Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
1177/0021998309345295
AND JIN-HO
CHOI
School of Mechanical and Aerospace engineering, Research Center for Aircraft Parts
Technology, Gyeongsang National University, Jinju, Gyeongnam 660-701, Korea
INTRODUCTION
is the assembly of many parts such as skins, stiffeners,
frames and spars, etc. These parts must be connected through joints: mechanical or
bonded. Mechanical joints with fasteners such as bolts, screws, or rivets are simple and
widely used when disassembly for maintenance is necessary. However, the fasteners themselves are an important source of weight increase and the fastener holes induce stress
concentrations and consequently reduce the strength of joints.
N AIRCRAFT STRUCTURE
K.-H. NGUYEN
ET AL.
Adhesive bonding is another joining method that has been increasingly used. In adhesively bonded joints, there is no fastener at all and, therefore, no stress concentration due
to fastener holes. The typical failure modes of bonded joints are cohesive failure, which is
failure within the adhesive, interfacial failure (failure along the interface between adherend
and adhesive), or adherend failure. Cohesive failure and interfacial failure are sometimes
referred to as bond line failure. Bond line failure is the typical failure mode of metallic
bonded joints. Adherend failure is mainly found in joints using composite laminate as the
adherend. The adherend failure modes in joints with composite adherend are very complicated, and include such failures as matrix failure, fiber failure, and interlaminar and
intralaminar failures.
The research on bonded joints has a long history and has been conducted experimentally
and/or numerically. Researchers have targeted stress analysis, failure mode and strength
prediction, strength improvement, and the effects of various parameters such as material,
geometry and bonding method, etc. One difficulty for failure load prediction in a bonded
joint comes from the presence of the singularity at the ends of overlapped area. Harris and
Adams [1] conducted finite element analysis to predict the failure mode of metallic singlelap bonded joints. They took into account the geometrical and material nonlinearity of
single-lap joints. Their failure prediction method was based on material strength. The
failure was assumed to occur when the maximum principal strain or maximum principal
stress at one Gauss point close to the singular position inside the adhesive layer attained
the ultimate stress or strain of the adhesive material. The method, therefore, is dependent
on mesh refinement near the singularity point. The authors first applied this method to
predict the failure mode and failure load of metal-to-metal single-lap joints [1] and then
extended their work to lap joints with composite adherends [2]. The point-based method
proposed by Harris and Adams was shown by Kairouz and Matthews [3] to be useful in
predicting the failure mode and strength of bonded single-lap joints made of cross-ply
laminated adherends.
To overcome the singularity problem, singularity parameter approaches [4,5] were also
used. A generalized stress intensity factor and a parameter called strength of the singularity were defined and used successfully in the method for the prediction of fractures. In
addition, the approach was used for the prediction of fatigue crack initiation in adhesive
bonds [6,7]. Ishii et al. [8] considered the concentrated multiaxial stress state in the adhesive layer in their analysis and proposed a method based on two-singularity parameters to
estimate the fatigue strengths of adhesively bonded joints made of carbon fiber reinforced
polymer and aluminum alloy. However, the singularity parameter approach did not consider the nonlinearity of the adhesive layer and therefore the obtained generalized stress
intensity factor may not be correct when the adhesive layer shows highly nonlinear behavior. The method also requires additional tests to define the fracture criteria using stress
singularity parameters.
Crocombe [9] proposed another method to predict the bond line failure of single-lap
bonded joints. Failure was assumed to occur as whole adhesive layers become plastic.
It was shown that this method can make a good estimation of joint strength for a
wide class of joints. In some cases, however, this approach can be incorrect because
local failure can occur before global yielding. However, the method contributed to establishing the concept of the damage zone method, in which a joint failure occurs after
adhesion in some area fails rather than after adhesion fails at a certain point. Clark and
McGregor [10] not only applied the point-based method but also proposed an approach
based on ultimate stress over a zone to predict the failure load and failure mode of
single-lap bonded joints. They showed that while the point-based method predicted lower
failure loads and failure initiation near the singularity, in contrast to observed experiments, the damage zone-based method gave good predictions of failure loads of different
joint geometries.
As bonded joints with composite adherend are used, interlaminar failure (delamination)
is usually found over the failure surfaces. Interlaminar failure is caused by weakness of
composite adherends in the through-the-thickness direction. Assuming that both cohesive
and out-of-plane adherend crack initiation of adhesively bonded joints will occur after a
damage zone develops, Sheppard et al. [11] proposed a damage zone model to predict the
joint failure loads of different geometries. Using their developed method, the authors
showed that failure load predictions of aluminum joints and composite joints were
within the experimental scatter range.
Based on the fact that the intra/interlaminar failures occur in the layer close to the
adhesive, Tong [12] conducted a 2D analysis and tried six stress-based criteria to predict
the failure loads of double-lap bonded joints. The stress was taken at the center of one ply
element or one adhesive element at the free end where stresses were highly concentrated.
The method is a point-based method and can be dependent on meshing near the singularity. In addition, 3D effects such as the free edge effect, the anticlastic effect, and the
bending-twisting coupling effect can play important roles in the failure initiation and
failure load [13]. To overcome the problem of singularity, Kim et al. [14] used a characteristic length method, which is widely used in failure prediction of mechanical joints.
2D analysis considering geometrical and material nonlinearity was conducted with the
use of a global yielding criterion for the adhesive layer and a quadratic delamination
criterion, which was proposed by Brewer and Palage [15], for the composite adherend.
Interfacial stresses in the specimens were calculated under the test failure loads. The characteristic length was determined as the distance from the overlapped end along the interface to a location of the finite element model in which the quadratic delamination criterion
was satisfied. The optimal joint strength was found and a new joint strength improvement
technique was also suggested.
Shin and Lee [16] performed a 3D analysis considering the thermal load of co-cured
single-lap and double-lap joints without any additional adhesive. They predicted failure
load of the joints considering two criteria: the Ye-delamination failure criterion and the 3D
TsaiWu failure criterion. In the case of single-lap co-cured joints, failure loads predicted
by the Ye-criterion were in good agreement with the experimental results. Otherwise, using
the TsaiWu criterion was better for failure load prediction for double-lap bonded joints.
