J O U R N A L O F M A T E R I A L S S C I E N C E 3 9 (2 0 0 4 ) 4729 – 4797

Review
Adhesively-bonded joints in metallic alloys,
polymers and composite materials: Mechanical
and environmental durability performance
A. BALDAN
Department of Metallurgical and Materials Engineering, University of Mersin, Ciftlikkoy,
Mersin/Turkey
E-mail: [email protected]

The factors affecting the mechanical and environmental durability (or stability), and
performance of the adhesively bonded joints in various adherends including metallic
alloys, polymers and composite materials are studied in detail. The primary function of a
joint is to transfer load from one structural member to another. In most bonded joints the
load transfer takes place through interfacial shear. At present, the use of adhesive bonded
joints are largely applied to secondary non-critical structures. Whereas the use of adhesive
bonding in primary structural applications has been somewhat limited because of the
difficulty in defining and predicting joint strength, and designing the joint geometry to
optimize strength and reliability. The determination of adhesive joint strength is
complicated primarily by the nature of the polymeric material itself. Since these problems
are mainly mechanical in nature, stress analysis is required to understand how the force
loads are distributed along the adherends and adhesive layer. Most structural engineers
consider the durability or stability of a joint to be fatigue related. This is only partly true for
adhesive bonds as most durability issues are driven by environmental resistance rather
than fatigue loads. The environmental resistance of an adhesive bond is determined by the
chemical bonds formed during cure of the adhesive and the resistance of the chemical
bonds to environmental degradation. Environmental resistance is fundamental to the
durability of a bonded joint or repair. Most in-service failures are caused by environmental
degradation of the interface between the bonding surface and the adhesive. Although the
use of adhesive bonding is increasing rapidly, there are still important issues which need to
be addressed in joint analysis, design, durability, and performance considerations.
Therefore, the study of joints usually involves consideration of (a) joint geometries, (b)
materials (i.e., adhesives and adherends), (c) loading conditions (i.e., static and dynamic
loadings), (d) failure modes (i.e., cohesive, adhesive or mixed failure modes), and (e)
temperature and moisture or environmental effects (humidity, solvents, corrosion,
temperature extremes, thermal cyling etc.). Therefore, in the present paper the adhesive
joints are critically assessed in terms of these factors which affect the durability and
performance of them.
There are two basic mathematical approaches for the analysis of adhesively bonded
joints: (a) closed-form or analytical model and (b) numerical solutions (i.e., finite element
analysis, FEA). In the closed-form approach, a set of differential equations and boundary
conditions is formulated. The solutions of these equations are analytical expressions which
give values of stresses at any point of joint. The analytical approach for the solution of
complex stress distributions in the joints has been progressively refined until recent times.
In the second approach, solutions of differential equations are obtained by numerical
methods or the continuum is represented by a discrete model at the outset. The solution of
these equations gives displacements at the determined points from which strains and
stresses can be obtained for any point within the model. Among the numerical methods,
finite element analysis (FEA) has been extensively used with success. The two- and
three-dimensional finite element analyses approaches have been extensively applied by
many workers to analyse the adhesive joints considering the linear and geometric
nonlinearities.
C 2004 Kluwer Academic Publishers

0022–2461

C 2004 Kluwer Academic Publishers 4729
1. Introduction blies often offer improved corrosion resistance through-
Almost every designed structure requires component out their useful lifetime. When used to bond polymers
members to be connected. The most structurally effi- or polymer-matrix composites, the adhesive can be se-
cient method of connecting the structures is with shear lected from the same family of materials to assure good
joints, which are either adhesively bonded or mechani- compatibility [11].
cally fastened [1]. In the design of the adhesive joints for Adhesives are currently used in many areas such as
engineering structures, strength, stiffness and life are automobile and aerospace industries. For example, ad-
considered to be the most important mechanical prop- hesive bonding of metallic structures was welcomed
erties. Under identical fatigue conditions, adhesively when it was first introduced by de Havillands, in Eng-
bonded joints are far superior to mechanically fastened land, on the Hornet and Dove in the middle 1940s (Ref.
joints [2]. Most structural engineers consider the dura- [10]). Since then, adhesive bonding of aircraft struc-
bility or stability of an adhesive joint to be fatigue re- tures has been in use and is still in use on current aircraft
lated. This is only partly true for adhesive bonds as most projects as a direct alternative to riveting [12]. In the
durability issues are driven by environmental resistance manufacture of automobiles the adhesives are almost
rather than fatigue loads [2].The main disadvantages always used as basic sealant materials or in non-critical
of mechanical connections are that they do not dis- secondary structures. In the manufacture of aircrafts the
tribute the load uniformly, thus resulting in large local use of adhesive bonded joints has also largely been lim-
stresses [3]. In comparison, adhesively bonded joints ited to secondary non-critical structures such as aero-
can achieve in excess of 80% of the tensile strength of dynamic fairings and wing panels. Therefore, the use
the weakest adherend even with a simple single-shear of adhesives in truly structural applications has been
configuration [4]. Therefore, adhesively bonded joint is quite limited. The reasons for these limitations are as
the most efficient method for joining both metallic and follows: (a) a concern about the fatigue and durability
non-metallic (i.e., composite-to-metal bonding) struc- behavior of bonded, strutural components over the ex-
tures where strength, stiffness, and fatigue life must be pected lifetime of the vehicle [13], and (b) secondly,
maximized at a minimum weight [5, 6]. For example, the fracture behavior of adhesive bonded joints, partic-
use of tough, moisture resistant, flow controlled, epoxy ularly those with dissimilar adherends (i.e., composite-
based adhesives to bond composites to composites, or to-metal), is still not well understood [5]. Since the
composites to metals offers a number of advantages [7]. adhesive joints must perform satisfactorily under ser-
What “successful bonding” implies is consistent vice conditions, which include static and dynamic load-
strength and long-term reliability [8]. To achieve ings and exposure to hostile environments such as wa-
consistent strength, the bondline thickness must be ter, petrol, other organic solvents, etc and, in many in-
controlled so that repeatable joints with similar perfor- stances, combinations of these conditions, may be ex-
mance characteristics are achieved [9]. Although adhe- perienced. As Jethwa and Kinloch [13] pointed out, it
sively bonded joints have many advantages over other is of prime importance for the adhesives user to be able
structural joining methods, particularly in their effi- to develop and recommend “adhesive systems” (i.e.,
cient load transfer in thin components and structural the substrate/surface pretreatment/adhesive) which will
repairs, the general application of adhesively bonded possess an adequate service life under the operating
joints has suffered due to their difficulty to inspect conditions which are to be experienced by the bonded
bondline quality following manufacture and in-service structure. This, in turn, leads to the need to understand
life [10]. As pointed out by Hart-Smith [10], the ma- the mechanisms of failure and to develop test methods
jor concern is degradation of the bondline mechanical (a) for developing and selecting adhesive systems, (b)
properties due to either poor surface preparation or in- for quality assurance, and (c) for predicting, quantita-
proper curing conditions. Because adhesives bond the tively, the expected service life [13].
entire joint area, good load distribution and fatigue re- The plastic yielding behavior of polymers such as ad-
sistance are obtained and stress concentrations (such hesives is well known to be sensitive to the hydrostatic
as those observed with screws, rivets, and spot welds) pressure (i.e., Refs. [14, 15]), in contrast to metallic ma-
are avoided. Similarly, because of the large amount of terials. To account for this pressure dependence, various
contact area that can usually be obtained, the total joint proposals have been made to modify the conventional
strength compares favorably with that produced by al- yield criteria by including a term that accounts for in-
ternative methods of joining or attachment [11]. Addi- fluences of hydrostatic stresses (i.e., Refs. [14–17]).
tives can be incorporated to enhance (a) strength, (b) However, due to the fact that the adhesive layer in a
increase flexibility, or (c) provide resistance to various bonded joint experiences not only a high stress con-
environments. centration near the end of the adhesive layer but also
Adhesives are generally inexpensive and frequently triaxial stresses [18] induced by the high constraint im-
weigh less than the fasteners needed to produce a parted by stiff adherends, experimental data of adhesive
comparable-strength joint. In addition, the adhesive can yielding are scarce. Consequently much of the work to
also provide [11]: (a) thermal and electrical insulation; date has been mainly concerned with adhesives in their
(b) act as a damper to noise, shock, and vibration; (c) neat from subjected to simple loading, and it is not
stop a propagating crack; and (d) provide protection clear whether existing yield criteria are applicable in
against galvanic corrosion when dissimilar metals are the context of bonded joints. It was recently found [14]
joined. By providing both a joint and a seal against that the conventional yield criteria widely employed to
moisture, gases, and fluids, adhesive-bonded assem- model adhesive, such as the modified Tresca criterion

