Yang et al.

Chinese Journal of Mechanical Engineering

Chinese Journal of Mechanical Engineering (2023) 36:10
https://doi.org/10.1186/s10033-023-00839-z

ORIGINAL ARTICLE Open Access

Progressive Failure Analysis of Composite/

Aluminum Riveted Joints Subjected
to Pull‑Through Loading
Yuxing Yang1, Yongjie Bao1*, Xueshu Liu2, Jinlong Wang1 and Fengming Du1

Abstract
Out-of-plane mechanical properties of the riveted joints restrict the performance of the wing box assembly of air-
plane. It is necessary to investigate the pull-through performance of the composite/metal riveted joints in order to
guide the riveting design and ensure the safety of the wing box assembly. The progressive failure mechanism of com-
posite/aluminum riveted joint subjected to pull-through loading was investigated by experiments and finite element
method. A progressive damage model based on the Hashin-type criteria and zero-thickness cohesive zone method
was developed by VUMAT subroutine, which was validated by both open-hole tensile test and three-point bending
test. Predicted load-displacement response, failure modes and damage propagation were analysed and compared
with the results of the pull-through tests. There are 4 obvious characteristic stages on the load-displacement curve
of the pull-through test and that of the finite element model: first load take-up stage, damage stage, second load
take-up stage and failure stage. Relative error of stiffness, first load peak and second load peak between finite element
method and experiments were 8.1%, − 3.3% and 10.6%, respectively. It was found that the specimen was mainly
broken by rivet-penetration fracture and delamination of plies of the composite laminate. And the material within the
scope of the rivet head is more dangerous with more serious tensile damages than other regions, especially for 90°
plies. This study proposes a numerical method for damage prediction and reveals the progressive failure mechanism
of the hybrid material riveted joints subjected to the pull-through loading.
Keywords Composite/metal joints, Riveted joints, Pull-through test, Progressive damage model

1 Introduction superior resistance to fatigue degradation [3–6]. Nowa-

With the continuous improvement of energy-saving, days, more and more complicated composite parts are
emission-reduction and lightweight requirements around designed and applied in aerospace industry and marine
the world, composite materials have gradually become industry [7, 8], which makes it a big problem to guarantee
one of the major industrial manufacturing materials [1, connecting performance of the joints during wing box
2], which has the advantages of light weight, high stiff- assembling for airplane. Riveting is one of the widely-
ness-to-weight ratio, high strength-to-weight ratio and used connecting methods in composite components
assembly for airplane [9–12], which has the advantages
of stable and reliable connection performance, easy auto-
*Correspondence: mation, wide application range, high reproducibility and
Yongjie Bao fast joining processes [13–15]. As has been reported,
[email protected]
1
Marine Engineering College, Dalian Maritime University, Dalian 116026,
Italian-based BBA Srl has manufactured a unique rivet-
China ing method for carbon fiber composite materials, Bulge
2
School of Automotive Engineering, Dalian University of Technology, Control Technology, which can be used where the hole
Dalian 116024, China
in the workpiece is up to 0.5 mm greater than the rivet

© The Author(s) 2023, corrected publication 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0
International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you
give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes
were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To
view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 2 of 10

