Composite Structures 71 (2005) 159–175

www.elsevier.com/locate/compstruct

Three-dimensional finite element analysis

of single-bolt, single-lap composite bolted joints:
Part II––effects of bolt-hole clearance
C.T. McCarthy 1, M.A. McCarthy *

Composites Research Centre, Materials and Surface Science Institute, Mechanical & Aeronautical Engineering Department,
University of Limerick, Limerick, Ireland

Available online 11 November 2004

Abstract

Three-dimensional finite element models have been developed to study the effects of bolt-hole clearance on the mechanical behav-
iour of bolted composite (graphite/epoxy) joints. In Part I of this two-part paper, the model was developed in the finite element code
MSC.Marc and validated against experiments and other finite element solutions. In this second part of the paper, results from the
model showing the effects of clearance in single-lap, single-bolt joints are presented and compared with experimental results. Both
quasi-isotropic and zero-dominated lay-ups are studied. It is shown that increased clearance leads to increased bolt rotation,
decreased bolt-hole contact area, and decreased joint stiffness. In single-lap joints, clearance is shown to cause three-dimensional
variations in the stress distribution in the laminate, and these variations are also dependent on the lay-up; the radial and tangential
stresses in each ply are shown to illustrate this point. Finally, a prediction of the failure onset load using HashinÕs failure criteria is
given and the results are compared with experimental results.

2004 Elsevier Ltd. All rights reserved.

Keywords: Composite; Bolted joints; Clearance; Failure onset; Finite element analysis

1. Introduction modelling. No previous three-dimensional study on var-

iable clearance has been found in the literature. In Part I
In this paper, three-dimensional finite element analy- of this two-part paper, a finite element model was devel-
sis is used to examine the effects of bolt-hole clearance oped in the non-linear finite element code MSC.Marc,
on the mechanical behaviour of single-lap bolted com- and validated against experiments and other finite ele-
posite joints. The single-lap, single-bolt joint studied ment solutions. In this second part of the paper, the
here is a standard configuration used for characterisa- model is used to examine the effects of clearance and
tion of mechanically-fastened joints in MIL-HDBK-17 again comparisons are made with experimental results.
[1,2] and in ASTM standard D 5961/D 5961M - 96 [3], To date, a number of studies have considered the ef-
and since clearance significantly alters the three-dimen- fects of clearance in pin-loaded joints (i.e. a single plate
sional stress state in single-lap joints, this configuration loaded by one or more pins), which is often considered
provides an interesting topic for three-dimensional representative of the central lap in a double-lap joint.
These studies have been mostly analytical [4–6] or
*
numerical [7–18] with just a few including experimental
Corresponding author. Tel.: +353 61 202222; fax: +353 61 202944.
E-mail address: [email protected] (M.A. McCarthy).
results [15,19,20]. From these studies, a number of con-
1
Present address: Materials Ireland, Department of Mechanical clusions have been drawn regarding the stresses at the
Engineering, University College Dublin, Dublin 4, Ireland. hole. The bolt-hole contact area has been shown to

0263-8223/$ - see front matter
2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2004.09.023
160 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

reduce significantly with small amounts of hole oversize, Table 1

resulting in increased peak bearing or radial stress Clearances in present study
[4,8,10,13,14,16]. The contact area has also been found Clearance code Nominal clearance (lm)
to increase with load in clearance fit joints, but not in C1 0
neat-fit joints [4,10,13,16]. The location of the peak cir- C2 80
cumferential stress has been found to vary with clear- C3 160
C4 240
ance, generally being found to be near the end of the
contact region [4,10,13]. Hyer et al. [4] also pointed
out that the direction of the peak circumferential stress load. The specimen geometry is based on an ASTM
changes with clearance, which could affect joint standard [3] and can be found in Part I. The composite
strength. The effect of clearance on the magnitude of laminates were manufactured from a carbon fibre/epoxy
the circumferential stress seems to depend on whether material (HTA/6376) and two different lay-ups were
friction is included in the analysis [10,13], and is not as used: one quasi-isotropic with stacking sequence [45/0/
significant as the effect on radial stress magnitude. Neg-
45/90]5s, the other zero-dominated with stacking
ative tangential stresses have also been reported at the sequence [(45/02/
45/90)345/02/
45/0]s. A nominal
bearing plane (in front of the bolt) when large clearances 8 mm diameter Titanium alloy bolt together with a steel
are present [10]. nut and washers were used to connect the two laminates.
In the above studies on pin-loaded specimens the The clearances under examination are shown in Table 1
strain and bolt-hole contact pressure have been quite rea- and represent percentage clearances of 0%, 1%, 2% and
sonably assumed to be constant through the thickness. In 3%; they are coded C1–C4 for reference. Clearances C1
single-lap joints however, it has been shown [21] that and C2 are within current aerospace tolerances, while
these quantities vary through the thickness, being great- C3 and C4 are slightly outside. The latter two are thus
est at the shear plane or faying surface. Therefore the ef- of interest in examining the possible effects of out-of-tol-
fects of clearance in single-lap joints can only be studied erance aerospace holes (or fasteners), and also in non-
with three-dimensional analysis. In addition, in the aerospace applications. The bolts were torqued to
majority of the above studies, homogeneous anisotropic 0.5 Nm (‘‘finger-tight’’) using a calibrated torque
material properties have been used for the laminates. wrench. Further details of the problem, and the finite
This is useful for an initial analysis of the effects of clear- element model can be found in Part I.
ance on laminate stresses, but is of no value for predicting
failure since the stresses in each ply are not determined.
Because of the increase in magnitude of the peak bearing 3. Joint deformation
stress, as well as the change in direction of the peak cir-
cumferential stress, many researchers have postulated The deformed shape of the single-bolt joint model
that increasing clearance should result in reduced load with a C1 (neat-fit) clearance is shown together with
capacity for the joint [4,10,14,16]; however only a few pa- the actual deformation in Fig. 1. It can be seen that
pers have included an analysis of failure [15,19], or experi- the finite element model displays similar deformation
mental evidence to support this claim [15,19,20]. characteristics to the experiment, including bolt tilting,
The present work seeks to address the above deficien- and secondary bending of the laminates, which is caused
cies in the literature, by using three-dimensional analysis by an eccentric load path through the joint. Actual out-
to study the effects of clearance in single-lap joints, of-plane displacements could not be compared because
including a presentation of stresses in each ply for differ- no measurements were taken in the experiment. Second-
ent lay-ups, and an analysis of failure initiation using ary bending and joint twisting were discussed in Part I of
HashinÕs failure criteria [22]. Comparisons with experi- this paper, so no more consideration is given to these is-
mental results are made. The analysis highlights some sues here.
of the potential pitfalls with using such models (e.g. A section through the mid-plane, showing the motion
due to the presence of singularities in the model) and of the washers is shown in Fig. 2. As can be seen, the
provides an evaluation of their capability to predict fail- washers slide relative to the bolt head and nut and lose
ure initiation. The present analysis can be considered as contact with the bolt head and nut in some locations.
a necessary pre-cursor to performing a more complex Additionally, the clearance between the bolt and washer
progressive failure analysis. opens up on one side of the bolt, but closes on the other.
This affects the through-thickness constraint applied to
the laminate, which as shown by Tong [23] can affect
2. Problem description the initial failure load of the joint. This highlights the
value of modelling the washers discretely and not con-
The joint under investigation is a single-lap, single- sidering them an integral part of the bolt, as is often
bolt configuration, subjected to a quasi-static tensile assumed [21].
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 161