4730
[15, 16], the modified von Mises criterion [15, 16], and mental resistance rather than fatigue loads. Therefore,
the linear Drucker-Prager criterion [16, 19], are unable the combination of an aqueous environment and me-
to characterize the yiels locus. To overcome this diffi- chanical loading such as cyclic-fatigue is a severe test
culty, Wang and Chalkley [14] have used the modified for any adhesive system. Practical tests have shown
Drucker-Prager/cap plasticity model that is commonly that properly designed and fabricated adhesive bonded
associated with geological materials is adopted in the structures do not exhibit fatigue failure [2].
plasticity of a film adhesive under multiaxial stresses, Work by Hart-Smith [21] has shown that bonded
which provides a good correlation with the experimen- joints may be designed such that the adhesive can sus-
tal data (see Section 8.1 for more detail about this work). tain loads greater than the unloaded strength of the par-
There is a growing trend to optimise the strength, ent material, ensuring that the adhesive will be able to
weight and durability of aircraft and spacecraft struc- sustain all possible load cases for the original struc-
tures by combining tradational metals with polymeric ture. As pointed out by Davis and Bond [2], such adhe-
composites [5]. Composites are more structurally effi- sive bonds should thus never fail, because the structure
cient than metals and do not experience galvanic cor- would have failed away from the joint before the load
rosion. Metals, however, have better damage tolerance necessary to fail the adhesive could be achieved. How-
and failure predictibility than composites and are un- ever, so many adhesive bonds fail in service. The reason
effected by the solvents and temperatures which tend is that they do not fail because of deficient design, nor
to degrade polymers. In order to optimise the benefits do they fail because of poor materials selection. The de-
offered by both types of materials, hybrid composite-to- ficiencies are usually in the processes themselves, not
metal structures are increasingly being developed [5]. necessarily in their implementation [2]. Deficient bond-
An example of these structures is an aircraft engine ing processes have been a significant contributor to the
strut containing a light-weight high strength carbon poor performance of adhesively bonded structures.
fiber/epoxy fairing joined to damage tolerant aluminum The past experimental data have shown that the con-
ribs. Although these structures provide an excellent tact thermooxidation can have the considerable effects
blend of material properties, their success depends upon on the stability performance of the adhesive joints (see
the integrity of the joints which connect them together for example Refs. [22–25]). The cohesive characteris-
[5]. tics of the boundary layers in the adhesive joints are
Although the use of adhesive bonding is increas- affected by the contact thermooxidation reactions of
ing rapidly, there are still important issues which need the polymer catalyzed by iron surface compounds [23,
to be addressed in joint analysis, design, and durabil- 24]. As pointed out by Kalnins and Ozolins [22], oxy-
ity considerations. The study of joints usually involves gen from several sources can take part in the contact
consideration [20] of (a) a variety of joint geometries, thermooxidation, namely, (a) oxygen which diffuses
(b) materials (i.e., adhesives, adherends), (c) loading through the polymer layer, (b) oxygen captured at the
conditions, (d) failure modes, and (e) temperature and interface, and, (c) finally, oxygen which is absorbed
moisture effects (or environmental effects). The analy- or chemisorbed by metal surface oxide. The main fea-
sis of adhesively bonded joints requires a reliable and tures of contact thermooxidation (oxygen uptake, car-
efficient tool (i.e., such as numerical analysis or finite bon dioxide evolution, change in the carbonyl group
element analysis, FEA, technique) to obtain stresses, content, and change in the layer weight) under condi-
strains, and fracture parameters. Most of the available tions of a free access of oxygen through the polymeric
methods require big investments in computer as well as layer are kinetically interrelated with the change in peel
designer time. strength [25]. As shown experimentally by Kalnins and
This paper critically assesses the progress in the Ozolins [22], the dependence of the rate of change of
mechanical and environmental durability and per- individual kinetic parameters of contact thermooxida-
formance of adhesively bonded joints in various tion as well as of peel strength on the thickness of the
technologically important adhesive-adherend systems adhesive layer and on the contact temperature can be de-
(i.e., such as metal-metal, metal-composite, metal- scribed by a simple equation (see Section 9.1 for more
polymer, polymer-polymer, composite-composite joint detail). The dependence of peel strength on the contact
systems). time t can be described by an expression based on the
assumption that the magnitude of peel strength is con-
trolled by two main competing processes [22], namely
1.1. Adhesive bond durability (a) the accumulation of oxygen-containing groups and
A common type of mechanical loading encountered the oxidation cross-linking which increase the peel
by structures such as in aerospace industry, especially strength values, on the one hand, and (b) the reactions
adhesively-bonded components, is cyclic-fatigue load- of oxidative destruction which decrease peel strength,
ing. For most materials, the presence of this type of on the other hand [25].
loading is found to lead to a much lower resistance to Water may affect both the chemical and physical
crack growth than under monotonic loading, and poly- properties of the adhesive and also the nature of the
meric adhesives are no exception to this observation. interface that exists between it and substrate [26].
Most structural engineers consider the durability of a Brockmann [27] studied the change in strength and de-
joint to be fatigue related. However, as pointed out by formability of the adhesive (usually a polymer-based
Davis and Bond [2], this is only partly true for adhesive material) layer under hot and humid environmental con-
bonds as most durability issues are driven by environ- ditions and concluded that, for aluminum adherends,

4731
Figure 1 Effect of moisture (i.e., wet and dry) on mechanical properties Figure 3 Effect of water uptake on joint strength (i.e., Butt joints) [28].
of adhesive joints [28].

