Untitled

RECENT ADVANCES IN STRUCTURAL JOINTS AND
REPAIRS FOR COMPOSITE MATERIALS

Recent Advances in Structural
Joints and Repairs for
Composite Materials
Edited by
LIYONGTONG
School of Aerospace,
Mechanical and Mechatronic Engineering,
University of Sydney, Australia
and
COSTAS SOUTIS
Department of Aeronautics,
Imperial College,
London, U.K.
SPRINGER-SCIENCE+BUSINESS MEDIA B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress.
ISBN 978-90-481-6319-9 ISBN 978-94-017-0329-1 (eBook)

DOI 10.1007/978-94-017-0329-1
Printed on acid-free paper
All Rights Reserved

© 2003 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 2003
Softcover reprint of the hardcover 1st edition 2003
No part of this work may be reproduced, stored in a retrieval system, or transmitted
in any form or by any means, electronic, mechanical, photocopying, microfilming, recording
or otherwise, without written permission from the Publisher, with the exception
of any material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work.
PREFACE
Joints in components or structures incur a weight penalty, are a source of failure and
cause manufacturing problems; therefore, whenever possible, a designer will avoid using
them. Unfortunately it is rarely possible to produce a construction without joints due to
limitations on material size, convenience and cost in manufacture or transportation and
the need for access in order to inspect or repair the structure. Such considerations apply
equally to joints between metallic components or between composite components.
Thus, an important requirement for the complete design of practical structures is the
development of attachment methods, joint designs, and the problem ofload introduction
in the composite assembly. Without properly thought and designed joints, we are
unable to take full advantage of the high stiffness and strength oflaminated composites.
Basically, there are two types of joint commonly employed with fibre-reinforced
materials: adhesive bonded joints and mechanically fastened joints. Welding is also a
possibility for thermoplastic composites, but this technique is not well developed for
load-carrying joints. In the following paragraphs, the first two types of joint are
discussed.
Adhesive joints are natural to consider for polymeric matrix composite materials
because many matrix resins are also good adhesives. When the matrix material of the
laminate is also used as the adhesive in the joint, excellent adhesion can result.
However, even with the excellent adhesion, the joint does represent a discontinuity in
the material, and resulting high stresses often initiate joint failure. Therefore, knowledge
of the stresses in joints is vital if we are to understand the failures that occur in practice
and hence improve designs and predict strength. Chapters 1, 2, 6 and 7 cover the topics
related to stress analysis, strength determination and design of adhesive bonded joints.
In structures where parts are removed for inspection or maintenance bolted joints
(Mechanically fastened joints) will be required. The behaviour of bolted connections,
for composite laminates made from unidirectional pre-preg material has been
extensively examined by several workers, who investigated a wide range of variables
such as lay-up, fastener type (screw, rivet, and bolt), friction effects, clearance and their
influence on the failure mode. A full theoretical description of the stresses in such a
joint must include their three-dimensional nature, a fact that has limited the analytical
treatment given to such connections. The prediction of failure loads, is for the moment
at least, only semi-empirical at best and any improvement will depend on the
development of failure criteria that are more generally applicable, together with an
easy-to-use three-dimensional stress analysis. Chapters 3, 4 and 7 describe stress
analysis, strength prediction and design of mechanically fastened joints.
The maintenance and repair of composite structures are always of main concern to
the end-users as well as the manufactures. On an aircraft, the need to repair a
component can arise from physical damage due to accidental impact, bird strike,
hailstones and lightning strike or from deterioration caused by the absorption of
moisture or hydraulic fluid. Once a defect has been found and it falls outside approved
allowable limits, one of the following actions should be taken according to its extent
and the constraints of operational conditions: temporary repair, permanent repair or
v
VI
replacement. The temporary repair (field repair or emergency repair) is usually a

preferred action when the operational limitations are severe and the temporary repair is
a safe option. The limitations are usually time, environment and facilities. Temporary
repair aims to restore the mechanical strength required to permit aircraft operation until
a permanent repair or replacement can be carried out.
Adhesively bonded repair is the most common type of repair carried out with
composite materials, because load transfer paths through mechanical fasteners can
cause local overloads and damage in the relatively brittle composites. Two types of
bonded repairs are mainly used; the highly refined and structurally efficient but
expensive flush patch repair and the external metal or composite patch. In the external
patch technique, the damaged material is removed by cutting a hole, the parent plate is
then cleaned and applied with filler and adhesive materials before the patches are
attached. In practice, the repair patch would always be tapered at the edges to reduce
peel stresses. This type of repair could also be used for a permanent repair in lightly-
loaded and relatively thin structures.
Flush, scarf-type, bonded repairs are used on critical, highly loaded, components
where load concentration and eccentricities, especially for compressive loading, must
be avoided. Thick monolithic structures lend themselves to such repairs since external
patch would cause excessive out-of-mouldline thickness and unacceptably high
bondline peel and shear stresses. In all these repair methods the main concerns are the
quality of surface preparation prior to bonding, the design of the composite patch and
the prediction of both strength and durability of the repaired configuration. Chapters 5
and 6 cover stress analysis, strength determination and design of adhesive bonded
repmrs.
LT is grateful to his postgraduate students, Quan Nguyen and Phu Due Nguyen, for
helping with preparation of camera ready copy of the book.
L. Tong and C. Soutis

Table of Contents
Preface v
1 Stress Analysis and Failure Assessment of Lap Joints

C.H. Wang and L.R.F. Rose
1.1 Introduction
1.2 A critical review of stress analysis and failure assessment 2
1.3 Ends effect 6
1.4 Stress concentration in adherends 17
1.5 Triaxial stresses and plastic yielding 20
1.6 References 23
2. Strength Determination of Adhesive Bonded Joints

L. Tong, JK. Spelt and G. Fernlund
2.1 Introduction 27
2.2 Failure of adhesive joints 28
2.3 Stress analysis of adhesive joints 35
2.4 Failure criteria and materials characterization 38
2.5 Strength determination for static loading 44
2.6 Strength determination for cyclic loading 57
2.7 Design for creep crack growth 60
2.8 References 62
3. Stresses in Mechanical Fastened Joints

F.L. Matthews and P.P. Camanho
3 .I Experimental behaviour 67
3.2 Semi-analytical and numerical methods of stress analysis 71
3.3 Strength prediction methods 82
3.4 Conclusions 95
3.5 References 96
4. Strength Determination of Mechanical Fastened Joints

F.-K. Chang and XL. Qing
4.1 Introduction 101
4.2 Progressive failure analysis 103
4.3 Numerical prediction 115
4.4 Verification and comparison 117
4.5 Effect of the clamping 126
4.6 Composite joints under combined bearing and axial-bypass loads 131
4.7 Conclusion 135
4.8 Acknowledgement 136
4.9 References 136
vii
V111
5. Strength Analysis of Adhesively Bonded Repairs

C. Soutis and F.Z. Hu
5.2 External Patch Repairs 142
5.3 Design ofbondedjoints 143
5.4 Strength prediction of external patch repair 147
5.5 Scarf patch repairs 161
5.6 Discussion and concluding remarks 168
5. 7 References 170
6 Adhesively Bonded Joints for Fibrous Composite Structures

L.J Hart-Smith
6.2 Adhesively bonded joint configurations 175
6.3 Adhesive stress-strain curves in shear 179
6.4 The non-uniformity ofload transfer in adhesively bonded joints 183
6.5 Double-lap and double-strap joints 187
6.6 Single-lap joints 191
6.7 Stepped-lap bonded joints 194
6.8 Scarf joints 200
6. 9 Adhesively bonded joints other than at splices 202
6.10 Flaws in adhesive bonds 204
6.11 Repair of fibrous composite structures 204
6.12 Rules ofthumb for designing structurally efficient joints 205
6.13 Concluding remarks 206
6.15 References 207
7 Design and Analysis ofBo1ted and Riveted Joints in Fibrous

Composite Structures
L.J Hart-Smith
7.2 Single-hole test coupons 212
7.3 Effects of fibre pattern on strength of bolted composite joints 224
7.4 Experimental test program
232
7.5 Correlation between test and theory 237
7.6 Joint efficiency charts for designing bolted composite joints 240
7. 7 Miscellaneous considerations 249
7.8 Rules of thumb for designing structurally efficient joints 250
7.9 Concluding remarks 251
7.11 References 253
CHAPTER 1: STRESS ANALYSIS AND FAILURE ASSESSMENT OF LAP JOINTS
C.H. WANG and L.R.F. ROSE

Aeronautical and Maritime Research Laboratory,
506 Lorimer Street, Fishermans Bend, VIC 3207, Australia
1.1 INTRODUCTION
Due to the differential straining in substrates, adhesively bonded joints inevitably

experience stress elevations, especially near the ends of overlap where load transfer
takes place. Since a bonded joint represents a multi-layer structure involving two
substrates and an adhesive layer, the stress states that exist at various levels in a bonded
joint are very complex. Accurate determination of the stresses at critical locations
within a joint would require detailed numerical computations using three-dimensional
elastic-plastic finite element analysis, or elaborate experimental techniques. Both of
these two methods are not very efficient and suitable for engineering designs,
parametric analysis, and optimisation. Consequently considerable effort has been
devoted to develop simple yet accurate analytical estimates for various lap joints,
including double overlap joint, single lap joint, and single-strap joint. Several good
texts are available, eg. Adams et al [I], and thus it is unnecessary to repeat the well-
known solutions here. Instead, this Chapter focuses on some aspects of stress analysis
issues pertinent to failure assessment of bonded joints, with emphasis on convenient
closed-form solutions that are suitable for engineering design.
The structure of this Chapter is as follows. The important results of plate-spring
theories and failure assessment of lap joints (some typical examples are shown in Figure
1.1) are first reviewed in Section 1.2, with a view to identifying gaps in the existing
methodologies. Section 1.3 presents some recent results on the stresses at the ends of
overlap, together with the comer stress-intensity factors determined using the asymptotic
matching method. Section 1.4 considers the stress concentration in adherend, which is
an important factor affecting the joint strength. The triaxial stress-state in the adhesive
layer and the implications on adhesive plastic yielding are discussed in Section 1.5.
L. Tong and C. Soutis (eds.).

Recent Advances in Structural Joints and Repairs for Composite Materials, 1-26.
© 2003 Kluwer Academic Publishers.
2 C.H. Wang and L.R.F. Rose
double lap joint
single lap joint a~ !=._I__A_d_h_e1re""nd•2-•'-----r--,__~

I!--_A_d_he_re_n_d_l_----1.l_t_~ a ,.
l+- 2 e -.~
single strap joint a~
Figure 1.1 Configurations of single lap, double lap, and single strap joints.
1.2 A CRITICAL REVIEW OF STRESS ANALYSIS AND F AlLURE ASSESSMENT
1.2.1 ADHESIVE STRESSES
In order to obtain closed-form solutions required to guide the design of lap joints, many
theories for bonded joints often treat the adhesive layer as infinite, unconnected shear
and tensile springs [2]. Examples of this type analyses include the classical work by
Goland and Reissner [3], Adams and Peppiatt [4], Hart-Smith [5-7]. In these analyses
both the shear stress and the peel stress in the adhesive layer vary strongly over the
length of the overlap, and attain their maximum values at the end of overlap. Assuming
the overlap length is sufficiently large as compared to the load transfer length [3], the
m,1ximum shear and peel stresses become independent of the length of overlap, and can
be expressed by the following equations [5],
(1.1)
(1.2)
with
Chapter 1 Stress Analysis and Failure Assessment of Lap Joints 3
(1.3)
where the subscripts i, o and A are used to indicate parameters pertaining to the inner
adherend, the outer reinforcement, and the adhesive. Here the parameters E and t denote
the Young's modulus and the thickness of the substrates and the adhesive, and the
parameter GA denotes the shear modulus of the adhesive. The parameter p denotes the
load per unit width. It is worth noting that for a given applied load, the peel stress
decreases at a faster rate than the maximum shear stress with the increase in bond-line
thickness.
For single lap joints, due to the load-path eccentricity, a significant bending exists
near the end of overlap. Furthermore, since the out-of-plane deflection is not
constrained, the distribution of the adhesive shear stress differs from that in double lap
joint. Similarly, suppose the overlap length is sufficiently large as compared to the load
transfer length, the maximum shear and peel stresses in balanced single-lap joints are
given by Goland and Reissner [3],
r
max
=a= (1+43k) [ 2~_!_ ]
Et
1/2
(1.4)
A
a
max
=a=~[6EA
2 Et
_!_]I/2[1+2J2AI[EtA]I/4]
Et
(1.5)
A A
with
k= cosh A£ (1.6)
cosh Af + 2-.fi sinh Af
3 2 a=
A-= -(1-v )-" (1.7)
2 Er
where e denotes the half length of the overlap. If geometrically non-linear deformation
is ignored, ie, k=O, and a= IE« 1.0, the maximum adhesive in a single lap joint is
about four times that exists in an equivalent double lap joint with identical adherend
and adhesive dimensions. It is interesting to note that, unlike in the case of double lap
joints, both the peel stress and the shear stress in a single-lap joint are proportional to
the reciprocal of the square root of the bond-line thickness.
It should be noted that the plate-spring method for lap joints suffers from three
major deficiencies. Firstly, the adhesive shear stress solution does not satisfY the exact
boundary condition of zero shear stress at the ends of the adhesive layer [4, 8-1 0].
Secondly, these solutions do not capture the complex stress singularity at the termini of
the adhesive layer [ 11-15]. The third problem is that the stress state in the adherends is
inevitably assumed to be uniform in the thickness direction, and hence stress
concentrations are ignored [10].
Attempts have been made to eradicate the first deficiency of the plate-spring
approach by treating the adhesive layer as a two-dimensional continuum [16], while
retaining the plate-theory approximation for the substrates. Although these refined
theories permit satisfaction of the stress-free surface condition at the ends of a square-
edged adhesive layer, the distance over which the adhesive shear stress attains its peak
value appears to scale with the substrate thickness, instead of the adhesive layer
thickness as indicated by finite element analysis [ 10].
Recent progresses in addressing the above mentioned three major deficiencies will
be presented in Section 1.3. Firstly, a simple correction to the shear stress distribution
along the mid-plane of the adhesive layer has been obtained based on an eigenfunction
expansion method, which captured the boundary layer effect near the end of an
adhesive layer. Secondly, the comer singularity will be determined using an asymptotic
matching method. It will be shown that this comer singularity offers a very promising
failure criterion of bonded joints under both static and fatigue loading conditions. The
improved adhesive shear stress solution permits determination of the stress
concentration in adherends.
1.2.2 JOINT FAILURE CRITERIA
Adhesive failure within a joint is usually assumed to be governed by a critical stress, or

a critical strain, sometimes coupled with a characteristic length [17]. Some examples
include Hart-Smith [5, 6], Adams [18, 19], Bigwood and Crocombe [20]. Strain energy
density [5, 21] and stress over a zone [22] have also been suggested for predicting joint
strength. However, it has long been noted that [4] that there is a large discrepancy
between theoretical predictions and experimental data. As discussed in the previous
section, both the maximum adhesive shear and peel stresses at the ends of a joint are
inversely proportional to the square root of the adhesive thickness. Therefore, for a
given applied load, the thicker the adhesive, the lower the stresses are. This means that
the conventional strain or stress based failure criteria would predict an increase in joint
strength (total failure load) with the increase in bond-line thickness. In other words, the
remote applied stress at failure is proportional to the square root of the adhesive
thickness,
(1.8)
Experimental results, however, exhibited exactly the opposite trend: the load carrying
capacity of joints decreases with bondline thickness [23-26].
Five sets of experimental results on single-lap joints [23-24, 26] and one set of data
on single-strap joints [27] are shown in Figure 1.2, together the predictions of stress-
based failure criteria. It can be seen that with the increase in bond-line thickness, joint
strengths decreased, contrary to the expectations of stress-based failure criteria. Also
shown in the figure are predictions based on fracture mechanics approaches (strain-
energy release rate and comer stress-intensity factor), which will be discussed in detail
later.
3r-------------------------~~---------.
Constant stress criteria
(jma:..= crr
- - • Crocombe & Moult

2
v-- 'V Harris & Fay, -30°C
G---El Harris & Fay, 20°C
G----<> Harris & Fay, 90°C
Strain-energy release rate criterion
- ~-~- -~~ ~ - - ---~""=-
0.5 Fillet comer S~re comer

QL-~--~--~--~--~~--~--~--~--~~
I 4 8 12
Normalised adhesive layer thickness tA/t 0
Figure 1.2 Comparison of experimental results of joint strength and theoretical predictions
Under fatigue loading, it has also been observed that fatigue crack growth rates and
fatigue endurance [26, 28-29] also exhibited anomalous behaviour with respect to
adhesive thickness. These results showed that on the basis of the maximum (shear or
normal) stress in the adhesive, joints with thinner adhesive lasted longer than joints
with thicker adhesive. For a given applied load, although joints with thicker adhesive
exhibited longer fatigue Jives than joints of thinner adhesive, the increase in fatigue
load is much less than what would be expected on the basis of constant stress.
Therefore it is clear that the conventional failure criteria, be it stress, strain, or strain
energy density based, are not geometry independent, hence the data obtained from one
configuration are not readily transferable to other types of joint, or the same joint with
different bondline thicknesses.
Fracture mechanics concepts have also been introduced for characterising bonded
joints [ 17, 30-33], but mainly to joints with crack-like defects, such as butt joints, edge
cracked cantilever beam and cracked-lap-shear specimens [17, 34]. The thickness of the
adhesive is basically ignored and the disband is treated as a crack, thus permitting the
use of linear elastic fracture mechanics parameters. Finite element analysis and
compliance approaches are the two main methods used in evaluating the energy release
rate. An alternative method of calculating the strain energy release rate is through the
J-integral [35-37, 46], leading to expressions of the strain-energy release rate in terms
of the maximum adhesive shear and peel stresses obtained from plate-spring theories of
lap joints,
2
G _ (Jmax t
I-2£A (1.9a)
A
2
G II ='l'max
--fA (1.9b)
2GA
and the total strain-energy release rate is the sum of the above two components.
Since the maximum adhesive shear and peel stresses are proportional to the
reciprocal of the square root of the bond-line thickness, referring to equations (1.1) and
(1.2), it can be readily shown that the strain-energy release rates given by (1.9a) and
(1.9b) are independent of bond-line thickness, and are dependent only on the
dimensions of the adherends and the applied load. Consequently, fracture mechanics
based failure criteria would predict that joint strength is independent of bondline
thickness, as indicated in Figure 1.2. When compared to the stress/strain based failure
criteria, this represents a considerable improvement, considering that joint strength has
been observed to decrease with bondline thickness. However, the fracture mechanics
methods are applicable only to joints with large disbond of length greater than or
comparable to substrate thickness. Since virtually all bonded joints are manufactured to
be free of large dis bonds, the physical basis of the above mentioned fracture mechanics
methods as applied to lap joints is not clear.
For disband-free bonded joints, the corner stress-intensity factor has been found to
be a promising criterion that can unity the failure loads of butt joints of varying
bondline thicknesses [38] and the failure loads of single lap joints [12]. Details of this
approach will be discussed in the next Section.
1.3 ENDS EFFECT
1.3.1 ADHESIVE SHEAR STRESS
According to the plate-spring theories for bonded joints, the adhesive shear stress
distribution near the ends of adhesive layer can be expressed as follows, assuming that
the overlap is sufficiently long so that the peaks at the two ends are isolated,
(1.10)
The reason that the plate-spring models predict a finite shear stress at the ends of
adhesive layer stems from the assumption of spring laws being position independent,
i.e. the shear stress is proportional to the relative longitudinal displacement between two
substrates, even at a stress-free surface. Nevertheless, despite this deficiency and the
large number of assumptions involved in the plate-spring models, solutions of the
adhesive shear and peel stresses have been shown to correlate well with finite element
analysis except very close to the ends of adhesive layer.
Since the ends of the adhesive layer are stress-free, the shear stress near the ends
must change from this simple distribution so as to satisfY the traction-free boundary
conditions. According to the St. Venant's principle, the perturbation in the stress field
is confined to a region near the ends (within a distance comparable in extent to the
layer's thickness). This means that for the case of square edge the finite shear stress
predicted by the plate-spring theories can be considered as residual stresses. It then
follows that the problem with stress-free edge condition can be solved by superposing
on the stress field derived based on plate-spring analogy a second stress field that just
negates the traction on the ends.
Since the adhesive layer generally has a Young's modulus much lower than the
substrates, it is not unreasonable to model the substrates as rigid clamps. Consequently
the problem can be considered as a strip clamped at two long edges and subjected a
uniform shear traction and zero normal traction at x=O, as depicted in Figure 1.3. From
a theoretical viewpoint, rigorous calculation of the shear stress near the square ends of
an adhesive layer requires a consideration of stress singularities at the comers of the
interface between the adhesive layer and the substrates. However, since the field in
which the stresses are dominated by a comer singularity is normally very small [39], the
effect of comer singularity can be ignored when determining the shear stress at the mid-
plane of the adhesive layer.
Using an eigenfunction expansion method, the shear stress at the mid-plane of the
adhesive layer has been found to be well approximated by the first order solution [10],
'Z' xy ( X, 0) -_ -rmaxe -2yxlt, COS ( 2 y X It A ) (1.11)
where y denotes the real part of the first eigen root (see Table 1 of Wang et al [10],
which is equal to 2.1535 for a Poisson's ratio of0.35.
Combining equations ( 1.11) and ( 1.1 0) leads to the following improved solution of
adhesive shear stress distribution in a bonded joint,
'Z'A(x)='Z'maxle-,& -e-2 yxlt, COS(2yx/tA)j (1.12)
which obviously satisfies the zero shear stress boundary condition at the end of
adhesive layer (x=O). This stress distribution is shown in Figure 1.4 together with the
results of finite element analysis detailed later, indicating a good agreement. This
solution will be utilized in Section 1.4 to determine the stress concentration in
adherend.
-rmax
Figure 1.3 A strip rigidly clamped along two long edges.
1.2 -------I------- -I------- - 1 - - - - - - - ~-------

I I I I
1.0 -----
I I I I
0.8 -- -~-- -[----- -- -[- ------ -[--- ---- ~---- ---

I
0.6 - - - - - -1------- -1------- -1--- - - - -

I I
_J- - - - - - -
I
I I I I
r I
I
I
I
I
I
I
I
I
I
I
I
0.4 , - - - - - - 1 - - - - - - __ j _ - - - - - - - 1 - - - - - - - _J_---- --
?-- - ~ Finite element

---Theory I
0.2 --~==.,..,..,...,:-=-::-::-:,.,...,=-::-~_----I-------
I
_J-------
(1/mm)
~=0.22
tA=O.l mm
o~~~~~~~~~~~~~~~~
0 1 2 3 4 5
x/(ti2)
Figure 1.4 Adhesive shear stress distribution including the end effect.
Chapter I Stress Analysis and Failure Assessment of Lap Joints 9
1.3.2 COR NER SINGULARITY
1.3.2.1 Square edge

It is well recognized that stress singularity exists at a comer between the adhesive and
the adherend [ 11-12, 15, 40]. Studies have also shown that for joints bonded with
brittle adhesives the intensity of the stress singularity at the interface comer may be
used to predict failure ofbondedjoints [12-13].
For the adhesive joint shown in Figure 1.5, there are two comer singularities at
points A and B. To apply the fracture mechanics approach, both the order and the
intensity of the comer singularity must be quantified in terms of joint geometry and the
applied load. While the order of singularity can be obtained from an asymptotic
analysis [11, 40], the magnitude of the stress-intensity factors requires matching the
asymptotic solution with the outer solution. An efficient method of carrying out this
asymptotic matching analysis is to employ the finite element method.
Lx
Figure 1.5 Finite element mesh for a double-lap joint.
Distributions of the three stress components, a xx, O)y, and r,y are shown in Figure 1.6,
indicating clearly the high stress elevations near the two comers.
Figure 1.6 Contour plots of normalized stresses: (a) axial stress a"' (b) peel stress Gyy, and (c)
shear stress r,y
In the case of square ends shown in Figure 1. 7, the stress intensity factor K is defined as
follows in terms of the stress component normal to the interface, CJee ,
( 1.13)
inner
adhe rend
adhesive A X
layer
outer
ad herend
Figure I. 7 Ends of adhesive layer, showing a square edge.
In this case, the shear stress distribution along the interface is [39],
(1.14)
where the parameter a denotes the order of comer singularity, which is shown in Fig.
1.8 for a square end and spew fillet.
The order of singularity at square end can be expressed as
a= 1.29v(I-0.768v) ( 1.15)
while the stress-intensity factor is given by
K = [A(V)CJmax + B(v)rmax ]t~ (1.16)
where CJmax and Tmax are the maximum adhesive peel stress and adhesive shear stress
given by equations ( 1.1) and ( 1.2) for double lap joints, and equations (1.4) and ( 1.5)
for single-lap joints. The parameters A and B depend solely on the Poisson's ratio of
the adhesive via the following equations [39],
A(v) = 0.836 - 2.23v + 6.29v 2 - 9.64v 3 (1.1 7)
B(v) =3.12 -15.8v + 40.lv 2 - 37.6v 3 (1.18)

i:l 0.5
,....._
'"0
s::
0
0.4
~;::l
C"
"'
'-'
0.3
-~
"3bl)
s::
·;;:; 0.2
\
Square end
....0 points A and B in Fig. 7
....0s:: •
()
0.1 a=1.29 v (1-0.768 v)
""'....0 •
0
"E
0 0
0 0.1 0.2 0.3 0.4 0.5
Poisson's ratio v
Figure 1.8 Order of singularity at square end for various Poisson's ratio.
Now the stress intensity factors at two comer points shown by the insert of Figure 1.9
can be expressed as follows, noting the difference in the sign of shear stress
(1.19a)
(1.19b)
Comparisons between finite element results and predictions based on equations (1.19a)
and (1.19d) are shown in Figure 1.9. It is seen that there is a good agreement between
the predictions and the finite element results. It is also clear from Figure 1.9 that the
comer singularity dominates a reasonably large region. For brittle adhesives whose
process zone at failure is smaller than the singularity field, the strength of the joint
would be entirely characterised the stress intensity factors discussed above. In this case,
it is interesting to note that the joint strengths when compared on the basis of the
maximum stresses obtained from the conventional one-dimensional theories would
exhibit a dependence on the adhesive layer thickness. In other words, if failure occurs
when the stress-intensity factor attains a critical value, the maximum adhesive stress
would decrease as the layer thickness increases, consistent with the various
experimental findings shown in Section 1.2.2.
Since the comer stress-intensity factor uniquely characterises the deformation at an
interface comer, it would serve as a failure criterion for bonded joints provided that the
size of the process zone is comparable to the comer singularity zone. Accordingly the
joint strength would decrease as the adhesive thickness increases, i.e.,
20
10
8
~ c,
to" 5
"'"'
~
0.
0
0
..<::: 2
comer c .
"0
.~
'"
§
0
z
0.5
symbols: FE results
line: predictions
0.0001 0.001 0.01 0.1
Distance along the interface ahead of corner x/h
corner cl
symbols: FE results
Iine: predictions
0.1 ~~--------------~~---------------
0.0001 0.001 0.01 0.1
Distance along the interface ahead of corner xlh
Figure 1.9 Distribution of(a) normal stress and (b) shear stress along the interface ahead ofthe
corners C 1 and C4 in a double-lap joint.
( 1.20)
where a is given by equation (1.15). For Poisson' s ratio of 0.35, the order of
singularity a is equal to 0.32. In this case prediction of the corner stress-intensity factor
criterion is shown in Figure 1.2. This prediction seems to provide a lower bound to all
the experimental results. This is most likely due to the ignorance of the spew fillet.
1.3.2.2 Spew fillet

Studies have shown that spew fillets formed during bonding at the ends of overlap may
have an important influence on the joint strength [1, 12, 41-42]. Formation ofthe spew
fillet not only modifies the stress distribution along the mid-plane of the adhesive layer
[4], but may also eliminate the corner singularity. Referring to Figure 1.10, while the
stresses at corner D are singular, the stress-state at point C depends strongly on the
angle </>. In the following this two points will be examined in turn.
inner
adherend Y
adhesi e
layer
( X
pew
fillet
Figure 1.10 Configuration of a spew fi Ilet.
Assuming that the Young's modulus of the adhesive is far smaller than the modulus of
the adherend, the order of the stress singularity at corner C under plane-strain
conditions can be obtained from the solution of Williams [40], and the results are
shown in Figure 1.11. It is clear that spew corners would be free of singularities when
the fillet angle is less than a critical angle, which is dependent on the Poisson's ratio of
the adhesive. The value of this critical angle is shown in Figure 1.12. The absolute
maximum angle at which no corner singularity would exist for any Poisson's ratio is 45
degrees. Therefore, if the spew fillet angle can be controlled to be equal or less than 45
degrees, stresses at corners A and B in Figure 1.10 would be regular.
To verity that the singularity associated with a spew fillet, a finite element analysis
has been carried out using the finite element mesh shown in Figure 1.5, with a spew
fillet shown in Figure 1.13. The spew fillet angle at both points A and B is equal to 45
degrees. Contours of the resulting stress components, <1xx. <1yy, and 'rxy are shown in
Figure 1.13, confirming that the stresses at points A and B are no long singular.
Consequently corner singularity exists only at corner C. Solutions of the order of
singularity and the stress-intensity factor for this comer can be found in Wang and Rose
[39]. It is worth pointing out that this comer singularity can be removed by rounding of
comer C (see Adams et al [ 1].
0.5
-a
o-r
0.4 4
1:l 4
0
·;: I
I
:;"'
Cl)
0.3 v=0.3 I
I
s::
·;;; I'
....... I v=O
0.2 I
...
0
<!)
I
I
"E I
0 I
I
0.1 I
I
I
I
I
0
0 30 60 90 120 !50 180
Fillet angle e (degrees)

Figure 1.11 Order of corner singularity at spew fillet (points A and B in Figure 1.1 0).
90
,-.._
"' 80
...
<!)
<!)
Cl)
<!)
'2.-
70
""Q)
Oil
s::
"'
E
;:::
60
E
·;;:
:2"'
50
40
0 0.1 0.2 0.3 0.4 0.5
Poisson's ratio v
Figure 1.12 Maximum fillet angle 1/> for no corner singularity.

Figure 1.13 Stress contours near spew fillet.

It is clear in Figure 1.13 that comer singularity exists only at point C of Figure 1.10. It
has been found that there are two singular terms for the stresses at comer C [39]. The
order of the stronger singular term is approximately equal to 0.41 for typical Poisson's
ratio of 0.35. Consequently the critical comer stress-intensity factor criterion would
predict the joint strength to decrease with adhesive thickness via the following relation,
(1.21)
which is also shown in Figure 1.2.

From the preceding analysis it is clear that comer singularities can be completely
removed if the end of adhesive layer can be shaped so that all interface angles are less
than 45 degrees. An example of such an end that would be free of comer singularity is
shown in Figure 1.14. Further work is required to clarifY the influence of adhesive layer
thickness on the joint strength.
inner
adherend
adhe ive
layer
outer
adherend
Figure 1.14 Adhesive end free of corner singularity.
1.4 STRESS CONCENTRATION IN AD HE RENDS
Since the adhesive shear stress peaks at the end of adhesive layer, significant through-
thickness stress concentrations exists in the loaded adherend [2, 10, 42-43]. This is
schematically shown in Fig. 1.15 for a double lap joint and a single lap joint. Such a
stress concentration can be extremely detrimental to joint strength, especially when the
overlap terminates near another stress concentration site, such as a fastener hole.
Quantitative methods for evaluating the stress concentration in adherends are often
critical to the design and assessment of bonded joints. For example, to avoid adhesive
failure and achieve the maximum strength, bonded joints are mostly designed so that the
bonding strength is greater than the substrate strength [5]. In this case, it is paramount
to be able to determine the strength of the substrates, taking into account of the stress
concentrations. Kairouz and Matthews [42] reported that the load carrying capacity of
adhesively bonded joints with cross-ply substrates having 90° surface layer would be
dictated by the maximum longitudinal stress occurring near the ends of overlap.
Recently there have been instances where unexpected fatigue failures of metallic
substrates occurred immediately outside the end of a reinforcement patch, possibly
caused by the interaction between the high stress concentration induced by bonded
reinforcement and fastener holes situated near the ends of overlap.
According to Wang et al [10], a first order estimate of the maximum longitudinal
stress (in the loaded adherend) at the end of overlap is given by the following
integration, noting equation ( 1.11 ),
-!= r(x) _ 2 i =e-f3tAz -e- 2 ~ cos(2~)

a xx max - --dx- - r max dz (1.22)
· o x n o z
Ad he ive
/.I_ . . . . _
[ ax.,
__\ !
\....... ..
__ /
···--..................-···
(a) Double-lap joint
/'
\
/_I_ ... . _ _ __ <rxx
\
!
\ . ..............................··· _.../
__
(b) ir:gle-lap joint

Figure 1.15 Through-thickness stress concentrations in adherends showing (a) double-lap joint
and (b) single-lap joint.
For a typical Poisson's ratio of 0.35, the parameter y is equal to 2.1535. In this case, the
maximum longitudinal stress is dependent only on the value of f3tA. Equation (1.22)
can be readily evaluated numerically, and the results are shown in Figure 1.16. It is
clear that a considerable stress concentration exists in the loaded adherend as a result of
the shear deformation in the adhesive layer.
::::
.9
~
~ 5
~
rJl
rJl
~rJl
"0
<!)
3
-~
"3
§
z 1
0.00001 0.0001 0.001 O.Dl 0.1
Figure 1.16 Normalised longitudinal stress in adherend.
As an example consider a double-lap joint representative of a unidirectional boron

patch bonded to an aluminium plate. The thicknesses of the inner adherend, the outer
adherend, and the adhesive layer are respectively 6 mm, 1.3 mm, and 0.1 mm. The
Young's moduli of the inner adherend and the outer adherend are 72 GPa and 210
GPA, while the shear modulus and the Poisson's ratio of the adhesive layer are 700
MPa and 0.3. The load transfer parameter f3 is equal to 0.18 mm- 1, thus giving rise to a
stress elevation of 3.27-rmax. The normalized longitudinal stress along the adherend-
adhesive interface is shown in Figure 1.17, indicating a maximum stress concentration
factor of 1.42. According to equation (1.1), the maximum shear stress 'X" max= 0.536a=.
Consequently the stress concentration (=a xx,max I a= ) is estimated to be 1. 75, which is
higher than the finite element results. The first order solution can be improved if the
effect of spew fillet is taken into account. Assuming that the presence of spew fillet
would reduce the peak shear stress by 20%, the analytical estimate would come very
close to the finite element solution.
,,p
1.4 o'\
a' '
D ~
p
p
1.2
J'l
p
1.0
-0.2 -0.1 0 0.1 0.2 0.3
Coordinate x/t.
I
Figure 1.17 Longitudinal stress along the plate-adhesive interface (with spew fillet).
1.5 TRIAXIAL STRESSES AND PLASTIC YIELDING
In analysing the stresses in a bonded lap joints, often only two stress components, the
peel stress and the shear stress, are considered. However, due to the constraint imparted
by the stiff adherends, the stress state in an adhesive layer is truly triaxial [1, 27, 46].
The lateral stress components parallel to the interface give rise a high hydrostatic
tension. This hydrostatic stress is important in determining the plastic yield behaviour
of polymers [44], affecting both the crazing mechanism but also the glass transition
temperature. In this Section the triaxial stress state within an adhesive layer is first
examined, and then the plastic yielding behaviour of adhesive under triaxial stresses is
discussed.
Referring to Figure 1.18 which shows a unit cell representing an adhesive layer
sandwiched between two stiff adherends, the two lateral stress components, a xx and
azz, can be expressed in terms of the peel stress a w (Wang and Rose, 1997; Adams et
a!, 1997),
(1.23)
Consequently the hydrostatic stress is

I l+v
p =-(
3
(j XX + (j \ '\' + (j zz) =
·· 3(1-v)
(jY'"
J
(1.24)
X
1-y
Adherends
Figure 1.18 A unit cell representing an adhesive layer sandwiched between substrates
To quantify the influence of hydrostatic stress on the plastic yielding behaviour of

polymer adhesive, a series of experiments have been carried out on a film adhesive
(FM73), under various combination of tension, compression, and shear loading [44].
The results are shown in Figure 1.19, where the equivalent shear stress is defined as
(1.25)
It can be seen that when the hydrostatic stress is negative (in compression), the
equivalent shear stress is approximately constant, indicating that the conventional von
Mises yield criterion is applicable. However, under tensile hydrostatic stress, the
equivalent shear stress decreases rapidly as the hydrostatic stress increases. These data
suggest that the modified von Mises yield criterion [45] is applicable for positive
hydrostatic stress,
(1.26)
where rys denotes the yield stress under shear (zero hydrostatic stress), and the
experimental data shown in Figure 1.19 suggest that the coefficient J.1 is approximately
1.13 for FM73 adhesive.
One important implication of the triaxial stress state in an adhesive layer is that it
would have a significant effect on the onset of plastic yielding. Denoting the shear stress
and the peel stress as r and cr , and recalling equations ( 1.1.23 ), the equivalent shear
stress can be written as
2]1/2
rm =r[1+_!_(1-2v (J) (1.27)
3 1-v r
where the ratio cr /r can be obtained from equations (1.1-1.2) for double-lap joint and
equation (1.4-1.5) for single-lap joint. Then the modified von Mises yield criterion can
be expressed as
2]1/2 + cr }
r) r
1 l-2v cr 1+v
r {[ 1+- - - - p- =r (1.28)
3( 1-v 3(1-v) ys
which is shown by the solid curve in Figure 1.20. For comparison purposes, two other
cases corresponding to the original von Mises yield criterion (!l=O) and an adhesive of
lower pressure sensitivity are also plotted in figure. It is clear that there is a substantial
reduction of the shear stress at yielding due to the effect of hydrostatic tensile stress.
This implies that if the design limit of a lap joint is the onset of plastic yielding, the
effect of peel stress on the allowable shear stress has to be considered. Likewise, if the
peel strength turns out to be the limiting factor, the design allowables for the peel stress
should also take into account the effect of hydrostatic tension.
m
40 • •
'
• ••• •
!
ll.
~
35 t shear • •----!hear tension
,.,"E 30 neat compression •

vi
(/)
compression
pure shear
•
•
....
Q)
25
1il • 'neat tension
....
«<
Q) 20
..c
(/)
1m= 38.6 + 1.13*p
c Q)
15
rn
.:::::::l 10 constrained tension - r/J •
w
0" •
5
-30 -20 -10 0 10 20 30
Hydrostatic stress p (MPa)
Figure 1.19 Influence of hydrostatic stress on yield stress of FM73 adhesive.

"'>. 1.0 ........_ --- ---

-.!::: ........_ 11=0
--
1-'
"0 ........_
~ 0.8 ........_
·:;;:. - ---~=0.2
1;j ........_
"'"'
<!) 0.6
!::::
"'til
<!)
..s:: 0.4
"'
"0
<!)
.:!l
a
"(;
0.2
z0 v=0.35
0
0 2 3
Ratio of peel stress to shear stress aft

Figure 1.20 Effect of triaxial stress on the value of shear stress at yielding
1.6 REFERENCES
1. Adams, R. D., Comyn, and Wake, W. C. (1997). Structural Adhesive Joints in

Engineering, Chapman & Hall, London.
2. Cornell, R. W. ( 1953). Determination of stresses in cemented lap joints, Journal of
Applied Mechanics Vol.20. 355-364.
3. Goland, M and Reissner, E. (1944). The stresses in cemented joints, Journal of
Applied Mechanics Vol.ll, Al7-A27.
4. Adams, R. D. and Peppiatt, N. A. (1974). Stress analysis of adhesive-bonded lap
joints, Journal of Strain Analysis Vol.9. 185-196.
5. Hart-Smith, L. J. (1973a) Adhesive-bonded double-lap joints, NASA Report CR-
112235.
6. Hart-Smith, L. J. (1973b) Adhesive-bonded single-lap joints, NASA Report CR-
112236.
7. Hart-Smith, L. J. (1982). Induced Peel Stresses in Adhesive-Bonded Joints.
Douglas Aircraft Company, 1982, MDC J9422A.
8. Gilibert, Y. and Rigolot, A. (1988). Determination of stress distribution in double-
lap joints, matched asymptotic expansions and conformal mapping, Adhesively
Bonded Joints: Analysis and Design, ASTM STP 98!, American Society for Testing
and Materials, Philadelphia. 145-159.

9. Roberts, T. M. (1989). Shear and normal stresses in adhesive joints, Journal of
Engineering Mechanics, Vol.115, 2460-2479.
10. Wang, C. H., Heller, M. and Rose, L. R. F. (1998) Substrate stress concentrations
in bonded lap joints, Journal of Strain Analysis, Vol.33, 331-346.
11. Hein, V. L. and Erdogan, F. (1971) Stress singularities in a two-material wedge.
International Journal ofFracture Mechanics 7. 317-330.
12. Groth, H. L. (1988). Stress singularities and fracture at interface comers in bonded
joints. International Journal of Adhesion and Adhesives 8. 107-113.
13. Reedy, E. D. Jr (1990). Intensity of the stress singularity at the interface comer
between a bonded elastic and rigid layer, Engineering Fracture Mechanics, Vol.36.
575-583.
14. Reedy, E. D. Jr (1993) Free-edge stress intensity factor for a bonded ductile layer
subjected to shear,.!. Appl. Mechanics, Vol.60, 715-720.
15. Akisanya, A. R. and Fleck, N. A. (1997). Interfacial cracking from the free-edge of
a long hi-material strip. International Journal of Solids and Structures Vol.34.
1645-1665.
16. Allman, D. J. (1977). A theory for elastic stresses in adhesive bonded lap joints,
Quarterly Journal of Mechanics and Applied Mathematics Vol.30. 415-436.
17. Kinloch, A. J. (1993). Adhesive and Adhesives, Chapman and Hall, London.
18. Adams, R. D. (1989) Strength predictions for lap joints, especially with composite
adherends, A review . .!. Adhesion, Vol. 30, pp.219-242.
19. Adams, R. D. (1992) Strength predictions in bonded joints- current thoughts and
research at the University of Bristol, UK,.!. Adhesion, Vol.37, pp.47-50.
20. Bigwood, D. A. and Crocombe, A. D. (1990) Non-linear adhesive bonded joint
design analyses, Int. .!. Adhesion and Adhesives, Vol. 10, No. 1, pp.31-41
21. Jones, R., Chiu, W.K. and Paul, J. (1993) Designing for damage tolerant composite
joints, Composite Materials, Vol.25, pp.201-207.
22. Clark, J. D and Mcgregor, I. J. (1993) Ultimate tensile stress over a zone: a new
failure criterion for adhesive joints,.!. Adhesion, Vol. 42, pp.227-245
23. Bennett, W. F. (1972) Non-destructive testing of adhesive bonding, paper
presented at the 9th Conference on Non-destructive Testing, Loughborough
University, UK.
24. Crocombe, A. D. and Moult, A. C. (1988) The effect of the adhesive thickness on
the strength of bonded joint, Proceedings of Adhesion 12, edited by K.W. Allen,
Elsevier Applied Science.
25. Hylands, R. W. and Sidwell, E. H. (1980) The effect of glue-line thickness on
bonded steel-to-steel joints,.!. Adhesion, Vol.l 0.
26. Harris, J. A. and Fay, P.A. (1992) Fatigue life evaluation of structural adhesives for
automotive applications, Int. .!. Adhesion and Adhesives, Vol. 12, No.1, pp.9-18.
27. Wang, C. H. and Rose, L. R. F. (1997a) Determination of triaxial stresses in
bonded joints, International Journal ofAdhesion and Adhesive, Vol.17, pp.l7-25.
28. Imanaka, M., Fukuchi, Y., Kishimoto, W., Okita, K., Nakayama, N. and Nagai, H.
(1988) Fatigue life estimation of adhesively bonded lap joints, .!. Engng. Mater.
Techn., Vol.llO, pp.350-354.

29. Krenk, S., Jonsson, J. and Hansen, L. P. (1996) Fatigue analysis and testing of
adhesive joints, Engineering Fracture Mechanics, Vol. 53, No.6, 839-872.
30. Mall, S., Johnson, W. S., and Everett, R. A. Jr. (1982) Cyclic debonding of
adhesively bonded composites, NASA Technical Memorandum 84577.
31. Johnson, W. S and Mall, S. (1985) A fracture mechanics approach for designing
adhesively bonded joints, Delamination and debonding of Materials, ASTM STP
876, W. S. Johnson Ed., American Society for Testing and Materials, Philadelphia,
pp.189-199.
32. Wang, C. H. (1997) Analysis of cracks in constrained layers, Int. J Fracture,
Vol.83, pp.l-17.
33. Wang, C. H. (1998) On the fracture of constrained layers, Int. J Fracture, Vol.93,
227-246.
34. Johnson, W. S. (1986) Stress analysis of the cracked lap shear specimen: an ASTM
round robin, NASA Technical Memorandum 89006.
35. Femlund, G. and Spelt, J. K. (l991a) Analytical method for calculating adhesive
joint fracture parameters, Eng. Fract. Mech., Vol.40, No.1, pp.ll9-132.
36. Femlund, G. and Spelt, J. K. (1991 b) Failure load prediction of structural adhesive
joints, part 1: analytical method, Int. J Adhesion and Adhesives, Vol. 11., No.4,
pp.213-220
3 7. Edde, F. and Verreman, Y. ( 1992) On the fracture parameters in a clamped cracked
lap shear adhesive joint, International Journal of Adhesion and Adhesive, Vol.12,
43-48.
38. Reedy, E. D. Jr and Guess, R. T. (1997) Interface comer failure analysis of joint
strength: effect ofadherend stiffness, Int. J Fracture, Vol.88, 305-314.
39. Wang, C. H. and Rose, L. R. F. (2000) Compact solutions for the comer singularity
in bonded lap joints, International Journal ofAdhesion and Adhesives 20, 145-154.
40. Williams, M. L. (1952). Stress singularities resulting from various boundary
conditions in angular comers of plates in extension. Journal of Applied Mechanics
19. 526-528.
41. Adams, R. D. and Harris, J. A. ( 1987). The influence of local geometry on the
strength of adhesive joints. International Journal of Adhesion and Adhesive 7. 69-
80.
42. Kairouz, K. C. and Matthews, F. L. (1993). Strength and failure modes of bonded
single lap joints between cross-ply substrates. Composites, Vol.24. 475-484.
43. Adams, R. D., Atkins, R. W., Harris, J. A., and Kinloch, A. J. (1986) Stress
analysis and failure properties of carbon-fibre-reinforced-plastics/steel double-lap
joints, J Adhesion 20, 29-53.
44. Wang, C. H. and Chalkley, P. (2000) Plastic yielding of a film adhesive under
multiaxial stresses, Int. J Adhesion and Adhesive, Vol.20, 155-164.
45. Bowden, P. B. and Jukes, J. A. (1972) The Plastic Flow of Isotropic Polymers,
Journal of Materials Science, Vol.7, 52-63.
46. Wang, C. H. and Rose, L. R. F. (1997b) Failure analysis of adhesively bonded

joints, Advances in Fracture Research, Vol.6, pp.3057-3064, Pergamon Press.
CHAPTER 2 STRENGTH DETERMINATION OF ADHESIVE BONDED JOINTS
L. TONG", J. K. SPELTb and G. FERNLUNDc

a School of Aero~pace, Mechanical and Mechatronic Engineering, The University of
Sydney, New South Wales 2006 Australia
b Department of Mechanical Engineering, The University ofTotonto, Ontario, Canada
c Department of Metals and Materials Engineering, The University of British Columbia,
Vancouver, Canada
2.1 INTRODUCTION
An adhesively bonded structural joint or repair is often one of the most important parts
in a load-carrying structure. Adequate design of structural joints and repairs is essential
for accomplishing the required performance of the structure because joints often impose
a significant weight penalty and play an important role in establishing structural
efficiency. To design efficient joints and to better utilize the advantages of adhesive
bonding, it is imperative to develop simple and reliable methodologies that can provide
accurate and reliable estimations of joint strength and service life. In this chapter, we
aim to address this important issue by: (a) discussing failure modes and mechanisms;
(b) summarizing failure criteria available; and (c) examining their application to
predicting joint strength and service life.
Structural bonded joints and repairs can fail at different location and in a variety of
failure modes. Failure can occur or initiate in the adhesive or in the adherend,
depending on geometrical configuration, materials of the adherends, adhesive used as
well as the manufacturing procedure. It is difficult to definitely describe and define all
the possible failure modes of adhesive bonded joints, particularly for bonded composite
joints or repairs. However, in order to describe various existing failure criteria, we can
classifY the failure modes of adhesive bonded joints into the following four general
groups: (i) adherend failure due to bending, tension or compression etc, (ii) adherend-
adhesive interfacial failure, (iii) cohesive failure, and (iv) out-of-plane adherend failure
due to delamination in composite adherends.
Adherend-adhesive interface failure often takes place at a macro scale when surface
preparation or material quality is poor. In this chapter it is assumed that all joints are
27
L. Ton); and C. Soul is (eds.).
RecentAdvances in Structural Joints and Repairs for Composite Materials, 27-66.
:[;) 2003 Kluwer Academic Publishers.
28 L. Tong, J. K. Spelt and G. Fernlund
manufactured to specifications, and consequently the adherend-adhesive interface

failure can not be predicted using theoretical techniques.
When joining polymer composite members, particularly thin-walled composite
members, an adhesive bond 1s often selected as the preferred way of structural joining
because the adhesive is similar to the polymer matrix. However, due to the laminated
nature of composite members and their relative weakness in the through-the-thickness
direction, the failure mechanism of adhesive bonded composite joints becomes more
complex than that of a metal-to-metal joint. Therefore, more factors need to be
considered in strength prediction of composite joints due to complexity in material
properties.
Accurate prediction of bonded joint strength and service life is one of the most
challenging tasks in structural design. First of all, it is not an easy job to with certainty
determine the accurate geotnetry of the adhesive spew fillet as it can significantly
change the stress concentrations at the end of adhesive. Secondly, it is difficult and
expensive to characterize adh~sive properties such as stress-strain curves and material
allowables including fracture toughness, especially when environmental effects are
present. One reason for this is the discrepancy between properties measured using
various test methods, although there are a number of standard testing methods as
described in [1,2]. Lastly, but not least, there is a lack of understanding and accurate
modelling of failure mechanisms in bonded joints. Joint failure mechanisms include
onset of initial failure, such as yielding, fracturing, and cracking, followed by stable
growth of crack(s) or development of yielding zone(s), and ultimately fast or
catastrophic crack propagation. The failure mechanism becomes even more complex
when one of the joining members is a laminated composite material. In this case,
additional factors, such as lay-up sequence, fiber orientation, material properties of each
ply, affect the failure mechanism.
There are a large variety of failure criteria in the literature that have been used for
strength prediction of adhesive bonded joints. As mentioned before, the methods of
predicting joint strength and service life can be divided into two major groups, one is
the strength of materials based approach and the other is the fracture mechanics based
approach. The former one has often been used to predict static or dynamic joint
strength, while the latter one has frequently been used to predict static strength and
service life for bonded joints. This chapter aims to present some of the failure criteria
available in the literature from both areas and to discuss their applications to selected
joints. It is certainly not possible to include all the existing failure criteria due to length
limitation.
2.2 F AlLURE OF ADHESIVE JOINTS
2.2.1 COMMONJOINTGEOMETRIES
Structural adhesive joints typically consist of thin substrates bonded together in a lap or
strap-joint geometry when the substrates are in the same plane. Examples of common
Chapter 2 Strength Determination of Adhesive Bonded Joints 29
lap and strap joint geometries are shown in Figure 2.1.
Figure 2.1 Single-lap, double-lap, single-strap, and double-strap joints (from top to bottom)
If the substrates meet at an angle, an L-joint or aT-joint is often used, see Figure 2.2.
Similar joint types are also used when the substrates are thin walled tubes, rather than
plates. Examples of tube joints are found in reference [I, 2].
i
i
!
Figure 2.2 Examples of L-and T-joints
The geometry of a structural bonded joint is mainly driven by the relatively low strength
of the polymeric adhesive compared to the strength of the bonded substrates. Structural
adhesives rarely have a strength that exceeds 100 MPa, which is typically substantially
less than the strength of the substrates. Because of this, the bonded area has to be
substantially larger than the load carrying cross-sectional area of the substrates to
prevent premature failure in the adhesive bond. This design feature is clearly seen in the
joints in Figure 2.1 and Figure 2.2.
2.2.2 SHEAR AND PEEL STRESSES IN THE ADHESIVE LAYER
For the joints shown in Figure 2.1 and Figure 2.2 the stress distribution in the adhesive
layer is highly non-uniform with high stress concentrations at the ends of the bonded
overlap. The two main reasons for this are the abrupt change in geometry at the end of
the bonded overlap and axial deformation (stretching) of the adherends. Figure 2.3
30 L. Tong, J. K. Spelt and G. Femlund
shows schematically the distribution of shear stresses in the adhesive layer for a lap
joint with elastic adhesive layer in the cases where i) the adherends are rigid, and ii) the
adherends are deforming axially.
Figure 2.3 Schematic of the shear stress distribution in the adhesive layer of a lap-joint with
elastic adhesive: i) rigid adherends, ii) deforming adherends.
If the substrates are rigid and don't deform axially, the shear stress in the adhesive layer
is uniform throughout the bonded overlap, curve i) in Figure 2.3. In reality, however,
the adherends will deform under load and the shear stress distribution will be non-
uniform, curve ii) in Figure 2.4. Since the area under the shear stress distribution curve
must equal the load applied to the substrate per unit width, a joint with deforming
adherends will always have a higher shear stress at the end of the overlap than a joint
where the substrates are rigid. The stress distribution in Figure 2.4 assumes that the
adhesive is elastic. If the adhesive layer is loaded beyond the yield point, the maximum
shear stress will reach a plateau value at some distance away from the end of the bonded
overlap. Analytical techniques for calculation of the stress distribution in an elastic-
plastic adhesive layer in a variety of joint geometries has been developed by Hart-
Smith, e.g. [7-9]
l ! +• • • • • +! l
M
Figure 2.4 Free-body diagram of a lap-joint showing shear stresses t , induced bending moments
acting on the substrates, M,and induced peel stresses, cr. Dashed line indicates neutral axis of
substrate.
Although the adhesive layer in lap and strap joints is nominally loaded in shear, there
will be induced peel stresses in the adhesive layer because load transfer between the two
substrates occurs via shear stresses at the adhesive/substrate interface rather than acting
through the neutral axis of the substrate. The off-set between the neutral axis and the
adhesive/substrate interface induces a bending moment in the substrates that tend to peel
the substrates away from each other, see Figure 2.4. The peel stresses in the adhesive
layer have a distribution similar to that of the shear stresses with a maximum at the end
of the bonded overlap [10]. The peel or transverse stresses are often tensile which has a
negative effect on the strength of the bondline and can cause delamination in fibre-
reinforced substrates with low transverse strength [4].
The high shear and peel stresses at the end of the bonded overlap are unfavorable
from a strength perspective and can be reduced by tapering the substrates towards the
end of the bonded overlap. A few common tapering techniques are shown in Figure 2.5.
Figure 2.5 Schematic of joints with tapered substrates
If cr* denotes the axial stress in the substrate away from the bonded overlap, I the length
of the tapered region, and t the thickness of the substrate (Figure 2.5), an approximate
expression for the shear stress, t, and peel stress, cr, in the adhesive layer is given by
[3]:
t"' cr*(tll) ; cr"' cr*(t/l) 2 (2.1)
Compared to joints with un-tapered substrates, tapering can give a substantial reduction
in both shear and peel stresses in the adhesive layer. However, tapering of the substrates
often add considerable cost to the manufacture of the joint since the substrates are
typically no more than a few millimetres thick. Tapering is more readily done with
composite substrates that can be manufactured with the taper included, in contrast to
metal substrates that needs to be machined to create the taper. Although tapering
reduces the high stress concentrations at the end of the bonded overlap, the end of the
overlap will still be the most highly stressed area of the bondline and failure will
subsequently initiate there.
2.2.3 TYPICAL FAILURE MODES AND M ECHA NISMS
The following discussion will focus on the bonding and failure of joints with high
strength substrates, where the strength and fracture toughness of the substrates are
substantially greater than that of the adhesive. This condition is almost always true for
joints with metal substrates, whereas for reinforced polymer substrates the stress
concentration at the end of the bonded overlap can cause delamination due to high
transverse stresses in the substrates [4].
If a lap or strap-joint bonded with a tough structural adhesive is loaded to failure, the
failure sequence in the bondline is as follows: yielding, damage initiation and growth,
followed by final fracture, see Figure 2.6. These events start at the end of the bonded
overlap and moves towards the centre of the joint.
F F
F
c
A OHI'~~/
A-8: Y1eld1ng
8 -C: Damage growth
C: Final fracture -
0
Figure 2.6 Schematic of the force-displacement (F-o) behaviour of a typical lap-joint and the
failure sequence of the adhesive layer.
The adhesive is elastic until the applied load, F, causes yielding at the end of the
adhesive layer. The yield zone then increases and extends towards the centre of the
joint. If the load is increased further, a crack will typically form at the spew fillet
perpendicular to the direction of maximum principal stress, or at one of the substrate-
adhesive interfaces.
In a tough adhesive, a series of microcracks then develop, starting at the end of the
overlap and continuing towards the centre of the joint. The microcracks then coalesce
into one macrocrack with a damage zone with microcracks ahead of it. When the load
is sufficiently high, the crack driving force exceeds the work of fracture, which includes
plastic deformation, microcrack formation and coalescence, and final fracture occurs.
The non-linearity of the force-displacement curve in Figure 2.6 is greatly exaggerated
for illustrative purposes. Depending on the joint geometry and the adhesive used, the
length of the stages shown in Figure 2.6 varies. For example, a very brittle adhesive
used in a joint where the crack driving force increases with increasing crack length may
have no detectable yielding or damage growth before final fracture occurs, whereas in
other cases final fracture occurs after a substantial amount of sub-critical crack growth.
Fracture tests where the bondline was monitored using a high magnification microscope
showed that aluminium lap and strap joints bonded with a mineral filled structural
epoxy, Cybond 4523GB, exhibited approximately 5 mm of sub-critical damage growth
and crack propagation prior to final fracture. Similar joints bonded with a rubber-
toughened epoxy, Permabond ESP 310, exhibited 50 mm of sub-critical crack growth
[6]
The fracture toughness of the in-situ adhesive layer in a lap or strap joint is typically
much higher than for the adhesive in bulk form . This is very advantageous for the
strength of structural bonded joints. One reason for this is the difference between the
desired and the allowed crack path in the adhesive layer. A crack in a homogeneous
material wants to propagate in a direction such that the crack tip is under mode I loading
[ 11 ]. In a lap or strap joint the direction of principal stress makes an angle of less than
45 ° with the length axis of the joint because of induced peel stresses, see Figure 2.7b.
t cr Direction of maximum
1 ~ principal stress
b)
T 1' Observed crack path
()
Figure 2.7 a) Typical micro-crack pattern; b) shear and peel stresses acting on the adhesive layer,
direction of maximum principal stress, and observed crack path.
Micro-cracks that are formed within the adhesive layer will thus propagate at an angle
towards the adhesive/substrate interface. When reaching the tougher interface/substrate
the crack is either arrested or will continue to grow under less favourable mixed mode
conditions along the interface. This crack arresting mechanism typically creates a large
number of angled cracks within the bond line that later coalesce and form a macro-crack.
The forming of a large number of smaller cracks and associated plastic deformation
allows for significant energy dissipation and creates a system with high fracture
toughness.
2.2.4 EFFECT OF JOINT TYPE AND JOINT GEOMETRY ON JOINT STRENGTH
Adhesive joints are in general structurally more efficient than joints using mechanical
fasteners because of reduction of stress concentrations. Although the stress distribution
in the adhesive layer is very non-uniform, structural adhesives exhibit a large plastic
shear strain, which reduces the effect of stress concentrations in the adhesive layer. The
advantage with joints using mechanical fasteners is that assurance of structural integrity
is easier to guarantee.
The strength of composite structures is typically determined by the strength of their
joints. When using mechanical fasteners, the strength of the joints are seldom even half
that of the surrounding structure, whereas when using adhesive joints, the joint can
normally always be made stronger than the surrounding structure by using tapering
techniques [4]. A schematic of how the bond strength is affected by joint type and
tapering is shown in Figure 2.8. In general, adhesive joints should always be designed
such that they are stronger than the surrounding structure to ensure some tolerance to
flaws and other imperfections of the adhesive bond.
Bond strength
Scarf joint
Tapered
strap joint
~t;;~
~0
<;;
~0
~'l>
~<:>
C?\j Double-strap joint
/-----------~~-------
Single-lap joint
Substrate thickness
Figure 2.8 Effect of joint type and geometry on joint strength, after L.J. Hart-Smith [9].
Single-lap joints are the least efficient joint type because the eccentricity of the load
puth causes bending of the substrates and high peel stresses in the adhesive layer. The
advantage with this joint type is that it is the easiest to make. Better performance is
achieved with double-lap and double-strap joints. Also in these joints, there are high
peel stresses that limit joint performance. Tapering of the adherends can reduce these
high peel stresses. The most efficient joints are scarf joints, which in principle allows
for a uniform stress distribution in the adhesive. By making the taper sufficiently long,
any thickness of substrate can theoretically be joined using a scarf joint without the joint
being the weak link. However, in reality the uniform stress distribution in the adhesive
make scarf joints susceptible to creep [3]. Instead of a continuous taper, the taper in the
scarf joint can be made in steps to facilit"te easier manufacturing.
2.3 STRESS ANALYSIS OF ADHESIVE JOINTS
2.3.1 INTRODUCTION
Stress analysis is an important task in the structural design of an adhesive bonded joint
in order to predict its failure strength and service life. Failure can occur in the adhesive
layer, the adherends and at adhesive-adherend interfaces. Failure is a progressive
phenomena during which a material gradually loses its load carrying capacity at
multiple scales, i.e., from micro cracking to macro crack formation and crack
propagation. To predict the failure strength of a structural joint under operational
loading and service conditions, it is important to obtain accurate information on stress
and/or strain distributions, particularly stress concentrations or even stress intensity
factor or energy release rates at a crack tip. It is also important to perform multi scale
stress analysis when it is desired to predict failure initiation.
There exists a large amount of information in the literature on stress analysis of
adhesively bonded joints. This section is not intended to provide an overview of the
available information on stress analysis. Readers who require further information are
referred to [1, 2]. For the purpose of completeness of this chapter, stress analysis
techniques are classified into two general categories, i.e., analytical approaches based
on mathematical simplifications for simple configurations and numerical methods using
the well developed finite element methods [4]. A brief review of stress analysis
techniques is given.
2. 3.2 ANALYTICAL APPROACHES
Figure 2.1 depicts the most common lap joints used in a variety of standard tests for
evaluating adhesive behaviour as well as in practical structures. The single-lap and
double-lap joints as shown in Figure 2.1 are simple in geometrical configuration and
representative of loading experience by structural adhesives in service, and have been
studied by many researchers [39,40,7,8,50]. The treatment of the adhesive and
adherends in terms of deformations, strains and stresses is the key issue in various
mathematical models.
The simplest model for a single lap joint is to model both adherends as rigid bodies
and the adhesive as subjected to pure shear. The uniform shear stress in the adhesive is
then
p
r=- (2.2)
2cb
where b and 2c are the width and length of the overlap region, and P is the tensile load
applied at the remote joint ends.
Volkersen [39] abandoned the rigid adherend assumption and assumed that the
adherends are allowed to deform elastically in the longitudinal direction. The
differential straining between the two adherends results in a non-uniform adhesive shear
stress distribution with a maximum at each end of the overlap. For a balanced single lap
joint with both adherends having the same thickness and Young's modulus, the ratio of
the maximum shear stress to the average shear stress, is given by
(2.3)
where E and Ga are the Young's modulus ofthe adherend and the shear modulus ofthe
adhesive respectively. t and T/ represents the thickness of the adherend and adhesive,
respectively. This is the well known shear lag model.
Go land and Reissner [40] presented a more complex and accurate solution for the
single lap joint of Figure 2.9 by following a two step approach. In the first step, the
effect of the adhesive layer is neglected as it is very thin compared to the adherend
thickness, and a global analysis of the stepped beam is performed by taking into account
large deformation of the adherends. The global analysis leads to development of
formuli for bending moment, transverse shear forces and longitudinal forces acting on
individual adherend at each end of the overlap. In the second step, it is assumed that (a)
both adherends undergo longitudinal and transverse bending deformations, and (b) the
adhesive transfer both shear and normal or peel stresses, which are constant across the
adhesive layer. A local analysis is performed for the overlap of the single lap joint, and
closed-form solutions for both shear and peel stresses are obtained as follows:
1
~.:_=-- { -(1+3k)
f3c cosh~.-
X ~t c +3(1-k)
}
(2.4a)
pt 8 t "h~
sm ~
2
~ ( .:. ) =_!_
l(R A- !5._2 + Ak' coshA-cosA-)cosh A-~cosA-~1
2
2
c c (2.4b)
p t L1 + (R 11l12 -+ILK
k 1 ,_, Sin
. h 1 . 1 ) "nh 1 X .
llSlll l l Sl
1 X
!l-Slll ! l -
2 c c
where
[3 2 =8 Gc !._ = 6~ !_ A- c
rI
4
E ·,., ' r E · ,.,. ' =
R 1 =cosh A- sin A-+ sinh A- cos A R 2 =sinh A- cos A -coshA-sinA
L1 =_!_(sinh2A+sin2A) k = 2M 0 = 1
2 pt 2 l+2fi coth(uJ)tanh(u 2 c)
, V0 c 1 T/ k
k =-=-(l+-)--
pt2 2 t 2
and
ul-
-j¥21-v
,
~ ,
Er'
(a)
Mo
-L-1_ ______.Jii7 N
T=pt
(b)
~- T=pl
Figure 2.9 (a) an undeformed single lap joint considered by Go land and Reissner [40] and (b)
free-body diagrams subjected to applied load T
Hart-Smith [8] and Oplinger [50] coupled the first and second step in the Goland and
Reissner's analysis and each proposed a formula for the factor k of the edge bending
moment acting on the continuous adherend at an overlap end.
For a balanced double lap joint with ideal linear elastic-plastic adhesive shear stress-
strain behaviour, Hart-Smith [7] decoupled the two governing equations for both peel
and shear stresses based on the following observations: (a) peak shear and peel stresses
occur at an overlap end; and (b) shear stress near an overlap end becomes constant after
yielding. It was shown that yielding can take place through the entire adhesive layer
when the overlap is short, and that an elastic trough exists in the middle of the overlap,
between the two plastic regions near overlap ends, when the overlap is long. lt is
believed that the existence of such an elastic trough may improve the fatigue life of a
bonded double lap joint.
2.3.3 NUMERICAL METHODS
There exist a large amount of information on stresses and deformations in various