Fig. 1. Deformation characteristics of a C1 clearance joint: (a)

experiment and (b) simulation.

Fig. 3. Deformation, rigid body motion and shear stress distribution

for the bolt at applied load of 5 kN. (a) C1 clearance joint and (b) C4
clearance joint.

is higher in the C4 case, which is due to the load being

transferred over a reduced contact area, as will be shown
in the next section.
Fig. 2. Movement of washers.

4. Effect of clearance on load take-up and joint stiffness

The deformation and rigid body motion of the bolt at
an applied load of 5 kN is shown for the C1 and C4 Fig. 4(a) shows the initial part of the load deflection
clearance cases in Fig. 3(a) and (b) respectively at a mag- curves obtained from experiments on C1 and C4 clear-
nification factor of 1 (actual deformation). As can be ance joints, with attempts at best-fit lines. The results
seen, bolt rotation occurs, with more rotation in the shown are for a quasi-isotropic lay-up. The conclusions
C4 case than in the C1 case (approximately 2 versus from this figure are:
1 measured graphically). At higher displacement mag-
nification levels, significant bolt bending is also evident • A reduction in slope (stiffness) due to increasing clear-
(not shown). Also shown in Fig. 3(a) and (b) are contour ance is clearly visible.
plots of the shear stress, s13, in the bolt, which is maxi- • The C4 joint has a slight tendency to stiffen as load
mum (excluding the singular stresses at the re-entrant increases, i.e. between 5 kN and 9 kN the curve is
edges) at the location of the joint shear plane. This is slightly above the straight-line. This is not exhibited
due to high load transfer across the bolt in this region. by the C1 joint. Above about 9 kN the slope drops
It is interesting to note that the shear stress in the bolt due to the development of damage in the joint.
162 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

Fig. 4. Load–displacement curves for a C1 and C4 clearance joint (quasi-isotropic lay-up): (a) experiment and (b) simulation.

• There is a delay in load take-up for the C4 clearance model shows a delay in load take-up approximately
joint that is slightly larger than the nominal clear- equal to the clearance, as in the experiments. Attempts
ance. This also occurred for both the C2 and C3 at best-fit straight lines are also shown in Fig. 4(b) and
clearance joints (not shown). Recall from Part I of it can be seen that both curves show some initial non-lin-
this paper that the bolts were initially centred in the earity, but after this, the C1 curve is essentially linear. In
hole using a special jig, so this delay is due to the fact contrast, the C4 curve shows a slight tendency to stiffen
that the clearance has to be taken up before the bolt with increasing load, as was observed experimentally.
makes contact with the hole. However, in the simulation, the stiffness does not drop
off at high loads, since the material properties in the
Fig. 4(b) shows the load–deflection curve for C1 and model were linear elastic. Comparing the two best-fit
C4 clearance single-bolt joints obtained numerically. lines, it can be seen that the models predict a reduction
Apart from the fact that the models are slightly stiffer in stiffness due to increasing clearance (as in the experi-
than the experiments (as outlined in Part I), the trends ments). The same observations were also evident for
in the experiments and simulations are similar. The joints with a zero-dominated lay-up.
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 163