as a function of immersion time. The rate of degrada-

the water stability is dependent on the adhesive and sur- tion in the butt joints is faster than in the single lap joints
face treatment. He pointed out [27] that the two most [28]. The silane improves the durability of both joints.
effective ways to increase stability were: (a) to change The single lap joints (grit blasted) shows a strength loss
the surface condition through special surface pretreat- of about 30% during the first 3 weeks, then the strength
ment, and (b) to use corrosion resistant primers. is recovered in the next 3–6 weeks before a further re-
As is generally known and accepted, adhesively duction [28]. For the butt joints, the loss in strength is
bonded joints will suffer a loss in joint strength when more significant in the first 3 weeks of aging.
exposed to high humidity, liquid water and/or high Applications for structural adhesive in a marine envi-
temperatures (see for example Fig. 1). It can be seen ronment will generally involve much thicker adherends
from Fig. 1 that the E-modulus (the slope of the lin- that used in some other industries. The effect of ag-
ear part), yield stress and ultimate stress decrease when ing on adhesively bonded joints in a wet environment
the adhesive becomes wet, while the strain to failure is particularly important for marine applications [29].
increases. In other words, moisture increases the flexi- A major concern is the sensitivity of the adhesive to
bility and ductility of the adhesive. Many studies have the effects of water. For marine applications [30], an
demonstrated this for many years (see for example Refs. important limitation that has been recognized is the
[28–30]. Abdel Wahab et al. [28] have investigated the degenerative effect that moisture may have upon the
degradation in the joint strength due to water uptake strength of an adhesively bonded joint. Such effects
in terms of the change in mechanical properties and are pronounced when the component is also subject to
swelling. Fig. 2 shows the effect of water uptake on conditions of high temperature [26, 31] and stress [32].
the swelling strain (i.e., this strain was measured as an As pointed out by Knox and Cowling [29], in contrast
increase in disc thickness at different intervals during to the technology used in the aerospace industry, the
the aging period). There is an evidence of swelling, development of adhesives for bonding steel in marine
which becomes significant as water uptake increases. applications has been governed not only by consider-
The maximum swelling strain at saturation reaches 3%. ations of the highest possible strength and good dura-
As shown in Fig. 3, the effect of water uptake on the bility performance, but more by the consideration of
strength of butt joint was determined by testing joints simple and economic process techniques. Accelerated
after different intervals of aging in water immersion at environmental testing is a means of estimating service
60◦ C [Note that two surface pretreatments were applied life or of providing data to rank influencing factors in
in this investigation: (a) grit blasting, (b) grit blasting terms of resistance to degradation [33]. The problem
+ silane]. This figure illustrates the strength retention is that such environmental testing does not necessarily
reflect the degradation mechanisms in the service en-
vironment. The generation of truly representative ad-
hesive bond durability data, with the conventional lap
shear joints, requires undesirably long periods of envi-
ronmental exposure. With developments in adhesives
and improvements in bonding technology, even longer
periods of exposure will be required, unless durability
testing can be modified. More severe environments at
higher temperatures and increased loads, whilst reduc-
ing test time, may give unrealistic failure modes which
would not predominate under service conditions [33].
As an answer to this problem, Arrowsmith and Maddi-
son [34] proposed a smaller bond area than standard and
predrilled holes to reduce the diffision path length and
therefore the length of environmental exposure time re-
quired. However, this was at the expense of the initial
Figure 2 Effect of water uptake on swelling strain [28]. strength and sensitivity.
4732
1.2. Basic adhesive joint properties
In a simple system, bonding at an interface is due to ad-
hesion between the adhesive and adherend. One talks
of adhesion when a measurable amount of mechanical
work is necessary to separate two surfaces of different
chemical composion or shape [35]. In order to get good
adhesion between adherend and adhesive the first es-
sential work is to ensure good molecular contact.The
determination of adhesion strength between compact
materials (i.e., such as between adherends and adhe-
sive) is still unsolved problem [36]. Polymeric adhe-
sion may be enhanced by grafting a chemical species
at the interface [37, 38]. In various industrial applica-
tions, multi-layer polymers are used to give better over-
all performance than the individual components alone
[36]. Typically, one polymer may exhibit high mechan-
ical strength whereas a second polymer may have the
properties of impermeability necessary for preventing
solvent diffusion. A combination will therefore give
both mechanical and chemical resistance, provided the
interfacial polymeric adhesion is adequate. Adhesion
mechanisms between contacting polymers are various,
depending on the type of polymers and contact condi-
tions [39]. Macromolecular chain interdiffusion is im-
portant for compatible polymers [40, 41] and chemi- (a)
cal bonding can lead to comparable interfacial strength
[42]. Overall separation energy, or energy of adhesion,
can be very high, considerably greater than expected
from considerations of interfacial bonds, and it is since
the pioneering work of Gent and Petrich [43] that bulk
losses due to polymeric deformation associated with
separation have been recognized. An upper limit for
the intrinsic energy of adhesion is set at ca. 20 Jm−2
(b)
when primary bonds dominate [44] (although it is usu-
ally much less), yet measured separation energy may
often be of the order of kJm−2 [45].
The basic adhesive properties are mainly [35] (a) ad-
hesive strength (debonding process), (b) conformobil-
ity (bonding process), and (c) cohesive strength. As de-
scribed by Creton (35), the three specific test methods
are used to determine these basic adhesive properties:
(a) peel (i.e., tensile transverse) tests (Fig. 4a) to deter-
mine adhesive strength (debonding process), (b) quick
tack tests (Fig. 4b) to evaluate conformability (bonding (c)
process), and (c) a shear test (Fig. 4c) to determine the
Figure 4 Basic tests for adhesive joints: (a) the peel tests for peel angles
cohesive strength of the adhesive. From industrial point
of θ = 90◦ and θ < 90◦ , (b) the quick tack test; Pr : pressure; T :
of view, the main concern is to obtain a value from each temperature, and (c) the shear test.
test for comparative purposes. Adherend tests (gener-
ally consisting of the mechanical seperation of the as-
sembly) are used in order to quantify the adherence
level between an adhesive and a substrate. It is now where wu and wl are the joint transfer deflections for
well known that the measured adherence or adhesion upper and lower adherends in overlap, respectively; u u
energy is a complex function of the adhesion (interfa- and u l are the longitudinal displacements for upper and
cial energy) and of the energy dissipated in the material lower adherends, respectively; ηa is the adhesive layer
during viscoelastic and deformation processes. thickness; E a and G a are the Young’s modulus and shear
The adhesive stresses, peel stress, σa , and shear stress, modulus of adhesive material, respectively. Due to the
τa , are useful basic parameters which can be used in the very thin adhesive layer, it is assumed that there is mini-
joint analysis. These expressions are given as [46] mal variation in the adhesive stresses along the adhesive
layer thickness.
wu − wl
σa = E a As illustrated in Fig. 4a, the peel test, which simu-
ηa lates the removal of an adhesive tape, is the standard
(1)
uu − ul indicator of the fracture energy of the adhesive bond
τa = G a
ηa (i.e., the debonding process) and is mainly sensitive to
4733
the viscoelastic losses in the adhesive [35]. A thin layer
of adhesive is applied to a backing tape which is usually
much stronger than the adhesive. The tape is then care-
fully bonded to an adherend (stainless steel polished to
2-µm roughness) and, finally, it is peeled in a manner
shown in Fig. 4a. The peel force per unit width of tape
measures directly the energy per unit area to separate
the adhesive bond. Unfortunately, peel test results are
also sensitive to the peel angle θ at which the tape is
removed and to the stiffness of the backing tape, mak-
ing it difficult to directly compare results from different
manufacturers.
The nominal peeling energy or energy release rate,
G p , for a crack propagating at or near to the interface,
can be calculated from the peel force, F, using the ex-
pression [47]

P(r − cos θ)
Gp = − W ηa (2)
wa

where P is the peel force, r is the extension ratio of the

peeled leg (i.e., r = 1 + e), where e is the maximum
strain in the peeled leg), θ is the peel angle, W is the
strain energy density in the leg, wa is the width of the
rubber (or adhesive) and ηa is the thickness of the rubber
in the unstrained state (see Fig. 4a). When extension in Figure 6 Peel rate versus peeling energy relationship from constant
the peeled leg is ignored, r → 1 and W → 0, and rate peel tests [48]: (
): Nylon, (•): Steel, (◦): Aluminum, and ( ):
peeling is carried out at 90◦ , Equation 2 simplifies [36, Ra ∼ 1 µm. Note that the bars indicate the minimum and maximum
48] to peeling energies calculated from the minimum and maximum peeling
force, respectively, recorded on a typical peeling force versus cross-head
separation trace shown in Fig. 5.
P
Gp = (3)
wa