adhesive bonded joints based on various types of numerical methods, such as the finite
element method [5]. The finite element method is a powerful tool which is capable of
calculating stresses, strains and displacements in most structural joints as long as
computer resources are available. Typical issues in finite element modelling of
adhesively bonded joints include meshing scheme, material properties and geometrical
nonlinearity. In the meshing scheme, selection of element types to model adherends
and adhesive layers plays an important role as there are stress concentrations or even
singularities at adhesive-adherend bimaterial interfaces. The adhesive spew fillet can
also have a remarkable effect on stress concentrations near an overlap end and thus
should be modelled when computing resources allows it. Some bonded joints, such as
the double lap, butt and thick adherend joints undergo small deformations and can then
be analysed with a reasonable accuracy using a small deformation formulation. Others
like the single lap joint, the lap shear joint and the peel test joint, experience intrinsic
large deformation or large rotation even when subject to a small load. There exist a
large number of finite element analysis results in the literature, which demonstrate the
importance of taking into account large deformation and the nonlinear material
characteristics of joining materials [20,54].
2.4 FAILURE CRITERIA AND MATERIALS CHARACTERIZATION
2.4.1 iNTRODUCTION
It is difficult to accurately and reliable predict the static strength and the service life of
adhesive joints in use. There are several reasons for this: i) the global behaviour of
adhesive joints is often non-linear under high loads due to the eccentricity in the load
path; ii) the local stresses at the end of the bondline are complex due to abrupt changes
in geometry, hi-material interfaces, and the presence of a spew fillet; iii) the strength of
the in-situ bondline is dependent on the strength of the substrate/adhesive interfaces and
the degradation thereof during environmental exposure, in addition to the strength of the
adhesive itself; iv) it is difficult to experimentally determine the in-situ strength of the
bondline. Therefore, the accuracy of the predicted strength of a joint subject to static
loading that is not subject to environmental exposure is significantly greater than the
accuracy of the predicted life of a joint in service that experiences cyclic loading and
environmental exposure.
To predict the strength of a joint, the applied "effective load" acting on the joint is
calculated and compared to some measure of the "strength" of the joint using a failure
criterion. There are many different ideas of what good measures of the "effective load"
and the corresponding "strength" of adhesive joints are. This is reflected in the wide
variety of failure criteria suggested in the literature. However, most of them fall into
two general classes: those based on strength of materials and those based on fracture
mechanics. In the following we will only discuss failure criteria addressing failure of the
bondline itself. Failure criteria for predicting failure of the substrates are discussed in
detail in reference [I].
2.4.2 STRENGTH OF J1ATERIALS BASED CRITERIA
This is the traditional approach where the stress-state in the adhesive layer is calculated
using analytical or numerical techniques and used in a yield criterion. Simplifications
are often made in regards to the stress-state in the adhesive layer, the constitutive
relation used for the adhesive, and the complexity of the yield criteria. Strength of
materials based criteria generally assume that failure occur when the maximum
adhesive stress or strain at the end of the bonded overlap reaches a critical value. A
failure criterion that has been used in the aerospace industry is the maximum strain
criterion proposed by Hart-Smith [12]. The criterion is based on the assumption that, by
proper design of the joint, the pee I stresses can be reduced to the point that they do not
contribute to failure. The bondline is thus assumed to be under pure shear. Other
examples of stress or strain based failure criterion can be found in Refs. [ 13-20]. Lately,
there has been an increased use of finite element analysis in the study of adhesive joints.
When predicting failure using finite element analysis, the failure criterion is often
expressed in terms of an effective stress or effective strain, according to a von Mises or
a Tresca criterion. However, accurate determination of the stress state in the adhesive
layer has shown to be nontrivial task because of geometric non-linearities, hi-material
interfaces, and singular points [21]. Instead of using the stress or strain state at a point as
a measure of the effective loading of the bondline, several researchers have proposed
criteria based on the stress or strain state of a zone, e.g. [20,3 7,49] or the strain energy
density [7].
Using strength of materials is the classical approach to the prediction of joint
strength, and is appropriate when predicting onset of yielding and the development of a
plastic zone in the adhesive layer. However, most structural adhesives exhibit a large
amount of plastic deformation, damage development and sub-critical crack propagation
prior to final failure, which indicates that a fracture or damage mechanics approach may
be suitable for these systems.
2.4.3 FRACTURE MECHAlv!CS BASLD CR!TtRIA
The second group of failure criteria is based on fracture mechanics, usually linear elastic
fracture mechanics (LEFM). Several studies have reported that the total strain energy
release rate correlates well with failure [22-32]. Other investigators, however, question
the application of LEFM to structural adhesive joints because many adhesives are
toughened and ductile, and thus violate the underlying assumptions of small scale
yielding in LEFM. One study showed that the critical strain energy release rate could
be used to predict failure in specimens with a very brittle adhesive, if the crack is loaded
in mode I [33]. The same study also showed that uncracked specimens could be treated
using fracture mechanics, by assuming an inherent flaw size due to the inevitable
presence of small bubbles and voids in the adhesive layer. The approach thus enables
the use of fracture mechanics to analyze adhesive joints without visible macroscopic
cracks. The presence of a large plastic zone around the crack tip has also been
addressed in the literature [34]. It was proposed that fracture in the bondline is
controlled by the plastic zone developed at the crack tip, and that a relation exists
between the fracture energy and the bulk properties of the adhesive. Another approach
assumes that there are regions with very high stress concentrations due to the bimaterial
interfaces in adhesive joints. This led to the proposal that a fracture criterion be based
on a generalized stress intensity factor derived from material induced stress singularities
at the ends of bonded joints [35, 36].
There are a large number of studies and proposed failure criteria in the literature.
Most of them are idealized and can only be applied to a limited number of joint
geometries and loading conditions. A major limitation of many of the proposed failure
criteria is that they neglect that the strength of an adhesive bondline is dependent not
only on the properties of the adhesive, but also on the properties of the adherends,
adherend pretreatment, and adhesive cure. When discussing the strength of the
bondline, one must consider the whole "adhesive system" consisting of adhesive,
adherends, pretreatment, bondline thickness, and adhesive cure schedule; i.e., all the
pertinent parameters that affect the strength of the in-situ bondline. In the following
section, we will discuss a fracture mechanics based failure criterion that has been
proposed and verified by the authors for a large number of joint geometries.
2.4.3.1 Mixed-mode fracture criterion

There are three main criteria for an engineering approach to failure load predictions of
adhesive joints: i) it has to be applicable to a large class of joints; ii) it should be able to
cope with various joint geometries, and iii) it should be simple to use. A large class of
lap and strap joints can be reduced to generalized elements using the concept of an
adhesive sandwich [10], where the bonded overlap is isolated from the surrounding
structure as a free-body. All of the joint types discussed in this chapter can be treated in
a uniform way using this technique. The reactions acting at the end of the free-body
sandwich can be determined using analytical or numerical techniques depending on the
complexity of the geometry.
If the loads acting on the undamaged sandwich element (Figure 2.1 0) are increased,
bondline failure will initiate at the end of the overlap and progress in the fashion
described in Section 2.2.3. As failure progresses, a macro crack embedded in a region of
plastic and damaged material with a damage zone ahead of it is developed. When the
macro crack is fully developed it will translate down the bondline under steady-state
conditions.
The nature of the crack propagation including micro-cracking and plastic
deformation is dependent on the mode of loading. It is significantly easier for a crack to
propagate under mode I (opening) conditions, compared to mode II (in-plane shear) or
mode III (out-of-plane shear) conditions. The resistance to crack propagation in the
bondline can be quantified in terms of the critical energy release rate, Gc. However, Gc
cannot be considered a unique property of the bondline since it is dependent on the
mode of loading of the adhesive joint. The plot of Gc as a function of the mode ratio
for a given adhesive system will be denoted the "fracture envelope." The phase angle,
If!= tan - 1 (~G 11 jG,), is typically used to denote the mode ratio for mixed mode I -
mode II conditions. It is more convenient to work in terms of energy quantities rather
than in terms of stress intensity factors as the physical interpretation of the latter
becomes more difficult if the crack propagates along one of the substrate/adhesive
interfaces. When we discuss the mode of loading or mode ratio in mixed-mode cases we
will always refer to the global mode of loading. With this we mean the mode the crack
tip would see if it propagated in a straight line at the centre of the bondline. The local
mode of loading experienced by the crack tip can be very different if the tip propagates
at an angle to the length direction of the bondline or at one of the substrate/adhesive
interfaces. However, one can argue that for a given adhesive system the global energy
release rate and the global mode of loading controls the fracture process in the bondline.
B)
C)
Damage zone Macro crack

Figure 2.10 A) Lap and strap joints showing adhesive sandwich elements; B) Adhesive sandwich
element without crack subject to general loading; C) Cracked adhesive sandwich with fully
developed damage zone ahead of the crack.
The proposed approach to fracture load prediction of joints subject to general (mixed-
mode) loading is based on the assumption that the stress and strain state in the adhesive
layer (effective loading) of a joint is uniquely characterized by the applied energy
release rate, G, and the mode of loading (mode I, II, and III). The approach is thus based
on a two-parameter failure criterion, and proceeds as follow: IdentifY an adhesive
sandwich element in the joint and determine the adherend reactions at the ends of the
adhesive sandwich using analytical or numerical techniques. If the loads in the
substrates are less than what is required to fail the substrates, calculate G and the mode
ratio. For the calculated mode ratio, obtain from the experimentally determined fracture
envelope the critical energy release rate, Gc If the calculated G is less than GC' the joint
will not fail, otherwise failure by crack propagation in the bondline will occur. The
application of the method to specific joint geometries is described in more detail in a
later section.
2.4.2.2 Characterization of the mixed-mode fracture toughness

The in-situ fracture toughness of the adhesive is greater than the bulk fracture toughness
due to stress shielding and the constraints by the substrates on the crack path. Several
tests have been proposed for mode I, mode II and mixed mode I - mode II fracture
characterization of adhesive joints and composite laminates: The Double-Cantilever-
Beam (DCB) test [38], the End-Notch-Fracture (ENF) test [21], the Cracked-Lap-Shear
(CLS) test [21], the Mixed-Mode-Flexure (MMF) test [41], the Asymmetric-Double-
Cantilever-Beam (ADCB) test [42], the Fixed-Ratio-Mixed-Mode (FRMM) test [43],
the Varying-Ratio-Mixed-Mode (VRMM) test [43] and the Mixed-Mode-Bending
(MMB) test [44]. However, none of these tests allows fracture testing in the entire
mode-ratio range from pure mode I to pure mode II. A simple load-jig that allows for
fracture characterization of adhesive joints over the entire range of mode ratios using a
single double-cantilever-beam specimen has been developed [45]. An important feature
of the load-jig is that balanced specimens (specimens with substrates of equal thickness)
can be used for the entire range of mode ratios. This avoids the controversy over how
the mode ratio should be calculated for unbalanced specimens [46]. Furthermore, the
mode ratio is virtually independent of the crack length, which eliminates a source of
experimental errors.
The mentioned load-jig was used to determine the fracture envelopes, Gc(\jf), for
two different adhesive systems [47,48]. Both adhesive systems consisted of: 7075-T6
aluminium adherends of height 12.53 mm, pretreated using the optimized FPL-etch
according to AS TM D2651-79. The adhesives used were the mineral filled epoxy
Cybond 4523GB (American Cyanamid), and the rubber toughened epoxy Permabond
ESP 310 (Nacan Products Ltd). An even bondline of thickness 0.4 mm was obtained
using teflon spacers for bondline control. The specimens were made in batches, cured
at 150°C for two and a half hours, and cut to a width of 20 mm using a table saw. To
measure the crack length, the adhesive was painted with diluted white fluid and the
crack was monitored through a travelling microscope. Gc was calculated from the
applied load, F, at the instant of crack extension [4 7,48].
The fracture tests showed that Gc was dependent on the type of starter crack for both
adhesive systems. However, it was found that Gc reached a steady-state value,
independent of the crack length, after a number of crack extensions at a given phase
angle. This is illustrated in Figure 2.11 where Gc is plotted as a function of the crack
length for a specimen loaded in mode II, using the Cybond 4523GB adhesive system.
The rising R-curve shown in the Figure 2.11 is consistent with the damage
development in the adhesive layer illustrated in Figure 2.6. Since the length of the
bonded overlap for most structural joints is significantly greater than the 4-10 mm the
crack requires to obtain steady-state conditions, the steady state value of Gc is the value
most relevant to fracture load predictions and will be the value presented in the
following two fracture envelopes.
For specimens using the Cybond 4523GB adhesive system, Gc was determined for
steady-state crack extension at eight different nominal phase angles, \jf. Figure 2.12
shows the average value of Gc, plus minus one standard deviation, at each phase angle.
Figure 2.13 shows the corresponding data for the Permabond ESP 310 adhesive system.
The number of data points used to determine the fracture envelopes was greater than
250 for both adhesive systems.
600
D D oD D D
500
~cP
D
400 D
01
~ u
300
0
200
100
D
Crack Length (em)
Figure 2.11 Critical energy release rate, Gc, as a function of crack length for Cybond 4523GB
adhesive under mode II loading.
700
600 T
D
T 1
500 T 1
D
D
1
N
~ 400 TD
u T 1
0 300 D
T
D
1
200 2 1
100
0
0 10 20 30 40 50 60 70 80 90
'V = atan VGJjTG] [deg.]
Figure 2.12 Fracture envelope, Gc(\jf), Cybond 4523GB adhesive system

6000 T
IJ
J.
5000
4000
N
,-..,
...
_§
.., .2
'-'
u
3000 ...
0
.2
2000
IJ
1000 IJ
IJ
0~-.-.,-.--r-,--r-.--.-.
0 10 20 30 40 50 60 70 80 90
'If = atan YGij7Tij" [deg.]
Figure 2.13 Fracture envelope, Gc(\jf), Permabond ESP 310 adhesive system.
The fracture envelopes shown in Figure 2.12 and Figure 2.13 can be expressed with
good accuracy in terms of the following polynomial curve-fit:
Cybond 4523GB: Gc (lf/) = 212- 2.561f! + 0.1431f1 2 -7.75 x l0-4lf/ 3 (2.5)
Permabond ESP 310: Gc (lf!) = 808- 8.86lf/ + 0.747lf/ 2 -6.51 X 10-4lf/ 3 (2.6)
where 'II is the phase angle in degrees. Experiments have shown these fracture
envelopes are not dependent on the thickness of the adherends [45].
2.5 STRENGTH DETERMINATION FOR STATIC LOADING
In this section, several illustrative examples are presented to demonstrate applications of

various fracture mechanics based and strength of materials based failure criteria in static
strength prediction of selected adhesive bonded joints.
2.5.1 METAL-TO-METALJOINTS
Consider the balanced single lap and lap shear joints under tension shown in Figure
2.14. Using the solution given by Goland and Reissner [40], the longitudinal membrane
forces and the bending moments in the continuous adherend at an overlap end can be
determined.
p
p
~
~
L,
I· ·I· ·I·
L,
·I
(a)
u
p
L
-~
I· ·I· ·I
(b)
Figure 2.14 Configuration fo r single-lap and lap-shear joints under tensile loading (a) Single-lap
joint, and (b) lap-shear joint
The single-lap and lap-shear joints can be viewed as being comprised of two identical
adherends which are only bonded together in the overlap region of length /. Assuming
that the virtual non-bonded part(s) carry no load, the void between the continuous
adherend and the virtual adherend can be deemed as a crack. The energy release rates
of mode I and II at both ends can be calculated using the longitudinal force and bending
moment at the overlap end and the following equations [52]:
A P t M M2
G __I ( - + - - ' )2 G _ _ ; (2.7)
II - 16 A 2D ' I - 4D
I I I
where A 1 and D 1 are the axial stiffness and bending stiffness of each adherend. For the
single-lap joint, M; (i= I ,2) is the bending moments acting on the continuous adherend at
the left and right end of the overlap respectively, whereas for the lap-shear joint, Mi is
the bending moment acting on continuous adherend at the right overlap end. The joint
strength can be determined using the measured in-situ envelopes [45, 47, 6] or using the
following failure criterion [52]
(!!...L_ ]a+ (G" l -I

G /C
a
G JIC
(2.8)
where a is a constant. G1 and Gu are the energy release rate of mode I and mode II
respectively, and G1c and Guc are the critical values of the mode I and II energy release
rates. Considered are two failure criteria given in equation (2.8) with a = 1 and a = 2
respectively.
Femlund et a! [45] and Papini et al [6] measured the failure loads for a set of
balanced lap-shear and single-lap joints in tension. They also characterized the in-situ
mixed-mode failure envelope for aluminium to aluminium bonding with two adhesive
systems, i.e., Cybond 4523GB and Permabond ESP 310. The in-situ mixed-mode
failure envelopes for the two adhesive systems are obtained through curve-fitting and
are given by equations (2.5) and (2.6). Tables 2.1 and 2.3 lists the measured and
predicted fracture loads for the lap-shear joints and single-lap joints. Numbers in
brackets in the last three columns are relative difference in percentage in relation to the
measured failure loads.
The results in Table 2.1 show that, between the two criteria, namely, a= 1 and
a= 2 criteria, the a= 1 criterion gives an average relative difference 5.6%, while the
a = 2 criterion yields an average relative difference of 17.1 %. Thus the a = 1 criterion
seems to be more applicable for predicting the static strength for the lap-shear joints.
This is because the a = 1 criterion fits best to the measured failure envelope when the
phase angle varies from 49.1 o to 90° for the lap-shear joints [6].
Table 2.2 compares the fracture loads predicted using the in-situ failure envelope
and the a = 1 and a = 2 failure criteria and the measured fracture loads for the single-
lap joints made of the 7075-T6 aluminium adherends and Cybond 4523GB epoxy
adhesive system [45]. It is noted that the relative difference between the measured and
those predicted using the a= 2 and a= 1 are 11.9% and 5.9% respectively. Similar to
the case of the lap-shear joint results in Table 2.1, the a = 1 criterion seems to be the
more appropriate for predicting the static strength of these joints than the a = 2
criterion.
Table 2.1 Comparison of predicted and measured fracture loads (N/mm) for equal
adherend Ia~ shear joints with t= 12.54 mm and a width of20 mm
Geomet!}' (mm) Measured and ~redicted fracture load using various ty~es of criteria
L I Measured* Eqn (2.5)** a =l a=2
160 187 1100 971 (-11.7) 965 (-12.3) 1174 (6.7)
165 185 968 962 (-0.6) 955 (-1.3) 1164 (20.3)
187 159 876 883 (0.8) 876 (0.0) 1071 (22.3)
182 154 915 874 (-4.5) 866(-5.4) 1059 (15.7)
162 180 935 953 ( 1.9) 947 (1.3) 1153 (23.3)
143 197 1092 1013 (-7.2) 1008 (-7.7) 1221 (11.8)
213 127 726 786 (8.3) 777 (7.0) 951 (31.0)
154 192 1044 988 (-5.4) 982 (-5.9) 1194 (14.4)
143 196 1081 lOll (-6.5) I 006 (-6.9) 1218 (12.7)
220 118 769 758 (-1.4) 750 (-2.5) 916 (19.1)
133 210 1226 1055 (-13.9) 1050 (-14.4) 1267 (3.3)
187 157 966 878 (-9.1) 871 (-9.8) 1065 (10.2)
242 101 720 703 (-2.4) 695 (-3.5) 846 (17.5)
168 184 960 957 (-0.3) 950(-1.0) 1158 (20.6)
240 112 695 735 (5.8) 727 (4.6) 888 (27.8)
*Data taken from Femlund eta! [45]
**Results obtained using failure envelopes given by Femlund et al [45]

Table 2.2 Comparison of predicted and measured fracture loads (N/mm) for equal
adherend single lap joints (t=12.5 mm, width was 20 mm).
Geometry (mm) Measured and predicted fracture load using various types of criteria
L1 I L, Measured* Eqn (2.5)** a= 1 a= 2
141 86 125 572 553 (-3.3) 546 (-4.5) 646 (12.9)
144 84 125 555 545 (-1.8) 539 (-2.9) 638 (14.9)
178 82 141 547 522(-4.6) 515(-5.8) 610(11.5)
176 86 138 525 527(0.4) 521(-0.8) 618(17.7)
245 75 80 438 419 (-4.3) 414 (-5.5) 492 (12.3)
265 55 80 391 385 (-1.5) 380 (-2.8) 449 (14.8)
216 98 88 497 475 (-4.4) 469 (-5.6) 561 (12.9)
203 62 135 551 470 (-14.7) 464 (-15.8) 549 (-0.4)
185 109 97 544 521 (-4.2) 514 (-5.5) 615 (13.0)
219 74 96 469 445(-5.1) 439(-6.4) 522(11.3)
167 127 96 565 565 (0.0) 558 (-1.2) 670 (18.6)
134 56 108 530 491 (-7.4) 485 (-8.5) 570 (7.5)
202 105 57 496 453(-8.7) 447(-9.9) 533(7.5)
196 Ill 94 559 514 (-8.1) 507 (-9.3) 607 (8.6)
216 92 82 525 458 (-12.8) 452 (-13.9) 540 (2.9)
148 115 106 574 576(0.3) 569(-0.9) 679(18.3)
148 115 106 595 576 (-3.2) 569 (-4.4) 679 (14.1)
148 115 106 587 576 (-1.9) 569 (-3.1) 679 (15.7)
*Data taken from Femlund et al [45]
**Results obtained using failure envelopes given by Femlund et al [45]
In Table 2.3, the measured and predicted fracture loads are tabulated for the single lap
joints made of the 7075-T6 aluminium adherends bonded with Permabond ESP 310
adhesive system [6]. It is shown that the failure loads predicted by using the a= 1
criterion and the a= 2 criterion exhibit an average relative difference of 13.4% and
5.4% from the measured ones for the joints. In this case, the a= 2 criterion seems to
be appropriate for predicting the strengths of the joints manufactured using this adhesive
system.
2. 5.2 COMPOSITE- TO-COMPOSITE JOIXTS
2.5.2.1 Examples using point-based criteria
(a) Composite double lap joints

Quasi static tensile tests were conducted for double lap joint specimens as schematically
shown in Figure 2.15 with aspect ratio lIt being kept as 30 and 20 respectively. When
the aspect ratio remained constant, overlap length varied with adherend thickness. The
material used was the T650/F584 graphite/epoxy prepreg and FM-300K film adhesive.
Joint specimens of25.4 mm width were cut from the large bonded panels.
Table 2.4 tabulates the adherend lay-ups, adherend thickness, and the overlap length
for all specimens. The mechanical properties for a typical ply are listed in Table 2.5
[53]. The tensile and shear stress-strain curves for the adhesive are depicted in Figure
2.16. The tensile stress-strain curve was measured using the neat adhesive tensile
specimens, while the shear-strain curve was measured using the thick adherend lap
shear test.
Table 2.3 Comparison of measured and predicted fracture loads (N/mm) for equal-
adherend sin~le la,e joints (t=l2.7 mm).
Geomet!)' (mm) Measured and ~redicted fracture loads using various ~~es of criteria
L, I L2 Measured* Eqn (2.6)** a=l a=2
142 80 140 1408 1514 (7.5) 1289 (-8.5) 1429 (1.5)
142 80 140 1352 1514 (12.0) 1289 (-4.7) 1429 (5.7)
142 80 140 1361 1514(11.2) 1289 (-5.3) 1429 (5 .0)
142 80 140 1405 1514 (7.8) 1289 (-8.3) 1429 (1.7)
142 80 140 1450 1514 (4.4) 1289 (-11 .1) 1429 (-1.4)
227 81 148 1478 1452 (-1.8) 1222 (-17.3) 1361 (-7.9)
227 81 148 1481 1452 (-2.0) 1222 (-17.5) 1361 (-8.1)
227 81 148 1524 1452 (-4.7) 1222 (-19.8) 1361 (-10.7)
254 54 148 1379 1284 (-6.9) 1083 (-21.5) 1195 (-13.3)
249 59 148 1373 1315 (-4.2) 1108(-19.3) 1225 (-10.8)
227 64 171 1284 1367 (6.5) 1157 (-9.9) 1279 (-0.4)
172 117 172 1654 1721 (4.1) 1458 (-11.9) 1638 (-1.0)
200 90 171 1477 1534 (3.9) 1297 (-12.2) 1445 (-2.2)
237 53 171 1308 1301 (-0.5) 1102 (-15.7) 1214 (-7.2)
197 82 192 1493 1496 (0.2) 1268 (-15.1) 1408 (-5 .7)
166 140 165 1768 1867 (5.6) 1581 (-10.6) 1791 (1.3)
203 104 164 1647 1615 (-1.9) 1362 (-17.3) 1526 (-7.3)
230 80 164 1457 1457 {0.0) 1229 {-15 .6) 1367 (-6.2)
* Data taken from Papini et al [6]
**Results obtained using failure envelopes given by Papini et al [6]
1•- --
I+- - - --245 mm-------~·1
r- 2/ ----l ~I
Figure 2.15 Geometric configurations of the balanced double lap joint specimen [53]
75
60
~ 45
0..
~
~
~
~ 30
Vi --Peel
-----·Shear
15
0
0.0 0.1 0.2 0.3 0.4 0.5
Strain
Figure 2.16 Stress-strain curves for FM300-K film adhesive [53]
Table 2.4 Parameters for balanced double lap joint specimens with lit of 30 and 20
(dimension in mm)
Specimen Doubler Central Adherend Overlap
Group Lay-up lay-up Thickens t length I
A2 [0,90]s [0,90]2s 1.12 33.5
82 [0 2 ,90]s [0 2,90]2s 1.67 50.1
C2 [0 2,90 2]s [0 2,90 2]2s 2.23 67.1
A3 [0,90]s [0,90]2s 1.12 22.4
83 [0 2,90]s [0 2 ,90]2s 1.67 33.4
C3 [0 2 ,90 2]s [0 2,90 2 ]2s 2.23 44.6
Table 2.5 Typical material properties of a T650/F584 graphite/epoxy lamina (nominal

ply thickness is 0.14 mm)
Longitudinal modulus EL 156.5 GPa
Transverse Modulus £1 and E 7. 15.65 GPa
Poisson ratio V u and V L7. 0.324
Poisson ratio V Tz (neat resin) 0.35

5.19 GPa
Shear modulus Gu and GLZ
1.53 GPa
Shear modulus GT7. (neat resin)
2337 MPa
Longitudinal tensile strength Xr
Xc 1585 MPa
Longitudinal compressive strength
46.6 MPa
Transverse tensile strength Yr and Zr
lnterlam inar shear strength R and S 126.9 MPa
All specimens were loaded at a constant crosshead speed of 1 mm/min in an ambient

environment. Inspection of the failure surfaces shows that the typical failure loci are
near the adhesive ends and within the surface ply of the central adherend as shown in
Figure 2.17a. Interlaminar cracking typically occurs near the top of the surface ply and
then either propagates in that ply or kinks into the next ply, with broken fibers. Figure
2.17b depicts the typical stresses acting on an infinitesimal element taken from the
critical areas marked by the circles in Figure 2.17a. Amongst these three stresses, the
normal tensile stress (}2 is believed to be the predominant one that causes delamination,
whereas the shear stress rxz and the axial tensile stress crx contribute to delamination
through interacting with the normal tensile stress (}2 •
The minimum, maximum and average values of the measured failure loads per unit
width are summarised in Table 2.6(a) for all specimens. Specimens in group A2 and A3
had the smallest failure loads because they had the thinnest adherends and hence the
weakest membrane strength. Specimens in group B2 and B3 had the largest failure
loads, and specimens in group C2 and C3 failed at the intermediate load levels. It is
also seen that joints with thin outer adherends have higher joint efficiency [53].
/.. cr,+ ·· ·.
······· ...'\.
L..__ __ X
(",bf')~
(a) (b)
Figure 2.17 (a) Typical failure loci (denoted by hollow circles), crack path (denoted by dotted
lines), and (b) definition of stresses in the circled loci for a double lap joint specimen [53]
Table 2.6(a) Measured failure loads (kN/m) for double lap joint specimens
Specimen A2 82 C2
Min 691.0 1241.7 1147.0
Max 860.1 1522.2 141 6.2
Average 781.6 140 1.6 1264.0
Av. joint efficiency 59.7% 53.5% 48.3%
Specimen A3 83 C3
Min 761.5 1280.6 1222.8
Max 951.5 1470.6 1405.9
Average 864.8 1393.6 1357.1
Av. joint efficiency 66.1% 53.2% 51.8%
* Average joint efficiency equals to the ratio of the maximum stress outside
the joint to its pertinent strength allowable.
The adhesive and adherend sandwich model described in [53] is used to perform one-
dimensional analysis. Calculated are the stresses in the adhesive as well as stresses in
the adherends in the test specimens. The adhesive is modelled as being linear and
nonlinear. In the nonlinear analysis, the true tensile and shear stress-strain curves of
Figure 2.16 are used as the adhesive material data. In the linear elastic analysis, the
Young's modulus and shear modulus are taken as the initial slopes of both curves. The
following failure criteria are utilized to predict joint strength using the peel and shear
stresses in the adhesive layer and the axial stress in the surface ply.
(2.9.1)
(2.9.2)
(2.9.3)
(2.9.4)
(2.9.5)
(2.9.6)
A nonlinear plane-strain finite element model was developed to analyze the stresses and
strains in the joints for specimen groups A2, B2 and C2 using MSC/NASTRAN [56].
Large deformations are considered using the available geometrical nonlinear facilities.
Composite adherends are modelled as being linearly elastic and using ply-by-ply
meshing scheme in which a fine mesh is near the end of adhesive layer [53]. The tensile
stress-strain curve for FM300-K adhesive depicted in Figure 2.16 is used in conjunction
with the Von Mises yielding criterion to model the nonlinear behaviour of the adhesive
layer. Table 2.6b tabulates the failure loads per unit width predicted using the six failure
criteria in equation (2.9.1-2.9.6) for the following four cases:
• one-dimensional predictions using linear adhesive model,

• one-dimensional predictions utilizing nonlinear adhesive material properties,
• finite element predictions using the stresses in the centre of the surface ply element,
and
• finite element predictions employingthe stresses in the centre of the adhesive layer
element.
The numerical results given by the one dimensional analysis reveal that the failure loads
predicted using the linear adhesive properties are much smaller than the measured,
while the failure loads predicted using the nonlinear adhesive properties are generally
close to the measured results. In the case of finite element predictions, the results
predicted using the stresses in the centre of the adhesive layer element are in a
reasonably good correlation with the experimental results.
Table 2.6(b) Predicted failure loads (kN/m) for double lap joints
One dimensional analysis results [53]
Specimen A2 & A3 B2 & B3 C2&C3
Predicted failure load using nonlinear adhesive properties
Criterion 2.9.1 1054.9 1798.4 1716.2
Criterion 2.9.2 807.9 1339.3 1248.6
Criterion2.9.3 1047.0 1782.1 1700.8
Criterion 2.9.4 800.2 1323.8 1230.1
Criterion 2.9.5 886.3 1521.1 1471.3
Criterion 2.9.6 671.0 1158.3 1050.1
Predicted failure load using linear elastic adhesive properties
Criterion 2.9.1 528.8 717.2 712.1
Criterion 2.9.2 470.4 648.3 648.0
Criterion 2.9.3 527.3 715.7 710.6
Criterion 2.9.4 468.7 646.5 646.2
Criterion 2.9.5 513.7 706.9 702.9
Criterion 2.9.6 454.2 636.1 636.7
Finite element analysis results (surface ply element)
Specimen A2 B2 C2
Criterion 2.9.1 1100.2 1865.6 1785.4
Criterion 2.9.2 835.1 1351.0 1244.2
Criterion 2.9.3 1090.9 1848.8 1766.9
Criterion 2.9.4 825.4 1334.4 1220.0
Criterion 2.9.5 894.8 1498.7 1422.7
Criterion 2.9.6 696.3 I 046.6 987.3
Finite element analysis results (adhesive element)
Specimen A2 B2 C2
Criterion 2.9 .1 1011.8 1642.5 1490.0
Criterion 2.9.2 710.5 1018.7 871.6
Criterion 2.9.3 1002.3 1627.9 1471.5
Criterion 2.9.4 701.1 990.6 857.4
Criterion 2.9.5 815.3 1317.3 1142.9
Criterion 2.9.6 611.3 817.6 765.7
(b) Composite panel-flange joints

Tong et al [54] investigated adhesive-bonded composite panel-flange joints in
compression. Figure 2.18 depicts the typical joint configuration representing a section
of the stiffened plate and the test set-up. The material used for the specimen is
T300/934 graphite/epoxy unidirectional tape with a nominal ply thickness of0.188 mm.
FM300-K film adhesive is used to secondary bond the panel and flanges. The adhesive
properties are £=2.4 GPa and V=0.3, and the nominal adhesive thickness is 0.12 mm.
Table 2.7 tabulates the lay-ups of the panel and flange for each specimen group. The
material properties for each ply are given in Table 2.8. Detailed information on
manufacturing and testing results can be found in Tong et al [54].
Panel All dimen ion inmm

Spec imen width 40 mm
I~ 40 .. 1.. 100
., , ,. 40 "' )lange
Figure 2.18 Adhesive-bonded skin-stiffener joint i ncompression
Table 2.7 Lay-ups of panel and flange in panel-flange joints

Specimen group Panel Flange
A (±45,0,90)s (±45,0,90)2s
B (±45,0,90)s (±45,0,90)s
c (±45,0,90)s (±45)s
D (±45,0,90)s (±45)2s
E (0,90,±45)s (0,90,±45)2s
f (0,90,±45)s (0,90,±45)s
G (0,90,±45)s (0,90)s
Table 2.8 T300/934 graphite/epoxy unidirectional tape

Mechanical properties Material strengths
EL = 127.5 GPa =
Xr 1586 MPa
Er = 9.0 GPa X c = 1517 MPa
E.= 4.8 GPa Yr = 80 MPa
Vu = VL, = 0.28 Yc = 80 MPa
vr, = 0.41 Z = 49 MPa
Gu = GL, = 4.8 GPa S= 96 MPa
Gr, = 2.55 GPa R=34.5 MPa
Finite element analysis is performed for all specimens using MSC/NASTRAN [56] and
the maximum normal stress criterion. Because the +45° and -45° plies in the laminates
of the bonded joints induce a coupling between bending and twisting, the geometric
symmetry of the joints can not be used to reduce the size of the FE models and the
assumption of plane strain is invalid. Consequently, a hybrid modelling strategy,

comprising the 4-node quadrilateral shell elements and 8-node solid elements, needs to
be used.
Figure 2.19 shows that there is a good correlation between the predicted and
measured load-displacement behaviour for the three specimens from group B. Table 2.9
tabulates the average measured and predicted failure load and displacement for all
specimen groups and shows there is a good correlation between the predicted and the
average measured failure loads. However, a poor correlation is noted between the
predicted and the measured failure axial displacements.
Table 2.9 Summary of the average buckling and failure loads and average deflections
(predicted values are given in brackets)
Specimen Buckling load (N) Failure load Axial deflection (mm)
Grou (N
A 1083 1188(1145) 2.3 (1.5)
B 928 l 053 (l 006) 3.3 (2.5)
c 632 686 (712) 15.7 (7.5)
D 821 923 (921) 4.2 (3.2)
E 1682 2051 (1953) 2.0 (1.2)
F 1469 1672 (1643) 3.0 (1.7)
G 1022 1278 (1310) 9.4 (3.7)
1200
1000
800
--Experimental B l
~
600
""""'0 - - - - - -Experimental B2
....l
400 - - - Experimental B3
200 e FEA
0
0 0.5 1.5 2 2.5 3 3.5
Axial displacement (mm)
Figure 2.19 Experimental and numerical comparison of load versus axial crosshead displacement
for specimen group B [54]
2.5.2.2 Examples using zone-based criteria
(a) Aluminiumjoints (Sheppard eta! [55})

Three groups of adhesive-bonded single lap joints were fabricated using aluminium
adherends and FM-300 film adhesive. All joints have the same free length of 38 mm,
overlap length of 13 mm, width of 25 mm, and adhesive thickness of 0.15 mm, except
for the adherend thickness as given in Table 2.1 0. The adhesive fillets formed at the
ends of the joints were removed using a soldering iron. The joints are designed to force
a cohesive failure for the purpose of validation of the damage zone model. All joints
were tested in an lnstron test machine under displacement control. Table 2.10 lists the
average failure load for the three joint groups.
In the finite element analysis, a geometric nonlinear analysis procedure was used to
correctly capture the large out-of-plane deflections experienced in some of the joints.
Material nonlinearity was also incorporated in each analysis to model the stress/strain
relation of the adhesive and the aluminium adherends. The Von Mises yield criterion
was used for the adhesive layer and the aluminium adherends. It was also assumed that
damage occurs when the Von Mises strain attains the ultimate value of0.2.
Table 2.10 Thickness (mm), average measured and predicted failure load (kN) of
adhesive-bonded aluminium single lap joints
Joint group Adherend Average measured failure Predicted failure load
thickness load
SLJI 1.6 I 0.4 10.4
SLJ2 3.2 11.4 11.9
SLJ3 10.0 13.7 13.4
Three- and four-node plate elements were used to create plane strain models for the
JOmts. Nine elements were used to model the adhesive layer in the through-the-
thickness direction at the end of the joint. A single 3-noded element is used at the ends
of the joints to model the small adhesive fillets and to eliminate the sharp comers at
these locations.
Numerical results show that the damage zone size is 0.0 I mm 2 corresponding to the
average measured failure load of SLJl joints. By setting the critical damage zone size
to be O.Olmm 2, the failure loads for the remaining two groups, SLJ2 and SLJ3, can be
determined. As shown in Table 2.1 0, the predicted failure loads are within 4% of the
average measured failure loads for all joint configurations. This represents a very good
correlation with the measured failure loads given the wide range of adherend thickness
used for the specimens.
(b) Composite joints (Sheppard et al [55])

The six groups of adhesive-bonded composite double lap joints described in Section
2.5.2.1 are examined again using the damage zone model. In addition, three groups of
adhesive bonded single lap joints are fabricated and then tested on an Instron machine
under the displacement control. Figure 2.20 shows the dimensions of the single lap
joints. Table 2.11 tabulates the lay-ups of the adherends, the overlap length and the
adherend thickness for the three groups of single lap joints. The materials used are the
same as those used for the double lap joints as given in Section 2.5.1. Table 2.12
summarizes the average measured failure loads for both single lap and double lap joint
specimens.
245 mm
~~
Width 25 mm
Figure 2.20 Adhesive-bonded composite single lap joint configurations
Table 2.11 Parameters of adhesive-bonded composite single lap joints

Joint Adherend lay-up Overlap length L Adherend thickness t
group (mm) (mm)
AI (0,90,90,0), 89.4 1.12
Bl (02,902,02)s 134.2 1.68
Cl (02,902,90,,02)s 178.8 2.24
Table 2.12 Comparison of the average measured and predicted failure loads for
composite single- and double lap joints using damage zone model
Joint group Average measured failure load Predicted failure load
kN (kN)
AI 16.25 13.1
Bl 21.0 23.5
C1 26.75 23.7
A2/A3 19.85/21.96 19.0
B2/B3 35.60/35.40 35.2
C2/C3 32.11 /34.47 35.4
To use the damage zone model for the composite JOmts, nine through-thickness
elements were used to model the surface adherend ply and three through-thickness
elements were used to model the adhesive layer at the end of the joint. In addition,
three elements were used to model the small adhesive fillets at the ends of the joints.
The detailed solution procedure can be found in [55]. By assuming that damage occurs
at an ultimate strain ofO.Ol, the critical size of the damage zone was determined to be
0. 0018 mm 2 . Using this critical damage zone size, the failure loads for all specimens
can be calculated and are tabulated in the third column in Table 2.12. Results in Table
2.12 shows that there exists a good correlation between the predicted and the average
measured failure loads, particularly for joints with thick adherends. For joints with thin
adherends, a 19% maximum difference is noted.
(c) Results using ultimate tensile stress over a zone

Clark and McGregor [49] proposed a failure criterion referred to as Ultimate Tensile
Stress over a Zone. They considered three joint configurations, i.e. single lap, double
lap and T-peel joints. The bulk adhesive properties used are £=3076MPa, v=0.3 and
the Ultimate Tensile Stress= 70MPa. For the adhesive fillet end shown in Figure 2.21,
it was postulated that failure occurs when the size of the zone with a maximum principal
stress exceeding 70 MPa attains 0.68mm. The critical zone size was determined using
the single lap joints, and then used to predict the failure loads for other joints made of
the same adhesive and adherend materials. Table 2.13 compares the predicted and
measured failure loads for the double-strap and T-peel joints. A good correlation is
found for different joint configurations.
trrs zone size 0.68 mm
Figure 2.21 Schematic of UTS zone size of 0.68mm.
Table 2.13 Comparison of predicted and measured strengths (N/mm) of various joint
configurations (Clark & McGregor, [49])
Joint configuration Predicted strength Measured strength
Single lap Calibration 320
Double-strap 70 I 740
T- eel 123.7 123
2.6 STRENGTH DETERMINATION FOR CYCLIC LOADING
2.6.1 FATIGUE MECHANISMS
Cracks can initiate and grow in adhesive layers under the influence of time-varying
loads that are much smaller than those that cause fracture under quasi-static conditions.
The mechanism for such fatigue cracking in adhesives is thought to begin with the
accumulation of damage due to local plasticity at sites of stress concentration. This
leads to microcrack formation and a damage zone similar to that described in Section
2.3 for quasi-static fracture. As fatigue continues, the microcracks coalesce into a single
main crack that advances as new microcracks continuously developing in the damage
zone ahead of it.
Fatigue in adhesive joints has been the subject of many investigations and the
references cited below are by no means exhaustive; they were selected to illustrate some
of the most common approaches and observations.
Studies with epoxy adhesives have shown that both the number of cycles (N) to
fatigue crack initiation, and subsequent crack propagation rates (da/dN, change in crack
length per cycle) correlate with the maximum strain energy release rate in the load
cycle, Gmax, [57,58] or AG=Gmax-Gmin [59]. The typical relation between Gmax and
daldN is typical of fatigue in other materials as shown in Figure 2.22 {sigmoidal s-
shaped curve}. Below a certain threshold value, G1h, cracks will not initiate, and
slightly above this point, the rate of crack growth increases very rapidly. This extreme
sensitivity to the applied energy release rate makes it difficult to predict total amounts of
crack extension accurately in loaded structures. Another complicating factor is the large
scatter typically found in fatigue data for adhesive joints; a range of studies have
measured fatigue crack propagation rates that vary by factors of2- 37 [57,60].
Log Gmax
Figure 2.22 Typical trend oflogarithmic crack growth rate, daldN, versus logarithmic maximum
energy release rate, Gmax· Threshold value of Gmax indicated as Gth·
It has been observed that fatigue crack initiation is not affected significantly by the local
geometry of the starting point; for example, similar initiation times were reported for
adhesive fillets, mode I precracks created by fast fracture, and precracks that resulted
from prior fatigue [57]. Crack initiation and growth in fatigue occur at Gmax values as
low as 10- 17% of the critical energy release rate, Gc, for slow fracture [57,58].
The effect of mode ratio on fatigue is dependent on the nature of the adhesive [57].
In some cases it can be neglected, and mode I fatigue crack growth data have been used
to predict crack growth in mixed-mode single-lap joints [58]. Because the quasi-static
critical energy release rate, Gc• is a strong function of the mode ratio, the fatigue trends
for a given adhesive can be different if the rate of propagation is reported versus
percentage of Gc as opposed to Gmax·
The effect of load frequency is dependent on adhesive ductility and strain-rate
sensitivity [59]. Adhesive creep can augment fatigue damage in low frequency testing.
2.6.2 FATIGUE FAILURE CRITERIA
The large scatter in most fatigue crack growth rates, their extreme sensitivity to the
applied strain energy release rate near the threshold, and uncertainty in the applied loads
often makes it difficult to base joint design on a prediction of the amount of crack
growth. A more practical approach is to design joints and surrounding structures to
keep the applied energy release rates below the threshold for fatigue crack initiation or
growth, Gth· This approach is further simplified by the observation that G1h appears to
be relatively insensitive to the mode ratio ofloading [57,58].
It is noted that, depending on the adhesive system, the presence of water can have
either an adverse effect on fatigue behaviour [61] or a positive effect due to
plasticization by water [62]. It is expected that changes in temperature or strain rate may
also affect fatigue behaviour since these alter the effective proximity to the adhesive
glass transition temperature. Therefore, the test conditions for fatigue data should
correspond as closely as possible to those expected in service.
2.6.3 DETERMINATION OF FATIGUE STRENGTH
Fatigue data can be generated using a variety of fracture specimens such as tapered or
straight double cantilever beams (DCB). Since G1h is not a strong function of mode
ratio, mixed-mode (asymmetric DCB) or mode I (symmetric DCB) tests can be used
equivalently, although there is evidence that a small amount of mode II can decrease
scatter by causing the crack to grow in a more defined path, closer to the more highly
strained adherend (see Section 4.2) [57].
Two approaches can be used to estimate Gth· Fatigue crack propagation rates
(da/dN) can be measured over a range of Gmax and then fitted to a curve that facilitates
extrapolation to zero crack growth at Gmax= Gth· If the data are in region II (Fig. 1), the
Paris law can be used: da/dN=C(Gmaxt, which is linear on a plot of log(daldN) -
log(Gmax). If the data span more than a single region of Fig. 1, then it may be useful to
use a relation that describes the entire sigmoidal curve, as described in ref. [58].
Alternatively, the Gmax corresponding to the onset of crack initiation (i.e. Gth) can be
measured directly by conducting experiments about region I (Fig. 1). Given the scatter
in most fatigue data, the two approaches should yield similar results [57].
2. 7 DESIGN FOR CREEP CRACK GROWTH
2. 7.1 CREEP MECHANISMS
Polymeric adhesives are viscoelastic materials and, as such, their fracture behaviour is
time and temperature dependent. In practical terms, this means that crack growth may
occur at sustained loads significantly lower than those producing quasi-static fracture as
measured using the approach in Section 4.2. The cracking pattern under conditions of
constant applied strain energy release rate, G, is similar to that described in Section 2.3
for quasi-static fracture; i.e. microcracks coalesce into a single main crack that advances
as new microcracks continuously developing in the damage zone ahead of it [63]. For
two single-part structural epoxies, however, it was observed that crack extension slowed
continuously at various mode ratios under constant G equal to 90% of the quasi-static
value, Gc [63]. This was interpreted as a self-toughening mechanism due to time-
dependent changes in the adhesive properties. A related form of self-toughening has
also been observed in quasi-static fracture [64,65].
2. 7.2 CREEP FAILURE CRITERIA
As in fatigue, a threshold value of G can be determined, below which creep crack

growth essentially stops. In work with two epoxy adhesives, this threshold value was
found to be a constant fraction of the quasi-static critical fracture energy, Gc,
independent of the mode ratio [66]. Therefore, testing at one mode ratio to measure the
fraction of Gc for zero (or an acceptably small) creep crack growth could be used to
estimate the threshold value of G at other mode ratios. Given the viscoelastic nature of
adhesives, the threshold value of G is expected to be a function of temperature.
2. 7. 3 DETERMINATION OF CREEP STRENGTH
A relatively convenient approach to measuring the threshold G for acceptably small

creep crack growth, is to use mode I wedge tests [66]. A stainless steel wedge is slowly
forced into a DCB specimen to generate a displacement, 8, (Figure 2.22) and crack
length, a, is recorded over time using a travelling microscope [66]. The applied mode I
energy release rate ( G=G1) may then be calculated at each crack length, a, using the
relation
G = 8 2 Eh\1- ha I h)\a + {3) 2

(2.1 0)
1 48(a 3 13+a 2 f3+af3 2 ) 2
where E is the tensile modulus of the adherends, ha and h, are the adhesive and adherend
Chapter 2 Strength Detennination of Adhesive Bonded Joints 61
thickness, respectively, and ~ is defined as
f3 = 0.667h{(l- h i hj[1 +(h i h0 )(2E I E" -1)]} 114 (2.11)
with Ea being the tensile modulus of the adhesive [66,67]. In cases where the adhesive
layer is thin and the adhesive is relatively stiff, as is true for most structural epoxy
joints, Equation (2.11) reduces to {3=0.667 h, and Equation (2.1 0) is equivalent to the
solution obtained for a cracked homogeneous DCB; i.e. one in which there is effectively
no adhesive layer [66].
As an example, Figure 2.24 shows the results of approximately 48 h of observations
plotted as crack growth speed, da!dt, versus the GIGc. Consideration of the relationship
between da!dt and the viscoelastic }-integral [66] leads to the conclusion that
(l
da I dt "" B ..!!.._
Gc
l/n
(2.12
where B and n are adhesive constants that, for the two epoxies of ref. [66], were found
to be independent of mode ratio, but are expected to be functions of temperature. In
order to detennine a threshold G corresponding to an acceptably small daldt, each of the
data sets of Figure 2.24 is fitted to Equation (2.12) and the resulting slopes, lin, are
averaged. The solid line was then drawn with this average slope and assigned an
intercept (shifted to the left) so that it lay just to the left of all the data points. This solid
line represents the design envelope giving the maximum predicted daldt for a given
fraction of Gc, and, as mentioned above, this was found to be independent of mode ratio
[63,66]. Therefore, if da!dt = 0.1 mmlday was acceptable in a particular application,
Figure 2.24 would predict that G should be kept smaller then 0.48Gc for this adhesive
system at room temperature.
L
ll
h . h,
h. h.
r__j 2ha
f a
o
Figure 2.23 Wedge loaded (mode I) DCB specimen. Note that is the relative displacement of the
DCB adherends; it is not necessarily equal to the wedge thickness (Figure 2 from ref. [66])
1()000
1000 Jl 00
c
a~00o0
00
c
i
100 ~
e 6,0 I
0 00
..
e
.._,
"0
10
~0
~ e8
0
0.1
0.01
"l
0
;g .....
c::i
CIO
c::i ~
G/GcE
Figure 2.24 Crack speed as a function of the applied energy release rate for the Permabond ESP
310 adhesive system. Each symbol refers to data collected from a single wedge test. The solid
line is slightly curved since it is a power-law function plotted on semi-log axes. (Fig. 10 from ref
[66])
2.8 REFERENCES
1. L. Tong and G.P. Steven, Analysis and Design of Structural Bonded Joints, Kluwer
academic publishers, 1999.
2. R.D. Adams, J. Comyn and W.C. Wake, Structural Adhesive Joints in Engineering,
2nd Edition, Chapman & Hall, London, 1997
3. D.W. Oplinger, Handbook of composites, ed. S.T. Peters, Chapman & Hall, 1998,
pp. 610-667.
4. L.J. Hart-Smith, Engineered Materials Handbook Vol. 1 Composites, ASM
International, 1987, pp. 479-495.
5. O.C. Zienkiewica and R.L. Taylor, The Finite Element Method, 41h Edition,
McGraw-Hill Book Company, England, 1989
6. M. Papini, G. Femlund, and J. K. Spelt, Effect of crack growth behaviour on
fracture load predictions of adhesively bonded joints, Composites Science and
Technology, 52 (1994), pp. 561-570.
7. L.J. Hart-Smith, Adhesive-bonded double-lap joints, NASA CR-112235, NASA
Chapter 2 Strength Detennination of Adhesive Bonded Joints 63
Langley Research Centre, Hampton Virginia, USA (1973).

8. L.J. Hart-Smith, Adhesive-bonded single-lap joints, NASA CR-112236, NASA
Langley Research Centre, Hampton Virginia, USA ( 1973).
9. L.J. Hart-Smith, Adhesive-bonded scarf and stepped-lap joints, NASA CR-112237,
NASA Langley Research Centre, Hampton Virginia, USA ( 1973).
I 0. D.A. Bigwood and A.D. Crocombe, Elastic analysis and engineering design
fonnulae for bonded joints, Int. J. of Adhesion and Adhesives, 9, (1989) pp. 229-
242.
11. J.W. Hutchinson and Z. Suo, Mixed mode cracking in layered materials, Advances
in Applied Mechanics, 29, ( 1992) pp 63-91.
12. L.J. Hart-Smith, Designing to minimize peel stresses in adhesive bonded joints,
Delamination and Debonding of Materials, ASTM STP 876, W.S. Johnson, Ed.,
American Society for Testing and Materials, Philadelphia, (1985) pp. 238-266.
13. J.A. Harris and R.D. Adams, Strength prediction of bonded single lap joints by non-
linear finite element methods, Int. J. of Adhesion and Adhesives, 4, (1984) pp. 65-
78.
14. R.D. Adams, The mechanics of bonded joints, International Conference on
Structural Adhesives in Engineering, IMechE Conference publications, 1986-6,
C 180/86, University of Bristol I London, ( 1986) pp. 17-24.
15. A.D. Crocombe, D.A. Bigwood and G. Richardson, Analyzing structural adhesive
joints for failure, Int. J. ofAdhesion and Adhesives, 10, (1990) pp. 167-178.
16. P. Czamocki and K. Piekarski, Non-linear numerical stress analysis of a symmetric
adhesive bonded lap shear joint, Int. J. of Adhesion and Adhesives, 3, ( 1986) pp.
157-160.
17. K. Matsui, Size effects on average ultimate shear stresses of adhesively bonded
rectangular or tubular lap joints under tension-shear, Int. J. of Adhesion and
Adhesives, I 0, (1990) pp. 81-89.
18. K. Matsui, Effects on nominal ultimate tensile stresses of adhesive bonded circular
or rectangular joints under bending or peeling load, Int. Jl of Adhesion and
Adhesives, 10, ( 1990) pp. 90-98.
19. H. Kyogoku, S. Uchida, and Y. Kataoke, CAD system for strength evaluation of
adhesively bonded joints, Advances in Adhesively Bonded Joints, American
Society of Mechanical Engineers Winter Annual Meeting, Chicago, Ill., (1988) pp.
117-128.
20. A.D. Crocombe, Global yielding as a failure criterion for bonded joints, Int. J. of
Adhesion and Adhesives, 9, (1989) pp. 145-153.
21. W.S. Johnson, Stress analysis of the cracked-lap-shear specimen: an ASTM Round-
Robin, J. of Testing and Evaluation, 15, (1987) pp. 303-324.
22. S.A. Hamaush and S.H. Ahmad, Fracture energy release rate of adhesive joints, Int.
J. of Adhesion and Adhesives, 9, (1989) pp. 171-178.
23. S. Mall and W.S. Johnson, Characterization of mode I and mixed mode failure of
adhesive bonds between composite adherends, Composite Materials: Testing and
Design (Seventh Conference), ASTM STP 893, J.M. Whitney, Ed., American
Society for Testing and Materials, Philadelphia, ( 1986) pp. 322-324.
24. S. Mall and N.K. Kochlar, Criterion for mixed mode fracture in composite bonded
joints, International Conference on Structural Adhesives in Engineering, !MechE
Conference publications, 1986-6, C 159/86, University of Bristol I London, ( 1986)
pp. 71-76.
25. S. Mall and N.K. Kochlar, Characterization of debond growth mechanism in
adhesively bonded composites under mode II static and fatigue loadings,
Engineering Fracture Mechanics, 31, (1988) pp. 747-758.
26. H. C. Cao and A.G. Evans, An experimental study of the fracture resistance of
bimaterial interfaces, Mechanics ofMaterials, 7, (1989) pp. 295-304.
27. Z. Suo, Failure of brittle adhesive joints, Applied Mechanics Reviews, 5, (1990) pp.
276-279.
28. M. D. Thouless, Fracture of a model interface under mixed-mode loading, Acta
Metallurgica et Materialia, 38, (1990) pp. 1135-1140.
29. Z. Suo, Debond mechanics of brittle materials, Scripta Metallurgica et Materialia,
25, (1991) pp. 1011-1016.
30. J.S. Wang and Z. Suo, Experimental determination of interfacial fracture toughness
curves using brazil-nut-sandwiches, Acta Metallurgica et Materialia, 38, (1991) pp.
1279-1290.
31. G. Femlund and J. K. Spelt, Failure load prediction of structural adhesive joints Part
1: Analytical method, Int. J. ofAdhesion and Adhesives, 11, (1991) pp. 213-220.
32. G. Femlund and J. K. Spelt, Failure load prediction of structural adhesive joints Part
2: Experimental results, Int. J. ofAdhesion and Adhesives, 11, ( 1991) pp. 221-227.
33. G.P. Anderson, S.H. Brinton, K.J. Ninow, and K.L. DeVries, A fracture mechanics
approach to predicting bond strength, Advances in Adhesively Bonded Joints,
American Society of Mechanical Engineers Winter Annual Meeting, Chicago, Ill.,
(1988) pp. 93-101.
34. S. M. Lee, An in-situ failure model for adhesive joints, J. of Adhesion, 18, (1985)
pp. 1-15.
35. H. L. Groth, A method to predict fracture in an adhesively bonded joint, Int. J. of
Adhesion and Adhesives, 5, (1985) pp. 19-22.
36. H. L. Groth, Stress singularities and fracture at interface comers in bonded joints,
Int. J. ofAdhesion and Adhesives, 8, (1988) pp. 107-113.
37. S.J. John, Predicting the strength of carbon fibre/epoxy composite joints.
Composites bonding, ASTM STP 1227, 1994 pp. 45-59.
38. E.J. Ripling, S. Mostovoy, and H.T. Corten, Fracture mechanics: A tool for
evaluating structural adhesives, J. ofAdhesion, 3, (1971) pp. 107-123.
39. 0. Volkersen, Die Niektraftverteiling in Zugbeansprunchten mit Konstanten
Laschenquerscritten, Luftfahrforsch, 15 (1938) 41-4 7.
40. M. Goland and E. Reissner, The stresses in cemented joints, ASME J. of Applied
Mechanics, 7 (1944) A17-27.
41. A.J. Russel and K.N. Street, Moisture and temperature effects on the mixed-mode
delamination fracture of unidirectional graphite/epoxy, Delamination and
Debonding of Materials, ASTM STP 876 , W.S. Johnson, Ed., American Society
for Testing and Materials, Philadelphia, (1985) pp. 349-370.
42. W.L. Bradley and R.N. Cohen, Matrix deformation and fracture in graphite-
reinforced epoxies, Delamination and Debonding of Materials, ASTM STP 876 ,
W.S. Johnson, Ed., American Society for Testing and Materials, Philadelphia,
(1985) pp. 389-410.
43. S. Hashemi, A.J. Kinloch, and G. Williams, Mixed mode fracture in fiber-polymer
composite laminates, Composite Materials: Fatigue and Fracture (Third Volume),
ASTM STP 1110, T.K. O'Brien, Ed., American Society for Testing and Materials,
Philadelphia, (1991) pp. 143-162.
44. J.R. Reeder and J.H. Crews Jr., The mixed-mode bending method for delamination
testing, AIAA Journal, 28, ( 1990) pp. 1270-1276.
45. G. Femlund and J. K. Spelt, Mixed-mode fracture characterization of adhesive
joints, Composites Science and Technology, 50, ( 1994) pp. 441-449.
46. M. Charalambides, A.J. Kinloch, Y. Wang and J.G. Williams, On the analysis of
mixed-mode failure, Int. J Fracture, 54, (1992) pp. 269-291.
47. G. Fem1und and J. K. Spelt, Mixed mode energy release rates for adhesively
bonded beam specimens, J Composites Technology & Research, 16, (1994) pp.
234-243.
48. M. Papini, G. Femlund, and J. K. Spelt, The effect of geometry on the fracture of
adhesive joints, Int. J of Adhesion and Adhesives, 14 (1994), pp. 5-13.
49. J.D. Clark and J.J. McGregor, Ultimate tensile stress over a zone: a new failure
criterion for adhesive joints, J. Adhesion, 42, 4 (1997) 227-245.
50. D.W. Oplinger, A layered beam theory for sinle lap joints, Army Materials
Technology Laboratory Report MTL TR91-23, 1991
51. L. Tong, Bond strength of adhesive-bonded single lap joints. Act Mechanica, 117,
(1996) 101-103
52. L. Tong, Strength of adhesively bonded single-lap and lap-shear joints, Int. J Solids
and Structures 35 (1998) 2601-2616
53. L. Tong, An assessment of failure criteria to predict the strength of adhesively
bonded composite double lap joints, J Reinforced Plastics and Composites, 16
( 1997) 698-713
54. L. Tong, A. Sheppard, Kelly D.W., et al, Effect of joint t1exibility in adhesively
bonded composite panel-t1ange, Composites Part B: Engineering, 29 (1998) 287-
298
55. A. Sheppard, D.W. Kelly and L. Tong, A damage zone model for the failure
analysis of adhesively bonded joints, Int. J. Adhesion and Adhesives, 18 (1998)
385-400.
56. G.J. Moore, MSC/NASTRAN- User's Guide, The MacNeal-Schwendler Co., 1994
57. M. Dessureault and J.K. Spelt, Observations of fatigue crack initiation and
propagation in an epoxy adhesive, Int. J Adhesion and Adhesives, 17 ( 1997), pp.
183-195.
58. A.J. Curley, J.K. Jethwa, A.J. Kinloch and A.C. Taylor, The fatigue and durability
behaviour of automotive adhesives. Part III: Predicting the service life, J. Adhesion,
66 ( 1998), pp. 39-59.
59. X.X. Xu, A.D. Crocombe and P.A. Smith, Fatigue crack growth rates in adhesive
joints tested at different frequencies, J Adhesion, 58 (1996), pp. 191-204.