Other research [1720] has utilized fracture mechanics to predict the failure loads of
bonded joints. An initial crack is usually assumed to exist in the adhesive, at the interface
of the adherend/adhesive, or in the composite adherend. The crack propagates as the
strain energy release rate exceeds a critical value. The strain energy release rate can be
computed by the virtual crack closure technique in conjunction with finite element analysis. However, this energy-based approach relies on the existence of a crack in the interface, and on the assumption of small-scale bridging and linear elasticity. If any of these
conditions are violated, an alternative approach such as cohesive zone modeling is
required [20]. The cohesive zone model, however, has some limitations such as mesh
sensitivity, lack of convergence, computing inefficiency, and so on. An improvement of
the cohesive zone model was done with the implementation of the discrete cohesive zone
model by Xie and Waas [21]. The authors showed that this model is not sensitive to the
mesh size and the load increment. Computation time was also reduced.
K.-H. NGUYEN
ET AL.
Progressive failure is another method to predict the strength of bonded joints. Apalak
and Apalak [22] performed a progressive damage analysis for unidirectional composite
single-lap joints in tension with the Hashin failure criteria. After the damage modes were
checked in each ply of the composite plate, degradation of failed plies was applied based
on a set of suitable rules.
The main objective of this article is to develop an efficient failure prediction method
based on material strength for single-lap joints with different geometries and materials.
While most previous researchers have conducted 2D analyses, a damage zone method
based on 3D failure analysis is proposed. Interlaminar failure is considered as the failure
of interply resin-rich layers inside the composite adherend. Adhesive is assumed to fail by
yielding and the material nonlinearity of aluminum is considered. Predicted failure loads
are compared with the experimental results in the authors previous paper [23].
(a)
Aluminum
Composite
tC
25
tA
40 mm
50 mm
(b)
50 mm
40 mm
70
Stress (MPa)
60
50
40
30
20
10
0
0.00
0.01
0.02
Strain
0.03
0.04
Tensile modulus
Shear modulus
Poissons ratio
Tensile strength
Shear strength
162
9.6
9.6
6.1
6.1
3.5
0.298
0.298
0.47
2552
43
43
94
94
40
AL2024-T3
FM73m
73
2.8
0.33
0.38
500
Stress (MPa)
400
300
200
100
0
0.00
0.01
0.02
0.03
0.04
0.05
0.06
Strain
Figure 2. Tensile stressstrain curve of (a) FM73m adhesive, and (b) aluminum 2024-T3.
The former was used to manufacture the thin specimens and the latter to make the thick
specimens. The aluminum adherend was made with anodized aluminum 2024-T3 and
bonded to a composite adherend using adhesive FM73m by Cytec. The mechanical properties of the composite unidirectional prepreg (USN125 by SK Chemicals), adhesive layer,
and aluminum adherend are given in Table 1. The thicknesses of the anodized aluminum
2024-T3 adherend are 1.58 mm and 3.01 mm. The stressstrain curve for the tension test of
the adhesive experimentally obtained is shown in Figure 2(a), and that of the aluminum is
shown in Figure 2(b) [24].
Detailed dimensions of the specimens are given in Table 2 and experimental failure loads
are shown in Figure 3. The experimental result shows that higher overlap length yields
K.-H. NGUYEN
ET AL.
b (mm)
Al
Composite
FM73m
Total
No. of specimens
15
20
25
30
35
40
15
25
35
1.58
1.58
1.58
1.58
1.58
1.58
3.01
3.01
3.01
1.68
1.68
1.68
1.68
1.68
1.68
3.38
3.38
3.38
0.112
0.123
0.143
0.132
0.137
0.199
0.168
0.187
0.193
3.372
3.383
3.403
3.392
3.397
3.459
6.558
6.577
6.583
5
5
5
5
5
6
6
6
6
25
21.6
18.2
20
14.2
16.4
16.5
14.7
15
12.8
11.9
10.7
10
0
FM15
FM20
FM25
FM30
FM35
a higher failure load. Increase of the specimen thickness also affects the failure load of
single-lap bonded joints. However, failure loads increase only 1232% as the thicknesses
of the specimens nearly double.
In a metal-to-metal joint, a bond line failure is the typical failure mode. However, the
weakness of the composite material in the out-of-plane direction leads to a different failure
mode. Typical failure surfaces of specimens are shown in Figure 4. The failure surface of
the specimens is dominated by the delamination of composite adherend, and intralaminar
failures are locally observed. In addition to out-of-plane failure of the composite adherend,
partial bond line failure is also found.
FM25
FM25D
Figure 4. Typical failure surface of the joint FM25 (left) and FM25D (right).
z
Y
X
0.4 mm
0.4 mm
K.-H. NGUYEN
ET AL.
Y
X
Distribute
uniformly
Fixed
Z
X
b
Figure 6. Boundary conditions for the finite element analysis.
can be considered as stresses on the interface between two adjacent plies or as stresses
inside these thin resin layers. Fenske and Vizzini [26], when investigating the delamination
of a plate subjected to an axial strain, found that the onset of delamination can be predicted by considering the failure of interply resin layers.
In this article, the composite adherend is modeled as a combination of orthotropic plies
and isotropic interply resin layers. An interply resin layer is modeled to exist between any
two adjacent orthotropic plies. Consequently, delamination is considered as the failure of
these interply layers. Moduli of the interply resin layer are assumed to be the same as the
matrix properties of an orthotropic lamina. A schematic cross section at the overlapped
area of the joint is shown in Figure 7. As in the work of Fenske and Vizzini, all interply
resin layers are assumed to have a constant thickness that is equal to 10% of one orthotropic ply thickness.
where DA and CDA are the total damage area inside a specimen and a critical damage
area, respectively.
The second approach is the weighted damage area method proposed by Choi and Chun
[27] to predict the failure load of mechanical joints. This method considers not only the
damage area but also the magnitude of the failure index by using a given failure criterion
and the geometrical effects. The joint is assumed to fail as the weighted damage area ratio
equals a critical value:
P n
FI DA
n
CDARn
2
DAR
AG
where DAR n, FI, DA, AG, n, and CDAR n are the damage area ratio, the failure index by a
failure criterion, the damage area, the area that is responsible for the geometrical effects,
the weighting power factor, and the critical damage area ratio, respectively.
45
0
Interply
resin layers
90
45
45
Adhesive layer
Y
Aluminum adherend
In the above two approaches, the damage area is the main concern. As a 3D analysis is
conducted in this article, the damage volume approach can be obtained by expanding the
damage area concept. Simply put, damage volume can be used instead of damage area.