With this geometry, peel speed, vp , is equivalent to ma- stopped or leg break occurred, were ignored (see Fig. 5)
chine crosshead speed, since it was checked that exten- . The remaining peaks on the trace were utilized for cal-
sion of the peel arm after seperation and under force P culating an average peeling force for the test piece (see
was negligible [36]. For example, Ansarifar et al. [48] Fig. 5). After the peel experiments are completed, force
have performed peel tests at an angle of 90◦ at 23◦ C values are placed in Equation 3 to calculate peeling en-
on the rubber to nylon 6,6 joint, and the results for the ergies G p for the test pieces. Fig. 6 shows the peeling
peel force P against the cross-head speed are shown rate variation as a function of the peeling energy deter-
in Fig. 5. In Fig. 5, the first peaks corresponding to mined using the above method.
the onset of crack growth, where the peeling force was The most widely used experimental parameters in
still rising, and the last peak corresponding to when test peel test are (a) the peeling rate, (b) temperature, and (c)
thickness of the adhesive. Fig. 7a shows the peel energy
of an SBR adhesive on a Mylar substrate as a function
of the peeling rate and at different temperatures [43]. It
is clear from the data that, similarly to the viscoelastic
properties, the peel energy data can be superimposed on
a single master curve using the Williams-Landel-Ferry
(WLF) time-temperature superposition [19], as shown
in Fig. 7b. Such a result is typical from peel tests of
pressure sensitive adhesive tapes and implies that the
measured energy of adhesion is quite sensitive to the
viscoelastic losses in the polymer used for the adhesive
layer. Another important experimental variable in peel
test is the thickness of the adhesive layer. Aubrey et al.
[50] found that for a poly (n-butyl acrylate) (PBA) the
90◦ peel strength (i.e., tensile transverse strength) was
independent of the adhesive thickness (in the 15–85 µm
Figure 5 Typical peel force versus cross-head separation trace for a con-
range) in the low peel rate regime, but that the rate at
stant rate peel test; trace produced at 60 mm/min cross-head speed after which they observed a transition from cohesive to fib-
peeling rubber from nylon 6,6 (Ref. [48]). Where P is the peel force, P3 rillar fracture to adhesive nonfibrillar fracture increased
and P1 are the maximum and minimum peeling force, respectively. with thickness.
4734
 (a)

Figure 9 Master curve of (temperature-corrected) peel energy, bT G p

versus (temperature-corrected) peel speed, log (aT vp ) (Ref. [36]).

the range of 0.1–100 mm min−1 . Note that the peel en-

ergy G p was calculated using the Equation 3. They have
constracted a master curve to combine these variables in
a single equation , in a manner of the classic Williams-
Landel-Ferry (WLF) treatment [19] developed for elas-
tomers above their glass transition. Therefore, they at-
tempted to combine the results of Fig. 8 by empirically
(b) using a time-temperature transition factor, aT :
Figure 7 Peel fracture energy against rate of peel for SBR adhesive on
a Mylar [43]. (a) Fracture energy versus the peel rate vp and (b) Master log aT = 0.11(To − T ) (4)
relation for peel fracture energy against rate of peel vp reduced 23◦ C.
where To is equal to 21◦ C, the reference temperature. T
is a temperature. A small, vertical shift correction was
Guiu and Shanahan [36] have investigated empir-
also applied to take into account effects due to thermal
ically and theoretically the effects of time, tempera-
expansion [19, 43, 51]. This thermal expansion correc-
ture and peel rate on the peel energy using two semi-
tion, bT , is equal to (To + 273) (T + 273) and is, con-
empirical polymers (i.e., EVOH and PE*) at various
sequently, quite small. Fig. 9 presents the master curve
peel angles. Fig. 8 summarises peel test results at tem-
of bT G p versus log (aT vp ) obtained for the tempera-
peratures in the range of 0–70◦ C and for peel speeds in
ture range of 0–70◦ C. Here G p is the overall energy,
and vp the peel speed. With a reference temperature
of To = 21◦ C, it can be seen that the master curve for
bT G p versus log (aT vp ) is most acceptable and shows an
approximately linear increase in the lower speed range,
with a gradient of ca. 0.06, consistent with the bulk
behavior of polyethyelene [52, 53].
Guiu and Shanahan [36] have also investigated this
problem theoretically. Polymeric relaxation phenom-
ena are associated with molecular motion resulting
from thermal agitation. It may be considered that the
shift factor aT represents the difference between the
characteristic relaxation time, t, at a given temperature,
T , and the equivalent relaxation time, to , at a reference
temperature To , such that t = aT to . For semi-crystalline
polymers, an Arrhenius type relation may be applied on
a dependence of t as a function of temperature for sec-
ondary transition (local macromolecular motion) [54,
55], such that


Figure 8 Peel energy, G p , versus peel rate, vp , at different temperature. Q
Abscissa being on a logarithmic scale [36].
t = to exp (5)
Rg T
4735
where Q is an activation energy, Rg is the ideal gas force of separation is taken as the measure of adhesion.
constant and T is (now) in Kelvin. Consequently, Gui Alternatively, a loop tack test has also been used where
and Shanahan [36] constructed following relation for the adhesive is put on a tape and the force necessary to
aT : unstick the loop of tape from the substrate is measured

 [35].
t Q 1 1 Among the tests probing the energy, the most widely
log aT = log = − (6)
to 2.3Rg T To used are the rolling ball test or the rolling cylinder
test where a standardized steel ball is allowed to roll
Provided the temperature difference (To − T ) is not too from an inclined plane on a sticky surface and the
great, with respect to To , Equation 6 may be simplified distance it rolls before stopping, or alternatively the
to largest ball that will not roll on the inclined plane,
is a qualitative indication of the energy involved in
Q bonding and debonding the adhesive from the adherend
log aT ≈ (To − T ) (7)
2.3Rg To2 [35].
Finally, in order to measure the resistance to shear,
which is precisely of the form of Equation 4. a shear test method is used (Fig. 4c). This test is par-
Fig. 10 shows log(aT ) versus T , both with empir- ticularly crucial for all permenant and semi-structural
ically obtained values and two lines. The dotted line applications where a large deformation with time un-
corresponds to the empirical fit, as given by Equation 4 der moderate stress is detrimental [35]. The adhesive
and the full curve represents Equation 6 after applica- is tested either in static shear, where a fixed load is ap-
tion of non-linear regression analysis. It can be seen plied and the time to failure of the adhesive bond is
that, although the empirical fit seems slightly better, recorded, or in dynamic shear, where a constant shear
application of Equation 6 gives acceptable agreement. rate is imposed and the force is monitored. This is ba-
A value for Q of 199 ± 5 kJmol−1 was obtained, which sically a creep test and will give an indication of the
is close to that given in the literature for polyethe- cohesive strength of the adhesive after a long time.
lene: 220 kJmol−1 [54]. Therefore, Guiu and Shanahan The important experimental parameters in shear test are
[36] suggested that the time and temperature effects the load and the temperature for the static shear, along
observed during peel are principally governed by (a) with the shear rate and the temperature in the dynamic
process(es) of bulk energy dissipation in the polyethe- test.
lene. The peel strength of the adhesively bonded joints are
As shown in Fig. 4b, the quick tack experiments affected by the contact thermooxidation [22, 25]. In or-
probe the instant adhesion and are an indication of the der to study the effects of contact thermooxidation on
ability of the material to form a contact very quickly the peel strength of adhesive joints, Kaelble [57] devel-
(i.e., conformability) to the substrate. The tests that oped a model of the peel mechanics for an elastic flex-
have been designed to probe this property can be di- ible adherend and a visoelastic adhesive layer bonded
vided into two categories [35]: (a) those measuring the to the rigid adherend (see Fig. 11a). It was shown that
force of separation, and (b) those measuring the en- the peel strength (i.e., tensile transverse strength) Ap
ergy. The main test probing the force is that of probe is proportional to the adhesive layer thickness ηa . This
tack [56], where a flat cylinder is brought into contact
with the adhesive film and subsequently removed at a
fixed crosshead speed. In this geometry, the maximum

(a)

(b)

Figure 11 Scheme of an adhesive joint [22, 57]: (a) an elastic flexible

Figure 10 Shift factor aT used to obtain the master curve at 21◦ C in adhesive [1] and viscoelastic adhesive layer [3] bonded to the rigid ad-
Fig. 9 (Ref. [36]): (♦): experimental, (------): 0.11 (To − T ), and (--------): herend [2] and (b) a flexible adhesive [4] bonded to steel [5] through the
Q/2.3Rg (1/T − 1/To ). boundary layer [6].