nut external size [16]. However, riveting requires hole the 0° unidirectional laminate for glass fiber reinforced
drilling so that reducing the strength of the connected aluminum alloy laminates.
parts, especially for composite laminates [17, 18]. Mean- In this paper, a composite/aluminum riveted joints sub-
while, mechanical properties of riveting, especially static jected to pull-through loading was studied. Both experi-
strength and fatigue life, is influenced by both geometri- ment and finite element method (FEM) were used to
cal features, die pressure, shape [19], etc. Therefore, it is analyse the failure mechanism in terms of load-displace-
necessary to investigate the bearing performance of the ment response, failure modes and damage propagation.
riveted joints in order to guide the riveting design and
ensure the safety of the joints. 2 Experiment
A lot of research works have been done about com- To investigate the riveting performance of compos-
posite riveting or composite/metal riveting, including ite skin and aluminum rib of airplane wing box, a pull-
joints load distribution and joint damage prediction, etc. through test of the composite/aluminum riveted joints
Sathiya et al. [20, 21] investigated a composite/metal lap was conducted according to ASTM D7332 standard
joints for the rivet load distribution and life estimation. [28]. As shown in Figure 1, the specimen was com-
It was found that the load shared by the rivet rows in a posed of a carbon fiber/epoxy composite laminate
composite/metal lap joints was not symmetric and there- (60 mm × 20 mm × 3.0 mm), a 2A12 aluminum alloy
fore are more susceptible to cracking and subsequent plate (60 mm × 20 mm × 4.0 mm) and a Ti6Al4V tita-
failure as the unequal distribution can cause some of the nium alloy ring groove rivet (diameter d = 4.0 mm). The
rivet loads to be high. Solmaz et al. [22] used the pro- composite laminate, whose volume percentage of the
gressive damage model to analyse the woven-type glass carbon fiber was 60%, has 24 plies with symmetric lay-up
fiber composite riveted joints and found that the fail- [45/− 45/90/90/0/90/0/90/90/− 45/45/90]s that yielded
ure occurring in the composite plates began around the laminate thickness of 3.0 mm (single layer nominal thick-
rivet hole and the catastrophic failure of these types of ness was 0.125 mm). The mechanical properties of the
joints resulted from fiber tensile failure. Pramanik et al. composite laminate, cohesive interface between plies,
[19] claimed that placing composite sheet on the top of 2A12 aluminum alloy and Ti6Al4V titanium alloy were
the aluminum alloy sheet during the riveting formation given in Tables 1, 2, 3. Kn, Ks and Kt are slope of the bilin-
was good to avoid crack growth. It was reported in Rao’s ear traction-separation response for the cohesive ele-
paper [23] that there were two distinctive failure modes ment. τn0, τs0 and τt0 are strength of the cohesive element.
for static loading and cyclic loading: (1) the lap-shear and GΙC, GΙΙC and GΙΙΙC are fracture toughness.
cross-tension joints failed due to rivet pullout of the bot- The composite/aluminum laminated drilling process
tom aluminum sheet in quasi-static loading; (2) in cyclic was conducted on a 3-axis CNC machine tool with car-
loading, the lap-shear joints failed due to kinked crack bide twist drill. After drilling, the specimens with mid-
growth in the bottom aluminum sheet, while the cross- dle hole diameter of 4.0 mm (tolerance 0–0.03 mm) were
tension joints failed due to rivet pulling out of the top selected and detected by the HCT-800 A-scan ultrasonic
carbon fiber reinforced polymer sheet. damage detector (Onend, China, resolution > 45 dB, fre-
Pull-through performance is an important property for quency 0.4–20 MHz, scanning range 0–10 m) to ensure
rivet joints, which focus on the through-thickness stiff- no delamination in the laminates.
ness and strength of the joints. Catalanotti et al. [24]
conducted an experimental and numerical study of the
pull-through damage in glass–fiber reinforced plastic Wing box Riveted joints
laminates and concluded that: (1) the type of resin did rivet gun

not affect the mechanical response of the joints when a rivet

sleeve
pull-through test was performed; (2) delamination of the Φ4
aluminum
composite
plies was the main failure mechanism. Gray and McCa-
rthy [25] developed an analytical bolt tension model for Φ6
through-thickness stiffness prediction of the composite
Typical structure
joints based on Rize-approximation method, which can Φ6
composite skin
predict stiffness well but cannot analyse the stress. Ma
[26] studied the effects of temperature, humidity, thick- 3
ness, fixture clamp length and bolt head diameter on the metal rib joint
unit: mm
pull-through strength of the braided composite lami-
nates. Liu [27] found that pull-through ultimate load of Figure 1 Composite/aluminum riveted joints specimen of airplane
the ± 45° cross-ply laminate was 25% higher than that of wing box
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 3 of 10

Table 1 Mechanical properties of the carbon fiber/epoxy

composite laminate [29, 30] Fixture

Mechanical property Value Unit

Longitudinal modulus E1 130 GPa

Transverse modulus E2 (E3) 7.64 GPa Specimen
Shear modulus G12 (G13) 3.7 GPa
Shear modulus G23 2.65* GPa
Poisson’s ratio ν12 (ν13) 0.32 –
Poisson’s ratio ν23 0.44* – 1 8
2 7
Longitudinal tensile strength XT 1750 MPa
3
Longitudinal compressive strength XC 1850 MPa 4
Transverse tensile strength YT (ZT) 80 MPa 5
Transverse compressive strength YC (ZC) 150 MPa 6
Shear strength S 112 MPa
Fixture diagram
Density ρ 1.69 g/cm3
Figure 2 Experiment set-up of the pull-through test for composite/
*The value is set according to the transversely isotropic assumption and aluminum joints
engineering experience