60. J. Schijve, Fatigue predictions and scatter, Fatigue Fracture Engineering Mater.
Struct., 17 (1994), pp. 381-396.
61. R.A. Dickie, L.P. Haack, J.K. Jethwa, A.J. Kinloch and J.F. Watts, The fatigue and
durability behaviour of automotive adhesives. Part II: Failure mechanisms, J
Adhesion, 66 (1998), pp. 1-37.
62. J.E. Ritter, J.C. Learned, G.S. Jacome, T.P. Russell and T.J. Lardner, Fatigue of
silane bonded epoxy/glass interfaces, Materials Reliability in Microelectronics I)(,
pp. 291-296. Eds. C.A. Volkert, A.H. Verbruggen and D.D. Brown, Mater. Res.
Soc. Proc. 563 (1999), Pittsburgh, PA.
63. D. Plausinis and J.K. Spelt, Application of a new constant G load-jig to creep crack
growth in adhesive joints, Int. J Adhesion and Adhesives, 15 (1995), pp. 225-232.
64. D.C. Phillips, J.M. Scott and M. Jones, Crack propagation in an amine-cured
expoxide resin, J Mater. Sci., 13 (1978), pp. 311-322.
65. R.A. Gledhill, A.J. Kinloch and S.J. Shaw, A self-toughening mechanism in
expoxide resins, J Mater. Sci. Lett., 14 ( 1979), pp. 1769-1772.
66. D. Plausinis and J.K. Spelt, Designing for time-dependent crack growth in adhesive
joints, Int. J Adhesion and Adhesives, 15 (1995), pp. 143-154.
CHAPTER 3 STRESSES IN MECHANICAL FASTENED JOINTS
F.L. MATTHEWS* and P.P. CAMANHO**

*Centre for Composite Materials, Imperial College of Science, Technology and
Medicine, University ofLondon, United Kingdom.
**Faculdade de Engenharia, Universidade do Porto, Portugal.
3.1. EXPERIMENTAL BEHAVIOUR
An appreciation of the experimental behaviour of mechanically fastened joints in

composite laminates forms an essential precursor to any theoretical study. A large part
of the experimental research that has been done is concerned with the determination of
the influence of geometric and other parameters on joint strength. The usual procedure
is to test specimens until final failure, varying the factor under study and keeping the
other factors constant. It is also possible to find in the literature investigations
concerned with the experimental determination of the strain/stress distributions near a
loaded hole using photoelastic and Moire techniques.
Tests are normally conducted on single-hole specimens, which have the geometry
shown in Figure 3 .1.
·u. . . -
0["1 --d _GJ+--j.
.
!
--..j---.L
Figure 3.1 Specimen geometry
There are five common failure modes in mechanically fastened joints made of
67
L. Tong and C. Soutis (eds.},
68 F.L. Matthews and P.P. Camanho
composite laminates, namely tension, shear-out, bearing, cleavage and pull-through, as

shown in Figure 3.2.
j Lo~ {€] 1 ________.

§ L--1_<D-
~))
L,__
Tension Shear-out Bearing
Cleavage
-~---
Pull-through
Figure 3.2 Failure modes
Associated with these failure modes, the planes shown in Figure 3.3 are defined. This
Figure also shows the circumferential co-ordinate direction, e, and the fibre orientation
angle, a.
Bearing plane
1 Load direction
2
~+a
hear-out plane
Tension plane
Figure 3.3 Definition of planes and angles
Hart-Smith [I] considered that net-tension failure occurs when the bolt diameter is a
large fraction of the laminate width. This fraction depends on the type of material and
lay-up used. Bearing failure occurs predominantly when the bolt diameter is a small
fraction of the plate width. This mode of failure leads to an elongation of the hole.
Shear-out failures can be regarded as a special case of bearing failures . For highly
orthotropic laminates this mode of failure can occur at very large end distances.
Chapter 3 Stresses in Mechanical Fastened Joints 69
Cleavage failures are associated with both an inadequate end distance and too few
transverse plies. Pull-through failure occurs mainly with countersunk fasteners or when
the thickness-to-diameter ratio is sufficiently high to precipitate failure.
3.1.1. INFLUENCE OF GEOMETRY
Several authors [1]-[5] have highlighted the importance of width (w), end distance (e),
hole diameter (d) and laminate thickness (t) on the joint strength.
Basically, as the width of the specimen decreases, there is a point where the mode of
failure changes from one of bearing to one of tension. This change leads to a
considerable drop in the load capacity of the joint. There is a similar relationship
between the end distance and the shear-out mode of failure. As the end distance
decreases the bearing failure mode changes to one of shear-out, with a corresponding
decrease in the joint strength.
Matthews eta!. [5] considered that all material combinations follow the same trend,
the only difference being the e/d and w/d values for which full strength is achieved. It
was suggested a maximum value oft/d of one be used in order to avoid bolt bending.
3.1.2. INFLUENCE OF FIBRE ORIENTATION
Collings [3] demonstrated that the tensile strength of a single-hole joint is strongly
dependent on ply orientation. Testing CFRP laminates balanced about the mid-plane,
the author concluded that in (0°/±a0 ) laminates the bearing strength was dependent on
the ratio of oo plies to ±a0 plies.
In other work, Collings [6] showed that in CFRP the addition of ±45° plies to a 0° or
90° laminate increased the bearing strength until the ±45° plies account for
approximately 75% of the total laminate thickness. The author considered that the
presence of ±45° plies reduces the stress concentration factor. The addition of either oo
or ±45° to a 90° laminate produces a similar change in the variation of the bearing
strength. Kretsis and Matthews [2] demonstrated the same effect in GFRP.
Hart-Smith [1], [4] concluded that shear-out failures are prevalent for fibre patterns
which are both rich in oo and deficient in 90° plies. The author concluded that shear-out
and bearing strengths are maximised for quasi-isotropic laminates. As a general rule, he
stated that there should never be more than 3/8 nor less than 1/8 of the fibres in any one
of the basic laminate directions: ±45°, oo and 90°.
3.1.3. INFLUENCE OF LATERAL CONSTRAINT
Collings [3] showed that the bearing strength ofCFRP could be improved by increasing
the lateral (i.e. through-thickness) constraint pressure around the loaded hole. For
(0°/±45°) laminates improvements from 60% to 170% according to hole size were
achieved with constraint pressures of up to 22 MPa. At higher constraint pressures little
further improvement was achieved. It was considered that lack of through-thickness
compression at the contact zone could lead to premature in-plane compressive failure.
Failure occurred by initiation of shear cracks at the hole edge and subsequent
propagation to the edge of the clamped region.
The same effect was found by Kretsis and Matthews [2] in GFRP. If the laminate is
restrained laterally, the zone under the washers develops shear cracks but is not allowed
to expand under compression. Therefore the lateral expansion is spread into a wider area
that lies outside the washer boundary.
Matthews et al. [5] considered that the joint strength depends on the through-
thickness restraint offered by the fastener. A plain pin will give the lowest bearing
strength and a fully tightened bolt the highest bearing strength. The bearing strength
obtained using a riveted joint would fall between these extremes.
Hart-Smith [1], [4] related the effect of bolt tightening to the higher bearing strength
of the laminate in the middle of a double-shear lap, compared with the laminates in the
outer laps. This is a result of the better confinement of the damage in the inner lap. This
effect leads to the conclusion that joint strength depends not only on the amount of bolt
clamping, but also on the area over which is applied. The influence of clamp-up is not
as pronounced for the other failure modes as it is for bearing failure. It can only add to
load transfer by shear stresses due to friction on the washer-laminate interface.
Care must be taken with the relief of clamp-up over the life of structures.
Shivakumar and Crews [7], [8] using a viscoelastic finite element analysis of a double-
lap joint predicted that clamp-up forces relax even for the room temperature-dry
condition. The same effect was studied by Hom and Schmitt [9] who showed that an
increase of temperature decreased the strength of a laterally constrained joint.
3.1.4. iNFLUENCE OF STACKING SEQUENCE
Quinn and Matthews [1 0] measured the pin-bearing strength of GFRP using specimens
with mid-plane symmetry. Eight different stacking sequences of laminates with 0°, 90°
and ±45° layers were studied. The joint strength was found to be dependent on the
stacking sequence. This effect is due to the through-thickness direct and shear stresses
that occur at the hole boundary. The magnitude of these stresses is dependent upon the
stacking sequence. The results suggested that placing the 90° layer at the surface
increases the bearing strength because this will produce a compressive through-
thickness direct stress that can inhibit delamination. Clearly, the interlaminar shear
stress also influences the failure characteristics.
This effect was also shown in CFRP by Collings [3] who concluded that less
homogeneous stacking sequences exhibited lower bearing strengths, due to the higher
interlaminar shear stresses that occur in such laminates.
3.1.5. INFLUENCE OF FASTENER TYPE
A wide range of fasteners is available to join metallic parts, but the particular
characteristics of composite joints limit the choice, if efficient designs are sought. Types
of fasteners typically used are screws, rivets and bolts, the choice of fastener being a
function of the particular application. The main drawback of rivet fastening is the
variation of lateral clamping obtained during installation. Bolts are required in heavily
loaded structures and when disassembly is required and are the most efficient way of
mechanically fastening CFRP.
Hart-Smith [1] associated the lower efficiency of countersunk fasteners, when
comparing with bolts, with the fastener rotation in single-lap joints, and with the fact
that the countersunk fastener head is less effective than the shank in transferring the
bearing load. This is due to the lower through-the-thickness strength of the composite,
when compared with the in-plane strength.
3.1.6. MULTI-FASTENERJOINTS
In the majority of practical applications, the load is transferred between two components
in multi-row fastener patterns. The preceding discussion has dealt with either single-bolt
joints or with individual bolts isolated out of a single row, by representing the latter as a
single bolt in a strip of a width equal to the bolt pitch. Collings [3] stated that this
approach is valid for sufficiently large pitches where there is no interference of stress
fields around the holes. In this case it was possible to predict the total load carried by
multi-hole joints using single-hole data. However, Matthews et al. [5] observed that
when the strip width was considered as the pitch of a row of bolts, interaction between
neighbouring holes could occur, resulting in a strength below that predicted from single-
hole tests. Hart-Smith [1], [4] concluded that a multi-hole joint did not offer a
substantial improvement over a single-hole joint.
3.2. SEMI-ANALYTICAL AND NUMERICAL METHODS OF STRESS ANALYSIS
Before attempting to predict the strength of bolted joints in composite laminates, an

accurate stress analysis of a region close to the hole boundary is required. Whilst in
metallic materials the localised yielding leads to a relaxation of the stress concentration
near the hole, allowing simplified stress analyses and design procedures, the brittle
nature of composite materials partially inhibits this phenomenon. Considering also the
possibility of high stress concentrations occurring for certain lay-ups, it is clear that
reliable methods of stress analysis provide the basis for composite joint design.
3.2.1. SEMI-ANALYTICAL METHODS OF STRESS ANALYSIS
3.2.1.1. Two-dimensional models

The determination of stress distributions in composite plates containing loaded or open
holes is usually based on the complex function method developed by Muskhelishvili
[11] and adapted to orthotropic materials by Lekhnitksii [12]. In order to analyse
layered composites this method must be combined with classical lamination theory.
Solution is achieved by determining a stress function, expressed in terms of analytical
functions of complex variables, that satisfies equilibrium and compatibility. The
problem is then reduced to the determination of two complex potentials that satisfY the
boundary conditions.
The above approach was used by de Jong [13] to determine the stress distributions
around a frictionless pin-loaded hole. The pin was assumed rigid, the uniformly
distributed load in the plate was applied at an idealised infinite distance, and a
cosinusoidal radial stress distribution represented the pin-hole interaction. The solution
was obtained as the combination of two load cases, a pin-loaded hole, with loads
applied at both edges of the plate, and an unloaded hole case. Results were obtained for
unidirectional, quasi-isotropic and (0 4°/±45°) CFRP laminates, using w/d ratios of 2.5, 5
and =. It was shown that the direct stress distribution at the hole boundary was highly
dependent on the lay-up and width used. The maximum value of the direct stress
occurred in the unidirectional laminate, for w/d=2.5, in the vicinity of the net-tension
plane. Qualitatively, this result is in agreement with experimental results [1]-[4]. The
decrease of the direct stress in the net-tension plane was more pronounced for the highly
orthotropic laminate, whilst in the bearing plane the opposite effect occurred. It should
be noticed that the minimum value of w/d for which the solution is valid is 2.5.
Therefore, for smaller values of w/d the results have to be corrected to account for
effects of finite widths.
In order to account for the effect of friction at the pin-hole boundary of a plate with
infinite dimensions, Zhang and Ueng [14] developed a method in which the stresses at
the hole boundary were obtained using Lekhnitksii's [12] method, using complex
potentials which satisfY prescribed boundary conditions. A solution for a homogeneous
orthotropic plate loaded by a rigid pin was obtained, giving the stress distributions at the
hole boundary for several friction coefficients for (0 4°/±45°), (±45°)s. (0 3°/±45°), and
(90 4°/±45°)5 laminates. It was concluded that these stress distributions were dependent
on the lay-up and on the presence of friction. As friction increased, the shear stress, O"re,
along the hole edge also increased whereas the radial direct stress, O"rn decreased in an
area near 8=0° (where 8 is defined in Figure 3.3). Zhang and Ueng [14] concluded that
the size of this area depends on the ratio of the moduli of elasticity along the two
principal axes. Near 8=0°, the hoop stress, cr88 , may vary gradually from tension to
compression or increase in compressive value as friction is increased. The maximum
value of cr88 also increased with increasing friction.
A detailed investigation into the effects of friction, pin elasticity and clearance on
the stresses near a pin-loaded hole was made by Hyer and Klang [15], [16]. Like in the
previous investigations, the elasticity problem was formulated in terms of complex
variable theory. Boundary and interface conditions were prescribed for the no-contact
region, no-slip region and slip region. The unknown coefficients of the complex Fourier
series that represented boundary tractions were determined at a finite number of points
on the circular boundary using a collocation and iteration procedure. Laminates with the
(0°), (90°), (0°/±45°/90°), and (0 2°/±45°)5 configurations were analysed. It was
concluded that pin flexibility was not as important a factor in the stress distribution in a
pin-loaded hole as clearance and friction, especially for perfect-fit joints. Like in de
Jong's [13] results, the degree of orthotropy strongly influenced the peak stresses and
the stress distributions. A comparison of results with the ones obtained using the often
assumed cosinusoidal stress distribution in the hole boundary was made and it was
concluded that the cosinusoidal stress distribution was not generally accurate. The
effects of friction were in agreement with the ones previously described [14]. The most
significant effects of increasing clearance was moving the peak circumferential stress
toward 8=0° and increasing the peak radial stress, this result being related to the
reduction of the contact surface for increasing values of clearance.
3.2.1.2 Three-dimensional models

Experimental results have shown that factors such as clamping pressure and stacking
sequence play a substantial role on the bearing strength of composite bolted joints. In
order to account for these effects, three-dimensional finite element models are required.
An innovative three-dimensional continuum model based on energetic
considerations (minimum potential energy principle) used in conjunction with spline
approximations has been proposed by larve [17]-[20]. The analysis of the composite
plate was based on work previously done on plates with open holes [ 17]. A curvilinear
transformation mapping the x-y plane of the plate into a region O:S:::p:S:::l, O:S:::<J>:S:::2n was
defined, and cubic spline approximations in curvilinear co-ordinates were obtained for
laminae displacements and interlaminar surface tractions at the ply top and bottom
surfaces. Subdivisions were introduced through the thickness of each ply and in the p
and <1> directions, and the vector of the three-dimensional spline approximation functions
was defined. Spline approximations for bolt displacement, and for the Lagrange
multipliers that impose the contact conditions were derived. Using such a procedure the
systems of equations to determine the unknown spline coefficients were obtained.
The validation of the model accuracy when predicting radial and hoop stress
distributions occurring in a composite bolted joint was performed by comparison with a
finite element solution and a good agreement was obtained. However, the verification of
interlaminar stresses is a more complicated task due to the lack of established baseline
results. Therefore, the predicted interlaminar strains were compared with Mollenhauer's
[21] experimental results obtained using Moire interferometry in laminates containing
an open hole and loaded in uniaxial tension. An excellent agreement between larve's
model (SVEL T) and experimental results was found for both in-plane strains [20] in the
vicinity of the hole and, as shown in Figure 3.4, out-of-plane strains [21] obtained in
different locations around the hole boundary.
Although these results were obtained for open-hole laminates, they are quite
relevant, as they provide a baseline for the validation of other models attempting to
capture three-dimensional stress states.
The model was also used in the analysis of a (0°/±45°/90°}, CFRP laminate loaded
by a rigid pin. Frictionless contact was assumed and the distributions of normalised
radial stress and polar displacement were obtained for zero clearance and 4% clearance.
It was concluded that a clearance fit leads to an increase of normalised radial stresses
and a reduction of the contact angle. The model proved also capable of capturing a
contact stress singularity, should one be present in the problem, as the predicted contact
stresses' amplitude for such a situation increased for more refined subdivisions.
3.2.1.3. Multijastener joints

The stress analysis of multi-fastener joints plays an important role, as in several
applications there are rows of fasteners and both bearing stresses and by-pass stresses
must be considered.
35000
30000
21<)00
20000
11<)00
10000
• 5000
S!
"
~-.5000
·10000
·15000
·20000
·25000
1.00 2.00 l .OO <-00
Distance from Upper Lamlnato Face (mm)
Note: The vonlcal dotted

llnu represent tho
ide.tllz.od ply bounderiea.
r
Figure 3.4 Shear strain at 75° for a CFRP laminate containing an open hole and loaded in tension
[21]
A method for the calculation of stresses in anisotropic plates with a row of equally
spaced pin-loaded holes was presented by de Jong [22] . Like in a previous work [13],
the solution of the problem was based on two load systems: a plate with a row of open,
unloaded holes and a plate with a row of pin-loaded holes. However, in contrast with de
Jong's previous work [13], the pin-hole interfacial stresses are now calculated rather
than assumed and the effect of friction is also considered. As expected, when decreasing
the pitch-to-diameter ratio the tangential stress concentration factor in the net-section
increased, but was almost constant at 8=0°. The radial stresses decreased at the latter
location. The effect of friction was in general agreement with the results obtained by
Hyer and Klang [16]. It should be noticed that the solutions provided by this model
yield stress distributions not applicable to joints with few pins or to the end pin of a
joint. Furthermore, the series expansions used in the solution of the load case of a plate
with a row of unloaded holes have limited areas of convergence around the central hole.
This means that accurate results can only be obtained within a circle of radius
approximately half the pitch. Therefore, the model is not applicable to joints with small
end distances as the traction free condition at the end of the joint affects the stresses
around the hole. This is not a major limitation, as minimum end distances e=3d are
recommended to avoid shear-out failures [1].
3.2.2. NUMERICAL METHODS OF STRESS ANALYSIS
The majority of the numerical models simulating composite bolted joints are based on
the finite element method. However, some models based on boundary elements have
also been developed.
The finite element method is one of the most powerful numerical methods used in
structural engineering problems. It can be interpreted as a process in which the
continuum is idealised as an assemblage of discrete finite elements connected at nodal
points on the element boundaries. In a displacement-based finite element formulation,
the governing equations are obtained in terms of the displacements [23], [24].
It is important to note that in displacement finite element models only the
displacements are continuous along the element boundaries. The strains and stresses
along the element boundaries take different values for adjacent elements, but are
continuous within each element. This means that compatibility is satisfied completely,
but equilibrium is only satisfied in the mean.
The nature of composite laminates adds extra complications when setting-up a finite
element model of a composite structure. Problems related to the definition of variable
fibre orientations within a ply, variation of fibre volume fraction, ply drop-offs,
through-thickness stresses occurring at free-edges, variation of material orientation in
geometrically non-linear problems, and failure modelling, require new approaches, as
well as the development of new types of elements [25].
When simulating composite bolted joints the contact between the bolt and the
composite must also be taken into account in the finite element models. The assumption
of a value for the contact angle between the bolt and the composite has the advantage of
simplicity but is not generally accurate since the contact angle varies non-linearly with
load for clearance fits. In order to determine contact surfaces and stresses the usual
procedure is to use iterative or inverse methods.
Initial iterative methods were based on beam or rigid link elements connecting the
nodes on the boundary of the laminate to a node located at the centre of the hole. The
76 F.L. Matthews and P .P. Camanho
nodes connecting the beam or rigid link elements which were in tension were assumed
to be in the non-contact zone (radial separation between pin and hole boundary) and
removed. Recent models have used constrained variational techniques [26], [27] to
address the contact problem. This is a general technique to address contact problems
and has the advantage of being also able to deal with frictional contacts and non-
uniform contact through the laminate thickness when used in conjunction with
three-dimensional models. As the variation of the contact surface represents a change on
the system boundary conditions, non-linear solution procedures are required.
Using this technique, the contact surface must be determined during the analysis.
For each load increment the boundary nodes are checked for contact with the bolt and,
when friction is present, the position of the contact nodes either in the slip or no-slip
region are determined. Contact problems including friction generally lead to a non-
symmetric coefficient matrix [24].
Although the contact problem is non-linear, inverse methods require only linear
finite element analysis, resulting in simpler models than the ones created using iterative
methods. Displacement constraint equations are specified along the bolt-hole interface
and stress boundary conditions are specified at the end of the contact arc and in the
region beyond the contact arc. A value for the contact angle is assumed and using the
previously defined conditions the corresponding load is calculated. The process is
repeated for a series of prescribed contact angles. It should be noted that this method is
only applicable when geometric and loading symmetries exist.
3.2.2.1 Two-dimensional models

Several authors [28]-[35] have considered pin-loaded composite plates in a state of
plane stress. Two-dimensional finite element models were created and classical
lamination theory [36] was applied to provide the required element elastic properties
and to determine layer stresses. Clearly, the effects of clamping pressure and stacking
sequence cannot be studied with this approach.
The effect of clearance on the stress distribution near the loaded hole was
investigated by Naik and Crews [28]. Using an inverse formulation, assuming a
frictionless contact and a rigid pin loading a quasi-isotropic laminate, it was shown that
the contact angle was a function of clearance. For constant bearing stress, increasing the
clearance decreases the contact angle. As shown in Figure 3.5 the clearance, cd, also
influences the value and position of the maximum radial and hoop stresses.
The maximum values of these stresses increases with increasing clearance. When
clearance is present, the relations of the radial and hoop stresses with bearing stress are
initially (small loads) non-linear, becoming linear with increasing bearing stresses. For a
perfect-fit joint, these relations are linear and the contact angle is constant.
Ramamurthy [29], [30] used an inverse technique to study the behaviour of pins
fitted with interference. This kind of joint is sometimes used to increase the fatigue life
of metallic structures. Assuming a frictionless contact and a rigid pin, it was concluded
that the maximum bearing stress varies nonlinearly with the load. The maximum
normalised radial and hoop stresses varied with the direction of the load. For high loads,
radial stresses are higher and hoop stresses are lower when compressive loads are
applied at the laminate border. No comparisons between stress distributions for the
cases of clearance, perfect-fit and interference were made.
A comprehensive study of geometrical effects on the stress distribution of a pin-loaded
plate was done by Crews et al. [31]. A frictionless contact and perfect-fit pin were
assumed and six laminates were considered: (0°/±45°/90°)" (0°), (90°), (0°/90°)8 , (±45°)8
and (0°/±45°)8 • The stress distributions around the hole boundary were represented,
showing the strong influence of anisotropy on both magnitude and location of the peak
hoop stress on the hole boundary. It was considered that the hoop stress, cr99 , the radial
stress, crm and the shear stress, O"xy. were associated with the tension, bearing and shear-
out failures respectively. The effect of w/d was predominant on the hoop stress. As a
result, the main influence of w/d will be in the tension failure mode. A tensile stress
concentration factor, K1b, was defined as the ratio of the maximum hoop stress, (cree)max•
that occurred near the end of the contact arc, to the bearing stress, Sb. It was shown that
for a constant value of end distance K1b increases with decreasing w/d. Decreasing e/d
increased both the hoop and shear stresses. The effect on the shear stress may be
important in laminates with low shear strengths, like the (0°) laminate. Both the shear
stress concentration factor, defined as the ratio of the maximum shear stress, (crxy)max• to
the bearing stress, Sb, and the tensile stress concentration factor, K1b, increase with
decreasing e/d. The edge distance only weakly influenced the peak values of bearing
stress. These conclusions are qualitatively in good agreement with experimental results
described in 3.1 [1]-[5].
1.6 r
0.8
0.0
-0.8
-1.6
60 120 180
8 (degrees)
Figure 3.5 Effect of clearance on hole boundary stresses [28]

78 F .L. Matthews and P .P. Camanho
The effects of friction, geometry, clearance and material properties on load distribution
were investigated by Wilkinson et a/. [32]. Using an iterative procedure previously
formulated [33] and assuming a rigid pin, the finite element model was substantiated by
comparing the results with Moire and strain gauge experimental results. It was
concluded that a change in contact friction from Jl=0.7 to J.1=0.4 had little effect on the
radial stress at the hole boundary. This result is in disagreement with the results
obtained by Zhang and Ueng [14] and Hyer and Klang [15]. As the stiffness of the
material increased (from glass fibres to boron fibres), the peak radial stress increased.
This effect explains partially the use of softening materials near pin loaded holes, as
described in [1], [4]. Like in Naik and Crews [28] work, it was shown that increasing
clearance increased contact radial stresses. For e/d<::4 the stress distribution was little
affected by variations in end distance. This result is qualitatively in agreement with
experimental results [3].
Using an iterative method, Eriksson [34] investigated the effects of clearance,
friction and bolt stiffness on the stress distribution near the hole boundary and arrived at
the same conclusions as Hyer and Klang [15]. In Eriksson's work, the stress
distributions were also obtained along an arc at a distance ao= 1.26 mm from the hole
boundary. This distance is associated with a strength criterion that will be described
later. As at the hole boundary, the stress distribution along this arc was strongly
dependent on the laminate properties. The same effect occurred for the shear stress on
the shear-out plane and for the bearing stress in the bearing plane. The effect of friction,
along the arc considered, followed the same trends as at the hole boundary for the hoop
stress, but had a different effect in the shear stress. At ao. the presence of friction
decreased the shear stresses.
All the models considering frictional contacts described above assumed a constant
value for the coefficient of friction between the composite and the pin in the whole of
the contact region. However, a variation of the coefficient of friction for different fibre
orientations at the pin-laminate interface would be expected. The dependence of
coefficients of friction on fibre orientation for CFRP unidirectional laminates in contact
with steel pins was experimentally determined by Xiao et al. [35]: the coefficient of
friction increased when the angle between the fibre orientation and the sliding direction
increased. Using these experimental results and Coulomb's friction law, a procedure to
calculate an equivalent coefficient of friction at the laminate edge was developed [35].
This procedure was used to define a relation between the coefficient of friction and the
position on the hole boundary, Jl=J.1(8), that was used in a two-dimensional finite
element model of a pin-loaded composite laminate. For (0°/±45°/90°)5 and (0°/902°/0°)s
laminates, the results were compared with a frictionless contact and with a model
considering a constant coefficient of friction, Jl=0.27. The differences between
frictionless and frictional contacts were in agreement with previous results [14]. The
difference between considering the coefficient of friction as a function of e or constant
occurred in a region close to 8=45°, due to the sliding conditions and to the relatively
large value of the radial stress. For the problem investigated, the variation of the
coefficient of friction with respect to laminate orientation was not critical.
3.2.2.2. Three-dimensional models

Three-dimensional models can provide a detailed simulation of the effects occurring at
the bolt-laminate neighbourhood. These models are capable of capturing through-
thickness stresses, and of assessing the effects of clamping pressure and stacking
sequence on the stress distributions occurring in a mechanically fastened joint.
However, three-dimensional finite element models have the drawback of being
computationally expensive and, when modelling a laminate using one or more three-
dimensional elements per layer (stacked brick models), problems related to the aspect
ratio of the elements can arise.
Matthews et al. [3 7] investigated the effect of the way the load is transmitted to the
plate on the stress distribution around a loaded hole in a (0°/±45°/0°)5 laminate. Using a
three-dimensional finite element model and assuming a perfect-fit rigid pin and
frictionless contact, three configurations were considered: pin-loaded hole, finger-tight
bolt and fully tightened bolt. Significant differences in stress distributions were
obtained. In the pin-loaded case, the maximum through-thickness tensile stress, cr"' was
found to be about 7% of the bearing stress. It was concluded that this effect can be
related to the experimental results, since failure occurs in the region of maximum crzz
due to splitting between layers. The distribution of <Jxx around the hole showed good
agreement with that obtained by de Jong [13]. For the finger-tight bolt, a general
reduction of crzz within the washer area was noticed. In a fully tightened bolted joint,
significant through-thickness compressive stress was present. The interlaminar shear
stress, cr"' was found to increase at the washer edge. This effect was related to the
experimental results where failure occurs by delamination at the washer edge. These
results are also consistent with experimentally observed relative strengths.
The effects of friction, clamping pressure, clearance and stacking sequence were
investigated by Chen et al. [38]. An incremental variational principle was formulated
and a general finite element technique satisfying contact conditions was established for
the three-dimensional contact stress analysis. Firstly, a (45°/0°/-45°/90°)5 CFRP laminate
subjected to bolt load and clamping pressure was analysed to evaluate the accuracy of
the model. Comparing with experimental results obtained by Smith et al. [39]
reasonable results were achieved. The previously described effects were studied in a
thick (0°/90°), or (90°/0°)5 GFRP subjected to a bolt load. The effects of friction,
clearance and bolt elasticity were in agreement with the results obtained by Hyer and
Klang [15], [16]. A much lower tensile interlaminar normal stress on the bearing plane,
crw was observed in the (90°/0°), laminate and this can prevent delamination. This result
helps to explain the higher bearing strength of laminates with 90° layers at the laminate
surface, as experimentally demonstrated by Quinn and Matthews [10]. As in the work of
Matthews et al. [37], the main etiect of the clamp-up applied by washers was to induce
lower interlaminar tensile direct stress or higher interlaminar compressive direct stress
near the hole. The effect of clamp-up was not investigated at the outer edge of the
washer.
Marshall et al. [40] investigated the effects of friction, clamping and laminate
stiffness on the stress distribution of (0°/90°)5 and (90°/0°)5 GFRP and CFRP laminates.
Results were obtained for pin-loaded holes and for bolted joints with several degrees of
clamping. The effect of friction was investigated in the pin-loaded hole. The finite
element model was validated comparing numerically determined strains with strains
obtained using strain gauges, on the bearing plane. The results were in reasonable
agreement. Only the frictionless and infinite friction cases were considered, showing the
reduction of axial interface stresses in the bearing plane when friction is present. This
result is consistent with previous analyses [15], [34]. It was concluded that increased
friction redistributes the load and correspondingly the position of the main load-carrying
fibres away from the bearing plane towards the net-tension plane. The stress
distributions of GFRP showed similar profiles as compared with CFRP, but with
marked drops in all stresses. These reductions are related with the lower stiffuess of
GFRP fibres. The effect of stacking sequence was in agreement with the results
obtained by Matthews et al. [37]. A general improvement in all the stress profiles was
obtained using bolted joints. The effect of clamping pressure was in good agreement
with the results obtained by Matthews et al. [37].
The effect of intralaminar non-linear shear behaviour on the modelling accuracy of
pin-loaded ((0°/90°)n)s and ((+45°/-45°)n)s GFRP laminates was investigated by Sarabian
[41] using a three-dimensional finite element model. The pin-plate interaction was
modelled using gap contact elements and a frictionless contact was assumed. Although
being a factor with small influence on the stress distribution near a pin-loaded hole [ 15],
[16], pin elasticity was considered. The non-linear shear behaviour was determim:d
from the AS TM D3 518-7 6 specification. This test allowed, using a least squares cun e
fit procedure, the derivation of a relation between the shear stress and the shear strain
that was implemented in the finite element model constitutive equation. In order to
assess the effect of this non-linear behaviour, a three-dimensional finite element linear
model was also created. It was found that both radial and hoop stresses at the hole
boundary were significantly affected by the inclusion of material non-linearity. A
reduction of the maximum shear stress was observed in the shear-out section of the
((0°/90°h,0°)s laminate and in the net section of the ((+45°/-45°)3 ) 5 laminate. Through-
thickness direct stresses were found to be negligible in comparison with in-plane
stresses, and increased when material non-linearity was assumed. Using the Moire
technique experimentally determined strains were found in the shear-out plane of the
((0°/90°)3,0°)s laminate and in the net and bearing planes of the ((+45°/-45°)3 ) 5 laminate.
The results obtained validated the importance of including non-linear material
behaviour in these laminates. However, one must be cautious when using the non-linear
shear stress-strain data from the ASTM D3518-76 type specimen. Much less non-
linearity is observed in other specimens, such as the ASTM D4255 type specimen or
1oo off-axis specimen.
3.2.2.3. Multi-fastener joints

The numerical investigations of stress distributions in multi-fastener joints are based on
two-dimensional finite element models. Both lines (parallel to the load) or rows
(perpendicular to the load) of fasteners have been considered.
Wang and Han [42] investigated the load proportioning in multi-row joint fasteners
using a two-dimensional finite element model. A single-lap joint of two symmetric:
laminates with several fastener configurations was considered. The fasteners were
modelled considering only the action of pure shear. Assuming linear elastic behaviour,
the joint behaviour was defined by the sum of two load cases and a factor~ was defined
as the ratio of load carried by a fastener to the average load. This factor represents the
unevenness of load distribution between fasteners. Not surprisingly, the results showed
that in the case of a line of fasteners, the first and last fastener are the ones subjected to
the highest loads. It was concluded that as the axial stiffness of the plates decreases or
as the total number of fasteners in the axial or transverse direction increases the load
distribution tends to be more uneven.
A comprehensive work about load proportioning was made by Griffin et al. [43].
Considering double lap joints, two independent two-dimensional finite element models,
one representing the composite inner lap and the other representing the steel outer laps,
were created. The laps interact by means of circular rigid surfaces and gap elements.
Two-row joints with three-hole, five-hole, seven-hole and nine-hole geometries with
shifted centres were analysed. It was found that load proportions were not substantially
affected by the level of applied load. As the number of fasteners increases, the load
proportioning becomes more equal. This result contrasts with Wang and Han [42]
results. The outboard row (row closest to the specimen end) carried a substantially
larger portion of the load in the narrow specimen, with a shift to the inboard row
carrying a slightly larger portion in the wider specimen. The different distributions of
radial contact stresses on outboard and inboard row holes were shown. Specially
instrumented pins and strain gauges were used to verifY the numerical results, and good
agreement was achieved.
In multi-row joints, fastener holes are subject to bearing loads and by-pass loads that
are reacted elsewhere in the joint. The effect of this type of loading on the stress
distribution of a frictionless hole loaded by a rigid pin was investigated by Naik and
Crews [44]. Using the previously developed inverse technique [28] and considering the
effect of bolt-hole clearance and a sufficiently large pitch to avoid interactions between
holes, tensile and compressive loads were analysed for a single hole. A coefficient ~
was defined as the ratio of bearing stress, Sb, to the by-pass stress, Snp· These stresses
are defined by:
(3.1)
pbp (3.2)
s np =--'-----
(w-d)t
where Pb and Pbp are respectively the load reacted at the hole and the by-pass load. This
coefficient was found to have a significant effect on the contact angle and peak stresses
around the hole. For a given bearing stress it was found that increasing the tensile by-
pass load (reducing ~) resulted in an increased contact angle <p, whilst increasing the
compressive by-pass loading had the opposite effect. When a tensile by-pass load was
applied, the peak value of radial stress was not very sensitive to Snp• whereas its
distribution showed the variation of the contact angle. With increasing Snp the hoop
stress rises significantly, its maximum being located a few degrees beyond the contact
region. In the case of compressive loading, the radial stress rises nearly proportionally
with Snp· As the absolute value of Snp increases the peak value of the hoop stress
becomes compressive. It was concluded that compressive loading can lead to dual
contact between the hole and the bolt. This contact occurred for values of ~ between 0
and -3. An important consequence of this kind of contact is that it allows load transfer
across the bolt and therefore reduces the stress concentration around the hole. It should
be noticed that these results were obtained for constant values of Sb. This means that
when Snp increases the total load applied to the joint also increases.
Using a novel approach, Kim and Kim [45] investigated two bolts in a line and in a
row of a double lap joint. Using extended interior penalty methods, the variational
formulation obtained was discretized using the finite element method. A frictionless
contact was assumed between a rigid pin and the laminate with symmetric stacking
sequence. Three laminates were considered: (0°), (90°) and (0°/±45°/90°)5 • The situations
of clearance and perfect fit were analysed for the (0°/±45°/90°), laminate, whilst
clearance was present in the others laminates. For the two holes in a line it was
concluded that the width and length of the plate and the pitch distance have a significant
influence on the distribution of contact pressure. For a clearance of 0.1% the hole
nearest to the applied force (under both bearing and by-pass stresses) experiences more
load and a wider contact area than the other hole. In a perfect fit situation the hole at the
bigger distance from the load has the widest contact area. The results obtained for two
holes in a row showed similar characteristics compared with those of single holes; the
contact area moves towards the symmetric axis of the plate and when the hole moves to
the left the location of maximum contact pressure moves to the right, and vice-versa.
3.3. STRENGTH PREDICTION METHODS
Several approaches have been used to predict the strength of mechanically fastened
composite laminates. Most of the methods developed are based on two-dimensional
models and only recently have methods considering three-dimensional models been
developed. The determination of joint strength depends on the definition of failure. This
d. i"iP.itwn can vary from the maximum load sustained by the joint to a criterion based on
the deformation of the hole. According to Godwin and Matthews [46] the majority of
authors seem to agree on the advisability of a definition based on the latter criterion, but
the values that are used vary significantly. Other definitions like the first peak or the
first non-linearity in the load-displacement plot, and the load at which cracking 1s
initiated, were suggested by Johnson and Matthews [47].
3.3.1. FAILURE THEORIES
Methods based on both hole boundary stress distribution and failure theories [48]-[51]
were among the first methods used. In these methods, the peak local stresses are used in
a failure theory to predict laminate strength.
Waszczak and Cruse [48] adopted the maximum stress, maximum strain and
distortional energy criteria. A frictionless contact and a cosinusoidal boundary radial
stress distribution were assumed. When the failure criterion was satisfied in one lamina,
that lamina was assumed to have failed and was removed from the analysis. The stress
distribution and failure criterion were then recalculated and the procedure repeated until
total laminate failure occurred. Using this procedure conservative results were obtained
and the failure modes were not predicted.
The typical conservative predictions obtained using failure theories are related to the
localised damage that occurs near the hole prior to laminate failure. This effect was
experimentally reported by Hart-Smith [1]. The latter author showed that the effective
stress concentrations at laminate failure are smaller than the elastic stress concentration,
and considered that the prime factor that leads to this stress concentration relief is
delamination. This effect is also related to the well-known hole-size effect that occurs in
laminates with unloaded holes and is characterised by a strength decrease for larger hole
sizes in laminates without finite width effects. Explanations for this effect were given by
Crews [52], who considered that the localised damage that occurs prior to failure
reduces the laminate stiffuess in that zone, and consequently reduces the stress
concentration, and is most effective for small holes. At such holes, the regions of high
stress concentration are small and readily influenced by the damage at the hole
boundary. A second explanation given by the author was that for large holes a higher
volume of material is highly stressed and the probability of containing a large flaw that
can lead to failure is higher than for small holes.
A similar explanation was given by Whitney and Nuismer [53] who related the hole
size effect to the stress distributions around holes of different sizes in infinite plates.
Although the computed stress concentration factors at the edge of the hole were the
same for all hole sizes, the stress distribution was much steeper in the case of the small
hole, making the stress concentration much more localised. For the large hole, a larger
volume of material is subjected to high stresses, leading to a higher probability of
containing a flaw. Strength prediction methods based on boundary stresses are simple to
use but do not account for the localized material response near the hole. As a result,
these methods underestimate the strength of the laminates.
3.3.2. TWO-PARAMETER METHODS
In order to account for the localised damage prior to failure previously described, two-
parameter methods have been applied. These methods are based on the Whitney-
Nuismer [53] failure criterion for unloaded holes and the parameters considered are the
unnotched tensile strength and a characteristic dimension. Two approaches were
proposed, the point stress and the average stress methods. In the first method it was
assumed that failure occurs when the direct stress in the direction of the load at a
distance dot away from the hole, measured in the tension plane, is equal to or greater
than the strength of the unnotched material. The second method considered that failure
occurs when the average stress over some distance aot equals the unnotched material
strength. These distances were considered to be a material property. These criteria were
formulated for the case of uniaxial tension where combined stresses play an
inconsequential role in the failure process. In cases where this cannot be assumed, these
criteria must be recast.
In later investigations, Nuismer and Labor [54], [55] applied the average stress
method to predict the strength of unloaded holes in tension [54] and of both unloaded
and loaded holes in compression [55]. The characteristic distance was now considered
to be laminate dependent. The predicted strengths for the tension case showed good
agreement with experimental results. The value of the characteristic distance for
compression was found to be larger than for tension. Good agreement between
experimental and predicted compressive strengths was found only for laminates with
unloaded holes.
Wilson and Pipes [56] used the point stress criterion to study the influence of both
e/d and w/d ratios in the shear-out strength ofloaded holes. The shear stress distribution
in the shear-out plane was obtained by a finite element model and then fitted to a curve
using a polynomial expression. A 180° contact angle and a frictionless contact were
assumed. The ratios w/d and e/d were then varied and the plots relating the maximum
shear stress with these ratios were obtained. This procedure allowed the definition of
parameters that are functions ofw/d and e/d and the inclusion of these parameters in the
representation of the stress profile. The characteristic distance dos was assumed to be a
function of the hole size. For a (45°/0°/-45°/02°/-45°/0°/45°/02°/90°)s laminate, good
agreement was found between predicted shear-out strength and experimental results
obtained for three fastener sizes. It should be noticed that this result could be partially
due to the fact that two of the fastener sizes considered were used to obtain the two
parameters required to define the characteristic distance. It was predicted that the shear
stress concentration decreases exponentially with increasing e/d and increases linearly
with increasing w/d.
A similar procedure was adopted by York eta/. [57] to investigate tension failures in
bolted joints. As the effect of e/d on the net tension stress profile was found to be
negligible, only the effect ofw/d in the tension strength was investigated.
These procedures have the drawback of being unable to predict the failure modes. In
order to predict the laminate strength, the modes have to be known a priori and, for
several lay-ups and laminate geometries, this is not straightforward.
3.3.3. COMBINED METHODS
Other methods consist of combining the concepts of the Whitney-Nuismer characteristic

distance [53] with a failure theory. Chang et a/. [58] used a two-dimensional finite
element model, assuming a frictionless contact, a rigid pin and a cosine normal load
distribution on the pin-hole boundary. The Yamada-Sun [59] failure criterion was then
applied together with a proposed characteristic curve. The Yamada-Sun [59] failure
criterion is based on the assumptions that just prior to laminate failure every ply has
failed due to cracks along the fibres, and that the shear strength of a symmetric cross-ply
laminate, with the same number of plies as the laminate under consideration, represents
the substantially higher shear strength of a lamina when it is in a laminate. This criterion
has the form:

(3.3)
Where cr 11 and cr 12 are respectively the longitudinal and shear stress in a ply, X is the ply
longitudinal strength and Sc the ply shear strength measured from a cross-ply laminate.
When er is equal to, or higher than, 1 in any ply, laminate failure is assumed to occur.
The characteristic curve was assumed to depend on the material and hole diameter and
is defined by rc (8) = d I 2 +dot + (doc +dot) cos e, where dot and doc are the
characteristic dimensions for tension and compression, respectively. These parameters
were obtained experimentally by measuring the tensile and compressive strength of
notched laminates. The location on the curve where er first reached 1, Sr, allows the
definition ofthe failure mode. For -15o:::;;er:::;;+l5° failure occurs in the bearing mode, for
+3oo:::;;ef:::;; +60° in the shear-out mode and for +75o:::;;er:::;;+90° in the tension mode. The
results were compared with experimental results obtained by Agarwal [60] and in most
cases the results agree with the data to within about 15%. As expected, the best results
were obtained for the quasi-isotropic laminate, where the assumption of a cosine normal
load distribution is acceptable. It should be noticed that the results are quite sensitive to
the values of the characteristic distances and that for laminates with a large proportion
of oo fibres the results are sensitive to the value of Sc. Due to the fact that the transverse
direct stress, cr22 , in the Yamada-Sun [59] failure criterion was assumed to be 0, this
approach is not suitable when the transverse strength of the lamina, Y, governs the
ultimate failure of the laminate.
Investigations based on the above described procedure have been performed by
several authors [61]-[64). Murthy et al. [61] used an inverse technique where a rigid pin
and a frictionless contact were assumed and the contact arc and pin-hole stress
distributions were calculated rather than assumed. The authors concluded that for a
(0°/905°)5 GFRP laminate the assumption of the cosine distribution leads to
underestimation of bearing failure and overestimation of shear and tensile failure loads.
Eriksson [62], using the two-dimensional finite element model previously described
[34], applied the point stress criterion to predict tension failures. It was assumed that
failure initiation occurs at points on the hole boundary where fibres are either tangential
or normal to the boundary. The characteristic distance should be calculated at each one
of these points. The concept of a characteristic curve together with the Y amada-Sun
[59] failure criterion was used to predict bearing failure. The author considered that
bearing failures initiate in the interval -45°:::;;e:::;;+45°. The shapes of the characteristic
curves obtained are strongly dependent on lay-up. No comparison with experimental
results was made and shear-out failure was not predicted. As tension failure generally
occurs in one defined plane, the points where the characteristic distances were
determined could have been limited.
The non-linear shear stress-shear strain behaviour was considered by Chang et al.
[63). The analysis follows that proposed in a previous work [58] including the non-
linear shear stress-shear strain relation proposed by Hahn and Tsai [64] in the Yamada-
Sun [59] failure criterion. For cross-ply and angle-ply laminates the strength predictions
using this method were more accurate than in the previous model [58] whilst for other
laminate configurations the linear and non-linear predictions were in very close
agreement. As in Sarabian's investigation [41], the results obtained by Chang eta!. [63]
highlight the importance of including the non-linear relation in the analysis of cross-ply
and angle-ply laminates.
3.3.4. FRACTURE MECHANICS
Methods based on linear elastic fracture mechanics (LEFM) have been developed for
both unloaded [65], [66] and loaded holes [67]. These methods are based on fracture
energies, this procedure being in accordance with Williams' [68] suggestion that, for
composites, energy release rates seems to be more physically meaningful than stress
intensities as a failure criterion.
In the case of unloaded holes, investigations have been made for traction [65] and
compression [66] loading. A damage zone model (DZM) was proposed, where damage
around the hole is represented by an equivalent crack with cohesive forces acting at the
crack surfaces. This crack represents matrix cracking and delamination in the traction
case [65], and fibre microbuckling and delamination in the compression case [66]. A
linearly decreasing relation between the cohesive stress and the crack opening, v, was
assumed, representing the increase in the extent of damage with increasing load. The
stress at the crack tip was assumed equal to the unnotched laminate strength, as shown
in Figure 3.6.
Damage Cohesive stresses
(J
<Jo
Equivalent crack
Figure 3.6 Damage zone model
Considering the traction case, when the maximum tensile stress at the notch reaches the
unnotched tensile strength of the laminate, cr0 , the equivalent crack is assumed to form.
With increasing loading, the crack opens, representing damage propagation. As shown
in Figure 3.6, this effect leads to a reduction of cohesive stress and to material softening.
The area under the cr-v relation is the total fracture energy, Gc, developed in the damage
zone. To obtain Gc the notched-to-unnotched strength ratio versus the apparent fracture
energy was calculated from the DZM, by using a two-dimensional finite element model
of a three-point bending specimen. The results were plotted versus the fracture energy,
and, for a given crack length, the value of Gc which gave the fracture load equal to
experimental results was chosen. This value was considered a material constant for a
certain lay-up.
Hollmann [67] extended the concept ofDZM to predict tension and shear failures in
bolted joints. A finger-tight, perfect-fit joint in (On°/90m0 /±45p 0 ) CFRP laminates was
considered. It was assumed that the shear stress-crack relative displacement relation had
the same form as in the tensile case (Figure 3.6), the area under the curve being the
shear fracture energy. Predefined crack paths were used along the tension and shear-out
planes and the fracture energies were obtained using a best fit to experimental data
procedure. Frictional effects were included in the analysis by considering three different
tangential boundary conditions, representing no friction, an intermediate amount of
friction and 100% friction (no sliding). Comparisons with experimental results showed
a good prediction of failure modes (shear-out in all specimens) and errors between
11.5% and -5.5%. However these results were obtained using a value of unnotched
shear strength, 'to, which was obtained using the same best fit to data used to determine
fracture energies. This procedure was justified by considering that 'to cannot be
determined from in-plane shear tests, as a constant stress distribution is not obtained.
Using 'to determined from in-plane shear tests, strength overpredictions of 40%-60%
were obtained. No significant differences in strength were obtained considering no
friction and 100% friction.
One of the limitations of the formulations based on DZM is that the mode of
propagation can change as the damage progresses and the self-similarity of crack
growth is not always assured.
3.3.5. PROGRESSIVE DAMAGE MODELS
Progressive damage models were developed to deal with the localised damage that
occurs prior to laminate failure in a more direct manner than the techniques previously
discussed. These models simulate damage initiation and growth, using elastic property
degradation models that are usually a function of the type of damage that occurred.
Models have been developed for both unloaded and loaded holes.
Shahid and Chang [69], based on a micromechanical analysis of a cracked lamina,
established constitutive equations relating material properties to extent of damage.
Damage accumulation was also taken into account by considering the ply strengths as a
function of crack density. Expressions for the transverse tensile and shear strengths as a
function of crack density were obtained on the basis of the theory of elasticity and
fracture mechanics. Other important features of the model are the inclusion of saturation
crack density, the fibre fracture process based on the fibre bundle theory, consideration
of matrix cracking-induced damage and account of the orientation of the neighbouring
plies when determining the stiffness of a cracked ply.
Based on an experimental characterisation of the bearing failure in composite

laminates [70] and on the above described procedure, Hung and Chang [71] developed a
novel method to predict the bearing strength in CFRP. Assuming a frictionless contact,
the two-dimensional form of Hashin's [72] failure criterion, taking into account the
variation of shear strength with crack density, was used to predict damage. Two
different ways to model bearing failure were used: for pin-loaded joints, it was assumed
that the joint could not sustain additional load when shear cracks occurred,
corresponding to damage accumulation at a characteristic distance from the hole
boundary. For joints with lateral constraints, the material degradation coefficients were
modified for the material under the washer surface. An empirical relationship between a
material degradation factor and the normalised clamping pressure was proposed. When
the accumulated damage reaches the edge of the washer, the stiffuess degradation
factors revert to the original values and failure occurs shortly after. Such a procedure
represents an indirect approach to model the effects of clamping pressure on the
laminate strength. The load transferred by friction between the washers and the laminate
was also considered in the analysis as a function of the clamping force. Good agreement
between predicted and experimental strengths and joint stiffuess was found.
Sun [73] proposed a three-dimensional finite element analysis based on the
progressive damage models developed by Chang et al. [69], [71] to study the effect of
clamping pressure on the bearing failure of double-lap composite joints. Eight-node
laminated brick elements were used and three configurations were proposed: pin-loaded
laminate, laminate with clamping pressure applied by washers, and laminate confined
between two metallic plates. The stiffuess of the bolt and metallic plates was taken into
account. The criteria used by Chang et al. [69], [71] to predict fibre compressive failure
was modified, and no longer includes the shear stresses. It was assumed that in-plane
matrix tension and compression failures have no effect on the out-of-plane properties,
whereas fibre failures degraded all material properties. In order to take into account the
bearing damage, it was assumed that the material under the lateral supports that failed in
fibre compression was incompressible. This hypothesis was implemented in the finite
element model using a penalty formulation. The predicted variation of the clamping
load with increasing load applied to the joint, showing a sharp increase in the clamping
load when bearing damage occurred, agreed reasonably well with experimental results.
Good agreement was found between joint strength predictions and experiments. Model
predictions showed an increase of strength with increasing bolt clamping force and,
more significantly, with increasing size of clamped area. The bolt stiffuess did not affect
the joint strength, which was expected taking into account the results previously
discussed [15], [16].
Tan [74], [75] developed a progressive damage model to predict failure in laminates
containing open holes loaded in tension [74] and compression [75]. Fibre failure was
predicted by a proposed quadratic criterion that utilises the longitudinal tensile and
compressive strength of a lamina and the lamina direct stress in the fibre direction.
Matrix cracking was predicted by the Tsai-Wu [76] criterion. Damage was represented
by a set of internal state variables, which are used to degrade E 11 , E22 and G 12
accordingly to the type of damage. The value of the internal state variable associated
with matrix tensile cracking was obtained from a micromechanical analysis of a cracked
lamina based on the theory of elasticity [77], [78]. The internal state variables associated
with matrix compressive cracking and fibre tensile and compressive fracture were
calculated using a parametric analysis. Total failure was assumed to occur when fibres
across the entire width of a ply are broken or when substantial matrix failure has
occurred in all the laminae. For CFRP laminates with several lay-ups predictions
correlated closely with experimental strengths. Damage progression basically agreed
with experimental results obtained by X-radiographic examination of the specimen at
several load levels.
The concept of internal state variables to simulate intralaminar damage proposed by
Tan and Nuismer [77], [78] was used in a three-dimensional finite element model of a
composite bolted joint in CFRP developed by Camanho and Matthews [79], [80]. As
shown in Figure 3.7, experimental results obtained in (0°/90°/±45°h CFRP specimens
loaded to several percentages of the failure load [81] showed a process of damage
accumulation leading to final failure that validates the use of progressive damage
models.
Figure 3.7 X-rays of specimen loaded to percentages of the failure load (bearing failure)
A finger-tight, double-shear joint in a (0°/90°/±45°h s carbon fibre/epoxy laminate was

modelled using ABAQUS [82] software. The three-dimensional progressive damage
model developed consisted of the following major components [80]: stress analysis,
damage prediction and damage-dependent constitutive equations.
Linear eight-node isoparametric three-dimensional solid elements were used to
model each ply of the laminate. This type of element assures that the contact forces are
consistent with the contact direction. The extent of the contact surface between the bolt
and the laminate was calculated throughout the analysis, rather than assumed, using the
Lagrange multipliers method [24], [26]. Following Hyer et al. [16] conclusions, the bolt
was modelled as a rigid body and the simulation of the finger-tight bolt was achieved by
restraining the through-thickness expansion of the nodes situated on the washer-
laminate contact surface. However, the through-thickness contraction of these nodes
was free .
Hashin's [72] three-dimensional failure criterion was used to predict damage within
a ply. This criterion can be used to predict fibre and matrix damage modes. It IS
established in terms of quadratic stress polynomials, in a piecewise form, accounting for
fibre and matrix failure separately:
Fibre tensile fracture, cr 11 ;?:0:
(3.4)
Fibre compressive fracture, cr 11 <0:

(3.5)
Matrix tensile or shear cracking, cr 22 +cr33 ;?:0:
(3.6)
Matrix compressive or shear cracking, cr22 +cr33 <0:
(3.7)
where X and Xc are respectively the tensile and compressive strengths in the fibre
direction ( 11 ), Y and Y c are respectively the tensile and compressive strengths in the
matrix direction (22) and Sc is the in-plane shear strength. Besides the failure criteria
used to predict the non-critical damage, a criterion to predict the load at which the joint
is no longer able to sustain the accumulated damage is required. In order to predict
tension and shear-out failure modes, it was assumed that total failure of the joint occurs
when the damage in the fibres extends to the laminate free-edge. Such an approach was
successfully used by other authors to predict failure in unloaded [74] and loaded holes
[83]. However, the approach proposed above cannot be used to predict total failure
when a joint fails by bearing and the global failure criterion was recast. Experimental
results [70], [81] have shownthat bearing failure is characterised by the accumulation of
subcritical damage mechanisms, leading to through-thickness shear cracks and final
failure of the joint. Bearing failure occurs shortly after the accumulated damage reaches
the washer's outer edge. It was then assumed that final bearing failure occurred when
the accumulated damage in the fibres reaches the outer edge of the washer.
When damage occurs, the material response was modified according to the type of
damage predicted using equations (3 .4)-(3. 7). The approach proposed assumes that the
presence of damage within an element has an effect on the elastic properties of that
element only. Following Tan's et al. approach [74]-[75], the effect of damage on the
stiffness of the material was represented using internal state variables DT and Df
(i=1,2,4), functions ofthe type of damage predicted. As a three-dimensional model was
developed, the procedure was extended to the through-thickness elastic properties [80].
Specimens that failed in the bearing, tension and shear-out modes were simulated
and the numerical predictions compared with experimental results. Figure 3.8 shows
through-thickness micrographs of the bearing (A-8) and shear-out (C-D) planes of a
specimen loaded to 7218 N. On increasing the load the specimen failed in tension.
B A
c D
B:
c
oo I
0 ' - - ---'1 mm Load
0 0. 125 mm
Figure 3.8 Specimen loaded to 7218 N
Figure 3.9 and Figure 3.10 show respectively the predicted damage in the 0° ply at 6580
N and at failure.
Matrix tensile cracking atrix compressive
n.
cracking
0° ply
~
~~a.df-3~---
~
_ ·- - ·- - . --
Fibre tensi le fracture Fibre compressive

fractu re
~·- · - K·· · - -·-- -- .... -·-· -
Figure 3.9 Predicted damage in the 0° layer of the tension specimen at 6580 N (90% of predicted
failure load)
Matrix tensile Matrix compressive

cracking
Figure 3.10 Predicted damage in the 0° layer of a fully failed tension specimen (7284 N)
The fibre compressive failure predicted by the numerical model in the ooplies of the
bearing plane is also visible in the micrograph shown in Figure 3.8. The model also
predicts the fibre tensile failure occurring in the ooplies in the shear-out plane and the
matrix cracks occurring in the off-axis plies in the bearing plane shown in Figure 3.8.
The predicted fractured surface across the specimen' s width was in agreement with
experimental results.
Figure 3.11 shows the load-displacement relations obtained in the experimental tests
and using the numerical model.
Lo~d tN)
8000
7000
6000
5000
~000
3000 --Experimen tal ( L VDT)
2000 - - 3D FF.modcl
1000
0
0.25 0.5 0.75 1.25
Boh di!!oplaccmcnt (mnt)
Figure 3. 11 Experimental and predicted load-displacement relations

