Professional Documents
Culture Documents
Experimental and Finite Element Studies of Thin Bonded and Hybrid Carbon Fibre Double Lap Joints Used in Aircraft Structures
Experimental and Finite Element Studies of Thin Bonded and Hybrid Carbon Fibre Double Lap Joints Used in Aircraft Structures
Experimental and Finite Element Studies of Thin Bonded and Hybrid Carbon Fibre Double Lap Joints Used in Aircraft Structures
Experimental and Finite Element Studies of Thin Bonded and Hybrid Carbon Fibre
Double Lap Joints Used in Aircraft Structures
PII: S1359-8368(15)00582-X
DOI: 10.1016/j.compositesb.2015.09.038
Reference: JCOMB 3805
Please cite this article as: Chowdhury NM, Wang J, Chiu WK, Chang P, Experimental and Finite
Element Studies of Thin Bonded and Hybrid Carbon Fibre Double Lap Joints Used in Aircraft Structures,
Composites Part B (2015), doi: 10.1016/j.compositesb.2015.09.038.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Experimental and Finite Element Studies of Thin Bonded and Hybrid Carbon Fibre
Double Lap Joints Used in Aircraft Structures
a b a b
Nabil M Chowdhury , John Wang , Wing Kong Chiu and Paul Chang
a
Department of Mechanical and Aerospace Engineering, Monash University, Clayton, Victoria 3800,
Australia
PT
b
Aerospace Division, Defence Science and Technology Group, 506 Lorimer Street, Fishermans Bend,
RI
Corresponding Author: Nabil M Chowdhury; nabil.chowdhury@monash.edu; +613 9905 1089; Monash
SC
Abstract:
Finite element analysis (FEA) is performed to verify the static and fatigue strength of mechanically
U
fastened, bonded and hybrid double lap joints. These joints are made from thin carbon fibre/epoxy
AN
laminates applied in aircraft structures. Several configurations are considered, including variations in
rivet array and the addition of bondline defects. Adhesive nonlinear material properties, rivet surface
contacts and frictional forces were included in the three-dimensional (3D) Finite Element (FE) models.
M
The Multicontinuum Theory (MCT) is used to simulate the progressive failure process and the stress
state for all specimens, whilst the strain energy release rate (SERR) as a function of crack length for
D
bonded and hybrid specimens are also compared. Results have shown the FE models are able to
TE
accurately predict the bonded, riveted and hybrid joint strengths. The position of the first row of
fasteners is critical in determining the crack growth rate. As the crack enters the fasteners’ clamping
EP
zone there is a significant drop in SERR resulting in a much slower crack growth rate, therefore
Keywords:
A. Carbon fibre; B. Damage tolerance; C. Finite element analysis (FEA); E. Joints/joining
AC
1.0 Introduction
Joints potentially act as weak links in structures, thus it is important to understand their strength and
the mode in which they fail. The three common methods of joining composite laminates together is
either through mechanical fastening, bonding or the combination of the two, called ‘hybrid’ joints.
Mechanical fasteners such as pins, rivets and bolts have commonly been used in the aerospace industry
for decades [1-4]. The key problem that arises through the use of mechanical fasteners is the high stress
1
ACCEPTED MANUSCRIPT
concentrations around the fastener holes which are more severe in composite laminates compared to
Adhesively bonded joints are structurally more efficient than mechanically fastened joints as they
perform better in distributing loads, eliminating a majority of the high stress concentration problems
seen in bolted joints [6,7]. One underlying weakness associated with bonded joints is its process control
PT
during repair applications. Improper processes would result in a weak bond that is not generally
possible to be detected by means of non-destructive inspection (NDI). The detrimental effect of some
RI
improper surface treatment may not even manifest in a significant reduction of the initial static strength
of the bond but in an adverse impact on the durability of the adhesive bond at service temperature and
SC
moisture environment and/or under fatigue loading [8].
The combination of mechanical fastening and bonding has been employed to safeguard against defects
U
within the adhesive layer which may cause premature or catastrophic failure [9]. It is only after the bond
AN
has failed where the fasteners begin to carry the remaining load in the joint. It is this safety factor that
analysis of bonded and bolted joints and concluded that hybrid joint configurations cannot achieve any
D
significant advantage over adhesive bonding in well-designed intact structures, however it may prevent
TE
defect/damage propagation. Sun et al. [11] showed how the use of ‘attachments’ can allow fasteners to
carry more load in a hybrid joint configuration; results were found experimentally and numerically using
EP
a two-dimensional model whereby the attachments increased joint strength by reducing peel stresses.
Furthermore, Kelly [12] investigated the load distribution in hybrid joints with a single bolt using FEA.
C
Hence in recent years the use of Finite Element (FE) methods to simulate the behaviour of composite
joints has increased particularly due to the advancement of many FE packages. However, most
AC
composite hybrid joint models so far reported are restricted to 3D models with a single fastener or 2D
models [9,13-17]. In order to be used as a versatile design tool, the analysis needs to be able to simulate
multi fasteners as practically used in a joint, and accurately capture the three-dimensional stress states,
material behaviour, bolt clamping, friction, secondary bending effects, load distribution as well as
contact interactions between surfaces [18]. In addition, the localised damage in peak stress regions can
reduce the laminates sensitivity to discontinuity. This results in highly conservative strength predictions
2
ACCEPTED MANUSCRIPT
based on initial failure around hole boundaries [20]. Hence the modelling must include the capability of
progressive failure predictions. From current literature there is no work known to the authors that has
This paper details 3D FEA to predict the static and fatigue strength of riveted, bonded and hybrid double
lap joints made from thin carbon fibre/epoxy laminates. Multiple rivets with different array patterns
PT
were used in the riveted and hybrid joints. For the bonded and hybrid joints, the effects of two different
types of bondline defects were also assessed. Nonlinear adhesive material property, rivet surface
RI
contact, friction forces and the Multicontinuum Theory (MCT) to simulate the progressive failure
process are all included in the FEA models. The full load-displacement curve and failure mode for each
SC
joint configuration are also predicted. In addition the strain energy release rate (SERR) as a function of
crack length for bonded and hybrid specimens are calculated to provide better understanding on how
U
rivets affect the fatigue resistance of joints. The FEA prediction was compared with the results obtained
AN
in the experiment conducted by the authors which was reported in an earlier publication [8].