To provide a more quantitative measure of the stiff- ance increases, joint stiffness decreases, and it is also evi-
ness changes due to clearance, the joint stiffness was dent that the finite element models provide an accurate
measured between 2 kN and 7 kN, for which the experi- prediction of this loss in stiffness.
mental load–deflection curves were essentially linear. The explanation for these variations in stiffness lies in
Tables 2 and 3 show the change in stiffness as the clear- the development of the contact area between the bolt
ance is varied from neat-fit (C1) to 240 lm (C4), for both and the laminate. Fig. 5 shows the growth of the contact
the experiments and the finite element models, for a area between the bolt and one of the laminates in the C1
quasi-isotropic and zero-dominated lay-up respectively clearance joint. It can be seen that the contact area gets
(for the zero-dominated lay-up only C1 and C4 clear- up to its final value quite quickly, with a contact angle of
ances were tested––hence the absence of results for C2 160–170 which is fairly constant through the thickness.
and C3 clearances in Table 3). It is evident that as clear- This angle is in close agreement with two-dimensional
analytical and finite element solutions [4,10,24]. In the
experiment, the bolt was found to leave a silver-coloured
Table 2 imprint on the inside of the hole as shown in Fig. 5(d)
Reduction in joint stiffness as a function of bolt-hole clearance–– (highlighted for clarity). This is an indication of the
simulations versus experiments (quasi-isotropic lay-up) maximum contact area that developed during the exper-
C1 C2 C3 C4 iment and was found to closely match that predicted by
Model stiffness (kN/mm) 31.56 30.14 28.96 27.98 the finite element model.
Percentage change from C1 –
4.5%
8.24%
11.34% In contrast, in the C4 joint, shown in Fig. 6, signifi-
(models) cant contact is not made until clearance is taken up,
Percentage change from C1 –
1.9%
7.3%
10.4%
and initial contact is over a very small contact arc. As
(experiments)
the load increases, the bolt tilts, and the contact area
grows quite gradually. Even at high loads, the contact
area is still much less than in the C1 joint with a value
Table 3 of 100–105 at the shear plane, reducing to 55–60 at
Reduction in joint stiffness as a function of bolt-hole clearance–– the free face of the laminate. Because the contact angle
simulations versus experiments (zero-dominated lay-up) varies considerably through the thickness, comparisons
C1 C4 with two-dimensional solutions are not possible. This re-
Model stiffness (kN/mm) 36.44 32.44 sult highlights the importance of examining the effects of
Percentage change from C1 –
10.98% clearance in single-lap joints three-dimensionally. Note
(models) that the contact area in the model again agrees well with
Percentage change from C1 –
10% the imprint left by the bolt in the experiment. The grad-
(experiments)
ual nature of the increase in contact area explains the

Fig. 5. Development of the contact area in the C1 joint: (a) initial contact, (b) intermediate, (c) final contact and (d) experiment.

Fig. 6. Development of the contact area in the C4 joint: (a) initial contact, (b) intermediate, (c) final contact and (d) experiment.
164 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

Table 4 bending clearly being important contributors. The pre-

Percentage increase in stiffness as a result of changing from a quasi- dicted difference in joint stiffness due to lay-up is in close
isotropic to a zero-dominated lay-up (as a percent of the quasi-
isotropic stiffness) for C1 and C4 clearance joints
agreement with the experimental results.
C1 (%) C4 (%)
Simulation 13 13.7
Experiments 11 11.9 5. Effect of clearance on stress distribution around the
hole boundary

continuing stiffening of the C4 joint with increasing The effects of clearance on the stress distribution in
load, while the lower final contact area explains the the laminate around the hole boundary are presented
lower stiffness of the C4 joint compared to the C1 joint. in this section. The results from models that used homo-
As a result of having more fibres aligned along the geneous anisotropic material properties for the lami-
loading direction, the zero-dominated joints would be nates are shown in Section 5.1. Even though these
expected to be considerably stiffer than the quasi-iso- results do not give the correct values for the stresses in
tropic joints (the un-notched zero-dominated laminates each ply, they provide a useful link with previous work
are approximately 42% stiffer in the axial direction than involving two-dimensional models that assumed homo-
the un-notched quasi-isotopic laminates––see Table 3 in geneous properties, and give an overview of the effects
Part I). Table 4 shows the percentage increase in joint of clearance without the complexity present in the lay-
stiffness as a result of changing the material from a ered stress results. The stresses in each ply are presented
quasi-isotropic to zero-dominated lay-up for both the in Section 5.2.
simulations and experiments. It can be seen that the dif-
ference in joint stiffness due to the change in lay-up is 5.1. Results from using homogeneous material properties
much less than for the un-notched laminates, which is
because stiffness of the laminate is only one contributor Fig. 7 shows the radial and tangential stresses at the
to the stiffness of the joint, with other effects such as bolt hole edge obtained from models with homogeneous

Fig. 7. Radial and tangential stresses in C1 and C4 clearance joints (quasi-isotropic lay-up) with homogeneous, orthotropic material properties, at
5 kN applied load. (a) Radial stress, C1 clearance, (b) tangential stress, C1 clearance, (c) radial stress, C4 clearance and (d) tangential stress, C4
clearance.
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 165

material properties (see Table 3 in Part I for these prop- this location due to edge contact between the tilted bolt
erties). The results in Fig. 7 are for the quasi-isotropic and the hole edge. The stresses at the shear plane in Fig.
lay-up. The applied load level for these graphs was 7 would thus increase without limit with increasing mesh
5 kN, because at this load experiments for all joint con- refinement, as was shown in Fig. 20 of Part I (the stres-
figurations indicated elastic behaviour without detecta- ses a small distance away from the shear plane are rela-
ble damage. The stress is seen to vary through the tively insensitive to the mesh as was also shown in Fig.
thickness being highest at the shear plane, which is in 20 of Part I). Therefore the maximum values of the stres-
agreement with [21]. The following effects of clearance ses in Fig. 7 can only be used for comparing different
are evident: clearance cases.