Equation (1) can be rewritten as follows:
DV CDV
where DV and CDV are the total damage volume inside a specimen and a critical damage
volume, respectively.
A weighted damage volume approach is given in Equation (4):
P n
FI DV
n
DVR
CDVRn
4
VG
where DVRn, FI, DV, VG, n, and CDVRn are the damage volume ratio, the failure index
by a failure criterion, the damage volume, the volume the relates to geometrical effects,
the weighting power factor, and the critical damage volume ratio, respectively.
It is noted that the damage area is the in-plane damage area in the adhesive, interply
resin layers, and orthotropic plies. The area AG and the volume VG should account for the
differences in geometry of the joints such as bonded length, thickness, and width. The
details of AG and VG are given in the next chapter. The weighting power factor n is an
integer such as 0, 1, or 2. When n equals 0, the failure index has no effect on the damage
volume ratio (DVR).
The general procedure to predict failure load of the joints was reported in Sheppard
et al. [11] and is rewritten here with a difference in the damage zone size, which may be the
size of the damage area, the damage volume, the damage area ratio, or the damage volume
ratio:
(1) Test one or more bonded joint(s) to record the failure load(s) and mode(s).
(2) Analyze the joint(s) under the experimental failure load(s) using an appropriate analysis tool.
10
K.-H. NGUYEN
ET AL.
(3) Use appropriate failure criterion and the relevant material allowable(s) to calculate the
damage zone size(s) in the joint and choose a value to be the critical value.
(4) Use the critical damage zone size calculated in the previous step to predict the critical
load of bonded joints with similar adherends, adhesives, and load paths.
FAILURE CRITERIA
Failure criteria selection depends on the failure mode of the specimens. The failure
mechanism of the bonded single-lap joint was out-of-plane failure and partial bond-line
failure. Based on the experimental failure modes, the Ye-delamination criterion [28] was
applied to the interply resin layers to take into account the interlaminar failure, and was
also applied to the orthotropic plies to account for the intralaminar transverse failure.
The Ye-delamination criterion is as follows:
8
2
2
2
>
33
13
23
>
>
< 2 2 2 1 33 0
Z
S13 S23
5
2
2
>
23
13
>
>
1
5
0
33
: 2
S13 S223
where 33 , 13 , 23 , Z, S13 , and S23 are the peel stress in interply layer or orthotropic ply,
out-of-plane shear stresses in interply layer or orthotropic ply, interlaminar normal
strength, and transverse shear strengths, respectively.
The adhesive used in this article shows a nonlinear stressstrain curve. Consequently,
the Von Mises strain criterion is applied to the adhesive layer:
p q
2
"VM
6
"1 "2 2 "2 "3 2 "3 "1 2
3
where e1, e2, and e3 are the principal strains.
11
200
180
0
20
40
60
80
100
120
140
160
180
200
160
140
120
100
80
60
40
1.0
20
0.0
2 X/b
0.5
0.5
1.0
/W
0.0
0.5
2Y
0
1.0
0.5
1.0
Figure 8. Von Mises stress distribution in the adhesive of the joint FM15 without considering material
nonlinearity.
180
Linear elastic adhesive
160
Nonlinear adhesive
140
120
100
80
60
40
20
0
1.0
5.0
0.0
0.5
1.0
2 X/b
Figure 9. Von Mises stress distribution along the center of the adhesive of joint FM15 with and without
considering material nonlinearity (Y Z 0).
12
K.-H. NGUYEN
ET AL.
Y
b
X
Figure 10. Damage area in the adhesive of joint FM15 according to Von Mises strain criterion.
The Von Mises stress distribution along the centerline in the X-direction of the adhesive
(Y Z 0) on the overlapped area (2X/b is from 1 to 1) considering the material nonlinearity is compared with the results of the linear elastic adhesive model in Figure 9. While
the stress in the elastic adhesive model shows peaks at the ends of the bonding area on
both overlap free ends, the stress of the nonlinear adhesive model is nearly constant in the
end areas. Consequently, the inner overlapped area (far from ends) carries a larger load
and then shows higher stress compared with the results of the elastic adhesive model.
Figure 10 shows the failure areas considering the material nonlinearity of the adhesive,
based on the Von Mises strain criterion. As expected, the failure of the adhesive is focused
over the end areas of the overlapped area where stresses are highly concentrated. The
failure area is slightly larger in the aluminum end area than in the composite end area.
Damage Area in Interply Resin Layers
Figure 11(a) shows the delamination failure index distribution by the Ye-criterion in
the interply layer between the first 45 and the first 45 layers of joint FM15 from the
adhesive layer. Obviously, along the ends of the overlap area and free side edges of the
composite adherend, the stress is highly concentrated. Consequently, a high failure index
and large damage area are predicted over the regions. Pagano [29] reported that the high
stress along the free side edges area can be attributed to the stress singularity along the
edges. Figure 12 shows the failure index in the same layer at various positions of X- along
the Y-axis. The figure obviously shows that the failure index is high at the overlapped area
ends (2X/b 1.0) and shows a peak value at the very limited free side edges.
The delamination failure index of the second (between first 45 and 90 layers) and
the third interply layers (between the first 90 and 0 layers) of joint FM15 are shown in
Figure 11(b) and (c), respectively. Similarly, the damage area and high failure index
are found near the ends of the bonded area and free side edges. It is interesting that
failure indices are very high along the free side edges of the non-overlapped region of
the composite adherend. However, delamination was not visually observed over these
13
Ye failure index
0
2
4
6
8
6
4
1.0
0.5
0.0
0
2.0
1.5 1.0
0.5
2 X/b
0.5
0.0
0.5
1.0
2Y/W
1.0
(b)
0
2
4
6
8
6
4
1.0
0.5
2
0.0
0
2.0 1.5
2Y/W
Ye failure index
0.5
1.0 0.5
2 X/b
0.0
0.5
1.0
1.0
Figure 11. Ye failure index in interply layers (a) between the first 45 and 45 layers, (b) between the first
45 and 90 layers, and (c) between the first 90 and 0 layers.
areas in the experiment. As noted by Pagano [29], it is guessed that there is a singularity
along the free side edges between two plies. Therefore, a high-density mesh was created at
the free side edges. Consequently, high stresses were obtained at these singular positions.