Table 2 Material properties of the cohesive interface plate (No. 7) was fixed by the lower part of the fixture
Material property Value Unit
assembly, including T-shape plate (No. 5), small plate
(No. 3) and two fasteners. Then the fixture assembly was
Kn 2170 MPa/mm clamped by the clamping chuck of the DNS100 electronic
Ks (Kt) 835 MPa/mm universal testing machine (Sinotest Equipment, China,
τn0 10 MPa maximum allowable load 100 kN, relative error of indica-
τs0 (τt0) 28.6 MPa tion ± 0.5%). Finally, the tensile load was applied in the
GΙC 0.3 N/mm form of displacement load with loading speed of 1.0 mm/
GΙΙC (GΙΙΙC) 0.6 N/mm min. During testing, both displacement and load of the
specimen were recorded by the testing machine. The
damages of the specimen were observed by the digital
microscope VHX-600E (Keyence, Japan, magnification
Table 3 Mechanical properties of the 2A12 aluminum alloy and
Ti6Al4V titanium alloy
20× to 5000×).
Three specimens in total were tested, whose load-dis-
Mechanical property Aluminum Titanium alloy Unit placement curves were shown in Figure 3 with four obvi-
alloy 2A12 Ti6Al4V
ous characteristic stages (first load take-up stage, damage
Elastic modulus E 71.7 108.0 GPa stage, second load take-up stage and failure stage). In Fig-
Poisson’s ratio ν 0.33 0.33 – ure 3, slope of the linear load take-up stage represents the
Density ρ 2.77 4.43 g/cm3 stiffness of the specimen, which is 1256.7 ± 61.3 N/mm
in average (from 500 N to 2000 N). The first load peak
was reached when a continuous crisp cracking sound of
Experiment set-up of the pull-through test reference fiber breaking was emitted and slight rivet sinking was
to ASTM D7332 standard [28] was shown in Figure 2. observed. The first load peak is 2406.4 ± 124.8 N in aver-
Firstly, the composite laminate and aluminum plate were age when displacement was about 2.0 mm. After the first
riveted by the titanium alloy ring groove rivet at the mid- load peak, it came to the damage stage. Fiber damages
dle hole. Secondly, the specimen was fixed to the fixture and delamination of the composite laminate occurred
by fasteners. The fixture was composed by 6 parts, whose close to the interface of two plates, which caused sud-
diagram was shown in Figure 2, including 2 T-shape denly load drop on the load-displacement curves. Rivet
plates (Part No. 1 and No. 5), 4 bolts and 4 nuts (Part sinking produced compressive force around the rivet
No. 2 and No. 6), 4 small plates (Part No. 3 and No. 4). hole, which has positive effect on preventing further frac-
Composite laminate (No. 8) was fixed by the upper part ture of the fibers so that the specimen could still bear
of the fixture assembly, including T-shape plate (No. 1), loads, and it came to the second load take-up stage with
small plate (No. 4) and two fasteners, while aluminum increase in load. Then it reached the second load peak
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 4 of 10

First load Damage Second load Failure (minimum mesh size ranges from 0.3–1.2 mm), con-
take-up stage take-up stage
stage stage sidering the computational cost and results accuracy,
3000
in-plane mesh size within the rivet head area (red cir-
First load peak Exp-1 cle in Figure 4) was set as 0.5 mm for both composite
Exp-2
Second load peak
2500
Exp-3 laminate and aluminum plate, while that of other areas
were 2.0 mm. There were 5 elements through the thick-
2000
ness for aluminum plate and 24 elements through the
thickness for composite laminate. In addition, insert-
Load (N)

1500
ing a cohesive layer between every two layers of com-
posite elements (totally 11 cohesive layers, as shown in
1000
Figure 4). Rivet was modelled as a one-piece part with
global element size of 0.5 mm.
500
For convenience and convergence, rigid body method
was used to apply boundary conditions. One circle ele-
0
0 1 2 3 4 5 6 7 ments within the fastener head of two loading holes
Displacement (mm) of the composite laminate were constrained as a rigid
Figure 3 Load-displacement response of the pull-through test for body related to a reference point RPc, while that of the
composite/aluminum joints
metal plate were constrained as another rigid body
related to another reference point RPm. Velocity load
was applied to the reference point RPc, which has no
Rigid body Laminate Rivet Metal plate other freedoms. All six freedoms of the reference point
for loading RPm were constrained. General contact for explicit
analysis with penalty tangential behavior (friction coef-
ficient of 0.1) was used to simulate interaction relation-
ships. In addition, to improve calculation efficiency,
y zx mass scaling was used for the whole model based on
o the restriction that mass changing percent less than 3%.
Cohesive layers
F Rivet head area To reveal the damage propagation mechanism of the
Figure 4 Finite element model of composite/aluminum riveted composite laminate, progressive damage model based on
joints VUMAT subroutine was used. Hashin-type criteria and
cohesive zone method were used to predict the damage
initiation and failure modes of the composite laminate.
Five kinds of failure modes are considered, which are
(2178.8 ± 279.0 N) when a loud sound similar to wood
fiber tensile failure, fiber compressive failure, matrix ten-
fracture was heard. In mesoscale, location of fiber frac-
sile cracking, matrix tensile crushing and delamination.
ture was random, which results in the differences of the
To predict the initial damage in composites, the linear
second load peak for different specimens. Since then,
elastic failure criteria are expressed as follows.
with several smaller load decreasing and increasing
Fiber tensile failure (FT), for ( ε1 > 0):
stages, it come to failure stage and the composite lami-
nate was damaged gradually, especially around the mid- 
ε1
2 
γ12
2 
γ13
2
2
dle hole. At the same time, rivet head began to penetrate + + = eFT .
XT /C11 S12 /C44 S13 /C55
the composite laminate, as displacement load increases,
(1)
the specimen could not bear any load because of serious
delamination and rivet-penetration fracture. Fiber compressive failure (FC), for ( ε1 < 0):
 2
ε1
3 Finite Element Model 2
= eFC . (2)
3.1 Finite Element Modelling XC /C11
In order to investigate the failure mechanism of the Matrix tensile cracking (MT), for ( ε2 > 0):
composite/aluminum riveted joints, finite element
model of composite/aluminum joints was established
 2  2  2
ε2 γ12 γ23 2
in Abaqus as shown in Figure 4. Three dimensional + + = eMT .
YT /C22 S12 /C44 S23 /C66
8-node linear reduced integration elements (C3D8R) (3)
with enhanced hourglass control method were used
Matrix tensile crushing (MC), for ( ε2 < 0):
for the specimen. After mesh dependence studying
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 5 of 10