4736
proportionality was confirmed experimentally and was 2. Adhesively bonded joints
established to be valid even for comparatively thick 2.1. Types of adhesive joints
(up to 25 µm) adhesive layers [58]. In accordance with Several simple (and standard) adhesively bonded joints
Kaelble’s model, the peel strength can be expressed as used in industry are shown [46, 60–63] in Fig. 12: (a)
Ap = ηa Z , where Z is a complex term which takes into single-lap joint, (b) double-lap joint, (c) scarf joint, (d)
account the strength deformation charcteristics of the single-lap joint with spew [61], (e) single-lap joint with
system (i.e., tensile and shear moduli of the adhesive, chamfer [61], (f) filleted single-lap joint [46, 61], (g)
boundary cleavage, and boudary shear stresses) and the recessed joint [62], and (h) tapered joint [63]. The spew
peel angle θ . In the case of T-peel tests, cleavage stresses can be considered as the result of the adhesive squeezed
predominates [59] and Z = σo3 /2E a , where σo is the out of the lap region at the moment of the joint manufac-
boundry cleavage stress and E a is the tensile modulus ture. A recessed (or a gap) joint is obtained by removing
of the adhesive. portions of the adhesive from the interior of the overlap.
Taking into account the origin of the gradient of The single-lap shear specimen is the most convenient
macromolecules changes in the polymer boundary one for testing quality of adhesive bond and that appli-
layer near the interface with metal and, as a result, the cation of adhesive on both rough sides of adherends is
existence of the respective gradient of the strength— proper way leading to the highest shear strength [64].
deformtion characteristics of the polymer, Kalnins and In the development of bonded joints for structures, a
Ozolins [22] have adopted the Kaelble approach in their simple joint can be fabricated first and tested for its
analysis of contact thermooxidation. They considered suitability in structures. The size of the joint can be
the residual boundary layer with thickness η as peculiar first estimated from a knowledge of the part sizes to be
“adhesive” which joins the metal with the either part of joined, allotted space for the joint, and a general idea
the polymer layer (see scheme in Fig. 11b): of how much overlap is required to carry the load. With
such knowledge, preliminary joint designs can be made
η σB2 that can be refined using an iterative analysis procedure.
Ap = (8) As pointed out by Maheri and Adams [65], while
2E a
the tensile lap-shear specimens are commonly used for
where σB is the tensile strength and E a is the tensile quality control, it is almost impossible to determine
modulus of the polymer in the boundry layer. They the adhesive properties from such tests owing to the
pointed out [22] that this is rather a foggy speculation. complex stress and strain distributions inside the joints.
The most serious thing is that the residual layer is not Therefore, it is necessary to know the the mechanical
uniform. According to the SEM data they have taken properties, particularly the stress-strain curve and mod-
the adhesive thickness η as an average value. In fact, uli in order to determine the stresses and strains in ad-
the peel strength Ap (the peel force P versus the width hesive joints in variety of configurations [66–70].
of the peeled off strip wa ) is the energy per unit rea of As stated above, the simplest and most popular so-
the debonded interface (see Fig. 11b). lution for adhesively bonding two sheets is the single-
lap joint. However, from the structural view point, the
P x single-lap joint has two harmful features [61]: (a) the
Ap = (9) offset of the two sheets causes a bending action in
wa x the joint, adding additional stress components; (b) the
stress distribution in the lap is not constant and dis-
where x is the elementary displacement of a delami- plays peaks at its ends. After the simplified solution of
nated front line, Px is the elementary work of frac- Volkersen [71] (developed for riveted joints but appli-
ture, and wa x is the elementry delaminated area. To cable also to adhesive bonds)and the more complete
exclude the dependence of the peel strength on the study of Goland and Reissner [72] (which accounted
thickness η, Kalnins and Ozolins [22] calculated the for the peel stress) much effort has been spent over six
peel energy characteristic as AV = Ap /η. In accor- decades to determine the stress field of the single-lap
dance with Equations 8 and 9, they developed the fol- joint and to obtain the “optimal” design of the joint.
lowing expression: In structural adhesive joints, the T-joint is one of typ-
ical joint geometry which is used to bond two plates
P x σ2 at a right angle (i.e., 90◦ ) ( see Fig. 4a) or some other
AV = = B (10)
wa x η 2E a angle (see Fig. 4a). The design and analysis of such
joints depend on the joint geometry, and also on the
where wa x η is the elementary volume of the adhe- relative stiffness of adherends and the adhesive [73].
sive. Consequently, AV is the peel energy per unit vol- Generally, the joints are subjected to load in the plane
ume of a polymer layer involved in the deformation fol- of the adherend or transverse to it. In general, poor re-
lowed by fracture, which depends only on the strength- sults are obtained when the joint is stressed such that
deformation characteristics of the adhesive (σB , E a ). (It tensile transverse (peel) stresses occur, as in the case of
should be mentioned that the ratio σB2 /2E a is equal to side loading [74]. Compressive transverse stresses do
the unit work of fracture of the elastic solid). Kalnins not generally cause failure. Compressive forces in the
and Ozolins [22] have used Equation 10 in their study direction of the sheet will only cause problems when the
of contact thermooxidation, which will be presented in sheet buckles [70]. T-joints have not been extensively
Secion 9.1. analyzed although they are used in many applications

4737
Figure 12 Types of adhesive joints used in industry: (a) single-lap joint, (where L is the joint overlap length), (b) double-lap joint, (c) scarf joint
(where L is the scarf length, and α is the scarf angle); (d) single-lap joint with spew (where α is ther spew angle), [61], (e) single-lap joint with
chamfer, (where α is the chamfer angle, η1 and η2 are the thicknesses of the adherends) [61], (f) filleted single-lap joint [46, 61], (g) recessed joint
(where total lap length = C + D+ recess; efeective lap length = C + D) [62], and (h) tapered lap joint, (α is the tapered angle, L 1 and L 2 are tapered
lengths) [63].

4738
such as in the automotive industry [75]. Adams et al. thick adherend tests (i.e., specified by ASTM D3983-
[70] showed variations of T-joints which may be en- 93) (see for example Fig. 13a). This analysis indicates
countered in practice. Shenoi and Violette [76] used that, within certain limit of overlap length and adherend
adhesively bonded composite T-joints in small boats thickness, bonded joints can be designed such that the
and investigated the influence of T-joint geometry on load capacity of the bond is greater than the unnotched
the ability to transfer out-of-plane loads. strength of the parent material. This specification sim-
The adhesive joints are very complex in nature since plifies joint design procedures in great extent [6]. As
they have different geometrical shapes. Therefore, there summarized in their recent paper by Davis and Bond
are still problems to be solved using the different meth- [2], an adhesively bonded joint may be designed by pre-
ods such as shape optimisation, numerical solutions and dicting (a) the load capacity P, and (b) the joint overlap
finite element analysis methods. length using the Hart-Smith approach [21]. Assuming
that peel effects and thermal stresses are ignored [2],
the load capacity P (the load which the adhesive can
sustain) of an elastic-plastic adhesive in a single over-
2.2. Mechanical durability design principles lap bonded joint is given by Hart-Smith [21] the lesser
In spite of the increased usage of adhesive bonding in value of:
structural applications, the determination of adhesion
strength and durability remains largely empirical in na- 


ture. This can result in over conservative design leading 1 E s(l) ηs(u)
P = 2ηa τa(p) γa(e) + γa(p) E s(l) ηs(u) 1 +
to structural redundancy or, conversely, to failure of ad- 2 E s(u) ηa
hesively bonded joints, typically due to the combined
effects of complex modes of loading and environmental
degradation. or
Many early bonded joint designs were based on the