However, the delamination should be predicted by the stress at some distance from the
singularity rather than at the singularity point itself because the singularity occurs in a very
limited area. Therefore, even though the failure index obtained is higher than unity at both
free edges far away from the bonding area, this does not mean that delamination occurs
there experimentally.
The peak failure index in Figure 11(a) is smaller than that in Figure 11(b) and (c). This
can be explained in terms of the compressive interlaminar peel stress at the free side edges
14
K.-H. NGUYEN
ET AL.
(c)
0
2
4
6
8
6
4
1.0
0.5
0.0
0
2.0
1.5 1.0
0.5
0.5
0.0
2 X/b
0.5
1.0
2Y/W
Ye failure index
1.0
1.4
1.2
Failure index
1.0
Y
1.0b 1.5b
0.8
X = 0.5b
X = 0.75b
X = 1.0b
0.6
0.4
b
0.75b
0.2
0.0
1.0
0.5
0.0
0.5
1.0
2 Y/W
Figure 12. Ye failure indices in the interply layer between the first 45 and 45 layers.
of the interface between the 45 and 45 layers. As shown by Yang and He [30], the free
side edges between the 45 and 45 layers of a composite plate of [ 45]S subjected to an
axial strain experience compressive interlaminar peel stresses. Consequently, this compressive stress reduces the delamination possibility and failure index.
The next interply between 45 and 45 layers is further from the adhesive layer than
the first interply. Therefore, the failure index and the total damage area in this interply
layer were found to be smaller than those in the interply layer between the first 45 and the
first 45 layer. Similar phenomena were observed in the case of the interply layer between
45 and 90 layers and also in that between 90 and 0 layers.
15
0
2
4
6
8
1.0
0.5
0.0
0
2.0
0.5
1.5 1.0
0.5
2 X/b
0.0
0.5
2Y/W
Ye failure index
1.0
1.0
(b)
0
2
4
6
8
6
4
1.0
0.5
0.0
0
2.0 1.5
2Y/W
Ye failure index
0.5
1.0 0.5
2 X/b
0.0
0.5
1.0
1.0
Figure 13. Ye failure index in the first (a) 45 layer, (b) 45 layer, (c) 90 layer, and (d) 0 layer of the joint FM15.
From the analysis, the failure area in the interply layers of the composite adherend,
which indicates interlaminar failure, is concentrated at both ends of the bonding area and
also at both edges of the composite adherend.
Damage Area in Orthotropic Plies
The Ye-criterion was also applied to the orthotropic plies to predict the intralaminar
transverse failure of joints. As shown in Figure 13, the peaks of the failure index
distribute at the ends of the bonding area (2X/b 1.0 and 1.0) and the free side edges
(2Y/W 1.0 and 1.0) of joint FM15. Maximum peaks are mainly found at the free side
edges, particularly, slightly outside the aluminum edge (where 2X/b is around 1.1 to
1.2). However, the region of high failure index at the free side edges is very limited
compared with the failure region over the end area of the joint. The failure index in the
16
K.-H. NGUYEN
ET AL.
(c)
Ye failure index
0
2
4
6
8
6
4
1.0
0.0
0
2.0
0.5
1.5 1.0
0.5
2X/b
0.0
0.5
2Y/W
0.5
1.0
1.0
(d)
0
2
4
6
8
6
4
1.0
0.5
0.0
0
2.0 1.5
2Y/W
Ye failure index
0.5
1.0 0.5
2X/b
0.0
0.5
1.0
1.0
90 orthotropic layer shown in Figure 14 is an example. The failure index and damage area
in the first 45 orthotropic layer are larger than those in the second 45 layers in the same
laminate, which are further from the adhesive layer. The other layers with the same fiber
angle show the same trend as that of the 45 layers. These phenomena are also found in the
other joints with different overlap lengths or joint thicknesses.
The summation of the damage area in the composite laminate and adhesive layer is
given in Table 3. For thin adherend joints (FM15FM40), interply failure contributes to
the total sum of the failure area more than does intralaminar failure of the orthotropic
plies. For thick joints (FM15DFM35D), however, intralaminar failure affects the most
joints, and is followed by adhesive and interplay failures. Considering these results, it can
be deduced that none of the three kinds of failures can be neglected when calculating the
failure zone. In joint FM30, the experimental failure load slightly deviated from the trend,
as shown in Figure 3. Because of the smaller failure load compared with those of the
17
Y
1.0b 1.5b
X = 0.5b
X = 0.75b
X = 1.0b
Failure index
4
b
0.75b
2
1.0
0.5
0.0
0.5
1.0
2 Y/W
Figure 14. Ye failure index in the first 90 layer of the joint FM15.
Interply layer
Orthotropic ply
Adhesive layer
Sum
57.4
55.8
65.8
54.9
85.4
80.8
90.4
216.9
208.2
89.4
80.9
63.8
39.8
49.6
45.4
141.6
328.5
280.3
117.5
160.0
178.0
133.0
187.0
166.3
106.2
235.5
256.9
264.3
296.7
307.6
227.7
322.0
292.5
338.2
780.9
745.4
adjacent joints FM25 and FM35, a smaller damage zone is predicted for FM25 than for
those other joints.
Failure Load Prediction
From the analysis results, the damage area in the interply layers, orthotropic plies, and
adhesive of each joint are obtained. As observed in the experiment, the failure modes are
mostly delamination and adhesive debonding. Although aluminum adherend experiences
plastic deformation, aluminum failure (cracking) was not found. Consequently, the
damage area affecting the bonded joint failure is assumed to be the sum of the transverse
damage area in the interply resin layers, orthotropic plies, and adhesive layers.
A critical step in predicting the failure load of bonded joints by the damage zone method
is to choose a critical damage area, which is the damage area corresponding to structural
failure. Figure 15 shows results of the failure load prediction when the critical damage
18
K.-H. NGUYEN
ET AL.
30,000
Experiment failure load
CDA = 307.6 mm2 (FM25)
CDA = 780.9 mm2 (FM25D)
25,000
20,000
15,000
10,000
5000
0
FM15
FM20
FM25
FM30
FM35
Joint
Figure 15. Predicted failure loads using damage area approach with two different CDA values.
areas of 307.6 and 780.9 mm2 are used, which are the damage areas at the failure loads of
joints FM25 and FM25D, respectively. Predicted failure loads when a critical damage area
(CDA) of 307.6 mm2 is used are closer to the experimental results, except for joints
FM25D and FM35D. On the other hand, when CDA is set at 780.9 mm2, better results
are found in joints FM25D and FM35D. This means that when a thicker specimens data
(FM25D) is used, big deviations are found in the thinner joint and vice versa.
