Composites: Part A 98 (2017) 192–206

Contents lists available at ScienceDirect

Composites: Part A
journal homepage: www.elsevier.com/locate/compositesa

Loading, support and geometry effects for pin-reinforced hybrid

metal-composite joints
Alex T.T. Nguyen a, Chamendra K. Amarasinghe a, Milan Brandt a, Stefanie Feih b,c, Adrian C. Orifici b,⇑
a
Additive Manufacturing Research Centre, School of Engineering, RMIT University, Melbourne, Australia
b
Sir Lawrence Wackett Aerospace Research Centre, School of Engineering, RMIT University, Melbourne, Australia
c
Joining Technology Group, Singapore Institute of Manufacturing Technology, 2 Fusionopolis Way, Singapore 138634, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: Reliable joining technologies are increasingly required for multi-material lightweight structures. For
Received 2 November 2016 metal-composite joints, there is significant potential to integrate through-thickness pins onto the metal
Received in revised form 18 March 2017 surface to improve bond strength. We investigated Selective Laser Melting manufactured Ti-64 adher-
Accepted 21 March 2017
ends with integrated pins bonded to carbon fibre-reinforced polymer composite, and characterised joint
Available online 4 April 2017
performance from pure tension to shear-dominated pin loading. Both single pin and multi-pin double
cantilever beam specimens were examined and correlated using experimental and finite element meth-
Keywords:
ods. Adherend support conditions affected the single pin pull-out process and related energy absorption
A. Hybrid
C. Finite element analysis (FEA)
up to 35% for all pin offset angles. The pin alignment with respect to the crack direction and fibre angle
D. Failure had little effect on joint performance. Pins with grooved surface features further increased energy absorp-
E. Joints/joining tion by 60% compared to smooth cylindrical pins. This work adds significant insight into pin-reinforced
hybrid metal-composite joints and their performance and optimisation in realistic structural scenarios.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Hybrid metal-composite joints are becoming more common in
many engineering structural applications especially in the aero-
space industry due to extensive use of carbon fibre-reinforced
polymer (CFRP) composites. One of the latest breakthroughs is
the implementation of hybrid titanium-CFRP composite fan blades
on modern aero engine models such as GE9X (GE Aviation) and
Advance (Rolls Royce), which offer 30% reduction in weight and
20% boost in efficiency/emission reduction [1]. In this system, the
composite fan blade is bonded to a titanium leading edge [2] for
improved damage tolerance under impact. The mechanical perfor-
mance of this joint category can be further enhanced with the
addition of surface features along the bondline through advanced
manufacturing techniques. Recent research by the authors has
employed the selective laser melting (SLM) technique to success-
fully produce functional titanium adherends integrated with hier-
archical groove features [3] or through-thickness reinforcement
(TTR) pins [4] for joining to CFRP composites.
The behaviour of the pins reinforcing the hybrid joint under
tensile (pull-out) loads has been characterised in detail by the
⇑ Corresponding author.
E-mail address: adrian.orifici@rmit.edu.au (A.C. Orifici).
authors [4], and established the relation between the pin geometry
parameters, pin/composite interface and failure mode, as well as
compared findings to similar research for carbon z-pinned com-
posites [5–10]. Separately, the behaviour of a pin-reinforced hybrid
joint in pure shear or shear-dominated loading has been investi-
gated by several researchers. Generally, depending on the type of
pin, three types of damage modes have been observed, including
pin fracture in the case of pins with tip features [11–14], pin defor-
mation accommodating the pull-out in the case of a straight cylin-
drical pin [14,15], and composite damage in the case of a triangular
pin [16]. These published studies show that in both pull-out and
shear the strength and damage mode of the pins directly influence
the overall damage pattern of any multi-pin joint, where pin frac-
ture and composite (adherend) fracture lead to catastrophic failure
while pin bending and pull-out result in a more progressive failure
process with higher energy absorption. However, the effect on the
pin pull-out process for different combinations of mode mixity has
not been characterised, and the variation in pin performance
between the limiting cases of pure tension pull-out and shear-
dominated loading is not well understood. Further, there have
been no studies considering the introduction of pin surface fea-
tures or alternative pin geometries that can enhance the energy
absorption during pull-out, without increasing the pin strength
to such an extent that catastrophic adherend fracture occurs.