2.0 Methodology
M
Static and fatigue tests are conducted on eight different specimen configurations highlighted in Table 1.
The composite adherends are manufactured from HexPly M18/1/G939 carbon fibre prepreg [23] in a
AC
satin weave configuration with 0° and 90° fibres; Table 2. All specimens are made in a double lap joint
configuration with the outer adherends five plies thick; [(0/90)/(45/-45)/(0/90)/(45/-45)/(0/90)] and the
inner adherend ten plies thick; [(0/90)/(45/-45)/(0/90)/(45/-45)/(0/90)]s. Lap joints are known to
produce severe stress concentrations around the ends of the overlap; to reduce this, the outer
adherends are manufactured with a 5mm long taper in the form of a ply drop as shown in Fig. 1.
Adherends are cut using a water jet to a size of 120mm x 57mm. Specimens with six fasteners have a
3
ACCEPTED MANUSCRIPT
pitch distance of 16mm [8]. The bond region is 50mm x 57mm and FM300-2K film adhesive is used with
PT
M18/G939 65 67 4.0 4.0 0.04 800 800 100 800 800
RI
σ12 2 2
ν γe γp GIC (kJ/m ) GIIC (kJ/m )
(MPa) (MPa) (MPa) (MPa)
FM300-2K 2400 840 0.4 94.2 54.4 0.055 0.580 1.3 5
SC
Table 4 Material properties of CR7621 Cherry MaxiBolt Rivets [27,28]
U
CR7621-U05-04 201 0.3 965.0 5.12 8.80 655
modelled due to symmetry halfway across the width of the specimen and by making the assumption
M
that the fasteners head and tail are the same, the model can be split midway through the thickness. Fig.
The adherends were modelled using orthotropic material properties defined in Table 2. The adhesive
was modelled using elastic-perfectly plastic material data specified in Table 3. A coefficient of friction of
EP
0.1 was used between the fasteners and composite plate as well as between contacting plate surfaces
[30,31]. This was applied using the master-slave algorithm with small sliding allowed between
C
contacting surfaces and a neat fit between the rivets and the adherends.
AC
A refined mesh using biased seeding was present around the fastener holes and at the ends of the
overlap. A coarser mesh was used further afield from the overlap. Specimens containing an adhesive
layer with no initial cracks had four elements placed through its thickness. A displacement was applied
to the left end of the outer five ply thick adherend in the longitudinal (1) direction; Fig. 1. The right end
of the middle adherend had a clamped boundary condition (u1, u2, u3 = 0). A symmetry condition was
also applied to the underside of the middle adherend and the plane halfway across the width of the
specimen.
4
ACCEPTED MANUSCRIPT
In order to accurately simulate the material failure behaviour, a micromechanical analysis method called
the Multicontinuum Theory (MCT) provided by Autodesk as a commercial plugin to Abaqus was utilised
[21,22]. Micromechanical analysis takes an additional step beyond the conventional laminate theory to
separate the stress and strain in the matrix and fibre from a Representative Volume Element (RVE). MCT
predicts failure at the fibre and matrix level by obtaining the volume averaged stress states in both the
PT
fibre and the matrix. Here, matrix failure is assumed to be influenced by all six of the matrix average
stress components in a 3D analysis, whilst a quadratic function is used to find the average stress of the
RI
fibre [21,22]. In contrast progressive damage models such as the Hashin criteria provided by Abaqus can
be computationally expensive and typically suffer convergence issues due to discrete stiffness
SC
reductions whilst not accounting for transverse stress and strain components which are accounted for in
MCT. By implementing this method it allows modelling specimen behaviour with far greater accuracy
U
than previously available. Due to this progressive damage model, fibre and matrix degradation values
need to be found through an iterative process. In the riveted case, bearing failure is a compressive
AN
mode of failure, hence requiring slightly higher fibre and matrix degradation values. Optimal values
ended up being 0.01 and 0.35 for the fibre and matrix respectively. This was obtained through
M
iteratively comparing FEA results with Configuration 2 experimental results.
Experimental load-displacement data were presented by Chowdhury et al. in [8]. These results are the
TE
raw force and displacement data provided by the Instron 100kN software interface. The displacement
transducer of the test machine will measure the total displacement that comprises of the deformation
EP
of the load train in addition to that of the test specimen. The true specimen deformation can only be
determined from the machine displacement transducer when the deformation of the load train is
isolated [29]. To do this, a calibrated extensometer was used on a simple coupon specimen made from
C
the same HexPly M18/1/G939 carbon fibre prepreg. An extensometer with a 25mm gauge is used to
AC
record specimen elongation when subjected to increasing load. A calibration factor is then easily
worked out between the cross head displacement and true displacement.
Configurations 1-4 are considered pristine specimens. These specimens are manufactured in a
controlled manner with no intentional bondline defects. Specimens containing bondlines have been
cured in ideal conditions in an autoclave and for the unique case - Configuration 4 was cured in an oven
5
ACCEPTED MANUSCRIPT
to test whether the clamping pressure provided by the rivets only - are sufficient in obtaining adequate
bond strength. Aircraft components are typically subjected to 3000µƐ [32,33] and hence further
analysis is carried out at this strain level as well as its final failure strain (defined by the specimens’
4.1.1 Configuration 1
PT
Based on the material properties and model setup previously discussed, Configuration 1 consisting of a
bonded joint with six rivets produced the following load-displacement curve; Fig. 2.