Consequently, it can be deduced that the damage zone approach is not robust without
considering the adherend thickness effect.
In the discussion of the damage areas given in Table 3, it was noted that the total
damage area of joint FM30 was out of trend and that the reason for this was the lower
experimental failure load. In the predicted failure load data shown in Figure 15 as well, the
same phenomenon is found for joint FM30. Moreover, it is found that the predicted
failure load of joint FM15D is out of trend compared with that of other thick adherend
joints (FM25D and FM35D). This is also attributed to the lower experimental failure load
of the joint. Observation of the two joints FM30 and FM15D suggests the possibility that
the failure loads of the joints were underestimated in the experiment.
As mentioned above, Figure 15 shows that adherend thickness should be considered as a
parameter affecting failure loads. Failure loads are also much different depending on the
critical damage area (CDA). To consider the magnitude of the failure index and the joint
geometry, Equation (2) was proposed as a failure criterion. In the equation, the failure
index is included as a weight factor for the damage area and the area AG, which is defined
in Equation (7):
p
tC bw if b w
AG
7
tC w
if b 4 w
19
25,000
20,000
15,000
10,000
5000
0
FM15
FM20
FM25
FM30
FM35
Joint
Figure 16. Predicted failure loads using weighted damage area approach with two different CDAR1 values.
where tC, b, and w are the thickness of composite adherend, overlap length, and joints
width, respectively.
Figure 16 shows the failure loads predicted using the weighted damage area method in
Equation (2) when the critical damage area ratio CDAR1 was set at 10.58 and 12.63, which
are the experimental failure load values corresponding to joints FM25 and FM25D,
respectively. Although the results using a CDAR1 of 10.58 show better agreement with
the experimental results than do the results obtained using the other value, except for the
thicker joints FM25D and FM35D, the results in both cases show quite good agreement
with the experimental failure loads. The relatively larger deviation between the predicted
and experimental failure loads is also found here for joints FM30 and FM15D.
Early in this investigation, the authors tried another definition of AG, which is shown in
Equation (8). However, the predicted failure loads show larger deviation from the test
results than when AG in Equation (7) was used:
p
w tC b if b w
p
AG
8
w tC w if b 4 w
where tC, b, and w are the thickness of the composite adherend, the overlap length, and the
joints width, respectively.
The weighted damage area ratios DARn and the weighted damage volume ratios by
Equation (4) DVRn of the joints with various weighting power factors n are summarized in
Table 4. In all the methods, the results of joints FM30 and FM15D are out of trend. The
differences between the experimental failure loads and those predicted by the weighted
damage area method are summarized in Table 5 when the weight power is n 1. As shown
in the table, predicted failure loads are always within 15.6% of experimental result
20
K.-H. NGUYEN
ET AL.
Table 4. Damage area, weighted damage area ratios, damage volume, and weighted
damage area ratios of the joints subjected to experimental failure loads.
DARn
ID
FM15
FM20
FM25
FM30
FM35
FM40
FM15D
FM25D
FM35D
Failure
load (N)
Damage
area (mm2)
nV0
nV1
10,676
12,874
14,737
14,202
16,380
16,513
11,926
18,185
21,591
264.3
296.7
307.6
227.7
322.0
292.5
338.2
780.9
745.4
8.22
7.99
7.41
5.49
7.76
7.05
5.26
9.41
8.98
11.27
10.99
10.58
7.63
11.69
10.43
6.87
12.63
12.01
DVRn
nV2
Damage
volume (mm3)
nV0
nV1
nV2
16.85
16.56
16.86
11.64
19.81
17.46
9.36
18.27
17.46
22.2
27.9
32.1
21.9
31.1
29.0
32.5
78.0
78.8
0.036
0.034
0.031
0.021
0.030
0.028
0.026
0.038
0.038
0.051
0.048
0.046
0.030
0.048
0.043
0.035
0.053
0.054
0.080
0.075
0.078
0.048
0.086
0.077
0.051
0.082
0.085
Table 5. Difference (%) between the predicted and experiment failure loads by the
weighted damage area method with n V 1.
CDAR1
11.27
(FM15)
10.99
(FM20)
10.58
(FM25)
7.63
(FM30)
11.69
(FM35)
10.43
(FM40)
6.87
(FM15D)
12.63
(FM25D)
12.01
(FM35D)
FM15
FM20
FM25
FM30
FM35
FM40
FM15D
FM25D
FM35D
Max
Min
0.0
0.8
1.4
10.4
1.0
1.7
12.0
4.5
1.8
12.0
4.5
1.3
0.0
0.7
9.8
1.6
1.3
10.0
5.2
2.6
10.0
5.2
2.4
1.1
0.0
8.8
2.4
0.6
9.0
6.2
3.7
9.0
6.2
11.8
10.4
8.2
0.0
9.6
5.5
0.7
13.6
11.8
0.7
13.6
0.4
1.8
2.3
11.3
0.0
2.4
9.5
3.5
0.7
11.3
3.5
2.8
1.4
0.5
8.5
2.7
0.0
7.2
6.6
4.1
8.5
6.6
14.8
13.4
10.9
2.0
11.9
7.5
0.0
15.6
14.0
0.0
15.6
2.7
4.0
4.2
13.2
1.5
3.9
11.1
0.0
1.9
13.2
0.0
1.2
2.6
2.9
12.0
0.4
2.9
10.1
2.6
0.0
12.0
2.6
regardless of the CDAR1 value. When the two joints FM30 and FM15D are not considered, the maximum deviation decreases to 6.6%.
The accuracies of the predicted failure loads using the weighted damage area method
with the other weighting power factors are shown in Tables 6 and 7. As n is set at 0, the
maximum deviation between the predicted and experimental results is slightly larger than
that when n is 1. On the contrary, the results for n 2 show slightly better agreement with
the experiment, where the maximum deviation is 15%. It should be also noted that the
maximum deviations are always related to joints FM30 and FM15D. When the two
joints FM30 and FM15D are not considered with n 2, the maximum deviation decreases
to 3.7%.
