅̅̅
fmt fmc
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Modified equivalent damage initiation strength (matrix)
〈σ22 〉2 + 〈σ33 〉2 + σ212 + σ213 + σ223 〈− σ22 〉2 + 〈− σ33 〉2 + σ212 + σ213 + σ223
σ0 = √̅̅̅̅̅̅ σ0 = √̅̅̅̅̅̅̅
fmt fmc

Note. The symbol˂> represents the Macaulay operator.

Appendix FLow-velocity impact test arrangements of T-joint specimens.

Test Impact energy (J) Repeated times Drop height (mm) Initial contact velocity (m/s)

Unpinned T-joint-8J 8 3 146.72 ± 0.38 1.70 ± 0.00


Z-pinned T-joint-8J 146.52 ± 0.88 1.69 ± 0.00
Unpinned T-joint-15J 15 272.20 ± 0.53 2.31 ± 0.00
Z-pinned T-joint-15J 271.90 ± 0.52 2.31 ± 0.00
Unpinned T-joint-25J 25 459.23 ± 0.73 3.00 ± 0.00
Z-pinned T-joint-25J 457.67 ± 0.18 2.99 ± 0.00

Appendix G. Material properties of unidirectional laminate of IM7/M91 and S35/YH69.

Material Category Parameter

IM7/M91 Density 1570 kg/m3


Stiffness properties E1 = 165 GPa; E2 = E3 = 10.3 GPa; ν12 = ν13 = 0.3; ν23 = 0.4
G12 = 6.3 GPa; G13 = 6 GPa; G23 = 3.7 GPa
Strength properties Xt = 2980 MPa; Xc = 1860 MPa; Yt = Zt = 82 MPa; Yc = Zc = 236 MPa
S12 = 110 MPa; S13 = 90 MPa; S23 = 40 MPa
Fracture energy Gft = 91.6 N/mm; Gfc = 79.9 N/mm; Gmt = 0.22 N/mm; Gmc = 1.1 N/mm
S35/YH69 Density 1870 kg/m3
Stiffness properties E1 = 115 GPa; E2 = E3 = 8.2 GPa; ν12 = ν13 = 0.32; ν23 = 0.41
G12 = G13 = 3.5 GPa; G23 = 2.4 GPa
Strength properties Xt = 1090 MPa; Xc = 760 MPa; Yt = Zt = 80 MPa; Yc = Zc = 145 MPa
S12 = 105 MPa; S13 = 95 MPa; S23 = 36 MPa
Fracture energy Gft = 62.1 N/mm; Gfc = 49.7 N/mm; Gmt = 0.17 N/mm; Gmc = 1.2 N/mm