• The peak radial stress increases with increasing clear- 5.2. Results from using layered material properties
ance (due to the load being distributed over a smaller
contact area). In this section, the stresses in each ply, through the
• The location of the maximum tangential stress varies thickness of the laminate for joints with different clear-
with clearance, generally being near the end of the ances (C1 and C4) and lay-ups (quasi-tropic and zero-
contact region. This also implies that the direction dominated) are presented. The stresses are plotted at
of the peak tangential stress changes. the centre of each ply and the applied load level for all
• The magnitude of the peak tangential stress increases figures was 5 kN.
slightly with increasing clearance. The radial stress distribution for the C1 clearance,
quasi-isotropic joint is shown in Fig. 9. The highest ra-
All of these findings are in agreement with previous dial stress occurs in layer no. 2, which is the second
two-dimensional studies on clearance [4,10,13]. In addi- ply from the shear plane and is orientated at 0 with
tion, it can be seen that compressive tangential stress ex- the loading direction. As can be seen, all the 0 plies
ists at the back of the hole (h ffi ±180) in both clearance are most highly stressed at the 0 location (the bearing
cases, and at the bearing plane (h ffi 0) in the C4 clear- plane), all the +45 plies are most highly stressed near
ance case. This was also noted in [10] and the reason for the +45 location and all the
45 plies are most highly
its occurrence is illustrated in Fig. 8, which shows the stressed near the
45 location. The 90 plies experience
deformed shapes of the C1 and C4 holes at high magni- their highest stresses at an angle less than ±90 which
fication. It can be seen that in both cases the hole de- varies through the thickness; the angle is less than
forms from a circular to a more oval shape leading to ±90 since the contact angle is less than 180.
compressive tangential stresses at the back of the hole, The tangential stresses around the hole boundary for
while in the C4 case a localised reduction in radius of the C1 quasi-isotropic joint are shown in Fig. 10. The
curvature under the bolt occurs (the hole locally folds tangential stress in each layer is positive except for some
in on itself) which leads to the compressive tangential plies near the free surface of the laminate at the back of
stresses at the front of the hole (h ffi 0). the hole (h ffi ±180). These negative stresses are due to
Finally before leaving these results it should be ovalisation of the hole as noted above. The highest tan-
pointed out that the stresses at the intersection of the gential stresses overall occur in all the 0 plies near the
hole edge with the shear plane are entirely mesh depend- 90 location (net-section plane). The highest tangential
ent, since as was shown in Part I, a singularity exists at stresses in the +45 plies occur near the
45 location

Fig. 8. Hole deformation at high magnification in C1 and C4 joints: (a) C1 clearance and (b) C4 clearance.
166 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

Fig. 9. Radial stress distribution in each layer of the C1 clearance quasi-isotropic ([45/0/
45/90]5s) joint (layer No. 1 is located at the shear plane).

Fig. 10. Tangential stress distribution in each layer of the C1 clearance quasi-isotropic ([45/0/
45/90]5s) joint (layer No. 1 is located at the shear
plane).

because the +45 plies are stiffest in the tangential direc- The radial stress distribution for the C4 clearance,
tion at this point. The opposite is also true for the
45 quasi-isotropic joint is shown in Fig. 11. Similarly to
plies. A second, but lower peak is also experienced in the the C1 joint (Fig. 9), the highest radial stresses occur
+45 and
45 plies at locations of 135 and
135 in the 0 layer nearest the shear plane (layer No. 2),
respectively. Again this is due to the plies being stiffest but the stress in this ply is much higher in the C4 case.
at these locations. Finally, the stresses in the 90 plies Interestingly, quite high stress levels exist in the C4 joint
are highest at the 0 location, once again due to their all the way through the thickness, whereas in the C1
high stiffness in the tangential direction at this location. joint (Fig. 9) the stresses tend to drop off to very low lev-
Thus, for the C1 quasi-isotropic joint, the highest tan- els as the free surface of the laminate (i.e. ply no. 40) is
gential stresses occur in each ply at locations along the approached. Fig. 11 shows that the radial stresses in the
hole boundary where they are stiffest in the tangential 0 plies are highest at the 0 location, as for the C1 case.
direction. However, differently from the C1 case, the +45 and
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 167

Fig. 11. Radial stress distribution in each layer of the C4 clearance quasi-isotropic ([45/0/
45/90]5s) joint (layer No. 1 is located at the shear plane).