RI
This shows that a specimen without a spew will fail after a displacement of 1.02mm with a peak load of
SC
57.1kN. This is 10.0% lower than experimental results. The sudden drop in load seen after a
displacement of 0.87mm is primarily due to a sudden increase in matrix damage using the MCT
progressive damage model [21]. The FEA results indicate a mixed mode failure as both the shear stress
U
in the bond and tensile stress in the first ply exceed their ultimate stress values. Here ‘first ply’ is
AN
defined as the initial ply that is in contact with the bondline in the specimen; hence each bondline can
have two respective ‘first plies’. Specimens previously manufactured contained a spew fillet
M
approximately 1.5mm in length. This was controlled through the use of high temperature resistant flash
tape. The spew fillet can alter the stress distribution in the adhesive layer and thus the model was
D
further modified to take this into account. This significantly reduced the peak shear stress at the end of
the adhesive layer by 17.4%. Initial failure now occurs through first ply failure with a remote strain of
TE
In a hybrid configuration, the presence of fasteners has very little influence on the joint behaviour when
the bond is still intact. The shear stress distribution (S13); Fig. 3(a) shows peak values at the ends of the
overlap. Further understanding on the behaviour of the rivets in the joint region is gained through
C
Fig. 3(b) shows the fastener load share for a joint with an intact bondline. The ‘rivet load’ stands for the
total load carried by all rivets in the joint. As expected there is minimal influence of fasteners in carrying
the load. As the applied load increases, there is a slight increase in fastener load share which reaches a
peak value of 0.21%. The fastener load share is seen to suddenly increase beyond a load of 50kN; this is
expected to greatly increase as bondline cracks begin to initiate. In addition to this, the amount of load
6
ACCEPTED MANUSCRIPT
transferred can be controlled by carefully selecting the adherend thickness, adhesive thickness, overlap
4.1.2 Configuration 2
Configuration 2, containing six rivets failed through bearing failure. This is a slow mode of failure and
thus preferable as it allows time for detection upon regular inspection intervals. The key disadvantage is
PT
its lower load sharing capability. The MCT progressive damage model applied to the composite
adherends allows capturing matrix and fibre damage in Configuration 2. A peak load of approximately
RI
34.1kN is reached followed by hole elongation; Fig. 4.
In contrast to Configuration 1, the shear and normal stress distribution in Configuration 2 produces peak
SC
values at the ends of the fastener holes. This is to be expected as the fasteners are the only means of
keeping the joint together. 53.40MPa is the highest normal stress (S33) with a remote strain of 280µƐ
U
(1.5mm displacement; Fig. 4).
AN
Fig. 5(b) shows the shear load share of the rivets. From the results it is clear that the outer fasteners (1
and 3, 4 and 6) play a larger role in the load sharing of the joint with the end fasteners (4 and 6) carrying
Two different models were analysed for Configuration 3; one model contained no adhesive spew near
the tapered region whilst the second contained an adhesive spew 1.5mm in length.
TE
When modelling Configuration 3 without a spew fillet, adhesive failure occurs before first ply failure.
Further modifying the geometry to contain a 1.5mm long adhesive spew reduces the adhesive shear
EP
stress by 16.1% to 45.65MPa. Unlike the previous model, this new model now fails at a higher load of
60.4kN with a remote strain of 14500µƐ (1.11mm displacement). The primary mode of failure in this
C
case is due to first ply failure at the tip of the tapered region. This failure matches the experimental
AC
whereby adherend failure occurred near the tapered region; reaching a maximum tensile stress value of
803.40 MPa and the adhesive reached a maximum shear strength value of 45.65 MPa.
The maximum shear stress value of this joint occurs during catastrophic failure with a stress of
45.65MPa. This peak value is found in an element adjacent to the middle adherend. Comparing these
result to Configuration 1 show close similarity. Peak values in the normal and shear stress distribution
curves are approximately the same although the presence of rivets results in a flatter shear stress
7
ACCEPTED MANUSCRIPT
distribution. Alongside this, the normal stress is much higher around the fastener holes in Configuration
1. Overall in order to reduce the maximum normal and shear stress values the ratio of the adherend
4.1.4 Configuration 4
Configuration 4 is a hybrid specimen containing three rivets arranged in a staggered array; Fig. 1(b) and
PT
cured in an oven as opposed to curing in an autoclave. This is done to see if the clamping pressure
RI
It must be noted that since the curing conditions have changed in this case, the adhesive strength of the
SC
FM300-2K film adhesive will change as well due to the different porosity level in the adhesive. The
results presented here assume the bond strength of the adhesive is unaffected. Fig. 8 shows a load-
displacement behaviour which slightly under predicts the peak load share of the joint. First ply failure
U
occurs near the tip of the tapered region, halfway across the width of the specimen. This is in very good
AN
agreement with the results presented by Chowdhury et al. [8] which shows adherend failure in the
central width area and adhesive failure toward the outer edges of the specimen.
M
Increasing the applied load in the joint, increases the peak shear stress values at the ends of the
adherends where maximum stress occurs at the ends of the tapered regions. This is where crack
D
initiation occurs. Fig. 9(b) shows the load carried by the rivets.
TE
Similar to Configuration 1, the rivets play minimal role in the joint structure when the bond is intact.
Due to the presence of a single fastener in Rivet row one (RR1), bondline cracking and defects arise at
EP
the corners of the bond region. This results in a corner wise peeling action. Comparing Fig. 3(b) with Fig.
9(b), Configuration 1 applies approximately twice the clamping force than Configuration 4. Hence this
C
shows fasteners play more of a role in the load share of a joint when the ratio between the adhesive
area and fastener hole area (AFHR) decreases. i.e. Configuration 1 has a AFHR value of 35.3, and
AC
Specimen Configuration 5-8 are considered defective specimens. These specimens contain known
defects such as initial bondline cracks or a weak bond achieved by under-curing the FM300-2K film
adhesive at a reduced temperature and curing duration. The main role of producing such specimens is
to understand whether or not they significantly affect the joint structure particularly in a hybrid
8
ACCEPTED MANUSCRIPT
configuration. Such defects can result in various complexities due to altered bondline properties etc.