ε2
2 
γ12
2 
γ23
2 Result comparison of stress-strain curves between the
2
+ + = eMC . open-hole tensile tests and the FEM are shown in Fig-
YC /C22 S12 /C44 S23 /C66
(4) ure 5(a). It is not necessary to add zero-thickness cohe-
sive layers for open-hole tensile model due to the fact
Where εi (i = 1, 2, 3) are normal strains, γij (i, j = 1, 2, 3; that cohesive layer only affects the mechanical behavior
i ≠ j) are shear strains, Cij (i, j = 1, 2, 3, 4, 5, 6; i = j) are along the out-of-plane direction. In Figure 5(a), average
components of the stiffness matrix. In any of the ele- slope value of the stress-strain curves of the open-hole
ments, when the criterion e ≥ 1 is satisfied, damage varia- tensile tests is 53241.4 MPa, while that of the FEM is
bles dk = 1 (k = FT, FC, MT, MC) will be satisfied, and the 66149.0 MPa with relative error between the FEM and
element will be damaged by reducing the stiffness matrix the tests of 24.2%; average ultimate stress value of the
as shown in Eq. (5). When there were fiber failure or out- open-hole tensile tests is 563.1 MPa, while that of the
of-plane matrix failure, the element was failed and it will FEM is 578.9 MPa with relative error between the FEM
be deleted. and the tests of 2.8%. The slope difference between the
d
C11 = (1 − dFT )(1 − dFC )C11 , FEM and the test is because bump points on the mechan-
d ical chucks were hard to embed into composite specimen
C22 = (1 − dFT )(1 − dFC )(1 − dMT )(1 − dMC )C22 , surface so that there was slight slip of the specimen dur-
d
C33 = (1 − dFT )(1 − dFC )(1 − dMT )(1 − dMC )C33 , ing tensile test in general. However, the FEM developed
d
C12 = (1 − dFT )(1 − dFC )(1 − dMT )(1 − dMC )C12 , in this study did not consider the slip phenomenon. Thus,
slopes of experimental curves were lower than that of the
d
C13 = (1 − dFT )(1 − dFC )(1 − dMT )(1 − dMC )C13 , FEM curve.
d
C23 = (1 − dFT )(1 − dFC )(1 − dMT )(1 − dMC )C23 ,
d
C44 = (1 − dFT )(1 − dFC )(1 − 0.9dMT )(1 − 0.5dMC )C44 ,
600
d
C55 = (1 − dFT )(1 − dFC )(1 − 0.9dMT )(1 − 0.5dMC )C55 , OHT-FEM
OHT-EXP1
d
C66 = (1 − dFT )(1 − dFC )(1 − 0.9dMT )(1 − 0.5dMC )C66 . OHT-EXP2
(5) OHT-EXP3
Stress (MPa)

400
OHT-EXP4
In order to simulate the delamination of the composite
laminate, zero-thickness cohesive interfaces were created
by offset method with COH3D8 elements. Bilinear trac-
200
tion-separation response was used for cohesive elements
during the simulation. Quadratic nominal stress crite-
rion is used to simulate the initiation of the mixed-mode
delamination. Once the delamination initiation criterion 0
is met, delamination begins to propagate according to the 0.000 0.003 0.006 0.009 0.012
delamination propagation criterion [31, 32]. Delamina- Strain
tion status is represented by the scalar stiffness degrada- (a)
tion variable (SDEG), which changes monotonically from 1500
3PBT-FEM-no cohesive layers
0 (no delamination) to 1 (delamination). The simulation 3PBT-FEM-cohesive layers
procedure stops as soon as either displacement load is 3PBT-EXP1
reached or the simulation fails to converge prematurely 3PBT-EXP2
Stress (MPa)