45 plies do not peak at their stiffest locations, but at to one in the quasi-isotropic joint. This is true for all the
an angle of approximately +15 and
15 respectively. 0 plies through the thickness. Interestingly, the stresses
This is due to the contact pressure being applied over in the ±45 plies are only slightly affected by the change
a reduced contact angle. The peak radial stress value in lay-up (compare Fig. 9 with Fig. 13). The tangential
for these plies is also increased compared to the C1 case. stress distribution is shown in Fig. 14. Again similar
Although difficult to visualise in Fig. 11, the 90 plies trends exist to those in the C1 quasi-isotropic joint.
experience very low levels of radial stress at any location The main difference is that the tangential stress in the
around the hole boundary with low peaks occurring at 0 plies at the 90 location are reduced, again due to
the 0 location. having two 0 plies stacked together.
The tangential stresses around the hole boundary for The radial and tangential stresses for the C4 joint
the C4 quasi-isotropic joint are shown in Fig. 12. Differ- with a zero-dominated lay-up are shown in Figs. 15
ently from the C1 joint (Fig. 10), the tangential stresses and 16 respectively. By comparing with the previous
are negative in a region bounded by the ±30 location in three-dimensional stress graphs, it is evident that the
all plies except for two 90 plies located nearest the shear change in stress state in going from a C1 joint to a C4
plane. In addition, the negative tangential stresses at the joint is the same for a zero-dominated joint as it is for
back of the hole (h ffi ±180) experienced in the C1 case a quasi-isotropic joint. Such characteristics as the in-
are more pronounced in the C4 case. This is in line with crease in radial stress in all the plies and the change in
the findings for homogeneous properties in the previous sign of the tangential stresses at the bearing plane are
section, and is due to the changes in the deformed shape also observed with increasing clearance in the zero-dom-
of the hole due to clearance, as shown in Fig. 8. The inated case. However, the negative tangential stresses at
peak tangential stress has also increased slightly and un- the bearing plane and the back of the hole are not as
like the C1 case, occurs in the +45 and
45 plies, not pronounced as in the quasi-isotropic case (compare
in the 0 plies. This is in line with the finding from the Figs. 16 and 12). The stresses are lower in the 0 plies
previous section and from [4,10,13] that the peak tan- in the zero-dominated case, which is due to the increase
gential stress shifts towards the bearing plane (0 loca- in the number of 0 plies sharing the load.
tion) in the presence of clearance.
The radial stress distribution for the C1 clearance,
zero-dominated joint is shown in Fig. 13. Similar trends 6. Effect of clearance on damage and failure initiation
to those observed for the C1 quasi-isotropic joint (Fig.
9) are seen in the zero-dominated joint. The main differ- To investigate the damage state in both the matrix
ence is that the peak stress in the 0 ply located nearest and fibres, HashinÕs failure criteria [22] were evaluated
the shear plane (ply no. 2) is reduced because there are around the hole boundary. The graphs in this section
two 0 plies to share the load at this location compared are all shown at the load level that was found to cause
168 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

Fig. 12. Tangental stress distribution in each layer of the C4 clearance quasi-isotropic ([45/0/
45/90]5s) joint (layer No. 1 is located at the shear
plane).

Fig. 13. Radial stress distribution in each layer of the C1 clearance zero-dominated ([(45/02/
45/90)345/02/
45/0]s) joint (layer No. 1 is located at the
shear plane).

the first occurrence of compressive fibre failure in each that first fibre failure may correlate with the first signi-
joint. As will be seen considerable matrix damage al- ficant loss in joint stiffness. To test this hypothesis, a
ready exists at these load levels, but it is postulated here comparison with experimental load–deflection curves is
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 169

Fig. 14. Tangential stress distribution in each layer of the C1 clearance zero-dominated ([(45/02/
45/90)345/02 /
45/0]s) joint (layer No. 1 is located at
the shear plane).

Fig. 15. Radial stress distribution in each layer of the C4 clearance zero-dominated ([(45/02 /
45/90)345/02/
45/0]s) joint (layer No. 1 is located at the
shear plane).

given in Section 7. The approach is similar to that used damage analysis of these joints will be presented in a
in [25], which postulated that damage initiation, as de- later publication. Note also that HashinÕs criteria were
tected by a change in stiffness of the specimen, may be evaluated at the second gauss point from the hole (i.e.
governed by the peak ply strain in the fibre direction. approximately 0.5 mm from the hole in the radial direc-
Note that to predict ultimate failure, a method such as tion). This was done to avoid the singular stresses in the
the point or average stress criterion [26] or progressive plies closest to the shear plane, near the intersection of
damage analysis should be used because damage causes the hole edge with the shear plane. This is reasonable
significant re-distribution of local stresses. A progressive since in practice, these singular stresses would not exist,
170 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

Fig. 16. Tangential stress distribution in each layer of the C4 clearance zero-dominated ([(45/02/
45/90)3 45/02/
45/0]s) joint (layer No. 1 is located at
the shear plane).

since contact would not occur over a line, but over an and C on the components of the strength tensor S de-
area due to local deformations and early formation of note material strength in tension and compression
matrix damage. respectively. Since the material is insensitive to the sign
Four different failure modes are considered as shown of the shear stresses, the superscripts are omitted on
below [22]: the shear strengths S12, S13 and S23.
The stress tensor at each gauss point around the hole
Tensile matrix failure mode, r22 + r33 > 0
boundary was extracted from the model output file. This
1 2 1 2 data is output from MSC.Marc in the global coordinate
ðr22 þ r33 Þ þ ðr
r22 r33 Þ
S 223 23
2
S T22 system and thus had to be transformed to obtain the
1 2 principal stresses in each ply. The transformed stress
þ ðr þ r213 Þ ¼ 1 ð1Þ tensor was used along with the strength data shown in
S 212 12
Table 5 to evaluate the failure criteria (Eqs. (1)–(4)).
Compressive matrix failure mode, r22 + r33 < 0 The strength data was obtained from an industrial part-
"
2 # ner in the research project BOJCAS [27] and the values
1 S C22 1 2 shown are typical mean values with no reduction due to
C

1 ðr22 þ r33 Þ þ 2 ðr22 þ r33 Þ
S 22 2S 23 4S 23 scatter or defects.
1 2 1 Fig. 17 shows the location of the first fibre compres-
þ 2
ðr23
r22 r33 Þ þ 2 ðr212 þ r213 Þ ¼ 1 ð2Þ sive failure in a C1 quasi-isotropic joint as predicted by
S 23 S 12
the above failure criteria. This occurred in the second
Tensile fibre failure mode, r11 > 0 ply from the shear plane, i.e., the first 0 ply, at a load

2 level of 9.43 kN or a bearing stress of 227.3 MPa. The
r11 1
T
þ 2 ðr212 þ r213 Þ ¼ 1 ð3Þ bearing stress, rbr, was calculated from the following
S 11 S 12
equation:
Compressive fibre failure mode, r11 < 0 Pb
rbr ¼ ð5Þ
r11 ¼
S C11 ð4Þ Dt
where rij, (i,j = 1, 2, 3) is the stress tensor and Sij, (i, where Pb is the bolt load, D is the hole diameter and t is
j = 1, 2, 3) is the strength tensor. The superscripts T the laminate thickness.