joint strengths and optimal overlap lengths determined
1 E s(u) ηa
from lap-shear tests. The incorrect assumption inherent P = 2ητa(p) γa(e) + γa(p) E s(u) ηa 1 +
in that approach is that the shear stresses in a bonded 2 E s(l) ηs(u)
joint are uniform, so that twice the overlap length pro- (11)
vides the joint with twice the load capacity [2]. How-
ever, early work by Goland and Reissner [72] showed
where E s(u) and E s(l) are the elastic modulus of the
that the adhesive shear stresses are non-uniform, with
upper and lower adherends, γa(e) , γa(p) are the elastic and
high shear stresses at the ends of the joint and a low
plastic shear strains of the adhesive, ηs(u) and ηs(l) are the
shear stress in the center of the joint. As Davis and
thicknesses of the upper and lover adherends, τa(p) is the
Bond [2] have underlined, the consequence of this is
plastic shear strength of the adhesive, ηa is the adhesive
that, for moderate to long overlap lengths, doubling the
thickness. Note that variables in Equation 11 are defined
overlap length of a joint simply adds to the low stress
in Fig. 13a and b as well. Similarly above equations
area in the joint and contributes nothing to the load
describe the load capacity for a double overlap joint. If
capacity of the joint.
the load capacity calculated in Equation 11 is greater
In many circumstances, the adhesive layer is de-
than the strength of the parent material, the adhesive
signed to carry a high level of stress, for example, in
will never fail, provided an adequate joint overlap is
bonded composite repairs to thick-section primary air-
designed.
craft structure [77] and hence is prone to yielding un-
The joint overlap requirements can be determined
der in-service loads. One recent development in bonded
simply by designing the overlap length (see Fig. 13c)
composite repair technology has been in the area of re-
such that the adhesive could sustain the parent material
pairs to curved surfaces, with a view to reducing the crit-
unnotched ultimate strength by plastic behavior within
ical stress levels. Bonded composite repairs to curverd
the adhesive. This is achieved by determining the re-
surfaces induce through-thickness stresses as well as
quired plastic zone overlap length lp from:
shear stresses in the adhesive [14]. Consequently, a
valid yield criterion is of considerable importance to
σsu ηs(u)
the design and analysis of bonded reinforcements. Due lp = for a single overlap joint (12)
to the constraint imparted by the stiff adherends, the 2τa(p)
adhesive layer in the bonded joint experiences a high
degree of constraint [78]. On the one hand, the resulting where σsu is the ultimate tensile strength of the ad-
triaxial constraint induces hydrostatic stresses that pro- herend. Additionally, generous allowance must be
mote crazing and void growth, leading to brittle failure. made for an elastic through to provide the joint with
On the other, the high degree of constraint implies that some measure of creep resistance [2]. In a bonded joint,
the adhesive layer is more resilient to shear deformation the elastic transfer length L/  is the distance neces-
than in the neat form [79]. sary for the elastic shear stresses to decay to near zero.
Hart-Smith [21] extended the Goland and Reissner A factor of three is applied to that elastic zone length to
analysis [72] of bonded joints to allow for plastic behav- provide the necessary allowance for creep resistance:
ior in the adhesive. Hart-Smith approach assumes that
the adhesive behaves ideally elastic-perfectly plastic 3
(see Fig. 13b). The data for design were obtained from le = (13)

4739
Figure 13 Mechanical design for an adhesive joint [2, 21]: (a) A thick adherend single-overlap joint (i.e., specified by ASTM D3983-93). (b) Shear
stress-shear strain curve for an ideally elastic-perfectly plastic adhesive. (c) Methodolgy for determination of overlap length for a bonded joint such
that the load capacity will be greater than the unnotched strength of the parent material. (d) Methodology for determination of overlap length for a
bonded joint such that the load capacity will be greater than the unnotched strength of the parent material.

where to the other) is the total of the elastic and plastic zone

 sizes (le + lp ). The overlap length (the required min-
Ga 1 1 imum distance between the edge of the joint to any
2 = +
ηa E s(l) ηs(u) E s(u) ηa discontinuity) is twice the transfer length. For a patch
or splice for a butt joint, the minimum patch or splice
where G a is the adhesive shear modulus. As shown in length must be twice the overlap length.
Fig. 13d, the required transfer length (the length neces- Joints designed using this approach will always be
sary to allow complete load transfer from one adherend stronger than the parent material [2]. The ability to