Thus the main emphasis of such configurations is to understand the joint strength through experimental
4.2.1 Configuration 5
The initial 2mm long crack in Configuration 5 spanned the entire width of the specimen. Due to non-
PT
symmetry in both bondlines, only half of the entire specimen is modelled in Abaqus. Compared to the
previous models discussed, the finite element model for Configuration 5 achieves a peak load of 57.0kN.
RI
Initial failure occurs through first ply failure, i.e. first ply of central adherend failed immediately aft of
the 2mm crack. The experimental results highlight that beyond a displacement of 0.7mm, the load
SC
gradually decreases in contrast to a fairly linear trend in previous configurations containing a bondline.
From Fig. 11, the shear stress distribution reaches a peak value of approximately 26.3MPa just prior to
U
failure. The normal stress distribution shows a small increase in stress 5mm from the edge of the
AN
tapered adherend. This is due to the ply drop off and hence sudden change in material stiffness.
Overall the results presented in this case are conservative in predicting joint failure. Due to plastic
M
deformation in the adhesive layer, joint stiffness reduces at the peak load limits.
4.2.2 Configuration 6
D
Configuration 6 contained a 2mm initial crack between the central adherend and adhesive layer. Six
rivets are placed in a square array to produce the hybrid configuration. Catastrophic failure occurred
TE
through first ply failure. The results once again are conservative compared to the experimental peak
load of 58.0kN. There is a peak shear stress of 26.6MPa following the crack front prior to catastrophic
EP
failure. Comparing the numerical load displacement graphs, Configuration 5 achieves a higher peak load
These configurations contains what is called a ‘defective semi-cured’ adhesive as it is cured at 90°C for
60 minutes as opposed to 121°C for 90 minutes , minimising the peak load share of the joint. From
experimental results conducted in [8], a peak load of approximately 17.3kN is achieved by a bonded
Although a semi-cured adhesive may represent a defective joint, there are numerous other ways to also
define a ‘defect’ in a structure. The presence of an initial crack and the use of expired adhesives are just
9
ACCEPTED MANUSCRIPT
some of many different methods of defining such a ‘defect’. As a result modelling each and every type of
joint will be a tedious task with some results already self-evident. Specimen Configuration 7 and 8 have
only been produced and tested to demonstrate one of many examples of a bondline ‘defect’. In this
case the ‘defective semi-cured’ adhesive is quite detrimental for a bonded joint configuration [8]. By
using a ‘hybrid’ configuration the likelihood of sudden catastrophic failure is minimised through the
PT
added residual strength of fasteners.
From Fig. 14, the combination of FEA results in Configuration 1 (perfect bond) and Configuration 2 (rivet
RI
only) show a type of load-displacement behaviour similar to Configuration 8 experimental results [8].
Adhesive failure occurs at a maximum shear stress (S13) of 19.06MPa with a remote strain of 3300µƐ at
SC
a load of 21.45kN. Following this, the rivets carry the remaining load in the joint structure resulting in
bearing failure.
U
4.3 Joint Configuration Summary
AN
Overall, from the eight different joint configurations experimentally tested and modelled using Abaqus
CAE, the use of the micromechanical MCT progressive failure model and non-linear adhesive properties
has resulted in close agreement between experimental and numerical results. Through this
M
investigation it has been found that modelling minor details such as the adhesive spew present at the
ends of the bondline are critical in achieving reliable results. The presence of the adhesive spew helps
D
reduce stress in the vicinity of the region therefore allowing a marginal increase in load carried by
TE
bonded and hybrid joint structures. The results are reflected in Fig. 2, 6 and 8.
Specimen configurations containing an adhesive has the highest peak normal and shear stress values at
EP
the ends of the overlap. This is the region where damage initiates from. Comparing the ‘pristine’
specimens (Configuration 1-4) show that there are no real advantages between bonded and hybrid
C
specimens under static testing. All of which achieve relatively the same stiffness and strength. A purely
AC
riveted specimen however has a reduced stiffness which fails over a large period of time due to hole
elongation. Overall the peak loads achieved by Configuration 1-4 using FEA are 60.2kN, 34.1kN, 60.4kN
Configuration 5 and 6 contained the 2mm initial crack. These specimens had a lower stiffness than its
pristine counterparts. Configuration 7 and 8 contained the defective semi-cured adhesive. The
numerical results showed that by utilising aspects of the load-displacement graph for Configuration 1
and 2, an upper and lower bound limit can be set for these defects. This may be used as a quick method
10
ACCEPTED MANUSCRIPT
of determining the strength limitations of such a joint. Overall the finding here conclude that fasteners
play a vital role in preventing catastrophic failure provided the adhesive has a shear strength less than
the bearing strength of the adherends. By carefully selecting the adhesive, adherend and fasteners used,
PT
Although several authors [9,35-40] have investigated hybrid joint structures for many years, very little
insight has been given as to why these joint structures have superior fatigue resistance over their
RI
bonded joint counterparts. This is a critical question that needs to be answered in the Aerospace
Industry. The drilling of fastener holes in fibrous materials has been a concern as it leads to (1) peeling
SC
of higher plies at the entry of the hole, (2) fibre wrenching and resin degradation on the wall of the hole
and (3) delamination on the last plies of the laminate [35-40]. This may result in fatigue crack initiation
U
and reduced strength [35]. However the holes for fasteners in a hybrid joint configuration help slow
AN
down crack propagation after initial failure of the adhesive [10]. This will result in improved fatigue life
compared to bonded joints [9,12] and ensure structural integrity even after complete adhesive failure
[38]. This work is conducted to further validate these points but in addition provide crucial information
M
on the effects of fastener positioning and layout on fatigue performance which is not widely seen in
literature. Without fully understanding how positioning fasteners in a bonded joint can improve fatigue
D
Results previously published by Chowdhury et al. [8] and Maofeng Fu et al. [34] have shown hybrid
joints have a longer fatigue life compared to bonded joints. Although under static cases there may not
EP
be any significant advantage between the two, hybrid joints have shown a 45% and 20% increase in
fatigue resistance compared to a purely fastened and a purely bonded joint respectively [8].