Table 5
Material strength data for HTA/6376 (from BOJCAS project [27])
S T11 (MPa) SC
11 (MPa) S T22 (MPa) SC
22 (MPa) S T33 (MPa) S T33 (MPa) S12 (MPa) S23 (MPa) S31 (MPa)
2200 1600 70 250 50 300 120 50 120
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 171

Fig. 17. Location of compressive fibre failure in the C1 quasi-isotropic and C4 quasi-isotropic joints.

Fig. 18. C1 quasi-isotropic ([45/0/
45/90]5s) joint–compressive matrix failure at a load of 9.43 kN or a bearing stress of 227.3 MPa.

Fig. 18 shows the extent of compressive matrix failure are as follows. Comparing Fig. 11 with Fig. 9, it can
at this load level and as can be seen, extensive failure in be seen that high compressive radial stresses exist all
this mode has occurred at the bolt-hole contact interface the way through the thickness in the C4 joint, which is
(h < ±85) starting at the shear plane and ending about 5 not the case in the C1 joint (hence the matrix failure
plies beyond the laminate mid-plane. These failures in all plies in the vicinity of the bearing plane). In addi-
coincide with the high compressive radial stresses at tion the compressive tangential stresses at the back of
these locations in Fig. 9. Matrix compressive failure at the hole are more pronounced in the C4 joint than in
the back of the hole (h ffi ±180) throughout the lami- the C1 joint (compare Fig. 12 to Fig. 10) resulting in ma-
nate thickness is also evident. The location of this dam- trix failures at this location also (note that Figs. 9–12 are
age concurs with the compressive tangential stresses at at a load level of 5 kN so the stresses at 7.6 kN would be
the back of the hole, as shown in Fig. 10. One feature higher than in these figures). No tensile fibre or tensile
to note is that the damage is not symmetric about the matrix failures were detected for this joint at this load
bearing plane (h = 0) in the off-axis plies. Very little ma- level.
trix tensile failure and no tensile fibre failures were de- Fig. 20 shows the location of compressive fibre failure
tected at this load level. in the C1 zero-dominated joint, which occurred at a load
The location of first fibre compressive failure in the level of 11 kN or bearing stress level of 264 MPa. Fibre
C4 quasi-isotropic joint was the same as that for compressive failure initiates in the 0 ply but also in
the C1 quasi-isotropic joint. However differently from the +45 ply at the shear plane. As can be seen in Fig.
the C1 joint, failure occurred at a joint load of 7.6 kN 13, the radial stresses in the +45 ply at the +45 loca-
or bearing stress level of 178 MPa. The extent of com- tion are of the same magnitude as the radial stresses in
pressive matrix damage at this load level is shown in the 0 ply at the 0 location and so failure is just as likely
Fig. 19 and as can be seen, the entire hole surface has to initiate in the +45 ply. The extent of compressive ma-
been detected to fail in this mode. The reasons for this trix damage at this load level is shown in Fig. 21. As can
172 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

Fig. 19. C4 quasi-isotropic ([45/0/
45/90]5s) joint–compressive matrix failure at a load of 7.6 kN or bearing stress level of 178 MPa.

Fig. 20. C1 zero-dominated ([(45/02/
45/90)345/02/
45/0]s) joint–compressive fibre failure at a load of 11 kN or bearing stress of 264 MPa.

Fig. 21. C1 zero-dominated ([(45/02/
45/90)345/02 /
45/0]s) joint–compressive matrix failure at a load of 11 kN or bearing stress of 264 MPa.

be seen this damage pattern is similar to the C1 quasi- quasi-isotropic joint (shown in Fig. 17) and occurred
isotropic case (shown in Fig. 18) but more damage is at a load of 8.3 kN or bearing stress level of 180 MPa.
present at the back of the hole in the zero-dominated Differently from the C1 zero-dominated case, failure in
case. This is most likely due to the higher load level in this mode was only detected in the 0 ply, due to the fact
Fig. 21. that the peak radial stress is much higher in the 0 plies
Finally, the first compressive fibre failure in the C4 than in the +45 plies (Fig. 15), unlike the C1 case (Fig.
zero-dominated joint was in layer no. 2 (0 ply closest 13). The extent of compressive matrix damage at this
to the shear plane) at the same location as in the C4 load level is shown in Fig. 22. As can be seen, the
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 173

Fig. 22. C4 zero-dominated ([(45/02/
45/90)345/02 /
45/0]s) joint–compressive matrix failure at a load of 8.3 kN or bearing stress level of 180 MPa.