http://dx.doi.org/10.1016/j.compositesa.2017.03.019
1359-835X/Ó 2017 Elsevier Ltd. All rights reserved.
 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206
In the broader body of knowledge on TTR pins such as carbon z-
pin reinforced composites, related knowledge on pin behaviour
under varying combination of pull-out and shear loading has been
presented in several publications. This is commonly investigated
using pull-out of pins that are manufactured at an ‘‘offset angle”
with respect to the loading axis. Cox et al. [17,18] developed a
semi-analytical model for pin traction under varying loading
conditions, where the shear loading caused an identified pin
‘‘snubbing” effect that was incorporated as an additional friction
component in pure pull-out loading. An experimental study by
Yasaee et al. [19] on single pin specimens with offset angles indi-
cated that only a limited range of angles produced stable pin
pull-out behaviour, i.e. below 11° for pins embedded in a unidirec-
tional laminate or 33° in a quasi-isotropic laminate. Above such
values, the pin fracture and energy absorption was similar to when
the pin was loaded under pure shear. M’membe et al. [20] and Car-
tie et al. [5] showed that the maximum load and energy absorption
increased with increasing offset angle when the pin was oriented
in the direction of the load. In terms of damage mechanisms, the
region around the pin cavity was crushed and the pin deformed
laterally to accommodate the pin pull-out process [5,21,22].
Despite these insights, no correlation to the ratio of pull-out and
shear stresses has been established, and no correlation has been
made between single pin pull-out and multi-pin joint performance.
Additionally, the range of offset angles (and hence ratio of pull-out
and shear stresses) has been relatively small due to the low shear
strength of the carbon z-pins, and a much broader characterisation
is required for titanium pins in a hybrid joint due to the strong pin-
composite bond [4]. Further, the change in joint behaviour depend-
ing on how the pin is inclined relative to the composite fibre archi-
tecture or the crack growth direction is not well understood, and it
is not clear in what circumstances joint behaviour is affected.
Another important aspect of joint performance that is relevant
to industrial structures is the support condition of the adherends.
This relates to whether the adherends are fixed as the pin is pulled
out, or whether the adherends and supporting structure can adjust
to the forces generated in the pull-out process. This is particularly
critical where the pins have an offset angle, as the additional
effects such as snubbing and pin bending cause forces that pro-
mote lateral displacement and rotation of the adherends. As such,
it is important to characterise the joint performance between the
two extreme conditions of adherends that are fixed or relaxed to
self-adjust, which has not previously been studied. Understanding
the effect of support condition in single pin specimens allows the
support condition in multi-pin joints to be identified, which then
relates to improved understanding of the corresponding damage
modes and joint performance. This also provides critical informa-
tion regarding the test method that is most appropriate for single
pin specimens in order to correctly correlate to multi-pin perfor-
mance. To the authors’ knowledge none of the multi-pin structures
previously studied in literature have had an assessment of the
adherend support conditions.
In this study, the performance of reinforcements in hybrid SLM
metal-CFRP composite joints is investigated in single pin speci-
mens and multi-pin joints. Single pin specimens of varying pin
geometry (length, diameter, offset angle and features) and adher-
end support condition (fixed and relaxed to self-adjust) are used
to characterise the composite microstructure and the response of
the single pin as it is pulled out of the composite under different
loading combinations. Our previous publication [4] used a similar
investigation method to successfully characterise the pin/compos-
ite interface strength and to determine the critical L/D aspect ratio
of the pin to achieve complete pull-out for straight pins. This study
further focuses on varying the pin offset angle (to control the ratio
of pull-out to shear stresses), adherend constraint, pin orientation
with respect to the load direction and fibre direction, and surface
193
features of the pin. Double Cantilever Beam (DCB) specimens in
unpinned and pinned configurations are tested experimentally
and analysed with finite element (FE) models. This work charac-
terises in detail the influence of the aforementioned joint parame-
ters on the pull-out performance of single-pin specimen, and the
findings are furthermore correlated with properties of multi-pin
joint configurations. The use of single-pin and multi-pin joints in
the same study also provides a key understanding of the correla-
tion between these joint types with regards to the damage modes,
joint performance and adherend constraint. The characterisation
across different pin geometries (pin diameter and length) and sur-
face features deepens our understanding of energy absorbing fea-
tures at different length scales.
2. Experimental methodology
2.1. Single pin pull-out test
The additive manufacture in form of selective laser melting
(SLM) was utilised to print components from Ti-64 in a build
chamber of 250 mm
250 mm
350 mm (SLM250HL, SLM Solu-
tions, Germany). Prior to printing, the chamber was filled with
Argon gas to avoid oxidation of the component during the manu-
facturing process. The platform was pre-heated to 200 °C to min-
imise build-up of residual stresses during manufacturing. The
adherends were printed with a layer thickness of 30 lm using a
YLR-Fibre-Laser at 175 W. The process parameters are listed in
Table 1. A porosity content of 0.3% ± 0.1% was obtained during
the printing process (3D X-ray Computer Tomography) with an
average surface roughness (Sa) of 10 lm ± 3 lm (Alicona IF-
EdgeMaster profilometer). Single pin specimens were printed on
a square platform with a centre pin positioned at various offset
angles ranging from 0 to 30° (Fig. 1a). The overall dimensional tol-
erance of the SLM-manufactured components was very high, with
an average tolerance of 0.1 mm for all pin dimensions. The tita-
nium adherend thickness was accurate within 0.05 mm. The com-
posite adherends were manufactured from T700 carbon/epoxy
pre-preg tape (VTM264, Advanced Composites Group) with unidi-
rectional lay-up for best match with the titanium arm stiffness. The
hybrid joints were autoclave co-cured according to manufacturer
recommendations for the composite (120 °C, 620 kPa, 1 h). A poly-
tetrafluoroethylene (PTFE) film of 0.1 mm thickness was inserted
at the interface of the titanium adherend and the composite to
act as a starter crack (see Fig. 1a). This allowed measurement of
the pull-out response of the pin, as the adhesion properties
between the two materials were characterised previously [4]. In
addition, the PTFE film thickness does not affect the overall pull-
out results as the film thickness is significantly smaller than total
interface area between the pin and the composite.
Following curing, aluminium tabs were adhesively bonded to
the surface of the adherends using a thinly spread layer of com-
mercial (420A) Araldite adhesive with room temperature cure. As
the tip of the pin is chamfered, there was no contact between
the pin and the applied adhesive. Two types of test setup, a fixed
constraint and a self-aligning or ‘‘relaxed” constraint, were devel-
oped to apply the tensile loading (see Fig. 1b). In the self-aligning
setup, the tabs were connected to a double lap joint with a single
bolt on each side of the joint, and the other side of the joint was
clamped in the machine grips. This allowed for relative horizontal
movement (along x-axis) of the two adherends to self-adjust or re-
orient during the pin pull-out process as the joints are loosely
bolted. In the fixed setup, the tabs were simply gripped in the test
machine and the adherends had no freedom to self-adjust.
Seventeen different specimen configurations for single pin
pull-out were investigated as shown in Table 2. The pins in all
194 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206

Table 1
SLM process parameters.

Laser power Layer thickness Scan speed Energy density Hatch type Hatch spacing Spot size Chamber temp
[W] [lm] [mm/s] [J/mm3] [lm] [lm] [°C]
175 W 30 710 68.5 Checker-board 120 80 200

D Fig. 2a). To investigate the mixed mode loading effect on pull-out

behaviour, an ‘‘offset angle” or ‘‘pin angle” was used to offset the
CFRP CFRP load axis and the pin axis. Four pin offset angles (see Fig. 1(a)) were
tested ranging from U = 0° to 30°. The degree of pull-out/shear
loading is characterised by the ‘‘mixed mode ratio” u, which is
defined by the ratio of shear load to total load [19,20,23] and
L

Pin Axis related to the pin offset angle U as follows:

Offset Angle ( ) sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P2x
R=0.5 mm u¼ ¼ sin U ð1Þ
P2x þ P2z
2.57 mm

PTFE as where Px and Pz are the applied pin load components in shear (per-
pre-crack pendicular to pin axis) and tension (along pin axis), respectively. As
shown in Table 2, the four offset angles correspond to a range of
Ti mixed mode loading conditions.
10 mm All specimens with cylindrical pin geometry were tested with
fixed and relaxed boundary conditions. To investigate the effect
of pin orientation with respect to the fibre direction, the pins were
inclined ‘‘along” (parallel to) and ‘‘across” (in-plane perpendicular
(a) to) the fibre direction of the composite. That is, with reference to
Fig. 1a, the fibres are horizontal when the pins are inclined along
the fibres, and directed out of the page when the pins are inclined
Machine Grip across the fibres. Three repeat specimens were tested for each con-
Fix All DOF figuration, except the 20° and 30° pins inclined across the fibre
direction in the fixed setup where only two specimens were tested
due to pin fracture in manufacture. The 10° pin inclined across the
fibre with relaxed test setup was not tested due to manufacturing
defects. Separately, three specimens of three pin geometry fea-
Double Lap Joint
tures, namely, a grooved surface, pyramid tip and helical pin were
investigated to assess the influence of pin surface geometry on
pull-out behaviour. Detailed geometrical parameters of each con-
Single Pin figuration are indicated in Fig. 2. All specimens were loaded under
Specimen displacement control at a rate of 0.5 mm/min. The opening load
was measured by the load cell, and the pull-out displacement
was corrected for machine compliance via an extensometer. Once
the pull-out tests were completed, the damage mechanisms within
Aluminium Tab
the specimens were analysed using Computer Tomography (CT)
scan (Phoenix v|tome|xs Industrial High-Resolution CT & X-ray
System). Typical scanning parameters were 50 kV, 400 A with a
voxel size of 11 mm.
Fix All DOF
2.2. Double cantilever beam hybrid joints

Z Double Cantilever Beam (DCB) specimens (ASTM D5528-13

Free Translation
Along X axis X Fig. 3. Specimens were printed via SLM with an inclination angle
(b)
Fig. 1. (a) Single pin specimen; (b) single pin test setups with relaxed (left) and
fixed (right) conditions. (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)
specimens were 4.5 mm in total length (Lt) and 1.0 mm in diameter
(D) with a 1.0 mm chamfer at the tip of the pin to assist the pinning
process. Given the chamfer of the pin, the effective pin length (L) is
3.5 mm, which is measured to the bottom of the chamfer (see
[24]) were manufactured to study the crack growth in the hybrid
multi-pinned joints under Mode I (opening) loading, as shown in
of 20° from the vertical axis of the building platform, which was
required to reduce geometrical distortion of the unsupported and
overhanging pin structures due to limited powder support [11],
and thermal residual stresses in the part due to insufficient heat
dissipation [4,11]. The same pin insertion and joint manufacturing
approach as for the single pin specimens was used. All pins were
oriented along the fibres (as previously discussed and shown in
Fig. 1a), and pin orientation was also defined as being ‘‘against”
or ‘‘along” the crack growth direction, as shown in Fig. 3.
Six configurations for DCB specimens were investigated as
shown in Table 3, with one unpinned baseline configuration and
 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206 195

Table 2
Single pin pull-out specimen configurations.