C
Finite element analysis using Abaqus CAE is used to model bonded and hybrid double lap joints. In
AC
these joints, significant stresses are present at the ends of the overlap resulting in ‘bondline cracking’.
The repeated loading and unloading of the joint eventually results in crack initiation and over time this
crack will grow across the length of the overlap. By modelling the bonded and hybrid joints with initial
cracks of various lengths, the effects of the fasteners (rivets) are found.
11
ACCEPTED MANUSCRIPT
The strain energy release rate (SERR) represents the energy available for crack growth. Under fatigue
cases due to the constant loading and unloading, there is also a constant release in energy. Hence
calculating the SERR for bonded and hybrid joint structures will provide an indication on its fatigue
performance. According to Paris’s Crack Growth model, the SERR ‘G’ is proportional to the crack growth
PT
rate ‘da/dN’. Thus the higher the SERR, the higher the rate of crack growth provided the critical G value
has been exceeded. There are three methods by which this can be calculated in Abaqus; the Contour-
RI
Integral method, VCCT and XFEM. Several trials of each method are conducted, out of the three models
the Contour-Integral approach is the easiest to implement in 3D models. An initial crack is position
SC
midway through the thickness of the 0.3mm thick FM300-2K film adhesive and is spread the entire
width of the specimen. Specimen setup is similar to that mentioned in Section 3.1 and Fig. 1(a) and (c).
U
Seven different crack lengths (2mm, 9mm, 14mm, 20mm, 25mm, 30mm, and 36mm) are positioned in
AN
the bond region. The nodes positioned aft of the crack tip are tied together using node to node contact.
Seeding is implemented in the mesh and refined around the crack tip with a minimum element size of
0.01mm (where bondline thickness is 0.3mm). Average SERR data is taken about the width of the
M
specimens (along the crack front). The individual Mode I and Mode II SERR (assuming Mode III is
negligible) are derived through post-processing the data using the displacement field method [41]. Fig.
D
5.1.1 Comparison of Bonded and Hybrid Specimens Containing Two Rows of Rivets
Part of Fig. 15 depicts the average SERR data for a bonded and hybrid specimen. From the figure, a
EP
bonded configuration has almost a constant SERR for an increasing crack length. Initially when the crack
length is small (<10mm) there is an increase in G, this is due to the higher peeling stresses present at the
C
end of the adherend in addition to the tapering which allows the adherends to separate easily. When
AC
the crack length is >30mm, there is an increase in G as a crack can easily propagate due to the reduced
bond area. In all cases the Mode II SERR are higher than the Mode I as the joint acts in shear.
For the hybrid configurations, when the crack length is <9mm the Mode I G values are smaller than a
bonded joint configuration. Hence crack initiation occurs at a later point in time for hybrid joints
compared to bonded joints. As crack length increases, Mode I and II G values decrease sharply as it
passes Rivet Row 1 (RR1). This is due to the clamping pressure provided by the head of the rivets and
also due to the added bearing strength from the shaft of the rivets. Furthermore the SERR for every
12
ACCEPTED MANUSCRIPT
point after the first row of rivets is much lower than that of a bonded joint. Hence it is this reduced
value which results in a lower crack growth rate and improved durability.
In the hybrid configuration, when the crack length approaches 25mm, the SERR no longer decrease as
sharply. Upon closer look at the FEA, the tensile load causes a peeling action between the two rows of
PT
This ‘opening action’ assists in initially increasing the crack growth immediately aft of a rivet. Further
RI
growth moves the crack front towards the second row of rivets which provides additional clamping
SC
5.1.2 Hybrid A - Effects of Removing the First Row of Rivets on Strain Energy Release Rate
The first row of rivets in a hybrid joint configuration results in a sudden increase in the SERR prior to
U
reaching RR1. To further understand the cause, the models are modified to contain the following:
• AN
Hybrid A – Hybrid configuration with the first row of rivets removed but containing all six fastener
holes.
•
M
Hybrid B – Hybrid configuration with the first row of rivets (RR1) and fastener holes removed.
Joint A assesses the influence of the fastener holes. Fig. 15 contains the SERR results. The behaviour is
D
quite similar to the standard hybrid joint configuration previously discussed. There is a rapid increase in
TE
SERR prior to reaching the first row of holes. This indicates that the sudden increase in SERR is not only
associated with the presence of rivets but in fact associated with the presence of fastener holes which
are known to be regions of stress concentration [11]. From Fig. 15, the average SERR indicates that
EP
‘Hybrid A’ is less optimal than a proper hybrid joint configuration. The fastener holes only act in
lowering the fatigue resistance of the joint. This highlights the criticality of first row fasteners and their
C
position. The presence of empty fastener holes results in greater localised displacement resulting in an
AC
5.1.3 Hybrid B - Effects of Removing First Row of Rivets + Holes on Strain Energy Release
Rate
Hybrid B is modelled in Abaqus to see the influence of second row rivets only. This is the same as
shifting the first row of rivets from 17mm to 33mm along the overlap length. Fig. 18 shows this type of
joint.
13
ACCEPTED MANUSCRIPT
Unlike the previous ‘Hybrid A’ Configuration, the absence of fastener holes in row one results in a SERR
rate behaviour very similar to that of a bonded joint configuration with the advantage of a marginally
lower SERR for small cracks. This highlights crack lengths less than 25mm are not influenced by the
rivets positioned 33mm away from the tapered region. As soon as the crack approaches the ‘second
row’ rivets (RR2), an overall increase in SERR is observed (Fig. 15). Under Mode II, a peak value of 405
PT
J/m2 occurs. This value declines as the crack progresses past each line of rivets; once again highlighting
the key advantage of these fasteners in reducing crack growth rate and thereby improving fatigue
RI
resistance. Although there is no significant advantage between ‘Hybrid B’ and a purely bonded joint, the
results do in fact suggest it is important to have fasteners positioned closer to the ends of the bondline
SC
than having fasteners positioned in the central regions of the bondline.