From the 3D finite element model, damage volume can be obtained and used instead of
damage area to predict the failure load. Basically, damage volume must be proportional to
the damage area. However, these factors are not linearly proportional because the interply,
adhesive, and orthotropic laminas have different thicknesses.
21
8.22
(FM15)
7.99
(FM20)
7.41
(FM25)
5.49
(FM30)
7.76
(FM35)
7.05
(FM40)
5.26
(FM15D)
9.41
(FM25D)
8.98
(FM35D)
FM15
FM20
FM25
FM30
FM35
FM40
FM15D
FM25D
FM35D
Max
Min
0.0
1.3
2.9
13.6
1.4
4.0
11.6
4.3
2.3
13.6
4.3
1.5
0.0
2.0
12.8
0.6
3.3
10.8
5.1
3.2
12.8
5.1
3.8
2.2
0.0
10.5
1.5
1.6
8.6
7.2
5.5
10.5
7.2
13.3
12.2
9.2
0.0
10.1
5.1
0.0
14.0
12.9
0.0
14.0
2.4
0.6
1.2
11.8
0.0
2.7
10.0
6.0
4.1
11.8
6.0
5.4
3.9
1.7
8.9
3.0
0.0
7.2
8.5
6.9
8.9
8.5
14.6
13.6
10.5
0.0
11.3
6.0
0.0
14.8
13.8
0.0
14.8
6.0
5.8
6.9
17.7
5.2
7.0
17.5
0.0
2.3
17.7
0.0
5.0
4.2
5.5
16.3
3.9
5.9
16.0
1.6
0.0
16.3
1.6
Table 7. Difference (%) between the predicted and experiment failure loads by the
weighted damage area method with n V 2.
CDAR2
12.85
(FM15)
16.56
(FM20)
16.86
(FM25)
11.64
(FM30)
19.64
(FM35)
17.46
(FM40)
9.36
(FM15D)
18.27
(FM25D)
17.46
(FM35D)
FM15
FM20
FM25
FM30
FM35
FM40
FM15D
FM25D
FM35D
Max
Min
0.0
0.1
0.2
7.4
3.0
0.2
11.0
3.3
1.5
11.0
3.3
0.6
0.0
0.5
7.1
3.3
0.5
7.1
3.7
2.0
7.1
3.7
0.2
0.1
0.0
7.5
3.0
0.2
7.3
3.3
1.5
7.5
3.3
9.4
8.4
7.3
0.0
9.1
5.7
1.9
11.4
10.5
1.9
11.4
3.8
3.9
3.0
10.6
0.0
2.2
12.0
1.3
3.7
12.0
0.0
0.7
0.9
0.5
8.1
2.4
0.0
7.9
2.3
0.4
8.1
2.4
14.8
13.5
11.5
3.9
12.6
8.9
0.0
15.0
14.4
0.0
15.0
1.8
2.0
1.4
9.0
1.7
1.0
11.6
0.0
1.0
11.6
1.7
0.7
0.9
0.5
8.1
2.4
0.3
10.5
2.3
0.0
10.5
2.4
Similar to the area AG, the volume VG, relating to geometrical effects, in Equation (4)
was defined in Equation (9).
t b w if b w
VG C
9
tC w w if b 4 w
where tC, b, and w are the composite adherend, the overlap length, and the width of
specimens, respectively.
Originally, the authors tried to find damage volume ratios with the volume VG of
thickness width overlap length regardless of the ratio of the overlap length to width.
However, it was impossible to find meaningful damage volume ratios with such a defined
volume VG. This means that the damage volume or area is not proportional to the overlap
length in this analysis method where the progressive stiffness degradation and damage
zone propagation are not considered.
22
K.-H. NGUYEN
ET AL.
30,000
Experiment failure load
CDV = 32.1 mm3 (FM25)
25,000
20,000
15,000
10,000
5000
0
FM15
FM20
FM25
FM30
FM35
Joint
Figure 17. Predicted failure loads using damage volume approach with two different CDV values.
Figure 17 shows the failure loads predicted using the damage volume method when the
critical damage volume is set at 32.1 and 78 mm3, which are the damage volumes of the
FM25 and FM25D joints, respectively. The predicted failure loads are much different
from the experimental result when CDV equals 78 mm3. Similar to the damage area
approach, the damage volume approach is not robust and the thickness difference
should be considered while predicting the failure load of bonded joints.
Figure 18 shows the predicted failure loads when CDVR1 equals 0.046 and 0.053, which
are the weighted damage volume ratios at the experimental failure loads of the FM25 and
FM25D joints, respectively. It is shown that when the thickness effect is considered, the
predicted failure loads show good agreement with the experimental values. The maximum
deviation between them is about 13%, as shown in the figure.
The predicted failure loads based on the weighted damage volume ratios are given in
Tables 810 with n 0, 1, and 2, respectively. As shown in the tables, the maximum
deviations are 17.8%, 16.6%, and 12.9%, respectively. The method with n 2 predicts
the failure loads the best. Once more, the maximum deviations are found when the damage
volume of joint FM30 is used as the critical value for failure evaluation.
Summarizing the finite element results, the damage volume ratio method gives the best
prediction of the failure loads. Among a total of nine joint specimens, the experimental
failure loads of joints FM30 and FM15D were out of trend and therefore resulted in large
deviations of failure load prediction. Without these two joints, the maximum deviation
was reduced to 4.1%, as shown in Table 10. It should also be noted that both the weighted
damage area and volume ratio methods are based on the 3D finite element analysis results.
To predict out-of-plane failure, 3D analysis is essential.
To investigate the effect of the mesh density of the model on the failure load prediction,
refined mesh models were created. The interlaminar stresses obtained at the free side
23
25,000
20,000
15,000
10,000
5000
0
FM15
FM20
FM25
FM30
FM35
Joint
Figure 18. Predicted failure loads using weighted damage volume approach with two different CDVR1
values.
Table 8. Difference (%) between the predicted and experiment failure loads by the
weighted damage volume method with n V 0.