Pin feature L [mm] D [mm] ðUÞ [°] u Pin constraint type Pin inclination orientation Pin volume [%]
Cylindrical 3.5 1.00 0 0 Relaxed, fixed – –
3.5 1.00 10 0.17 Relaxed, fixed Along fibres, against fibres –
3.5 1.00 20 0.34 Relaxed, fixed Along fibres, against fibres –
3.5 1.00 30 0.50 Relaxed, fixed Along fibres, against fibres –
Grooved 3.5 1.00 0 0 Fixed – –
Pyramid 3.5 1.00 0 0 Fixed – –
Helical 3.5 1.00 0 0 Fixed – –

Fig. 2. Single pin specimen pin geometries. (a) Cylindrical, (b) grooved, (c) pyramid, and (d) helical.

ten subsequent configurations of varying pin geometry. Two pin
sizes were used i.e. large pins (L = 3.5 mm, D = 1.0 mm) with 1.0%
volume content and small pins (L = 2 mm, D = 0.5 mm) with 0.5%
volume content. The volume content of different pin sizes was con-
trolled by using a spacing of 8.23 mm and 6.27 mm for the large
and small pins respectively. Each pin size was tested with two dif-
ferent offset angles (0° and 20° for large pins, 0° and 10° for small
pin). All pins were inserted along the fibre direction and against the
crack growth direction, and the large pins at 20° were also investi-
gated along the crack direction (see Fig. 3). Two specimens per con-
figuration were tested.
The overall dimensional tolerance of the SLM-manufactured
components is good, with process-related geometry deviations dis-
cussed in detail in reference [4]. The pin axis of the specimen is
straight and the misalignment angle is estimated to be 0° prior
to the insertion of pins into the CFRP composite adherend. Detailed
post-test inspection of the pin cavity on the single pin specimens
found a small misalignment angle of 2–3° between the cavity (or
pin) and the loading axis of the test machine. This was the result
of the joint assembly process, specifically in the manual insertion
process and the bonding of the loading tabs. This misalignment
is still much less than typical values of 14° for CFRP z-pins [7].
The first row of pins was placed at the start of the bonded
region ahead of the starter crack. The thickness of the composite
adherend in the unpinned specimen was calculated to equal the
bending stiffness of the metal adherend using Eq. (2) [24] to ensure
symmetric deformation.
sffiffiffiffiffiffiffiffiffiffi
1 ETi
tCFRP ¼ 3
t Ti ð2Þ
ECFRP
where t is the total adherend thickness, and E represents the elastic
modulus in the specimen longitudinal direction of the composite
(ECFRP) or metal (ETi) adherend. An equivalent thickness of the com-
posite adherend (tCFRP) in an unpinned specimen can be determined
to be 2.5 mm for a metal adherend (tTi) of 2.57 mm thickness. After
curing, the thickness of the composite reduced to 2.43 mm due to
resin bleeding. The as-manufactured thickness was used in numer-
ical models. For the pinned configurations, the composite adherend
thickness was equivalent to the total pin length. In the case of the
196 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206

Fig. 3. DCB specimen; Cross-section AA showing pin orientation against and along crack direction. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)

Table 3 3.2. Finite element model

Double cantilever beam specimen configurations.

L [mm] D [mm] ðUÞ [°] Pin inclination orientation Pin volume [%] Three-dimensional models of the DCB specimens with geomet-
– – – Unpinned – ric non-linearity and progressive interface damage were evaluated
3.5 1 0 – 1.0 in Abaqus/Standard 6.14. The key aspects of the FE model and all
3.5 1 20 Along crack 1.0 material parameters are summarised in Fig. 4 and Table 4. Zero
3.5 1 20 Against crack 1.0 thickness cohesive elements (COH3D8) [25] were embedded along
2 0.5 0 – 0.5
the interface of the titanium and the composite adherend to model
2 0.5 10 Against crack 0.5
the pre-defined crack path. The total number of nodes and ele-
ments of the numerical model was approximately 63,000 and
51,000, respectively. The pre-crack region was modelled with
large pinned configuration, an additional ten composite layers were
duplicated, but disconnected nodes. Cohesive elements along the
co-cured with the metal adherend to ensure equal bending stiffness
crack path had an average length of 0.05 mm, which is smaller than
of both arms. Moreover, due to anticipated high fracture toughness,
the critical cohesive zone length to ensure accurate convergence of
another ten composite layers were furthermore added to each
the damage modes. The penalty stiffness K was determined using
adherend arm of the large pinned configuration to avoid the frac-
[26]:
ture of the arm during DCB testing. The total thickness of the com-
posite arm after curing is approximately 6.6 mm ± 0.1 mm. Without E33
the additional laminate reinforcements, the arms of the DCB speci- K ¼ bp ð4Þ
t
mens break without significant delamination growth [9,10].
The bonded aluminium tabs were loaded under displacement where E33 is the adherend through-thickness elastic modulus, t is
control at a rate of 2 mm/min. The crack length was measured dur- the adherend thickness, and bp is a parameter used to set the pen-
ing the test with a travelling optical microscope. Using the modi- alty stiffness.
fied beam theory method, the mode I interlaminar fracture As described in a previously published study by the authors for
toughness (GIc) can be calculated according to the ASTM test stan- the same material system [4], the basic adhesion properties of the
dard [24]: unpinned specimens are based on resin fracture and fibre bridging.
To capture the simultaneous damage mechanisms of resin fracture
3Pd and fibre bridging, we utilize two superposed sets of cohesive ele-
GIc ¼ ð3Þ
2bða þ DaÞ ments along the bond line, treating these two damage events as
continuous processes. A bilinear and exponential cohesive law is
where P is the applied load, d is the opening displacement, b is the
used for the resin fracture and fibre bridging, respectively. Table 5
specimen width, a is the crack length and Da is a correction factor
summarises the required input data.
determined from test compliance.
The discrete bridging traction mechanism of the pins during the
pull-out process can be modelled using 3D solid pin geometry [27],
3. Analysis methodology non-linear springs [11,12] or cohesive elements [7,13,16]. In this
work, individual non-linear springs (SPRINGA) [25] were used at
3.1. Maximum load and pull-out energy the actual location of each metal pin, hence resulting in discrete
toughening elements rather than a continuous contribution. The
In this work, the load-carrying capability of the pin configura- same numerical approach was applied in our previous publication
tion is evaluated in terms of the maximum load, Pmax. The failure [4] to successfully evaluate the fracture toughness of pin-
energy (or absorbed energy) E of a single pin associated with the reinforced hybrid joint with straight pins. Tabular spring law input
pull-out process can be calculated from the area under the traction was defined for the numerical model based on respective experi-
load versus pull-out displacement curve. mental pin pull-out data.
 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206 197

Displacement Control
Coupled In All DOF Pins Spacing
Based On Pin Non-Linear Spring (SPRINGA)
Volume Density Presenting Pin Pull-out Process
Pre-crack

CFRP Composite (C3D8) Superposed Cohesive Elements (COH3D8)

Presenting Crack Front and Fibre Bridging Mechanisms
Titanium (C3D8)

CFRP Composite
Adherend

Interface Elements with

Resin Fracture Property
Interface Elements with
Fibre Bridging Property
Springs with Single Pin
Pull-out Property
SLM Titanium
Adherend

Fig. 4. DCB FE modelling scheme (a) methodology and representative elements; (b) typical damage mechanisms of crack propagation process (4). (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)