Figure 3.11 shows that the stiffness of the joint was accurately represented. The
predicted failure loads for the different failure modes were in good agreement with
experimental results. However, the model underpredicted the strength (taken as the first
load drop-oft) in the case of bearing failures. This effect can be related to the
experimental results [81] that have shown the presence of de laminations for this failure
mode.
The progressive damage model was able to accurately predict failure modes, joint
stiffness and strength. The model develop@d 1I1m provides information for defining
failure loads based on other design constraints, e.g. mrudmum hole deformation or load
at which matrix cracking first appears. The model was able to reproduce the
catastrophic characteristics of shear and tensile failures, showing a confined amount of
damage at loads close to failure. In bearing failures, the model simulated damage
progression occurring under the washer, but ndt the unconstrained damage growth
developed in regions lying outside the washer edge. It was considered that delamination
should be included in the analysis if the material response is sought in this region and to
improve the predictions for bearing failure modes.
3.3.6. MULTI-FASTENERJOINTS
Some of the strength prediction methods previou!'!ly described have also been applied to
predict the strength of joints with several loaded holes.
A comprehensive experimental and numerical investigation on the effects of the
bearing stress-to-by-pass stress ratio, ~'was performed by Naik and Crews [50], [51].
This ratio was previously defined [44] by the expressiotts (3.1) and (3.2). ~was assumed
to depend on the joint stiffness and configuration. Assuming that one fastener in a
multi-fastener joint can be represented by a single fastener coupon subjected to both
bearing and by-pass loads, an experimental test system that could simultaneously apply
these loads was used. With this system, it was possible to apply a constant value of ~
until damage onset occurred. The damage onset was detected by the first non-linearity
in the load-displacement curve. The bearing-by-pass diagram obtained for a
(0°/45°/90°/-45°h,CFRP laminate is shown in Figure 3.12.
The symbols (o) represent measured Sb and Snp values for damage onset, and NT,
TRB, NC and CRB mean respectively net-section tension, tension reacted bearing, net-
section compression and compression reacted bearing. It can be seen that the tension
case with net-tension damage can be represented by a straight line. It was concluded
that local stresses responsible for damage onset consist of a component due to bearing
and a second superimposed component due to by-pass loading. The failure envelope
obtained is similar to the one obtained by Hart-Smith [4]. It should be noticed that for
compressive loading the damage for the filled hole case initiated at a higher stress than
for the open hole case. This fact was due to the dual bolt-hole contact that occurred in
the filled hole case, allowing load transfer across the hole. For all-by-pass loading (~=0)
the specimens failed soon after damage onset, whilst for all bearing load (~=±oo) the
specimens failed at considerably higher loads than the damage onset load.
Using the previously described inverse technique [28], the NASTRAN finite element
code was used to determine lamina strains [50] and stresses [51] around the hole
boundary. The bearing and by-pass stresses applied to the model corresponded to the
experimentally determined damage onset stresses for the modes previously investigated.
In almost all cases the strains and stresses around the hole boundary were greater than
the unnotched ultimate strains and stresses. As previously discussed, these differences
are typical of methods based on hole stress or strain distributions used together with
failure theories. As damage onset was found to be governed by fibre failure, in order to
predict strength it was assumed that damage onset occurs when the peak £ 11 in a ply
reaches the experimentally determined critical value for each failure mode. Using such
procedures the calculated damage onset stresses and failure modes for several values of
~ agreed reasonably well with experimental results. However, the accuracy was
dependent on ~ and for compressive loading unconservative predictions can occur.
CRB Bearing stress, Sb, MPa TRB

p 600 pp
p
~b
p~ •oo
\3
b
p.
l pa
" " ' '· ~00

~. ·
I/;
~
200 400
By-pass st ress, Snp, MPa
Figure 3.12 Bearing-by-pass diagram for damage onset strength [51]
Hung and Chang [83] developed the progressive damage models previously described
[69], [71] to predict the strength of bolted joints under multiaxialloads. A rigid bolt and
a frictionless contact were assumed and an iterative procedure was used to calculate the
contact angle. Hashin's [72] failure criterion was used to predict five basic damage
modes: matrix cracking (cr 22 ~0), fibre-matrix shearing (cr 11 ~0), fibre fracture (cr 11 ~0),
matrix compressive failure (cr22 <0) and fibre compression-shear failure (cr 11 <0). For
joints under uniaxial combined bearing and by-pass loads, good agreement for NT, NC,
TRB and CRB failures was found between model predictions and the experimental
results obtained by Naik and Crews [50], [51]. The non-linear relation between by-pass
stresses and strength shown in Figure 3.12 was justified by the results showing a
reduction in fastener-hole contact region resulting from higher compressive by-pass
stresses. Due to the unavailability of experimental results for biaxial loading, model
predictions could not be validated for this case. Results showed a decrease of strength
for CRB failures with increasing transverse by-pass tensile loads. However the opposite
effect occurred for NT failures.
3.4. CONCLUSIONS
From the description of the stress analysis methods, it is clear that semi-analytical
techniques have the advantages of not requiring high computing capabilities and
significant pre-processing and of providing solutions in an efficient amount of time.
However, some of the solutions are available for infinite dimensions, requiring finite
dimension correction factors. Numerical methods of stress analysis are a general tool for
determining stress distributions, allowing the consideration of factors that have
importance on the joint performance. Furthermore, numerical models are adequate to
use with all the strength prediction methods that have been developed.
The models to determine the stress distribution around a loaded hole should take
into account the effects of clearance or interference on the contact area, as well as the
form of the contact stress distribution, which should be calculated instead of being
assumed. Although the use of mechanically fastened joints in cross-ply and angle-ply
laminates is not recommended, the stress analysis of such laminates should include an
appropriate non-linear shear stress-shear strain relation.
The majority of the models developed to predict strength are based on two-
dimensional models. However, such models have some drawbacks: two-dimensional
failure criteria are applied on the hole boundary, where three-dimensional stresses that
may contribute to the damage mechanisms are present. Related to this fact, two-
dimensional models are clearly unable to explain the experimental results that
highlighted the importance of stacking sequence and lateral constraint on the joint
strength. Considering these facts, we can conclude that three-dimensional models are
general procedures to accurately predict the strength of mechanically fastened joints,
avoiding the use of empirical parameters to account for different stacking
sequences/clamping pressures. When using finite element techniques, the drawback
related to model size can be minimised by using three-dimensional elements only in the
bolt-bearing region, using substructuring techniques or using two- and three-
dimensional elements properly connected.
There is no definitive method to predict joint strength. Although some of the
methods described achieved good results for some laminate configurations and failure
modes, extrapolation to other situations is usually not possible. Methods based on
boundary stresses and failure theories are simple to implement but do not take into
account the localised damage and therefore underestimate the joint strength.
Furthermore, such methods are unable to predict failure modes.
The accuracy of both two-parameter and combined methods is dependent on
experimentally determined values that should be obtained for each material and
geometry. Two-parameter methods do not take into account the effect of interaction
between stresses on the joint strength and can lead to significant errors in predictions.
Fracture mechanics methods, based on the DZM, provide an indirect and simple
expression of the complex micromechanical failure process. However, the experimental
determination of fracture energies is required for each material system, using a best-fit-
to-data procedure and pre-defined crack paths are required. This approach must be
further developed to predict bearing failures.
Progressive damage methods have the advantages of being able to determine
damage mechanisms, the direction of damage propagation, failure modes and both
ultimate and residual strengths. Progressive damage models will have increasing
importance for the investigation of tougher composites that can develop a significant
amount of damage prior to total failure. Most of the approaches consider that the elastic
properties are reduced to zero in the damage zones, whilst a few recent models include a
gradual reduction of elastic properties. Further developments for the damage-dependent
constitutive equations implemented in progressive damage models and able to simulate
the damage mechanisms occurring in composite bolted joints are required. These
developments could be based on micromechanical models, based on stress analyses in
the presence of damage, or on damage continuum models, where the damage states arc
represented by a damage tensor and the composite treated as a continuum.
3.5. REFERENCES
1. Hart-Smith, L. J.: Mechanically-fastened joints for advanced composites··

phenomenological considerations and simple analysis, Douglas Paper 6748
(1978), 1-32.
2. Kretsis, G. and Matthews, F. L.: The strength of bolted joints in glass fibre/epoxy
laminates, Composites 16 (1985), 92-105.
3. Collings, T. A.: The strength of bolted joints in multi-directional CFRP laminates.
Composites 8 (1977) 43-54.
4. Hart-Smith, L. J.: Design and analysis of bolted and riveted joints in fibrous
composite structures, Douglas Paper 7739 (1986), 1-15.
5. Matthews, F. L., Kilty, P. F. and Godwin, E. W.: Load-carrying joints in fibre
reinforced plastics, Plastic and Rubber Process and Applications 2 ( 1982), 19-25.
6. Collings, T. A.: On the bearing strengths of CFRP laminates, RAE Technical
Report 82033, (1982), 1-23.
7. Shivakumar, K. N. and Crews, J. H.: Bolt clampup relaxation in a graphite/epoxy
laminate, NASA TM83268 (1982).
8. Shivakumar, K. N., and Crews, J. H.: An equation for bolt clampup relaxation in
transient environments", NASA TM 84480 ( 1982).
9. Hom, W. J. and Schmitt, R. R.: Influence of clamp-up force on the strength of
bolted composite joints, AIAA Journal32 (1994), 665-667.
10. Quinn, W. J. and Matthews, F. L.: The effect of stacking sequence on the pin-
bearing strength in GFRP, J. Composite Materials 11 (1987), 139-145.
11. Muskhelishvili, N. I.: Some Basic Problems of the Mathematical Theory of
Elasticity, Noordhoff International Publishing, Leyden, 1977.
12. Lekhnitskii, S. G.: Theory qf Elasticity of an Anisotropic Body, Mir Publishers,
Moscow. 1981.
13. de Jong, T.: Stresses around pin-loaded holes in elastically orthotropic or isotropic
plates, J. Composite Materials 11 (1977), 313-331.
14. Zhang, K. D. and Ueng, C. E. S.: Stresses around a pin-loaded hole in orthotropic
plates, J. Composite Materials 18 (1987), 432-446.
15. Hyer, M. W. and Klang, E. C.: Contact stresses in pin-loaded orthotropic plates,
International J. Solids and Structures 21 (1985), 957-975.
16. Hyer, M. W., Klang, E. C. and Cooper, D. E.: The effects of pin elasticity,
clearance and friction on the stresses in a pin-loaded orthotropic plate, J.
Composite Materials 21 ( 1987), 190-206.
17. Iarve, E.: Spline variational three-dimensional stress analysis of laminated
composite plates with open holes, International J. Solids and Structures 33 ( 1996),
2095-2118.
18. Iarve, E.: Three-dimensional stress analysis in laminated composites with fasteners
based on the B-spline approximation, Composites-Part A 28 (1997), 559-571.
19. Iarve, E.: Stress analysis in laminated composites with fastener holes, Proceedings
of the American Society for Composites 1Oth Technical Conference (1995),
Technomic Publishing Co., Lancaster, 408-419.
20. Iarve, E. and Scaff, J.R.: Stress analysis of open and fastener hole composites
based on three-dimensional spline variational technique, AGARD Conference
Proceedings 590 (1997), Advisory Group for Aerospace Research and
Development, Florence, 4.1-4.12.
21. Mollenhauer, D.H.: Interlaminar Deformation at a Hole in Laminated Composites:
A Detailed Experimental Investigation Using Moire Interferometry, Ph.D. Thesis,
Virginia Polytechnic Institute and State University, 1997.
22. de Jong, T.: Stresses in pin loaded anisotropic plates, AGARD Conference
Proceedings 427 ( 1987), Advisory Group for Aerospace Research and
Development, Madrid, 5.1-5.17.
23. Zienkiewicz, 0. C. and Morgan, K.: Finite Elements and Approximations, John
Wiley and Sons, New York, 1983.
24. Bathe, K. J.: Finite Element Procedures, Prentice-Hall, New Jersey, 1996.
25. Matthews, F.L., Davies, G.A.O., Hitchings, D. and Soutis, C.: Finite Element
Analysis of Composite Materials and Structures, Woodhead Publishing Ltd,
Abington, 2000.
26. Crisfield, M. A.: Non-Linear Finite Element Analysis of Solids and Structures,
Vol.2, John Wiley & Sons, Chichester, 1997.
27. Zhong, Z.-H. and Mackerle, J.: Static contact problems- a review, Engineering
Computations 9 (1992), 3-37.
28. Naik, R. A. and Crews, J. H.: Stress analysis method for a clearance-fit bolt under
bearing conditions, AIAA Journal24 (1985), 1348-1353.
29. Ramamurthy, T. S.: New studies on the effect of bearing loads in lugs with
clearance fit pins, Composite Structures 11 (1989), 135-150.
30. Ramamurthy, T. S.: Recent studies on the behaviour of interference fit pins in
composite plates, Composite Structures 13 (1989), 81-99.
31. Crews, J. H., Hong, C. S. and Raju, I. S.: Stress-concentration factors for finite
orthotropic laminates with a pin-loaded hole, NASA TP 1862 (1981).
32. Wilkinson, T. L., Rowlands, R. E. and Cook, R. D.: An incremental finite-element
determination of stresses around loaded holes in wood plates, Computers and
Structures 14 (1981), 123-128.

33. Rowlands, R. E., Rahman, M. U., Wilkinson, T. L. and Chiang, Y. 1.: Single and
multiple-bolted joints in orthotropic materials, Composites 13 (1982), 273-279.
34. Eriksson, L. 1.: Contact stresses in bolted joints of composite laminates, Composite
Structures 6 (1986), 57-75.
35. Xiao, Y., Wang, W.-X., Takao, Y. and Ishikawa, T.: The effective friction
coefficient of a laminate composite, and analysis of pin-loaded plates, J. Composite
Materials 34 (2000), 69-87.
36. Jones, R. M.: Mechanics of Composite Materials, Taylor and Francis, U.S.A.,
1975.
37. Matthews, F. L., Wong, C. M. and Chryssafitis, S.: Stress distribution around a
single bolt in fibre-reinforced plastic, Composites 13 ( 1982), 316-322.
38. Chen, W. H., Lee, S. S. and Yeh, J. T.: Three-dimensional contact stress analysis
of a composite laminate with bolted joint, Composite Structures 30 (1995), 287-
297.
39. Smith, P. A., Pascoe, K. J., Polak, C. and Stroud, D. 0.: The behaviour of single-
lap bolted joints in CFRP laminates, Composite Structures 6 (1986), 41-55.
40. Marshall, I. H., Arnold, W. S. and Wood, J.: Observations on bolted connections in
composite structures, Composite Structures 13 (1989), 133-151.
41. Sarabian, S. M.: An experimental and finite element investigation into tbe
nonlinear material behaviour of pin-loaded composite laminates, MTL Report TR
91-2 (1991), U.S. Army Materials Technology Laboratory, Massachusetts, 1-265.
42. Wang, S. and Han, Y.: Finite element analysis for load distribution of multi-
fastener joints, J. Composite Materials 22 (1988), 124-135.
43. Griffin, 0. H., Hyer, M. W., Cohen, D., Shuart, M. J., Yalamanchili, S. R. and
Prasad, C. B.: Analysis of multifastener composite joints, J. Spacecraft and
Rockets 31 (1994), 278-284.
44. Naik, R. A. and Crews, J. R.: Stress analysis method for clearance-fit joints with
bearing-by-pass load, AIAA Journal29 (1991), 89-95.
45. Kim, S. J. and Kim, J. H.: Finite element analysis of laminated composites with
contact constraint by extended interior penalty methods, International J. for
Numerical Methods in Engineering 36 (1993), 3421-3439.
46. Godwin, E. W. and Matthews, F. L.: A review of the strength of joints in FRP-part
1: mechanically fastened joints, Composites 11 (1982), 155-160.
4 7 . Johnson, M. and Matthews, F. L.: Determination of safety factors for use when
designing bolted joints in GRP, Composites 10 (1979), 73-76.
48. Waszczak, J. P. and Cruse, T. A.: Failure mode and strength prediction of
anisotropic bolt bearing specimens, J. Composite Materials 5 (1971), 421-425.
49. Tang, S.: Failure of composite joints under combined tension and bolt loads, J.
Composite Materials 15 (1981), 329-335.
50. Naik, R. A. and Crews, J. H.: Ply-level failure analysis of a graphite/epoxy
laminate under bearing-bypass loading, NASA TM 100578 (1988), 1-35.
51. Crews, J. H. and Naik, R. A.: Combined bearing-bypass loading on a
graphite/epoxy laminate, Composite Structures 6 (1986), 21-40.
52. Crews, J. H.: A survey of strength analysis methods for laminates with holes, J.
Aeronautical Society ofIndia 36 (1984), 287-303.
53. Whitney, J. M. and Nuismer, R. J.: Stress fracture criteria for laminated composites
containing stress concentrations, J Composite Materials 8 (1974), 253-265.

54. Nuismer, R. J. and Labor, J.D.: Applications of the average stress failure criterion:
part !-tension, J Composite Materials 12 (1979), 238-249.
55. Nuismer, R. J. and Labor, J.D.: Applications of the average stress failure criterion:
part 11-compression, J Composite Materials, 13 ( 1979), 49-60.
56. Wilson, D. W. and Pipes, R. B.: Analysis of the shearout failure mode in composite
laminates, 1st International Conference on Composite Structures ( 1981 ), 34-49.
57. York, J. L., Wilson, D. W. and Pipes, R. B.: Analysis of the net-tension failure
mode in composite bolted joints, J Reinforced Plastics and Composites 1 (1982),
141-152.
58. Chang, F. K., Scott, R. A. and Springer, G. S.: Strength of mechanically fastened
composite joints, J Composite Materials 16 (1982), 470-494.
59. Yamada, S. E. and Sun, C. T.: Analysis of laminate strength and its distribution, J
Composite Materials 12 ( 1978), 275-284.
60. Agarwal, B. L.: Static strength prediction of bolted joints in composite materials,
AIAA Journal18 (1980), 1371-1375.
61. Murthy, A. V., Dattaguru, B., Narayana, H. L. and Rao, A. K.: Stress and strength
analysis of pin joints in laminated anisotropic plates, Composite Structures 19
(1991), 299-312.
62. Eriksson, L. I.: An analysis method for bolted joints in primary composite, AGARD
Conference Proceedings 427 (1987), Advisory Group for Aerospace Research and
Development, Madrid, 6.1-6.19.
63. Chang, F. K., Scott, R. A. and Springer, G. S.: Failure strength of nonlinearly
elastic composite laminates containing a pin loaded hole, J Composite Materials
18 (1984), 464-477.
64. Hahn, H. T. and Tsai, S. W.: Nonlinear elastic behaviour of unidirectional
composite laminate, J Composite Materials 7 ( 1973), 102-110.
65. Aronsson, C. G.: Strength of carbon/epoxy laminates with countersunk hole,
Composite Structures, 24 (1993), 283-289.
66. Soutis, C., Curtis, P. T. and Fleck, N. A.: Compressive failure of notched carbon
fibre composites, Proceedings R. Soc. London 440 (1993 ), 241-256.
67. Hollmann, K.: Failure analysis of bolted composite joints exhibiting in-plane
failure modes, J Composite Materials 30 (1996), 358-383.
68. Williams, J. G.: Fracture mechanics of anisotropic materials, in Application of
Fracture Mechanics to Composite Materials, K. Friedrich, ed., Elsevier,
Amsterdam, 1989.
69. Shahid, I. S. and Chang, F. K.: An accumulative damage model for tensile and
shear failures of laminated composite plates, J Composite Materials 29 ( 1995),
926-981.
70. Wang, H. S., Hung, C. L. and Chang, F. K.: Bearing failure of bolted composites
joints: part 1- experimental characterization, J Composite Materials 30 (1996),
1284-1313.
71. Hung, C. L. and Chang, F. K.: Bearing failure of bolted composite joints: part 11-
model and verification, J Composite Materials 30 (1996), 1359-1400.
72. Hashin, Z.: Failure criteria for unidirectional fiber composites, J Applied
Mechanics 47 (1980), 329-334.
73. Sun, H. T.: Strengths of Composite-to-Metal Double-Lap Bolted Joints, Ph.D.
Thesis, Stanford University, 1997.

74. Tan, S. C.: A progressive failure model for composite laminates containing
openings, J. Composite Materials 25 (1991), 556-577.
75. Tan, S. C. and Perez, J.: Progressive failure of laminated composites with a hole
under compressive loading, J. Reinforced Plastics and Composites, 12 (1993),
1043-1057.
76. Tsai, S. W. and Wu, E. M.: A general theory of strength for anisotropic materials,
J. Composite Materials 5 (1971), 58-80.
77. Nuismer, R. J. and Tan, S. C.: Constitutive relations of a cracked composite
lamina, J. Composite Materials 22 (1988), 306-321.
78. Tan, S. C. and Nuismer, R. J.: A theory for progressive matrix cracking in
composite laminates, J. Composite Materials 23 (1989), 1029-1047.
79. Camanho, P. P. and Matthews, F. L.: Delamination onset prediction in
mechanically fastened joints in composite laminates, J. Composite Materials 33
(1999), 906-927.
80. Camanho, P. P. and Matthews, F. L.: A progressive damage model for
mechanically fastened joints in composite laminates, J. Composite Materials 33
( 1999), 2248-2280.
81. Camanho, P. P., Bowron, S. and Matthews, F. L.: Failure mechanisms in bolted
CFRP, J. Reinforced Plastics and Composites 17 (1998), 205-233.
82. ABAQUS 5.6 User's Manuals, Hibbitt, Karlsson and Sorensen, Inc., Pawtucket,
U.S.A., 1996
83. Hung, C. L. and Chang, F. K.: Strength envelope of bolted composite joints under
bypass loads, J. Composite Materials 30 (1996), 1402-1435.
CHAPTER 4 STRENGTH DETERMINATION OF MECHANICAL FASTENED
JOINTS
F.-K. CHANG and X.L. QING

Department of Aeronautics and Astronautics, Stanford University, Stanford, CA 94305,
USA
4.1 INTRODUCTION
Joining by mechanical fasteners is one of the common practices in the assembly of

structural components. However, mechanical fasteners introduce complicated stress
field near the bolt hole area in the structures. As a consequence, mechanical fastened
joints are frequent sources of failure in aircraft and spacecraft structures. It is well
known that they can severely reduce the load carrying capability of the components by
more than fifty percent [1 ]. Optimal design of joints improves not only structural
integrity and performance, but more importantly, it considerably minimizes the weight
of the structures and hence, can increase the load-carrying capability.
Fiber reinforced laminated composite materials have been widely used in aircraft
and spacecraft structures because of their high strength and stiffness to weight ratios.
Due to the complex failure modes of composite materials, the mechanical joining of
structures made of composite materials demands much more rigorous design knowledge
and techniques than those currently available in the traditional methodology for metallic
joints. Damage in bolted composite joints can initiate at an early loading stage and
accumulate inside the laminates as the load increases. The accumulation of damage and
the mode of failure strongly depend upon the material, ply orientation, laminate
thickness, joint geometry, and loading condition, etc.
Considerable analyses and experiments on bolted composite joints have been
performed to evaluate the effect of material properties and ply orientation as well as
layup on the joint response and failure in the literature [2-24]. There are in general three
basic failure modes in bolted composite joints: net-tension, shear-out, and bearing (see
Figure 4.1 ). Net-tension failure is associated with fiber and matrix tension failures, due
to stress concentrations, while shear-out and bearing failures result primarily from the
shear and compression failures of fiber and matrix. The first two failure types tend to
101
L. Tong and C. Soutis (eds.),
·© 2003 Kluwer Academic Publishers.
102 F.-K. Chang and X. L. Qing
fail catastrophically. The bearing failure is more progressive and may not result in total
reduction of the load-carrying capability of the joints. A combination of any of the three
modes may also occur in the joint.
It has been demonstrated experimentally that the strength and response of bolted
composite joints that failed in bearing mode are very sensitive to clamping area, initial
clamping force, bolt stiffuess, etc [25-40]. Lateral supports and clamping pressure
significantly affect the bearing strength. For the best performance, the initial clamping
force recommended by manufacturers can be quite different for different types of bolts.
Undoubtedly, the lateral constraining effect on a bolted composite joint is a three-
dimensional problem. But most of the existing analyses are based on a two-dimensional
plane stress analysis, which cannot consider such effects. Hung and Chang [36, 371
characterized the bearing damage mechanism and mechanics due to clamp-up loads in
bolted composite joints. Three-dimensional stress analyses based on pin joints have
been studied to model the through the thickness effect [41-50]. Recently, Chang and his
associates [51-54] developed a computer code 3DBOLT/ABAQUS for predicting tht:
strength and failure response of bolted composite joints. The constitutive model
implemented in the code includes a progressive damage analysis for the damaged
material and a volume conserving technique that accounts for the incompressible:
behavior of damaged material.
•
Net-Tension Shear-Out Bearing
* {
:\
)
\.._ /
Applied
Load +
Figure 4.1 Illustration of the three failure types in mechanically fastened composite joints
In practice, mechanically fastened joints are subjected to bypass loads. It has also been
demonstrated experimentally that axial-bypass loads can significantly affect both the
strength and failure mode of bolted composite joints [20]. An axial-bypass load could
significantly reduce the strength of bolted joints and change the failure mode from
noncatastrophic bearing failure to catastrophic tension failure. Therefore, in order to
Chapter 4 Strength Determination of Mechanical Fastened Joints 103
optimally design bolted joints in composite structures, the relationships between the
bypass loads and the strength and failure of mechanically fastened composite joints
should be thoroughly understood. However, it is very difficult and expensive to
conduct appropriate experiments to accurately assess the bearing response of bolted
composite joints under multi-axial bypass loads [55]. Accordingly, the development of
an adequate analytical model for establishing such relationships is essential. Based on
the damage accumulation models [37], an investigation was performed to study the
strength and failure of bolted composite joints subjected to bypass loads by Hung and
Chang [56]. The strength envelope for a bolted composite joint under multi-axial bypass
loads was generated. The predictions of bolted composite joints subjected to an axial
bypass load were compared with available experimental data. The envelope can
provide information on the strength as well as the failure type of the joint as functions
of the bypass loads and is, therefore, very useful in joint design.
In this chapter, a progressive damage analysis will be summarized based on the
studies [34-37, 52, 53, 57]. Then, the effect oflateral support and clamping pressure on
the strength and failure response of bolted composite joints, and the strength envelope
of bolted composite joints under bypass loads, are presented.
4.2 PROGRESSIVE FAlLURE ANALYSIS
Based on the damage mechanisms, a progressive failure analysis model was developed
by Chang and his associates [34, 37, 52-54]. The model is composed of three major
parts:
• Constitutive modeling
Appropriate constitutive relations were developed on the ply level to estimate the
effect of local damage on the laminate stiffness.
• Damage Estimator
Failure criteria were selected to predict the damage accumulation in composites as a
function of the applied load.
• Stress Analysis
Nonlinear finite element analyses were used to calculate the stresses and
deformations of bolted composite joints.
4.2.1 ACCUMULATED DAMAGE PREDICTION
A set of damage accumulation criteria for predicting damage accumulated in bolted

composite joints were proposed by Shahid, Hung and Chang [34, 37]. Six failure
modes were considered in the literature [54]: matrix tension, fiber-matrix shearing, fiber
tension, matrix compression, fiber-matrix compression, and fiber compression. Two-
dimensional in-plane stresses were selected to predict damage and the corresponding
failure mode.
(1) Matrix Tension

( J( J
_!!_1.._
~(tfJ)
+ ~ >1
S(tfJ) - ,
(J2 >0
(2) Fiber-Matrix Shearing

[~J+~'-J>I
xt s(tfJ) '
CJ1>0, az+a3>0
(3) Fiber Tension (::)'>I, (Jl >0
(4) Matrix Compression

(~: J+(;;>)'>I, (J2 <0
(5) Fiber-Matrix Compression (~J +(;;)J >I, (Jl <0
(6) Fiber Compression (~J >I, (Jl <0
where a 1 and a 2 are the in-plane normal stresses parallel and normal to the fiber
direction of the ply under consideration, respectively, a 3 is the out-of-plane nomml
stress ofthe ply, and a6 is the in-plane shear stress of the ply. The corresponding on-
axis coordinate system for a ply is shown in Figure 4.2. X 1 and Xc are the tensile and
compressive strengths of the ply along the fiber direction, respectively. ¥; (tfJ), Yc and
S(tfJ) are the transverse tensile strength, transverse compressive strength and shear
strength of the ply, respectively. Since compression does not create matrix cracks,
except under shear loads, the transverse compressive strength Yc , is not a function of
the matrix crack density <)>.
Figure 4.2 The on-axis coordinate system used in the damage analysis
It is important to point out that both~ (tfJ) and S(tfJ) are not constants and may vary from
ply to ply in the laminate and depend on the crack density. Both Y, (l/J) andS{l/J) can be
determined based on the assumption of conservation of energy through fracture

mechanics. The expression of effective ply strengths :r; (cp )and S(cp )can be found in
reference [34].
Based on experimental observations [53], the test results showed that the lateral
compressive load can suppress the fiber-matrix splitting. Accordingly, it is postulated
that the occurrence of fiber-matrix shearing will only occur whena 2 +0' 3 > 0.
For a given stress state, if the ply stresses satisfY any one of the criteria, the mode of
failure and the state of accumulated damage are predicted on a ply-by-ply level.
Consequently, the criterion can be used to monitor the damage progression in laminated
composites if the corresponding constitutive equations for a ply are developed.
4.2.2 CONSTITUTIVE MODELING
Once the damage and the corresponding failure mode are predicted, the resulting
residual properties of the damaged material have to be determined in order to model the
full response of bolted composite joints. A 2-D material degradation model was
developed by Shahid, Hung and Chang [34, 37]. In order to simulate accurately the
bearing response of bolted composite joints in real structure, a 3-D material degradation
model was adopted by Sun, Qing and Chang [52-54].
4.2.2.1 Matrix Tension

Under in-plane tensile and shear loads, matrix cracks are progressively generated in
laminated composites, which degrade laminate properties. Based on the theory of
elasticity and the concept of continuum damage mechanics, Shahid and Chang [34]
derived the effective constitutive relationships that relate the stiffuess of each lamina in
a symmetric laminate as a function of the matrix crack density in its ply. These
relationships can be extended to the 3-D on-axis coordinate system of the ply as follows
{a}=[C(¢)] {e} (4.1)
a, c" (¢) c" (¢ )dml C"(¢) 0 0 0 e,

a, c"(<P)d,, C, (¢ )dmt Cn(¢) 0 0 0 e,
or a) c),(¢) c, (¢) c33(¢) 0 0 0 e3 (4.2)
a. 0 0 0 c •• (¢) 0 0 2£.
a, 0 0 0 0 c, (¢) 0 2£,
a. 0 0 0 0 0 c •• (¢) 2£6
where [C(cf>)] is the effective degraded stiffuess related with the matrix cracks.
When matrix tension has occurred, the stiffuess matrix is degraded as follows
ell (ct>) C,, (cj> )dmt ell (cp) 0 0 0

c,,(cp)dmt C, (cj> )dmt c23 (ct>) 0 0 0
CJI (cp) c" (cp) c33 (ct>) 0 0 0 (4.3)
[c(cp )]
0 0 0 c •• (cp) 0 0
0 0 0 0 css (cp) 0
0 0 0 0 0 c,, (cp)
where dmt is the matrix cracking failure degradation factor, and is defined as follows
when
when
where t:g is defined as the ply transverse tension failure strain corresponding to the
saturated crack density ¢0 of a unidirectional ply in the laminate under consideration
and is given by
Eg=Yt(cf>o)
E,
where ~ (¢0 ) is the ply transverse tensile strength at saturated crack density and £ 22 IS
the ply transverse Young's modulus.
4.2.2.2 Fiber-Matrix Shearing

For fiber-matrix shearing, the degraded on-axis ply stiffness matrix can be written in the
3-D stiffness matrix as follows
ell (cp )dfip C12 (cp) C, 1 (cp) 0 0 0

c" (cp) en (cp) c, (cp)
1 0 0 0
c,, (cp) c,, (cp) c,3(cp) 0 0 0 (4.4)
[c(cp )]
0 0 0 c .. (ct>) 0 0
0 0 0 0 Css (cp )dF 0
0 0 0 0 0 c" (ct> )d1,
where dfs, dfsp are the fiber-matrix shearing failure degradation factors. d fi is defined
as follows
0
df, = 1 when Y12 < Y12
0
d 1s = 0 when Y12 ~ Y12
where y1°
2 is defined as the ply shear failure strain corresponding to the saturated crack
density ¢0 of a unidirectional ply in the laminate under consideration and is given by

y~, = s~~:)
where S(¢0 ) is the ply shear strength at saturated crack density and Gxy is the ply shear
modulus; and d fsp is defined as follows
dl,p = Yn
where n is the number of plies in the ply group under consideration. df,P is used to
describe the degradation of ply stiffness in the fiber direction due to fiber-matrix
splitting failure mode in the laminate. Based on experimental observations, it is
postulated that the reduction of fiber stiffness due to fiber-matrix splitting failure is
inversely proportional to the thickness of the ply group under consideration.
4.2.2.3 Fiber Breakage

For fiber breakage, the degraded on-axis ply stiffness has the following form
c, (cp)df c" (cp )d I c,3(cp)dl 0 0 0

c, (cp}tl c" (cp )dt c,3 (cp )dt 0 0 0
c],(cp)dl c], (cp )d 1 ell (cp )d, 0 0 0 (4.5)
[c(cp )]
0 0 0 CH (¢ )d I 0 0
0 0 0 0 c,(cp)d, 0
0 () 0 0 0 c66 (cp )dt
where dr is defined as follows
d 1 =I when Ar < 8"

d/=0 when A 1 28 2
where A 1 is the accumulated fiber breakage area over which the stress is either equal to
or higher than the tiber longitudinal strength X, of the composite, and 8 is the criteria
fiber interaction zone which is defined as an effective length of a fiber break on the
surrounding fibers [58].
4.2.2.4 Matrix Compression

For matrix compression, the degraded on-axis ply stiffness takes the following form
ell cl2dmc ell 0 0 0

C2ldmc c22dmc en 0 0 0
[c(¢)] ell c32 c33 0 0 0 (4.6)

0 0 0 c •• 0 0
0 0 0 0 Css (¢) 0
0 0 0 0 0 c66 (¢ l
where d me is the matrix compression failure degradation factor, and is defined as

follows
dmc =1 when £2 < £~
dmc =0 when £ 2 ~£~
where e; is defined as the ply transverse compression failure and is given by

c Yc
£2 = -
£22
where Yc is the ply transverse compressive strength and £ 22 is the ply transverse
Young's modulus. Note that because compression creates no crack in the matrix except
under shear loads, the material properties, except for the shear modulus, are not
functions of the crack density cf> •
4.2.2.5 Fiber-Matrix Compression

For fiber-matrix compression, the degraded on-axis ply stiffness has the following
expression
ell c,2 CJJ 0 0 0

c21 c22 c" 0 0 0
[c(¢)] ell c12 CJJ 0 0 0 (4.7)

0 0 0 c •• 0 0
0 0 0 0 C55 (¢ )d fmc· 0
0 0 0 0 0 c66 (\i'J)d fmc
where dfmc is the fiber-matrix compression factor, and is defined as follows
djmc =1 when
dfmc =0 when
°
where y 12 is defined as the ply shear failure strain corresponding to the saturated crack
density !f>0 of a unidirectional ply in the laminate under consideration and is given by
Yio2 = S(!f>o)
Gxy
where S(l/>0 ) is the ply shear strength at saturated crack density and Gxy is the ply shear
modulus.
4.2.2.6 Fiber Compression

For fiber compression, the degraded on-axis ply stiffuess can be expressed as follows
Clldfc C"d"' C 13 dt<· 0 0 0

C2Idfc C22 d 1,. C 23 df<· 0 0 0
c3Idk c3,d"' c33dr, 0 0 0 (4.8)
[c(~ )]
0 0 0 C 44 d 1, 0 0
0 0 0 0 C55 (~ )d fc 0
0 0 0 0 0 c66(~)d fc
where d fc is fiber compression factor, and is defined as follows
dk = 1 when t: 1 < t:t'

d 1c = 0 when t: 1 ~ t:t
where t:t is defined as the ultimate longitudinal compression strain and is given by
c XC
t:l = -
Ell
where Xc is the longitudinal compression strength of composite and Ell IS the

longitudinal Young's modulus.
Notation:
aI a II c:I c:II
a2 (1::!2 c:, f22
a3 a33 c:3 £33
a. a23 £4 en
a, al) c:, £13
a6 (j 12 c:6 fl2
4.2.3 THREE-DIMENSIONAL BEARING EFFECT
4.2.3.1 Failure Mechanisms and Mechanics

A set of typical load-displacement curves of double-lap joints was generated by Sun and
Chang [52], as shown in Figure 4.3. The three curves in this Figure represent three
different types of double-lap joint configurations: pin joint, bolted joint with washers
and bolted joint without washers. Without any lateral supports, the failure load of the
pin joint was found to be the lowest. With the lateral constraints, the joints sustained a
higher bearing failure load than the pin joint. The bolted joint without washers had the
highest failure load. Tested specimens were sliced for damage inspection. Figure 4.4
shows a picture of a damaged specimen sliced from the center and a schematic
description of the bearing damage mechanisms in double-lap bolted joints. Visible shear
cracks were formed by the accumulated in-plane compression failure in each individual
ply of the laminate. Shear cracks appeared to be the primary failure mode of bearing
damage and caused significant loss of stiffuess. As shown in Figure 4.5, once the
damaged material resulting from compression failure propagates to the laminate surface,
it leads to joint collapse if no lateral constraints exist. In conclusion, if there were lateral
constraints to the bearing-damaged region of the specimen, the initiation and
propagation of shear cracks could be suppressed, resulting in a higher bearing load.
Therefore, it was anticipated that bolted joints with lateral constraints could sustain a
higher load than those without.
10000
TS00/3900-2
..-._
"'
::9
'-'
]
0
....:l 5000
"1:1
.....v
•
~
DwiD=2
Pin joint
• [(0/±45/90)3]s
fc = 800 lbs
0
0 0.1 0.2 0.3
• Displacement, 8 (in)
Figure 4.3 Load-displacement curves of double-lap joints under variable sizes of lateral clamped
areas (1 inch= 2.54 em, I lb =0.454 kg) [52]
Chapter 4 Strength Determination of Mechanical Fastened Joints Ill
Bearing load
•+1tf1
T
I
__!.._ shear
cracks
Figure 4.4 A picture and schematic description of the bearing damage mechanisms in double-lap
bolted composite joints [52]
Double-Lap Joint
Increase
bearing load
Lo of
washer support
Figure 4.5 A schematic description of the damage mechanisms of double-lap bolted composite
joints with washer supports [53]
Experimental observations seem to indicate that the mechanisms of bearing failure in

bolted composite joints occur stage by stage as damage propagates into the composites.
Within each stage, a set of shear cracks will be formed for joints with low initial
clamping pressure due to accumulated damage in individual plies of the laminates.
Shear cracks may not be formed for joints with high initial clamping pressure. This is
because the higher the lateral clamping pressure is, the less transverse expansions in a
laminate will be likely. As a result, clamping force suppresses the development of shear
cracks. However, it is believed that clamping pressure does not change the nature of the
mechanisms of bearing failure under lateral constraints, but only restrains the damaged
material from forming the visible shear cracks. Hence, the increase of bearing strength
in the experiments for joints with the increase of initial clamping pressure can probably
be attributed to the friction forces between center laminates and side panels or washers.
However, the nature of the bearing failure mechanisms remains unchanged
Clearly visible in the bolt bearing experiments was that as the damaged material
grew outside the constrained area provided by washers, the damaged material swelled
and bulged laterally because of lacking lateral constraint to suppress the damaged
material from spreading apart. This kind of phenomenon can be described with thl!
analogy of a pile of sand. Sand is similar to the damage material with very low stiffness.
It easily spreads apart if there is no bucket to hold it, but it can sustain significant
compressive pressure as it is put into a bucket. Figure 4.6 illustrates this analogy.
Without
side support
With
side support
Figure 4.6 Sand is used as analogy for the crushed composite material under lateral constraints
[52]
Experiments demonstrated that bearing damage was limited to a region underneath the
bolt-hole contact area. Failure of bolted joints in bearing mode results from the
formation of shear cracks associated with fiber compression failure in the bolted region
[36, 51-53].
It was observed that shear cracks greatly reduce the stiffuess of composite materials
in the area where bearing damage takes place, and are associated with fiber compression
failure. Along the paths of the shear cracks, extensive fiber kinking takes place.
Therefore, the fiber compression failure mode helps trigger the through-the-thickness
shear cracks and results in significant stiffness loss in the damaged material.
However, the joint strength could be improved significantly by adding lateral
supports, such as washers and side plates, to a bolted joint. Lateral supports offer
constraints on the damaged material when the bearing damage occurs.
4.2.3.2 Hypothesis for Damaged Material under Lateral Constraint

The reason why lateral clamping is so critical to the joint strength is the volume
conserving behavior of the damaged material residing in the bearing area. Although the
material has been damaged and has lost most of its stiffuess under compressive load, it
is still constrained by the surrounding intact material and by the lateral support offered
by the clamping and is thus capable of enduring considerable compressive stresses. For
composite material in the bolted region that failed in the tiber compression modes, it
was assumed that the damaged material became incompressible due to the lateral
clamping constraints. The incompressible condition assumed that the change of a unit
volume due to compression from one strain state to another is zero, i.e.,
H+lV-"V =~V=O (4.9)
and
~v
£, =-=£1 +£, +£3 (4.10)
, "V -
Eq. (4.10) can also be written as follows
(4.11)
t:, =(I I I 0 0
e" is the volumetric strain of composite material. For an incompressible material, the
change of volumetric strain£v is zero. This means that the material is not allowed to
expand or shrink but is permitted to change its shape as long as its volume remains a
constant.
4.2.3.3 Imposing Volumetric Constraint by Penalty Method

In order to model the volume conserving behavior of the bearing-damaged material
under consideration, a 3-D penalty formulation [59] was incorporated in the current
finite element analysis. Consider a motion of a continuum body that deforms from the
known current configuration n V to the state of next configuration n + 1V due to the
applied external loads. Following the method of the Updated Lagrangian formulation
[60], a variational form based on total potential energy of the body can be
approximately written as [51,53]
(4.12)
where n Ciikl is the tangent moduli of the material at current configuration "V . n+~Eii IS
the incremental strain tensor associated with infinitesimal displacements and n+~1h IS
the rotational strain tensor associated with finite deformation at step n+ 1. n+l~ is the
surface traction at current configuration "V . "aii is the Cauchy stress tensor in the
current configuration "V .
By adding the constraining Equation (4.1 0), the total potential energy Equation (4.12)
associated with volumetric constraint II can be written as follows
(4.13)
The quantity p is a 'penalty number' required for solving the constrained equations.
4.2.3.4 Complete Damage Model with the Addition of Volumetric Constraint

When bearing damage is predicted by the fiber compression criterion and the
accumulated damage area is under lateral supports, e.g., by washers, side plates, etc., the
vclume conserving mode comes into play. By having the volume of the damaged
material preserved under lateral supports, the mechanically fastened joints are able to
carry higher bearing load and thus achieve better performance.
As shown in Figure 4.7, the volumetric constraint for bearing-damaged composite
materials, once developed, was added to the existing progressive damage model
described earlier to complete the material modeling in the failure analysis of composite
bolted joints. Once a damage mode is predicted by failure criteria, the material
properties are updated accordingly. If fiber compression happens and the damaged
material is under constraint, the volume conserving technique is imposed upon tlw
damaged material.
4.3 NUMERICAL PREDICTION
Finite element analysis has been adopted to analyze the failure of bolted composite
joints including open-hole and filled-hole laminates, double-lap bolted joints and single-
lap bolted joints [52-54]. The commercial finite element package ABAQUS was
chosen to complete the task. The 3-D solid eight-node brick layered element [61] was
utilized to avoid the lengthy computation process of modeling each ply with a single
element through the laminate thickness. The package uses internal contact elements
[62] developed to handle the contact between: (i) metal (washers, side plates) and
composite, (ii) metal bolt and composite, (iii) composite and composite, and (iv) metal
bolt and metal plates. Friction between these joint parts was also considered in the
predictions.
The proposed progressive damage model was implemented in a computer code
3DBOLT to interface with ABAQUS/Standard [54]. During the calculations,
ABAQUS passes information regarding strain distribution to the user's subroutine. The
constitutive model in the user's subroutine then calculates the stress distribution based
on the progressive damage analysis. The accumulated damage criteria estimates the
state and extent of damage with given stresses. The constitutive model then degrades
material properties according to the damage modes. Finally, the stresses are passed
back to the finite element solver in ABAQUS for the next load increment. Figure 4.8
shows the analysis scheme.
Two sequential steps were performed in the computer predictions of the bolted joint
tests. In the first step, the metal bolt was pre-strained in order to generate internal axial
tension stress to simulate the clamping condition. This was necessary because as a bolt
is tightened, it is essentially under tension while the washers and side plates are under
out-of-plane compression. An artificial coefficient of axial thermal expansion in the
aJ
bolt, 3 , was used to accomplish the task. The axial thermal strain t:J3 in the bolt
induced by the temperature change can be obtained as follows
(4.14)
where A.T is temperature change in the bolt and should be negative in order to model
the lateral clamping. In the subsequent step, an incremental displacement was applied
on the edge of the composite plate to simulate the loading condition.
The geometries of the composite plates for net-tension and bolt-bearing predictions
are the same as shown in Figure 4.9. Figure 4.10 shows some typical 3-D meshes of
different configurations of bolted joints used in the calculations to simulate the joint
tests. For open-hole and filled-hole laminates, and double-lap bolted joint, only a
quarter of the specimen was simulated due to symmetry. For single-lap bolted joint,
only half of the specimen was simulated. In all cases, one element layer was used
through half the thickness of the multi-directional laminate plates. The load cell
mounted on the side of the fixture was also carefully simulated in some configurations.
Stress Analysis
• 3D stress analysis
. . - - - - - - - - - - P I • Contact
• Calculate deformation &
stress
+
Damage Prediction
Increase loads
or displacement
•Matrix cracking
•Fiber-matrix shearing
•Fiber breakage
Volumetric
Conservation mode •Matrix compression
•Fiber-matrix compression No damage
•Fiber compression
+
Update properties
• Degraded ply stiffuess
•Effective ply strength
+
Joint strength
•Failure load
•Type of failure
•Clamping force
•Etc.
Figure 4.7 Flowchart of the progressive damage analysis
Preprocessing
Mesh, Loading
Material Properties
Commercial Finite Element Package

(ABAQUS)
Figure 4.8 Interaction of user's material subroutine with ABAQUS

w w
Figure 4.9 Geometries of composite plate for tension (left) and bolt-bearing (right) test
4.4 VERIFICATION AND COMPARISON
In order to verity the proposed damage model, the data obtained from the experiments
were compared with the predictions from the 3DBOLT/ABAQUS code. This paper
summarizes only the comparison for the open-hole and filled-hole laminates, the metal-
composite-metal (M/C/M) double-lap joints and the composite-composite (C/C) single-
lap joints. Using the code, the damage progression and failure modes in composites
during loading could also be generated, in addition to the load-deformation curve and
the clamping force history. The visualization of damage progression and a complete
joint response from the predictions could significantly assist engineers to optimally
design bolted composite joints. The material properties used in the simulations for
TS00/3900-2 graphite/epoxy unidirectional prepreg tapes are listed in Table 4.1.
4.4.1 OPEN-HOLE AND FILLED-HOLE LAMINATES
Using the model, numerical predictions of tensile strengths of open-hole and filled-hole
laminates were generated for TS00/3900-2. Typical comparisons for the open-hole and
filled-hole laminates between test data and numerical predictions are shown in Figure
4.11. Figure 4.12 shows the effect of clamping force on the tensile strength of filled-
hole laminates. Test results are denoted as circles, and the predictions are represented
by solid lines in the figures.
a)
1
\ I I
' '. I I
(b)
,_ \t-\-+\~\I-)-\..\-.J\3E} H
Fs' \4 ]
(c)
(d)
(e)
Figure 4.10 Finite element mes
hes for (a) open-hole laminat
lap joint, (d) single-lap joint e, (b) filled-hole laminate, (c)
with one bolt, and (e) single-l double-
ap joint with two-bolts
Chapter 4 Strength Detennination of Mechanical Fastened Joints 119
Table 4.1 Material Properties Used m Numerical Simulation for T800/3900-2

Graphite/Epoxy Prepreg Tapes
Material Properties Symbol (units*) Value
Longitudinal Young's modulus E 11 (msi) 23.2
Transverse Young's modulus E22 (msi) 1.30
Transverse Young's modulus E33 (msi) 1.30
Poisson's ration 0.28
Poisson's ration 0.28
Poisson's ration V23 0.36
In-plane shear modulus G12 (msi) 0.90
Out-of-plane shear modulus G 13 (msi) 0.90
Out-of-plane shear modulus Gn (msi) 0.50
Shear nonlinearity a 11664
Longitudinal tensile strength Xt (ksi) 412
Longitudinal compressive strength Xc (ksi) 225
Transverse compressive strength Yc (ksi) 24
Ply thickness h (in) 0.00645
Mode I fracture toughness G1c (in-lb/in 2) 0.86
Mode II fracture toughness Guc (in-lb/in 2 ) 2.7
Fiber interaction zone (in) o 0.055
* 1 msi=6.9x10 9pa
1 ksi=6.9xl0 6pa
1 in=2.54 em
1 in-lb/in 2= 17.86 m-kg/m 2
Some typical test data are shown in Figure 4.13. The strengths of the filled-hole
specimens were compared with those of the open-hole specimens with the same ply
orientation and geometry for various ply orientations. For this series oftests, the clamp-
up load was 682 kg ( 1500 lbs ). The presence of a bolt with a clamp-up load of 682 kg
affects the notch strength of laminates containing a circular hole. For A, B, C, and D
laminates of Figure 4.13 (Group I}, no reduction of strengths was found for filled-hole
specimens as compared to open-hole specimens. However, for those laminates labeled
E, F, G, H, and I ofFigure 4.13 (Group 2), the notch strengths of filled-hole laminates
were lower than those of the open-hole laminates.
Overall, the results of the predictions show that for the laminates (Group 1) that are
not prone to fiber-matrix splitting and delamination, the model predictions agree well
with the data. However, for those laminates that are prone to tiber-matrix splitting and
delamination, the predictions are not as accurate. The model can underestimate the
notched strength by as much as 20% for some of the open-hole cases. But the
predictions become better for the loading conditions at which the strength of the
laminates are reduced by the lateral clamping force.
200 2$0
[(0190)4)1 [(4510101-4510190)]s
160 200
i I
.-..
120 t i 150 l-. J
i
;§, 80
1
100
40 - - Simulllion 50
1:1} i Elqlorimalt.
J~
0 0
0 10 0 10
2$0
((45~-45mi9M)I90)]1
200 t
ll
B
150 f J Material: TB00/3900-2
D=0.2S inches, WID=4
Washer: Stainless Steel
100 CF: ISOOibs
50
~
0
0 4 10
Washer Size/Hole Diameter (Dw/D)
Figure 4.11 Effect ofwasher size on tensile strength of filled-hole laminates [54]
2$0 250
1(11190)4)1
[(4WIOI~)I
'"" lOO 200

~ I
i 50 150 I
~ 00
1
100
·~
~ 50 - - Sialolllioo
50 D=0.2Sin
zis I illpooiaa DwiD-2

W/0&4
0 0
0 SOD 1000 ISOD 2000 0 SOD 1000 1SOO 2000
Clamping Force (lbs)
Figure 4.12 Effect of clamping force on tensile strength of filled-hole laminates [54]
250
Open-Hole Tension
200 Filled-HoleTeosion (DwiD=2, ISOOibs Clamp-up)

.-.
......
g
-ros:l ISO
g
en
s:l ..
,g
.. ~
f
100
"'~ s
I
·.·.·
..
f~
:::::
.. .
E-< .,
~ ~
~
!:t .re. ::•,
50 ~ ~
N' N
6' ;. ~ ~
~ :
~ ~~ ~ :~ ~ ~ ~"'
:
~
on •,
~ ·. ~ ~ ~ $:::::
:·:·:·
·:·:·:
·.·.·. ::::: ~ +I ••
~·.
...:·
.......,
::::: ~ "'~
0 .•.•, ..
r--
A B c D E F G H I
Group I
~
Laminate Configuration (TS00/3900-2)
Group II ~I
Figure 4.13 Comparison of notch strength between open-hole and filled-hole. 0=0.635 em (0.25
in). (I ksi=6.9xl0 6 pa) [54]
By examining the radiographs of tested specimens, it was found that the failure patterns
of specimens within each group were quite similar, but were significantly different
between the two groups. In Group 1, as shown in Figure 4.14, damage was mostly
confined near the stress concentration areas and propagated from the hole region toward
the free edges as a function of the applied load. The damage consisted mostly of matrix
cracking and fiber breakage. However, in Group 2, damage was widespread over the
laminates as shown in Figure 4.15 due to two additional damage modes: delamination
and fiber-matrix splitting, which were not clearly evident in Group 1. A schematic
description of the damage pattern for each laminate group is shown in Figure 4.16. For
Group 1, damage is localized to the stress concentration areas before final failure. For
Group 2, edge delamination may occur in both open-hole and filled-hole specimens in a
early loading stage. The edge delamination grew throughout the width of the open-hole
specimens. In addition, a fiber-matrix splitting mode occurred along the zero degree
plies (in the direction of the applied load) emanating from the edge of the hole in Group
2 laminates upon loading.
GROUP! Layup: ((±45/0/90/)2]s (%±45-%0=25%)
a b c d e
a,b: Open-Hole Tension; c,d,e: Filled-Hole Tension, Dw/D=2, Finger Tight
Figure 4.14 Damage pattern of the net-tension failure of damage resistant laminates [57]
Figure 4.17 shows the predicted image of damage in the first 0° ply for [(0/±45/90)21,
and the third oo ply for [( 45/0/0/-45/0/0/90/0/0/90)2] 5 open-hole laminates. The
predicted damage pattern resembled the X-radiograph (see Figure 4.14) well for
[(0/±45/90)2 ] 5 • The major dominated failure modes that were predicted were fiber
breakage and matrix cracking. For [(45/0/0/-45/0/0/90/0/0/90h]s, it is worth noting that
extensive fiber-matrix shearing and matrix cracking failures were predicted by the code
before final failure as compared to the corresponding X-radiograph (see Figure 4.15).
Note that no delamination analysis was included in the model. It is believed that the
localized fiber-matrix splitting failure around the hole can considerably reduce the stress
concentration and consequently, enhance the load-carrying capability of the laminates.
Unfortunately, the proposed model, which was based on a continuum mechanics
approach could not take into account the stress reductions due to localized cracking
such as fiber-matrix splitting. A refined model or fracture model would be needed to
address such a local phenomenon.
As shown in Figure 4.12, the clamp-up load can improve or reduce the notch
strength of laminates, depending upon the failure modes and damage progression in the
laminates. Based on the prediction, it was found that friction was the primary factor
contributing to the increase in failure loads for those laminates which are not prone to
fiber-matrix splitting. However, the tensile strength could be reduced, by adding clamp-
up, to suppress the fiber-matrix splitting mode for those filled-hole laminates that are
prone to delamination and fiber-matrix splitting.
GROUP2 Lay-Up: [45/0/0/-45/(0/0/90/)2]s (%±45-%0 = -40%)
a b c d e f
a,b,c: Fill ed-Hole, Dw/0 =2, 3000 lbs Clamp-Up; d,e,f: Open-Hole Tension
Figure 4. 15 Damage pattern of the net-tension failure of damage prone laminates [57]
FAILURE MODES
FIBER BREAKAGE DELAMINATION FIBER-MAlRIX SPLITTING

MA1RIX CRACKING
GROUP 1 GROUP2
Figure 4.16 Typical tensi le failure modes in composite laminates containing a circular hole [57]
TS00/3900-2 TS00/3900-2
[(0/±45/90)2], [ (45/0/0/45/0/0/90/0/0/90) ],
0 =0.25 inches 0 =0.25 inches
W/D=4 W/D=4
Figure 4.17 The predicted image of damage in the first 0° ply for [(0/±45/90) 2] , (left) and the third
0° ply for ((45/0/0/-45/0/0/90/0/0/90)2], (right) open-hole laminates (D=0.25 inches, W/D=4) [54]
4.4.2 METAL-COMPOSITE-METAL DOUBLE-LAP JOINT
The comparison of load-deformation response between test data and numerical

predictions for M/C/M double-lap joints under a fully clamped condition (no washers)
is shown in Figure 4.18. The circles indicate test data, while the solid line represents the
numerical prediction. No significant load drop was found. For the cross-ply laminate,
the load reached a plateau after excessive deformation. However, the load continued
increasing gradually for the [(0/±45/90) 3] , laminate. The calculated clamping force
history as a function of applied load for both joints is also presented in Figure 4.19. The
predicted clamping force agreed very well with test data. Figure 4.20 shows the
comparisons of failure load between test data and predictions for double-lap joints with
different washer sizes. The failure load was determined by the first occurrence of either
a significant load-drop or offset value at 8% of bolt diameter.
4. 4. 3 COMPOSITE-COMPOSITE SINGLE-LAP JOINT
Figure 4.21 shows the comparison of the load-deformation curves between the
predictions and test data. The circles indicate test data. The solid line represents the
numerical prediction where the bolt was considered to be elastic, and the dash line
represents the prediction, which considered the bolt to have an elastic-plastic behavior.
The reason for consideration of the plastic deformation for bolts was because permanent
bolt bending was found from some experiments upon unloading the single-lap joints.
The constitutive equation of the elastic-plastic bolt was modeled based on the Mises
yield surface model and the isotropic hardening rules [63].
12000 12000
T800/3900-2 ((0/±45/90h],
[(0/90)6];
9000 9000
0=0.25 inches
~
"' W/D=S,E/0=6
'0 CF:800 lbs
"
.2 6000
0000
6000
'0
.!o!
c..0. 3000
3000
~ 0 Test Data
Simulation
0.04 0.08 0.11 0.16 0.2 0.16 0.2
Displacement (in)
Figure 4.18 The comparisons of load-displacement curves for the [(0/90)6], and [(0/±45/90) 3],
composite double-lap joints under fully clamped condition (no washers) [54]
12000 !2000
TS00/3900-2 [(0/±45/903];
9000 [(0/90)6]; 9000
-
r
""
@. 0=0.25 inches ooo 0
'0
~
W/0=8, E/0=6
2 6000 6000
000
'0
.~
0. 0
0.
~
g
<( 3000 3000
0 rest data
Simulation
0
~
0
0 400 800 1200 1600 2000 0 400 800 1200 1600 2000
Figure 4.19 The comparisons of applied load-clamping force curves during the course of loading
[54]
The prediction that considered plastic deformation of the bolt showed much better
agreement with the experimental data than the simulation that considered only elastic
deformation for the bolt. Figure 4.22 shows the comparisons of failure load between
test data and predictions for single-lap joints with different initial clamping loads. The
failure load was determined by the offset value at 8% of bolt diameter.
4.5 EFFECT OF THE CLAMPING
As shown in the previous sections, lateral clamping can affect the strength of bolted
composite joints. Using the computer code, both the clamped area and the clamping
force were evaluated to determine the degree of their influence on the strength of bolted
joints. The material system used for this study was TSOOH/3900-2 graphite/epoxy. The
failure load was determined by the first occurrence of either a significant load-drop or
offset value at 8% of bolt diameter.
·~~---------------------------.
TS00/3900-2 0 Testda!a
MICJM
((01±45190),), 1SJ Simulation
7500 0=0.25 inches.WID---8, EID=6
...... CF: 800 lbs
"'
~
1 5000
~
~ 2500
Pin joint Dw/0=2 Dw/0=3 Fully clamping
Washer Size
Figure 4.20 Comparisons of failure load between test data and simulations for double-lap joint>
with different washer sizes [54]
8000
T800/3900-2
[(0/±45/90)21< I
6000 D- 0.25 inche
,.-._
WfD- 8, E/D- 6
"'
..0
:::::.- CF: OOlbs
'"0
4000
..s
~
'"0
-~ 0 Test data
0.. 2000
0.. Sim ulat ion (ela tic bolt)
<!;
Simu lation (p lastic bo lt)
0
0 0.05 0.1 0.15 0.2 0.25
Displacement (inch)
Figure 4.21 Load-deflection curves for [(0/±45/90) 2] , single-lap bolted composite joints [54]
4.5.1 EFFECT OF THE SIZE OF CLAMPED AREA
Figure 4.23 shows the effect of washer size on the strength of M/C/M double-lap joints.
The joint geometry was D=0.635 em (0.25 inches), W/D=8, E/D=6. The width of the
joints was particularly selected to ensure that failure would occur in a pure bearing
mode. Clearly, as the washer diameter-to-hole diameter (Dw/D) ratio was greater than
or equal to 3, the joint strength reached a plateau for both the finger-tight and the 363 kg
(800 lbs) clamping force conditions. The "finger-tight" condition was equivalent to a 23
kg (50 lbs) clamping force. The results strongly suggest that a minimum value ofDw/D
greater than 3 shall be recommended for bolted composite joints which may fail in a
bearing mode.
~~.-------------------------~
T80013900-2
[(0/±45190)J. C1C
~ 7500 D=0.25 inches
W/D=8, EID=6
@_
~
E3 Testdata
,.g 5000
!SJ Simulation
]
&! 2500
Finger tight 400 800
Initial clamping force (lbs)
Figure 4.22 Comparisons of failure load between test data and simulations for single-lap joints
with different initial clamping force [54]
As demonstrated in the previous sections, clamping can reduce the notch strength of
filled-hole laminates under uniaxial tensile load. The filled-hole tension condition is
normally referred to as the I 00% bypass condition in bolted joint tests, while the bolted
joint test that has been performed in the study is typically referred to as the 0% bypass
condition. Since bolted joints (at 0% bypass) could also fail in a tension mode similar
to that of filled-hole laminates (100% bypass), it was desired to determine if the strength
reduction in filled-hole laminates due to the clamp-up would also appear in bolted joints
which fail in a tension mode.
Accordingly, numerical predictions were performed for both filled-hole laminates
and bolted joints as shown in Figure 4.24. The width of the joints was particularly
selected to ensure the joints would fail in a tension mode. The geometry of both
double-lap joint and filled-hole laminate was D=0.635 em (0.25 inches), W/D=2.5.
Tension failure was the failure mode in both the double-lap joint and the filled-hole
laminate.
The results of the study clearly showed that unlike filled-hole laminates, bolttxl
joints (0% bypass) continue to benefit from the clamp-up in a tension mode, although
filled-hole laminates produced higher failure loads than bolted joints. At a Dw/D ratio
greater than 3, the joint strength reached a plateau. Overall, it is recommended that a
Dw/D ratio greater than or equal to 3 be used for bolted joint design.
10000
TS00/3900-2
[(0/±45/90),],
0=0.25 inches, WII}.;8, EID=6
7500 Bearing Failure Mode
~--··················
5000
- CF:800lbs
2500
-.. CF:Fr
::e
rl)
_.
-g 0
0
....:l 10000
G)
.E
·~
[(0/90),],
~ 7500
5000
2500
~--------------------
'
0
0 2 4 6 8 10
Dw/D
Figure 4.23 Simulation for effect of washer size on strength of double-lap joints (bearing is
dominant failure mode) [54]
200CO~----------------------~
[(0/±45190),]s FilledHo1e [(45/0/0/-45/010/90/0/0190)]s
16000 Double-lap
12000
··-------------·--··
--------------
8000
"'"'
"'
:9
'-'
4000 /
~
0 0
~
] 200CO
[(45/0/0/-45/0/90/0/90)]s
·;
16000
"'"' Tension Failure Mode
~eri&:TS00/3~2
Filled Hole Laminates
12000 D=0.25 inches, WID=2.5
Clamping Force: 1500 lbs
8000 Double-lap Joint
4000
---------------· D=0.25 inches, WID=2.5
Clamping Force: 1500 lbs
/
0
0 2 4 6 8 10
Dw/D
Figure 4.24 Simulation for effect of washer size on strength of double-lap joints and filled-hole
laminates (fiber breakage is dominant failure mode) [54]
4.5.2 EFFECT OF THE iNITIAL CLAA4PING FORCE
The effects of the initial clamping force were predicted for both [(0/±45/90) 3]s double-
lap joints and [(0/±45/90)2], single-lap joints. The geometry of both double-lap and
single-lap joints was D=0.635 em (0.25 inches), W/D=8, E/D=6. In the predictions of
double-lap joints, no washers were inserted between the metal side plates and the center
composite plate. Various levels of initial clamping force were applied. The "finger-
tight" condition was equivalent to a 23 kg (50 lbs) clamping force again. Based on
experiments [57], a value of 0.13 was selected as the coefficient of friction between the
composite and steel plates while a value of 0.36 was selected as the coefficient of
friction between composite plates.
Figures 4.25 and 4.26 show the effect of the initial clamping force on the failure load of
double-lap and single-lap bolted composite joints, respectively. The failure load of the
double-lap pin joint without lateral supports was also plotted for comparison. Clearly,
the study showed that the initial clamping force had little effect on the failure load as
long as at least a "finger-tight" initial clamping force was applied.
1~ ,---------------------------~
TS00/3900-2
[(01±45/90)3],
-;;;- 8000 0=0.25 inches, WID=8
f--------------------~~
,D
c
~6000
j
~inger-tigbt
~ 4000
·o;
~ 2000 ~n joiDt
0+-----~----r---~~--~----~
0 200 400 600 800 1000
Figure 4.25 Simulation for effect of amount of initial clamping force on failure load of double-lap
bolted composite joints [52]
1~
TS00/3900-2
~
[(0/±45/90h),
.-.. 8000
0=0.25 inches, WID=8
"'
,D
c
'0 6000
j"' ( Finger-tight
!::! 4000 E/0=6
::>
~
~
2000
0
0 200 400 600 800 1000
Figure 4.26 Simulation for effect of amount of initial clamping force on the failure load of single-
lap bolted c..>mposite joint [54]
Additional parametric study, such as effect of joint geometry and effect of number of
holes, and procedures for design of bolted composite joints can be found in references
[52-54].
4.6 COMPOSITE JOINTS UNDER COMBINED BEARING AND AXIAL-BYPASS

LOADS
4.6.1 VERIFICATION AND COMPARISON
Based on the two-dimensional material degradation model, a computer code designated

as BJSTAT was developed for analyzing the response of bolted composite joints under
multiple bypass loads [56] . The progressive failure model is also verified with the
ultimate strength of joints under combine bearing and axial-bypass loads.
For joints under bearing and axial-bypass loads, two types of loading configurations
were previously defined in the literature [20]: tension-reacted bearing (TRB) and
compression-reacted bearing (CRB). The notations for both configurations are shown in
Figure 4.27. In both cases, a uniform load Pgross is applied to the bottom edge of the joint
and a uniaxial bypass load P\:rass is reacted on the top of the joint. The reacted bypass
load is in the opposite direction to the applied gross load, and the ratio between the two
is kept constant during the loading process. The bearing load carried by the fastener can
be determined from the difference between the two loads.
~ ......,
Pgross Pgross
Geometry Tension-Reacted Compression-Reacted

Bearing (TRB) Bearing (CRB)
Figure 4.27 Illustration of the two configurations for combined bearing-bypass loading
The comparison between the predictions and the data obtained from the Boeing
Company for T800H/3900-2 bolted composite joints under a combined bearing and
axial-bypass load is presented in Figure 4.28. The same testing configuration was used
as the one proposed by Naik and Crews [20]. The fastener diameter was 0.952 em
(0.375 inches) and the washer diameter was 1.636 em (0.644 inches). The ply
orientation of the laminate was [45/90/-45/0/45/90/45/02/45/-45/0 2/-45/90/45/0/-
45/90/45h, which is an unsymmetric layup. Due to the limitation of current analysis.
asymmetric layup, [45/90/-45/0/45/90/45/0i45/45/0 2/-45/90/45/0/-45/90/45]r wa~
chosen to estimate the response of the joint
In Figure 4.28, the bearing stress (PbearingiDH) was normalized with respect to the
bearing strength of the joint under a uniform gross load only, and the axial-bypass stre~s
(P\ypass /WH) was normalized by the filled-hole tension strength of the joint. The filled
hole tension condition was defined when the gross load and the axial-bypass load wer~
both in tension and had the same magnitude. In addition, bearing stress and axial-bypass
stress are defined as positive if the direction ofloading is upward.
Data from Boejne Co. Nonnalized

TSOOH/3900-2 Axial-Bypass Stress
D=0.375" Dw=0.644"
WID=S EID=S UD=IO
FHT
1
Nonnali.zed
-1 Bearing Stress
--~----.-----~----~----~--
0 1
e Data- TRB
-1 0 Data- CR.B
-Model
FHC
Figure 4.28 Comparison between the predicted and the measured bolt bearing strength of joints
under combined bearing and axial-bypass loads. [56)
Under compression-reacted bearing (CRB), both the data and the predictions indicate
that the joint strength decreases in a nonlinear fashion as the compressive bypass stress
increases. Based on the model, it was found that the fastener-hole contact region was
smaller resulting from the compressive loads acting in opposite directions [55]. This
decrease in the contact angle was believed to be the major reason causing the reduction
in compression-reacted bearing strength [20]. The predicted damage progression of a
joint under TRB and CRB could be found in reference [56].
4.6.2 STRENGTH ENVELOPE OF A JOINT UNDER UNIAXIAL LOADS
Based on the progressive model, predictions on the strength and the types of failure of a
bolted composite joint under all possible uniaxial loading conditions were shown in
Figure 4.29. The joint was under the finger-tight condition and the material system
chosen for the simulation was T800H/3900-2 composites. The ply orientation of the
laminate was [(90/±45/0h]s and the geometry ofthejoint was 0.635 em (0.25 inches) as
the hole diameter and 1.27 em (0.5 inches) as the washer diameter. The width to hole
diameter WID and the end margin to hole diameter E/D ratios are 8 and 4, respectively.
$.
TSOOH/3900·2 0 =41.25" Dwf0=2 - - NET-TENSION
((901±4SIO)l]s Wf0=8 E/D=-4 Lf0&8 - BEARING
,,.., Normallud
Axlai· Bypass Stress 1p41•
k<lri"f •ui•J
1.0 FliT
''061 <0> ...................
................... ..._ troll
~
T . r - = - -...__- + 0
_-1"""'"1 - _T_RB.._ _ _ :®~
: l.O Normalized
Bearine Stress
CRB
$ -1.0 @
Figure 4.29 Predicted strength envelope of a bolted composite joint under combined bearing and
axial-bypass loads [56]
As mentioned before, the bearing stress and the axial-bypass stress are normalized with
respect to the bearing strengths of the joint under a uniform gross load only and a filled
hole tension strength, respectively. Both bearing stress and axial-bypass stress are
defined as positive if they are upward.
In the figure, the bearing strength of the joint under a single tensile load is indicated
by point A and the strength under a single compressive load is indicated by point F,
which is slightly lower than A. The axial-bypass stress at point C represents the filled
hole tension strength of the joint and point G indicates the filled hole compression
strength. The current analysis may not be able to predict the filled hole compression
strength due to the possible failure mode which is not considered in this investigation.
Accordingly, simulation was not performed for this loading configuration.
Under tension-reacted bearing (A-B-•••-E), the predicted bearing strength of the
joint decreases linearly with the increase in the bypass stress and the failure mode of the
joint changes from bearing to net-tension. Unlike TRB, bearing strength of the joint
under CRB (F-G-H) decreases nonlinearly with the increase of the bypass stress. As
discussed before, this nonlinear reduction may be attributed to the decrease in the bolt-
hole contact angle caused by the compressive loads acting in opposite directions.
At point E, the joint is only loaded by the bypass load in tension and the normalized
bearing stress is equal to one. Between E and F, the uniaxial loads are acting upward
and the negative bearing stress changes from TRB due to the axial-bypass load (point E)
to CRB due to the compressive gross load (point F). A similar transition in the joint
strength is shown from TRB at point A to CRB at point H where the uniaxial loads are
all acting downward and the resulting bearing stress is positive. For a given joint
geometry and ply orientation of the laminates, the corresponding strength envelope can
be constructed using the computer code. Accordingly, the interaction effects between
the bearing and axial-bypass loads on the joint response can be studied.
4.6.3 EFFECTS OF TRANSVERSE-BYPASS LOAD
For the same joint configuration, predictions on the bearing strength under biaxial
bypass loads are shown in Figure 4.30. Similar to the axial-bypass stress, the transverse-
bypass stress is also normalized with respect to the filled hole tension strength of the
joint. The tensile transverse-bypass load was applied uniformly to both lateral edges of
the joint in equal magnitude but opposite directions. During the incremental loading
process, the transverse-bypass load to gross load ratio (P \ypas/P gross) was kept constant
until the desired level of transverse-bypass load was reached. After the desired level was
reached, the gross load was increased continuously up to the final failure of the joint
while the transverse-bypass load remained fixed.
In summary, the calculations seemed to indicate that, at the bearing failure mode, the
joint strength decreased with the increase of the transverse-bypass stress. This effect is
most noticed under the CRB condition as indicated by F-G-H in Figure 4.30. However,
the transverse-by stress may also increase the bearing strength at a given axial-bypass
load under the TRB condition, if the joint fails in the net-tension mode as shown by B-
C-D in the figure.