damage is confined to a narrower region when com- experimental results in Fig. 23. The experimental results
pared to the C1 zero-dominated joint (Fig. 21) at the are plotted as the bearing stiffness versus the bearing
bolt-hole contact interface and less damage is evident stress. The bearing stiffness is the slope of the bearing
at the back of the hole. This result is also in marked con- stress–bearing strain curve, averaged with a moving win-
trast to the result for the C4 quasi-isotropic joint (Fig. dow size of 7 points (as suggested in [20]). The bearing
19) and is due to the lower value of the compressive tan- stress was calculated from the load cell, using equation
gential stresses in zero-dominated case (compare Fig. 16 (5). The bearing strain was calculated according to the
with Fig. 12). For this joint, no tensile matrix or tensile ASTM standard [3] from a pair of extensometers placed
fibre failures were detected at this load level. around the overlap region of the joint (see [20] for fur-
ther details):
ðd1 þ d2 Þ=2
7. Predicted failure initiation loads in comparison to ebr ¼ ð6Þ
experimental results KD
where d1, d2 = displacements in extensometers 1,2, and
The stresses predicted to cause first fibre failure deter- K = 1.0 for double shear tests and 2.0 for single-shear
mined above from all the joints are shown together with tests (so K = 2.0 here). The graphs in Fig. 23 show where

Fig. 23. Experimental joint bearing stiffness variations with increasing applied bearing stress, together with model predictions of first fibre failure.
(a) Quasi-isotropic, C1 clearance, (b) quasi-isotropic, C4 clearance, (c) zero-dominated, C1 clearance and (d) zero-dominated, C4 clearance.
174 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175

loss of joint stiffness occurs more clearly than the origi- stress distribution; these variations are also dependent
nal stress–strain curves would. Several repeats of each of on the lay-up. When homogeneous material properties
the four cases in Fig. 23 were performed, and the graphs were used in the models, clearance resulted in
presented represent average behaviour seen in each increased peak radial stress, a shift of the location of
group of specimens. The figures show quite clearly (as the peak tangential stress towards the bearing plane,
was noted earlier) that the stiffness gets up to its maxi- slight increase in the peak value of the tangential stress,
mum value quite quickly in the C1 cases (Fig. 23 (a) and compressive tangential stress at the bearing plane.
and (c)) but more gradually in the C4 cases (Fig. 23 • The radial and tangential stresses in each ply were
(b) and (d)). They also confirm that the maximum recovered from models using layered solid elements.
stiffness attained in the C4 cases is less than in the cor- As clearance increased, the peak radial stress
responding C1 cases, and that the stiffness of the zero- increased significantly in all plies. The peak tangential
dominated joints is greater than the quasi-isotropic stress also increased and a change in sign was observed
joints. The notion of an extended ‘‘linear region’’ is also at the bearing plane (h = 0). All these findings are
seen to be something of a myth, since the stiffness in all consistent with the findings from using homogeneous
cases is constant only over a very small range of applied properties and from previous two-dimensional studies
stress, if at all. on clearance [4,10,13]. An interesting finding was that
In general, the predictions of the bearing stress at the peak radial and tangential stresses in the 0 plies
which first fibre failure occurs obtained from the model were lower in the zero-dominated joints than in the
coincide quite well with the imminent occurrence of a quasi-isotropic joints because two 0 plies are stacked
sharp drop off in stiffness in the experimental results. together and thus share the load.
The models correctly predict that the stiffness drops off • HashinÕs failure criteria [22] were evaluated around
earlier at higher levels of clearance. The curve that fits the hole boundary to investigate the damage state
the predictions least well is the zero-dominated joint with at a load level where first fibre compressive failure
a C1 clearance. In this case a gradual loss of stiffness starts was detected. The criteria were evaluated at 0.5 mm
to occur at a very low stress level (about 125 MPa); how- from the hole boundary to avoid the mathematical
ever, the stress predicted for first fibre failure does coin- singularity present at the intersection of the hole edge
cide with a change to a much sharper rate of stiffness with the shear plane. For all joint configurations it
reduction. In the zero-dominated lay-up, the pairs of was found that considerable matrix damage had
adjacent 0 plies initially provide a very stiff foundation occurred at both the contact interface and the back
for the bolt to press against, but perhaps this stiffness is of the hole at these load levels. The matrix damage
lost early on due to buckling of these plies or delamina- was primarily due to compressive radial stresses in
tion from the surrounding plies. A delamination initia- front of the bolt and compressive tangential stresses
tion analysis will be reported in a later publication. at the back of the hole. The extent of this matrix dam-
age may affect the subsequent propagation of damage
in the laminate and should be investigated using pro-
8. Concluding remarks gressive damage analysis. A progressive damage anal-
ysis of these joints will be reported on in a later
In this paper, three-dimensional finite element analy- publication. Very little tensile fibre or matrix failure
sis was used to examine the effects of bolt-hole clearance was detected at these load levels for all joint configu-
on the mechanical behaviour of single-lap bolted com- rations, which was expected since the joints were sized
posite joints, and comparisons were made to experimen- to fail in the bearing failure mode.
tal results. No previous three-dimensional study on • The predicted load levels for first fibre failure were
variable clearance has been found in the literature. The compared with experimental results for the variation
main findings from the study are summarised below. in joint stiffness with increasing load level. In general
it was found that the predicted load for first fibre fail-
• Clearance results in increased bolt rotation, decreased ure coincided well with onset of a sharp drop in the
bolt-hole contact area, and decreased joint stiffness. experimental joint stiffness. The zero-dominated con-
Joints with clearance tend to gradually increase in figuration with neat-fit clearance exhibited early loss
stiffness with increasing applied load, whereas neat- of stiffness, which may be due to out-of-plane motion
fit joints do not. Three-dimensional finite element of the pairs of adjacent 0 plies or delamination of
analysis has proved capable of quantifying these these plies from surrounding plies, which was not
effects with a high degree of accuracy in comparison detected by the present analysis.
to experimental results.
• In single-lap joints, the stress distribution in the lami- The results presented provide an exposition of the
nate is non-uniform through the thickness, and clear- main effects of clearance in single-lap composite joints,
ance causes three-dimensional variations in this which should provide a useful reference for future more
 C.T. McCarthy, M.A. McCarthy / Composite Structures 71 (2005) 159–175 175