1000
[33]. 3PBT-EXP3
3PBT-EXP4

3.2 Finite Element Method Validation 500

In order to validate the proposed FEM based on the
Hashin-type criteria and zero-thickness cohesive zone
method, typical standard loading tests, including open-
0
hole tensile test according to ASTM D5766 and three- 0.000 0.005 0.010 0.015 0.020
point bending test according to ASTM D7264, were Strain
conducted with same materials as described in Section 2, (b)
whose stress-strain curves were compared with that of Figure 5 Result comparison of stress-strain curves between the tests
the FEM. and the FEM for: a open-hole tensile test; b three-point bending test
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 6 of 10

Result comparison of stress-strain curves between the stiffness between the FEM and experiments is 8.1%. The
three-point bending tests and the FEM without cohe- first peak load of the experiments in average is 2406.4 N,
sive layer and the FEM with cohesive layer are shown predicted first peak load from the FEM is 2327.7 N, and
in Figure 5(b). In Figure 5(b), average slope value of the the relative error of ultimate load between the FEM
stress-strain curves of the three-point bending tests is and experiments is − 3.3%. The second peak load of the
84256.5 MPa, while slopes of the FEM without cohe- experiments in average is 2178.8 N, predicted second
sive layer and that with cohesive layer are 81996.6 MPa peak load from the FEM is 2409.9 N, and the relative
(− 2.7%) and 81973.4 MPa (− 2.7%); average ulti- error of ultimate load between the FEM and experiments
mate stress value of the three-point bending tests is is 10.6%. Above results indicate that the proposed FEM is
1016.3 MPa, while that of the FEM without cohesive layer effective to predict the load-displacement response of the
and that with cohesive layer are 1026.1 MPa (0.96%) and pull-through test for composite/metal riveted joints.
1021.0 MPa (0.46%). The close simulation results validate Damages of the specimen from the experiment and that
that the method of inserting multi-layer cohesive lay- from FEM are shown in Figure 7. The experiment speci-
ers has little effect on the load-displacement responses. men was mainly broken by rivet-penetration fracture
However, the model with cohesive layers can visualize the (Region 1) and delamination (Region 5). There are obvi-
delamination between composite plies as an interlaminar ously fiber fracture (Region 2), matrix failure (Region 3)
fracture phenomenon, while the model without cohesive and fiber buckling (Region 4) damages in Figure 7(a). Pre-
layers cannot visualize similar phenomenon [4]. dicted damages of the laminate from FEM when displace-
It can conclude that the proposed FEM based on the ment Uz is equal to 3.4 mm were shown in Figure 7(b).
Hashin-type criteria and cohesive zone method is effec- Rivet-penetration fracture, fiber failure, matrix failure,
tive to analyse the mechanical behaviors of the laminated fiber buckling as well as delamination can be seen in Fig-
composites subjected to tensile or flexural loading. ure 7(b), which have good agreement with experiment
results. Figure 7(c) shows the final failure status of the
4 Results and Discussion specimen predicted by the FEM (Uz = 5.0 mm), in which
4.1 Load‑Displacement Response and Predicted Damages
The validated finite element method was then used to
analyse the mechanical behavior of the composite/metal (a)
Failure modes:
riveted joints. The load-displacement curve predicted 1
by the FEM with cohesive layer was compared with that 1-Rivet penetration
from the experiments, as shown in Figure 6. It can be
seen that load-displacement curve from the FEM agrees 2-Fiber failure
well with that from the experiments for all 4 character- 2
istic stages exhibited similar trends. The stiffness of the 3-Matrix failure
experiments in average is 1256.7 N/mm, predicted stiff- 5
3
ness of the FEM is 1359.1 N/mm, and the relative error of 4-Fiber buckling
Len X50

4 200 um 5-Delamination

3000 (b)
Joint-EXP1
Joint-EXP2 Rivet-penetration fracture
2500
Joint-EXP3
Joint-FEM Fiber & matrix failure
2000
Load (N)

1500

1000
(c)
500
Fracture of the plies
due to delamination
0
0 1 2 3 4 5
Displacement (mm) Figure 7 Damages of the experiment specimen and that of the
Figure 6 Comparison of load-displacement curves from finite element model: a damages of the experiment specimen; b
experiments and finite element method damages of the FEM; c final failure status of the specimen
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 7 of 10