Note that the computer code 3DBOLT/ABAQUS can also be used to study the
response of bolted composite joints under multiple bypass loads.
T800HJ3900-2 0 : 0.25" Dw/D=2

[(901±45/0)J]s W/D--8 E/0=4 UD:8
Nonnallnd
Axial-Bypass Stres.s
Normalized
~~~----r-------4-------~----~~- BearingS~
0
-1.0 <OJ
Normllliud Transverse-Bypass Stress = 0

Normalized Transverse-Bypass Stress= 0.667
Figure 4.30 The predicted strength envelopes of a bolted composite joint under combined bearing
and multiple bypass loads [56]
4.7 CONCLUSION
An experimental and analytical investigation was presented to study the response of

bolted composite joints. The failure mechanisms and failure modes in laminated
composites were characterized. A progressive damage model was developed to predict
the strength and failure response of bolted composite joints.
A computer code 3DBOLT/ABAQUS has been developed. The constitutive model
implemented in the code includes a progressive damage analysis for the damaged
material and a volume conserving technique that accounts for the incompressible
behavior of damaged material due to bearing failure inside the bolted joints. The
computer code can be used for characterizing the response of the laminates for the
following configurations: (i) open-hole tension, (ii) filled-hole tension, (iii) double-lap,
joints with and without lateral clamp-up loads, and (iv) single-lap joints with and
without lateral clamp-up loads
It has also been demonstrated that the progressive model is capable of predicting the
strength and the failure mode of joints under combined bearing and axial-bypass loads.
It is important to point out that the failure of bolted composite joints is still not
completely understood yet, but many problems can be solved by the computer code
3DBOLT/ABAQUS.
4.8 ACKNOWLEDGEMENT
The support of the Boeing Commercial Airplane Group and the Federal Aviation
Administration for the project is gratefully appreciated.
4.9 REFERENCES
1. Hart-Smith, L. J., Bolted Joints in Graphite/Epoxy Composites, NASA Cr-144899,

National Aeronautics and Space Administration, Washington, DC, January 1977.
2. Agarwal, B. L., Static Strength Prediction of Bolted Joints in Composite Material,
AIAA/ASME/ASCE/AHS. 20th Structure, Structural Dynamics and Material
Conference, St. Louis, MO, April, 1979, pp. 303-309.
3. Oplinger, D. W., and Gandhi, K. R., Stresses in Mechanically Fastened Orthotropic
Laminates, Proceedings of The Conference on Fibrous Composites in Flight
Vehicle Design, Dayton, Ohio, 21-24 May, 1974, Air Force Flight Dynamic~
Laboratory Report No. AFFDL-TR-74-1 03, pp. 811-842.
4. Oplinger, D. W., Analytical Studies of Structural Performance in Mechanically
Fastened Fiber-reinforced Plates, Proceedings of The Army Symposium on Solid
Mechanics, 1974: The Role of Mechanics in Design - Structural Joints. Army
Materials and Mechanics Research Center, Watertown, Mass., Report No. AMMRC
MS 74-8, pp. 211-42.
5. Garbo, S. P. and Gallo, R. L., Strength of Laminates with Loaded Holes, presented
at The 5th DOD/NASA Conference on Fibrous Composites in Structural Design,
New Orleans, LA, January, 1981.
6. Matthews, F. L., Joining Fiber-reinforced plastics, London, Elsevier Applied
Science, 1987.
7. Chang, F. -K., Scott, R. A. and Springer, G. S., Failure of Composite Laminates
Containing a Pin Loaded Holes - Method of Solution, Journal of Composite
Materials, Vol. 18, 1984, pp. 255-278.
8. Chang, F. -K., Scott, R. A. and Springer, G. S., Failure Strength of Nonlinear
Elastic Composite Laminates Containing a Pin Loaded Holes, Journal of
Composite Materials, Vol. 18, 1984, pp. 467-477.
9. Waszczak, J. P. and Cruse, T. A., Failure Mode and Strength Predictions of An

isotropic Bolt-bearing Specimens, Journal of Composite Materials, Vol. 5, 1971,
pp. 421-425.
10. Soni, S. R., Stress and Strength Analysis of Bolted Joints in Composite Laminates,
Composite Structures, I. H. Marshall, Ed., Applied Science Publishers, New Jersey,
1981 ,pp. 50-62.
11. Soni, S. R., Failure Analysis of Composite Laminates With a Fastener Hole, Joining
of Composite Materials, Special Technical Publications 749. Published by ASTM,
Philadelphia, PA, USA, pp. 145-164.
12. Tsujimoto, Y. and Wilson, D., Elasto-Plastic Failure Analysis of Composite Bolted
Joints, Journal ofComposite Materials, Vol. 20, 1986, pp. 236-252.
13. Wilson, D. W. and Tsujimoto, Y., On Phenomenological Failure Criteria for
Composite Bolted Joint Analysis, Composite Science and Technology, Vol. 26,
1986, pp. 283-305.
14. Lessard, L. B. and Shokrieh, M. M., Two-dimensional Modeling of Composite
Pinned-joint Failure, Journal of Composite Materials, Vol. 29, No.5, 1995, pp. 671-
697.
15. Eriksson, L. I., On The Bearing Strength of Bolted Graphite/Epoxy Laminates,
Journal ofComposite Materials, Vol. 24, 1990, pp. 1246-1269.
16. Eriksson, L. 1., Contact Stresses in Bolted Joints of Composite Laminates,
Composite Structures, Vol. 6, 1986, pp. 57-75.
17. Chiang, Y. J. and R. E. Rowlands, Finite Element Analysis of Mixed-mode Fracture
of Bolted Composite Joints, Journal of Composite Technology Research, Vol. 13,
No.4, Winter 1990, pp. 227-235.
18. Tsai, M. Y. and Morton, J., Stress and Failure Analysis of a Pin-loaded Composite
Plate: An Experimental Study, Journal of Composite Materials, Vol. 24, 1990, pp.
1101-1120.
19. Chang, F. -K. and Chang, K. -Y., Post-Failure Analysis of Bolted Composite Joints
in Tension or Shear-out Mode Failure, Journal of Composite Materials, Vol. 21,
1987, pp. 809-827.
20. Crews, J. H., and Naik, R. A., Combined Bearing and Bypass Loading on a
Graphite/Epoxy Laminate, Composite Structures, Vol. 6, 1986, pp. 21-40.
21. Hart-Smith, L. J., Mechanically-Fastened Joints for Advanced Composite-
Phenomenological Considerations and Simple Analysis, Douglas paper 6748,
1978, 1-32.
22. Hart-Smith, L. J., Bolted Joints in Graphite/ Epoxy Composites, NASA Cr144899,
National Aeronautics and Space Administration, Washington, DC, January, 1977.
23. Smith P. A., Pascoe K.J., Polak C. and Stroud D.O., The Behavior of Single-lap
Bolted Joints in CFRP Laminates, Composite Structures, Vol. 6, No. 1-3, 1986, pp.
41-55.
24. Nekrasov Yu. A., A Study of The Strength of Bolted Joints Between Metal and
Composite Plates, Soviet Machine Science, No.3, 1985, pp. 87-91.
25. Stockdale, J. H. and Matthews, F. L., The Effect of Clamping Pressure of Bolt
Bearing Loads in Glass Fiber-Reinforced Plastics, Composites, Vol. 7, 1971, pp.

34-38.
26. Crews, J. H., Bolt-Bearing Fatigue of a Graphite/Epoxy Laminates, Journal of
Composite Materials, ASTM STP 749, K. T. Kedward, Ed., American Society for
Testing and Materials, Philadelphia, PA, 1981, pp. 131-144.
27. Matthews, F. L., Roshan, A. A. and Philips, L. N., The Bolt Bearing Strength of
Glass/Carbon Hybrid Composites, Composites, Vol. 13, 1982, pp. 225-227.
28. Collings, T. A., On The Bearing Strength ofCFRP Laminates, Composites, Vol. 13,
1982, pp. 241-252.
29. Collings, T. A. and Beauchamp M. J., Bearing Deflection Behavior of a Loaded
Hole in CFRP, Composites, Vol. 15, No. 1,1984, pp. 33-38.
30. Cooper, C., and Turvey, G. J., Effects of Joint Geometry and Bolt Torque on The
Structural Performance of Single Bolt Tension Joints in Pu1truded GRP Sheet
Material, Composite Structures, Vol. 32, 1995, pp. 217-226.
31. Graham, U., Winsom, M. R. and Webber, P.H., A Novel Finite Element
Investigation of The Effects of Washer Friction in Composite Plates With Bolt-
filled Holes, Composite Structures, Vol. 29,1994, pp. 329-339.
32. Herrington, P. D., and Sabbaghian M., Fatigue Failure of Composite Bolted Joints,
Journal of Composite Materials, Vol. 27, No.5, 1993, pp. 491-512.
33. Akay, M., Bearing Strength of As-cured and Hydrothermally Conditioned Carbon
Fiber/Epoxy Composites Under Static and Dynamic Loading, Composites, Vol. 23,
No.2, March 1989, pp. 101-108.
34. Shahid, I. and Chang, F.-K., An Accumulative Damage Model for Tension and
Shear Failure of Laminated Composite Plates, J. of Composite Materials, 1995, 29
(8), 926-981.
35. Shahid, I. and Chang, F.-K., Predicting Scaling Effect on the Notched Strength of
Prepreg and Fiber Tow-Placed Laminated Composites, J. of Composite Materials,
1995,29 (8), 1063-1095.
36. Wang, H. -S. Hung, C. -L. and Chang, F.-K, Bearing Failure of Bolted Composite
Joints. Part I: Experimental of Characterization, Journal of Composite Materials,
Vol. 30, No. 12, 1996, pp. 1248-1313.
37. Hung, C. -L. and Chang, F. -K, Bearing Failure ofBolted Composite Joints. Part II:
Model and Verification, Journal of Composite Materials, Vol. 30, No. 12, 1996, pp.
1359-1400.
38. Morgan, M. E. and Beckwith, S. W., Bolt Torque Loading and Radial Gap Effects
on Thick-wall Composite Joint Strength, Proceedings of The 30th National SAMPE
Symposium, March 19-21, 1985, pp. 1321-1334.
39. Hom, W. J. and Schmitt, R. R., Influence of Clamp-up Force on The Strength of
Bolted Composite Joints, AIAA Journal, Vol. 32, No.3, 1993, pp. 665-667.
40. Smith, P. A., Ashby, M. F. and Pascoe, K. J., Modeling Clamp-up Effects in
Composite Bolted Joints,Journal of Composite Materials, Vol. 21, No. 12, 1987, pp.
878-897.
41. Camanho, P. P. and Matthews, F. L., Stress Analysis and Strength Prediction of
Chapter 4 Strength Detennination of Mechanical Fastened Joints 139
Mechanically Fastened Joints in FRP: a review, Composites Part A, Vol. 28A,

1997, pp. 529-547.
42. Matthews, F. L., Wong, C. -M. and Chryssafitis, S., Stress Distribution Around a
Single Bolt in Fiber-reinforced Plastic, Composites, Vol. 13, 1986, pp. 316-322.
43. Chen, W. -H., Lee, S. -S. and Yeh, J. -T., Three-dimensional Contact Stress
Analysis of a Composite Laminate With Bolted Joint, Composites Structures, Vol.
30, 1995, pp. 287-297.
44. Marshall, I. H., Arnold, W. S. and Wood, J., Observations on Bolted Connections
in Composite Structures, Composites, Vol. 13, 1989, pp. 133-151.
45. Hyer M. W., Klang, E. C. and Cooper, D. E., The Effects of Pin Elasticity,
Clearance and Friction on The Stresses in a Pin-loaded Orthotropic Plate. Journal of
Composite Materials, Vol. 21, 1987, pp. 190-206.
46. Barboni, R., Gandenzi, P. and Carlini, S., A Three-dimensional Analysis of Edge
Effects in Composite Laminates With Circular Holes, Composite Structures, Vol.
15, 1990, pp. 115-136.
47. Iarve, E., Stress Analysis in Laminated Composites With Fastener Holes,
Proceedings of The American Society for Composites I Oth Technical Conference,
Technomic Publishing Co., Lancaster, PA, 1995, pp. 408-419.
48. Iarve, E., Three-dimensional Stress Analysis in Laminated Composites With
Fasteners Based on The B-spline Approximation, Composites Part A, Vol 28A,
1997, pp. 559-571.
49. Hassen, N. K., Mohamedien, M.A. and Rizkalla, S. H., Finite Element Analysis of
Bolted Connections for PFRP Composites, Composites Part B, Vol 27B, 1996, pp.
339-349.
50. Lessard, L. B., Compression Failure in Laminated Composites Containing an Open
Hole, Ph.D. dissertation, Stanford University, 1989.
51. Sun, H.T., Chang, F.-K. and Qing X.L., The Response of Composite Joints with
Bolt-Clamping Loads, Part I: Model Development, Journal of Composite Materials,
Vol. 36, No. I, 2002, pp.47-68.
52. Sun, H.T., Chang, F.-K. and Qing X.L., The Response of Composite Joints with
Bolt-Clamping Loads, Part I: Model Verification, Journal of Composite Materials,
Vol. 36, No. I, 2002, pp.69-92.
53. Fu-Kuo Chang, Xinlin Qing, Hsien-Tang Sun and Yanmin Yan, Damage tolerance-
based design of bolted composite joints, Final Report to the Boeing Commercial
Airplane Groups and the Federal Aviation Administration, Aeronautics and
Astronautics Department, Stanford University, 1999.
54. Qing, X., Sun, H.-T., Dagba, L., and Chang, F.-K., , Damage tolerance-based
design of bolted composite joints, Composite Structure Theory and Practice, ASTM
STP 1383, P. Grant, Ed., American Society for Testing and Materials, West
Conshohocken, PA, 2000, pp.243-272.
55. Hoa, S.V., Lucking, W.M. and Lefebvre, D., Biaxial Testing of Bolted
Graphite/Epoxy Plates, ICCM, Madrid, 1993.
56. Hung, C.L. and Chang, F.K., Strength Envelope of Bolted Composite Joints under
Bypass Loads, J. of Composite Materials, 30 (13), 1996, pp.1402-1434.

57. Yan, Y., Wen, W.-D., Chang, F.-K., and Shyprykevich, P., Experimental study on
clamping effects on the tensile strength of composite plats with a bolt-filled hole,
Composite: Part A 30, 1999, pp.1215-1229.
58. Hahn, H. T., Nonlinear Behavior of Laminated Composites, Journal of Composite
Materials, Vol. 7, 1973, pp. 257-271.
59. Zienkiewicz, 0. C., The Finite Element Method, 4th ed., McGraw-Hill, 1989.
60. Bathe, K. J., Finite Element Procedures in Engineering Analysis, Prentice-Hall,
Inc., New Jersey, 1982.
61. Hibbit, Karlsson & Sorensen Inc., ABAQUS/Standard Theory Manual, Hibbit,
Karlsson & Sorensen Inc., Pawtucket, Rhode Island, USA, 1995.
62. Hibbit, Karlsson & Sorensen Inc., Writing UMATs and Vumats, Volume II, Hibbit,
Karlsson & Sorensen Inc., Pawtucket, Rhode Island, USA, 1995.
63. Donald C. Stouffer and L. Thomas Dame, Inelastic Deformation of Metals, models,
mechanical properties and metallurgy, JOHN WILEY & SONS, INC, 1996.
CHAPTER 5 STRENGTH ANALYSIS OF ADHESIVELY BONDED REPAIRS
C. SOUTIS"" and F.Z. HU

Imperial College of Science. Technology & Medicine
Prince Consort Road. London SW7 2BY, UK.
5.1 INTRODUCTION
The use of fibre reinforced plastics has grown in recent years in aerospace, automotive
and marine structures. The maintenance and repair of these structures are always of
main concern to the end-users as well as the manufactures. On an aircraft, the need to
repair a component can arise from physical damage due to accidental impact, bird
strike, hailstones and lightning strike or from deterioration caused by the absorption of
moisture or hydraulic fluid [I, 2]. Once a defect has been found and it falls outside
approved allowable limits, one of the following actions should be taken according to its
extent and the constraints of operational conditions: temporary repair, permanent repair
or replacement. The temporary repair (field repair or emergency repair) is usually a
preferred action when the operational limitations are severe and the temporary repair is
a safe option. The limitations are usually time, environment and facilities. Temporary
repair aims to restore the mechanical strength required to permit aircraft operation until
a permanent repair or replacement can be carried out.
Adhesively bonded repair is the most common type of repair carried out with
composite materials, because load transfer paths through mechanical fasteners can cause
local overloads and damage in the relatively brittle composites. Two types of bonded
repairs are mainly used; the highly refined and structurally efficient but expensive flush
patch repair and the external metal or composite patch [3-6]. In the external patch
technique, the damaged material is removed by cutting a hole, the parent plate is then
cleaned and applied with filler and adhesive materials before the patches are attached. In
practice, the repair patch would always be tapered at the edges to reduce peel stresses.
• Corresponding author. c.soutis@ic.ac.uk.
141
L. Tong and C. Soul is (eds.).

Recent Advances in Structural Joints and Repairs.for Composite Materials. 141-172.
142 C. SOUTIS & F. Z. HU
This type of repair could also be used for a permanent repair in lightly-loaded and
relatively thin structures.
Flush, scarf-type, bonded repairs are used on critical, highly loaded, components
where load concentration and eccentricities, especially for compressive loading, must be
avoided. Thick monolithic structures lend themselves to such repairs since external
patch would cause excessive out-of-mouldline thickness and unacceptably high
bondline peel and shear stresses [7, 8]. In all these repair methods the main concerns arc
the quality of surface preparation prior to bonding, the design of the composite patch
and the prediction ofboth strength and durability of the repaired configuration.
In this chapter, the compressive behaviour of bonded (external and scarf-type)
repairs is discussed. Some recent analytical and numerical studies are presented to help
site-engineers to define the repair configurations (patch size, shape, stiffness) and
predict the efficiency of the repairs.
5.2 EXTERNAL PATCH REPAIRS
The external patch technique is simpler to apply and less critical in nature than a scarf
type repair. In this approach the load is taken over and around the damaged area. Any
bending strains due to eccentric load path must be considered in the patch design. The
patch must also be capable of withstanding the high peel and shear stresses, which
develop at the edges of the overlap. The specimen configuration examined here is an
idealised case that simulates a situation where the impact damaged laminate has been
repaired by drilling a hole and then external patches are attached on one or both sides of
the plate, as shown in Figure 5.1.
Figure 5. 1 External patch repair

Chapter 5 Strenth Analysis of Adhesively Bonded Repairs 143
As a preliminary analysis, this repair may be modelled as a single or double strap joint,
and the so called 'shear lag model' [4, 9] could be used together with the maximum
shear strain criterion in the adhesive, to identifY the optimum patch size, thickness and
elastic modulus. This model can also examine the effect of the adhesive thickness and
shear properties on the joint strength.
5.3 DESIGN OF BONDED JOINTS
The primary purpose of a joint is to transfer load between the two items being joined.
In most bonded joints the load transfer takes place through interfacial shear. The
interfacial shear gives rise to high interlaminar stresses in the adhesive layer.
Knowledge of the stresses in joints is vital if we are to understand the failures that occur
in practice and hence improve designs and predict strength. Even relatively simple
theories can be useful if they allow the important parameters to be identified.
5.3.1 'SHEA R L AG' STR ESS A NALYSIS
Consider a double lap joint as shown in Figure 5.2. The bonding length is L and an
axial load P is applied. The thickness of the parent plate, the repair patches and the
adhesive layers are tp, tR and tA, respectively. The Young's moduli of the parent plates
and the repair patches are Er and ER. The shear modulus, failure shear stress and shear
strain of the adhesive material are GA, 1: 5 and Ys, respectively. The problem is simplified
by making the following assumptions: (a) the parent plate and repair patches are
predominantly subjected to direct stress along the x-axis, (b) the adhesive layer is
mainly subjected to shear stress in the y-z plane, (c) the parent plate and the repair
patches are elastic and the adhesive layer is elastic-perfectly plastic, as shown in Figure
5.3, and (d) the maximum shear strain criterion is used to determine the failure of the
adhesive material.
... I~ lr----+----.1
P ._I'' t I , Y I.... P
..,. 1.....__ _ ___.1
~
Figure 5.2 Schematic drawing of a double-lap joint
0 y, Shear strain
Figure 5.3 Shear stress- strain relation of the adhesive layer
The detailed analytical solution and a Fortran programme that estimates the stress
distributions and the ultimate failure load of a double-lap joint are given in an internal
technical report produced by the authors [10]; a copy can be obtained by contacting the
authors directly.
Here, only the output stress results are presented. In particular the following
significant parameters that affect the stress distribution and joint strength are discussed:
overlap length, patch thickness/stiffuess and strength properties of the adhesive
material.
5.3.1.1 Optimum Overlap Length

The first task of the stress engineer is to decide the size of the patch. Figure 5.4 shows
the shear stress distribution in the adhesive layer of joints with short (L=5 mm), medium
(L=lO mm) and long overlap length (L>20 mm) under ultimate loading. The actual
magnitude of the stress depends on many geometric and material property parameters,
such as the thickness and length of the adhesive layer compared to the corresponding
values for the adherend material, flexibility of adhesive and type of load to be
transferred. With short overlap joints, all adhesive material is under high shear stress.
For longer overlap joints, the plastic adhesive zones developed at the ends of the bonded
overlap and separated by a lightly loaded elastic region (trough) carry the great majority
of the load. Because the load transfer zones occur at the ends and eventually reach a
constant length, no increase in joint strength will be achieved once these zones are fully
developed. Figure 5.5 illustrates the joint failure stress for different overlap lengths
obtained by applying the maximum shear strain criterion on the adhesive (elastic-
plastic model).
There is no point in increasing the overlap beyond this critical value, L=== 12 mm,
since no significant enhancement in strength, crmax , will result. However, considering
various effects such as imperfect bonding, patch delamination, environmental effecb
and a safety factor, the limiting overlap for current carbon fibre-epoxy systems is
around 30tR, where tR is defined in Figure 5.2; large flaws in the middle of such a joint
would impose no loss of strength since there is no load being transferred there. A thin
laminated plate with a 20 mm open hole will need a 60-80 mm diameter patch.
50
f.r55 .12 GPa
E,.=41 .34 GPa
40
n r.
0 I
I
r. G...= l.26 GPa
tp=3 mm
tR=2 mm
~ Ji I
.I
tA=O.l mm
6 30 \
.....· I
• I
I
.
't,=40MPa
I
~
I
~ I y,=0.05
..,5 20
;
ii .II I
Vl
I
. I
I - - L=40mm
---· L=20mm
·+·!
i
I I
10
I
-·- L=IOmm
.............. L=S mm
i
I
0
.... /
0 5 10 15 20 25 30 35 40
y -coordinate (mm)
Figure 5.4 Shear stress distribution in the adhesive layer under ultimate load.
200
a....
160
~ E,= 55 .12 GPa

i"'e 120 Eo. = 41.34 GPa
."
G,.. = I.26GPa
tp=3 mm
~
80
ta"'2mm
.s IA=O.I IDDl
.,~ 't,=40 MPa
0::
y,=0.05
40
0
0 5 10 IS 20 25 30 35 40
Bonding length (mm)
Figure 5.5 Remote failure stress of a double-lap joint as a function of overlap length.
5.3.1.2 Optimum Patch Thickness

The influence of the membrane stiffuess (i.e. the product of elastic modulus and
thickness) of patches is demonstrated in Figures 5.6 and 5.7. As expected, a joint with
soft patches (patches are thin or with low elastic modulus) has low strength, Figure 5.6.
200~----------------------------------------,
160
~
~
~
120
~
~
]"
..·············································································
.·..
l:l 80
~ Er55.12GPa
~
GA=I.26 GPa
t,.=3 nun
40 IA=O.I nun
1,=40MPa
y,=0.05
0 5 10 15 20 25 30 35 40
Bonding length (nun)
Figure 5.6 Remote failure stress of a double-lap joint versus bonding length (different patch
membrane stiffness values are examined).
5.3.1.3 Effect ofAdhesive Material

The strength properties of the adhesive certainly influence the strength of the joint.
Gr~nerally, using adhesive materials with high shear properties (strength, modulus) will
produce stronger joints, and the ultimate adhesive shear strain affects the joint strength
more than the ultimate adhesive shear stress does [4, 10]. Figure 5.8 shows that using
adhesive materials with slightly higher shear strain can enhance the joint strength
significantly.
Since high localised stresses are developed at the end zones of the overlap, further
investigations are focused on the small area near the patch edges. Figure 5.9 indicates
that the high shear strain can be markedly reduced by increasing the adhesive thickness
in the edge area.
Therefore, a joint with patches tapered from inside (with locally thicker bondline),
Figure 5.10, can substantially reduce the stress concentration in the adhesive layer and
therefore enhance the joint strength.
However, it is important to remember that good adhesive bonds can be produced
only in a small range of thickness (typically 0.125 mm to 0.25 mm) since thick bonds
tend to be porous and weak while ultrathin bonds are too stiff and brittle.
200-r------------------------------------~
150
~"'
"'
~ 100 L=40mm
P,=55.12 GPa
~ G..=I .260Pa
~ lp=3 mm
~
50 t.. =O.I mm
.,~ 't,=40 MPa
c::
y,=0.05
0
0 0.2 0.4 0.6 0.8 1.0
(ER lJt )/ (Ep ~)
Figure 5.7 Remote failure stress of a double-lap joint with a large bonding length (L = 40mm)
against patch membrane stiffness
5.4 STRENGTH PREDICTION OF EXTERNAL PATCH REPAIR
Although the 2-D analysis of the 'shear lag' model is useful to optimise the design of
the patch repair, it is only able to calculate the failure load of a lap-joint, but not that of
the patch repaired laminate (3-D configuration). In fact, the parent plate can still carry
load after losing the support of the patches (plate with an open hole). Therefore, the
two-dimensional (2-D) analytical model would underestimate the strength of the
repaired laminate and a 3-D stress analysis should be performed. However, this is
difficult to solve analytically and in the following section a finite element (FE) method
is used to calculate stresses developed in the patch, adhesive layer and parent laminate.
The stress results are used to identify critical regions where damage is more likely to
occur and a fracture model is employed to predict the strength of the repaired panel.
Comparing the strength of the repaired laminate with those of the notched and
unnotched laminates, the repair efficiency can be evaluated.
200
~ 175
g"'
"'
~ 150
]
~
L=40mm
.,8
0
Ep=55.12 GPa
~
125 E..=41.34 GPa
0 11 =1.26 GPa
tp=3 mm
IR=2mm
10 20 30 40 50 60
Shear strength of adhesive (MPa)
Figure 5.8 Influence of the ultimate adhesive shear strain on joint failure stress
0.08 -r--------------------,
Adhesive thickness
0 .06 t,..(mm)
0.05
c
0.10
"(;i 0.15 (1, _ . : 150 MPa
l:l
0.20 L=40mm
.,:a"'
.c
0.04 0 .25 Ep=55.12GPa
(ll Iit1=41 .34 GPa
0 11= 1.26 GPa
tp=3 mm
0.02 IR=2 mm
't,=40MPa
0 5 15 20
y - coordinates (mm)
Figure 5.9 Influence of the adhesive thickness on the maximum shear strain at the end zone oftht
overlap.
5.4.1 FINITE ELEMENT MODEL FOR A REPAIRED LAMINATE
Consider a symmetric laminate with a hole of diameter d, external patches bonded on

both faces and subjected to a uniaxial compressive loading, as shown in Figure 5.1. The
x-y plane of the Cartesian co-ordinate system lies in the mid-plane of the laminate and
the origin is at the centre of the hole. A quasi-isotropic lay-up [(±45/0/90h]s is selected;
the total length of the carbon fibre-epoxy panel is 100 mm long by 50 mm wide, the
hole diameter is 10 mm and the laminate thickness is 3 mm. Due to the symmetry of
loading, hole location and lay-up, only one-quarter of the laminate is modelled, Figure
5.11, using the FE77 finite element package [12].
The analysis is based on displacement formulation employing a curved
isoparametric 20-node element; four elements have been used through the thickness of
the parent laminate and the patch while the adhesive has been modelled with only one
element. The laminate and the patches are treated as homogeneous, elastic, orthotropic
materials while the adhesive layer (Araldite 2011) of thickness tA=0.1 mm is modelled
by using isotropic elements; all relevant stiffness properties are presented in Table 5.1.
ct c
r=
=112
,.
t t
..-:-----
~-----
)
II I'
0.127 mm
(Nominal)
0.5tol .Omm J Slope of I in 10
Figure 5.10 Tapering of the edges of the overlap to relieve adhesive shear and peel stresses [9].
FigureS. II Finite element mesh
Table 5.1. Parameters and properties for running the FE77 and PATCH code.
Parent:
Width: W = 50 mm
Thickness: tp = 3mm
Hole radius: r = 5 mm
Young's modulus: Ep= 55.12 GPa ( = Ex = Ey), E, = 9.3 GPa
Shear modulus: G,y = 21 GPa, Gxz = Gyz = 4.4 GPa
Vxy = 0.315, Vxz = Vyz =0.175
Unnotched strength: O"un = 593.2 MPa
Fracture toughness: K 1c = 42.5 MPa ..{r;;
* The subscripts x, y and z denote the transverse, loading and
thickness directions.
Patches:
Thickness: tR = 2mm
Patch radius: R = 17.5 mm
Same elastic properties as parent material.
Adhesive:
Thickness: tA = 0.1 mm
Young' s modulus: EA = 3.4 GPa
Shear modulus: GA = 1.26 GPa
Shear strength: 'ts = 40 MPa
Shear failure strain: Ys = 0.05
5.4.1.1 Stress Results

The stress concentration factors (SCF) in the parent, patch and adhesive material are
summarised in Table 5.2 for a repaired plate with external round patches. In the parent
laminate, the normal stress cr), is the predominant component, which may cause failure.
Two critical positions suffering high stresses are identified; point A at the hole edge,
along the x-axis and point B at the edge of the overlap, along the y-axis (the load
direction), see Figure 5.11 and insert in Table 5.2. The stress magnitude at points A and
B depends very much on the thickness of the patch. As the patch thickness increases the
SCF at point A is reduced but when it exceeds a certain value, tR> 1.5mm (0.5 tr), high
stresses appear at point B. The patch is subjected to relatively low stresses. The
maximum value (339 MPa) of the dominant stress component, crn, is less than the
remote loading (350 MPa), when the patch thickness is 1.5 mm (half of the parent
plate). Therefore, patches are generally safe if they are not too thin (greater than 40% of
the parent plate thickness).
The adhesive/adherend interface is another location where failure may occur. The
interlaminar shear stress, 1:,/, is the main stress component to initiate failure. High shear
stress concentration at the overlap edge may cause debonding. Only the narrow zones
at the interface edges transfer load from the parent laminate to patches and most
adhesive materials are in low stress state. High peel stresses, cr"' can also contribute to
the final failure; they may cause the parent plate to delaminate because of its poor
through-thickness strength. According to extensive experimental observations [12, 13]
and the finite element stress results, the failure mechanisms of the external patch
repaired laminates discussed in this work are shown schematically in Figure 5.12.
Damage initiates by adhesive plastic deformation and patch debonding at the edges
(point B) due to high shear and peel stresses. As the patch loses partially its load
carrying capacity, the local stress at the hole boundary of the parent plate (point A)
increases causing 0° fibre microbuckling. Under increasing applied load, fibre
microbuckling surrounded by delamination propagates like a line-crack across the
specimen width. Final failure occurs when the buckling length reaches a critical length
of 2 - 3 mm depending on material system and lay-up, very similar to the failure pattern
observed in a plate with an open hole [14, 15].
5.4.2 COHESIVE ZONE FRACTURE MODEL
Soutis and coworkers [ 14, 15] investigated the compressive failure of carbon fibre-
epoxy laminate containing a circular hole and developed a fracture-toughness model
[14], which successfully predicts the notched compressive strength of any orthotropic
CFRP laminate. In this section, the original model is briefly introduced and then applied
to the problem of a patch repaired laminate.
5.4.2.1 Plate with an Open Hole

Consider the compressive failure of a multi-directional composite panel containing a
circular hole. The damage development is described in Figure 5.13 . At low loads, no
damage is evident either on the faces or in the bore of the hole. At between 75% and
80% of the failure load microbuckling occurs in the 0° plies accompanied by matrix
cracking and delamination in the off-axis plies. At higher loads, microbuckling
propagates in a stable manner as a line crack until it reaches a critical length lcr when
catastrophic failure occurs.
Table 5.2 Stress concentration factors, critical locations of round external patch repair
specimens and the influence of patch thickness, tR.
Parent Adhesive Patch
® ®
B
,-,
I ' '\
{I A(JA\I
\ I
'
'_.-' I B
B
ft.y
14 x
(cryy )max 't..,., (cryy)max
t,. (cr.,Jmax {<lvon -~Uses)max
(mm)
a remote 0 remote crrunote 0 remote
Locations
cr..,m«e
at''A"' at "B"'
at "C'
0 3.19 . . .
(no patches)
0.5 2.03 1.13 0.250 0.119 0.461 ''C' 1.62

0.8 1.73 1.19 0.294 0.174 0.539 ''B" 1.34
1.0 1.57 1.22 0.31 4 0 .204 0.575 " B"' 1.21
1.3 1.40 1.27 0 .338 0 .240 0.6 16 "B" 1.05
1.5 1.31 1.29 0.350 0 .261 0.637 "B" 0.968
1.7 1.24 1.32 0.362 0.279 0.656 uon 0.901
2.0 l.l6 1.34 0.377 0.302 0.681 'lB" 0.821
25 1.06 1.38 0.396 0.332 0 .71 4 " B" 0.727
3.0 0.99 1.40 0.411 0.354 0.738 " B" 0.666
Soutis and Fleck [14] have modelled the initiation and propagation process using the
stress distribution at the edge of the hole CYyy(x, 0), and linear elastic fracture mechanics.
The buckled zone (0° fibre microbuckling, delamination and matrix cracking in the off-
axis plies) at the edges of the hole is mathematically replaced by a through thickness
crack (line crack) with no traction on the crack surfaces. It is assumed that:
+
(b) Debondingtdelaminatioo
+
(a) Intact patches (c) Micro-buckling initiation
initiation underneath patch
~ +
(d) Micro-buckling
propagation
k lcr
~
~lcr
(c) Fracture when micro-buckling

reaches a critical length
Figure 5.12 Failure mechanisms of patch repaired laminate under compression
(i) Average stress criterion: Microbuckle cracks initiate and grow stably over a distance
e from the hole edge when the average stress over the distance ereaches the unnotched
strength of the laminate, O'un :
) fr+l
CJun = f ,. CJ yy(x,O)dx (I)
The stress distribution crl 0 (x,O) can be calculated analytically or by using finite element
analysis. The applied stress normalised by the unnotched strength for stable growth of
buckling is presented in Figure 5.14 as a function of damage length, e,
normalised by
the hole radius, r.
(ii) Fracture mechanics criterion: Unstable fibre microbuckling occurs when the stress
intensity factor (SIF), K 1, at the tip of the buckled region is equal to the material fracture
toughness, Kc.The stress intensity factor at the tip of a micro buckle of length efrom the
hole edge is expressed as:
(2)
where cr~ is the applied axial stress and f(£/r) is the correction factor for symmetrical
cracks emanating from a circular hole in an anisotropic material. The values of f(€i 1)
can be calculated by FE analysis.
Figure 5.13 Failure process of a notched laminate under uniaxial compression.
Here, the crack-driving force is defined by K" which describes the dominant stress and
deformation field in the vicinity of the crack tip. Crack advance is thus restrained by
lowering of the applied load. When K 1=Kc the crack growth becomes unstable and
eqn.(2) can be written as:
(3)
Having obtained the unnotched compressive strength Gun and the in-plane fracture
toughness Kc of the laminate, the notched compressive strength and the critical buckle
length f!cr for any hole radius r can be predicted. The general procedure is to plot the
ratio of cr~lcrun given by equations (1) and (3), versus fir. Then the failure strength can
be obtained from the point where the two curves intersect, Figure 5.14. This point alsu
provides the critical buckle length f!cr·
1.0
0.8
0.6
o-
o..,
0.4
Kr = KlC
Unsl.able growth
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0

lfr
Figure 5. 14 Strength and buckled length prediction of a notched laminate (HTA/913C [±45/02]3 5 ,
panel width W=50 mm, hole radius r = I0 mm).
The fracture mechanics approach is justified as the microbuckled zone resembles a

crack and the damage zone associated with the microbuckle is small in extent compared
with other specimen dimensions. The in-plane fracture toughness K1c was measured for
six different T800/924C multidirectional laminates by performing a series of tests on
centre-cracked compression specimens [ 14, 16]. These specimens were 245 mm long by
50 mm wide with central slits perpendicular to the applied load and pre-sharpened by a
razor blade. The crack length-to-width ratios ranged from 0.1 to 0.6 and the slits were of
sufficient width (about 1 mm) to prevent contact of the slit faces under compressive
loading. An anti-buckling device was used to prevent Euler bending and the tests were
performed on a screw-driven test machine at a displacement rate of 0.017-mm s·'. The
specimens behaved in an elastic-brittle manner and the fracture toughness was
computed from the failure load; the orthotropic K calibration factors were obtained from
a finite element analysis [ 14]. It was concluded that K1c was independent of initial crack
size and could be considered as a laminate property. Typical values for the carbon fibre-
epoxy system examined were in the region of 40 to 50 MPa m 112 , depending on lay-up.
For theoretical predictions of the compressive toughness and microbuckling propagation
in composites the reader should refer to some recent work by Fleck et al [ 17, 18] and
Soutis et al [ 19].
1.0
0.8 an r
-=1--
w
0.6
O'a
0'1111
0.4
.p fracture model
-----------7----- .......................
0.2 -------......
~=- .....' ' ',,~
0.0
0.0 0.2 0.4 0.6 0.8 1.0
r/W
Figure 5.15 A comparison of strength predictions by the maximum stress, ductile failure and
Soutis-Fleck fracture models (HTA/913C [±45/0/90hs)
Figure 5.15 shows an example ofthe notched strength predictions obtained for a carbon
fibre-epoxy HTA/913C, [±45/0/90hs laminate by the fracture model, bounded by the
simple failure criteria of the ideally brittle response (notch sensitive) and the ideally
ductile behaviour (non-insensitive). The strength that estimated by the current model is
in good correlation with the test results. The predictions using the maximum stress
criterion (notch sensitive curve) are rather low and those assuming ductile failure arc
much higher than the measured strengths. The 'ductility' displayed by the composite is
due to the development of local damage near the hole in the form of fibre
microbuckling, delamination, matrix plasticity and matrix cracking in the off-axis plies.
5.4.2.2 Patch Repaired Laminate

Experimental observations [12, 13] indicate that in the patch-repaired laminate under
uniaxial compression the critical failure mechanism is 0° fibre microbuckling, which is
similar to the open hole case. In both configurations, microbuckling initiates when tht'
local compressive stress parallel to the 0° fibres at the hole edge equals the unnotched
strength of the laminate. As the remote load is increased the microbuckle grows in
length before becoming unstable at a critical length tcr= 2-3 mm. However, as tht:
patches and adhesive reduce the stress concentration at the edge of the hole in the parent
plate, both the fibre micro buckling initiation and final fracture of the repaired laminate
are delayed to higher applied stresses.
Due to this similarity in damage development, the fracture model developed for the
notched laminate can be employed to predict the failure load of the patch-repaired
laminate. Of course, the effect of the patch and adhesive plastic deformation need to be
considered in the calculations of the stress distribution, cryy(x,O) and stress intensity
factor, Kl( f), required in the model. The derivation of these parameters is explained in
the following paragraphs.
i) Determination of stress distribution

In the external patch repair approach the load is taken over and around the open hole
(damage area). The patch is also capable of withstanding the normal (peel) and shear
stresses that develop at the edges of the overlap. In the stress analysis, the contribution
of the patches and adhesive can be replaced by forces which are transferred through the
adhesive, so that a 3-D patch bonded structural problem becomes a 2-D open hole
problem, as shown in Figs.16.
The key step is to find the forces transferred by the adhesive. The double lap joint
analysis described in section 3 (Figure 5.2) indicates that the load transfer zones occur
at the overlap ends separated by a lightly loaded elastic region (trough), Figure 5.4.
Therefore, it is assumed here that the adhesive transfer forces develop at the hole edge
(Q 1) and at the patch edge (Q 2), Figure 5.17. The Q 1 and Q 2 forces vary along the hole
edge with the angular position, a, following the equations,
(4)
where a is the angle to x-axis, F 1 and F2 are the maximum forces along they-axis when
a=90° . F1 and F2 are determined by using the shear stress distribution under adhesive
ultimate load, shown in Figure 5.17.
F 1 and F2 are simply the area under the adhesive shear stress curve (f.tdx) divided by
the parent plate thickness tp. Q 1 and Q 2 are defined as the forces per unit laminate
thickness and arc length. Figure 5.16 illustrates the approach for the stress distribution
o 1 ~(x,O) calculation; the patch and adhesive resistance is represented by the equivalent
forces Q 1 and Q 2 . The stress distribution O")y(x,O) of the patch-repaired laminate is
determined by a linear superposition of the three stress components, 0)y 0(x,O), 0yy 1(x,O)
and o1y2(x,O), that correspond to a finite plate with an open hole under the remote load
ow, the hole edge force F 1 and the patch edge force F 2, respectively, i.e.,
The stress functions f)yo(x,O), fy) 1(x,O) and fyy 2(x,O) are the stress distributions when o=
=F 1 =F 2 =1 and obtained from a 2-D elastic finite element analysis, using the FE77
package [12].
ii) Determination of the stress intensity factor (SIF)

Figure 5.18 shows the superposition method for the stress intensity factor calculation.
The stress intensity factor K 1( f) of the patch repaired laminate with cracks is determined
by a linear superposition of three stress intensity factors, K 10 ( f), Kll (f) and K 12 ( f), for a
laminate with cracks emanating from a hole under three load cases: the remote load cr'··.
the hole edge force F 1 and the patch edge force F2 .
(6)
where k10 , k 11 and k 12 are the stress intensity factors when cr~ =F 1=F 2=1 and determined
by a simple 2-D FE analysis, Table 5.3 .
•
An external patch
repaired laminate
•
An open-hole laminate
subjected to end load
An open-bole laminate
subjected to bole-edge load
+
An open-bole laminate subjected
to patch-edge load
Figure 5.16 Stress distribution,cryy(x,O), for an external patch repaired laminate.

r' ;<
: h"'Adhesive
, ! arent
X I~ P
(a) Schematic drawing of a double-lap joint
: (Fo t,.) ea: (~t,.)
0 L
(b) Shear stress distribution in the adhesive layer under the ultimate load
Figure 5.17 Calculation of F 1 and F2 load parameters by a 2-D lap joint analysis
Table 5.3 Basic stress intensity factors for HTA/913C [+45/-45/0/90hs (W=50 mm, r=5
mm
l/r 0.1 0.2 0.4 0.6 0.8 1.0 1.2
/(mm) 0.5 1.0 2.0 3.0 4.0 5.0 6.0
kw(/) 3.452 4.377 5.256 5.769 6.173 6.547 6.920
-k,,(l) 2.291 2.821 3.208 3.350 3.421 3.468 3.511
-kdl) 2.856 3.615 4.329 4.739 5.061 5.359 5.657
iii) Strength prediction

Having obtained the stress distribution cryy(x,O) and stress intensity factor K 1( t'}, the final
failure load crt and the critical buckling length ecr of the external patch-repaired
laminate can be calculated in a similar way to that shown in Figure 5.14. Equations (5)
and (6) are substituted in equations (l) and (3), respectively, in order to obtain the
damage initiation/growth and final fracture of the repaired configuration.
As an example, consider a finite CFRP multidirectional laminate (HTA/913C, [+45/-
45/0/90]35), 100 mm long by 50 mm wide and 3 mm thick, containing a circular hole of
10 mm in diameter and repaired by two 3 5 mm diameter patches of the same material,
HTA/913C, [+45/-45/0/90hs. (2 mm thick). Table 1 gives the basic parameters and
properties of parent, patches and adhesive used for running the FORTRAN program
PATCH [10, II]. From the 2-D double lap joint analysis, the load parameters F 1 (=57.6
MPa) and F 2 (= 74.5 MPa) are obtained, Figure 5.17. A 2-D finite element elastic
analysis is performed to derive the basic stress distributions, fyy0(x,O), fyy ,(x,O) and
fyy 2(x,O), and stress intensity factors, k 10(/), kii(l) and k 12(/), see Figure 5.19 and Table 3,
respectively.
-o- +
___
.___ _. ~<0
-M-
•
K~Q
•
L..----....1 Ku(l)
An external patch An open-hole cracked An open-hole cracked

repaired cracked laminate under axial load laminate subjected to hole-
laminate edge load
+
An open-hole cracked
laminate subjected to patch-
load
ed.~te
Figure 5.18 Stress intensity factor, K1(l), for an external patch repaired laminate
calculated by linear superposition.
In Figure 5.20, the stress distribution cryy(x,O) calculated by the linear superposition
approach (damage model) is compared to that obtained from a 2-D elastic FE analysis .
As the adhesive is assumed purely elastic in the FE calculation, it appears that the
adhesive carries much higher load than its capacity and therefore lower stresses an:
developed in the parent plate, as shown by the dot-dash line in Figure 5.20. The stress
distribution corresponding to the open hole case is also showed for comparison
purposes. The result of the linear superposition approach that accounts for the adhesive
plastic deformation lies between the two elastic solution curves of the repaired and
open-hole laminate.
The final failure strength ( cr; ) and the critical microbuckling length U'cr) of the
patch repaired laminate are shown in Figure 5.21 and compared to the open-hole
laminate. For this case, the patch-repaired laminate fails at 353 MPa compared to 258
MPa for the open hole, which is almost 40% higher. Experimental data and theoretical
results for two composite systems (HTA/913C and T800/924C) are shown in Table 5.4;
the strength predictions are in general I 0% to 15% lower than measurements,
suggesting that the model will lead to a slightly conservative design.
3
fnO(lt,O)
~ /
~ ~-
2
/ l t,O) -frll (ll,O)
--
I
\
!'.....r-.L /
..........
'--. i--
0
-1
0 2 4 6 8 10 12 14 16 18 20
lt -r (mm)
Figure 5.19 Stress distributions of parent plates along x-axis subjected to a unit load: cr~= I MPa,
F 1= 1 MPa and F ~= l MPa.
5.5 SCARF PATCH REPAIRS
In recent years, demands on technology development for permanent field repairs of

composite structures have increased considerably. Thick monolithic structures lend
themselves to such repairs since external patch would cause excessive out-of-mouldline
thickness and unacceptably high bondline peel and shear stresses [7, 8]. Scarf joints
between identical adherends will have a uniform distribution of adhesive shear stress
and hence will show a higher strength than the lap joint types.
Design methods for scarf-type bonded repairs require criteria to predict both strength
and durability. In this section, a three-dimensional stress analysis is presented and the
stresses developed in a flu sh scarf repaired laminate, Figure 5.22, under uniaxial
162 C. SOUTIS & F. Z.HU
compressive loading are discussed. Predictions are made of the optimum scarf angle and
likely points of failure.
Table 5.4 Compressive strengths of notched and repaired laminate plates

Specimen Measurement Prediction
crr!MPa <Jri<Jun <JriMPa <Jri<Jun
HTA/913C Notched 315.0 0.531 257.4 0.434
[±45/0/90hs
Gun=593.2
Repaired 405.1 0.683 353.0 0.595
T800/924C Notched 270.7 0.469 256.0 0.444

[±45/0/90hs
Gun=578.8
Repaired 399.4 0.692 368.1.0 0.636
4
I I
Open bole
3 /
Oyy \/
\
\
Damage Model
2 \ /
'
/ --- --
/epair (FB), elastic
~ ...._... __
-- f--L.
v·-- ---
--- -- -- -- - --
--· -
0 2 4 6 8 10 12 14 16 18 20
x-r (mm)
Figure 5.20 <Tyy(x,O) stress distributions along the x-axis for a HTN913C quasi-isotropic
laminate.
1.0
Failure load
0.8
0.6
_______ ...
-----
Unstable growth:
0.4 I )-
Kt = Kte
I
: (Open-hole)
'\
I
I
I
I
Damage I
0.2 I
initiation I
I
I
I
lcr
I
0.0 +---r--~-.--~~F---,-'"""T---r---,--""T"--,
0.0 0.\ 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
II r
Failure occurs at intelllection point:

(a) Open-hole plate: crT I cr WI =0.434, lair= 0.364, fa. = 1.82 mm
(b) Repaired plate: crT/crUD =0.595,t.,./r=0.394, 1a =l.97 mm
Figure 5.21 Calculation of strength and buckled length of a notched and repaired laminate
(HTA/913C (±45/0/9Qh 5 , O'un= 593 MPa, W= 50 mm, r=S mm and R=l7.5 mm).
Figure 5.22 Specimen geometry of the flush scarf-type bonded repair

5.5.1 ADHESIVE BONDED SCARF JOINT AND OPTIMUM SCARF ANGLE
The joint of interest has identical adherends, use a relatively brittle adhesive and have
small scarf angles, Figure 5.23.
Figure 5.23 A schematic of scarf joint geometry
For this simple case, the semi-empirical analysis [20] predicts that the optimum scarf
angle for a maximum strength joint is a function of the adhesive shear strength 'ts and
laminate strength O'un. given by
(7)
For small 9, the failure stress, Sgfi ofthe 2-D scarf joint is determined by
(8)
where KA and KL are the stress concentration factors (SCFs) in the adhesive and
adherent, respectively. Based on the correlation with experimental data [20, 21J,
KA = 2.88 and KLz2.35, for 9::; 10°. The load carrying capability of the adhesive and
the composite plates in the joint is plotted in Figure 5.3 as a function of scarf angle.
The optimum scarf angle occurs when the adhesive failure load is equal to the
laminate failure load. For the composite system examined ( 'l's =40 MPa and O"un =454
MPa) eqn (8) results to 90p1=4°.
5.5.2 FLUSH SCARF PATCH REPAIR
The scarf patch repaired laminate, shown in Figure 5.22, is generally a three-
dimensional (3-D) problem which is over simplified by the scarf joint model described
earlier. In a scarf joint, all loads are transferred through the adhesively bonded interface.
In fact, the parent plate can still carry load after losing the support of the patch (a plate
with a tapered open hole). Therefore, the 2-D analysis would underestimate the strength
of the repaired laminate and a 3-D stress analysis should be performed.
Chapter 5 Strenth Analysis of Adhesively Bonded R epairs 165
400
l
~
.."
!!
300
]
1! 200
·a
...~
0
-s 100
""g
"'
0
0 2 3 4 s 6 7 8 9 10
Scarf angle (Degree)
Figure 5.24 Fail ure stress vs. scarf a ngle for acomposite scarf joint
5.5.2.1 Elastic FE Analysis

The FE-77 finite element package [1 2] is employed t o determine the 3-D stress field in
the repaired l aminate, using a n isoparametric 8-node solid element. Due to the
odelled, F igure
symmetry of loading and hole ol cation, only o ne-quarter of the p late is m
5.25 .
Because a high stress concentration is expected near the bonded interface, a fine
mesh refinement is required in this a rea. The smallest element size in the radial
direction is 0.01 25 mm. The Celion/LARC-1 60 c omposite laminate ( length=304.8 mm,
width=l52.4 mm, t=2.27 mm) and the patch, [±45/0/90hs, are treated as homogeneous,
elastic materials with the same properties5: Exx=E,y=53 .8 GPa, Ezz= l l.3 GPa, Gxy=20.5
GPa, Gxz=G,-z=4.85 GPa, V,, =0.3 1, Vxz=Vvz=0. 19 and compressive strength (}un=454
MPa. The e poxy adhesive layer of thickness tA =0. 129 mm has the fo llowing stiffness
and strength properties [21] :E=3.40 GPa, G= l.26 GPa, v=0.35 and rs=40 MPa.
5.5. 2.2 Stress Results

It is found that the dominant stress components are the in-plane stress (}x in the parent
laminate and the shear stress r tangent to the tapere d bond surface. The stress contours in
the parent plate, patch and adhesive layer are p ottedl for a scarf angle of 3° in Figure
5.26. Other in-plane and through-thickness stress c omponents a re rel atively s mall and
could be neglected in the failure load calculations. Significant o:,. stress concentration
occurs at the scarf tip of the parent p lateand the patch, Figure 5 2. 6a. The patch is
subjected t o relatively low stresses, Figure 5 2. 6b, and the shear stress in the adhesive is
quite uniformly distributed in the radial direction, Figure 5.26c. The predicted peak
stress positions coincide with those observed experimentally [21].
Figure 5.25 Finite element mesh for a scarf patch repaired laminate
The axial stress distribution, o;,, along the x-axis (load axis) has a steep gradient near the
tip of the scarf (x=R, y=O), suggesting the existence of stress singularity, and
approaches the applied remote stress, Sg. within eight plies (1 mm). Since a
mathematical stress singularity exists at the scarf tip due to stiffuess discontinuity, the
computed stresses are taken near but not at the tip of the scarf, Figure 5.27. The stresses
closest to the scarf tip are at r-R=0.5h; that is at a distance of 0.0625 mm. As the
distance from the tip (r-R) increases, the in-plane stress o;, is rapidly decreases. The
largest compressive stress concentration factor (K1=1.7) occurs at <p=0° (r-R=0.5h, y=O)
and is reduced down to 1.06 at <p=90° (0, R) for all radial locations, compared to 3.6 for
the case with an open hole; o;, shows no significant variation along the y-axis. Figure
5.27 suggests that o;, is large in the region -60°:5:<p:5:60° at r-R<5h. The shear stres~
't varies mainly in the circumferential direction; its maximum value occurs at <p=0° along
the x-axis and becomes zero along they-axis. Its magnitude is nearly proportional to the
scarf angle; for 8=9°, Tis almost three times higher than the value obtained for 8=3°.
5.5.2.3 Strength Prediction
The stress analysis above, has shown that the dominant stress components are crx and 1
for the parent laminate and the adhesive layer, respectively. The maximum stress failure
criterion, for K1=1.7 at distance r-R=0.5h and O"un =454 MPa, predicts a residual strength
of267 MPa, which is about 25% less than the measured value [21]. A SCF value taken
at distance r-R=2h would produce a better agreement. This suggests that stress
redistribution, due to material non-linearities, occurs in the repaired region before final
failure and using an elastic SCF can substantially underestimate the failure load.
Min A B C 0 E Max
0.95 0.99 1.01 1.03 1.05 1.10 2.90
Load axis
.. .. s,
Mm
(a) CixiSg in the parent palte
BCDEMax C D E Max
1.01 1.03 1.05 1.10 1.33 0.03 0.04 0.048 0.052
Min
Max
Min
(b) cr)Sg in the patch (c) t/Sg in the adhesive layer
Figure 5.26 Finite element stress contours in the scarf patch repaired laminate
Previous work[5, 22] has shown that the approach for predicting failure in such
laminates is that of averaging the stresses over a distance from the tip of the scarf,
suggesting that the exact values of the stresses at the tip are not too important. The
average stress failure criterion (ASFC) assumes that failure occurs when the average
stress over a length d0 from the scarf tip equals the ultimate strength of the material. The
average of the stress component o:, is defined as:
- __1_ fR+do dx (9)
ax- R ax
do
and failure in the parent laminate occurs when <Jx=<Jun and in the adhesive when 1:=1:,.
Assuming a characteristic length d0=1 mm, the failure load of the scarf patch repair
is obtained as a function of the scarf angle, Figure 5.28; for the composite system
examined, the optimum scarf angle is almost 7° compared to 4° obtained for the scarf
joint. The experimental strength data [21] is in good agreement with the theoretical
predictions.
The value of the adjustable parameter d0 is based on the correlation with
experimental data and does not result from a mechanics analysis. It accounts for
material non-linearities and plasticity of the adhesive that reduce local peak stresses in
the repaired region, stress redistribution mechanisms which are not considered in the
elastic FE analysis.
1.6
-o- (r·R)=0.5 h
Cl
U) -<>- (r-R)=h
--;: 1.4 -a-(r-R)=2 h
\?
-<>- (r-R)=3 h
-IZ-(r-R)=10h
1.2
0 10 20 30 40 50 60 70 80 90
<p (Deg)
Figure 5.27 Circumferential distributions of OxiSg in the parent plate for 9=3°.
5.6 DISCUSSION AND CONCLUDING REMARKS
The analytical/numerical approaches presented in this chapter can improve the design of
external patch repairs and predict their strengths quite accurately. For the cases
examined, the optimum overlap length of the patch is approximately 12-15 mm. The
optimum patch membrane stiffness (i.e. the product of elastic modulus and thickness) is
half of the parent stiffness. Overstiff patches are harmful because they produce higher
peel and shear stress concentrations at the edge of the overlap and increase the weight or
the structure. Tapering the patch edges and increasing the adhesive thickness near the
edge reduce the local stresses substantially and therefore increase the failure load. Using
the optimum patch configuration bonded repairs can recover almost 80% of the
undamaged laminate strength [ 11 , 13].
600 -Load bound

······ Adherend failure
---Adhesive failure
500 ~ Test data
400 0
Sg.f
\
(MPa) 300
\
200 "'
100 Scarf joint
0
0 2 4 6 8 10
Scarf angle, e (Deg)
Figure 5.28 Failure stress vs. scarf angle for the scarf joint and the scarf patch repaired laminate.
A characteristic length dO= I mm is used to correlate the test data [21].
A fracture mechanics based criterion with damage initiation and growth governed by a
stress-based criterion has been applied successfully to predict initiation and propagation
of damage in the form of 0° fibre micro buckling and also the compressive strength of
CFRP laminates repaired with external patches. The model treats the damage zone at the
edges of the patched hole as a through-thickness crack (line-crack) with no traction on
its faces and takes as its input the compressive unnotched strength and the in-plane
fracture toughness of the laminate. Shear lag stress analysis and simple superposition
procedures are employed to assess the contribution of the patch and adhesive. The
strength predictions are compared with experimental measurements; it is shown that the
predictions are conservative and errors are acceptable ( 10% to 15% lower than the
measured values). Further work is required to examine different laminate stacking
sequences, environmental effects and the response of such repairs under fatigue loading.
Simple analytical and numerical models were also presented to determine the
optimum geometry and compressive strength of flush composite repairs. To account for
material non-linearities and plastic deformation of the adhesive the average stress
failure criterion (ASFC) is used with the FE stress distributions to estimate the ultimate
strength of the 3-D repaired configuration. This avoids the need for a non-linear
analysis, saving on computing time and memory requirements. The 2-D scarf joint
analysis underestimates the strength of the scarf patch repair by more than 40% and
predicts an optimum scarf angle of 4° compared to almost 7° obtained from the ASFC.
A 4° scarf angle would remove too much undamaged material and weaken the repaired
laminate.
5.7. REFERENCES
1. Baker, A.A. (1990) Repair techniques for composite structures, Chapter 13,
Composite Materials in Aircraft Structures, ed. by D.N.Middleton, Longman
Scientific&Technical, Longman Group UK Ltd., 207-227.
2. Myhre, S.H. and Labor, J.D. (1981) Repair of advanced composite structures,
Journal ofAircraft, 18(7), 546-552.
3. Hu, F.Z. and Soutis, C. (1997) Strength prediction and design optimisation for
external patch repairs in composite structures, 4th International Conference on
Deformation and Fracture of Composites, 24-26 March 1997, UMIST, UK,
Proceedings pp.609-618.
4. Soutis, C and Hu, F. Z. (1997) Design and performance of bonded patch repairs of
composite structures. Proc. Instn. Mech Engrs. Vol.211, Part G, 263-271.
5. Soutis, C. and Hu, F.z. (1997) Repair design of composites and efficiency of scarf
patch repairs. Proc. ICCM-11, Gold Coast, Australia, 14-18th July 1997. VI,
pp.395-404.
6. Soutis, C. and Hu, F.Z. (2000) Repair design of composites and efficiency of scarf
patch repairs. AIAA Journal, 38 (4), 737-740.
7. Hart-Smith, L.J. (1973) Adhesive-bonded scarf and stepped-lap joints. NASA CR-
112237.
8. Matthews, F.L. (ed.) (1987), Joining fibre-reinforced plastics, Elsevier Applied
Science.
9. Hart-Smith, L.J. (1980) Further developments in the design and analysis of
adhesive-bonded structural joints, American Society for Testing and Materials.
Symposium on Joining of Composite Materials, ASTM STP-749, Mineapolis.
Minnesota.
10. Hu, F.Z. and Soutis, C. (1996) Analysis and optimisation of bonded patch repairs
in composite structures, Imperial College, Department of Aeronautics, Progress
Report, 72 pages.
11. Hu, F.Z. and Soutis, C. (2000) Strength prediction of patch-repaired CFRP
laminates loaded on compression, Comp. Sci. & Techn. 60 (7), 1103-1114.
12. Hitchings, D. (1995) Finite element Package FE77 user's manual, Imperial
College of Science, Technology and Medicine.
13. Soutis, C., Duan, D-M. and Goutas, P. (1999) Compressive behaviour of CFJU>
laminates repaired with adhesively bonded external patches. J. Comp. Struct., 45
(4), 289-309.
14. Soutis, C and Fleck, N. A. (1991) Static compression failure of carbon fibre
T88/924C composite plate with a single hole, J. Compos. Mater. 24(5), 536-558.
15. Soutis, C. (1994) Damage tolerance of open-hole CFRP laminates loaded in
compression, Composites Engineering, 4 (3), 317-327.
16. Soutis, C., Curtis, P.T. and Fleck, N.A. (1993) Compressive failure of notched
carbon fibre composites, Roy Soc Lond. A, 440, 241-256.
17. Fleck, N. A. (1997) Composite failure of fibre composites, Advances in Applied
Mechanics, 34 , Academic Press, eds., J. W. Hutchinson & T. Y. Wu, 43-118.

18. Sutcliffe, M. P. F. and Fleck, N. A. (1996) Microbuckle propagation in fibre
composites',__Acta Metallurgica & Materialia, 921-932.
19. Soutis, C. and Curtis, P.T., (2000) A method for predicting the fracture toughness
of CFRP laminates failing by fibre microbuckling. Composites Part A, 31, 733-
740.
20. Advanced composite design guide, AFML Wright-Patterson Air Force Base,
United States Air Force, Vol.ll, Analysis, 1973.
21. Jones, J.S. and Graves, S.R., (1984) Repair techniques for Celion/LARC-160
graphite/polyimide composite structures, NASA-CR-3794.
22. Hu, F.Z., Soutis, C. and Edge, E.C. (1997) Interlaminar stresses in composite
laminates with a circular hole. Composite Structures, 37 (2), 223-232.
NOTATION
d Hole diameter
D Patch diameter
Er Elastic modulus of the parent laminate
Ertr Membrane stiffness of the parent laminate
ER Elastic modulus of the patch
ERtR Membrane stiffness of the patch
F1 Load parameter in Q 1=F 1*sina
F2 Load parameter in Q 2=F 2*sina
fYyO(x,O) Stress distribution of open-hole laminate subjected to remote load a== 1
foyl(x,O) Stress distribution of open-hole laminate subjected to face load Q 1=sina
foy2(x,O) Stress distribution of open-hole laminate subjected to face load Q2=sina
GA Shear modulus of the adhesive
KA.L Stress concentration factors in a scarf joint
K, Stress concentration factor
KI(l) Stress intensity factor of patch repaired plate
Kro(/) Stress intensity factor of open-hole laminate subjected to remote load a=
K11(l) Stress intensity factor of open-hole laminate subjected to face load Q 1
K!2(/) Stress intensity factor of open-hole laminate subjected to face load Q 2
K1 Stress intensity factor
K1c Fracture toughness
kw(l) Stress intensity factor of open-hole plate subjected to remote load a== 1
k!I(/) Stress intensity factor of open-hole plate subjected to face load Q 1=sina
k!2(l) Stress intensity factor of open-hole plate subjected to face load Q2=sina
L Bonding overlap length
l Crack length at the hole edge
p Applied load
Ql Load applied on the hole surface of the parent plate
Q2 Load applied on a circular face of the parent plate around the hole centre
with a radius ofR (i.e. patch radius)
R Patch radius
Sg Remote applied stress
Sg, r Failure stress of a scarf joint/repair
r Hole radius
Thickness of the adherend
tv Thickness of the parent lamiante
tR Thickness of the patch
tA Thickness of the adhesive layer
W Width of the parent laminate
Greek Symbols
a Anti-clockwise angle to the loading axis around the hole centre
Ye Maximum elastic shear strain of the adhesive
Ys Ultimate shear strain of the adhesive
e Scarf angle
8ovt Optimum scarf angle
<P Circumferential direction in a scarf repair
O"r Failure strength
O"max Maximum remote failure stress
O"n Compressive strength of a notched laminate
cr~ Remote stress at the specimen ends
O"un Compressive strength of an unnotched laminate
O"yy(x,O) Stress distribution near the hole edge of patch repaired laminate subjected to
remote load cr~
O"yy0(x,O) Stress distribution of open-hole laminate subjected to remote load cr~
O"yy 1(x,O) Stress distribution of open-hole laminate subjected to face load Q1
O"yy2(x,O) Stress distribution of open-hole laminate subjected to face load Q2
O"xx(O, y) Stress distribution near the scarf patch repair
'ts Shear strength of the adhesive
Abbreviations
2-D Two dimensional
3-D Three dimensional
CFRP Carbon fibre reinforced plastics
FEA Finite element analysis
SIF Stress intensity factor
CHAPTER 6 ADHESIVELY BONDED JOINTS FOR FIBROUS COMPOSITE
STRUCTURES 1
L. J. HART-SMITH
The Boeing Company, Long Beach, Cal!fornia, USA
6.1 INTRODUCTION
The use of adhesive bonding in aircraft structures dates back to the dawn of aviation
history. Wooden structures contained glued scarf joints in the primary members, wing
spars were often constructed from laminated strips of wood to improve the mechanical
properties, and plywood skins and shear webs contained considerable glue in
themselves and transferred their loads to the primary structural members entirely
through bonded joints. In the infancy of the development of airframes, mechanical
fasteners were used only at areas of relatively high load intensity (for which the glued
joints were not considered strong enough) and to provide manufacturing breaks to