Table 4 (horizontal lengths H3 and H4 and vertical length V 1 and V 2 ) define

Material properties of VTM264 [31] and Ti-6Al-4V [32]. E and G are elastic and shear the size of the localised fibre waviness region (Fig. 5). The cross-
modulus; m is Poisson ratio; subscripts 1,2,3 are directions in a ply-based coordinate sectional surface of the specimen was sanded with several grit
system of fibre, in-plane transverse and out-of-plane transverse, respectively; XT, ZT
and S12 are in-plane tension, out-of-plane tension and in-plane shear strength; tply is
sizes ranging from 200 to 2000 and each of the aforementioned
thickness of the composite ply. parameters was measured using microscopy for three specimens
of each configuration. The coefficient of variation (CV) was less
VTM264 Ti-6Al-4V
than 5% for all sets of measurements. These parameters were
E11 (MPa) 120,000 E (MPa) 110,000 previously identified as symmetric and dependent on the pin
E22 (MPa) 7500 G (MPa) 42,500
E33 (MPa) 7500 m 0.32
dimension in the same direction. A fitting correlation of
G12 (MPa) 3900 H3 = H4 = 3.5Dh and V1 = V2 = 3Dv was established for all investi-
G13 (MPa) 3900 gated pin sizes. The fibre waviness was confined within the region
G23 (MPa) 2300 of V 1 and V 2 and reduced linearly from its maximum around the
m12 0.32
pin to zero at the length V 2 . A resin-rich zone including pure resin
XT (MPa) 2459
ZT (MPa) 45 zone and broken fibres formed in the region between the fibre
S12 (MPa) 85 waviness region and the pin as shown in Fig. 5. The horizontal
tply (mm) 0.21 length of the resin-rich zone was defined by H1 and H2 , and it
was found that H1 = H2 = 3Dh.
When an offset angle is introduced in the direction of the fibres,
the horizontal length of the resin-rich zone H1 and fibre waviness
4. Results and discussion
zone H3 is reduced at about H1 = 2.3Dh and H3 = 2.8Dh for all three
offset angles. Other parameters of the microstructure remain sim-
4.1. Composite micro-structure surrounding the pin inserts
ilar to the 0° pins. When the offset angle is across the fibres, the
vertical length V1 is increased to approximately V1 = 3.8Dv while
Inserting the through-thickness SLM pins into the composite
other parameters remain unchanged. A summary of microstructure
material causes micro-structure disruptions similar to those seen
parameters is given in Table 6.
for CFRP z-pins (see for example the work of Chang et al. [28]).
The microstructure of the composite was described in detail in a
previous publication by the authors [4], and the model is extended 4.2. Effect of loading and support for a single pin along the fibres
here to account for the different pin offset angles. Fig. 5 shows a
schematic of the altered in-plane cross-sectional profile indicating The experimental results of the single pin specimens oriented
regions of fibre waviness, broken fibres and a resin-rich zone along the fibre direction with varying pin offset angles are pre-
around each inserted pin. For 0° specimens, four parameters sented in Fig. 6 and Table 7, where Fig. 6 presents the traction load
198 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206

Table 5
Cohesive model parameters.

Cohesive model parameters Metal-composite joint

Resin fracture (cohesive input 1) Fibre bridging (cohesive input 2) Composite
Normal strength t n (MPa) 42.2a 1.3a 44.5a
Normal fracture energy GIc (J/m2) 200a 800a 1000a
Normal displacement d (mm) – 6.6a
Shape parameter a – 11b
Normal stiffness Kn (MPa) 35,700b 35,700b
Shear strength t s (MPa) 81a 4b 85c
Shear fracture energy GIIc (J/m2) 1100b 1650b 2750c
Shear stiffness Ks (MPa) 35,700b 35,700b
Mixed-mode ratio g 1.21c 1.21c
Penalty stiffness bp 10b 10b
a
Calculated from experimental data.
b
Calibrated to fit with experimental data.
c
Based on literature data [26,33–35].

Resin-Rich Zone V1

Broken Fibres Pure Resin Broken Fibres

Dv

Dh

V2

Decreasing
Fibre Waviness

H1 H2

Y H3 H4

Fig. 5. Single pin specimen in-plane cross-section (4). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)

Table 6 For all tests the maximum load was relatively insensitive to test
Fibre waviness region parameters.
setup and presence of shear stresses, with only a small (10%) vari-
Pin orientation U (°) H1/Dh H2/Dh V1/Dv V2/Dv ation in the relaxed test setup with changing offset angle, and rel-
0 2.9 3.0 3.1 3.0 atively constant maximum load when changing to the fixed setup
Along fibres 10 2.2 2.8 2.9 3.0 or considering the fixed setup with increasing offset angle. All tests
20 2.4 3.0 3.1 3.0 were in general characterised by two-stage load-displacement
30 2.3 2.9 3.0 3.0 behaviour (shown in Fig. 6a and b): (i) elastic behaviour up to max-
Across fibres 10 2.8 2.8 3.8 3.1 imum load; (ii) gradual reduction in load until the pin was pulled
20 2.8 2.9 3.9 2.8 completely out of the composite.
30 2.9 2.8 3.8 3.0 For the pure pull-out specimens (0° pin) an exponential reduc-
tion in load was seen in both relaxed and fixed test setups (see
Fig. 6a and b). This is caused by the partially melted particles on
versus pull-out displacement results for each support condition, the surface of the SLM pin. During pull-out these particles break
and Table 7 summarises all results. Fig. 7 shows computed tomog- off from the pin surface and subsequently reduce the friction coef-
raphy (CT) scans of the specimens after testing for the two setups, ficient for pull-out, resulting in an exponential load-displacement
where images (a)–(d) are at the base of the pin cavity in the com- curve as the friction forces dissipate with increasing pull-out. This
posite adherend, and images (e)–(f) show the entire pin and top of type of behaviour and damage mechanism had been characterised
the metal adherend. in reference [4]. Although the behaviour was similar across the two
All specimens exhibited pull-out failure mode as the length to test setups, the fixed setup had an absorbed energy 27% higher
diameter (L/D) aspect ratio was below the transition value of 3.5 than the relaxed test setup in pure pull-out. The difference in total
established in previous study [4]. Good repeatability was seen energy absorption is directly related to the damage mechanisms of
across all results, with CV values between 7% and 11% for maxi- each configurations, which are further discussed the following
mum load and failure energy. paragraphs.
 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206 199

Fig. 6. Experimental results: Opening traction load vs crack opening displacement of single pin specimens, pins along the fibres, with (a) relaxed and (b) fixed test setups;
post-test images of specimens and tabs, with (c) relaxed and (d) fixed test setups. (For interpretation of the references to color in this figure legend, the reader is referred to
the web version of this article.)