hole of the composite laminate has been totally broken by Matrix compressive crushing (MC) firstly appears at
rivet head and delamination has been propagated to the elements under the edge of the rivet head in the mid-
whole laminate. The comparison results indicate that the dle plies when Uz = 1.4 mm. And it propagates slowly
proposed FEM is effective to predict damages of the pull- to the top surface through the thickness of the laminate.
through test for composite/metal riveted joints. Between the first load peak and second load peak, matrix
compressive crushing begins propagating rapidly along
the circumferential direction under the edge of the rivet
4.2 Progressive Failure Mechanism of the Joints head at the top surface, which is the main cause of the
Based on the proposed FEM, failure mechanism of the load fluctuation when displacement reaches 2.1 mm.
composite/aluminum riveted joints subjected to pull- Delamination (SDEG) firstly appears at inter-
through loading was investigated. face between ply-4 and ply-5 with small size when
For first load take-up stage on the load-displacement Uz = 1.4 mm. Serious delamination appears at inter-
curve of the FEM, load increases linearly as Uz increases face between ply-16 and ply-17 as displacement reaches
from 0 mm to 0.7 mm, and there are no damages for the 1.8 mm and propagates very rapidly along the longitudi-
specimens. As shown in Figure 8, matrix tensile cracking nal direction and transverse direction, which causes the
(MT) firstly appears when Uz = 0.7 mm, and it propa- suddenly load decrease after the first load peak. After the
gates rapidly along the radial direction in the ply and second load peak point, multi-layer delamination occurs
along the thickness direction outside the ply. After the and propagates, which is the other cause of the suddenly
first load peak point, matrix tensile cracking has propa- decrease of the load after the second load peak. Then,
gated to the transverse edge of the laminate. After the as the displacement increases from 2.4 mm to 3.0 mm,
second load peak point, the rivet head penetrate through the rivet head penetrates through the laminate layer by
the top surface of the laminate, and matrix tensile crack- layer and contacts with the lower layer, which contrib-
ing propagates more rapidly and almost reaches its maxi- utes to close the interface splits caused by delamination
mum value. so that load increases to another small peak again. Until
Fiber tensile damage (FT) firstly appears close to the displacement reaches 3.4 mm, the rivet head has already
interface between the laminate and the aluminum plate penetrated through 8 layers of the laminate, and the lami-
when Uz = 1.1 mm, and it propagates very slowly. Until nate is fractured.
the rivet head penetrates through the laminate, fiber ten- After above analysis, dominant failure modes and their
sile damage rapidly propagates at the top surface close locations in laminate was illustrated in Figure 9. The
to the edge of the rivet head, which is one of the main length of the line represents the degree of damage. The
causes of the suddenly decrease of the load after the sec- longer the line, the more serious the damage. It can be
ond load peak. concluded that: (1) the laminate is progressively broken

x: Longitudinal direction z: Loading direction

3500
z SDEG MT SDEG MT
3000 y x

2500 FT FT MC
MC SDEG
2000
Load (N)

1500 FT
MT
ply-24
1000 ply-1
FT SDEG MT MC
500

0
0 1 2 3 4 5
Displacement (mm)
Figure 8 Progressive damage propagation of the specimen from the finite element model
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 8 of 10

Delamination
Fracture
Fiber tensile failure 4 mm
Matrix tensile failure
Propagation direction
Radius Circumferential

45 ply-24
-45 2 mm
90
90
0
90
0 Rivet
90
90
-45
45
90
90
Composite 45
-45

direction
Loading
laminate 90
90
0
90
0
90 Longitudinal
90 direction
-45 ply-2
45 ply-1
Figure 9 Schematic illustration of dominant failure modes and their locations in laminate

mainly by rivet-penetration fracture and delamination shown in Figure 10. When Uz = 1.5 mm, as Figure 10(a)
between ply-16 and ply-17; (2) the material within the shows, the specimen is in the stage of stiffness and it
scope of the rivet head is more dangerous with more seri- deforms elastically as the displacement load increases.
ous damages; (3) damages are more serious for the 90° When Uz = 2.4 mm, as Figure 10(b) shows, the compos-
plies than other plies; (4) tensile damages are the main ite laminate begins to delaminate close to the interface of
failure modes. the composite/aluminum plates due to high tensile stress.
To validate the predicted damage propagation results, When Uz = 3.9 mm, the rivet has completely penetrated
three characteristic points (Uz = 1.5 mm, 2.4 mm, and through the top surfaces of the composite laminate, and
3.9 mm) on the load-displacement curve were selected, meanwhile the delamination propagates to the whole
whose damages were compared with the experiments, as laminate. At the three characteristic points, all predicted

Rivet penetration
Delamination Delamination

(a) Uz=1.5 mm (b) Uz=2.4 mm (c) Uz=3.9 mm

Figure 10 Comparison of damage propagation of the specimens between experiment and finite element model when displacement Uz is equal
to: a 1.5 mm; b 2.4 mm; c 3.9 mm
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 9 of 10