facilitate storage and repair. One is tempted to suggest that our grandfathers'
understanding of the respective uses of adhesive bonding and mechanical fastening in
aircraft structures seems to have been lost with the passage of time.
At all stages of development since then, the science of predicting the strength of
adhesively bonded joints has always been adequate. There are virtually no known
instances of failures associated with improperly proportioned joints. Unfortunately,
there are far too many well-known service problems involving the interaction between
moisture and the glue (or resin). Such problems have occurred with wooden structures
using the early casein glues. They recurred later with some U.S. adhesively bonded
metal structures, particularly those with perforated honeycomb cores that had not been
treated to enhance their resistance to corrosion. Even today, moisture absorbed within
composite laminates has an adverse effect on secondary bonding and on the second-
1 This chapter is an updated version of a paper presented to the International Symposium on Joining And
Repair Of Fibre-Reinforced Plastics, at the Imperial College, London, on September I 0 and II, 1986 and
published in the proceedings, Joining Fibre-Reinforced Plastics, edited by F. L. Matthews, Elsevier Applied
Science, England, 1987, pp. 271-311.
173
L. Tong and C. Souris (ed1'),
174 L. J. Hart-smith
stage curing of stiffeners against a precured skin. It should be noted that all of thes~
problems are beyond analysis. They are not load-sensitive, and the only solution is to
eliminate the problems by use of proper processing and verification techniques and by
the correction of processing specifications, not to establish some knock-down factor
and tolerate the existence of such conditions. In particular, it is necessary to add som~
form of durability testing, under peel stresses in a hostile environment, to the quality
assurance program for bonding composite structures that is equivalent to the wedge-
crack test now standard for metal bonding because the lap-shear test coupons alone
were shown to be inadequate (see Reference 1). References 2 and 3 explain how release
agents (particularly silicone) transferred from some peel plies and pre-bond moisture
from a number of sources has been responsible for interfacial separations in service and
during fabrication (without any damage to the adhesive itself) between adherends. All
of the strength analyses and design techniques discussed here assume that such
durability issues have been resolved by proper processing, bearing in mind that many
specifications do not ensure durable bonds.
The introduction during the 60s of the advanced fibrous composites - boron-epoxy
and carbon- (or graphite-) epoxy - led to a resurgence of interest in the design and
analysis ofbondedjoints. This has been particularly so for highly loaded joints between
composites and stepped titanium plates because of a desire to avoid the reduction in
structural efficiency associated with bolt holes in such brittle materials. However, there
has been a subsequent recognition that such all-bonded, thick, fibrous-composite
structures are all but impossible to repair to the original structural capabilities.
Therefore, the extra-efficient composite structures are likely in the future to be confined
to one-shot applications, such as missiles, while re-useable vehicles like manned aircraft
are likely to make more use of mechanical fasteners at areas of high load intensity.
Nevertheless, the material presented in this chapter covers both possibilities.
Our discussion starts with a description of the various joint geometries, each of
which is appropriate for a different load intensity. There is a discussion of the pitfalls of
trying to correlate the measurements of bonded test coupons designed to force a failure
in the adhesive with the performance of structural bonded joints in which the objective
is to ensure that the adhesive itself never fails. The elimination of induced peel stresse~
by tapering the edges of the adherends is explained. Also, the distinction between
repairable and nonrepairable bonded structures is identified. Some adhesive bonding is
used primarily to increase structural efficiency by the use of laminated structures rather
than to transfer loads. Further applications of adhesive bonding, as between skins and
stiffeners, transmit very little load through the bonds (except at the ends of the
stiffeners) provided that the structure remains intact, but the bonds become very
important with respect to containing any local damage. One pleasing characteristic of
adhesively bonded joints is that both testing and detailed analyses have confirmed that
they have considerably greater tolerance to local flaws than is generally recognized.
Chapter 6 Adhesively Bonded Joints for Fibrous Composite Structures 175
6.2 ADHESIVELY BONDED JOINT CONFIGURATIONS
Adhesively bonded joints are used in many different configurations, of which the most
common are shown in Figure 6.1. Their relative uses are placed in perspective by
Figure 6.2 from which it is clear that there is no purpose in using an unnecessarily
complex joint for the lower load intensities. Conversely, it is obviously hopeless to
expect that the simpler and cheaper configurations could ever sustain high load levels,
no matter how much quality workmanship was employed in their manufacture.
REVERSE-TAPERED DOUBLE-LAP JOINT

BONDED DOUBLER (ALSO APPLICABLE TO SINGLE-LAP JOINTS, ETC .. )
UNSUPPORTED SINGLE-LAP JOINT

DOUBLE-STRAP (BUTT) JOINT
SINGLE-STRAP JOINT
TAPE RED-STRAP JOINT
TAPERED SINGLE-LAP JOINT

(ALSO APPLICABLE TO DOUBLE-LAP JOINTS, ETC.)t
r ~
t--=:::::!:==------~~
}
STEPPED-LAP JOINT
DOUBLE-LAP JOINT SCARF JOINT
Figure 6.1 Aadhesively bonded joint types
BONDED
JOINT
STRENGTH
#V
.j} BENDING OF ADHERENDS
~ ·----~~~~~--~DU~E!TO~E~C~C~EN~TR~IC~L~O~AD~P~A~TH
SINGLE-LAP JOINT
ADHEREND THICKNESS
Figure 6.2 Relative uses of different bonded joint types

The simplest and most common joint is the bonded doubler which is often not even
considered to be a structural joint. Yet, as is shown in Figure 6.3, the load transf~r
through the adhesive in load-sharing bonded doublers is just as intense, at the ends, as
that in a full-transfer bonded joint. Perhaps the equivalence of these situations has
escaped attention because bonded doublers are usually confined to thin-gauge structure
whereas bonded joints can also be applied to much thicker and more heavily loaded
structures. Bonded doublers are sometimes used to provide local thickening for
countersunk fasteners in thin skins, and to provide resistance to acoustic fatigue of thin
metal structures. They have also been used, particularly at de Havillands, in England, at
the Fokker works in the Netherlands, and at SAAB in Sweden, to increase the structural
efficiency of stiffened structures subject to shear or compressive loads. In that context,
stiffener flanges bonded to the skins act much like a doubler as far as the adhesive is
concerned.
• SAME ADHESIVE STRESS DISTRIBUTION IN EACH CASE
ADHESIVE~
SHEAR
, -=rl%
STRESS, f
~------~~----~--------~----
DOUBLER
ADHESIVE~
SHEAR
STRESS, 1'
• SAME MAXIMUM ADHESIVE SHEAR STRAINS FOR SAME ADHERENDS AT SAME LOAD
Figure 6.3 Similarity between adhesive stresses in doublers and joints
The unsupported single-lap joint can never be as strong as members that have been
bonded together. Nevertheless, with an adequate overlap-to-thickness ratio (of at least
50 to 1 and preferably as much as 100 to 1) the gentle transverse deflections under
tensile load can relieve the eccentricity in load path and develop structural efficiencies
high enough to compete with the reduced net section of mechanical splices.
Unsupported single-lap joints are also efficient for in-plane shear transfer but should
never be used for compressive loads; the initial eccentricity becomes progressively
worse as the load is increased and the joint should be stabilized in such cases.
The single-strap joint in Figure 6.1 barely qualifies as a joint - a better description
would be a severe built-in stress concentration, unless a transverse support is provided
in the middle of the splice plate. The abrupt eccentricity where the skins butt together is
necessarily associated with a high bending moment in the middle of the splice plate. In
addition, the same area is subject to high adhesive peel stresses and an associated
tendency to delamination of fibrous composite adherends. The analysis of peel stresses
in that and other joints is given in Reference 4.
Thicker single-lap joints need the ends of the adherends tapered, as in Figure 6.1,
primarily to prevent a premature failure by induced peel stresses before the full
potential shear strength of the adhesive can be developed. A secondary benefit is an
actual increase in shear strength, provided both ends are tapered, because of a reduction
in the severity of the peak in the shear stress distribution.
It is appropriate at this point to introduce the concept of the three distinct failure
modes by which adhesively bonded joints can fail. The strongest joints do not fail in the
adhesive at all; they fail outside the joint area at a load level as high as 100 percent of
the strength of the adherend. The next strongest joints are those in which the load that
can be developed is limited by the shear strength of the adhesive. The weakest failure
mode is associated with failure of the adhesive under peel loads, or with the even more
premature failure by delamination of fibrous composite adherends.
The strength of the joint is proportional to the laminate thickness (to the first power)
if it fails outside the joint under direct tension, compression, or in-plane shear in the
absence of bending. The shear strength of the adhesively bonded joint, on the other
hand, is only proportional to the square root of the laminate thickness. The peel strength
drops off even faster, being proportional to the quarter power of the laminate thickness.
These effects are explained in Figure 6.4. Consequently, seemingly similar adhesively
bonded joints will fail by different modes at quite different structural efficiencies,
primarily as a function of the thickness of the joint. The reason for this is that, despite
the ability to vary all other dimensions of the joint, adhesive layers are most efficient
only throughout the thickness range of 0.005 to 0.010 inch (0. 125 to 0.25 mm).
Thicker bond lines could theoretically develop higher strengths, but it is virtually
impossible to make them without intolerable levels of flaws or porosity. Heat-cured
adhesives reach such low viscosity during cure that they will flow out of thick gaps
between the adherends. On the other hand, locally thickened adhesive layers at the
ends of the overlap have been shown to very effective for paste-adhesives that are so
thixotropic as to remain in place while curing. Simply tapering the outer adherends
locally, down to a tip thickness of 0.020 ± 0.010 inch (0.5 ± 0.25 mm) with a taper
angle of one-in-ten, is a very effective means of suppressing premature failures of
adhesively bonded joints; such tapering virtually prevents failures by induced peel
stresses.
Laminates of somewhat greater thickness than those for which single-lap joints are
adequate require two bond surfaces to transfer the strength of the adherends. The
transition in thickness is not exact, being a function of the adhesive as well as of the
thermal environment, but typically is of the order of0.06 to 0.07 inch (1.5 to 1.75 mm)
for aluminium alloy adherends or for near quasi-isotropic carbon-epoxy laminates.
Lower thicknesses would apply for stronger steels or unidirectional composite members.
The optimum overlap-to-thickness ratio for double-lap or double-strap joints (see Figure
6.1) is of the order to 30-to-1 for ductile adhesives used with the structure of subsonic
transportaircraft. Despite the lack of any obvious gross eccentricity in load path, such
joints are also subject to peel strength cutoffs, as explained in Figure 6.5.
Merely tapering the outside of the splice plates (while leaving the basic members at
a uniform thickness), however, will not create any additional shear strength in the joint.
(Neither will it usually cause any decrease in the shear strength, but it could if the
overlap were too short.) The tapered strap joint (Figure 6.1) employs deliberate
thickening of the splice plate at its middle to achieve simultaneously the alleviation of
peel stresses and an increase in shear strength of the adhesive joint. The uniform
double-strap joint (with locally tapered ends as explained earlier) is limited to the
joining of about 118 to 3/16 inch (3 to 4.5 mm) of aluminium alloy or equivalent. The
use of the tapered splice straps might push that as high as 1/4 inch (6.35 mm) without a
significant increase in manufacturing difficulty.
NEEDS
TAPERED
EDGES
STRENGTH
THICKNESS
BOND STRONGER
THAN ADHERENO
ADHEREND STRONGER
THAN BOND
Figure 6.4 Relative severity of adhesive shear and peel stresses
STRESSES ACTING ON OUTER AOHEREND B.
-~J----r===~J
_[_·-·-·-·-~~
A.
~
-·--·-·-·-·---·-·-·-·-·-·-·-·-· -
c.
A, B, AND C INDICATE FAILURE SEQUENCE
Figure 6.5 Peel stress failure of thick composite joints

Only the stepped-lap or scarf joints should be considered for adhesively bonding
members more than 114 inch (6.35 mm) thick. The idea that the theoretical scarf joint
can always develop the entire strength of the laminate is usually violated by the need
for a finite thickness at the tips of the tapers. One researcher has even calculated that a
tip thickness of only 0.001 inch (0.025 mm) is equivalent to a stress concentration of 25
percent in the elastic adhesive shear stress distribution. For this reason, the analysis of
scarfed joints is often better performed by a stepped-lap joint program rather than one
for scarf joints. 2 This point is discussed in Reference 5, which explains also when a
stepped-lap joint should not be treated as an approximation of a scarf joint. The key
difference is that the shear strength of a scarf joint continues to increase as the scarf
angle is decreased, while the strength of a stepped-lap joint with a fixed number of steps
does not continue to increase indefinitely as the overlap is increased. The prime factor
in determining the strength of a stepped-lap joint is the number of steps, provided that
some gross error has not been made in the dimensioning of some single critical step.
6.3 ADHESIVE STRESS-STRAIN CURVES IN SHEAR
The basis of the analysis of the shear strength of adhesively bonded joints is the
nonlinear stress-strain curve of the adhesive layer in shear. This refers to the
characteristics, as measured on the napkin-ring or thick-adherend test specimens, that
are described in Figures 18 and 19 of Reference 6. Such characteristics are illustrated in
Figure 6.6 for both ductile adhesives (as used on subsonic transport aircraft) and brittle
adhesives (as needed for the much higher service temperatures developed in the vicinity
of engines or by supersonic fighters). It is shown in Reference 7 that the shear strength
of adhesively bonded structural joints can be expressed uniquely by the strain energy to
failure per unit bond area of the adhesive layer rather than by any of the individual
properties such as peak shear stress. Therefore, even for the brittle adhesive in Figure
6.6, the majority of the load transfer is accomplished by the nonlinear behaviour; in the
case of the ductile adhesive, the contribution of the linearly elastic behaviour may be as
little as 10 percent. Consequently, any perfectly elastic analysis would require a very
large correction factor in order to correlate with test results.
The adhesive stress-strain curves are quite sensitive to the environment, as can be
seen from Figure 6. 7. Note, however, that the area under each of those curves is much
more nearly the same than are any of the individual properties such as peak shear stress
or strain to failure. Consequently, the strengths of structurally proportioned joints would
be insensitive to the environment, even though measurements on the standard lap-shear
test coupons would suggest otherwise. The pitfalls of trying to relate the performance of
bonded structural joints to results from test coupons are so substantial that Reference 6
was written just on that issue. Indeed, there is almost a complete lack of any one-to-one
~ The finite-difference analysis techniques developed by Oplinger (see Reference 8)

include adherends that are tapered to a finite, rather han zero, tip thickness, so they can
be applied directly.
correlation, even though certain test data are obviously needed as the basis for design.
A strong case is made in Reference 6 that a totally different approach is needed to
characterize adhesives in such a way that a rational selection can be made for any given
application. Some of the reasons for this will become evident from the discussions in
this article.
10 ROOM-TEMPERATURE
CHARACTERISTICS
TORSION·RING
SHEAR STRESS
(KSI)
0.5 1.0 1.5 2.0

SHEAR STRAIN (lN./IN.)
Figure 6.6 Adhesive stress-strain curves in shear (10 ksi = 70 mn/m2)
10r--------------------------------.
·67"F
SHEAR STRESS
+180°F
(KSI)
4
NYLON·EPOXY ADHESIVE (250"1' CURE
oL---~--~----~--~----~--~--~
0
SHEAR STRAIN (lN./IN.)
Figure 6.7 Effect of temperature on adhesive stress-strain curves in shear (10 ksi = 70 rnn!m2 )
Various linear and nonlinear adhesive models are shown in Figure 6.8. The elastic-
plastic model passes through the same failure stress and strain and has the same strain
energy to failure as the actual stress-strain curve. It permits accurate predictions of the
ultimate strength of bonded joints, but does not give an accurate representation of the
internal stresses associated with much lower loads. The perfectly elastic model is to be
preferred at very low load levels. Improved accuracy can be obtained by use of a
modified model for intermediate loads, as shown. The peak stress and strain in the
adhesive are again matched, and the elastic portion of that model is selected by equating
areas again. Theoretically, a new model should be prepared for each load level, but that
is obviously impractical. Considerable simplifications would result if the single bilinear
characteristic shown in the Figure could be used throughout. Unfortunately, the
complexity of the associated mathematics is increased greatly. Nevertheless, such an
explicit closed-form solution for uniform double-lap joints (Reference 9) has
established that any two-straight-line representation having the same failure stress and
strain as well as the same strain energy to failure would predict precisely the same joint
strength. A refinement such as the bilinear model would therefore not enhance the
accuracy of ultimate strength predictions but would simplify the computation of the
internal stresses associated with specified lower loads.
MOOEL FOR INTERMEDIATE LOADS
~
LASTIC.PLASnC
EI...ASTIC.PLAS TIC MODEL
ADHESIVE
SH EAR
STRESS,
-r
ADHESIVE SHE AR STRAIN, 7
Figure 6.8 Representations of adhesive nonlinear shear behaviour
Adhesive characteristics can also be affected by porosity, which occurs as a result of the
bond layer being slightly thicker than nominal; gross bond flaws result if the gap
between the adherends becomes so large that the adhesive is not retained by either
capillary action or external dams. Reference 10 contains examples of stress-strain curves
for porous bonds in both ductile and brittle adhesives. Interestingly, it is shown there
that naturally occurring porosity does not usually result in the failure of those portions
of the bonds that would be suspect on the basis of ultrasonic inspection. Porous bonds,
being thicker than nominal, are also softer. Therefore, they transfer some of the load
they should have transmitted to adjacent thinner bonds which thus become more highly
loaded than they would be in a uniform adhesive layer. Actually, porous bonds are
rarely created in areas of naturally high load transfer, as at the ends of a bonded
overlap; rather, they tend to occur away from those edges, in areas in which there would
be little load to transfer even if the bond were not porous there.
The considerable straining of the ductile adhesives beyond the knee in their stress-
strain curves is typified by Figure 6.6. However, it is not considered good design
practice to use all that strength for normal design conditions, even if the minimum strain
is adequately restricted. Adhesives do not actually deform plastically in the manner
suggested in Figure 6.8. It is now known that this is merely a convenient mathematical
model. What actually happens is explained in Figure 6.9. A series of microcracks
(hackles) develops at 45° to the plane of the adhesive layer under the action of the
tensile component of the shear load. The size and density of these hackles increases
with applied load. The actual softening of the adhesive layer is caused by the elastic
bending of the ligaments tying the adherends together. Unloading does not follow the
classical linear offset line of ductile metals. Instead, there is almost complete recovery
to the original reference point. The cracks do not heal themselves, of course; the
adhesive layer remains permanently softened. Eventually, the hackles can link up if the
load is high enough, resulting in total failure. However, the real feature governing the
design process is the setting of the design limit load to not exceed the knee in the stress-
strain curve because permanent damage does not occur until that point is exceeded. In
that event, since the bonded joint strength, for long-overlap joints, is proportional to the
square root of the adhesive strain energy in shear, design ultimate load must not strain
the adhesive beyond 1.625 times the strain at the knee, as explained in Figure 6.1 0.
LOAD
<=
LOADS BELOW ELASTIC LOADS BEYOND ELASTIC LOADS WELL BEYOND ELASTIC
CAPABILITY OF ADHESIVE; CAPABILITY OF ADHESIVE; CAPABILITY OF ADHESIVE;
NO HACKLES HACKLES BEGIN TO FORM HACKLES SPREAD AND
EVENTUALLY LINK UP
Figure 6.9 Tensile hackle failures associated with shear load on adhesive
DESIGN LJMIT LOAD MUST NOT

STRAIN ADHESIVE BEYOND KNEE
IN STRESS·SlRAIN CURVE
MAXIMUM ADHESIVE STRESS AND

ADHESIVE STRAIN ASSOCIATED wrTH FALURE
SHEAR OF AOHERENOS OUTStOE THE JOIHT.
STRESS , ~ AOHESrvE SHOULD WITHSTAND A
50-PERCEHT HIGHER OAD BEFORE
FA.II.URE. THIS RE.SRVE IS NEEDED
FOR DAMAGE TOLERANCE; AN
CORRESPONDiNG ELASTlC· ADHESIVE 800 WEAKER THAN THE
PLASTIC MODEL FOR DESIGN AOtiEA.ENDS COULD ACT LIKE A
UL.TIM.ATE LOADS WEAK ~INK FUSE
MODEL MATCHES TRUE MODULUS

LOAD .. .JAREA UNDER CURVE
u2sr. AOHESJVE SHEAR STRAIN, T
NOTE THAT DESIGN PROCESS MUST ACCOUNT FOR NONLINEAR ADHESIVE BEHAVIOR,
BUT A PRECISE STRESS-SRAIN CURVE IS NOT MANDATORY. AN APPROXIMATION,
BASED ON A SIMILAR ADHESIVE, Will USUALLY SUFFICE.
Figure 6.10 Modeling of adhesives for design of shear joints
The considerable nonlinear adhesive behaviour is available for load redistribution due
to local manufacturing imperfections of various forms. The strength of such bonded
joints in practical structures would be greatly diminished if this reserve of strength were
not available.
Restricting the peak strain in the adhesive to the knee in the curve for normal
conditions has the effect of limiting the adherend thicknesses for which the simpler
joint configurations are satisfactory; the use of carefully designed stepped-lap joints,
even with this restriction, still permits the transfer of substantial loads. Such joints
between boron-epoxy or carbon-epoxy laminates and titanium step plates keep the skins
attached to the tails and wings of several modem fighter aircraft.
Nevertheless, experience in Australia, at the ARL (now AMRL) and RAAF in
bonded composite repairs of cracked metallic structures has shown that it is possible to
strain the adhesive to twice the strain at the knee in Figure 6.10 for frequently occurring
fatigue loads. There is not so much )attitude in designing bonded patches as there is at
the initial design stage. Patch designs are often constrained even more by an irrational
resistance to the removal of any additional material to make stronger (multi-step)
bonded joints than would be possible otherwise. Again, the RAAF has been a pioneer
in confirming the benefits of rejecting such an attitude.
6.4 THE NON-UNIFORMITY OF LOAD TRANSFER IN ADHESIVELY BONDED

JOINTS
The most important single thing to learn about the design and operation of structural
adhesively bonded joints is that the load transfer is not, and must not be, uniform. This
requires what must seem to those intent on optimizing joint proportions and performance
mathematically to imply a need for what is to them an unnecessarily long bonded

overlap. Nevertheless, the need is very real; it is explained in Figure 6.11. In the short-
overlap test coupon (Figure 6.11 b), the adhesive is seen to be uniformly stressed and
nearly uniformly strained. Indeed, that is the usual intent of the use of such a specimen
- to know the exact state of stress and strain of the adhesive when it fails. No such
claims could be made for the adhesive in the much longer overlap in Figure 6.11 a.
There is considerable variation in both the adhesive stress and adhesive strain
throughout the bonded overlap.
c B
ADHESIVE
SHEAR
STRAIN
'
G
~ D
II :{
(A) REALISTIC OVERLAP FOR STRUCTURAL JOINT
ADHESIVE
SHEAR
STRAIN
)
DO II
ADHESIVE
SHEAR
STRESS
(b) SHORT OVERLAP TEST COUPON
Figure 6.11 Nonuniform stresses and strains in bonded joints
The importance of the variation is that, provided the minimum adhesive shear strain at
point A is sufficiently low, the bonded joint will be able to survive more than one or
two load cycles or more than a few minutes of high sustained load. The short-overlap
joint is incapable of doing that because the minimum adhesive shear strain, at the
middle of the overlap, is so high that the joint would fail rapidly because of creep
rupture. That is not to imply that there would be no adhesive creep, under sustained
load, between the adherends at points F and Gin the long overlap joint. Obviously, the
adhesive shear strain at points C and B is so high that creep must inevitably be there.
How then does such a bonded structural joint not fail in the same manner as a short-
overlap coupon? The reason is that while the creep occurs, it does not accumulate.
Consider what happens when the load is removed after one hour or 8 hours, or even
15 hours which represents the longest flight durations of commercial transport aircraft.
It is fair to assume that there has been no creep between points D and E across the
middle of the bonded overlap, and that neither adherend has changed length due to creep
between points D and F or E and G. Otherwise there would be no long-life bonded

structural joints. Yet what appears to have been creep has been measured
experimentally between points F and G as long as the load was maintained. What has
happened is that the visco-elastic behaviour of the adhesive in the vicinity of point C
causes an elastic redistribution of the stresses within the intervals FD, DE, and EG.
Thus, when the load is removed, elastic residual stresses are induced in the adherends
and, as these try to relieve themselves, they push the adhesive at point C back towards
its original unstrained state. There is no mechanism to accomplish that in the short-
overlap joints because of the creep that would have occurred under load between the
equivalent of points D and E.
It should be clear now that in the design of adhesive-bonded joints it is more
important to limit the minimum adhesive shear strain than it is to limit the maximum,
except as noted above. The deep elastic trough between the plastic load transfer zones J
in Figure 6.11 should not be regarded as an inefficiency to be eliminated by "improved"
design. Much research remains to be done to establish a scientific basis for setting that
minimum value of adhesive shear strain or stress. As a matter of record, the design
procedures employed on the Primary Adhesively Bonded Structure Technology
(PABST) program (Reference 11) were based on restricting the minimum shear stress
to no more than 10 percent of the maximum at the ultimate load level. That
corresponded with an even smaller fraction for the normal operating stresses. Those
double-strap joints that are characterized in Figure 6.12 performed without failure under
both full-scale testing and artificially severe coupon testing for four years in a hot/wet
environment under slow-cycle testing. The need for slow-cycle testing -half an hour or
more per cycle - arises because even the short-overlap test coupon appears to have an
infinite fatigue life if tested at 30 cycles per second (the load is being removed so
rapidly that the adhesive has no opportunity to creep).
The transition between the extreme behaviours illustrated in Figure 6.11 is
explained in Figure 6.13 as a function of the bonded overlap for double-lap joints. The
joint strength is initially proportional to the short overlap, as at A. Then, as the overlap
is increased still further, the strength increases only very slightly as the elastic trough is
developing, at B. Finally, no matter how much longer the overlap is made, as at C, the
joint strength remains constant and so does the maximum adhesive stress and strain.
The minimum adhesive stress decreases asymptotically towards zero, and the load
transfer zones keep the same width but move steadily further apart.
The design of such joints is thus reduced to the simple task of identifying the overlap
for which that minimum stress has become sufficiently low. Any further increases in
overlap cannot increase the joint strength. This process is explained in Figure 6.14.
(The results of such calculations are given in Figure 6.1 0. which was developed for the
adhesive bonding of aluminium alloy adherends during the PABST program.) The
results can be summarized by the simple approximation that each bonded overlap is
roughly 30 times as long as the thickness of the inner adherend. Much the same ratio
would apply also to typical cross-plied carbon-epoxy laminates. The slightly shorter
overlaps, for essentially the same strength, that were given as optimum designs in Figure
16 of Reference 7, Figure 11 of Reference 12, and Figure 3 of Reference 9 should now

be considered superseded. The 3/A. in Figure 6.14 replaces an increment of 1/A. used in
those references before the role of the elastic trough had been fully understood. Also,
the conservative 6/A. increment used during the PABST program to establish the design
overlaps as a function of adherend thickness has since been replaced by 51 A, now that
credit is given for the small load transferred through the elastic trough in the adhesive
shear-stress distribution.
CENTRAL SHEET THICKNESS 0.040 0.050 0.063 0.071 0.080 0.090 0.100 0.126
t;(IN.)
-·
SPLICE SHEET THICKNESS 0.025 0.032 0.040 0.040 0.050 0.050 0.063 0.071
t 0 (IN.)
RECOMMENDED OVERLAP 1 1.21 1.42 1.68 1.84 2.01 2.20 2.39 2.84
£(IN.)
STRENGTH OF 2024 · T3 2600 3250 4095 4615 5200 5850 6500 8125
ALUMINUM (LB/IN.)
POTENTIAL ULTIMATE 7699 8562 9628 10,504 10,888 11,865 12,151 13,910
BOND STRENGTH (LB/IN.)2.3
'BASED ON 1&0DF DRY OR 14IJOF/100-PERCENT RH PROPERTIES NEEDING LONGEST OVERLAP.

VALUES APPLY FOR TENSILE OR COMPRESSIVE IN-PLANE LOADING. FOR IN.PLANE SHEAR LOADING.
SLIGHTLY DIFFERENT LENGTHS APPLY.
28ASED ON - liCJOF PROPERTIES GIVING LOWEST JOINT STRENGTH AND ASSUMING TAPER OF OUTER
SPLICE STRAPS THICKeR THAN 0.050 IN. STRENGTH VALUES CORRECTED FOR ADHEREND
STIFFNESS IMBALANa.
3FOR NOMINAL ADHESIVE THICKNESS 11 • 0.0051N. FOR OTHER THICKNESSES, MODIFY STRENGTHS
IN RATIO ./•1011116.
~d ct'
P- JJ=~=Q_~,=----4~1~Ct~.:=!,~t-P
Figure 6.12 Design overlaps used for PABST skin splices
p
B !MAXIMUM (CONS~
ANT)
zp - ----r:=====::::J: : :
._l
1 !STRENGTH ZONE CONSTANT-

A (TRANSITION! WIDTH
LOAD-
~ TRANSFER
ZONES
EFFECT OF LAP LENGTH ON
ADHESIVELY BONDED JOINT STRENGTH (b) INTERMEDIATE OVERLAP
2P -L..l ·==::::J:
: :
-_pp
[]
(a) SHORT OVERLAP (c) LONG OVERLAP
Figure 6.13 Influence oflap length on bond stress distribution

t,
ADHESIVE
SHEAR STRESS
DISTRIBUTION
t--------I~----E-LA-ST-IC-T-RO_U_G_H----~ L T,
~'""'iii
• PLASTIC ZONES LONG ENOUGH TO TRANSFER ENTIRE ADHEREND STRENGTH

• ELASTIC TROUGH WIDE ENOUGH TO PREVENT CREEP AT ITS MIDDLE
• DESIGN OVERLAP = I 01 + I 02 + (5/.t)
• CHECK ON ADEQUATE STRENGTH BY NOT EXCEEDING MAXIMUM ADHESIVE SHEAR STRAIN
Figure 6.14 Design of double-lap bonded joints
6.5 DOUBLE-LAP AND DOUBLE-STRAP JOINTS
The pioneering analysis of adhesively bonded joints is that by Volkersen (Reference

13). That linear analysis, and its nonlinear development by this chapter's author in
Reference 12, rely on continuum mechanics analyses, and are expressed in terms of one
mathematical variable with no account of the minor variations in stress through the
thickness. That level of analysis had led to an extremely thorough understanding of the
stress state within such bonded joints. References 14 and 15 provide comprehensive
illustrated explanations of the behaviour of such joints. They cover the origin of the
nonuniformity in the load transfer across balanced double-lap joints, and the strength
losses associated with adherend stiffness imbalances or thermal mismatches, and the
grossly dissimilar behaviour of short-overlap test coupons and long-overlap structural
joints. The consequences of refining such analyses to account for variation in the
adhesive stresses across the thickness of the layer are discussed in Figure 4 of
Reference 15, where there is an explanation that justifies neglecting such effects in
actual joints because of the naturally occurring fillets in the adhesive.
There is also a significant shear-lag effect that varies the adherend stresses across the
splice plate thickness over the point where the skins butt together in double-strap joints.
This effect has been investigated in depth in relation to the use of adhesively bonded
boron-epoxy patches to retard the growth of cracks in metallic structures (see Reference
16). The effect was also sufficiently pronounced during the fatigue testing of aluminium
alloyadherends during the PABST program to result in a strong bias towards a failure in
the middle of the splice plate rather than in the central sheet at the other end of the bond
where the adherend stresses were nominally equal. This led to an increase by half a
gage in the thickness of the splice plates, with respect to a nominal thickness half that of
the basic member. The same kind of reinforcement is recommended even more strongly
for bonding fibrous composites, because of the low interlaminar shear stiffness of th~
resin matrix.
The key characteristics of the state of adhesive stress and strain in an adhesively
bonded double-lap joint are the peak adhesive shear strain, the minimum adhesive shear
strain, and the peak induced adhesive peel stress. The latter will probably delaminate
the composite adherend, as shown in Figure 6.5, before the adhesive could peel apart.
The detail modifications (Figure 6.15) made to eliminate the peel-stress problem when:
necessary are quite simplr yet effective. The tip thickness of the splice plate should not
exceed 0.020 inch (0.50 riun) for composites. The local thickening ofthe adhesive layer
makes the glue more flexible there, reducing the peel stresses by half or more and also
reducing the peak adhesive shear strain without any loss in the shear strength of the
joint. The upper limit on layer thickness is needed because, otherwise, heat-cured
adhesives would run out of that region under capillary action unless there were special
dams. (There is no such problem with room-temperature-curing paste adhesives, for
which the technique of thickening the adhesive layer at the ends of the overlap has
increased joint strengths by as much as a factor of I 0 for the fittings in tubular space
structures.) The degree of such thickening is not at all critical, as is explained in
Reference 15. The adherend tapering on the outside is not critical either, and this is
shown in Figure 6.16. The critical condition is then moved to the other end of the joint
and remains constant because there are no modifications there.
5 ~· -2-.
c·-----· II =
ADHESIVE LAYER
I
0.10T00.201N.
SLOPE OF 11N 10
0.020 TO 0.040 IN. FOR ALUMINUM,
0.010 TO 0.030 IN. FOR FIBER-POLYMER COMPOSITES
Figure 6.15 Tapering of edges of splice plates to relieve adhesive peel stresses
.t
ADH IJIVE THICK NESS CO NST ~NT
At 0.001 THAOUGHDUT
0.011
II Y·•· ·'
ADHESIVE STRAIN AT RIGHT END OF

JOINT DECRESES WITH MORE TAPER
Figure 6.16 Insensitivity of adhesive bonded joint strength to modifications at one end of joint
only
Once any possible premature failures by induced peel stresses have been eliminated, it
is possible to consider the adhesive shear stresses alone. Figure 6.17 shows the
maximum and minimum adhesive shear stresses as functions of bonded overlap.
(Figures 2 to 4 of Reference 6 show the additional influence of adherend thickness.)
Note that there is an abrupt transition in the adhesive behaviour between those overlaps
at points A and B. The minimum adhesive shear strain is almost as high as the
maximum for short-overlap joints and those configurations should be used only for
certain test specimens and never for structural bonds. To the right of those transitions,
the peak adhesive shear strain for the thin and moderately thick adherends is restricted
to values insufficient to permit failure by the limited strength of the adherends that
would inevitably fail first.
The design exercise is thus reduced to the establishment of that overlap, to the long
side of such a transition, for which the minimum adhesive shear strain is sufficiently
low. Any further increase in overlap could not add to either the strength or the life of
such a joint. Those calculations form the basis for the recommendations in Figure 6.12.
(The computations on which Figures 6.12 and 6.14 are based were actually performed
for aluminium alloy-bonded construction but, since the adherend modulus E has a value
that is representative of the range of practical fiber patterns for HTS graphite-epoxy
shown in Figure 21 of Reference 17, the overlaps are also representative of those for
fibrous composites.) Figure 6.12 can be summarized by the statement that the overlap
should be about 30 times the central adherend thickness. Varying the adhesive ductility
or the strength of the fibrous reinforcement has much more of an effect on the maximum
thickness of adherends that should be bonded than on this overlap-to-thickness ratio.
ADHESIVE
SHEAR
ADHESIVE SHEAR STRESS DISTRIBUTIONS
STRAIN, Y
ELASTIC ADHESIVE
r...
SHEAR STRAIN
r.,.
BONDED OVERLAP, i
LOAD TRANSFER
ZONES MOVED
FORMATION FURTHER APART
OF ELASTIC
TROUGH
BONDED OVERLAP, l
Figure 6.17 Influence of overlap on maximum and minimum adhesive shear strains in bonded
joints
The reason there is an upper thickness cutoff in Figure 6.12 can be understood from
curves C in Figures 2 to 4 of Reference 6 for 0.25-inch (6.35-mm) thick adherends.
There is no abrupt transition in adhesive behaviour the way there was in curves A and B
for the 0.063- and 0.125-inch (1.6- and 3.1-mm) thick adherends, no matter how long
the overlap, and consequently an imperfection in the bond between the thicker
adherends could cause a catastrophic unzipping even if analysis suggested that there was
adequate strength for a nominally perfect bond. The calculations on which Figure 6.12
is based were performed for that environment, usually the hot/wet one, which needs the
longest overlap. The actual calculations are particularly simple, as shown in Figure
6.14. The adhesive properties are represented by the peak shear stress rp, the elastic
shear modulus G, and the thickness, 1]. The suggested check to determine that bond
strength is adequate to fail the adherends outside the joint implies an upper limit on the
adherend thickness for which this simple joint configuration should be used. Although
Figure 6.14 implies an equal transfer of load through the two plastic zones at the ends
of the joint, the adherend tapering shown in Figure 6.15 for thicker adherends alters the
widths of those plastic zones in such a manner that the combination of the two zones is
essentially constant; this is because the total load is still limited by the adherend
strength. Therefore the technique in Figure 6.14 still applies, even though the actual
adhesive stress and strain distributions would be modified, as shown in Figure 6.16.
6.6 SINGLE-LAP JOINTS
The design of single-lap joints in fibrous composite structures is actually easier than for
double-lap joints, even though the mathematical analysis of the former is more difficult.
Goland and Reissner were the first to account for the out-of-plane bending associated
with the eccentricity in the load path in what is now considered one of the all-time
classic bonded joint analyses (see Reference 18). Along with many other researchers,
the author of this chapter tried to improve their estimate of the bending moment at the
end of the overlap (which is the critical boundary condition for all her variables in the
analysis) but, as pointed out by Oplinger in Reference 19, introduced a mistake of his
own. Both of the mistakes were finally corrected in Reference 20, which established
that the simplified analysis by Goland and in Reissner that did not satisfy the
requirement of zero bending moments at the ends of the adherends, was nevertheless
numerically realistic for practical joint geometries. In Reference 21, their linear elastic
analyses were extended to nonlinear adhesive behaviour in the form of an elastic-plastic
model.
Figure 6.18 illustrates the key to the difference between the behaviour of
unsupported single-lap joints and double-lap bonded joints. The single-lap joints incur
significant bending moments in their adherends in conjunction with the average
(uniform) stress component due to the applied loads. The presence of such bending
stresses requires such long overlaps to permit the adherends to bend gently rather than
abruptly that the possible criticality of the adhesive in shear becomes secondary.
Actually, the same load path eccentricity induces more severe peel stresses in the
single-lap joints than occur in double-lap joints. Again, beyond the tapering of the
adherends shown in Figure 6.13, the only mechanisms with which to reduce the peel
stresses are to increase the overlap or to provide transverse support at the ends of the
overlap to react against the eccentricity in load path. Such supports would be
absolutely necessary to stabilize single-lap joints under compressive shear loading.
Figures 6.19 and 6.20 show how the structural efficiency of adhesively bonded
single-lap joints varies as a function of joint geometry for ductile and brittle adhesives,
respectively. These composite adherends are assumed to have no peel stress relief and
therefore show considerable decreases in strength in the right hand sides of the figures.
Not a single failure of the ductile adhesive in shear is predicted in Figure 6.19, which is
why the design of such joints is so simple. The brittle adhesive in Figure 6.20 is weaker,
as well as more brittle, and a small area of shear failures is predicted in the upper right
comer. The effect of the use of a brittle adhesive is to decrease the thickness of
adherends that can be bonded well by single-lap joints. For the thin adherends, in the
left of the figures, the efficiency is limited by adherend bending only and is strongly
dependent of the Ut ratio. A value of 80, corresponding with a structural efficiency of
90 percent, was adopted for the PABST program. Note that for the typical fit ratio of 8
used in the standard lap-shear test coupon with a half-inch (12.7-mm) overlap and
aluminium alloy adherends 0.063 inch (3.2 mm) thick, the structural efficiency is
limited to less than 40 percent by that abrupt eccentricity.
LOW LOAD LEVEL
DEFORMATIONS OF ADHERENDS UNDER "INTERLAMINAR" FAILURE OF FILAMENTARY

MODERATE LOAD COMPOSITE ADHERENDS
ff PLASTIC HINGES
- Cl ::J_c==::JI-
===:::::'~~~
HIGH LOAD LEVEL (METAL ADHERENDS) PERMANENTLY DEFORMED (METAL) ADHERENDS
AFTER FAILURE OF ADHESIVE
Figure 6.18 Failure of single-lap bonded joints (brittle and yielding adherends)
Of special significance is that neither Figure 6.19 nor Figure 6.20 would predict that the
failure of such a test coupon would in any way be influenced by the shear properties of
the adhesive. Yet that coupon is the very specimen that is most widely (mis)used in the
belief that it does characterize the shear strength of adhesives and permit meaningful
comparisons. The specimen also is the basis of many MIL specifications. The reasons
for its continued use for such purposes have not been clear for at least two decades.
although it should be acknowledged that that specimen still serves as a practical quality
control coupon. In Reference 6, it is explained that the apparent shear strength
indicated for adhesives by that specimen is influenced strongly by the adherend
thickness and material, and that invariably the failure is by peel rather than by shear.
Actually, the use of such data to aid in adhesive selection has become extremely
difficult and misleading for newer, stronger adhesives. The author suggests in
Reference 6 that a more meaningful characterization of adhesives for selection would
be a rating in which adhesive A could fail a given adherend material outside the joint up
to a thickness of 0.05 inch ( 1.25 mm), say, while the stronger adhesive B night fail the
same material up to a thickness of 0. 10 inch (2.5 mm). In neither case would there be a
failure of the adhesive because bonded joints should be designed so they do not fail.
Such an approach to characterizing adhesives would be unambiguous and remove all of
the. difficulties of trading off the higher peak shear-stress-at-failure of brittle adhesives
against the much greater strains to failure of ductile adhesives.
1.0
0.9
0.8
0.7
0.6
ADHESIVE PEEL
JOINT FAILURES (TYPICAL)
EFFICIENCY 0.5
11
0.4
0.3
0.2
0.1
0
0.01 0.2 0.5
ADHERE NO THICKNESS t (IN.)
Figure 6.19 Joint efficiency chart for single-lap composite joints (ductile adhesive)
1.0
0.9
0.8
0.7
0.6
JOINT
EFFICIENCY 0.5
11
0.4
0.3
0.2
0.1
0.02 0.05 0.1 0.2 0.5

ADHEREND THICKNESS t (IN.)
Figure 6.20 Joint efficiency chart for single-lap composite joints (brittle adhesive)
The design of structural single-lap joints is then seen to be reduced to the specification
of an adequate overlap ( e/t in the range 50 to 100) and to the determination of an upper
limit on thickness (of the order of 0.06 inch or 1.5 mm) even after the peel stresses have
been alleviated by tapering the ends of the adherends. These recommendations refer in
particular to applications to subsonic airframe structures. The same techniques would
apply for the different adhesives used in the extreme environment of outer space or of
the blast-off of a rocket, for example, but the thickness cutoff would probably be more
severe and the adherend tapering could become more critical.
6.7 STEPPED-LAP BONDED JOINTS
At the time when the so-called advanced composites (boron-epoxy and carbon-epoxy)
were first being introduced into aircraft structures more than 30 years ago, little was
known about designing bolted joints in such materials. In the case of boron--epoxy, it
might fairly be added that drilling holes was sufficiently difficult to discourage that
approach anyway! Also, the use of adhesive bonding offered the chance to obtain
higher structural efficiencies, and to use less of what was then an extremely costly
material. That was the environment that led to the derivation of powerful analytical
computer programs with which to design composite-to-titanium stepped-lap bonded
joints in the early '70s.
The first such analysis seems to have been the elastic derivation by Corvell i
(Reference 22), followed soon after by the nonlinear (elastic-plastic adhesive) program
A4EG (Reference 23) developed by the author of this chapter. The latter program has
been developed much further (References 24 and 25) and now provides for non-
uniform adhesive thicknesses and mechanical properties and also accounts for local
flaws and porosity. Many worked examples derived by A4EI, the latest version, are
contained in References 26, 25, and 5. Joints of this kind have found production
applications in the tails of the F-14, F-15, F- 16, and F-18 as well as in the wing of the
F-18. There also are numerous applications of such joints in research and development
programs. The current state of the art is about one-nch (2.54-cm) thick laminates,
transferring load intensities of about 30,000 pounds per inch (5.25 MN/m) of width.
However, that is not a theoretical upper limit; production applications demanding
stronger joints of this type have not arisen yet.
The strength of these bonded joints is known to be particularly sensitive to a number
of fac!~rs. Perhaps the most significant is the need for adherend stiffness balance from
or_e end of the joint to the other, as explained in Figure 6.21. The lower joint has almost
twice the strength of the upper one, even though this has been obtained by removing
material rather than by adding reinforcement. The length and thickness of the end step
on the titanium are very important, as shown in Figure 6.22. The yielding of the
titanium at the root of that step (see preliminary design A in Figure 6.22) has shown up
as a design problem in two forms. That step would fatigue off if made too long - more
than about 114 to 3/8 inch (6 to 9 mm)- and result in a secondary failure because of the
reduced effective thickness of the composite laminate.
An extreme case of poor detailing of that area, in which a one-inch (2.5-cm) end
step was actually tom off under static load, is shown in Figure 6.23.That the solution to
this problem is not the thickening of the end step should be evident from Figure 6.24.
In the load-introduction area of that specimen, a bundle of all-zero-degree boron

filaments 0. 1 -inch (2.5-mm) thick had been terminated abruptly just beyond the end of
the step plate in the foreground of the illustration. The application of load caused a
delamination at each side of that bundle of strong, stiff filaments because the resin
matrix was not strong enough to transfer so much load in shear without failure.
c:::=====t3
TITANIUM
CONSTANT·THICKNESS (STIFFNESS-UNBALANCED) JOINT

J TITANIUM
CONSTANT- STIFFNESS (VARIABLE-THICKNESS) JOINT
ADHESIVE SHEAR STRESS DISTRIBUTION ADHESIVE SHEAR STRESS DISTRIBUTION
Figure 6.21 Stepped-lap bonded joints (strength improved by matching stiffnesses)
A tip thickness of 0.030 inch (0.75 mm) is more appropriate for the materials used on
aircraft today, and even then the core plies should not be cut off but be diverted around
a low-modulus triangular wedge about 0.5-inch (1.3-cm) long. This is shown in Figure
6.25 for the similar detail of a scarf joint with finite tip thickness. The last step on the
other end of the joint must have thin outer adherends to prevent delamination due to
induced peel stresses. The composite there should be no more than 0.030-inch (7.5-mm)
thick on each side of the step plate for the same reasons that the ends of the splice plates
of Figure 6.15 are tapered. That end step should also be limited to 3/8 to 112 inch (9.4 to
12.7 mm) in length to prevent the thin steps from being overloaded.
The remaining design variables are the length and number of steps. Of these, the
dominant effect is the number of steps. The length effect, for a given number of steps, is
extremely slight, as can be seen from the three joint analyses in Figure 6.26. Each joint
had five steps and all thicknesses were the same at the corresponding stations. An
increasein overlap by a factor of four is seen to increase the strength by only 14 percent.
This highlights the large difference between the mechanism of load transfer in scarf
joints (for which such an increase in area would have had a substantial effect on the joint
strength) and stepped-lap joints. The reason the stepped-lap joints are insensitive to the
overall joint length, as the sum of a series of uniform steps, is that each of those steps 1s
governed mathematically by precisely the same differential equation that governs
double-lap joints. The insensitivity of the strength of those joints to overlap is shown in
Figure 6.13.
TENSION
rFA.It.U~E IN COMPOSITE AT 3. 181 IIIH/ m UB,216LBI IN.)
.._\~ {YIELD IN TITANIUM AT 2.077 kN / m 11 1.866 lB / IN l
{!§~A:,~~~~~=:::=:J} /f.~
FAILURE IN COMPOSITE AT 2 974 t!Ht m (llfi.g97l8/ IN )J
COMPRESSION
~OT E tHAt TITANIUM END STEPS WERE ALREADY SHORtENED OURlNG PRELIMINA~"f OEStGN.
WITH UNIFORM STEPS 1.91 em 10. 7:io IN I LONG THROUGHOUT. PRtM.-fUFU: FAliGUE FAILURE
WOULD OCCUI1AT A. FOLLOWED BY FAILU RE Of CO~P'0$1TE AT THE SAME cREDUCED)
SECTION .
(A) PRELIMINARY DESIGN 0 1 2 l
SCAU~
TENSION 0 (IN_J I
rFAILURE IN COMPOSITE liT 3 , 182 ~N / m (18. 180 L8/ IN)
,t/~
6AI-4V TITANIUM TENSION
It~
COMPRESSION LFAILURE IN COMPOSITE AT 3.182 kN/ m !18.182l8/ JN _)
NO YIUDIHG OF TITAI'IUM
(8) OPTIMIZED DESIGN
DUCTILE ADHESIVE CURW AT '!10 K (l5Q.FJ
STRENGTHS CALCULATED AT ROOM TEMPERATURE
STIIfNGfN OF COMPOS4TE ADHEAEND OUTSIDE JOINT - l.lla kN/ft'[l8,216 LB / IN .)
POTENTIAL - D SHEAR STRENGTH WOULD EXCEED 4.070 kN/ "' l2l,2S7 L811N.J IN EVERY CASE
SHOWN IF ADHERENDS WERE SUffiCifNTLV STIIONG
Figure 6.22 Optimization of details in stepped-lap bonded joints
Figure 6.23 Failure of stepped-lap adhesive bonded joint

ADHESIVE BON.DED SPLICE
MECHANICALLY FASTENED SCARF JOINT

THAT WAS THE IHTENOED TEST SECTION
INITIAl. fALURE AT "AwBECAUSE 'TH.CKNESS "'B"'IS EXCESSIVE AHD

LOAD IN FIBERS "C " C.UNOT BE UNLOADED THROUGH RESIN MATRIX
FINAL FAR...URE AT "O" lS BY NET·SECTION TEHSJON ON THE TOP FACE
AAD SHEAROUT (NOT SHOWNI ON THE LOWER SURFACE
Figure 6.24 Premature failure by delamination of stepped-lap bonded joint
\"fl~--
-- -·- ·- ·- - - - - --
CENTERUHE
ADHESIVE
---~· ---~
Figure 6.25 Lay-up details for composite scarf joints
The strength increases associated with maximizing the number of steps in the joint are
substantial, as shown by a comparison of Figures 9, 8, and 6 of Reference 25. The
strength is maximized by having a separate step for each ply (0.005-inch or 0.13-min
thick) and such a procedure may be appropriate for the highest load transfers between
thick members. As a good starting point in design for typical composite laminates that
are either quasi-isotropic or only slightly orthotropic, each step should be about 0.5-
inch (12.7-mm) long and the thickness increment should be in the range 0.02 to 0.03
inch (0.5 to 0.8 mm) on each side of what is usually a titanium step plate. The exact
increment should, of course, be a precise match for the corresponding increment of

laminate thickness- for example, four layers of unidirectional tape at 0.005 inch (0. 13
mm) each would require 0.020-inch (0. 5 l -mm) steps while two layers of satin-wem e
cloth at 0.013 inch (0.33 mm) each would require 0.026-inch (0.66-mm) steps.
Yma:~~ = 0.2
Pelastic·plastic = 1,955 kN/m
(11,172 LB/IN.)
ADHESIVE
SHEAR STRESS (KSI)
MN/m2
MN/m2
(b) OVERLAP = 5.1 em (2.0 IN.)

8
6
4 (KSI)
MNfm2
(c) OVERLAP = 10.2 em (4.0 IN.)

NOTE: ALL CROSS SECTIONS IDENTICAL AND ALL STEP LENGTHS PROPORTIONAL
Figure 6.26 Limited effect of overlap on strength of stepped-lap bonded joints
Just as with double-lap joints, it is vital that at least one of the steps near the middle of
the total bonded overlap contain a very lightly stressed deep elastic trough to prevent
failure of the joint by creep. That step should probably be in the range 0.75 to 1.0-inch
(19 to 25-mm) long and its adequacy should be checked as part of the analysis. That is
why program A4EI prints out not only the high adhesive stresses at the ends of each
overlap but also the low stress at the middle of each step. That is taken to be indicatiw
of the precise minimum that is neither computed nor located.
It should be noted that the suggestions above are intended only to provide the
starting point of the design process using the A4EG or A4EI computer programs. There
is still usually room for improvement by minor tweaking of the design once the internal
stress distribution has been characterized. It is still desirable to establish that the
potential shear strength of the adhesive would exceed by at least 50 percent the strength
of the weaker adherend outside the joint. Otherwise the joint could become the weak
link in the chain and suffer catastrophic disbonding as the consequence of only a minor
local flaw, as explained in Figure 6.27. For this safeguard it is acceptable to count on
the entire length of the adhesive stress-strain curve to failure without any restriction on
the peak strain developed under normal operating loads.
PROPAGATE DISBOND
IF BOND DOES NOT FAIL
FATIGUE CRACK IN META L IS LI KELY
Figure 6.27 Redistribution of load at flaws in bond
Practical experience with bonded stepped-lap joints has taught that the layer of fibers or
filaments immediately adjacent to the metal plate should never be perpendicular to the
load direction. Otherwise, a much weaker joint results because of the tendency of such
layers to split transversely and roll within themselves. This is easily understood by
considering an axially loaded test coupon. The preferred fiber direction then obviously
would be zero degrees aligned along the specimen. A 90° ply would be quite
unsuitable, with an intermediate performance coming from a +45 ° or -45 ° degree layer.
Even with a large panel loaded primarily by in-plane shear, the same zero degree
direction is to be preferred because whichever 45 ° layer would be optimum for shear in
one direction would be the worst possible for shear in the opposite direction . This
consideration defines where the oo plies are to be located within the stack-up, and
therefore influences the basic laminate outside the joint as well.
As the load transfers increase to the level of tens of thousands of pounds per inch
(several MN/m), the joint proportions become sufficiently critical to be of concern from
the point of view of manufacturing tolerances. Fabrication experience has already
established that it is important not to overlap the end of the composite layers on to the
next step or to stop them significantly short on their own steps. Otherwise, the laminate
will be wrinkled within the joint area and will probably suffer internal delaminations
during cyclic loading. The delaminations, in tum, can redistribute the internal load
transfer so as to cause a reduction in joint strength or life. Such problems become more
acute as the incremental step thicknesses become larger. Indeed, for step thicknesses
significantly greater than those recommended here, the author's analyses will not
suffice. It then would become necessary to use an analysis that includes a check on the
possibility of delaminations within the composite adherends (see Reference 27). Another
important use of stepped-lap bonded joint analysis is the establishment of the sensitivity
of the joint strength to minor deviations from the nominal design. This task becom.:s
progressively more important after the strength has been raised by repeated
optimization of the detailed dimensions.
This discussion has been confined to symmetric stepped-lap joints laid up on each
side of a central step plate with the assembly co-cured and bonded together to ensure a
good fit and the absence of warpage. A single-sided stepped-lap bonded joint between
dissimilar materials would warp so badly from a 350°F (177°C) cure as to probably
break apart during the cool down. If that problem were removed by the use of a room-
temperature-cured adhesive or thermally compatible materials, a single-bond surface
stepped-lap joint could be analyzed as one side of a double-bond joint. While there is
some eccentricity in load path, that eccentricity is inevitably minor if the outermost
steps have been restricted in thickness adequately to minimize any induced ped
stresses.
6.8 SCARF JOINTS
It will seem strange to some that there was any need to have conducted research into the
design of adhesively bonded scarf joints, particularly to those who still think of the
joints as a means of transferring load uniformly across the bonded interface. Actually.
that simplistic characterization applies only to the transfer of load between t\\ o
precisely identical members, both of which have perfect feather edges at their tips. The
two dominant effects which cause real bonded scarf joints to deviate from that ideal are
stiffness imbalance between the adherends and finite thicknesses at their ends.
Adherend thermal mismatch between adherends of different materials is a further
significant source of loss of structural efficiency.
Figure 6.28 shows how, for sufficiently long scarf lengths, the ratio of average to
peak adhesive shear strength asymptotes towards the lesser ratio of adherend
stiffnesses. The critical location for the adhesive is usually at the point A shown. Most
structural joints have scarf angles so low that the design is at or beyond the right end of
the diagram. Interestingly, even adherend thermal mismatch which weakens such joints
(and has the maximum loss around AJi = 2) cannot overpower those asymptotic trends
for large values of AL Furthermore, adhesive ductility is effective in raising that ratio,
Tav/rmax• only for the shorter overlaps also. Reference 23 contains a thorough derivation
of the elastic-plastic transfer of load across idealized scarf joints. The design of such
joints is simplified tremendously by the use of Figure 6.26.
The practical difficulties of designing real scarf joints center around the thickness of
the tips of the adherends. Any attempt to actually create a feathered edge over a long
length will inevitably result in weaknesses of the type shown in Figure 6.29 due to
bending and wrinkling of the metal adherends. The upper joint shown failed not in the
thin basic section but across the tip of the internal doublers, where the nominal stress
was only half as high, in repeated tests. The simpler design shown with external doublers
was consistently stronger even though it had only half the bond area. If one accepts the
need for a finite tip thickness, say 0.02 to 0.03 inch (0.25 to 0.38 mm), one is then
forced to analyze the joint by an approximation as a stepped-lap joint in order to allow
for the stress concentrations around the tip of the adherends. Indeed, the McDonnell
Aircraft Company does not even keep the scarf-joint analysis programs in its active
computer library.
.~ ~ - ---J
IDENTICAL ADHERENDS·~
I 1.0
~ -:::--
---
0,9
0.8
~~ ~ 0,7
..._i
~ .........
""'~~ ~ '-._~~ 0.6
. o.s
J(... r= F"'
""'"
o.•
~ o. 3 ~'
I'
"0:3 = =
"'
!>,! 0,2 --.:;-::::: =
~ --;;-:....
o: ::::::.:::
~ o. I
IADHERENI ~""'"'nNAI. ","'~~si,';'j 1 o I m
"<1 00.5 10 :?0 30
~
Figure 6.28 Elastic strength ofscarfbondedjoints (effect ofadherend stiffness imbalance)
The use of the stepped-lap joint programs to analyze scarf joints requires a fairly fine
grid, even in those areas that are not critical. This can be seen from Figure 6.30, which
represents progressively more detailed models of the same scarf joint. The finite tip
thickness was only 0.010 inch (0.25 mm), yet there is still an appreciable stress
concentration there in the adhesive, even with the 32-step approximation. By assessing
the entire output of the program instead of studying just the peak stresses, one can
rapidly learn to make an accurate model with grids in which the fineness is varied as
needed.
While stepped-lap joints and scarf joints have some very pronounced differences in
their governing differential equations, one problem they share in common is the
tendency for the thin tip of the stiffer member to break off in fatigue. With reference to
Figure 6.28, the load transfer is concentrated at the left end of any unbalanced joint so
the buildup of load in the thicker member is not uniform. The load can, in fact, build up
faster than the strength. Somewhat surprisingly, more failures of the adherends in Figure
6.28 would occur at the tips of the stronger adherends rather than in the weaker
adherend just outside the joint.
IHT'EAHAUY BONDED DOUBLERS (IDEAL}
~s: & ==-- - \

t BENT UPS ON ENDS OF DOUBLERS
LONG CRACK
BWHT UPS ON EHOS OF DOUBLERS
(JOGGLED PUES IN COMPOSrTE)
INT£AN.At.LY BONDI!D DOUII..EAS
(A.EPAODUCED FROM "C'T\JAL PHOTOGRAPHS)
EXTEANAUY BONDED DOUBUR$ (ACTUAL)
Figure 6.29 Internal and external bonded doublers for composite laminates
One potential problem with bonded scarf joints which is frequently overlooked is that
the mathematically ideal form has a uniformly strained adhesive (ignoring the tip
effects) and that, in consequence, the only way to restrict the minimum strain to prevent
creep failures is to restrict the upper strain also. That leads to very long scarf joints with
extremely small scarf angles, which is contrary to the concern expressed above about