14
J. Zhou et al. Composites Science and Technology 245 (2024) 110316

References [22] H. Liu, B.G. Falzon, W. Tan, Predicting the Compression-After-Impact (CAI)
strength of damage-tolerant hybrid unidirectional/woven carbon-fibre reinforced
composite laminates, Compos. Part A-Appl. 105 (2018) 189–202, https://doi.org/
[1] H. Liu, J. Liu, Y. Ding, Z.E. Hall, X. Kong, J. Zhou, et al., A three-dimensional
10.1016/j.compositesa.2017.11.021.
elastic-plastic damage model for predicting the impact behaviour of fibre-
[23] H. Tuo, Z. Lu, X. Ma, J. Xing, C. Zhang, Damage and failure mechanism of thin
reinforced polymer-matrix composites, Compos. Part B-Eng. 201 (2020), 108389,
composite laminates under low-velocity impact and compression-after-impact
https://doi.org/10.1016/j.compositesb.2020.108389.
loading conditions, Compos. Part B-Eng. 163 (2019) 642–654, https://doi.org/
[2] S. Yan, X. Zeng, A. Long, Meso-scale modelling of 3D woven composite T-joints
10.1016/j.compositesb.2019.01.006.
with weave variations, Compos. Sci. Technol. 171 (2019) 171–179, https://doi.
[24] H. Tuo, Z. Lu, X. Ma, C. Zhang, S. Chen, An experimental and numerical
org/10.1016/j.compscitech.2018.12.024.
investigation on low-velocity impact damage and compression-after-impact
[3] H. Zhang, R. He, B. Hou, Y. Li, H. Cui, W. Yang, Artificial hail ice impact damage of
behavior of composite laminates, Compos. Part B-Eng. 167 (2019) 329–341,
laminated composite T-joint with stitching reinforcement, Compos. Struct. 278
https://doi.org/10.1016/j.compositesb.2018.12.043.
(2021), 114714, https://doi.org/10.1016/j.compstruct.2021.114714.
[25] W. Jia, W. Wen, L. Fang, Low-velocity impact and post-impact biaxial residual
[4] S. Yan, X. Zeng, A. Long, Experimental assessment of the mechanical behaviour of
strength tests and simulations of composite laminates, Compos. Struct. 235 (2020),
3D woven composite T-joints, Compos. Part B-Eng. 154 (2018) 108–113, https://
111758, https://doi.org/10.1016/j.compstruct.2019.111758.
doi.org/10.1016/j.compositesb.2018.08.007.
[26] Z. Cheng, J. Xiong, Progressive damage behaviors of woven composite laminates
[5] J. Zhou, Z. Zhao, Y. Shi, X. Wang, X. Chen, C. Shan, Hail impact responses and
subjected to LVI, TAI and CAI, Chin. J. Aeronaut. 33 (10) (2020) 2807–2823,
residual tensile properties of CFRP T-joints, Chin. J. Aeronaut. (2023),
https://doi.org/10.1016/j.cja.2019.12.015.
S1000936123000614, https://doi.org/10.1016/j.cja.2023.03.005.
[27] Z. Cheng, W. Tan, J. Xiong, Progressive damage modelling and fatigue life
[6] M.M. Thawre, K.N. Pandey, A. Dubey, K.K. Verma, D.R. Peshwe, R.K. Paretkar, et
prediction of Plain-weave composite laminates with Low-velocity impact damage,
al., Fatigue life of a carbon fiber composite T-joint under a standard fighter aircraft
Compos. Struct. 273 (2021), 114262, https://doi.org/10.1016/j.
spectrum load sequence, Compos. Struct. 127 (2015) 260–266, https://doi.org/
compstruct.2021.114262.
10.1016/j.compstruct.2015.03.026.
[28] Z. Cheng, W. Tan, J. Xiong, Modelling pre-fatigue, low-velocity impact and post-
[7] J.-B. Bai, C.-H. Dong, J.-J. Xiong, C.-Y. Luo, D. Chen, Progressive damage
impact fatigue behaviours of composite helicopter tail structures under multipoint
behaviour of RTM-made composite T-joint under tensile loading, Compos. Part B-
coordinated loading spectrum, Thin-Walled Struct. 176 (2022), 109349, https://
Eng. 160 (2019) 488–497, https://doi.org/10.1016/j.compositesb.2018.12.069.