damages from the FEM are in good agreement with the Authors’ Information
Yuxing Yang, born in 1991, is a PhD at Marine Engineering College, Dalian Mari-
experiment results, which indicates that abovementioned time University, China. He received his doctoral degree from Dalian University
progressive failure mechanism results of composite/ of Technology, China, in 2019. His research interests include mechanics of
metal rivet joints subjected to pull-through loading pre- composite materials and composite marine propeller.
Yongjie Bao, born in 1980, is a professor at Dalian Maritime University, China.
dicted by the proposed FEM are effective. He received his doctoral degree from Dalian University of Technology, China, in
2010.
Xueshu Liu, born in 1977, is an associate professor at Dalian University of
Technology, China. He received his doctoral degree from University of Tokyo,
5 Conclusions Japan, in 2012.
In this paper, a composite/aluminum riveted joints of Jinlong Wang, born in 1989, is a PhD at Dalian Maritime University, China. He
airplane wing box subjected to pull-through loading was received his doctoral degree from Dalian University of Technology, China, in
2019.
studied. The specimen consists of a carbon fiber/epoxy Fengming Du, born in 1988, is an associate professor at Dalian Maritime
composite laminate, 2A12 aluminum plate and Ti6Al4V University, China. He received his doctoral degree from Dalian University of
rivet were tested and their load-displacement responses Technology, China, in 2016.
were analysed according to characteristic points. To Funding
predict the damage initiation and failure modes of the Supported by National Natural Science Foundation of China (Grant Nos.
composite laminate, a progressive damage model based U21A20165, 52205515, 52105431), Applied Basic Research Program of
Liaoning Province of China (Grant No. 2022JH2/101300221), Dalian Science
on the Hashin-type criteria and zero-thickness cohesive and Technology Innovation Fund of China (Grant No. 2022JJ12GX033),
zone method was developed using VUMAT subrou- National Key Research and Development Project of China (Grant No.
tine in Abaqus and validated by open-hole tensile test 2020YFB2009805), and China Postdoctoral Science Foundation (Grant Nos.
2020M680937, 2020M670734)
and three-point bending test, which can predict 5 types
of failure modes (fiber tensile failure, fiber compressive Data availability
failure, matrix tensile cracking, matrix tensile crushing The data that support the findings of this study are available from the
correspondingauthor upon reasonable request.
and delamination). Load-displacement response, fail-
ure modes and damage propagation predicted by the Competing Interests
finite element method were all in good agreement with The authors declare no competing financial interests.
the experimental results for the composite/metal riv-
eted joints. At last, progressive failure mechanism was Received: 27 January 2021 Revised: 7 June 2022 Accepted: 6 January
revealed using the proposed finite element method. 2023
Published: 26 January 2023
Major conclusions are as follows.

(1) There are 4 obvious characteristic stages (first load

References
take-up stage, damage stage, second load take-up [1] N Zimmermann, P H Wang. A review of failure modes and fracture analy-
stage and failure stage) on the load-displacement sis of aircraft composite materials. Engineering Failure Analysis, 2020, 115:
curve for the pull-through test of the composite 104692.
[2] D K Rajak, D D Pagar, P L Menezes, et al. Fiber-reinforced polymer
rivet joints; composites: Manufacturing, properties, and applications. Polymers, 2019,
(2) The proposed finite element method can predict 11(10): 1667.
stiffness, first load peak and second load peak well [3] A Katunin, P Sul, A Łasica, et al. Damage resistance of CSA-doped PANI/
epoxy CFRP composite during passing the artificial lightning through the
with relative error of 8.1%, − 3.3% and 10.6% com- aircraft rivet. Engineering Failure Analysis, 2017, 82: 116-122.
pared to experiment results; [4] Y Yang, X Liu, Y-Q Wang, et al. A progressive damage model for predicting
(3) Specimen was mainly broken by rivet-penetration damage evolution of laminated composites subjected to three-point
bending. Composites Science and Technology, 2017, 151: 85-93.
fracture and delamination between ply-16 and ply- [5] G Mittal, K Y Rhee, V Mišković-Stanković, et al. Reinforcements in multi-
17; scale polymer composites: Processing, properties, and applications.
(4) The material within the scope of the rivet head is Composites Part B: Engineering, 2018, 138: 122-139.
[6] M Li, W Luo, Y Chen, et al. Full-field strain distribution and failure char-
more dangerous with more serious tensile damages acteristics of CFRP-repaired steel structures. Engineering Failure Analysis,
than other regions, especially for 90° plies. 2020, 115: 104664.
[7] X Meng, Y Huang, Y Xie, et al. Friction self-riveting welding between
polymer matrix composites and metals. Composites Part A: Applied Science
Acknowledgements and Manufacturing, 2019, 127: 105624.
Not applicable. [8] M S H Fatt, D Sirivolu. Marine composite sandwich plates under air and
water blasts. Marine Structures, 2017, 56: 163-185.
Author contributions [9] N Kharghani, C G Soares, N G Tsouvalis. Experimental and numerical study
YY write the article and analysed the data. YB proposed the idea of experi- of the bolt reinforcement of a composite-to-steel butt-joint under three-
ments and assisted to conduct the experiments. XL provided experimental point bending test. Marine Structures, 2019, 63: 384-403.
conditions. JW and FD contributed to review the article. All authors read and [10] Y Li, Y Ma, M Lou, et al. Advances in spot joining technologies of light-
approved the final manuscript. weight thin-walled structures. Journal of Mechanical Engineering, 2020,
56(6): 125-146. (in Chinese)
Yang et al. Chinese Journal of Mechanical Engineering (2023) 36:10 Page 10 of 10