fatiguing the thin tips of scarfed adherends. A two-step scarf of graduated thicknes~
buildup would seem to be an appropriate solution to this problem, as discussed in
Reference 23.
6.9 ADHESIVELY BONDED JOINTS OTHER THAN AT SPLICES
While modem military aircraft with advanced composite structures contain many highly
loaded adhesively bonded joints, most of the even more extensive use of metal bonding
in certain commercial transport aircraft is in areas other than classical joints.
Both the de Havilland and Fokker companies have made extensive use of bonding.
rather than riveting, stiffeners to skins with the prime objective of improving the static
structural efficiency of panels loaded by compression or shear. The associated increases
in fatigue lives and resistance to corrosion are like frosting on the cake since it usually
is the static ultimate design condition that establishes the weight of such structures .
While it is true that the companies' choice of applications has been enhanced because
they don't make the largest of the transport aircraft, it is somewhat surprising that the
applications of metal bonding on the other side of the Atlantic have had such a different
emphasis. The less widespread use of metal bonding in American aircraft structures has
concentrated more on the local thickening of thin skins to permit countersinking and on
the enhancement of the fatigue lives, in severe acoustic environments, of structures that
are of minimum gauge and have considerable static strength margins. In all of these
applications, there has been no need for more than a cursory analysis of the strength of
the adhesive bonds because of the minimal load transfer. For that reason, such
reinforcements are usually made integral with the basic structure in fibrous composites.
STEPS IN APPROXIMATION
PEAK SHEAR STRAIN COMPUTED
ACTUAL SHEAR STRAIN. UNIFORM THROUGHOUT
4-STEP APPROXIMATION
ADHESIVE
SHEAR
STRAIN
TRUE
SCARF-JOINT
:N~~~.7'<l'ff;,\-+..J~f.+-l~'""f'lo(-l~ CONSTANT-
STRAIN
SOLUTION
0~------------------------------_J
I-----EXTENT OF BONDED-SCARF OVERLAP---
NOTE: BEST ANSWERS FOR A GIVEN TOTAL NUMBER OF STEPS ARE ACHIEVED WITH A VERY
FINE GRID NEAR THE ENDS AND A COARSER GRID IN THE INTERIOR
Figure 6.30 Influence of modeling precision on analysis of scarf joint
The PABST program (Reference 11) at Douglas identified a further area for the
application of adhesive bonding in which good analysis was needed. This area of
interest is the damage tolerance of adhesively bonded structures, such as stiffened
panels, in which the bond loads are quite low provided that the metal elements are
intact. However, if one element is broken by impact damage or by fatigue cracking
from one of the remaining rivet holes, the question arises whether a catastrophic
widespread disbanding will follow or such initial damage will remain confined. This
topic is addressed in Reference 28, where it is shown that the likelihood of failure to
contain the damage is remote in most of the structure that would be employed on
fuselages or tail structures. This issue would become quite complex in fibrous
composite construction because any initial damage would more likely be followed by
delaminations within the basic material rather than by disbanding between the
members.
6.10 FLAWS IN ADHESIVE BONDS
The variable-adhesive feature of the A4EI joint analysis program has permitted a
comprehensive assessment of the effects on the strength of bonded joints of flaws and
porosity in adhesives or even the effects of moisture absorption or drying out.
References 5, 10, and 23 contain several examples depicting such effects. It was found
in most cases that adhesive bonds are much more tolerant of flaws than is generally
recognized. The critical location in a bonded joint is usually at one end of the overlap
and flaws tend not to occur there; they form more often in the interior, where air
bubbles can be trapped. Now, the interior of the joint is usually lightly loaded anyway
and, even neglecting that, local flaws were found to have no effect on the strength of
bonded joints until they became so large as to shift the critical location away from the
edge of the joint to a location adjacent to the flaw. When this happened, the load
redistribution was so great that a substantial loss of strength occurred. This is shown
particularly well in Figures 40 to 43 of Reference 10.
The history, and even current practice, of metal bonding is replete with instances of
the structurally unnecessary repairs of flaws in bonded structures which have
accomplished only a reduction in the service life of the structure by breaking the surface
protection that had been afforded by the primer, anodize, or etch. A more
comprehensive discussion on the adhesive-bonded repairs of metal structures is
contained in Reference 29.
A closely related topic is the use of adhesive bonding and mechanical fastening
together to provide fail-safe characteristics. This topic is discussed thoroughly in
Reference 23. The highlights are that, for thin and moderately loaded structures.
adhesive bonding provides a fail-safe load path to protect the structure in the vicinity of
the fastener holes. Conversely thick, highly loaded adhesively bonded structures need
mechanical fasteners to protect the bond from unzipping catastrophically due to any
small initial damage. The vast differences between the behaviour of these combination~
for thick and thin structures stems from the different relative strengths of the basic
laminates and the bonded joints, as shown in Figure 6.4. The fail-safe load path change~
at the point at which the adherends become stronger than the bond. The grossly
dissimilar stiffnesses of the load paths through the bond and fasteners result in the two
not acting together when both load paths are intact. The bond always transfers the load
because it is so much stiffer. However, the combination is shown to be particularly
useful once damage has been incurred by one or another alternative load path.
6.11 REPAIR OF FIBROUS COMPOSITE STRUCTURES
The most vital step in repairing a fibrous composite structure using adhesive bonding is
the thorough drying out of the laminate to be repaired prior to hot bonding. Otherwise,
even a small amount of water in the laminate will have a devastating effect on the
strength of the adhesive or of the co-cured patch. This problem has been known to exist
for years (as explained in Reference 30) but has only recently attracted the attention it
merited (Reference 31 ). The difficulty of drying the laminate can be slight for a thin
skin on a control surface or fairing - the drying can be done using a small hot-air gun--
but it could be horrendous for a thick component that may require a few days of gentle
heating in an oven. The difficulty of removing large, thick components from an aircraft
and stripping them of heat-sensitive equipment, in conjunction with the problem that an
adequate scarfing out around any initial damage would leave little of the original
structure remaining, has led to a recognition of the need for mechanical repairs in such
cases. That, in tum, implies a need to provide for mechanical repairs in the initial design
stage, as was done with the AV-8B Harrier.
The other important step in the execution of adhesively bonded repairs, or in initial
manufacture for that matter, is the thorough cleaning of the surfaces to be bonded. The
need for mechanical abrasion, as by grit blasting, is made very clear in Reference 32.
Removal of peel plies is not a sufficient preparation for bonding. It is also important
that the surface not be contaminated by "cleaning solvent," as is often officially
prescribed. Scrubbing with an abrasive cleanser like Ajax is preferable. Finally, the
adequacy of the surface preparation should be verified by the standard water-break test
(Reference 33), just as for metal bonding.
Failure to adhere to these simple safeguards will nullify whatever effort was
expended in the analysis and design of bonded repairs. The problem is directly akin to
the widespread corrosion problems that were encountered in the U.S. by the
combination of the use of clad 7075 aluminium alloys with inadequate surface
preparation and the first generation of 250°F (120°C) cured epoxy adhesives that were
too susceptible to the absorption of moisture.
6.12 RULES OF THUMB FOR DESIGNING STRUCTURALLY EFFICIENT

JOINTS
• Bonded joints must always be designed to be stronger than the adjacent structure.
Otherwise, the bond can act as a weak-link fuse and unzip catastrophically from a
local defect.
• Bonding works best for thin structures.
• Thick bonded structures need complex stepped-lap joints to develop adequate
efficiency.
• Thick structures are impractical to repair by bonding so, except for one-shot and
throwaway structures, original designs should be bolted together to permit bolted
repmrs.
• When there is no need for repair, as in missiles and unmanned aircraft, bonding
permits extremely high structural efficiencies to be obtained, even on thick
structures.
• Proper surface preparation is a "must"! Beware of "cleaning" solvents and peel
plies. Mechanical abrasion is more reliable.
• Ensure that the surface preparation for bonding is one that ensures that the glue will
adhere to the adherends. Do not continue to use processing techniques for which
past experience has shown that the glue did not stick to the surfaces. Interfacial
failures between adhesive, primer, and substrate are irrefutable proof that the
adhesive was never stuck in the first place, no matter what short-term coupon tests
and non-destructive inspections might suggest. Only cohesive, or interlaminar,
modes offailure are acceptable, no matter what the specification might say.
• The best analyses are absolutely worthless unless one can rely on the glue staying
stuck.
• Be aware of the need to dry laminates before doing bonded repairs.
• The key to durability of bonded joints is that some of the adhesive must be lightly
stressed to resist creep.
• Bonded joint overlaps are usually sized by the hot/wet environment.
• Bonded joint strength is usually the least in cold environments, where adhesives are
brittle.
• Care is needed to taper ends of bonded overlaps- down to 0.5 mm (0.020 inchl
thick with a 1-in-10 slope - to minimize induced peel stresses that would cause
premature failures.
• Adhesives work best in shear - and are poor in peel. But composites are even
weaker in interlaminar tension.
• Design of simple uniformly thick, (for near quasi-isotropic carbon epoxy) bonded
splices is very simple - about 30t overlap in double shear, 80t overlap for single-
lap joints, and 1-in-50 slope for scarf joints.
• Design of stepped-lap joints for thick structures needs a nonlinear analysis
program.
• Adhesives are well characterized by the thick-adherend test coupon, generating
complete nonlinear stress-strain curves in shear.
• Acknowledge that the simplistic design rule whereby a bonded joint strength is
rated as the product of the total bond area and some fictitious uniform bond shear
"allowable" became obsolete once aircraft were no longer built of wood and fabric,
and the glue was no longer stronger than the members being joined.
6.13 CONCLUDING REMARKS
The purpose of this chapter has been to provide the designer with some of the finer
details in the proportioning and analysis of various adhesively bonded joints in a
context in which the appropriate uses of each configuration become apparent.
The analysis methods discussed utilize an elastic-plastic shear model for the
adhesive. The unsuitability for design, or even for adhesive selection, of the standard
specifications of adhesive "strength" is explained and a case is made for a more rational
approach to the problem.
Only the foundations of the design process are presented here because the
completion of the task often requires computer programs and, in any case, the sizes of
the joint can vary with the fibrous reinforcement material, with the fiber pattern, and
with the resin matrix. The simpler joints, of single-lap and double-lap configurations,
can be designed accurately using no more than simple analytical formulae.

An extensive reference list at the end of this chapter contains the sources of more
specific information related to the sizing and analysis of the various kinds of bonded
joints, using the methods described here. While not all identified here, there are many
production applications of adhesively bonded joints that have been designed using these
methods. The various computer programs in the A4E series, of which the early ones are
freely available while the later ones can be obtained only on request to the USAF, have
been used extensively throughout the aerospace industry for many years now.
6.14 ACKNOWLEDGMENT
The origins of most of the material in this article are four U.S. government-sponsored
investigations, in which the author participated, at the Douglas Aircraft Company (now
part of the Boeing Company), at Long Beach, California. The first is the NASA
Langley Research Center study, Analysis and Design of Advanced Composite Bonded
Joints, performed between 1971 and 1973 under contract NASI-11234. All of the
material produced by that study, as well as from the third and fourth contractual
projects identified below, has been released by the U.S. Government for unlimited
distribution. The second project, jointly managed and directed by the Air Force Flight
Dynamics and Materials Laboratories, was the Primary Adhesively Bonded Structure
Technology (PABST) program, F33615-75-C-3016, conducted between 1975 and
1980. Some of the test data generated during this contract that was prepared by others,
still has a restricted circulation, but all of the published works in which the author's
contributions have been documented have been released for unlimited distribution.
Additional relevant work was accomplished under Air Force Flight Dynamics
Laboratory contract F33615-79-C-3212, Design Methodology for Bonded-Bolted
Composite Joints, during the period 1979 to 1981 (Volume II of that report, containing
the computer codes, was never circulated, but the codes have since been released for
general distribution). Further applicable information was developed under Air Force
Materials Laboratory contract F33615-80-C-5092, Adhesive Layer Thickness and
Porosity Criteria for Bonded Joints, which extended from 1980 to 1982.
It is probable that none of these developments would have occurred had not the
original concept of an elastic-plastic model for adhesives in shear been sponsored under
IRAD funding by the Douglas Aircraft Company throughout 1969 and 1970 to develop
those ideas to a marketable level.
6.15 REFERENCES
1. Hart-Smith, L. J., 'The Bonded Lap-Shear Test Coupon - Useful for Quality
Assurance, but Dangerously Misleading for Design Data", McDonnell Douglas
Paper MDC 92K0922, presented to 38th International SAMPE Symposium &
Exhibition, Anaheim, California, May 10-13, 1993; in Proceedings, pp. 239-246.
2. Hart-Smith, L. J., Redmond, G, and Davis, M. J., "The Curse of the Nylon Peel
Ply", McDonnell Douglas Paper MDC 95K0072, Proceedings of the 41st Int.
SAMPE Symposium and Exhibition, Anaheim, Mar 25-28, 1996 pp. 303-317.
3. Hart-Smith, L. J., "Effects of Pre-Bond Moisture on Interfacial Failures in Glued
Composite Joints- And What To Do About It," Boeing Paper, presented to MIL-
HDBK-17 Meeting in San Diego, March 30 -April 2, 1998.
4. Hart-Smith, L. J., "Induced Peel Stresses in Adhesive-Bonded Joints", Douglas
Aircraft Company, McDonnell Douglas Corporation, Report MDC-J9422A, August
1982, USAF Contract Report AFWAL-TR-82-4172.
5. Hart-Smith, L. J., "Further Developments in the Design and Analysis of Adhesive-
Bonded Structural Joints", Douglas Aircraft Company, McDonnell Douglas
Corporation, Paper 6922, presented to ASTM Conference on Joining of Composite
Materials (STP 749), Minneapolis, Minnesota, Aprill980.
6. Hart-Smith, L. J., "Differences Between Adhesive Behavior in Test Coupons and
Structural Joints", Douglas Aircraft Company, McDonnell Douglas Corporation,
Paper 7066, presented to ASTM Adhesives Committee D-14 meeting, Phoenix,
Arizona, March 1981.
7. Hart-Smith, L. J., "Design and Analysis of Adhesive-Bonded Joints", Douglas
Aircraft Company, McDonnell Douglas Corporation, Paper 6059A, presented to
Air Force Conference on Fibrous Composites in Flight Vehicle Design, Dayton,
Ohio, September 1972, USAF Conference Proceedings AFFDL-TR-72-130.
8. Oplinger, D. W., Paper on finite-difference analysis codes, FAA research, Privak
Communication.
9. Hart-Smith, L. J., "Adhesive-Bonded Double-Lap Joints", Douglas Aircraft
Company, NASA CR-112235, January 1973.
10. Hart-Smith, L. J., "Effects of Flaws and Porosity on Strength of Adhesive-Bonded
Joints", Douglas Aircraft Company, McDonnell Douglas Corporation, Report
MDC-J4699, November 1981, USAF Contract Report AFW AL-TR-824172.
11. Hart-Smith, L. J., "Adhesive-Bonding of Aircraft Primary Structure," Dougla~
Aircraft Company, McDonnell Douglas Corporation, Paper 6979, presented to SAl-
Aerospace Congress and Exposition, Los Angeles, California, October 1980.
12. Hart-Smith, L. J., "Analysis and Design of Advanced Composite Bonded Joints''.
Douglas Aircraft Company, NASA CR-2218, January 1973.
13. Voikersen, 0., "Die Nietkraftverteilung in Zugbeanspructen Nietverbindungen mit
Konstanten Laschenquerschnitten", Luftfahrtforschung 15, 4-4 7, 193 8.
14. Hart-Smith, L. J., "Adhesive-Bonded Joints for Composites -Phenomenological
Considerations", Douglas Aircraft Company, McDonnell Douglas Corporation,
presented to Technology Conference Associates, Conference on Advanced
Composites Technology, El Segundo, California, March 1978.
15. Hart-Smith, L.J., Effects of Adhesive Layer Edge Thickness on Strength of
Adhesive-Bonded Joints, Douglas Aircraft Company, McDonnell Douglas
Corporation, Report MDC-J4675, May 1981, USAF Contract Report AFWAL-TR-
82-4172.
16. Jones, R., and Callinan, R. J., "Developments in the Analysis and Repair of
Cracked and Uncracked Structures", in: Finite Element Methods in Engineering;
Proceedings of the Third International Conference, Sydney, Australia, July 26,
1979, pp. 231-245.
17. Hart-Smith, L. J., "Design and Analysis of Bolted and Riveted Joints in Fibrous
Composite Structures", Douglas Aircraft Company, McDonnell Douglas
Corporation, Paper 7739, presented to International Symposium on Joining and
Repair of Fibre-Reinforced Plastics, Imperial College, London, Sep 10-11, 1986.
18. Go land, M., and Reissner, E., The Stresses in Cemented Joints, J Appl. Mech. 11,
pp. A17-A27, 1944.
19. Oplinger, D. W., "Effects of Adherend Deflections in Single-Lap Joints", Int. J
Solids & Structures, 31, pp. 2565-2587, 1994.
20. Hart-Smith, L. J., "The Goland and Reissner Bonded Lap Joint Analysis Revisited
Yet Again- But This Time Essentially Validated", Boeing Paper MDC OOK0036,
to be published under the auspices of the CRAS program.
21. Hart-Smith, L. J., "Adhesive-Bonded Single-Lap Joints", Douglas Aircraft
Company, NASA CR-112236, January 1973.
22. Corvelli, N., and Saleme, E., "Analysis of Bonded Joints", Grumman Aerospace
Corporation, Report ADR 02-01-70. 1, July 1970.
23. Hart-Smith, L.J., "Adhesive-Bonded Scarf and Stepped-Lap Joints", Douglas
Aircraft Company, NASA CR-112237, January 1973.
24. Smith, M. K., Hart-Smith, L. J., and Dietz, C. G., "Interactive Composite Joint
Design", Douglas Aircraft Company, McDonnell Douglas Corporation, USAF
Technical Report AFFDL-TR-78-38, April1978.
25. Hart-Smith, L.J., "Design Methodology for Bonded-Bolted Composite Joints",
Douglas Aircraft Company, McDonnell Douglas Corporation, USAF Technical
Report AFWAL-TR-81-3154, February 1982.
26. Hart-Smith, L.J., "Adhesive Layer Thickness and Porosity Criteria for Bonded
Joints", Douglas Aircraft Company, McDonnell Douglas Corporation, USAF
Technical Report AFWAL-TR-82-4172, December 1982.
27. Dickson, J. N., Hsu, T. M., and McKinney, J. M., "Development of an
Understanding of the Fatigue Phenomena of Bonded and Bolted Joints in Advanced
Filamentary Composite Materials", Vol. 1, Analysis Methods, Lockheed Georgia
Company, AFFDL-TR-72-64, Vol. 1, June 1972.
28. Hart-Smith, L.J., Adhesive Bond Stresses and Strains at Discontinuities and Cracks
in Bonded Structures", ASME, J Eng. Mat. Tech., Vol. 100, pp. 16-24, 1978.
29. Hart-Smith, L. J., "Design and Analysis of Bonded Repairs for Metal Aircraft
Structures", Douglas Aircraft Company, McDonnell Douglas Corporation, Paper
7089, presented to International Workshop on Defense Applications of Advanced
Repair Technology for Metal and Composite Structures, Naval Research
Laboratory, Washington, D. C., July 1981.
30. Hart-Smith, L. J., "Design and Analysis of Bonded Repairs for Fibrous Composite
Aircraft Structures", Douglas Aircraft Company, McDonnell Douglas Corporation,
Paper 7133, presented to International Workshop on Defense Applications of

Advanced Repair Technology for Metal and Composite Structures, Naval Research
Laboratory, Washington, D.C., July 1981.
31. Myhre, S. H., Labor, J. D., and Aker, S. C., "Moisture Problems in Advanced
Composite Structural Repair", Composites, Vol. 13, No.3, July 1982, pp. 289-297.
32. Parker, B. M., and Waghorne, R. M., "Surface Pretreatment of Carbon Fibre-
Reinforced Composites for Adhesive Bonding", Composites, Vol. 13, No. 3., July
1982, pp. 280-288.
33. Hart-Smith, L. J., Ochsner, R. W., and Radecky, R. L., "Surface Preparation of
Fibrous Composites for Adhesive Bonding or Painting", Douglas Service
Magazine, Vol. 41, First Quarter 1984, pp. 12-22; reprinted in Canadair Service
News, Vol. 14, No.2, pp. 2-8, Summer 1985.
CHAPTER 7 DESIGN AND ANALYSIS OF BOLTED AND RIVETED JOINTS IN
FIBROUS COMPOSITE STRUCTURES 1
L. J. HART-SMITH
The Boeing Company, Long Beach, California, USA
7.1 INTRODUCTION
The purpose of this chapter is to explain the principles behind the design and analysis of
bolted or riveted joints in fibrous composite structures. Because each manufacturer of
composites creates material as parts are fabricated, no two manufacturers of composite
hardwarecan develop the same structural allowables, even from what nominally appears
to be the same resin and fibre ingredients. Minor differences in cure cycles, resin
content, bagging techniques, lay-up environment, storage conditions, supplier shipping,
and the like cause statistically different mean strengths and scatter factors. There are,
therefore, no universal allowables such as those in MIL-HBK-5 for metal alloys,
because there are two sources of variability, from the material suppliers and from those
who produce the final products. Nevertheless, it should be noted that the great bulk of
what has previously been mistakenly believed to be variability in composite material
strengths has now been established to be the result of poor test-specimen design,
fabrication, and testing. The true variability in fibre/polymer strengths is no greater than
for typical metal alloys. Moreover, the statistics used to establish A- and B-allowables
have been applied in a totally inappropriate way. Only a fibre, the matrix, or an
interface can fail, not a homogenized lamina, because the latter is only a mathematical
concoction created to simplify structural analyses. [It has never had any relevance to
failure (strength) criteria for composites of materials.] The typical test coupon contains
something on the order of a million fibres, and a million resin-filled interstices between
the fibres. Under these circumstances, only the mean values of the measurements have
1 This chapter is an updated version of a paper presented to the International Symposium on Joining And
Repair Of Fibre-Reinforced Plastics, at the Imperial College, London, on September I 0 and II, 1986 and
published in the proceedings, Joining Fibre-Reinforced Plastics, edited by F. L. Matthews, Elsevier Applied
Science, England. 1987, pp. 227-269.
211
L. Tong and C. Soutis (eds.).
212 L.J. Hart-Smith
any statistical significance. Variations between the means established on multiple test
coupons made from the same roll of material and cured simultaneously must, by
definition, be assigned to something other than true material variability. It is vital that
these variabilities be associated with their true causes because these causes are not the
same in test coupons as they are in complete structures.
Additionally, each manufacturer of composite aircraft structures is certified
independently and the processing method, once approved, is not transferrable to other
manufacturers' sites without re-certification at each new facility. Given this
background, it would not be appropriate to try to document precise design-allowable
strengths here. In any event, new and improved materials are coming on the market
continually.
The approach followed here is to explain the philosophy behind the design and
analysis processes from a general point of view that does not need to be changed for
each different composite material. The variables considered are related to the physical
behaviour, to identify which factors must be accounted for and which may be neglected
for any particular material, based on what is known of its behaviour. One purpose of
such a presentation is to define the minimum test program that can be used to
adequately characterize bolted joints in any composite material.
Any material allowables quoted here are usually sufficiently close for preliminary
designs, and a considerable part of the structures of subsonic transport aircraft can be
assessed within the framework outlined here. (The unresolved issue of the correctness
of unnotched reference strengths is immaterial here, since this effect would
automatically be compensated for by a change in the empirical correlation factors
relating test and theory. The so-called "notched" strengths will be unchanged since the
stress concentrations at holes have usually been found to over-ride those (such as tabs
and edge-effects) that have caused such misunderstanding when there have not been
any more dominant features to restrict the variability in measurements.) Other
references should be studied for such information as bearing strengths at elevated
temperatures. However, once such data have been obtained to establish a new set of
coefficients, the analysis methods contained here have been shown to be effective in
those situations also.
The material in this chapter is organized in the following sequence. The
characteristics of single fastener joints are described first. Next, the effects of fibre
pattern and lay-up sequence are discussed. This is followed by an outline of a minimum
test plan for bolted joints in new composite materials. This, in tum, is followed by an
explanation of a theory developed to explain the behaviour of singly bolted composite
joints. Then the theory is generalized to multi-row bolted joints, which leads to the
concept of joint structural efficiency. The concluding sections cover design procedures
and miscellaneous considerations, including a set of rules-of-thumb.
7.2 SINGLE-HOLE TEST COUPONS
Many of the data on which the design of bolted joints is based have been accumulated
from the testing of simple coupons having a single fastener, in double or single shear.
Chapter 7 Design and Analysis of Bolted and Riveted Joints 213
The variables considered usually include fibre pattern, fastener size, laminate thickness,
edge (or end) distance, width, and sometimes clamping torque. Theories developed to
explain these observations and to predict the performance of more complex structural
joints are customarily checked out first on single-hole data. Therefore, these simple
joints provide a logical starting point for this work.
There are various modes of failure for single fastener joints and the most frequently
occurring ones are shown in Figure 7.1. Of these, only two (bearing and tension-
through-the-hole) are considered desirable, in the sense that all the others fail
prematurely at lower loads. For example, the use of too small a bolt diameter results in
excessive bolt bending and an inefficient, highly nonuniform bearing stress distribution,
or, worse still, the bolt head may dig in and pull through the laminate, as shown. Shear-
out and cleavage failures result from either too small an edge (end) distance, too
orthotropic a laminate pattern, or insufficient dispersion of the differently oriented plies.
The distinction between bearing and tension failures is established largely by the joint
geometry - particularly the width-to-diameter ratio, wid - and, to a lesser degree, by the
fibre pattern. Much of the subsequent discussion will be concerned with only these two
failure modes.
A further consideration is the difference between loaded and unloaded holes, as
shown in Figure 7 .2. It is customary in design to allow for a higher stress concentration
at a loaded hole than at an unloaded hole . More importantly, a distinction must be made
between bearing and nonbearing, or bypass, loads to be able to characterize the internal
loads in multi-row bolted joints. The bearing load at each fastener is defined by the
shear load transferred by that particular fastener. Any remaining load bypasses that
fastener, to be reacted elsewhere.
CLEAVAGE· TENSION FAILURE BEARING FAILURE BOLT FAILURE
Figure 7.1 Modes offailure for bolted joints in advanced composites

214 L.J. Hart-Smith
t TOTAL LOAD
BEARING CUT ·OFF
BEARING
LOAD
TENSION-THROUGH-
THE-HOLE FAILURES
BYPASS LOAD
Figure 7.2 Bearing/bypass load interaction for loaded bolts in advanced composites
The strength of single-hole bolted joints as a function of the joint geometry can be
characterized by curves of the type shown in Figure 7.3. This information has been
extracted from Reference I and the explanation of its derivation is given below. Figure
7.3 has actually been prepared for quarter-inch (6.35-mm) bolts in double shear in a
quasi-isotropic pattern in high-tensile-strength (rather than high-modulus) carbon-
epoxy, but the general form would be the same for almost any fibrous composite.
Curves of this type could be derived based purely on test results, using no theory at all.
There are bearing failures at the left side of the picture, where the strip width (or
fastener pitch) is large in comparison with the hole diameter. As the fasteners arc
moved closer together (or the strip is narrowed), there is a change in failure mode t\)
tension-through-the-hole. Beyond the peak, there is a further change in the precise
nature of the tensile failure mode but, by then, the fasteners are too close together to be
of practical interest.
For laminates with a relatively high bearing strength, the regime of tension failures
includes a plateau across a range of d/w ratios before the strength drops off steadily due
to loss of tension area at larger dlw ratios. Those materials having a relatively lower
bearing strength would not exhibit that plateau - the strength versus diameter-to-width
ratio curve would have a sharp peak at some intermediate d/w ratio. Figure 7.3 shows
also the losses of strength due to short-edge-distance effects -low e/d or e/w ratios.
It is evident from Figure 7.3 that the peak strength occurs either at a geometry
associated with the tension failure mode or at an abrupt transition between bearing and
tension failures, and that, in either case, that peak strength is associated with a definite
wid value, which is usually closer to 3 than the frequent design practice of 4 to 5. Any
effort to impose on a bolted composite joint design an arbitrary requirement of failure by
bearing only (by using a sufficiently small value of dlw) in order to ensure a less
catastrophic failure mode must necessarily be associated with a significant reduction in

joint strength and, hence, in structural efficiency. 2 This applies to both single-row and
multi-row bolted joints in fibrous composite structures. Indeed, because of the influence
of bypass loads, it is almost impossible to design a multi-row bolted joint that will fail in
bearing rather than net-section tension. Even then, the fasteners in each row would need
to be so far apart, by ever increasing distances as more rows of fasteners are added, that
it would be very difficult to raise the gross-section laminate strain above that which
could be attained with a single-row bolted or riveted joint with the fasteners at the
optimum pitch. The problem with designing multi-row joints to fail in bearing is that
the total number of fasteners can be increased very little, if at all, so the load transferred
cannot be increased, either. This aspect of the joint design process needs to be quantified
better because of a commonly stated preference for bearing failures rather than the
unforgiving tensile failures. In the past, the choice has often been made with no
appreciation of the necessarily different ultimate joint strengths. This is somewhat
surprising in view of the widespread knowledge of bearing-bypass curves (discussed
later) that clearly show steadily decreasing gross-section strains with sub-critical bearing
loads, culminating in a drastic loss of strength when the full bearing stress is developed.
JOINT EFFICIENCY
JOINT STRENGTH
BASE LAMINATE
STRENGTH
Figure 7.3 Bolted joint efficiencies for composite laminates as functions of joint geometry
Curiously, some aircraft design organizations follow a practice of deliberately designing to high bearing
stresses, and lower average net-section tensile stresses, for metallic structures in the belief that this will result
in more benign failure modes. On the contrary, while this is the case for brand new structures, the higher
bearing stresses result in the earlier initiation of fatigue cracks at the fastener holes so that, later in the life of
such structures, the residual strength is limited by many such cracks which would be absent from alternative
designs, with closer fastener spacing and lower bearing stresses for a common load per unit length.
216 L.J. Hart-Smith
The physical explanation of the plateau and optimum joint proportions in Figure 7.3 is
revealed in Figure 7.4, which presents some results of classical analyses of the stress
fields around loaded bolt holes in isotropic plates. This analysis is for a single loaded
hole in an infinite plate (and is not to be confused with the better-known classical
solution for a stress concentration factor of 3 at an unloaded circular hole in an
isotropic plate). The key characteristic is that the peak tension stress alongside the bolt
hole is roughly equal to the average bearing stress. In other words, the only way to
restrict the peak tension stress is to limit the average bearing stress. When finite-width
effects are added, this need is achieved by having the bolts far closer together than is
customary for maximizing the static strength for ductile metal alloys, as is discussed
later in the context of optimized design for multi-row joints.
The bearing strength of carbon-epoxy laminates has long been known to be
improved by torquing the bolts tighter. Furthermore, there is a dramatic difference
between the bearing strength with regular torqued fasteners and with pin-loaded holes
having no clamp-up whatever. This effect is illustrated in Figure 7.5, extracted from
Reference 2, showing how the bearing strength of carbon-fibre-reinforced plastics is
doubled by finger-tight torquing of the bolts. The reason for this increase in strength is
that the initially damaged material can still withstand compressive loads under the bolt
shank, provided there is nowhere for the material to be displaced to.
BEARING STRESSES T
4P
7rdt
Figure 7.4 Hoop- and bearing-stress distributions around loaded bolt hole in isotropic plate
The initial damage is sometimes detected as a blip on a load-deflection curve that

continues linearly to much higher loads, as shown in Figure 7.6, or it often may only be
heardduring the test. As this damage spreads, powdered composite material is displaced
laterally and induces higher tensile loads in the fastener, compressing the composite
material ever more tightly. With a pin-loaded hole with neither fastener head nor nut, on
the other hand, the initially damaged material brooms O~Jt and does not continue to
sustain load. This, in tum, increases the bearing load on the remaining material, which
then fails at an accelerating rate as the load is increased.
140r---------------------------------.
120 I- ~
8
100 I- 8
BEARING STRENGTH T3001N5208 CARBON-EPOXY LAMINATES

(KSI)
e8 QUASI-ISOTROPIC L.AM,INATE PATTERN
BOLT DIAMETER = 0.25 INCH
TESTS CONDUCTED
0 AT ROOM TEMPERATURE
I I
PIN CONNECTION PROTRUDING-HEAD BOLT
Figure 7.5 Comparison between bearing strengths for pin loading and regular (torqued) bolts (100
ksi = 700 mn/m2 )
LOAD
Figure 7.6 Load-deflection curves for bolted composite joints
The use of flush (countersunk) fasteners permits some clamp-up, but considerably less
than under a washer or protruding bolt head, so the benefits of clamp-up are not as
218 L.J. Hart-Smith
evident with countersunk fasteners. In addition, of course, there is a much highi:r

bearing stress associated with the reduced shank area.
Figure 7.7 shows how still greater increases in bearing strength can be obtained by
applying even higher torques on the fasteners, as discussed in Reference 3. Reference 4
documents similar phenomena with glass-fiber-reinforced plastic laminates. Reliance on
the benefits of torquing beyond finger tightness (to constrain any damage) has been
questioned because some of the additiunal clamp-up will be relieved by creep over thi:
life of the structure. Investigations into the rate of loss of this clamp-up have been
conducted (Reference 5) and the decay rate is shown there to be fairly slow. The
drawback in taking advantage of the added strength associated with the extra bolt
clamp-up is that it would take only one under-torqued bolt to impose a significant loss
of static strength. The equivalent problem in metal structures- that of one loose bolt in
a pattern of interference-fit bolts - causes only a loss of fatigue life, not of static
strength.
200
BEARING STRENGTH (KSI)
100 HTS CARBON EPOXY

QUASI-ISOTROPIC LAMINATE PATTERN
0.25-INCH DIAMETER BOLTS
~ PIN-LOADED HOLE
NORMAL TORQUES NORMAL METAL

FOR COMPOSITES BOLT TORQUES
FINGER- TIGHT
oL
BOLT TORQUE
Figure 7.7 Effect of bolt torque on bearing strength of fibrous composite laminates. (100 ksi ='
700 mn/m 2)
The effect of increases or decreases of bearing strengths, in relation to tensile strengths.

on bolted composite strengths is characterized in Figure 7.8, a modification of Figuri:
7.3. For carbon-epoxy laminates having a plateau of failure strengths in the tensik
failure mode, a higher bearing strength cannot increase the joint strength; the measured
increases are due mainly to additional load transfer by friction and to an effective
increase in the laminate strength in tension because of the lateral compression associated
with the higher torque and tension in the fastener. However, a substantial loss of bearing
strength, caused for example by the failure to tighten the nut down in contact with the
laminate is shown in Figure 7.8 to cause a drastic shift in the failure mode as a function
of the wid ratio. This strength reduction is always present, albeit to a lesser extent than
shown, whenever flush (countersunk) fasteners are used.

Fibreglass-epoxy laminates tend to have a higher ratio of tension-to-bearing
strengths than do carbon-epoxy laminates. Thus, there is virtually no plateau in the
joint-strength-versus-geometry relationship depicted in Figure 7.8. The peak strength
for bolted fibreglass joints is obtained at the abrupt transition between bearing and
tension-through-the-hole failure modes, again at a wid ratio closer to 3 than 4.
Additional bolt clamp-up is now seen to provide additional strength and an upward shift
in the optimum wid ratio. This effect would be seen in carbon-epoxy composites only
for laminates with excessive orthotropy or inadequate dispersion of the plies because
the weakest link would then be in the resin.
50
FIBREGLASS·EPOXY
40 TENSION STRENGTH pt
.~
30
JOINT STRENGTH
(KSI)
20
10
0.2 0.4 0.6 0.8 1.0

d/wRATIO
Figure 7.8 relation between bearing and tension failures for carbon-epoxy and fibreglass-epoxy
laminates as functions of joint geometry (50 ksi = 350 mn/m2)
There is another significant effect associated with clamp-up in a bolted composite joint.
Testing (Reference 6) has shown consistently that, all other factors being equal, the
bearing strength of the laminate in the middle of a double-shear sandwich is greater
than that of the laminates on the outer faces. This is purely the result of better
confinement of any initial damage. It is appropriate, therefore, to design splice plates to
be slightly thicker (by about 20 percent) than a simple balanced design would suggest,
as shown in Figure 7.9. It is also helpful to use large, stiff washers under the fastener
head and nut to improve the bearing strength of splice plates.
Indeed, it has been shown in the material of Reference 3, and by testing elsewhere,
that the use of such washers can actually alter the failure mode (see Figure 7.1 0). There
often is no visible bearing damage at all in the high bearing stress area under the
220 L.J. Hart-Smith
washers, with the damage initiating by delamination at the edge of the washer.
Reference 7 contains clear photographs and sections of the same kind of behaviour with
glass-fibre-reinforced plastics.
EXCELLENT CLAMP-UP IN CENTRAL REDUCED CLAMP-UP IN SPLICE PLATE

LAMINATE SANDWICHED BETWEEN SPLICES AS SOLT BENDS UNDER LOAD
p
-
p
THE BOLT-BENDING EFFECT SHOWN IS ACTUALLY MINIMAL FOR d = t BUT

BECOMES PROGRESSIVELY MORE SEVERE FOR SMALLER BOLT DIAMETERS
THE CONSEQUENCES OF THIS EFFECT OF BOLT BENDING IS EVEN MORE
PRONOUNCED FOR COMPRESSIVE LOADS
Figure 7.9 Need for reinforcement of composite splice plates
y---,
NO DAMAGE IN AREA OF
HIGHEST BEARING STRESS
VIEWY·Y
NOTE:
ONE WOULD ALSO EXPECT SHEAROUT PLUGS OF THE SAME WIDTH AS
THE WASHER, RATHER THAN OF THE BOLT SHANK DIAMETER, FOR HIGHLY
ORTHOTROPIC LAMINATES UNDER EQUIVALENT CIRCUMSTANCES.
Figure 7.10 Composite bearing failures at the edge of a washer rather than at the bolt shank for a
highly torqued fastener
The author has observed that most bolted joints designed for fibrous composite
structures have a bias towards the use of fasteners of less-than-optimum diameter. This
happens because the shear strength of steel and titanium fasteners is much greater than
most composites can develop because of their limited bearing strengths. The
consequence of using too small a fastener is usually substantial bolt bending long
before laminate failure at a significantly lower load than the laminate could have
withstood had a larger and stiffer fastener been installed in the same strip of composite.
The fastener diameter can be almost half the strip width (or bolt pitch) before the loss of
net-tension area becomes more significant than the benefits of a reduction in bearing
stress. As a rough rule of thumb, steel or titanium fasteners should have about the same
diameter, d, as the central laminate thickness, t, as defined in Figure 7.9. The small
weight penalty of heavier fasteners is more than offset by the increased operating stress
that can then be justified for the entire composite laminate. This is just as true for
structural repairs in service as for splices made during initial fabrication. (There is a
common misconception throughout much of the aircraft industry that the "optimum"
design has zero margins with respect to all possible failure modes. 3 Bolted and riveted
joints are often designed to be simultaneously critical in net-section strength, bearing
strength, shear-out strength, and fastener shear and bending strengths. In fact, this is
what is scientifically referred to as a constrained optimal result, and is far weaker than
for a true optimum, for which only one - or at the most two - conditions are critical,
even for ductile metal structures. In the context of fibrous composite structures, this
situation is most frequently encountered as the result of the high cost of blind fasteners,
for which the strength margin will never be less than a factor of 2 for a well designed
joint. This is wrongly perceived as an opportunity to reduce cost by using smaller
fasteners to increase the bearing stress. Doing so, however, means increasing the
thickness of the composite components, to reduce the bearing stress and restore what
would otherwise have been a negative margin on the strength of the laminate. If this
procedure eliminates all margins, the perception will have been created that there is no
excess strength and that this is therefore a minimum-cost design. That is not the case, of
course; ironically, since the cost of pre-preg is so high, the cost may be even higher
than the earlier design with "too high" a fastener cost.)
The fastener rotation associated with single-shear joints decreases the strength of
bolted composite joints, compared with double-shear configurations, by two
mechanisms. These different bearing strengths must be accounted for in design. Figure
7.11, taken from Reference 2, gives an indication of the reduction in bearing strength
due to bolt rotation and bolt bending without any of the bending of the laminate that
normally would be associated with the eccentricity in load path~ Such laminate bending
3 Other unappreciated consequences of a structure that truly has zero-margins for all conditions is that, by
definition, there can be no damage tolerance and no possibiity of repair other than by replacing the component
if the loads during and after the damage had been insufficient to cause failure.
4 It is significant that much the same decrease in strength, by 20 percent, was observed for the splice plates of
the double-shear joints tested in Reference 6. The mechanism is probably the same in both cases, being the
reduction in clamp-up that is shown in Figure 7.9.
222 L.J. Hart-Smith
imposes a severe reduction on the strength of single-fastener single-shear test coupons,

so much so that designers will not permit that condition to arise in aircraft structures.
Single-row single-shear joints are used only for in-plane-shear load transfer. Any
single-shear joint transferring tensile lap shear should always have two or more rows of
fasteners to react against the eccentricity in load path, as shown in Figure 7.12.
140,---------------------------------,
TESTS CONDUCTED AT
ROOM TEMPERATURE
s
~
120-
100-
80
T3001N5208 CARBON-EPOXY LAMINATES
BEARING STRENGTH BO -
QUAS~ISOTROPIC LAMINATE
(KSI)
60- 0.25-INCH-DIAMETER LIGHTLY TORQUED BOLTS
401- NOTE:
THE EFFECT SHOWN IS FOR BOLT ROTATION ALONE, AS AT EACH
BOLT IN A SPAR-TO.SKIN JOINT. THE ADDITIONAL WEAKENING DUE
20f-
l
TO ADHEREND BENDING, AS AT A SINGLE-ROW SPLICE, HAS BEEN
EXCLUDEr-FROM THESE TESTS.
0~------~------------~------~
SINGLE SHEAR DOUBLE SHEAR
Figure 7.11 Difference between bolt bearing strengths under single and double shear. (100 ksi =
700mn/m2)
'
-£:::£ ====:=:::;::1~~\===~%'-
A. SEVERE LAMINATE BENDING WITH SINGLE-ROW
'*
OF FASTENERS IMPOSING AN ABRUPT ECCENTRICITY
~b__...oo!
- ..~C=='=dB.~*~:=:::J·-
~ -j' ;
B. REDUCTION IN BENDING MOMENTS BY GRADUAL
ECCENTRICITY ACHIEVED BY LARGE bit RATIO
(NOTE THAT COMPRESSIVE LOADING IS WORSE, IN BOTH CASES.)
Figure 7.12 Alleviation of eccentricity in single-shear joints by multiple bolt rows

The effects discussed above for carbon-epoxy and fibreglass-epoxy bolted joints are
exhibited also by boron-epoxy and even metallic structures, with minor shifts in the
optimum wid ratios. The analysis and design of bolted or riveted joints in conventional
ductile metal alloys are much simpler than those for fibrous composites because of the
nearly uniform tensile stresses across the net section through the fasteners that is
developed prior to failure. The analysis technique is summarized in Figure 7.13 and is
based on an assumption of virtually complete relief of the stress concentration factors --
by yielding in the metal alloys and by massive intra- and inter-laminar failure of the
resin (with no fibre damage) in fibreglass-epoxy laminates. A superior analysis for
fatigue of metallic joints would account for the nonuniform stresses along the lines
explained here; it would also indicate the need to reduce bearing stresses to retard crack
formation and prolong fatigue lives. Likewise, the tensile operating stresses in fibreglass
laminates should be restricted in order to control the size of the delaminations around
bolt holes and thus retain much of the compressive strength for other load conditions.
1.0 ~------------,
' ' ~ NET AREA LOSS
0 _8 '' JOINT STRENGTH IS THE LESSER OF

' BEARING STRENGTH P,,9 = F,,, x d x t
' AND TENSION STRENGTH P, = F'" x (w-d) x t
0.6
JOINT EFFICIENCY
0.4
''
FIBERGLASS-EPOXY'
0.2 (STRESS CONCENTRATIONS' '
EFFECTIVELY REDUCED TO NEAR ZERO
BY WIDESPREAD DELAMINATIONS)
0 L---L--L--~-~--~
0 0.2 0.4 0.6 0.8 1.0
d/wRATIO
Figure
7.13 Simple fully plastic design procedures for bolted joints in ductile metals and fibreglass-
epoxy laminates. (fur is ultimate bearing strength; ftu is ultimate tensile strength)
This discussion has referred to tensile shear loading of each fastener. A bolted
composite joint behaves differently under compressive shear loading because of the
dissimilar stress trajectories (see Figure 7 .14). The bearing strength of most composite
laminates is usually found to be higher in compression than in tension. For that reason,
and because of the difficulty of stabilizing test specimens under compression, relatively
few compression bearing tests are performed. Designs are frequently based on tensile-
224 L.J. Hart-Smith
bearing allowables, to be conservative, whenever compressive bearing data are not

available. An exception is made for Kevlar@ fibre reinforcement because of its known
weakness in compression. The compressive bearing strength is influenced by bolt
torque and the presence or absence of washers in much the same manner as for tensile
bearing strength.
ttttttttttttt
, rl Tr: : :1
-r,,11,,1111
I I II I I I II 1
I
I
I
I'
I I II 1 I I I II
I I I I I I I
I 1 I 1 II I 1
I 1 1 II I I
"''!'"''
'''OJ"',,
1111
1/// 1 "'''
ll]ll
Ill\ II II
',','...... .____
\~~== . . .:/
\ \ \ \ '-.. ,///I I
///1
::=~. ....... . .,./
TENSILE LAP SHEAR COMPRESSIVE LAP SHEAR
Figure 7.14 Stress trajectories around bolts for tensile and compressive lap shear
7.3 EFFECTS OF FIBRE PATTERN ON STRENGTH OF BOLTED COMPOSITE

JOINTS
One of the advantages claimed for fibrous composite structures is that orthotropic
properties can be tailored for any specific application. That design philosophy would
imply a need for bolted-joint strength data for a large variety of laminate patterns
Design of composite structures then would be quite a complex and costly task.
However, the motivation for such a philosophy derives mainly from stiffnes~
considerations; the desire to also develop adequate strength, particularly at bolt hole~
and cutouts, tends to restrict the choice of fibre patterns to those that do not deviate very
far from a quasi-isotropic pattern. As a general rule, there never should be more than
three-eighths nor less than one-eighth of the fibres in any one of the basic laminate
directions- 0°, + 45°, -45° and 90°. These fractions refer particularly to highly loaded
unidirectional tape laminates that otherwise tend to split too easily, causing wide-spread
unrepairable damage.
Most, if not all, of the major U.S. aerospace companies have similar limits in their
design procedures. However, those who have yet to discard their laminate optimization
computer programs still occasionally design structures that violate this principle and, on
test, show a distinct unwillingness to accept the load transfer at their joints that had
been expected.Thesemaximumand minimum fractions are sometimes violated on lightly
loaded minimum-gauge structure in which the only alternative is to use a nonstandard

thinner ply in the laminate. Of course, the choices for all-bonded or integrally stiffened
components on one-shot missiles are not constrained at all by any need for repairability
or damage tolerance.
These restrictions on using practical laminate patterns for aircraft are necessary
because today's high-temperature-cured brittle resin matrices are not quite adequate to
develop the full potential of the high-strength fibres and filaments. It is, therefore,
necessary to maximize the number of resin interfaces between changes in direction of
the layers of fibres; obviously, specifying a significantly orthotropic lay-up pattern
prevents this because parallel plies are inevitably clustered together to at least a
minimum extent. Similarly, the use of thicker plies or the unnecessary bunching
together of thinner plies has been found by test to impose significant reductions in
laminate strengths (which, incidentally, are not accounted for in most laminate strength
theories because they fail to address matrix failures in a scientific manner). It should be
noted that this problem is not to be solved by improving the adhesion between the resin
and the fibre - what are needed are tougher, stronger resins, that fail at a much greater
strain. Once, when composites where in their infancy, carbon-epoxy laminates were
made with improved sizing on the fibres. The laminate strengths were reduced dramatically
because the most highly loaded fibres simply broke instead of pulling out of the resin
over a short length to alleviate the locally high fibre strains. This ability of an imperfect
bond between the resin and the fibres to average out local stress concentrations results
in more of the fibres being loaded up before the final failure. This fibre/resin dis bond is
vital to the relief of the stress concentrations induced around bolt holes and cutouts in
fibrous composites,as is explained in Figure 7.15, taken from Reference 2.
Recent research into loaded and unloaded holes in composite laminates has
uncovered additional mechanisms that occur before final failure. The highly stressed
area in the immediate vicinity of the hole is surrounded by fibres subjected to far lower
loads, so there is a capacity to redistribute load, once the first fibres have broken, to
other fibres without immediately breaking them, too. In addition, in the immediate
vicinity of the hole, the ±45° fibres have considerable reserve capacity when the first 0°
fibres are loaded to failure, as in Figure 7.15. Also, in the zones where the ±45° fibres
first become critical, along the axes at ±45° to those shown in Figure 7.15, the load in
the 0° fibres drops to zero where they are interrupted by the hole. Consequently, it is
probable that catastrophic failure will be deferred until fibres in both directions have
been broken (not necessarily at the same point) and the ability to redistribute loads in a
softened zone akin to the plastic zones developed in ductile metals has been destroyed.
Indeed, there is photographic evidence of broken fibres around holes in deplied
laminates subjected to subcriticalloads.
The logical preference for near-isotropic fibre patterns in basic unnotched laminates
is even more pronounced at bolt holes. This can be seen in Figure 7.16, in which the
bearing strengths of high tensile strength (HTS) carbon-epoxy laminates are plotted for
various fibre patterns. Additional data are given in Reference 2, while Reference 8
confirms the existence of the presence and height of this plateau for HTS/914 carbon-
226 L.J. Hart-Smith
epoxy laminates. There is an obvious peak-strength plateau around the quasi-isotropic

pattern in Figure 7.16. This is particularly important since the bearing strength is thus
insensitive to small fibre pattern changes in that area, as would occur in laminate
thickness transitions. Those patterns with an excess (more than 50 percent) of 0° plies
do not develop the full bearing strengths because of premature failure by shear-out.
even at large edge distances (8 to 10 times the diameter). The location of the transition
in failure mode is known to vary with the fibrous reinforcement and probably also
varies with the resin content.
Figure 7.17 presents the corresponding characterization of what has rather
inappropriately been called the shear-out strength. It can be seen that, despite the shmt
edge distance (eld = 2) and large widths (wid= 8 to 12), all of the failures in the lower
part of the diagram were by tension-through-the-hole. The only shear-out failures
occurred in the excessively orthotropic patterns at the top of the diagram. It may seem
that the absence of shear-out failures for the testing in the lower portion of Figure 7 .l fi
would imply that an e/d ratio of 2 is adequate. That is not so because those failures
occurred at bearing stresses only about 80 percent as high as for e/d ratios of 8, which is
the value beyond which no further increase can be realized. An edge distance of 3d is
shown in Figures 37 and 38 of Reference 2 to develop some 90 percent of the full
bearing strength of the laminate. Figures 16 and 17 were constructed from test data on
about 15 specific fibre patterns.
~REMOTE
NO DELAMINATION STRESS
) DELAMj ATION AND

MATRIX CRACKING
BOLT HOLE \
f-~
HIGH PEAK
STRESSES IN o•
l NO CHANGE IN
FIBER STRESS
T HROUGHOUT
DELAMINATEO
AREA , /
FIBERS ADJACENT
T O HOLE REDUCED PEAK FIBER
STRESS BECAUSE
STRAIN SPIKE IS
AVERAGED OVER A
FINITE LENGTH
Figure 7.15 Stress concentration relief in fibrous composites by delaminations

100
90
BEARING STRESS CONTOURS (KSI)
90
~
70
60
PERCENT
50 BOLT DIAMETER = 1/4 INCH
O• PLIES (TORQUED BOLTS)
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
PERCENT ±45° PLIES
Figure 7.16 Bearing stress contours for bolted carbon-epoxy joints (100 ksi = 700 mn/m2)
100
90 SHEAROUT STRESS CONTOURS (KSI)
90
SHE ROUT FAILURES
70 BOLT DIAMETER= 1/41NCH
e/d=2
60 (TORQUED BOLTS)
l
PERCENT NO SHEAROUT FAILURES
50
O• PLIES
40
30
20
oL-~~L-~~~~~--~~~~~
0 10 20 30 40 50 60 70 80 90 100
PERCENT ±45• PLIES
Figure 7.17 Shear-out stress contours for bolted carbon-epoxy joints (100 ksi = 700 mnlm2)
228 L.J. Hart-Smith
The apparent shear-out strengths are not material properties at all and are known to vary
with the edge distance. Whereas an e/d ratio of 2 is considered to be the usual minimum
for metal alloys, a value of 3 is more appropriate for near isotropic fibrous composite
laminates. (This value of 3 is appropriate for single fasteners or a seam of fasteners
subjected to a load perpendicular to the end of the laminate. A lesser eld ratio of2.5 can
be justified for running shear loads parallel to the edge when there is only a negligible
tensile load to consider.) No matter how large the edge distance is, it is not sufficient for
highly orthotropic laminates, as can be seen from Figures 17 and 18. The absence of all
90° (or all ± 45°) plies in combination with an excess of oo plies aligned along the load
direction results in failures by shear-out of parallel plugs, which are often but not
always precisely as wide as the bolt hole (see Figure 7.20). These failures are premature
in the sense that laminates closer to the quasi-isotropic pattern can sustain much higher
gross-section stresses without failure. Such shear-out failures have been observed with
laminates made from unidirectional tapes in which the reinforcements have been boron,
high-modulus carbon, several high-tensile-strength carbons, and fibre glass. Woven
fabrics should not be as susceptible to this problem because of both the intralaminar
weaving and the great difficulty of actually making an excessively orthotropic laminate
out of material that inherently normally contains roughly 50 percent of cross-plies in
each layer.
30,-------------------------------~
p-~
0
20 ~r--
\ AVCO 5505 BORON·EPOXY
SHEAROUT
\ (0/45/0/-45)
STRESS, \
T, (KSI) \ 0 CURRENT DOUGLAS TESTS
\ IJ PRIOR DOUGLAS TESTS
\
10 \ A GENERAL DTNAMICS TESTS
' ', ''

.......... ',
..............
RECTANGULAR HYPERBOLA-)'"".,..,.;:-..,..,
(CONSTANT LOAD LINE) =.J' "" ""--.!1.. .......... ..
0~--~----~----~----~----~--~
0 5 10 15 20 25 30
EDGE DISTANCE RATIO, eld
Figure 7.18 Shear-out stress as function of edge distance (10 ksi = 70 mn/m2)
The coupons in Figure 7.19 are reacted at the large central hole and pulled separately on
each test bolt hole in turn. The test shown in the top right of the Figure 7.19 was stopped
prior to complete failure. It is evident that the initial failure is confined to the immediate
vicinity of the bolt. The shear-out failures are only about one bolt diameter long and the
edge of the laminate shows no sign of the protruding shear-out plug. Similar but distinct
failure modes observed at the McDonnell Aircraft Company, (now part of The Boeing
Company), in St. Louis, did not exhibit a single plug being sheared out. In some of their
tests (at a much shorter edge distance), the individual bundles of 0° plies were sheared
out after delaminating from the ± 45° and 90° plies which remained essentially
undamaged as shown in Figure 7.21.
~ ·I I
Figure 7.19 Shear-out failures for bolted composite joints in highly orthotropic composites
• SHEAROUT FAILURES FOR LARGE EDGE DISTANCES

·CHARACTERISTIC NEAT PARALLEL PLUG SHEARED OUT AT LOW LOAD
· FAILURE LOAD INSENSITIVE TO EDGE DISTANCE
• PREVALENT FAILURE MODE FOR FIBER PATIERNS CONTAINING TOO MANY 0" PLIES
Figure 7.20 Shear-out failures for bolted joints in composites
Having alluded to the strength imbalance between the high-strength reinforcements and
brittle resin matrices cured at 350°F ( 180°C), it is appropriate to record the highlights
230 L.J. Hart-Smith
from two series of tests that pertain to this subject even though they did not need any
bolt holes to develop the stress concentrations needed to induce premature failure. A
series of square quasi-isotropic laminates was made at Douglas Aircraft Company,
(now part of The Boeing Company), Long Beach, starting with basically 0.005-inch
(0.13-mm) unidirectional tape (T300/N5208). All laminates contained the same total
number of plies and were, therefore, of the same thickness. However, one specimen was
made with a change in fibre direction after each ply (a "homogeneous" lay-up) - to
maximize the number of effective resin interfaces - while the others were laid up two
plies at a time, four plies at a time, and finally eight at a time, corresponding to a 0.040-
inch ( 1.0-mm) tape. After the square laminates were trimmed carefully using a diamond
slitting wheel, the edges of the laminates made from bundles of eight plies and four
plies were so badly split by the contraction of the resin during the cool-down after cure
as to experience resin damage substantial enough to be visible to the naked eye. Even
those laminates made from two-ply bundles suffered damage that was visible under a
low magnification eyepiece. Only the specimen laid up one ply at a time was considered
free from such internal delaminations. These phenomena also arise on the "edges" that
exist in the form of bolt holes and cut-outs. They reinforce the need for complete
composite failure criteria, such as in Reference 9, that are formulated at the constituent
level and which include all stresses and strains in the matrix, including the major
increments customarily ignored that exist within each lamina.
TEST COUPON
m..OCK$ OF 0" PLIES

SHI!'AAE.O OUT SEPARATEL.'f
Figure 7.21 Shear-out failures at bolt holes in laminates with clustered parallel plies
The second series of tests was performed at NASA's Langley Research Center, Virginia
(Reference 10) and consisted of tension-tension fatigue testing of laminates containing
three 0.005-inch (0. 13-mm) layers of 90° fibres as a bunch sandwiched between
surface layers each having four plies in the± 30° direction. This laminate was deliberately
selected to aggravate the problem of resin weakness at a change in fibre direction.
Despite notch-free sides and the absence of any stress riser such as a hole, by the end of
the test the laminate had been reduced to three distinct wafers and the 90° core had been
repeatedly split transversely between the fibres. However, not a single fibre had been
broken, proving even by the exception the remarkable fatigue strength of fibrous
composite laminates.
The significance of these delamination failures is that in bolted composite joints those
resin interfaces are needed to shear the load as the fibres change direction around the
hole. In doing so, there already is considerable intralaminar and interlaminar resin
failure even with thoroughly interspersed thin-ply lay-ups, as indicated in Figure 7.15.
Examples of such delaminations are shown in Reference 4 and have been reported in
ASTM proceedings in the USA also. It is standard practice at McDonnell Aircraft
Company (now part of Boeing) to define the onset of such delaminations as the limit-
load capacity of bolted composite structures. The consequent load redistribution has
always been found to be sufficient to generate at least an additional 50-percent greater
load capacity prior to catastrophic failure, which is consistent with the test data in
Figure 7.5. The ultimate failures of the pin-loaded holes correspond with the onset of