doi.org/10.1016/j.tws.2022.109349.
[8] M. Barzegar, M.D. Moallem, M. Mokhtari, Progressive damage analysis of an
[29] J. Zhou, Y. Shi, Y. Zuo, C. Shan, Z. Gu, Experimental investigation into influences
adhesively bonded composite T-joint under bending, considering micro-scale
of Z-pin and deltoid on structural properties and damage tolerance of CFRP T-
effects of fiber volume fraction of adherends, Compos. Struct. 258 (2021), 113374,
joints, Compos. Part B-Eng. 237 (2022), 109875, https://doi.org/10.1016/j.
https://doi.org/10.1016/j.compstruct.2020.113374.
compositesb.2022.109875.
[9] J. Cheng, Y. Xu, W. Zhang, W. Liu, A review on the multi-scale simulation of Z-
[30] Z. Hashin, Fatigue failure criteria for unidirectional fiber composites, J. Appl.
pinned composite laminates, Compos. Struct. 295 (2022), 115834, https://doi.org/
Mech. 48 (4) (1981) 846–852, https://doi.org/10.1115/1.3157744.
10.1016/j.compstruct.2022.115834.
[31] A. Faggiani, B.G. Falzon, Predicting low-velocity impact damage on a stiffened
[10] B. Gong, W. Ouyang, M. Nartey, H. Wang, K.D. Potter, H.-X. Peng, Minimizing the
composite panel, Compos. Part A-Appl. 41 (6) (2010) 737–749, https://doi.org/
in-plane damage of Z-pinned composite laminates via a pre-hole pin insertion
10.1016/j.compositesa.2010.02.005.
process, Compos. Sci. Technol. 200 (2020), 108413, https://doi.org/10.1016/j.
[32] A. Hillerborg, The theoretical basis of a method to determine the fracture energyG
compscitech.2020.108413.
F of concrete, Mater. Struct. 18 (4) (1985) 291–296, https://doi.org/10.1007/
[11] B. Liao, J. Zhou, J. Zheng, R. Tao, L. Xi, T. Zhao, et al., Effect of Z-pin inclined angle
BF02472919.
on the impact damage suppression effectiveness for cross-ply composite laminates,
[33] M.L. Benzeggagh, M. Kenane, Measurement of mixed-mode delamination fracture
Compos. Part A-Appl. 142 (2021), 106264, https://doi.org/10.1016/j.
toughness of unidirectional glass/epoxy composites with mixed-mode bending
compositesa.2020.106264.
apparatus, Compos. Sci. Technol. 56 (4) (1996) 439–449, https://doi.org/
[12] B. Gong, H. Wang, M. Nartey, W. Ouyang, J. Ye, S. Qu, et al., Fatigue life
10.1016/0266-3538(96)00005-X.
assessment and damage evolution in Z-pinned laminates, Compos. Sci. Technol.
[34] P.P. Camanho, C.G. Davila, M.F. de Moura, Numerical simulation of mixed-mode
221 (2022), 109328, https://doi.org/10.1016/j.compscitech.2022.109328.
progressive delamination in composite materials, Materials 37 (16) (2003)
[13] L. Francesconi, F. Aymerich, Damage mechanisms in the CAI failure of thin z-
1415–1438, https://doi.org/10.1177/0021998303034505.
pinned composite laminates, Compos. Part A-Appl. 158 (2022), 106991, https://
[35] Y. Shi, T. Swait, C. Soutis, Modelling damage evolution in composite laminates
doi.org/10.1016/j.compositesa.2022.106991.
subjected to low velocity impact, Compos. Struct. 94 (9) (2012) 2902–2913,
[14] H. Zhang, H. Ding, D. Yang, Q. Xu, Y. Ma, Y. Bi, Experimental investigation of Z-pin
https://doi.org/10.1016/j.compstruct.2012.03.039.
insertion angles on the mechanical behavior and failure mechanism of compression
[36] J. Zhou, B. Liao, Y. Shi, Y. Zuo, H. Tuo, L. Jia, Low-velocity impact behavior and
after impact for CFRP laminates, Compos. Struct. 304 (2023), 116476, https://doi.
residual tensile strength of CFRP laminates, Compos. Part B-Eng. 161 (2019)
org/10.1016/j.compstruct.2022.116476.