CDVR0
0.036
(FM15)
0.034
(FM20)
0.031
(FM25)
0.021
(FM30)
0.030
(FM35)
0.028
(FM40)
0.026
(FM15D)
0.038
(FM25D)
0.038
(FM35D)
FM15
FM20
FM25
FM30
FM35
FM40
FM15D
FM25D
FM35D
Max
Min
0.0
1.8
4.3
16.0
4.6
6.1
4.8
1.3
1.3
16.0
1.3
3.3
0.0
2.7
14.4
3.0
4.4
3.5
3.0
3.1
14.4
3.3
5.8
2.7
0.0
12.1
0.7
2.3
1.7
5.3
5.4
12.1
5.8
17.6
12.2
10.5
0.0
9.9
5.8
6.8
13.9
14.1
0.0
17.6
6.7
3.6
0.5
11.2
0.0
1.5
1.0
6.1
6.2
11.2
6.7
8.9
5.5
2.5
9.2
2.1
0.0
0.6
7.9
8.0
9.2
8.9
11.1
7.4
4.5
7.2
4.0
1.7
0.0
9.6
9.7
7.2
11.1
0.2
3.7
5.8
17.5
6.0
7.7
5.9
0.0
0.4
17.5
0.0
0.5
4.0
6.1
17.8
6.3
8.0
6.2
0.7
0.0
17.8
0.0
edges can be affected by the mesh density there. Therefore, the number of elements along
the width of the joints increased (from 20 to 30 elements) and the smallest width (along Ydirection) of a single element was reduced to 0.024 mm. The procedure of failure load
prediction using damage volume ratio was done again to predict the failure load of joint
FM25, which experimentally fails at 14,737 N. The results are given in Table 11. As shown
in the table, the critical damage volume ratios can be slightly changed but the predicted
24
K.-H. NGUYEN
ET AL.
Table 9. Difference (%) between the predicted and experiment failure loads by the
weighted damage volume method with n V 1.
CDVR1
0.051
(FM15)
0.048
(FM20)
0.046
(FM25)
0.030
(FM30)
0.048
(FM35)
0.043
(FM40)
0.035
(FM15D)
0.053
(FM25D)
0.054
(FM35D)
FM15
FM20
FM25
FM30
FM35
FM40
FM15D
FM25D
FM35D
Max
Min
0.0
2.2
2.4
12.6
1.3
0.6
3.8
2.1
1.7
12.6
2.1
3.7
0.0
0.9
11.1
0.1
1.8
2.8
4.0
3.6
11.1
4.0
4.6
0.4
0.0
10.5
0.7
2.3
2.3
4.8
4.4
10.5
4.8
16.6
11.7
9.4
0.0
9.4
9.5
4.3
13.9
13.8
0.0
16.6
3.8
0.4
0.9
11.1
0.0
1.8
2.8
4.0
3.6
11.1
4.0
6.7
2.2
1.3
8.9
2.0
0.0
1.3
6.5
6.2
8.9
6.7
12.5
7.5
5.8
4.4
6.1
6.8
0.0
11.0
10.7
4.4
12.5
0.3
3.2
3.3
13.5
2.1
0.1
4.4
0.0
0.5
13.5
0.5
0.3
3.7
3.8
14.0
2.5
0.4
4.7
0.4
0.0
14.0
0.4
Table 10. Difference (%) between the predicted and experiment failure loads by the
weighted damage volume method with n V 2.
CDVR2
0.080
(FM15)
0.075
(FM20)
0.078
(FM25)
0.048
(FM30)
0.086
(FM35)
0.077
(FM40)
0.051
(FM15D)
0.082
(FM25D)
0.085
(FM35D)
FM15
FM20
FM25
FM30
FM35
FM40
FM15D
FM25D
FM35D
Max
Min
0.0
1.3
0.6
9.7
1.4
1.2
3.8
2.1
2.4
9.7
2.4
1.8
0.0
0.5
8.5
2.4
0.3
3.0
3.7
4.1
8.5
4.1
0.8
0.8
0.0
9.3
1.7
0.9
3.5
2.7
3.1
9.3
3.1
12.0
9.5
8.3
0.0
8.9
5.5
2.1
12.1
12.9
0.0
12.9
1.4
2.8
1.9
11.0
0.0
2.1
4.6
0.3
0.5
11.0
0.5
1.2
0.4
0.1
8.9
2.0
0.0
3.3
3.2
3.6
8.9
3.6
10.9
8.5
7.4
1.4
8.2
4.8
0.0
11.3
12.0
1.4
12.0
0.5
1.9
1.1
10.2
0.9
1.6
4.1
0.0
1.7
10.2
1.7
1.3
2.7
1.8
10.9
0.4
2.0
4.5
0.4
0.0
10.9
0.4
Table 11. Predicted failure load of the joint FM25 obtained by weighted damage volume
method with two different meshes.
Original mesh
CDVR1
0.051
0.048
0.035
0.053
0.054
(FM15)
(FM35)
(FM15D)
(FM25D)
(FM35D)
Refined mesh
Predicted
failure load (N)
15,095
14,867
13,882
15,228
15,290
CDVR1
Predicted
failure load (N)
Different of predicted
failure loads from
different meshes (%)
0.050 (FM15)
0.046 (FM35)
0.036 (FM15D)
0.054 (FM25D)
0.054 (FM35D)
15,051
14,875
13,822
15,311
15,291
0.3
0.05
0.4
0.5
0
25
failure loads of joint FM25 obtained using two meshes are very close, showing only 0.5%
of the maximum difference. Consequently, the original mesh density is believed to be
suitable to predict the failure loads of the joints.
CONCLUSION
A weighted damage area method and volume ratio method were proposed to predict the
failure loads of single-lap bonded joints with dissimilar composite-aluminum materials. In
the 3D finite element analysis, interply resin layers were assumed to simulate the delamination. Geometric and material nonlinear effects were included in the analysis. The Yecriterion and Von Mises strain failure criterion were applied for the composite adherend
and adhesive, respectively. Analysis results were compared with the experimental data for
nine different bonding lengths or adherend thicknesses. The damage zone method based
on the simply calculated damage area or volume did not predict the failure load accurately.
When the intensity of the failure index and geometrical effects were considered with the
weighting power factor n 2, however, the damage volume ratio method predicted failure
loads within a deviation of 13% from experimental values. When a damage zone method is
used, it is very important to set the critical damage zone for failure evaluation. The more
accurate the experimental data that are obtained, the more accurate the critical damage
zone definition; as a result, failure load prediction is possible.