Overall, the results here clearly show the advantages of using a hybrid joint configuration compared to a
U
bonded joint. It is more advantageous to place fasteners closer to the ends of the overlap to arrest
AN
sudden crack propagation and improve fatigue resistance. However there is a limit as to how close the
fasteners are placed to the ends, as this is where high stress concentrations are present in the bondline.
This will reduce load carried by the fasteners and will result in shear out failure [10,42]. Thus the stress
M
behaviour of the joint must first be analysed followed by placing fasteners soon after the high stress
6.0 Conclusion
TE
The use of the micromechanical MCT progressive failure model and the nonlinear adhesive material
properties enabled the comparison of fastened, bonded and hybrid joint structures. Both fastened and
EP
conclude that under static test conditions at room temperature there are no real advantages of hybrid
C
joints over properly manufactured bonded joints. Through conducting this analysis, close comparison
AC
between experimental and numerical results is achieved. Findings show that the higher the load carried
by the joint, the larger the role played by the fasteners in a hybrid configuration; the overall significance
In addition, previously very little insight using FEA has been given as to why hybrid joint structures have
superior resistance over their bonded joint counterparts. However through this work it has been shown
the SERR method using FEA provides an effective way of comparing their fatigue performance and may
be used as a tool to optimise fastener placement to enhance hybrid joint durability before manufacture.
14
ACCEPTED MANUSCRIPT
Results here show the presence of fasteners in a hybrid configuration significantly reduce Mode I and II
SERR by more than 83% and 27% respectively compared to a purely bonded joint; provided the crack
has surpassed the initial row of rivets. The additional clamping pressure provided by the rivets also
helps in reducing this value. Further investigating the effects of fastened, bonded and hybrid joint
configuration for thicker joint structures such as step lap joints still require further research. Analysing
PT
these joints through finite element models will assist in primary composite joint repair certification.
Acknowledgments
RI
This research was carried out partially under financial support from the Australian Defence Materiel
Organisation and also as part of a CRC-ACS research program, established and supported under the
SC
Australian Government’s Cooperative Research Centres Program. The authors would like to thank Dr
Andrew Gunnion from CRC-ACS for reviewing this paper and for his helpful comments/suggestions.
U
AN
M
D
TE
C EP
AC
15
ACCEPTED MANUSCRIPT
References
1. T. Ireman, T. Nyman and K. Hellbom, “On Design Methods for Bolted Joints in Composite Aircraft
Structures,” Composite Structures, vol. 25, pp. 567-578, 1993.
2. J. J. Khudhayer and A. R. Othman, “Fiber Reinforced Composite Structure with Bolted Joint - A
Review,” Key Engineering Materials, Vols. 471-472, pp. 939-944, 2011.
3. T. A. Collings, “The Strength of Bolted Joints in Multi-Directional CFRP Laminates,” Composites, vol.
8, no. 1, pp. 43-55, 1977.
4. Wei, J., Jiao, G., Jia, P., & Huang, T. (2013). The effect of interference fit size on the fatigue life of
PT
bolted joints in composite laminates. Composites: Part B, 62-68.
5. L. J. Hart-Smith, “Bolted and Bonded Joints,” in ASM Handbook Volume 21, 1998, pp. 167-176.
6. Soykok, I. F. (2015). End geometry and pin-hole effects on axially loaded adhesively bonded
RI
composite joints. Composites: Part B, 129-138.
7. A. Baker and R. Chester, “Bonded Repair Technology for Aging Aircraft,” RTO AVT Specialists'
Meeting on "Life Management Techniques for Ageing Air Vehicles", Manchester, United Kingdom,
SC
2001.
8. N. M. Chowdhury, W. K. Chiu, J. Wang and P. Chang, “Static and Fatigue Testing Thin Riveted,
Bonded & Hybrid Carbon Fiber Double Lap Joints used in Aircraft Structures,” Composite Structures,
vol. 121, pp. 315-323, 2015.
U
9. G. Kelly, “Quasi-Static Strength and Fatigue Life of Hybrid (bonded/bolted) Composite Single-Lap
Joints,” Composite Structures, vol. 72, pp. 119-129, 2006.
10.
11.
1985.
AN
L. J. Hart-Smith, “Bonded-Bolted Composite Joints,” Journal of Aircraft , vol. 22.11, pp. 993-1000,
Prediction of Hybrid Composite Double Lap Joints,” Composite Structures, pp. 2916-2926, 2010.
14. A. Barut and E. Madenci, “Analysis of bolted-bonded composite single-lap joints under combined
TE
in-plane and transverse loading,” Composite Structures, vol. 88, pp. 579-594, 2009.
15. M. Y. Solmaz and T. Topkaya, “Progressive Failure Analysis in Adhesively, Riveted, and Hybrid
Bonded Double-Lap Joints,” The Journal of Adhesion, vol. 89, pp. 822-836, 2013.
16. A. Pirondi and F. Moroni, “Clinch-Bonded and Rivet-Bonded Hybrid Joints: Application of Damage
EP
Models for Simulation of Forming and Failure,” Journal of Adhesion Science and Technology, vol. 23,
pp. 1547-1574, 2009.
17. Fiore, V., Alagna, F., Galtieri, G., Borsellino, C., Di Bella, G., & Valenza, A. (2013). Effect of curing
time on the performances of hybrid/mixed joints. Composites: Part B, 911-918.
C
18. Z. Kapidzic, L. Nilsson and H. Ansell, “Finite element modeling of mechanically fastened composite-
aluminum joints in aircraft structures,” Composite Structures, vol. 109, pp. 198-210, 2014.