Table 7 particles remaining in the pull-out cavity can be seen in both cases.
Single pin specimens, pins along the fibres, experimental results. For the relaxed test setup, the pin cavity diameter in pure pull-out
Test setup U (°) u Pmax, N (CV) E, J (CV) Failure mode was approximately 1.08 mm, which is very close to the diameter of
the pin (1.05 mm) prior to bonding. In the case of the fixed setup in
Relaxed 0 0 595 (7%) 438 (6%) Pull-out
10 0.17 650 (7%) 713 (5%) Pull-out pure pull-out, the diameter of the pull-out cavity increased to
20 0.34 647 (6%) 819 (3%) Pull-out + spalling approximately 1.25 mm. Cavity widening is caused by the small
30 0.50 604 (2%) 846 (5%) Pull-out + spalling misalignment angle between pin axis and load axis as shown in
Fixed 0 0 632 (3%) 555 (7%) Pull-out Fig. 7a. Cavity widening absorbs additional energy and results in
10 0.17 663 (3%) 578 (3%) Pull-out a notable 27% increase in absorbed energy for this boundary condi-
20 0.34 641 (1%) 616 (3%) Pull-out + spalling tion. In contrast, with the relaxed test setup, the adherends can
30 0.50 607 (7%) 568 (11%) Pull-out + spalling
adjust horizontally during pull-out to minimise energy absorption.
For the specimens with mixed mode loading, the difference in
damage mechanism between the relaxed and fixed setups is more
As the mixed mode loading conditions were introduced with apparent, and the absorbed energy behaviour is quite different to
the use of non-zero pin offset angles, the load-displacement beha- the pure pull-out case. Fig. 7c and e shows cross-sectional CT scans
viour and the corresponding absorbed energy varied significantly of the composite and the titanium adherend in the relaxed setup,
between the two setups. For the relaxed specimens, the descending and Fig. 7d and f shows typical angle pin specimens in the fixed
branch of the load-displacement curve changed from exponential setup. In the relaxed setup, the main damage mechanism was resin
to linear. This produced a significant increase in total energy crushing due to bearing stresses. This was accompanied by hori-
absorption even for small mixed mode ratios with 63%, 87% and zontal movement of the two adherends, as shown in Fig. 6c. The
93% increases for the 10°, 20° and 30° pins respectively. In contrast, fact that the specimens displaced horizontally relative to each
the fixed specimens showed only a small variation in absorbed other during the pull-out process shows that freedom in displace-
energy with increasing mode mixity, with 11% increase for the ment is critical to the joint behaviour and rotational freedom has a
20° pins and no change for the other pin angles. The load- rather insignificant effect on the joint. No visible plastic deforma-
displacement curve for these specimens showed two regions: an tion was seen on the titanium pin. In contrast, for the fixed setup,
exponential region as in pure pull-out, and a plateau region at no horizontal displacement was permissible or observed (see
low loads (see Fig. 6b). Fig. 6d), and part of the composite adherend was crushed and
The difference in pull-out response between the two test setups spalled out during the pull-out process. Additionally, the pin was
is a result of the way the specimen responds to shear stresses, and observed to be plastically deformed after being pulled out (see
the subsequent damage mechanisms. For the pins in pure pull-out, Fig. 7f). So, despite being more restrained and experiencing
Fig. 7a and b shows CT scans of the composite adherend after test- obvious loss of material in the pin-contacting resin-rich area, the
ing for the two boundary conditions, where the partially melted fixed setup only led to a small increase in absorbed energy with
200 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206
Fig. 7. CT scan cross-sections single pin specimens, pins along fibres: Left images from relaxed test setup, right images from fixed setup. (a and b) 0° pin cavity; (c and d) 20°
pin cavity; (e and f) 20° pin. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
increasing mode mixity. In contrast, despite the adherends being
allowed to displace and rotate, the damage mechanisms promoted
in the relaxed setup absorbed a significantly higher amount of
energy relative to the pure pull-out case.
The damage mechanisms are illustrated schematically in Fig. 8.
Fig. 8a and b shows the 0° pin and offset angle pin at maximum
load, indicating the local shear stresses along the pin-composite
interface from the tensile load and the local bearing stresses cre-
ated by the shear stresses for the angled pins. For the relaxed setup
(Fig. 8c), the bearing stresses force the composite to displace hor-
izontally, which keeps the pin in contact with the composite
around the pin diameter. The increase in contact pressure also
increases the friction that resists pin pull-out (often referred to
as ‘‘enhanced friction”). At low offset angles of 2–3° due to manu-
facturing constraints, the bearing stresses are small and the addi-
tional friction component contributes little to the failure energy.
However, this contribution is more profound with increasing offset
angle, which leads to higher maximum loads and absorbs more
energy during the pull-out process for the 10° pin shown in
Fig. 6. Further increasing the offset angle, the bearing stress is
becoming more prominent and forms a localised stress concentra-
tion region at the wedge created by the acute angle made between
the pin and the resin pocket. For the 20° and 30° pins, prior to
reaching maximum load, the tip of this localised stress concentra-
tion region fails leading to slight loss of contact between the pin
and the composite. Thus, maximum load is reduced slightly with
increasing offset angle from 10° to 30° angle. Following maximum
load, as the resin crushes due to the bearing stresses, a majority of
the crushed material remains in the pin cavity and in contact with
the pin (see Fig. 7c), thus the ‘‘enhanced friction” continues to
resist pull-out throughout the process.
On the other hand, for the fixed joints, the bearing stresses
crush the adjacent resin material to increase the pin cavity
diameter and the pin pulls away from the cavity causing a loss of
contact with approximately half of the pin cavity (see Fig. 8d).
The crushing damage is localised and concentrated at the wedge
region, and the crushed region spalls out due to the lack of support
from the loss of contact. This process continues until at large pull-
out displacements with a large spalled region the bearing stresses
no longer crush the resin but instead the pin starts to bend and
plastically deform. This is indicated by a change of gradient in
the pull-out curve at large pull-out displacements, or the ‘‘plateau”
region of the curve. The bent pin also causes resin crushing on the
non-contact side of the pin cavity, as shown in Fig. 8d, which was
also observed experimentally. A steeper reduction in load is seen at
the end of the pull-out curve as the deformed pin loses contact
with the pin entirely, rather than a gradual end due to progres-
sively reducing friction and crushing.
4.3. Effect of loading and support for a single pin across the fibres
Fig. 9 shows the pull-out traction versus displacement curve for
pins inserted across the fibres. The key features of the curves are
recorded in Table 8 for both boundary conditions. Good repeatabil-
ity was seen across all configurations, with CV values below 10% in
maximum load and energy, except for the 30° pin specimen in the
fixed setup where the CV for failure energy was 19%. In general, for
small to moderate bearing stresses the pull-out behaviour was
similar to pin insertion along the fibres. However, for the 30° pin
with an equivalent mode mixity of 0.5 (see Eq. (1)), fracture in
the composite adherend was seen with a rapid loss of load as
shown in Fig. 9.
The insertion of the pin across the fibres means that the pin is in
contact with fibres, which increases the stiffness of the pin cavity
in comparison with the resin pocket cavity created for pins
inserted along the fibres. This leads to differences in the damage
 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206 201

Fig. 8. Stress state of the interface: at maximum load for (a) 0° pin and (b) angle pin; during pull-out process for the (c) relaxed and (d) fixed setups. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 9. Opening traction load vs crack opening displacement of single pin specimens, pins against the fibres, with (a) relaxed and (b) fixed test setups, experimental results.
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