ACKNOWLEDGMENTS
This work was supported by a Korea Research Foundation Grant funded by the
Korean Government (KRF-2008-J01001) and second BK21 project at Gyeongsang
National University.
REFERENCES
1. Harris, J.A. and Adams, R.D. (1984). Strength Prediction of Bonded Single-lap Joints by
Nonlinear Finite Element Methods, International Journal of Adhesion & Adhesives, 4(2): 6578.
2. Adams, R.D. (1989). Strength Predictions for Lap Joints, Especially with Composite Adherends:
A Review, Journal of Adhesion, 30: 219242.
3. Kairouz, K.C. and Matthews, F.L. (1993). Strength and Failure Modes of Bonded Single-lap
Joints between Cross-ply Adherends, Composites, 24(6): 475484.
4. Groth, H.L. (1988). Stress Singularities and Fracture at Interface Corners in Bonded Joints,
International Journal of Adhesion & Adhesive, 8(2): 107113.
5. Hattori, T. (1991). A Stress Singularity Parameter Approach for Evaluating Adhesive Strength of
Single-lap Joints, Japan Society of Mechanical Engineers International Journal, 34(3): 326331.
(Series I).
6. Lefebvre, D.R. and Dillard, D.A. (1999). A Stress Singularity Approach for the Prediction of
Fatigue Crack Initiation in Adhesive Bonds Part I: Theory, Journal of Adhesion, 70: 119138.
7. Quaresimin, M. and Ricotta, M. (2006). Life Prediction of Bonded Joints in Composite
Materials, International Journal of Fatigue, 28: 11661176.
8. Ishii, K., Imanaka, M., Nakayama, H. and Kodama, H. (1999). Evaluation of the Fatigue
Strength of Adhesively Bonded CFRP/Metal Single and Single-step Double-lap Joints,
Composite Science and Technology, 59: 16751683.
26
K.-H. NGUYEN
ET AL.
9. Crocombe, A.D. (1989). Global Yielding as a Failure Criterion for Bonded Joints, International
Journal of Adhesion & Adhesives, 9(3): 145153.
10. Clark, J.D. and Mcgregor, I.J. (1993). Ultimate Tensile Stress over a Zone: A New Failure
Criterion for Adhesive Joints, International Journal of Adhesion & Adhesives, 42: 227245.
11. Sheppard, A., Tong, L. and Kelly, D. (1998). A Damage Zone Model for the Failure Analysis of
Adhesively Bonded Joints, International Journal of Adhesion & Adhesives, 18(6): 385400.
12. Tong, L. (1997). An Assessment of Failure Criteria to Predict the Strength of Adhesively
Bonded Composite Double-lap Joints, Journal of Reinforced Plastics and Composites, 16:
698713.
13. Tsai, M.Y., Morton, J. and Matthews, F.L. (1995). Experimental and Numerical Studies of a
Laminated Composite Single-lap Adhesive Joint, Journal of Composite Materials, 29(4):
12541275.
14. Kim, K.S., Yi, Y.M., Cho, G.R. and Kim, C.G. (2008). Failure Prediction and Strength
Improvement of Uni-directional Composite Single-lap Bonded Joints, Composite Structure,
82: 513520.
15. Brewer, J.C. and Lagace, P.A. (1988). Quadratic Stress Criterion for Initiation of Delamination,
Journal of Composite Material, 22: 11411155.
16. Shin, K.C. and Lee, J.J. (2000). Prediction of the Tensile Load-bearing Capacity of a Co-cured
Single-lap Joint Considering Residual Thermal Stresses, Journal of Adhesion Science and
Technology, 14(13): 16911704.
17. Fernlund, G. and Spelt, J.K. (1991). Failure Load Prediction of Structural Adhesive Joints
Part 2: Experimental, International Journal of Adhesion and Adhesives, 11(4): 221227.
18. Wahab, M.M.A. (2000). On the Use of Fracture Mechanics in Designing a Single-lap Adhesive
Joint, Journal of Adhesion Science and Technology, 14(6): 851865.
19. Cheuk, P.T. and Tong, L. (2002). Failure of Adhesive Bonded Composite Lap Shear Joints with
Embedded Precrack, Composites Science and Technology, 62: 10791095.
20. Li, S., Thouless, M.D., Waas, A.M., Schroeder, J.A. and Zavattieri, P.D. (2006). Mixed-mode
Cohesive-zone Models for Fracture of an Adhesively Bonded Polymer-matrix Composite,
Engineering Fracture Mechanics, 73: 6478.
21. Xie, D. and Waas, A.M. (2006). Discrete Cohesive Zone Model for Mixed-mode Fracture Using
Finite Element Analysis, Engineering Fracture Mechanics, 73: 17831796.
22. Apalak, Z.G. and Apalak, M.K. (2006). Progressive Failure Modeling of an Adhesively Bonded
Unidirectional Composite Single-lap Joint in Tension at the Mesoscale Level, Journal of
Thermoplastic Composite Material, 16: 671702.
23. Seong, M.S., Kim, T.H., Nguyen, K.H., Kweon, J.H. and Choi, J.H. (2008). A Parametric Study
on the Failure of Bonded Single-lap Joints of Carbon Composite and Aluminum, Composite
Structure, 86: 135145.
24. Rice, R.C., Jackson, J.L., Bakuckas, J. and Thompson, S. (2003). Metallic Materials Properties
Development and Standardization (MMPDS), DOT/FAA/AR-MMPDS-01.
25. MSC (2005). MARC User Manual.
26. Fenske, M.T. and Vizzini, A.J. (2001). The Inclusion of In-plane Stresses in Delamination
Criteria, Journal of Composite Materials, 35: 13251342.
27. Choi, J.H. and Chun, Y.J. (2001). Failure Load Prediction of Mechanically Fastened Composite
Joints, Journal of Composite Materials, 37(24): 21632177.
28. Ye, L. (1988). Role of Matrix Resin in Delamination Onset and Growth in Composite
Laminates, Journal of Composite Science and Technology, 33: 257277.
29. Pagano, N.J. (1989). Interlaminar Response of Composite Materials, Elsevier, Amsterdam,
New York.
30. Yang, H.T.T. and He, C.C. (1984). Three-dimensional Finite Element Analysis of Free Edge
Stresses and Delamination of Composite Laminates, Journal of Composite Materials, 28(15):
13941412.