AC
19. A. C. Orifici, I. Herszberg and R. S. Thomson, “Review of methodologies for composite material
modelling incorporating failure,” Composite Structures, vol. 86, pp. 194-210, 2008.
20. J. Wang, A. Banbury and D. Kelly, “Evaluation of Approaches for Determining Design Allowables for
Bolted Joints in Laminated Composites,” Composite Structures, vol. 41, pp. 167-176, 1998.
21. M. R. Garnich and A. C. Hansen, “A Multicontinuum Approach to Structural Analysis of Linear
Viscoelastic Composite Materials,” Journal of Applied Mechanics, vol. 64, no. 4, pp. 795-803, 1997.
22. Key, C. T., Schumacher, S. C., & Hansen, A. C. (2007). Progressive failure modeling of woven fabric
composite materials using multicontinuum theory. Composites: Part B, 247-257.
23. HEXCEL, “HexPly M18/1 180ᵒC Curing Epoxy Matrix Product Data,” 2011.
24. Cytec, “FM300-2 Modified Epoxy Resin Film,” Cytec Engineered Materials, 2011.
16
ACCEPTED MANUSCRIPT
25. C. E. Materials, “FM 300-2 Film Adhesive Technical Data Sheet,” 2011. [Online]. Available:
http://www.cytec.com. [Accessed 2012].
26. M. J. Donough, A. J. Gunnion, A. C. Orifici and C. H. Wang, “Critical Assessment of Failure Criteria
for Adhesively Bonded Composite Repair Design,” in 28th International Congress of the
Aeronautical Sciences, Brisbane, 2012.
27. C. Aerospace, “Cherry Maxibolt Blind Bolt,” Cherry Aerospace, Santa Ana, California, 2007.
28. A. Materials, “Aircraft Materials,” Cherry Aerospace, 2013. [Online]. Available:
http://www.aircraftmaterials.com/data/alstst/A286.html. [Accessed 16 October 2013].
29. Chalkley, P. D., & Chiu, W. (1993). An improved method for testing the shear stress/strain
PT
behaviour of adhesives. International Journal of Adhesion and Adhesives, 237-242.
30. P. D. Herrington and M. Sabbaghian, “Factors Affecting the Friction Coefficients between Metallic
Washers and Composite Surfaces,” Composites, vol. 22, no. 6, pp. 418-424, 1991.
RI
31. L. Tong, “Bearing Failure of Composite Bolted Joints with Non-uniform Bolt-to-Washer Clearance,”
Composites Part A, vol. 31, no. 6, pp. 609-615, 2000.
32. W. Johnson, G. K. Yamauchi and M. E. Watts, “NASA Heavy Lift Rotorcraft Systems Investigation,”
SC
Ames Research Center, Hampton, VA, 2005.
33. A. Baker, S. Dutton and K. Donald, “Properties of Composite Systems,” in Composite Materials For
Aircraft Structures, Reston, Virginia, American Institute of Aeronautics and Astronautics, 2004, p.
275.
U
34. M. Fu and P. K. Mallick, “Fatigue of Hybrid (adhesive/bolted) Joints in SRIM Composites,”
International Journal of Adhesion and Adhesives, vol. 21, pp. 145-159, 2001.
AN
35. M. Ouellet and A. Vadean, “Design Improvement of Hybrid Composite Joints By Axiomatic Design,”
in The Seventh International Conference on Axiomatic Design, Worcester, 2013.
36. R. Matsuzaki, M. Shibata and A. Todoroki, “Improving performance of GFRP-aluminum single lap
joints using bolted-co-cured hybrid method,” Composites Part A: Applied Science and
M
Manufacturing, vol. 39, pp. 154-163, 2008.
37. R. Zitoune and F. Collombet, “Numerical prediction of the thrust force responsible of delamination
during the drilling of the long-fibre composite structures,” Composites Part A, vol. 38, pp. 858-866,
D
2007.
38. J. P. Davim, P. Reis and C. C. Antonio, “Experimental study of drilling glass fiber reinforced plastics
(GFRP) manufactured by hand lay-up,” Composite Science and Technology, vol. 64, pp. 289-297,
TE
2004.
39. Singh, A. P., Sharma, M., & Singh, I. (2013). A review of modeling and control during drilling of fiber
reinforced plastic composites. Composites: Part B, 118-125.
EP
40. T. Sawa, T. Kobayashi and T. Fujii, “Strength of combination joints of an adhesive with bolts (T-
flange adherends subjected to an external bending moment),” JSME International Journal, Series 1:
Solid Mechanics, Strength of Materials, vol. 32, no. 3, pp. 411-419, 1989.
41. T. L. Anderson, “Mathematical Equations of Linear Elastic Fracture Mechanics,” in Fracture
C
Mechanics Fundamentals and Applications - Second Edition, Texas, CRC Press, 1995, pp. 101-116.
42. J. L. Clarke, Structural Design of Polymer Composites: EUROCOMP Design Code and Handbook, E &
AC
FN Spon.