mechanisms, which are shown in Fig. 10. For the relaxed setup crushing, but conversely a small decrease (11%) in absorbed energy
showing pull-out failure (20° pin), no visible crushing was seen as crushing is an energy absorbing mechanism. For the fixed test
due to the stiffer pin cavity (Fig. 10a), which led to a small increase setup, the stiffer pin cavity did not affect the spalling damage
(8%) in maximum load as the contact area was not reduced by (Fig. 10b), and the maximum load and absorbed energy were very
202 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206
Table 8
Single pin specimens, pins across fibres, experimental results.
Test setup U (°) u Pmax, N (CV)
Relaxed 20 0.34
30 0.50
Fixed 10 0.17
20a 0.34
30a 0.50
a
Only 2 specimens tested.
Fig. 10. CT scan cross-section of pins along fibres: (a) 20° pin, relaxed setup; (b) 20° pin, fixed setup. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)
similar to pins inserted along the fibres. For large bearing stresses
(30° pins) in both test setups, the stiffer pin cavity increased the
stress concentration in the composite causing composite fracture,
which significantly reduced the absorbed energy by 69% and 82%,
respectively, for the relaxed and fixed boundary condition.
4.4. Mode I crack growth of unpinned and multi-pinned hybrid joints
Fig. 11 shows results for load-displacement responses and
R-curves (fracture toughness versus crack length) for all DCB spec-
imen configurations. The overall adhesion response of unpinned
and pinned configurations with small pins (D = 0.5 mm,
L = 2.0 mm) at 0° pin angle has been characterised in detail in a
previous study [4]. The unpinned specimen was shown to have
an initial fracture toughness of 0.2 kJ/m2 (resin-controlled) with
an approximately 30 mm fully developed fibre bridging zone and
a steady-state fracture toughness value of 1.0 kJ/m2. Further details
on characterisation of the unpinned specimens are discussed in
reference [3].
For all pinned configurations, the mechanisms during crack
propagation involve a combination of resin fracture, fibre bridging,
and pin bridging traction via pull-out processes, as previously
demonstrated for 0° pins [4]. As the crack starts to initiate from
the starter crack, the first row of pins acts to provide bridging trac-
tion as follows: the crack front is first arrested by the elastic region
of the pin pull-out process, under increasing opening displace-
ment, the pins start debonding, allowing the crack to then propa-
gate until it is temporarily arrested by the second row of pins.
The crack growth is accompanied by gradual pin pull-out behind
the crack front, which absorbs a significant amount of energy
and results in increasing the mode I fracture toughness. This pro-
cess repeats up to the third row of pins for small pin configurations
and the fifth row of pins for large pin configurations (see Fig. 12).
At this point the maximum load is reached, and the bridging zone
along the crack length is fully developed. Following this, steady-
state crack growth with stick-slip behaviour is seen. The crack slips
when a row of pins furthest away from the crack front is com-
pletely pulled out. This is usually reflected by a small drop in load,
which is regained when the crack advances and arrests at the next
row of pins.
The maximum load and steady-state fracture toughness (Gc,ss)
values are summarised in Table 9, and show good repeatability
with coefficients of variation below 6% for all tested configurations.
E, J (CV) Observed failure mode
699 (9%) 730 (3%) Pull-out + crushing
527 (2%) 259 (8%) Composite fracture
640 (2%) 610 (5%) Pull-out
631 (5%) 600 (6%) Pull-out + crushing
579 (1%) 100 (19%) Composite fracture
Integrating small pins (D = 0.5 mm, L = 2.0 mm) at 0° offset
angle increases the maximum load, fracture toughness at crack ini-
tiation and steady-state fracture toughness to 270%, 250% and
377% of the respective unpinned specimen values. In the case of
the large pins (D = 1.0 mm, L = 3.5 mm), the initiation and steady-
state fracture toughness value is increased by 200% and 900%.
The initiation fracture toughness for DCB specimens with large
pins (D = 1.0 mm, L = 3.5 mm) appears to be less than for small pins
(D = 0.5 mm, L = 2.0 mm) and is due to the larger compliance con-
tributed by the increased thickness of the adherends arm [29]. As
the adherend thickness of the large pin specimens is much larger,
comparison of the maximum loads is not feasible.
Comparing the performance of the angled pins with 0° pins, the
experimental results show that the effect of offset angle on crack
growth responses is very subtle and the crack propagation process
is identical to the 0° pins. Further, when the offset angle is set
against the crack direction, negligible increase of 2% in steady-
state fracture toughness and small reduction of 5% in maximum
load is recorded for both small and large pins configuration. For
the large pins (D = 1.0 mm, L = 3.5 mm) configuration, when the
offset angle is set along the crack direction, both of the maximum
load and fracture toughness value is recorded in Table 9 with a
reduction of approximately 9% and 5%.
Fig. 13 shows CT scans of the composite adherends for the DCB
configuration after testing. In comparison with the single pin spec-
imens in Fig. 7, a key observation is that the behaviour of the pins
in the DCB specimens correlates to behaviour for fixed support
conditions. From Fig. 13a, the 0° pins in the DCB specimens showed
an increase in pin cavity diameter from 1.05 mm (equivalent to pin
diameter prior to pull-out) to 1.28 mm with a small loss of contact
area at the root of the cavity. Similarly, from Fig. 13b and c, the
angled pins set against and along crack direction both showed
spalling and loss of contact area in the wedge region. Pin plastic
deformation was observed for the 20° large pinned specimens
while 10° pinned specimens showed no deformation. These dam-
age mechanisms are characteristic of fixed support conditions seen
in the single pin specimens subjected to additional bearing stress
shown in Fig. 7d and f. The fact that the 0° pins DCB specimens
have bearing damage characteristics, despite being perfectly
straight, is speculated to be attributed to the inherent rotation of
adherend arm in a DCB test [30], which subjects the 0° pins to a
degree of bearing stress during pull-out. Indicated in the single
pin section, this ‘‘enhanced friction” phenomenon has a crushing
 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206 203

2000

Mode I Fracture Toughness (kJ/m2)

12
1750 Pinned: D=1.0, 0o 300
10
1500 Pinned: D=1.0, 0o
Opening Load (N)

250
1250 8

1000 200
6
750 150 Pinned: D=0.5, 0o
20 30 40
4
500

250 Pinned: D=0.5, 0o 2

Unpinned
Unpinned
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60 70 80
Opening Displacement (mm) Crack Length (mm)

(a) (b)
2000

Mode I Fracture Toughness (kJ/m2)

12
1750 Pinned: D=1.0, 20o
Against Crack
10
1500
Pinned: D=1.0, 20o
Opening Load (N)

1250 Against Crack 8

Pinned: D=1.0, 20o
1000 Along Crack
6

750
Pinned: D=1.0, 20o
Along Crack 4
500 Pinned: D=0.5, 10o, Against Crack
FE spring input scaled from D=1.0, 10o pin
2 Pinned: D=0.5, 10o
250
FE Spring input scaled from D=1.0, 10o pin
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60 70 80
Opening Displacement (mm) Crack Length (mm)
(c) (d)
Fig. 11. DCB specimen results, experiment (solid lines) and numerical (dotted lines: relaxed setup; dashed lines: fixed setup). (a) Load-displacement. (b) R-curve of unpinned
(D = 0), pinned at 0° (D = 0.5 and D = 1.0), and (c) Load-displacement. (d) R-curve of pinned specimens at 10° (D = 0.5) and pinned at 20° (D = 1.0). (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 12. DCB specimen features and damage mechanisms. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)
204 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206

Table 9
DCB specimen results.

Experiment, (CV) Numerical, relaxed pull-out law Numerical fixed pull-out law
(% difference to experiment) (% difference to experiment)
Pmax [N] Gc,ss [kJ/m2] Pmax [N] Gc,ss [kJ/m2] Pmax [N] Gc,ss [kJ/m2]
Unpinned 126 0.99 136 0.97 136 0.97
(0.4%) (3.8%) (8%) (3%) (8%) (3%)
D = 0.5, 0° 349 3.61 335 3.71 356 3.90
(3.6%) (5.2%) (4%) (3%) (2%) (7%)
D = 1.0, 0° 1600 8.99 1465 7.12 1585 8.47
(1.5%) (5.1%) (8%) (21%) (1%) (6%)
D = 0.5, 10°, 331 3.72 437 5.72 338 3.76
Against crack (2.0%) (5.1%) (32%) (55%) (2%) (2%)
D = 1.0, 20°, 1519 9.04 1882 11.88 1578 8.99
Against crack (3.8%) (4.0%) (24%) (32%) (4%) (1%)
D = 1.0, 20° 1455 8.51 1882 11.88 1578 8.99
Along crack (1.3%) (4.1%) (29%) (40%) (8%) (6%)