4740
design bonded joints such that the adhesive is never the moisture diffusion through adhesive is predominant
the weakest part of the joint has significant design im- if there was no pre-existing micro-cracks and debonded
plications. If the adhesive will not fail due to loading, areas at the interface although this is still an area of
then there is no requirement to perform detailed de- some debate. Although some studies have shown that
sign of the joint for the various structural load cases water diffusion in polymers is non-Fickian [87, 88],
usually required to demonstrate structural integrity, be- most researchers assume Fickian diffusion model, for
cause there is no load case greater than material ultimate simplicity, in order to evaluate the rate of diffusion in
for the parent material. All of the above design princi- adhesives [82, 89].
ples rely implicitly on the integrity of the bond being The major disadvantages of adhesive bonding [11,
sustained in service, i.e., the bond is durable. 28, 29, 48, 80, 81, 90] are: (a) There is no universal ad-
hesive. Selection of the proper adhesive is often compli-
cated by the wide variety of available options; (b) Most
industrial adhesives are not stable above 180◦ C. Oxida-
2.3. Advantages and disadvantages tion reactions are accelerated, thermoplastics can soften
of adhesive bonded joints and melt, and thermosets decompose. While some ad-
Adhesive bonding offers many advantages over the hesives can be used up to 260◦ C, elevated temperatures
classical fastening techniques such as welding, rivet- are usually a cause for concern; (c) High-strength ad-
ing and mechanical fastening. It has a high resistance hesives are often brittle (poor impact properties). The
to fatigue and as a concequence the life-cycle mainte- toughness of an adhesive joint may decrease consider-
nance costs are significantly reduced [28]. The substan- ably under impact loading conditions. Resilient ones
tial reduction in weight that can be achieved by the use often creep. Some become brittle when exposed to
of adhesive bonding is an important advantage, espe- low temperatures; (d) Long-term durability and life ex-
cially for lightweight structures. In joining lightweight pectancy are difficult to predict; (e) Surface preparation
composites, adhesive bonding is the most appropriate and cleanliness, adhesive preparation, and curing can
joining technique. Almost any material or combination be critical if good and consistent results are to be ob-
of materials can be joined in a wide variety of sizes, tained. Some adhesives are quite sensitive to the pres-
shapes, and thicknesses [11]. For most adhesives, the ence of grease, oil, or moisture on the adherend surfaces
curing temperatures are quite low, seldom exceeding to be joined. Surface roughness and wetting charac-
180◦ C. A substantial number cure at room temperature, teristics must be controlled; (f) The joint can not be
or slightly above and can provide adequate strength for disassembled readily. Assembly times may be greater
many applications. As a result, very thin or delicate than for alternative methods, depending on the curing
materials such as foils can be joined to each other or to mechanism. Elevated temperatures may be required,
heavier sections. Heat-sensitive materials can be joined as well as specialized fixtures; (g) The joint design is
without damage and heat-affected zones are not present thickness-limited; (h) Only shear loading is acceptable;
in the product. When joining dissimilar materials, the (i) It is difficult to determine the quality of an adhesive-
adhesive provides a bond that can tolerate the stresses bonded joint by tradational nondestructive techniques,
of differential expansion and contraction [11]. although some inspection methods have been devel-
Therefore, the advantages of using adhesively oped that give good results for certain types of joints;
bonded joining methods over conventional mechani- (j) Many structural adhesives deteriorate under certain
cally fastened joints can be summarized as [11, 28, operating conditions. The adhesive can be subjected to
80, 81]: (a) few parts in the joint. Adhesive bonding environmental effects. Environments that may be par-
assembly can simplify the assembly process, increase ticularly hostile include ultraviolet light, ozone, acid
production and quality and reduce production cost.(b) rain (low pH), moisture, and salt. Thus durability and re-
full load transfer can readily be achieved, (c) the joint liability of a joint over an extended service lifetime may
is fatigue resilient, (d) the method of construction also be questioned. The determination of adhesion strength
seals the joint, (e) a stiffer connection is produced, (f) and durability remains largely empirical in nature; (k)
the connection is light weight, (g) a smooth contour re- Some adhesives contain objectionable chemicals or sol-
sults, (h) the action of the adhesive provides corrosion vents, or produce them upon curing; and (l) The thermal
resistance between the adherends, and (i) no open hole residual stresses can be induced.
stress concentrations are created.
Despite these advantages, the use of adhesive bond-
ing is not yet widespread. As pointed out by Abdel
Wahab et al. [28], this is mainly due to its low durability 2.4. Stress analysis of adhesive joints
when the structure is exposed to hostile environmental The use of adhesive bonding is increasing rapidly, but
conditions. The effect of moisture on the strength of there are still important issues which need to be ad-
adhesively bonded joints is significant due to the de- dressed in joint analysis, design, and durability con-
terioration of the adhesive layer and the interface. The sideration. The durability study of joints involves con-
strength of a joint is decreased dramatically when aged siderations [91] of (a) a variety of joint geometries,
in hot/wet environments. This has been demonstrated (b) materials (i.e., adherends and adhesives, primers),
in metal/epoxy joints [82, 83] and composite bonded (c) loading conditions [i.e., static loadings (pure load-
joints [10, 84, 85]. In a metallic joint, moisture diffuses ings such as shear, tension, and compression or com-
into the joint through the adhesive layer and wicking plex loading such as bending) or dynamic (or im-
through the interface. It was reported [86] that in general pact) loadings], (d) failure modes (i.e., adhesive failure,
4741
cohesive failure, mixed failure), and (e) environmen- a large set of data in order to obtain reasonably accu-
tal effects (i.e., temperature and moisture effects). The rate results. Although the convenience of modern finite
analysis of adhesively bonded joints requires a reli- element programs has significantly improved analysis
able and efficient tool to obtain stresses, strains, and efficiency, a large investment in engineering time and
fracture parameters. Most of the available methods re- computer resources is still required for many practi-
quire big investments in computer as well as designer cal problems. Finite element modelling can be a useful
time. tool in the understanding of the behavior of adhesive
The increased application of adhesive joints was ac- joints, especially when it is combined with an extensive
companied by the development of mathematical mod- experimental programme.
els to analyze the behavior of those joints. There are As mentioned previously, the single-lap joint has
two basic approaches for the analysis of adhesively been used extensively in many studies due to its sim-
bonded joints [i.e., 91, 92]: (a) closed-form or ana- plicity and practical application. One of the first finite
lytical model, and (b) numerical solutions (i.e. finite element analyses of a single-lap joint was published
element analysis, FEA). Analytical models such as the by Wooley and Carver [98]. They used plane stress
classical work by Goland and Reissner [72] usually in- elements, performed linear analysis, and modeled the
volve some simplifying assumptions. However, some adhesive layer with two rows of elements, thus obtain-
of these techniques yield closed-form solutions which ing the stress variation across the adhesive thickness.
generally involve some simplifying assumptions. All Many other two-dimensional (2-D) models of single-
are limited to a certain range of geometrical or load- lap joints followed this work. Similar linear analyses
ing conditions. In the closed-form approach, a set of were done by Guess et al. [99]. They used a more re-
differential equations and boundary conditions is for- fined mesh, which allowed them to identify significant
mulated. The solutions of these equations are analyt- stress gradients across the adhesive thickness. Harrison
ical expressions which give values of stresses at any and Harrison [100] did a simplified analysis in which
point of joint. Several analytical models have been de- they considered rigid adherends and studied the adhe-
veloped for analysis of adhesively bonded joints. The sive deformations. Adams and Peppiatt [101] used the
analytical approach for the solution of complex stress plane strain triangular elements to model a single-lap
distribtions in the joints has been progressively refined joint and considered the adhesive layer to have either a
until recent times [i.e., 61, 93–97]. In general terms it square edge or a spew fillet.
makes use of the plate theory to model the sheets while The stress analysis of adhesively bonded joints is
some simplified assumption is made for the adhesive complex because the joints rarely see simple or pure
layer behavior. The analysis is bidimensional, since it loading of any type (e.g., shear, tension or compres-
involves a strip of unit width in transverse direction, sion) but, rather, are usually subjected to some bending.
assuming plane strain conditions. Volkersen [93] im- Details stress analysis of balanced joints (for example,
proved his previous work by including the peeling stress such joint given in Fig. 12a) can be found in studies such
and assuming a relationship between shearing strain as by Goland and Reissner [72] and Oplinger [102]. It is
and displacement more correct than that of Goland and therefore useful to give some mathematical expressions
Reissner [72]. Segerlind [94] pointed out that as the in order to better understand the effects of bending load-
lap length increases the stresses are in general reduced, ing on the adhesive joint behavior, as this subject was
but the stress peaks at the lap ends are more marked. recently reviewed by Li and Lee-Sullivan [46] in their
Conversely, in the case of lap length the stresses are ev- joint stress analysis for an idealized geometry of the
erywhere higher but more uniform. Renton and Vinson unfilleted single lap joint using the analytical solution
[95] developed an analytical solution for the case of an approach.
adhesive joint of orthotropc plates. Ojalvo and Eidinoff In general, the bending moment at the end of the
[96] accounted for the variation of the stresses through overlap, Mo , can be related [46] to the applied in-plane
the adhesive thickness and investigated on the adhesive loading, P, and the adherend thickness, ηs , by the bend-
layer thickness effect. Bigwood and Crocombe [97] ing moment factor, k, where
developed an analytical solution capable to describe
not only simple lap joints but also more complicated

ηs
geometries. Mo = P k (14)
In the second approach, solutions of differential 2
equations are obtained by numerical methods or the
continuum is represented by a discrete model (i.e., fi- It is noted that P is the applied tensile force per unit ad-
nite element analysis model) at the outset. A system of herend width. The physical significance of Equation 14
algebraic equations is derived, commonly from energy may be explained as follows: The two opposing forces,
functions. The solution of these equations gives dis- P, applied to the joint are not co-linear, so there will be
placements at the determined points from which strains some bending applied to the joint in addition to the in-
and stresses can be obtained for any point within the plane tension. As the applied tensile force increases, the
model. Numerical methods provide a general tool for overlap correspondingly rotates to reduce the bending
anaysis of arbitrary geometries and loading conditions. moment, i.e., k decreases with increasing ηs (Ref. [46]).
Among the numerical methods, finite element analysis In order to improve the accuracy of the FEM simu-
(FEA) has been extensively used with success; how- lation in the numerical analysis the following longitu-
ever, this kind of analysis requires the generation of dinal stress values at the top/bottom surface points for

4742
  
the bending moment Mo and its factor k at the overlap 3ηa 
8 G a ηs 1 + 4 ηs 6P
end may be used [46]: µ2 = +
E  ηa E  ηs

σbending I η3 σbending η2 σbending P
Mo = = s = s (15a) ξ=
t/2 6ηs 6 D
Mo ηs σbending E  ηs2
k= = (15b) (D =
1
P ηs 3P 12 , individual adherend bending stiffness), and
2
Es
E =
where I = ηs3 /12is the centroidal moment of inertia 1 − νs2
per unit adherend width, σbending is the bending stress
(E s is the adherend Young’s modulus under plane strain
at the ovelap end.
conditions, and νs is the adherend Poisson’s ratio). The
Li and Lee-Sullivan [46] have also reviewed the an-
adhesive stress expressions for Oplinger model can be
alytical solutions for an idealized geometry of the un-
found in Ref. [104].
filleted single-lap joint since most analytical solutions
in the literature are derived on the assumption of an
idelized geometry of the unfilleted single-lap joint. The 3. Defects in adhesively bonded joints:
available expressions for the bending moment factor k Modes of failure
at the overlap end are supplied in Refs. [102–104]. Tsai The determination of adhesive joint strength is compli-
and Morton [105] discussed the discrepancies and con- cated primarily by the nature of the polymeric material
troversies in these theoretical solutions. According to itself, in that the mechanical behavior of the adhesive
Li and Lee-Sullivan [46], among the models reviewed, is sensitive [106] to (a) moisture condition, (b) strain
the k model offered by Oplinger [102] seems to be, rate (i.e., loading conditions; simple loadings such as
by far, the most complete and therefore, they have ap- shear and peel stresses and complex loading such as
plied this model to their recent work [46]. Oplinger’s bending), (c) temperature, and (d) other environmental
[46, 102] expression for the bending moment factor conditions. The second major complication with regard
kop is