[11] H Zhao, J Xi, K Zheng, et al. A review on solid riveting techniques in

aircraft assembling. Manufacturing Review, 2020, 7: 40.
[12] B Jiang, Q Chen, J Yang. Advances in joining technology of carbon fiber-
reinforced thermoplastic composite materials and aluminum alloys. The
International Journal of Advanced Manufacturing Technology, 2020, 110(9):
2631-2649.
[13] A Galińska, C Galiński. Mechanical joining of fibre reinforced polymer
composites to metals—A review. Part II: Riveting, Clinching, Non-
Adhesive Form-Locked Joints, Pin and Loop Joining. Polymers, 2020, 12(8):
1681.
[14] G Meschut, M Merklein, A Brosius, et al. Review on mechanical joining
by plastic deformation. Journal of Advanced Joining Processes, 2022, 5:
100113.
[15] D Krassmann, E Moritzer. Development of a new joining technology for
hybrid joints of sheet metal and continuous fiber-reinforced thermoplas-
tics. Welding in the World, 2022, 66(1): 45-60.
[16] C Aldridge. Riveting technology for carbon fibre and composite material.
Surrey: Mack-Brooks Exhibitions, 2016 [2020-12-21]. https://​www.​fastf​
ixtec​hnolo​gy.​com/​autom​otive/​rivet​ing-​techn​ology-​for-​carbon-​fibre-​and-​
compo​site-​mater​ial.
[17] S D Thoppul, J Finegan, R F Gibson. Mechanics of mechanically fastened
joints in polymer–matrix composite structures – A review. Composites
Science and Technology, 2009, 69: 301-329.
[18] D Liu, Y Tang, W L Cong. A review of mechanical drilling for composite
laminates. Composite Structures, 2012, 94: 1265-1279.
[19] A Pramanik, A K Basak, Y Dong, et al. Joining of carbon fibre reinforced
polymer (CFRP) composites and aluminium alloys – A review. Composites
Part A: Applied Science and Manufacturing, 2017, 101: 1-29.
[20] S Sathiya Naarayan, D V T G Pavan Kumar, S Chandra. Implication of
unequal rivet load distribution in the failures and damage tolerant design
of metal and composite civil aircraft riveted lap joints. Engineering Failure
Analysis, 2009, 16(7): 2255-2273.
[21] D V T G P Kumar, S S Naarayan, S K Sundaram, et al. Further numerical and
experimental failure studies on single and multi-row riveted lap joints.
Engineering Failure Analysis, 2012, 20: 9-24.
[22] M Y Solmaz, T Topkaya. Progressive failure analysis in adhesively, riveted,
and hybrid bonded double-lap joints. The Journal of Adhesion, 2013, 89:
822-836.
[23] H M Rao, J Kang, G Huff, et al. Impact of specimen configuration on
fatigue properties of self-piercing riveted aluminum to carbon fiber
reinforced polymer composite. International Journal of Fatigue, 2018, 113:
11-22.
[24] G Catalanotti, P P Camanho, P Ghys, et al. Experimental and numeri-
cal study of fastener pull-through failure in GFRP laminates. Composite
Structures, 2011, 94: 239-245.
[25] P J Gray, C T McCarthy. An analytical model for the prediction of through-
thickness stiffness in tension-loaded composite bolted joints. Composite
Structures, 2012, 94: 2450-2459.
[26] W Ma. Study on the pull-through characteristics of carbon/epoxy braided
composite laminates. Nanchang: Nanchang University, 2016. (in Chinese)
[27] Y Liu. Analysis on mechanical properties of glare laminates with countersunk
bolted joints subjected to out-of-plane loading. Harbin: Harbin Institute of
Technology, 2018. (in Chinese)
[28] American Society for Testing and Materials. ASTM D7332/D7332M:
Standard test method for measuring the fastener pull-through resistance
of a fiber-reinforced polymer matrix composite. PA, West Conshohocken:
ASTM International, 2016.
[29] Y Shen, Y Xiao. Effects of fretting damage on preload relaxation in bolted
composite joints. Acta Materiae Compositae Sinica, 2019, 36: 400-409.
[30] Y He, Y Xiao, Z Su. Effects of surface contact on the dynamic responses of
delaminated composite plates. Composite Structures, 2019, 229: 111378.
[31] J F Chen, E V Morozov, K Shankar. Simulating progressive failure of
composite laminates including in-ply and delamination damage effects.
Composites Part A: Applied Science and Manufacturing, 2014, 61: 185-200.
[32] G Giuliese, R Palazzetti, F Moroni, et al. Cohesive zone modelling of
delamination response of a composite laminate with interleaved nylon
6,6 nanofibres. Composites Part B: Engineering, 2015, 78: 384-392.
[33] G Kolks, K I Tserpes. Efficient progressive damage modeling of hybrid
composite/titanium bolted joints. Composites Part A: Applied Science and
Manufacturing, 2014, 56: 51-63.