Fig. 13. CT scan cross-sections of DCB specimens: (a) 0° pin; 20° pin (b) against
crack and (c) along crack path. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)
effect on the interface resulting in a widened cavity and a small
spalled area at the root as observed in Fig. 13a.
The experimental results show only a small to negligible differ-
ence in crack growth behaviour from the pin alignment, with
around 5% lower values of maximum load and fracture energy
for pins aligned along the crack growth direction compared to
alignment against the crack growth direction. Experimental work
from other researchers [5,20] found significant changes to failure
mode and energy absorption for end notched flexure (ENF) speci-
mens under mode II loading. In those specimens, changing the
pin alignment meant the axial component of the pin load was a
compressive load for the case where the pin angle is against the
shear load direction. Here, the axial pin load is tensile regardless
of the pin alignment, and the overall pin conditions and load com-
ponents are similar in both cases. Fig. 13b and c shows CT scans
from the two alignment methods, which shows the damage mech-
anisms in the composite were very similar for both configurations.
The larger spalled area shown in Fig. 13b in the offset against crack
direction specimen is also indicative of higher energy absorption in
the pull-out process; hence the slightly higher recorded fracture
toughness.
Fig. 11 and Table 9 also show numerical analysis results, where
two sets of results are presented as the nonlinear pin pull-out law
(used in the spring elements) was taken from the single pin spec-
imens for either the relaxed or fixed test setup. The numerical
results for the 0.5 mm diameter pins were determined by scaling
the pin pull-out law characterised for 1.0 mm diameter pins, where
the scalability of the pull-out laws has been previously demon-
strated [4]. The numerical analysis does not account for the pins
inserted along or against the crack, so the same results are
presented for both cases. The differences between the numerical
and experimental results are also shown as a percentage of the
experimental result.
The comparison between experimental and numerical results is
highly illustrative of the support condition in the DCB tests. The
numerical results for the analyses that used the pin pull-out law
from the single pin specimen with fixed support conditions
resulted in significantly better correlation to the test results than
the relaxed test setup. This is because of the close correlation in
damage mechanisms of the pin pull-out process between single
pin and multi-pin joint (see Figs. 7 and 13). A similar trend is seen
across all configurations of pins; the predictions using the fixed
pull-out law are within 10% of all experimental results for load
and fracture toughness. Furthermore, the pull-out law from the
single pin specimens shows small differences in fracture toughness
response between pin alignment against and along the crack direc-
tion (within experimental standard deviation), which suggests
Mode I fracture response is not sensitive to pin alignment. More
broadly, the comparison between experimental and numerical
results confirms the excellent predictive capabilities of the numer-
ical analysis methodology.
In the previous work [4], the relaxed test setup was used to
characterise the pull-out performance of single pin specimens
and predict the DCB mode I response with thin pins (D = 0.5) only.
As shown in this work, the differences between the two support
conditions are negligible for the thinner pins, as the contribution
of bearing stress becomes much smaller due to the smaller bearing
 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206
area between the pin and the composite. However, the difference is
profound for the large pins (D = 1.0 mm, L = 3.5 mm) as the bearing
area increases significantly, thus, the overall pull-out energy is
enhanced considerably even at a small mixed mode ratio. For the
multi-pin DCB configuration, fixed support during pin pull-out is
proven to accurately predict the joint properties. However, in real
joint structures, the typical support condition is likely to vary
between relaxed and fixed, thus characterising pin behaviour
under both conditions provides key insight into the range of possi-
ble performance.
4.5. Effect of pin geometry features on single pin pull-out and mode I
response
It is speculated that the effect of pin pull-out can be optimized
further for the same pin density by designing additional pin surface
features. Experimental pull-out load-displacements results for the
single pin specimens with pins using different geometry features
(grooved, pyramid, helical) are presented in Fig. 14 and sum-
marised in Table 10. Results for the pin with no features were pre-
sented previously (as D = 1.0, 0° pin, fixed setup) but are repeated
here and labelled ‘‘cylindrical” for comparative purposes. It should
also be noted that the investigated pins are vertically straight so
the pins are neither along nor against the CFRP fibre direction. Sim-
ilar to the previous tests, the CV is within 11% across all variables
indicating high experimental repeatability.
In terms of single pin pull-out, cylindrical pins and grooved sur-
face pins exhibited progressive pull-out while pins with pyramid
tip and helical shape led to composite fracture as shown in
Fig. 14. Despite the difference in failure mode, all the pins studied
had higher maximum load and pull-out energy than the original
cylindrical pin as indicated in Table 10.
The pull-out response of the grooved pin was similar to the
cylindrical pin, however, the behaviour following maximum load
was more linear than exponential. The maximum load increased
by 14%, and significantly the energy absorption increased by 60%
as a result of the linear behaviour following maximum load. This
increase in load-carrying capability is expected to be caused by
the 30% increase in contact area for the pin-composite bond and
mechanical interlocking from the surface features. This demon-
strates that the use of pin features leads to a three level hierarchy
900
Composite Fracture
800
700
Opening Traction Load (N)
600
500
400
300
Pulled-out
200
100
0 tem is produced to substantially enhance the joint performance
0.5 1.0 1.5 2.0 2.5
Pull-out Displacement (mm)
Fig. 14. Opening traction load vs crack opening displacement of single pin
specimens, 0° pin, fixed setup, varying pin geometry. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of
this article.)
205
Table 10
Single pin specimens with geometry features, experimental results.
Pmax, N (CV) E, J (CV) Observed failure mode
Cylindrical 632 (3%) 555 (7%) Pull-out
Grooved surface 720 (3%) 886 (4%) Pull-out
Pyramid tip 755 (5%) 729 (3%) Composite Fracture
Helical 760 (2%) 581 (11%) Composite Fracture
of energy absorbing mechanisms at three different length scales:
the pins, the surface features and the pin-composite bonding.
For the pyramid tip and helical pin, cracking in the composite
was seen and heard, which caused a reduction in slope of the
load-displacement curve. Following this, there was a small
increase in load, before further cracking led to rapid loss of load.
For the pyramid pin the cracking was observed in the composite
around the pin tip (see Fig. 14), and for the helical pin the cracking
was midway along the pin. The maximum load of the pyramid and
helical pins were both around 20% higher than the cylindrical pin.
In this respect, the use of geometry features increases the pin
strength such that the composite adherend strength becomes the
limiting factor for joint behaviour in both cases. In terms of energy,
the helical pin absorbed a comparable amount of energy to the
cylindrical pin, while the pyramid pin showed a 30% increase over
the cylindrical pin. This correlates well with the reported
increasing in joint strength and ductility for this type of failure
mode [16].
5. Conclusion
An investigation into pull-out behaviour of pin-reinforced CFRP
composite combining experimental and numerical procedure was
conducted to investigate the pin pull-out response under varia-
tions in support conditions, mixed mode ratio, pin alignment and
pin geometry. Mixed mode loading ratios from 0 to 0.5 were
generated through the use of offset pin angles from 0° to 30°.
The experimental analysis indicates that adherend support con-
ditions influence the pin pull-out behaviour significantly.
Comparatively, relaxed support conditions exhibit increasing
energy absorption with increasing mode mix loading, which is
characterised by horizontal displacement between the adherends.
On the other hand, the fixed support specimens showed loss of
contact with a large spalled area and pull-out energies remained
similar for all cases. When the pins are oriented across the fibres,
similar pull-out behaviour was seen up to 20° for both test condi-
Cylindrical
tions; above which the bearing pressure exceeded the composite
Grooved
strength leading to catastrophic failure and sudden loss in load-
Pyramid
carrying capability of the joint.
Helical
By correlating pin behaviour in single pin and multi-pin joints,
it was shown that the pin support condition in the DCB configura-
tion corresponds to the situation with fixed adherends. This was
seen by correlation in the damage mode and also through the
use of numerical analysis. In addition, the mixed mode loading
effect is relatively subtle in the DCB configurations with similar
crack propagation responses and energies compared to the 0° pins
at different mixed mode ratio and pin size. Furthermore, the joint
properties are not influenced by the direction of pin offset angle
with respect to the crack direction. Finally, by introducing surface
features onto the surface of the pins, a hierarchical adhesion sys-
3.0 3.5
by a further 60% increase in energy absorption.
The substantial experimental data in this study provides key
insight for the optimal design of pin-reinforced hybrid joints. The
work shows that the hybrid joint performance is not compromised
under mixed mode loading and pin insertion orientation as long as
the pin L/D aspect ratio remains below the critical value of 3.5, and
206 A.T.T. Nguyen et al. / Composites: Part A 98 (2017) 192–206