complex studies involving progressive damage analysis, [13] Ko HY, Kwak BM. Contact analysis of mechanically fas-
the results of which can sometimes be difficult to inter- tened joints in composite laminates by linear complementarity
problem formulation. Composite Structures 1998;40(3–4):187–
pret. It has been found that bolt-hole clearances only 200.
slightly outside those used in the aerospace industry [14] Lanza Di Scalea F, Cappello F, Cloud GL. On the elastic
have a significant effect on the stiffness and failure onset behaviour of a cross-ply composite pin-joint with clearance fits.
load of composite bolted joints and thus clearance Journal of Thermoplastic Composite Materials 1999;12:13–22.
should be taken into consideration when designing com- [15] Pierron F, Cerisier F, Grediac M. A numerical and experimental
study of woven composite pin-joints. Journal of Composite
posite joints. Materials 2000;34(12):1028–54.
[16] Chen WH, Lee SS, Yeh JT. Three-dimensional contact stress
analysis of a composite laminate with bolted joint. Composite
References Structures 1995;30:287–97.
[17] McCarthy CT, McCarthy MA, Padhi GS. 9th Annual Confer-
[1] DODSSP, Polymer Matrix Composites, MIL-HDBK-17, DOD- ence, Association for Computational Mechanics in Engineering,
SSP, Naval Publications and Forms Center, Standardization University of Birmingham, 8–10 April 2001, p. 111–14.
Documents Order Desk, Building 4 D, 700 Robbins Ave., [18] McCarthy MA, McCarthy CT, Padhi GS. 9th Annual Confer-
Philadelphia, PA 19111-5094. ence, Association for Computational Mechanics in Engineering,
[2] Shyprykevich P. Characterization of bolted joint behaviour: MIL- University of Birmingham, 8–10 April 2001, p. 123–26.
HDBK-17 Accomplishments at standardization. Journal of [19] DiNicola AJ, Fantle SL. Bearing strength behaviour of clearance-
Composites Technology and Research 1995;17(3):260–70. fit fastener holes in toughened graphite/epoxy laminates. In:
[3] ASTM standard D 5961/D 5961 M––96, Standard test method for Camponeshi Jr ET, editor. Composite Materials: Testing and
bearing response of polymer matrix composite laminates, 1996. Design, ASTM STP 1206, vol. 11. Philadelphia: American Soci-
[4] Hyer MW, Klang EC, Cooper DE. The effects of pin elasticity, ety for Testing and Materials; 1993. p. 220–37.
clearance, and friction on the stresses in a pin-loaded orthotropic [20] McCarthy MA, Lawlor VP, Stanley WF, McCarthy CT. Bolt-hole
plate. Journal of Composite Materials 1987;21(3):190–206. clearance effects and strength criteria in single-bolt, single-lap,
[5] Kradinov V, Barut A, Madenci E, Ambur DR. Bolted double-lap composite bolted joints. Composites Science and Technology
composite joints under mechanical and thermal loading. Interna- 2002;62:1415–31.
tional Journal of Solids and Structures 2001;38:7801–37. [21] Ireman T. Three-dimensional stress analysis of bolted composite
[6] Fan WX, Qiu CT. Load distribution of multi-fastener laminated single-lap joints. Composite Structures 1998;43:195–216.
composite joints. International Journal of Solids and Structures [22] Hashin Z. Failure criteria for unidirectional fiber composites.
1993;30(21):3013–23. Journal of Applied Mechanics 1980;47:329–34.
[7] Chutima S, Blackie AP. Effect of pitch distance, row spacing, end [23] Tong L. Bearing failure of composite bolted joints with non-
distance and bolt diameter on multi-fastened composite joints. uniform bolt to washer clearance. Composites Part A 2000;31:
Composites Part A 1996;27 A(2):105–10. 609–15.
[8] Rowlands RE, Rahman MU, Wilkinson TL, Chiang YI. Single [24] deJong T. Stress around pin-loaded holes in elastically orthotropic
and multiple bolted joints in orthotropic materials. Composites or isotropic plates. Journal of Composite Materials 1977;11:
1982;13(3):273–9. 313–31.
[9] Pradhan B, Kumar R. Stresses around partial contact pin-loaded [25] Naik RA, Crews Jr JH. Ply level failure analysis of a graphite/
holes in FRP composite plates. Journal of reinforced plastics and epoxy laminate under bearing-bypass loadings. In: Garbo SP,
composites 1984;3(1):69–84. editor. Composite Materials: Testing and Design, ASTM STP
[10] Naik RA, Crews Jr JH. Stress analysis method for a clearance-fit 1059, vol. 9. Philadelphia: American Society for Testing and
bolt under bearing loads. AIAA Journal 1986;24(8):1348–53. Materials; 1990. p. 191–211.
[11] Ramamurthy TS. New studies on the effect of bearing loads in [26] Whitney JM, Nuismer RJ. Stress fracture criteria for laminated
lugs with clearance fit pins. Composite Structures 1989;11:135–50. composites containing stress concentrations. Journal of Compos-
[12] Kim SJ, Kim JH. Finite element analysis of laminated composite ite Materials 1974;8:235–65.
plates with multi-pin joints considering friction. Computers and [27] McCarthy MA. BOJCAS: bolted joints in composite aircraft
Structures 1995;55(3):507–14. structures. Air and Space Europe 2001;3/4(3):139–42.