17
ACCEPTED MANUSCRIPT
(a)
(b)
PT
(c)
RI
SC
Figure 1 Three major geometry configurations (tabs used for clamping shown for representation only); (a) Configuration
3 – Bonded Double Lap Joint, (b) Configuration 4 – Hybrid Double Lap Joint with three rivets, (c) Configuration 1 – Hybrid
Double Lap joint with six rivets; RR1 – Rivet Row One (closest fastener row to tapered edge); RR2 – Rivet Row Two
(furthest fastener row from tapered edge)
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Experiment
Finite Element Analysis - ABAQUS
Finite Element Analysis with 1.5mm Spew - ABAQUS
70
60
50
Load (kN)
40
PT
30
20
10
RI
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Displacement (mm)
SC
Figure 2 Load-displacement comparison between experimental and FEA for Configuration 1; displacement represents
the specimens total elongation for the given load
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
50 0.25
S33 @ 3000µƐ MR Rivet Load/Failure Load
40 S33 @ 13800µƐ MR
S13 @ 3000µƐ MR 0.20 Rivet Load/Applied Load
20 0.15
PT
10
0 0.10
-10
0.05
RI
-20
-30 0.00
0 10 20 30 40 50 0 10 20 30 40 50 60
SC
(a) Overlap Length (mm) (b) Load (kN)
Figure 3 Comparison of stress and load distribution in Configuration 1; (a) Stress distribution (b) Rivet load share; (S33 -
Normal Stress, S13 - Shear Stress, MR – Middle Region)
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Experiment
Finite Element Analysis - ABAQUS
50
40
Load (kN)
30
PT
20
10
RI
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Displacement (mm)
SC
Figure 4 Load-displacement comparison between experimental and FEA for Configuration 2; displacement represents
the specimens total elongation for the given load
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
60 7
S33 @ 215µƐ MR
S33 @ 280µƐ MR Rivet 1 & 3
50 6
S13 @ 215µƐ MR Rivet 2 & 5
S13 @ 280µƐ MR
5 Rivet 4 & 6
PT
30
3
20
2
RI
10 1
0 0
0 10 20 30 40 50 0 5 10 15 20 25 30 35
(a) Load (kN)
(b)
SC
Overlap Length (mm)
Figure 5 Comparison of stress and load distribution in Configuration 2; (a) Stress distribution (b) Rivet load share; (S33 -
Normal Stress, S13 - Shear Stress, MR – Middle Region)
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Experiment
Finite Element Analysis - ABAQUS
Finite Element Analysis with 1.5mm Spew - ABAQUS
70
60
50
Load (kN)
40
PT
30
20
10
RI
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Displacement (mm)
SC
Figure 6 Load-displacement comparison between experimental and FEA for Configuration 3; displacement represents
the specimens total elongation for the given load
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
50
S33 @ 3000µƐ MR
40 S33 @ 14500µƐ MR
S13 @ 3000µƐ MR
30 S13 @ 14500µƐ MR
Stress (MPa)
20
10
PT
-10
-20
-30
RI
0 10 20 30 40 50
Overlap Length (mm)
Figure 7 Comparison of stress in Configuration 3; (S33 - Normal Stress, S13 - Shear Stress, MR – Middle Region)
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Experiment
Finite Element Analysis - ABAQUS
Finite Element Analysis with 1.5mm Spew - ABAQUS
70
60
50
Load (kN)
40
30
PT
20
10
RI
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Displacement (mm)
SC
Figure 8 Load-displacement comparison between experimental and FEA for Configuration 4; displacement represents
the specimens total elongation for the given load
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
50 0.16
S33 @ 3000µƐ MR Rivet Load/Failure Load
40 S33 @ 13300µƐ MR 0.14
S13 @ 3000µƐ MR Rivet Load/Applied Load
20 0.10
PT
10 0.08
0 0.06
-10 0.04
RI
-20 0.02
-30 0.00
0 10 20 30 40 50 0 10 20 30 40 50 60
SC
(a) Overlap Length (mm) (b) Load (kN)
Figure 9 Comparison of stress and load distribution in Configuration 4; (a) Stress distribution (b) Rivet load share; (S33 -
Normal Stress, S13 - Shear Stress, MR – Middle Region)
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Experiment
Finite Element Analysis - ABAQUS
70
60
50
Load (kN)
40
30
PT
20
10
RI
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Displacement (mm)
SC
Figure 10 Load-displacement comparison between experimental and FEA for Configuration 5; displacement represents
the specimens total elongation for the given load
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
30
S33 @ 3000µƐ MR
S33 @ 13800µƐ MR
20
S13 @ 3000µƐ MR
S13 @ 13800µƐ MR
Stress (MPa)
10
PT
-10
-20
0 10 20 30 40 50
RI
Overlap Length (mm)
Figure 11 Comparison of stress in Configuration 5; (S33 - Normal Stress, S13 - Shear Stress, MR – Middle Region)
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Experiment
Finite Element Analysis - ABAQUS
70
60
50
Load (kN)
40
30
PT
20
10
RI
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Displacement (mm)
SC
Figure 12 Load-displacement comparison between experimental and FEA for Configuration 6; displacement represents
the specimens total elongation for the given load
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
30
S33 @ 3000µƐ MR
S33 @ 8300µƐ MR
20 S13 @ 3000µƐ MR
S13 @ 8300µƐ MR
Stress (MPa)
10
PT
-10
-20
RI
0 10 20 30 40 50
Overlap Length (mm)
Figure 13 Comparison of stress in Configuration 6; (S33 - Normal Stress, S13 - Shear Stress, MR – Middle Region)
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
40
Load (kN)
30
PT
20
10
RI
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Displacement (mm)
SC
Figure 14 Load-displacement comparison between experimental and FEA for Configuration 8; a combination of
Configuration 1 and 2 FEA results
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
400
PT
300
200
RI
100
SC
0
0 5 10 15 20 25 30 35 40
Crack Length (mm)
U
Figure 15 Available average energy for crack growth at different crack lengths subjected to a 3000µƐ displacement for a
Bonded, Hybrid, Hybrid A (Fasteners in row 1 removed) and Hybrid B (Fasteners and Holes in row 1 removed)
AN
Configuration; *Applicable to hybrid joints only; (RR1 – Rivet Row One, RR2 – Rivet Row Two)
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Figure 16 Separation of adherends with a 25mm long crack (5x deformation scale)
PT
RI
U SC
AN
M
D
TE
EP
C
AC
ACCEPTED MANUSCRIPT
Figure 17 Schematic of ‘Hybrid A’ with first row of rivets removed and all fastener holes still present
PT
RI
U SC
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Figure 18 Schematic of ‘Hybrid B’ with first row of rivets and fastener holes removed
PT
RI
U SC
AN
M
D
TE
CEP
AC