provides a highly accurate numerical analysis procedure to predict
properties of hybrid joints for different configurations.
Acknowledgements
The authors acknowledge the efforts of Mr Aaron Pateras for
manufacturing of SLM structures, Mr Robert Ryan for support dur-
ing composite lay-up and curing, and Mr. Peter Tkatchyk and Mr.
Julian Bradler for providing assistance during mechanical testing.
One of the authors, A.T.T. Nguyen acknowledges financial support
through a RMIT PhD International Scholarship (RPIS). One of the
authors, S. Feih, acknowledges the support from the Agency for
Science, Technology and Research and the Science Engineering
Research Council of Singapore through the Additive Manufacturing
Centre Initiative (SERC Grant no. 142 68 00088).
References
[1] Gardiner G. Aeroengine composites, Part 2: CFRPs expand. Composites World;
2015.
[2] Rolls-Royce to create composite technology hub in Bristol [press release]; 2015
[18 March 2015].
[3] Nguyen ATT, Brandt M, Orifici AC, Feih S. Hierarchical surface features for
improved bonding and fracture toughness of metal–metal and metal–
composite bonded joints. Int J Adhes Adhes 2016;66:81–92.
[4] Nguyen ATT, Brandt M, Feih S, Orifici AC. Pin pull-out behaviour for hybrid
metal-composite joints with integrated reinforcements. Compos Struct
2016;155:160–72.
[5] Cartié DDR, Cox BN, Fleck NA. Mechanisms of crack bridging by composite and
metallic rods. Compos Part A: Appl Sci Manuf 2004;35(11):1325–36.
[6] Cartié DDR, Troulis M, Partridge IK. Delamination of Z-pinned carbon fibre
reinforced laminates. Compos Sci Technol 2006;66(6):855–61.
[7] Mouritz AP. Review of z-pinned composite laminates. Compos Part A: Appl Sci
Manuf 2007;38(12):2383–97.
[8] Pegorin F, Pingkarawat K, Daynes S, Mouritz AP. Influence of Z-pin length on
the delamination fracture toughness and fatigue resistance of pinned
composites. Compos Part B: Eng 2015(0).
[9] Pingkarawat K, Mouritz AP. Improving the mode I delamination fatigue
resistance of composites using z-pins. Compos Sci Technol 2014;92:70–6.
[10] Pingkarawat K, Mouritz AP. Comparative study of metal and composite z-pins
for delamination fracture and fatigue strengthening of composites. Eng Fract
Mech 2016;154:180–90.
[11] Parkes PN, Butler R, Meyer J, de Oliveira A. Static strength of metal-composite
joints with penetrative reinforcement. Compos Struct 2014;118:250–6.
[12] Philip NP, Richard B, Darryl A. Growth of damage in additively manufactured
metal-composite joints. In: Proceedings of 15th European conference on
composite materials, Venice, Italy.
[13] Philip NP, Richard B, Darryl A. Fatigue of metal-composite joints with
penetrative reinforcement. In: 54th AIAA/ASME/ASCE/AHS/ASC structures,
structural dynamics, and materials conference. structures, structural
dynamics, and materials and co-located conferences. American Institute of
Aeronautics and Astronautics; 2013.
[14] Ucsnik S, Scheerer M, Zaremba S, Pahr DH. Experimental investigation of a
novel hybrid metal–composite joining technology. Compos Part A: Appl Sci
Manuf 2010;41(3):369–74.
[15] Graham DPRA, Baker D, Smith PA, Watts JF. A hybrid joining scheme for high
strength multi-material joints. In: Proceedings of 18th international
conference on composite materials, South Korea.
[16] Wang X, Ahn J, Kaboglu C, Yu L, Blackman BRK. Characterisation of composite-
titanium alloy hybrid joints using digital image correlation. Compos Struct
2016;140:702–11.
[17] Cox BN. Snubbing effects in the pullout of a fibrous rod from a laminate. Mech
Adv Mater Struct 2005;12(2):85–98.
[18] Cox BN, Sridhar N. A traction law for inclined fiber tows bridging mixed-mode
cracks. Mech Adv Mater Struct 2002;9(4):299–331.
[19] Yasaee M, Lander JK, Allegri G, Hallett SR. Experimental characterisation of
mixed mode traction–displacement relationships for a single carbon
composite Z-pin. Compos Sci Technol 2014;94:123–31.
[20] M’Membe B, Gannon S, Yasaee M, Hallett SR, Partridge IK. Mode II
delamination resistance of composites reinforced with inclined Z-pins. Mater
Des 2016;94:565–72.
[21] Rugg KL, Cox BN, Ward KE, Sherrick GO. Damage mechanisms for angled
through-thickness rod reinforcement in carbon–epoxy laminates. Compos Part
A: Appl Sci Manuf 1998;29(12):1603–13.
[22] Rugg KL, Cox BN, Massabò R. Mixed mode delamination of polymer composite
laminates reinforced through the thickness by z-fibers. Compos Part A: Appl
Sci Manuf 2002;33(2):177–90.
[23] Allegri G, Yasaee M, Partridge IK, Hallett SR. A novel model of delamination
bridging via Z-pins in composite laminates. Int J Solids Struct 2014;51(19–
20):3314–32.
[24] ASTM. Standard test method for mode I interlaminar fracture toughness of
unidirectional fiber-reinforced polymer matrix composites ASTM D5528-
13. West Conshohocken: ASTM International; 2013.
[25] Simula. ABAQUS version 6.14 documentation collection; 2015 [updated April
20th].
[26] Turon A, Dávila CG, Camanho PP, Costa J. An engineering solution for mesh size
effects in the simulation of delamination using cohesive zone models. Eng
Fract Mech 2007;74(10):1665–82.
[27] Feih S, Nguyen ATT, Ullah I, Brandt M, Orifici AC. Numerical simulation of pin
pull-out for hybrid joints: comparison of detailed 3D versus simplified 2D and
1D modeling approaches. In: South East Asia Simulia regional users
conference. Singapore: Simulia; 2016.
[28] Chang P, Mouritz AP, Cox BN. Properties and failure mechanisms of z-pinned
laminates in monotonic and cyclic tension. Compos Part A: Appl Sci Manuf
2006;37(10):1501–13.
[29] Robinson P, Das S. Mode I DCB testing of composite laminates reinforced with
z-direction pins: a simple model for the investigation of data reduction
strategies. Eng Fract Mech 2004;71(3):345–64.
[30] Tong L, Sun X. Bending effect of through-thickness reinforcement rods on
mode I delamination toughness of DCB specimen. I. Linearly elastic and rigid-
perfectly plastic models. Int J Solids Struct 2004;41(24–25):6831–52.
[31] Group AC. ACG VTM 260 series; 2006. Available from: http://www.lavender-
ce.com/wp-content/uploads/pds1154-vtm260-issue5.pdf.
[32] Bianchi F, Zhang X. A cohesive zone model for predicting delamination
suppression in z-pinned laminates. Compos Sci Technol 2011;71
(16):1898–907.
[33] Knaupp M, Baudach F, Franck J, Scharr G. Mode I and pull-out tests of
composite laminates reinforced with rectangular z-pins. J Compos Mater
2014;48(23):2925–32.
[34] Cui H, Li Y, Koussios S, Zu L, Beukers A. Bridging micromechanisms of Z-pin in
mixed mode delamination. Compos Struct 2011;93(11):2685–95.
[35] Dai S-C, Yan W, Liu H-Y, Mai Y-W. Experimental study on z-pin bridging law
by pullout test. Compos Sci Technol 2004;64(16):2451–7.