300–313, https://doi.org/10.1016/j.compositesb.2018.10.090.
[15] T.M. Koh, S. Feih, A.P. Mouritz, Experimental determination of the structural
[37] X.K. Li, P.F. Liu, Experimental analysis of low-velocity impact behaviors of carbon
properties and strengthening mechanisms of z-pinned composite T-joints, Compos.
fiber composite laminates, J. Fail. Anal. Prev. 161 (2017) 1126–1130, https://doi.
Struct. 93 (9) (2011) 2222–2230, https://doi.org/10.1016/j.
org/10.1007/s11668-017-0350-z.
compstruct.2011.03.009.
[38] J. Zhou, B. Liao, Y. Shi, L. Ning, Y. Zuo, L. Jia, Experimental investigation of the
[16] J. Hoffmann, G. Scharr, Mechanical properties of composite laminates reinforced
double impact position effect on the mechanical behavior of low-velocity impact in
with rectangular z-pins in monotonic and cyclic tension, Compos. Part A-Appl. 109
CFRP laminates, Compos. Part B-Eng. 193 (2020), 108020, https://doi.org/
(2018) 163–170, https://doi.org/10.1016/j.compositesa.2018.03.004.
10.1016/j.compositesb.2020.108020.
[17] S.Z.H. Shah, S. Karuppanan, P.S.M. Megat-Yusoff, Z. Sajid, Impact resistance and
[39] F. Bianchi, T.M. Koh, X. Zhang, I.K. Partridge, A.P. Mouritz, Finite element
damage tolerance of fiber reinforced composites: a review, Compos. Struct. 217
modelling of z-pinned composite T-joints, Compos. Sci. Technol. 73 (2012) 48–56,
(2019) 100–121, https://doi.org/10.1016/j.compstruct.2019.03.021.
https://doi.org/10.1016/j.compscitech.2012.09.008.
[18] P. Rozylo, H. Debski, T. Kubiak, A model of low-velocity impact damage of
[40] I. Lapczyk, J.A. Hurtado, Progressive damage modeling in fiber-reinforced
composite plates subjected to Compression-After-Impact (CAI) testing, Compos.
materials, Compos. Part A-Appl. 38 (11) (2007) 2333–2341, https://doi.org/
Struct. 181 (2017) 158–170, https://doi.org/10.1016/j.compstruct.2017.08.097.
10.1016/j.compositesa.2007.01.017.
[19] A.H. Baluch, O. Falcó, J.L. Jiménez, B.H.A.H. Tijs, C.S. Lopes, An efficient
[41] F. Guo-dong, L. Jun, W. Bao-lai, Progressive damage and nonlinear analysis of 3D
numerical approach to the prediction of laminate tolerance to Barely Visible
four-directional braided composites under unidirectional tension, Compos. Struct.
Impact Damage, Compos. Struct. 225 (2019), 111017, https://doi.org/10.1016/j.
89 (2009) 126–133, https://doi.org/10.1016/j.compstruct.2008.07.016.
compstruct.2019.111017.
[42] R. He, Y. Gao, L. Cheng, H. Cui, Y. Li, Fracture toughness for longitudinal
[20] T. Ouyang, R. Bao, W. Sun, Z. Guan, R. Tan, A fast and efficient numerical
compression failure of laminated composites at high loading rate, Compos. Part A-
prediction of compression after impact (CAI) strength of composite laminates and
Appl. 156 (2022), 106834, https://doi.org/10.1016/j.compositesa.2022.106834.
structures, Thin-Walled Struct. 148 (2020), 106588, https://doi.org/10.1016/j.
[43] Y. Zhao, J. Zang, J. Ben, J. Yang, X. Wan, W. Wang, X. Chen, Revealing the
tws.2019.106588.
interlaminar shear failure behavior of unidirectional laminate under combined
[21] J. Zou, Z. Lei, D. Liu, X. Hong, Y. Shang, R. Bai, Rapid prediction of compression
compression-shear loads, J. Mater. Sci. Technol. 157 (2023) 110–119, https://doi.
after impact properties of composite structures: an equivalent strategy for impact
org/10.1016/j.jmst.2023.01.049.
damage, Compos. Struct. 307 (2023), 116655, https://doi.org/10.1016/j.
compstruct.2022.116655.

15

You might also like