damage (i.e., limit load) around the torqued bolts. The additional increment of load
carried by the torqued bolts is the result of through-the-thickness clamp-up, which
retards the spreading of initial delaminations. (The U. S. Navy has long held a more
conservative view that the first matrix failures defined laminate ultimate design
strengths. However, their acceptance of the constituent-based failure model based on
the work of Gosse (see Reference 9) is leading to a re-appraisal of this position now
that there can be more confidence in theoretical predictions of strength from physics-
based models than was possible from earlier abstract-mathematical models relying on
artificially homogenized "composite materials.")
As larger and thicker composite structures are studied and being built, there is
considerable pressure on engineering designers to reduce fabrication costs by
thickening the basic plies - after all, how can composites compete with a single slab of
aluminum alloy for a wing skin that is 10 feet wide, 100 feet long, and more than half
an inch thick, particularly if the composite skin alone (not to mention the stiffeners)
entails the laying up of more than I 00 layers of composite tape or about 50 layers of
woven fabric? As long as the resin remains the weak link in the composite, it is
probable that these pressures will not achieve an appreciable reduction in the number of
plies to be laid up but will, instead, merely double the thickness and weight of large
bolted composite structures! In addition to this, the design philosophy of the aircraft
regulatory agencies throughout the world has been to allow no permanent structural
damage at less than limit load for conventional ductile metal alloys, How would they
cope with thick-plied composite structures that contained millions of small but
detectable internal delaminations before such an aircraft was even assembled? Worse
yet, any trend towards thickened plies would have the effect of increasing the minimum
gauges (or cost by using what would by then be nonstandard thin plies) on those lightly
stressed components such as control surfaces and fairings that are widely used today.
Significantly, today, economies of scale from making fibres in 12K bundles are evident
in thin tape layers that used to be made only from 3K tows of fibres, by flattening the
232 L.J. Hart-Smith
bundles before impregnating or weaving them. This represents a vast improvement over
the older and far thicker woven fabrics made from 12K tows that led to what came to be
known as the Kevlar<.W-Nomex(~ problem. It is possible to reduce cost without
degrading performance.
It seems to the author that a far more fruitful approach for large subsonic transport
aircraft would be to move away from the brittle (350°F- or 180°C-cured) resins that are
full of microcracks towards ductile (250°F- or l20°C-cured) resins of much greater
toughness, and to accept these associated manufacturing savings instead. It is
interesting to note in this regard that the first, and for many years the only all-composite
aircraft already certified by the FAA - the Windecker Eagle - is made from a room-
temperature-cured wet-lay-up resin and has not suffered measurable degradation in
decades ofuse'.
Before leaving the related subjects of laminate fibre pattern and lay-up sequence, a
few words on the provision for repair in service are warranted. At one extreme are the
optimized highly orthotropic laminates, with local modifications giving closer to quasi-
isotropic patterns and local buildups in thickness in the areas of fastener seams. By
including as well the use of glass softening strips in basically carbon-epoxy laminates,
Douglas Aircraft (now part of Boeing) once reduced the stress concentrations at bolt
holes sufficiently to demonstrate by test structural efficiencies in excess of 90 percent
of the unnotched laminate strength (see Reference 11). Unfortunately, thick laminates -
one-fourth inch (6.3 mm) or more- of such fibre patterns are to all intents and purposes
unrepairable and, therefore, quite unsuitable for most aircraft applications. The other
extreme is exemplified by the carbon-epoxy wing on the AV-8B Harrier. This wing
skin has been made from the quasi-isotropic pattern (25% 0°, 50% ± 45°, 25% 90°)
throughout, precisely to permit simple bolted repairs - even under battlefield conditions
-without the need to seek help from the manufacturer for each and every location of
damage.
7.4 EXPERIMENTAL TEST PROGRAM
Various experimental investigations into bolted joints in fibrous composite laminates

have repeatedly shown that the most efficient mechanical joints are associated with
near-quasi-isotropic fibre patterns. One might expect then that the analysis of such
joints would involve little more than the determination of the appropriate geometric
stress concentration, particularly since those unnotched laminates exhibit virtually no
nonlinear behaviour prior to failure. Such an approach has been shown to be far too
conservative. Fibrous composites exhibit considerable nonlinear behaviour in the
immediate vicinity of stress concentrations and free edges, as discussed in Reference I.
This is because the composites have distinct fibre and matrix phases and, while each
behaves essentially linearly until failure, local failure of the interface between the two
can cause significant modification of the internal stress distributions, as explained in
' The Beech Starship was the next such certified aircraft.
Figure 7.15. (It should be noted that the individual constituents of the composites of
materials remain linear to failure. It is only when viewed at the lamina or laminate
level, that seemingly nonlinear behaviour appears.)
Further "nonlinear" behaviour prior to eventual failure is associated with local splitting
of the resin between the fibres in a single layer and with local de laminations between
the plies at a change in fibre orientation. Those effects could actually be accounted for
with a linear analysis by modeling the composites not as homogeneous orthotropic
materials but as individual constituents and interfaces. The complexity of such an
idealization will continue to prevent its widespread use. Nevertheless, the usual analysis
of fibrous composites as single-phase materials should not be misinterpreted as
implying that they actually behave that way just prior to failure. In applying Gosse's
strain-invariant models for matrix failures (Reference 9), the primary structural analyses
continue to be made at the customary homogenized lamina level. This is followed by
automated post-processing, element by element where needed, in which the differences
between the transverse strains in the matrix, fibres, and laminae are properly established
before the failure criteria are applied. There has never been any need for the large
finite-element models to include discrete fibres.
The interpretation of bolted joint data in References 1 and 2 makes it clear that
analyses based on either perfectly elastic or fully plastic behaviour would be seriously
erroneous. The actual results were found to be roughly half way between those widely
different extremes, as shown in Figure 7.22. The correct behaviour could not be
predicted on the basis of a minor perturbation of either the linear elastic or fully plastic
analysis. The key to correlating the test results with the analyses in References 1 and 2
was an experimentally determined correlation factor.
0.7 . - - - - - - - - - - - - - - ,
0.6
0.5
JOINT 0.4
STRUCTURAL
EFFICIENCY
[P/(F,"wt)] 0.3
0.2
0.1
0.2 0.4 0.6 0.8 1.0

RATIO OF BOLT DIAMETER TO STRIP WIDTH, dlw
Figure 7.22 Relation between strengths of bolted joints in ductile, fibrous composite and brittle
materials
234 L.J. Hart-Smith
Another analytical approach for bolted composite joints is to be found in Reference 12.
That analysis of the stress field around the bolt holes is linear, but the prediction of
failure is made not at the surface of the fastener, where the failure initiates, but at a
small offset (where failure does not occur)- on the order of 0.020 inch (0.5 mm) away
from the hole. The philosophy behind that approach is that any nonlinear behaviour
prior to limit load will be confined to the immediate vicinity of the hole (even though
damage may spread more than a diameter away from the hole just prior to failure). A
perfectly linear elastic analysis should apply beyond that zone. The magnitude of that
offset must be established empirically on the basis of specific test data. This approach is
just as empirical as the C-factor method presented here- the needed offset dimensions
are NOT characteristic material properties, even though they are often portrayed as
such. The utility of this approach rests on how may different joint geometries can be
characterized by a single offset dimension. Experience has shown that a new dimension
is needed for each different failure mechanism, as one should intuitively expect. When
this approach was used for the wing skins of the McDonnell F/A-18 and AV-8B
aircraft, it was found that a single offset dimension covered the many non-bearing
critical cases of possible tension-through-the-hole failures; tests indicated a need for
redesign at the few locations that were bearing critical, instead.
This need for a major correlation factor, rather than just a minor perturbation, with
respect to some basic analysis means that the reference analysis can be quite simple and
still achieve the same end results as a more complex analysis combined with a slightly
smaller empirical modification. It seems logical, therefore, that the reference analysis
should be that for linear elastic isotropic materials, since so many such analyses arc
already available. This is the thinking behind the interpretation of the test data reported
in Reference 2 in a form that permitted the test results to be generalized to other
configurations for which specific test results were not available.
The efficacy of this approach has been demonstrated in Reference 6 in which large
multi-row bolted composite joints were designed on the basis of single-row test data.
Before-the-fact predictions of joint strength were often within a few percent of the test
results (except for some joints that needed reanalyzing because of premature
delamination of the splice plates at spot faces around the bolt holes); the best of thi..~
joints so designed attained gross-section failure strains of 0.005 at load levels of about
50,000 lb/inch (88 kN/m).
The first step in planning a test program to characterize the behaviour of bolted
joints in a particular composite laminate is the selection of the fibre pattern and lay-up
sequence. Wherever possible, this should be from within the area identified in Figure
7.23, throughout which the joint response has consistently been found to be well
behaved. The designer is sometimes forced into the other area by such considerations as
minimum-gauge requirements and the area identified should not be regarded as
surrounded by abrupt strength losses. However, that other area combines significant
strength losses with known anomalies in behaviour and generally requires more
extensive testing and more detailed analyses. As an example, consider the simple
problem of an unloaded bolt hole in a large composite panel under uniaxial load.
Existing test data (References I and 2) show that for the quasi-isotropic fibre pattern,
the theoretical isotropic stress concentration factor of 3.0 had been reduced to 1.5 for
carbon-epoxy laminates. For two different laminates having three-eighths of the fibres
in the 0° (load) direction, the corresponding stress concentration factor had been
determined by test to be I. 75 at failure of the composite. (This increased stress
concentration factor is associated with an even higher unnotched laminate strength,
leading to a net strength gain of about 15 percent with respect to the quasi-isotropic
laminate.)
• PERCENTAGE
100 0 90• PLIES
FISER PATTERNS TO BE AVOIDED,
PART ICULAR FOR STRUCTURES W ITH
HIGH FASTENER LOADS
EXPANDED DESIGN REGIME

FOR LIGHTLY LOADED
STRUCTURES
RECOMMENDED FIBER
PERCENTAGE PATTERNS FOR HIGHLY
o• PLIES LOADED STRUCTURES
10 20 30 40 50 60 70 80 90 100
PERCENTAGE ~45• PLIES
Figure 7.23 Selection of lay-up pattern for fibrous composite laminates
On the basis of these known results, a stress concentration factor of 2.0 had been
predicted for a tape laminate containing 50 percent oofibres and 50 percent 90° fibres -
a laminate that does not seem to be too far outside the shaded area in Figure 7.23. The
measured strength corresponded to a net-section stress-concentration factor very close
to 1.0. The reason was long (3d) cracks between the oofibres originating at the tangents
to the hole that effectively unloaded the laminate throughout a strip as wide as the hole.
Not a single fibre had been broken by the tests and only extremely small interply
delaminations were detected. Similar behaviour has been reported in the literature for
other orthotropic laminates on a fairly regular basis. The problem with not knowing
when these matrix failures occurred is that they severely limit any subsequent load
capacity with respect to transverse loads.
Having identified the fibre pattern, or patterns, to be tested for a particular fibrous
material, the next step is to identify the minimum set of joint geometries to be tested. A
useful starting point is the single composite laminate sandwiched between two metal
236 L.J. Hart-Smith
splice plates and loaded in double shear to develop the highest possible strengths
which can subsequently be modified by knockdown factors based on other tests found
in the literature. Supplementary tests should be run to establish the reduced strengths
for flush fasteners, for example, and to verify the strengths of specific highly loaded
multi-row joints after they have been designed based on data from single-hole tests.
Depending on the needs of each design application and the presence or lack of some
pertinent data base, the tests may have to be run for a single fastener diameter or for a
range of sizes. The area of softening caused by the delamination and splitting of the
resin is related to the ply thickness and fibre diameter and, therefore, decreases
proportionally as the fastener (or cut-out) diameter is increased.
For carbon-epoxy laminates, the ultimate tensile bearing strengths are achieved for a
laminate having a width, w, and edge distance, e, both equal to eight bolt diameters, d.
The transition between bearing and tension-through-the-hole failures has been found to
occur at a strip width somewhere in the range of 5 to 6 bolt diameters. Test strips
narrower than 8d will consequently develop lower bearing stresses at failure (in
tension). The laminate thicknesses should be roughly equal to the bolt diameter except
that proportionally smaller thicknesses will be needed for small bolt sizes to prevent
excessive bending or failure of the bolts. If the design is subject to geometric
constraints that preclude an adequate edge distance or, in particular, if the laminate
pattern lies outside the preferred area in Figure 7.22, one should repeat the bearing test
at a short edge distance of 2d or so. If the fibre reinforcement is KevlarcRJ, which is
weak in compression, or if the load spectra warrant it, it may be appropriate to repeat
the above tensile bearing test in compression along the lines described in Reference 2.
Glass fibres have much higher strains to failure than the so-called advanced composites,
so the specimen width and edge distance could perhaps be reduced to 6d.
The next tests cover tension-through-the-hole failures for both loaded and unloaded
holes. For the loaded bolt holes, the strips of carbon-fibre-reinforced plastic should be
about 3d wide, corresponding with the highest possible single-hole joint strengths. For
glass fibres, a width of 2.5d, or perhaps as little as 2d, might be more appropriate. These
particular tests should be as free as possible from any influence of bearing failure.
Again, he laminate should be sandwiched between metal clevises and the bolt should be
only lightly torqued. It has not yet become usual to run these tests under compressive
loading as well as tensile loading because, for compressive loading, the stress
concentration factor is reduced by the load transferred through the fastener in bearing.
In the absence of specific test data, the author advocates the use of a stress-concentration
allowance for compression that is half as severe as for tension. Experience suggests
that, just as for ductile metal alloys, the compressive design cases for bolted joints arc
usually less severe than for tensile shear loading. Unloaded hole tests should be run at a
strip width of 4d, taking advantage of light clamp-up by a bolt in the hole. The precise
fit of the fastener in any of the single-hole 5 tests described above is known not to be critical
5 In bolted or riveted joints with multiple fasteners,

sloppy holes can easily result in load sharing far different
than predicted on the basis of net-fit fastener/hole combinations. So the issue of hole cleamces cannot be
ignored for complete structures.
because of the substantial non-catastrophic damage sustained by the laminates prior to

eventual failure. These unloaded-hole data are needed in the design of multi-row bolted
joints, in which the bolt rows are typically 5d apart for the critical row in carbon-epoxy
or 4d apart in fibre glass-epoxy. If all of the tensile tests above have failed with a visibly
similar failure mode, preferably clearly by tension-through-the-hole, the need for
further testing will be minimized. The results should be compared against known results
in the literature to identify any anomalies or need for retesting.
Experience with the newer toughened resin matrices has exposed a further potential
failure mode, that of through-thickness cracking under compressive in-plane fatigue
loads that has all of the hallmarks of traditional fatigue cracks at holes in metallic
structures subjected to tensile loads. These compression-induced cracks in composite
laminates occur at gross-section strain levels that were impossible to reach with the
older brittle resin matrices. Whereas the local stress-concentration relief alongside the
holes due to matrix failures is benign for tensile loads, since it allows the fibers to reach
a higher gross-section strain before they break, the same damage destabilizes
compressed fibers, allowing them to buckle and break. The first such fracture sheds
load to immediately adjacent fibers. When they fail, the adjacent fibers are forced to
accept the load until they fail too, leaving in their wake a "fracture" surface of broken
fibres. This failure mechanism is clearly triggered by the same micro-damage as
appears under tensile loads in the presence of bolt holes and cut-outs. However, the
gross-section strain levels at which the compression cracks would occur will be raised
above the ultimate strains-to-failure for tensile loads because some of the bypass loads
can be transmitted through the bolt, then, instead of having to all be diverted around it,
as happens with tensile loads. While this new compression-failure mechanism is known
to exist, because other researchers discovered it, the author is unaware of test coupons
or programs used to characterize it; so it is not discussed further here.
Based on the test program outlined above, one might need as little as three test
geometries for each bolt diameter and material combination. The reasons for
recommending the specific joint geometries above will become apparent in the
following section, in which the reduction of these data is discussed.
7.5 CORRELATION BETWEEN TEST AND THEORY
The hypothesis on which the author's analysis of bolted composite joints, contained in
References 1 and 2, is based is that the stress concentration relief due to resin damage
prior to fibre breakage is proportional to the intensity of the stress concentration factor
itself. In other words, the stress concentration factor kt.c on the net section of the
composite laminate is proportional to (and less severe than) the elastic isotropic stress
concentration factor k1e for an ideal joint having the same geometry. This is explained in
Figure 7.24, which compares hypothetical but typical stress concentration factors
observed at failure in quasi-isotropic carbon-epoxy laminates with the corresponding
238 L.J. Hart-Smith
theoretical predictions for perfectly elastic isotropic materials. Actual plots of results
such as these are given in References l, 2, and 6. It is immediately evident that there is
substantial relief from the theoretical predictions; this is expressed more explicitly by
the higher strengths shown in Figure 7.22. The reasons for picking the specific values
of the wid ratios are also evident. In the right side of the figure, the results for the
bearing failures (large wid ratios) should lie consistently above the line drawn through
the tension-through-the-hole failures by a steadily increasing increment.
STRESS CONCENTRATION
FACTOR k,e
ON NET SECTION OF 3
FIBROUS COMPOSITE
LAM INATE AT FA ILURE
3 4 5 6 7 8
THEORETICAL ELASTIC STRESS CONCENTRATION FACTOR k"
OPEN HOLES ~
APPROXIMATE { • I I I
wid VAl UES l OADED HOLES .L2- - - ' - ,_ _ ! _ _. l __ _ l __ _. L __ _L
Figure 7.24 Relation between stress concentration factors observed at failure of fibrous composite
laminates and predicted for perfectly elastic isotropic materials
Actually, the consistency of the loaded and unloaded hole data is not always as good as
shown in Figure 7.24. However, there are as many sets of results for unloaded holes
d::..tin~tly above the line through the loaded hole data as below. Also, some of those
results substantially below that line are known to have a markedly different failure
mode.
Relationships of the type shown in Figure 7.24 have been prepared for different
carbon-epoxy materials, for different fibre patterns, and for different fastener diameters.
The results are subject to many influences - particularly to bolt clamp-up - and the
scatter is too great to justify the fitting of anything other than a straight line through the
data. Nevertheless, this approach has made it fairly simple to predict reliably the
strength of bolted joints that have not been tested, on the basis of a very limited number
of tests. In Reference 13, this method ha ~ even been extended to in-plane-shear loads on
multi-row joints, for which no test data are yet available. The influence of composite
material orthotropy on the stress concentration factor might have been accounted for by
modifying the abscissa k1e, by an appropriate factor. However, provided that the failure
mode and location are the same, the same result is achieved by absorbing that additional
factor into the slope of the line in Figure 7.24.
Thus,
ktc -1 = C(k1c -1) , (7.1)
where the coefficient C provides the necessary correlation between test and theory. As
an example, C has a value of about 0.25 for quasi-isotropic carbon-epoxy laminates
made from 0.005-inch (0.13-mm) thick tape in brittle (350°F- or 180°C-cured) resins. A
thorough quantitative treatment of this aspect of the problem is given in References 1
and 2. The factor Cis sensitive to hole size, being approximated by the formula
(7.2)
when the fastener diameter d is expressed in inches and e is the standard exponential
function. C is equal to 0 for zero-diameter holes, indicating the absence of any stress
concentrations.
The expressions for the stress concentration factors k1e are quite simple. For
example, for an unloaded hole of diameter din a finite strip of width w,
(7.3)
The corresponding stress concentration factor for an infinite plate with a row of
unloaded holes at a pitch p can be expressed by the empirical formula
d )1.5
kte =1+2( 1-- (7.4)
•
w
Likewise, for a loaded hole in a finite-width strip, with the center of the hole a distance
e from the edge (end) ofthe laminate,
(7.5)
in which
8 = ( :· - 1) for eI w ~1 (7.6)
8 =0 for eI w z1 (7.7)
For a row of loaded holes in an infinite panel, at a pitch p,

240 L.J. Hart-Smith
kte =E+o.s(1-~)e
d p ' ""E
d '
(7.8)
in which
0 =( ~ -1) for eI p ~ 1, (7.9)
0 =0 for e I p 2': 1 (7.1 0)
The stress concentration factor at failure is computed as
k = Ftu(w-d)
tc p '
(7 .11)
in which P is the failing load and Ftu the unnotched ultimate tensile strength of the
laminate. (Any improved higher estimate of Ftu would clearly be automatically
compensated for by a proportionately higher value of ktc, so this theory is not tied to any
specific unnotched failure criterion.)
With these formulae, together with the bearing strength cutoff, it now becomes
possible to predict the behaviour of each fastener in a multi-row joint, even without
testing such more complex joints. Tests of this type which have been run already (see
References 1 and 2) have shown a linear interaction between the bearing and bypass
loads (as explained in Figure 7.2) from as long ago as 1970 when what appears to have
been the first such tests were run at the Convair division of General Dynamics in San
Diego. That linear interaction is the basis of the unkinked straight line in Figure 7.2 for
narrow strips.
For wider strips, the kinked straight line can be formed by combining the bearing
strength cutoff with a theoretical extension of the formulae above to the unmeasurable
tensile load that the laminate could have endured had it not failed first in bolt bearing.
The BJSFM analysis method (Reference 12) developed at McDonnell Aircraft (no'W
part of Boeing) also predicts a straight line for each segment of the bearing-bypas~
interaction, with any kink associated with a change in failure mode.
Similar bearing-bypass interactions can be developed for compressive loading also.
as shown in Figure 7.25, again on the basis of very few tests. References 6 and 14
contain a more detailed discussion of this aspect. There are apparently no tests at all to
confirm the corresponding assessment for in-plane shear loads postulated in Reference
13.
7.6 JOINT EFFICIENCY CHARTS FOR DESIGNING BOLTED COMPOSITE

JOINTS
The combination of the formulae above and the correlation factor C, together with the
bearing strength cutoffs, makes it possible to predict the joint strengths for any geometry
of single-row or multi-row bolted joints subjected to tensile or compressive double-

shear loads. When such joint strengths are divided by the unnotched gross-section
laminate strengths, one would predict the corresponding joint efficiency. Such
assessments have shown that the most carefully designed fibrous composite joints will
achieve an efficiency of no more than 40 to 50 percent, no matter how many fasteners
or how much design finesse or analysis effort is expended. To place that number in
perspective, the usual design range for ductile metal alloys is an efficiency of 70 to 80
percent. The lower numbers in Figure 7.13 refer to single-row joints.
CONSTANT TOTAL
BEARING STRESS
CONSTANT
TOTAL LOAD
~ COMPRESSION
p~~M'~
;/~
B~/-
Figure 7.25 Outer envelope of bearing/bypass load interactions
Now, ifthatjoint efficiency were to be multiplied by the ultimate fibre strain to failure,
the strength of bolted joints in fibrous composites could be characterized in a manner
that is very easily related to the conventional design process for composites. This has
been done in Figure 7.26 (adapted from Reference 1), which encompasses the entire
gamut from unloaded holes at the top to the transfer of 100 percent of the load on a
single fastener at the bottom. The interval between covers all possible forms of multi-
row bolted joints. Figure 7.26 was actually prepared for 0.25-inch (6.35-mm) bolts in
quasi-isotropic HTS carbon-epoxy laminates, but is quite representative of the
characteristics of other fibre patterns and material. (Note, however, that fibreglass
242 L.J. Hart-Smith
laminates would not exhibit the tensile-failure plateau on the top of the single-hole
curves, but would be more triangular in form.)
Many important conclusions can be drawn from Figure 7.26. As a reference, the
ultimate unnotched fibre strain can be taken to be 0.009, so that the strongest single-
hole bolted joint, at d/w = 0.3, has a joint efficiency of 40 percent and a gross-section
failure strain of about 0.0035. (Higher strains would apply for the newer carbon fibres,
and tests with which to establish the corresponding values of C and the joint efficiency
have indicated that the tensile strengths may be increased in proportion to the strain-to-
failure but that there has been no increase in compression and bearing strengths.
Consequently, the bearing-failure cut-offs would then be proportionally lowered in
Figure 7.26, reducing the width of the plateau and requiring that the fasteners be even
closer together to attain the maximum possible strengths.) The associated failure mode
for this peak strength is tensile and is inherently stronger than any single-bolt bearing
failure, which would require smaller values of d/w (<0.2) to realize. That optimum
single-bolt joint is hard to improve upon without considerable complexity being added
to the joint. For example, in References 1 and 2 there is an account of testing joints that
were identical to the single-bolt tests in all other regards except that they contained two
bolts in tandem. The test results consistently showed only a 10-percent increase in
strength, which is quite consistent with the predictions of Figure 7.26 in which, for the
same 3d width, the bearing stress had been halved by sharing the load between two
bolts. This seemingly surprising result is caused by the fact that the stress concentration
factor on the bypass load is very nearly as high as that for the bearing load.
0.006 r-----=--:=--===:::-::::::=::-::::::-=::===----,
r - Ob.;l Fbrv FOR MULTI-ROW BOLTED JOINTS (FAILURE
0 BY BEARING/BYPASS INTERACTION)
STRONGER BOLTED JOINT (w/ d = 5), WITH 60

0.005 MULTIPLE ROWS OF BOLTS TO REDUCE
BEARING STRESS
STRONGEST POSSIBLE 50
SINGLE-ROW BOLTED
0.004 JOINT (w/d= 3)
40 GROSS SECTION
GROSS SECTION
STRUCTURAL
STRAIN, AWAY 0.003
EFFICIENCY
FROM HOLES
30 (PERCENT)
0.002
20
0.001
10
~~-~~~~-~~-~-~~~ 0
0.5 0.6 0.7 0.8 0.9 1.0
RATIO OF BOLT DIAMETER TO STRIP WIDTH, d/w
Figure 7.26 Influence of bolted-joint design on structural efficiency of carbon-epoxy composite

structures
Figure 7.26 refers to quasi-oisotropic fibre patterns. There is a mild effect of nearby
fibre pattern on joint strength, terminated by a dramatic loss of strength for excessively
orthotropic patterns. Figure 7.27 summarizes the influence of fibre pattern on joint
strengths, being terminated at a 50-percent 0° fibre content because of the premature
shear-out failures associated with laminates containing even more 0° plies. It is evident
that bolted joint strengths are less sensitive to 0° fiber content than unnotched laminates
are. It is also apparent that even the most efficient bolted joints will have a cross-
section strength no higher than about half the corresponding unnotched strength.
That 40-percent best efficiency for single-bolt composite joints is not competitive
with major aluminum alloy primary structures on a weight basis - and the equivalent
value for an orthotropic laminate with three-eighths of the fibres in the direction of the
principal load is barely competitive. One can avoid this problem on small aircraft by
eliminating the major splices as is done on the one-piece wings of the AV-8B Harrier
and LearFan aircraft. Or one can employ an efficient but costly titanium stepped-lap
adhesively bonded splice at the wing root, as on the wing of the F/A-18 aircraft. Local
reinforcement could permit simple one-row bolted splices at the time of manufacture
(actually, two rows or one staggered row would be needed to stabilize compressive
loads) but at the cost of repairability of the basic thinner laminate. Figure 7.26 identifies
the only possible way of overcoming this limitation- by moving up and to the left from
the location of the best single-bolt joint. That move necessarily involves multiple rows
of fasteners, an increase in the bolt separation (a smaller value of d/w, down to 1-in-5 or
1-in-6), but more importantly, a drastic reduction in the bearing stress to no more than
25 percent of what the laminate could withstand. Moving the lines ofbolts further apart
seems at first to be incompatible with a major reduction in the bearing stress that
normally would be accomplished by using fasteners of a larger diameter (equivalent to
moving the bolts closer together). Fortunately, this challenge is not as daunting as it
first seems because the bearing stress needs to be reduced only at the most critically
loaded bolt stations. The other stations are rendered less critical by the progressive
transfer of load out of the basic laminate into the splice plates. The pitch-to-diameter
ratio can be reduced and the bearing stress increased away from the most critical
fastener row, as the remaining load in the laminate is reduced. The last row of fasteners
can have a d/w ratio of only 3-to-1, since that is optimum when there is no bypass load
left. This makes the splice plate critical there, since both bypass and bearing load are
thereby maximized. However, splice plates represent such a small fraction of the total
weight that it is more important to minimize the weight of the structure as a whole,
rather than of the splice plates in isolation.
The possibility of achieving joint efficiencies slightly in excess of 50 percent using
the approach described above has been verified by the test results reported in Reference
6. While difficult in terms of joint complexity, that represents a substantial 25-percent
weight saving for each entire laminate (with respect to the best single-row joints) for
the trivial penalty of a slightly heavier splice. The significance of this seemingly minor
accomplishment is that, for the first time, there has been an actual demonstration that
major composite structures large enough to need mechanical splices can, in fact, be
244 L.J. Hart-Smith
made lighter than equivalent well-designed aluminum alloy structures while not giving
up the possibility of practical bolted repairs. The attainment of such improved
efficiencies for four-row bolted joints relied upon the accurate load-sharing
computations permitted by the nonlinear computer program A4EJ (see Reference 14) in
conjunction with bilinear fastener stiffnesses deduced from test results and a formula
proposed in Reference 6. It would be premature to say that the design of major
mechanical splices in composite primary aircraft structure is as yet straightforward.
100
ANALYSIS IN TERMS OF
B·BASIS ALLOW ABLES
UN NOTCHED
80 LAMINATE STRENGTH
TYPICAL CRITICAL SOLT

IN MULRI·ROW JOINT
60 (wid= 5, 11,~ = 0.25 F.,)
GROSS
SECTION
STRESS
(KSI) 40 --
,/
20
OPTIMUM SINGLE· ROW BOLTED JOINTS (wl d =3)
CHART APPliCABLE FOR BOLTS UP TO 0.375 INCH IN DIAMETER.
LARGER BOLTS ARE ASSOCIATED WITH PROGRESSIVELY LOWE LAMINATE STRESSES.
0 o~------1~2.-5------2~5-.o-------37~.5-------5~o.-o---
PERCENT 0" PLIES
Figure 7.27 Influence of fibre pattern on bolted composite joint strength, for structural efficiency
of carbon-epoxy composite structures
Nevertheless, some definite trends have been identified by analytical studies of such
joints. Figure 7.28 shows the bolt load distributions for four different four-row bolted
joints that were analyzed to compare their strengths. Somewhat surprisingly, the
weakest joint was the scarf joint, with both the basic laminate and splice plates tapered.
This weakness was due to the loss of thickness at the first fastener station, with respect
to the basic laminate outside the joint. The joint with uniformly thick members was
predicted to have the second highest strength. That joint having tapered but
unreinforced splice plates sandwiched around a uniform skin was found to be the
second weakest. With respect to the joint having untapered splice plates, all that the
tapering accomplished was to reduce the load transfer through the outermost bolts while
not changing the criticality of the most highly loaded bolts at the middle of the splice
plates, adjacent to where the skins butt together.
The strongest joint analyzed combined tapering of the splice plates with their
reinforcement while leaving the basic laminate at a uniform thickness. The objective of
such a design is to maximize the load transfer on the last bolt in the skin, where there is
no bypass load, while minimizing the load transfer on the first bolt, where the bypass
load is greatest. The variables with which to accomplish this are the amount of
reinforcement, the minimum taper thickness, and selective variation of fastener
diameter. It should not be surprising that this joint configuration was shown to be
superior. It has long been known that the techniques for achieving high static joint
strengths in fibrous composites are precisely the same as those employed to develop
long fatigue lives in metal aircraft structures - and this superior composite joint has all
of the same features as the wing skin splice at the side of the fuselage on the DC-I 0.
BOLT LOAD DISTRIBUTIONS 4·

ROW BOLTED JOINTS
*
14
I CONFIGURATION 4
=O.OOSIN./IN.
12 I o t;ross
I I
10
a
··..
/
··.•.
--*I
I-fl..
::......I···....
.· .c
INDIVIDUAL
~ ··'*"' ..··
I . ··.. ..........
.~ a
CONFIGURATION 3
tiJross 0.0045 lN./IN.
BOLT LOADS 0.50 IN. 1 2 3 4 0.211N.
'
(1000 LBS)
-~!: i i i i tt:=
.6 .6. CONFIGURATION 1 CONFIGURATION 2
[Jo ............. q CONFIGURATION 2 Eg,0 , . - 0.00451N./IN.
o- .. --o CONFIGURATION 3
~--tl CONFIGURATION 4
CONFIGURATION 1
BOLT ROW NUMBER €gross 0.00411N./IN.
Figure 7.28 Effect of joint configuration on bolt load distribution (1,000 lb~4.448 kn, I in=25.4
mm)
The design of highly loaded joints such as described in Figure 7.28 is a specialized task
requiring data not yet generally available. However, it has become apparent that not all
of the fastener load-deflection characteristics need to be specified as accurately as one
might first think. Obviously, these stiffness characteristics are needed to establish the
correct share of the load transferred through each fastener - and they need to be
nonlinear if the predictions are to come close to matching measured joint strengths.
However, they do not need to be precise. There are several published formulae with
which to estimate the elastic stiffness of each installed fastener. It is the subsequent
(secondary) stiffness, past the proportional limit, that is not generally available. Figure
7.29 indicates that this omission need not be critical. It is customary to limit the total
permissible deformation of the hole by a characteristic 2-percent offset deflection, so the
remainder of the load-deflection curve is immaterial for design purposes. (It would be
246 L.J. Hart-Smith
needed for closely matching tests taken beyond design ultimate loads to final failure,
however.) The primary effect of the nonlinearity in the analyses is to limit the additional
load transferred by the most highly loaded fasteners while those relatively unloaded
fasteners accept more load within their linear elastic capabilities. The total load carried
by all of the fasteners is affected very little by each of the possible secondary stiffnesses
shown in Figure 7.29, which is why precision is not needed. The author would suggest
a secondary stiffness only 20 percent as high as the primary one. Even a flat secondary
slope would give reasonable predictions. In the absence of complete stress-strain
curves when only the failure loads of test coupons are recorded, the author would
suggest setting the knee in the load-deflection curve in Figure 7.29 at 85 percent of the
measured maximum - except if the bearing data were from pin-bearing tests with no
clamp-up at all, in which case an estimate 50 percent higher might be more realistic.
BEARING FAlLURE, (w/d =3)
LIMIT OF
ELASTIC
BEHAVIOR T--4--,:=-- CLOSE
APPROXIMATIONS
LOAD
/ ULTIMATE TEST·
COUPON FAILURES
/ FAR BEYOND
PRACTICAL DESIGNS
/
/
..4,..._ 2-PERCENT DIAMETER OFFSET
/
./
DEFLECTION
Figure 7.29 Idealized nonlinear fastener load-deflection characteristics
Aircraft structures contain many more lightly loaded fasteners, and quick, simple
procedures have been developed throughout the industry for such cases as the running
shear loads transferred by the bolts joining wing skins to spars. Several companies have
accomplished such tasks by means of a chart ofthe form of Figure 7.30 but with minor
variations in the values shown (see, in particular, Reference 15). If this diagram is
interpreted in the context of Figure 7.26, it can be seen that the allowable design region
is safe to use without deeper insight into the effects of the joint variables and that, with
a few simple restrictions on the d/w ratios, the 0.004 strain cutoff could well be raised
to 0.005 (for tensile loads only; there is no basis for doing so in the case of applied
compressive shear loads). Conversely, doing so would imply a need to ensure that there
were no abnonnalities in the fastener bearing stresses - e.g. , due to material thickness
changes or some very loose bolt transferring its load into the two adjacent bolts - so it
should be said that the procedures in Figure 7.30 are well founded, even though higher
strains will be needed to make highly loaded primary composite aircraft structures
competitive with tomorrow's improved metal airframes. Recent fatigue tests of a long-
overlap single-lap splice in a new low-cost metallic fuselage barrel (Reference 16) have
confinned how easy it is to eliminate the multiple cracks at fastener holes that have
plagued older designs with overlaps that were minimized through a widespread pre-
occupation with sub-optimum solutions. (The role of bending stresses caused by the
eccentricity in load path had not been fully understood.) It is now clear that today's
extensive activity in relation to the safety and continued operation of what are referred
to as ageing aircraft need not be an issue for future metallic aircraft.
0.005 1-----------------~
0.004
GROSS.SECTIO N
0.003 --- - -,
STRAIN e , (lN./IN.) I
0.002 ~~~m~~~t I
I
0.001 I
I
20 40 60 80 100 120
BEARING STRESS, 0'. ,0 (KSI )
Figure 7.30 Design technique for bolted carbon/epoxy structures (I 00 ksi = 700 mn/m2)
The design of all but highly loaded primary structural bolted or riveted joints in fibrous
composite laminates is thus seen to be straightforward in most cases. The treatment of
the exceptions is discussed in the following section. After the choice of material has
been made, the first question to be resolved is the selection of laminate lay-up pattern.
A good general-purpose pattern is the quasi-isotropic pattern (25%0°, 50% ± 45 °,
25%90°). If there is a dominant load direction, identified with the oodirection, a more
appropriate laminate would contain about three-eighths of all the fibres in that 0°
direction. The associated choice between 50% ±45 ° with 12\12% 90° or 37Yz% ±45 °
with 25% 90° would then be detennined by the relative severity of perpendicular loads
(in the 90° directions) or of torsional requirements (in the ±45 ° directions).
248 L.J. Hart-Smith
The next question to be resolved is whether or not there needs to be provision for repair
of the structure. If so, the selection of the best laminate pattern in the area of the joints
also defines the laminate pattern and thickness in the spaces in between. Experience has
indicated that the opposite approach of first selecting some idealized pattern for the
basic laminate usually compromises the strength of the joints and hence the integrity
and repairability of the structure.
However, if there is no need for either damage tolerance or provision for repair,
only those plies needed in the basic laminate need be continuous - the others in the
joint area then become local build-ups. In such one-shot applications, the use of glass
softening strips can provide even greater structural efficiencies for basically carbon-
fibre-reinforced laminates. Glass softening strips involve the substitution of local strips
of oo carbon fibres by strips of glass fibres about four hole diameters wide. No carbon
fibres are cut since those in the ± 45° and 90° directions remain continuous. Tests of
such hybrid laminates have confirmed the virtual elimination of stress concentrations at
bolt holes. The improved strengths of bolted joints in such materials are characterized
by the test results in Reference 2, which also explains how to deduce the stress
concentration alleviation factor on the basis of extremely limited test data.
The thickness of the basic laminate in the presence of lightly loaded fasteners is then
determined by the local load intensity in conjunction with the laminate stiffness and
strain level of 0.004 from Figure 7.30. For bearing stresses as high as 50 ksi (350 MPa),
the laminate strain would be restricted to 0.0035. If that meant thickening the laminate
in such an area, a tradeoff would be made between the weight of such thickening and
the weight of larger fasteners to reduce the bearing stresses. As a general rule, the
fastener diameter should be about the same as the laminate thickness, being greater for
laminates of about 1/4 inch (6.36 mm) or less, and less for laminates thicker than about
3/4 inch (19.05 mm). The shear strength of steel or titanium fasteners usually should not
even be close to critical, although the stiffness of the fasteners and their bending yield
strength can be of particular importance to the load sharing in multi-row joints (see
Reference 6). The apparent excessive shear strength of fasteners in well-designed
bolted composite joints has at times Jed to the redesign of the joint in order to use
smaller and lighter fasteners (to remove that "excess" strength margin) at the expense of
a considerable portion of the overall joint strength. That tendency needs to be resisted;
the associated reduction in laminate strength is clearly evident in Figure 7.30.
Finally, the strength of composite laminates containing bolt holes can be improved
by clamping the laminate well between large washers and protruding-head fasteners
wherever possible. Likewise, some conservatism should be exercised whenever such
clamp-up is impossible. Reliance on the extra strength associated with severe, rather
than lighter, bolt torques should be avoided except in tightly controlled special
applications.
The straightforward design procedure above becomes more complicated in cases
which, for various reasons, violate the rules for maximum possible efficiency in bolted
composite joints. Such cases usually will require additional testing. Some of those cases
are discussed in the next section.
7. 7 MISCELLANEOUS CONSIDERATIONS
Deviations from the straightforward approaches reported above can sometimes occur
due to truly unavoidable constraints. A number of these and their consequences are
discussed here. Specific tests are often necessary in such cases to generate data for
design allowables.
The most frequent limitation in joint strength is due to the need for flush fasteners to
minimize aerodynamic drag. The strength of such bolted joints in composites is reduced
both by the nonuniform bearing stress associated with the usually concurrent rotation of
the fastener under single shear and by the reduction in additional strength after the
initial failure owing to the greatly reduced clamp-up. While no completely general
explanations have been found - so specific tests are usually necessary - it was shown in
Reference 17 that a reasonable strength prediction could be made for near-isotropic
laminates by ignoring the bolt head completely and considering that the shank was able
to develop a bearing stress of about 100 to 120 ksi (700 to 800 MPa), just as for double-
shear joints. Obviously, this approximation would break down in the extreme case of
knife-edged countersinks. The countersunk head should penetrate no more than one-
third to one-half of the laminate thickness, depending on the operating bearing stress on
the shank. It is quite wrong to analyze for bearing failures on the basis of a uniform
bearing stress acting on an area equal to the product of the fastener diameter and
laminate thickness -unless the so-called bearing-stress allowable has been reduced way
below normal to compensate for the ineffectiveness of the laminate adjacent to the
countersunk fastener head. On the other hand, the full thickness of the laminate is
available for reacting bypass loads.
Another problem arises for structures so lightly loaded that it is not necessary to use
as many as eight of the minimum-gauge tape plies to sustain the loads. In the same
category are structures in which there is no significant shear load on the fasteners
because the loads acting on the structures are mainly normal, rather than in-plane.
Sometimes, tension-head (rather than shear-head) fasteners are needed to prevent the
bolt head from being pulled through the laminate. That condition can arise under both
normal pull-off and pure-shear loadings. Such larger bolt heads would require local
thickening of the laminate. Such cases require specific testing of any bolted joints, if the
weight of the structure is critical, because the failure modes are often not as well
behaved as are those of the preferred fibre patterns (Figure 7.23). Consequently, bolted
joints in such composite structures fail at lower local stresses.
If a fibrous composite structure has been designed as a replacement for an existing
metal structure, an undesirably low edge distance often is caused by mating with the
adjacent structure. This condition is particularly bad if it is associated with a fibre
pattern outside those recommended in Figure 7 .21. Reference 2 gives some idea of the
reduction in joint strength due to this condition. As a guide, the bearing strength for a
2d edge distance is only about 80 percent as high as for a large (8d) edge distance. The
short edge distance causes the stress trajectories in Figure 7.13 to bunch up much closer
to the fastener, reducing the net-section tension strength as well.
250 L.J. Hart-Smith
Considerable experience has shown that well-designed bolted composite joints are not
sensitive to cumulative damage under tensile fatigue loads. Another way of expressing
this is that those factors that reduce the fatigue strength of joints in ductile metal
structures (and have little effect on their ultimate strengths) impose a substantial
reduction on the static ultimate strength of fibrous composite structures. The same
techniques to relieve stress concentrations are needed in both situations. Actually,
gentle tensile-tensile fatigue of composites tends to increase the residual strengths of
composite structures, as shown in Figure 7.31, because of progressive relief of the
stress concentrations. Tests at NASA Langley and elsewhere have shown that the
strength can increase asymptotically towards the unnotched net-section strength. That is
to say, there is a hole-out factor with no stress concentration (a notch-insensitive
situation) just as for ductile metals. Obviously, one cannot rely on preconditioning to
increase the joint strength, but at least there usually is no concern about loss of strength
under such conditions throughout the life of the structure. Unfortunately, the same
cannot be said of the life of a composite laminate containing loose bolts pounding back
and forth in their holes under reversed (tension-compression) fatigue loading. It is most
important that the fasteners in composite laminates should not be loose and that the
laminates should not be damaged by the installation of fasteners with an excess1ve
interference.
RESIDUAL
JOINT
STRENGTH
CYCLES (LOG SCALE)
Figure 7.31 Fatigue of notched composites or bolted composite joints
7.8 RULES OF THUMB FOR DESIGNING STRUCTURALLY EFFICIENT JOINTS
• Design the joints first and fill in the gaps adterwards. Optimizing the "basic"
structure first compomises the joint design and results in low overall structural
efficiency.
• The best bolted joints can barely exceed half the strength of unnotched laminates.
• Optimum single-row joints have approximately three-fourths of the strength of
optimum four-row joints.

• Joints proportioned to attain the highest tension strength through the net section are
stronger than those designed to fail at the highest bearing stress.
• Many bolted composite joints contain too many bolts, spaced too far apart, and the
diameters are too small to permit maximizing the strength of the laminate.
• Rated shear strength of fasteners should not be a factor in design. Bolts need to be
sized to restrict the bearing stresses in the laminate.
• Peak hoop tension stress around bolt holes is roughly equal to the average bearing
stress.
• Bolt Bearing strength is sensitive to through-the-thickness clamp-up.
• Splice plate stresses should be lower than in skins to prevent delaminations.
• Bolt bending is more significant for composites than for metals, because composite
members are thicker (for a given load) and more sensitive to non-uniform bearing
stresses (because ofbrittle failure modes).
• Bolt diameter should typically be about the same size as the skin thickness.
(Slightly more for thin skins and slightly less for thick laminates.)
• Optimum wid ratio for single-row joints is about 3-to-1.
• Optimum wid ratio for multi-row bolted joints varies from row to row along the
length of the joint; wid= 5 at the first row to minimize load transfer there, and wid
= 3 at the last row to maximize load transfer there, with wid= 4 for intermediate
bolt rows.
• Bolted joint strength varies far less with percentage of zero-degree plies in fiber
pattern than does unnotched laminate strength.
• Best fiber patterns are fully interspersed (parallel plies not bunched together) and
have at least 12i5 percent of the plies in each of the four directions- 0°, +45°, -45°
and 90°.
7.9 CONCLUDING REMARKS
It is suggested in Reference 4 that the author is in favor of a "continued reliance on art,

skill, and empiricism" in the design of bolted composite joints. In the decades that have
passed since the writing of the original and more explicit final paragraph in Reference 1
- which the author of Reference 4 summarized so succinctly - considerable progress
has been made in this subject. Yet today the design of highly efficient multi-row bolted
joints is still very much an art, even though the scientific computer programs designed
to determine the load sharing and to assess the combination of the bearing and bypass
loads provide invaluable tools in that task. Furthermore, the considerable supply of test
data for precise elastic analysis of bolted composite joints has not prevented the
reporting of tension-through-the-hole strength as a material (net-section) allowable
depending only on the material and fibre pattern (as would be the case for a ductile
metal) as if it were independent of the joint geometry (the wid ratio in particular) and
the bearing/bypass load interaction. In a similar vein, the reduction in bearing stress
252 L.J. Hart-Smith
developed for those joint geometries associated with tension failures has also been
overlooked. Worse, these same errors have been perpetuated in a widely distributed
composites design manual. Perhaps the authors of Reference 4 might have added that
the present author is even more in favor of a physical understanding of the phenomena
involved as the best possible pre-requisite for designing efficient bolted composite
joints- or any other structures, for that matter.
In any event, a need remains for considerable skill in interpreting and integrating the
bulk of largely uncoordinated testing of bolted composite joints, with many
investigators interested only in specific aspects of the problem. Finally, no accurate
predictions of the strength of bolted composite joints have yet been made without
reliance on either a substantial empirical correction factor or the benefit of directly
applicable test data. And, given the widespread failure to appreciate the need to rely on
only laminate theories accounting for separate fibre and resin constituents and possible
failures at the interface, this situation is likely to persist. In other words, science alone
still cannot get the job done without a considerable dose of art, skill, and empiricism!
The purpose of this chapter has been to provide an integrated framework within
which the behaviour of bolted composite joints can better be understood. The analysis
and design of bolted joints in conventional ductile metal alloys benefit from gross
yielding of the metals to such an extent as to completely mask any imperfections in the
rather simple analyses used for that task. The design and analysis of bolted joints in
fibrous composite laminates will always remain a far more complex task because of the
extremely low strain to failure of such materials. Nevertheless, the "nonlinear"
behaviour that is created by the non-catastrophic local failures between the resin and
fibre phases of composite materials provides a considerable increase in strength beyond
that which could be achieved by equivalent elastic homogeneous materials.
Although it does not provide a large source of specific composite material data, this
chapter has identified relatively simple procedures to cover the design and analysis of
bolted composite joints throughout the regime in which they are simultaneously the
strongest and the best behaved.
7.10 ACKNOWLEDGMENT
While this chapter alludes to information from several sources, the primary source of
the material presented here has been three U.S. Government-sponsored investigations in
which the author participated. The first is the NASA Langley Contract NAS 1-13172,
Bolted Joints in Graphite-Epoxy Composites, which was reported in NASA CR-144899
in January 1977 and forms the basis of Douglas Paper 6748A, November 1978. The
second is the USAF Flight Dynamics Laboratory Contract F33615-79-C-3212, Design
Methodology for Bonded-Bolted Composite Joints, to be found in AFWAL-TR-81-
3154, dated February 1982, and of which the computer codes in Volume II were not
available for unlimited distribution. These codes have since been released for general
distribution. The third research program was the NASA Langley Contract NASI-16857,
Critical Joints in Large Composite Aircraft Structure, documented inter alia in Douglas
Paper 7266, January 1983, and in NASA CR-3710, August 1983.

The author also takes this opportunity to thank his many friends throughout the
international aerospace community who have kindly provided him with insight into
their own research and problem areas and have thereby contributed to the understanding
of the subject conveyed in this chapter.
7.11 REFERENCES
1. Hart-Smith, L. J., "Mechanically-Fastened Joints for Advanced Composites --

Phenomenological Considerations and Simple Analyses", Douglas Aircraft
Company Paper DP 6748A, November 1978. Also contained in Fibrous
Composites in Structural Design, edited by E. M. Lenoe, D. W. Oplinger, and J. J.
Burke, Plenum Press, New York, 1980.
2. Hart-Smith, L. J., "Bolted Joints in Graphite-Epoxy Composites", Douglas Aircraft
Company, NASA Langley Report NASA CR-144899, January 1977.
3. Crews, J. H. Jr., "Bolt Bearing Fatigue of a Graphite/Epoxy Laminate", Joining of
Composite Materials, ASTM STP 749, K.T. Kedward, Ed., American Society for
Testing and Materials, 1981, pp. 131-144.
4. Godwin, E. W.; Matthews, F. L.; and Kilty, P. F., "Strength of Multi-Bolt Joints in
GRP", Composites, Vol. 13, No.3, pp. 268-272, July 1982.
5. Shivakumar, K. N., and Crews, J. H. Jr., Bolt Clamp-up Relaxation in a
Graphite/Epoxy Laminate, NASA TM 83268, January 1982.
6. Nelson, W. D.; Bunin, B. L.; and Hart-Smith, L. J., "Critical Joints in Large
Composite Aircraft Structure", Douglas Aircraft Company Paper DP 7266, January
1983; also NASA CR-3710, August 1983.
7. Matthews, F. L.; Wong, C. M.; and Chyrssafitis, L., "Stress Distribution Around a
Single Bolt in Fibre-Reinforced Plastic", Composites, Vol. 13, No.3, pp. 316-322,
July 1982.
8. Collings, T. A., "On the Bearing Strengths of CFRP Laminates", Composites, Vol.
13,
No. 3, pp. 241-252, July 1982.
9. Hart-Smith, L. J. and Gosse, J. H., "Characterizing The Strength Of Fiber-Polymer
Composites Using Mechanistic Failure Models," Boeing Paper MDC OOK0050, to
be published in AIAA Text Book on Composite Materials being revised/edited by
Murray Scott.
10. O'Brien, T. K., "Characterization of Delamination Onset and Growth in a
Composite Laminate", NASA Langley Technical Memorandum 81940, January
1981.
11. Nelson, W. D., et. al., "Composite Wing Conceptual Design", Douglas Aircraft
Company, USAF Contract Report AFML-TR-73-57, March 1973.
12. Garbo, S. P., and Ogonowski, J. M., "Effect of Variances and Manufacturing
Tolerances on the Design Strength and Life of Mechanically Fastened Composite
Joints", McDonnell Aircraft Company, USAF Contract Report AFWAL-TR-81-
254 L.J. Hart-Smith
3041, Vols. 1-3, April1981.

13. Hart-Smith, L. J., "Analysis Methods for Bolted Composite Joints Subjected to In-
Plane Shear Loads", McDonnell Douglas Paper MDC 96K0086, presented to
AGARD 83rd Structures and Materials Panel, Bolted/Bonded Joints in Polymeric
Composites, Specialists Meeting, Florence, Italy, September 2-3, 1996; published
in AGARD Conference Proceedings, AGARD CP-590, Bolted/Bonded Joints in
Polymeric Composites, January 1997, p 8-1 to 8-11.
14. Hart-Smith, L. J., "Design Methodology for Bonded-Bolted Composite Joints",
Douglas Aircraft Company, USAF Contract Report AFW AL-TR-81-3154,
February 1982.
15. Garbo, S. P., "Effects of Bearing/Bypass Load Interaction on Laminate Strength",
McDonnell Aircraft Company, USAF Contract Report AFWAL-TR-81-3144,
September 1981.
16. Hart-Smith, L. J., "Low-Cost Riveted Lap Joints That Last Forever", Boeing Paper
MDC OOK0094, presented to 2000 USAF ASIP Conference, San Antonio, Texas,
December 4-7, 2000.
17. Thompson, C. E., and Hart-Smith, L. J., "Composite Material Structures- Joints",
Douglas Aircraft Company IRAD Report MDC-J0638, July 1971.

Untitled

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Untitled

Uploaded by

Copyright:

Available Formats

RECENT ADVANCES IN STRUCTURAL JOINTS AND

REPAIRS FOR COMPOSITE MATERIALS

SPRINGER-SCIENCE+BUSINESS MEDIA B.V.

ISBN 978-90-481-6319-9 ISBN 978-94-017-0329-1 (eBook)

Printed on acid-free paper

All Rights Reserved

replacement. The temporary repair (field repair or emergency repair) is usually a

L. Tong and C. Soutis

1 Stress Analysis and Failure Assessment of Lap Joints

2. Strength Determination of Adhesive Bonded Joints

3. Stresses in Mechanical Fastened Joints

4. Strength Determination of Mechanical Fastened Joints

5. Strength Analysis of Adhesively Bonded Repairs

6 Adhesively Bonded Joints for Fibrous Composite Structures

7 Design and Analysis ofBo1ted and Riveted Joints in Fibrous

C.H. WANG and L.R.F. ROSE

Due to the differential straining in substrates, adhesively bonded joints inevitably

L. Tong and C. Soutis (eds.).

double lap joint

single lap joint a~ !=._I__A_d_h_e1re""nd•2-•'-----r--,__~

single strap joint a~

1.2 A CRITICAL REVIEW OF STRESS ANALYSIS AND F AlLURE ASSESSMENT

1.2.1 ADHESIVE STRESSES

1.2.2 JOINT FAILURE CRITERIA

Adhesive failure within a joint is usually assumed to be governed by a critical stress, or

- - • Crocombe & Moult

- ~-~- -~~ ~ - - ---~""=-

0.5 Fillet comer S~re comer

Normalised adhesive layer thickness tA/t 0

1.3 ENDS EFFECT

1.3.1 ADHESIVE SHEAR STRESS

'Z' xy ( X, 0) -_ -rmaxe -2yxlt, COS ( 2 y X It A ) (1.11)

Figure 1.3 A strip rigidly clamped along two long edges.

1.2 -------I------- -I------- - 1 - - - - - - - ~-------

0.8 -- -~-- -[----- -- -[- ------ -[--- ---- ~---- ---

0.6 - - - - - -1------- -1------- -1--- - - - -

?-- - ~ Finite element

1.3.2 COR NER SINGULARITY

1.3.2.1 Square edge

Figure 1.5 Finite element mesh for a double-lap joint.

Figure I. 7 Ends of adhesive layer, showing a square edge.

K = [A(V)CJmax + B(v)rmax ]t~ (1.16)

A(v) = 0.836 - 2.23v + 6.29v 2 - 9.64v 3 (1.1 7)

B(v) =3.12 -15.8v + 40.lv 2 - 37.6v 3 (1.18)

0.0001 0.001 0.01 0.1

Distance along the interface ahead of corner x/h

Distance along the interface ahead of corner xlh

1.3.2.2 Spew fillet

Figure 1.10 Configuration of a spew fi Ilet.

Fillet angle e (degrees)

Figure 1.12 Maximum fillet angle 1/> for no corner singularity.

Figure 1.13 Stress contours near spew fillet.

which is also shown in Figure 1.2.

Figure 1.14 Adhesive end free of corner singularity.

1.4 STRESS CONCENTRATION IN AD HE RENDS

-!= r(x) _ 2 i =e-f3tAz -e- 2 ~ cos(2~)

(a) Double-lap joint

(b) ir:gle-lap joint

Figure 1.16 Normalised longitudinal stress in adherend.

As an example consider a double-lap joint representative of a unidirectional boron

-0.2 -0.1 0 0.1 0.2 0.3

1.5 TRIAXIAL STRESSES AND PLASTIC YIELDING

Consequently the hydrostatic stress is

To quantify the influence of hydrostatic stress on the plastic yielding behaviour of

t"' cr(tll) ; cr"' cr(t/l) 2 (2.1)