 ηa  tanh 1 
µ c  ηa  
tanh(ξ ls ) R3 Ro 1 + ηs + C2 − 8R4 tanh µη2sc C1 1 + ηs + C2
ηs
kop = (16)
tanh
µ1 c
ηs
√  µ1 c 
tanh (ξ ls ) R3 Ro − 8R4 C1 µ2 c
+ 8Ro (1 − C1 ) tanh ηs
tanh ηs

where ls is the outer adherend length, ηa the adhesive to adhesive joint strength prediction is the failure pro-
thickness, ηs the adherend thickness, c the half-overlap cess, and therefore by implication, the failure criterion
section length. to use. Identifying and interpreting the locus of failure
Where and crack propagation behavior are significant aspects
in evaluating the mechanical properties of adhesively
 bonded joints, and have been interests for many years.
12PB 2 While it is believed that materials always fail at the
Ro = , R3 =   ,
G a ηs 3ηa  Ro2 weakest location, the locus of failure, while closely re-
4 1+ 4 ηs + 4
lated to material properties (such as tensile strength),
1 quality of adhesion at the interface, and fracture tough-
R4 = ness of the bonds, depends also on the stress state at
R3
the crack tip. The locus of failure is closely related
where PB is the applied tensile force per unit adherend to the direction of cracking. The allowable loads on
width for the bottom surcace points, G a the shear mod- a joint are the loads at which micromechanical dam-
ulus adhesive material. age (or defects) first occurs that will eventually lead to
macromechanical damage. Thus the micromechanical
damage can be the basis for the selection of ultimate-
9Pηa 1 + ηηas + GPa ηηa2 load-prediction techniques and the prediction of failure
C1 = −     3Pη s
modes of the joints. Defects in adhesively bonded joints
4 1 + 43ηηas G a ηs2 4 1 + 3η
4ηs
a
+ a
G a η2 s can be characterized in various ways such as (a) bond-
ηa Pηs
3 Pηa 1+ ηs + G a ηs2
line defects [i.e., 10, 80, 107, 108], and (b) the modes
C2 =  3ηa  3P ηa
, of micromechanical damage at the joint (i.e., the failure
G a ηs2 4 1+ 4 ηs + G a ηs2 characters) [i.e., 2, 10, 60, 64, 106, 109]. The predic-
 tion of failure in adhesive joints is of great importance
6P
E  ηs in the use of bonding for structural applications. The
µ1 =  
3 ηa  3 Pηa
certification process requires calculations to be made
4 1+ 4 ηs + G a ηs2 regarding the strength of critical bonds in the structure,
4743
as well as experimental tests to prove these calcula- mode, i.e., some cohesive in the primer layer and some
tions.The mode of failure is recognized as the vital is- interfacial between nylon and primer. In each case, the
sue, regardless of the amount of any crack (disbond) primer layer had been cohesively the weakest aspect of
growth. A poor bonding surface would be denoted by the joint. Under constant load, bond failed in a time-
even small areas of adhesion (interfacial) failures, al- dependent manner at or near the interface. The time-
though it usually results in complete interfacial fail- dependent failure of the bond during the peeling test
ures. For example, failures confined within the matrix, was caused by a time dependent cavitation of the rub-
between the adhesive layer and adjacent fibers in com- ber in the highly stressed region of the peel front, where
posites, would normally be regarded as indicative of strands of rubber were observed to form in tension per-
acceptable bond strength and durability, assuming that pendicular to the bond plate. In fact, the eventual rupture
the bond also passed its short-term static strength re- of these strands caused the time-depended failure of the
quirements [10] (which is one way of verifying that bond.
the adhesive was exposed to acceptable thermal profile A common contributor to premature interfacial fail-
during cure). ures is some form of pre-bond moisture [10]. For
Ansarifar et al. [48] have studied the bonding failure example, moisture can be present in an undried com-
of rubber to steel, aluminum and nylon 6,6 substrates by posite laminate cured long before it was bonded, as
means of peel tests. They observed that under constant condensation on the surface of adhesive film not stored
load bond failure occurred in a time-dependent manner or thawed out properly, or absorbed within adhesive
at or near the interface, and is cohesive within the rub- film which had been left out for too long in the lay-up
ber and very close to the covercoat. This is caused by room. Water absorbed by the nylon filaments before the
a time-dependent cavitation of the rubber in the highly original laminates are cured with the peel ply in place
stressed region of the peel front. Slow time-dependent is also identified as a contributing problem [10]. Ad-
bond failure observed in the rubber by Ansarifer et al. verse effects of moisture are alleviated, or sometimes
[48] attributed entirely to ozone attack has been re- totally eliminated, by such good venting that all water
ported in the past [47]. Constant load peel tests cover- is sucked off during the cure instead of being trapped at
ing a range of peeling energy were carried out at dif- the interface between the adhesive and the composite
ferent ozone concentrations ranging from 0.25 (atmo- components.
spheric) to about 20 parts per hundred millions [47]. Xu et al. [111] have recently investigated the effects
This study showed that below a certain energy level, of hot/humid environment on the failure mechanisms of
there was very little growth along the bond and the peel galvanized steel to polypropylene (PP) composite lap
rate was essentially independent of peeling energy, but shear joint specimens under the Cataplasma aging tests
increased significantly as a function of ozone concen- and cyclic moisture/temperature tests, with and with-
tration. However, from the time-dependent rate of crack out primer. After the lap shear testing, they observed
growth against the peeling energy (see Fig. 6) Ansarifar that without primer, the locus of failure of the aged
et al. [48] suggested that failure was mainly due to the joints shifted close to the galvanized steel/adhesive in-
physical rupture of the rubber at the peel front and was terface at the edges of the joints. With primer, the joints
not caused by ozone attack. primarily failed at the primer/galvanized steel inter-
Ansarifar et al. [49] have also studied the bonding face. Using the Environmental Scanning Electron Mi-
rubber to nylon 6,6 substrates with commercial organic croscope (ESEM), the locus of failure was observed
solvent-based or water-based bonding agents and tested to be partially within the adhesive close to the adhe-
by means of peel tests to evaluate the bonding failure. sive/galvanized steel interface and partially at the gal-
Peel tests were performed under constant load at an vanized steel interface in the edges of the joints. Corro-
angle of 90◦ and at ambient temperature (∼23◦ C). Us- sion of zinc layer on the galvanized steel was found in
ing the newly developed constant load peel test-method the interfacial failure region. In the central region of the
[48, 110] (see Section 4.2 for more detail about this test adhesive joints, the failure was still at the adhesive/PP
method), bond failure produced over the wide range of composite interface. Clearly, the adhesion at the adhe-
peel rate. There were two distinct stages of failure dur- sive/PP composite interface degrades more slowly than
ing peeling, irrespective of the rate of peel. They were: at the adhesive/galvanized steel interface under Cata-
(a) stick stage and (b) slip stage. In the stick stage [48], plasma environment. The loss of adhesion in a high
no crack-growth occurred as the rubber adherend was moisture environment was mainly due to corrosion of
extended, and the peel angle θ remained at about 90◦ the galvanized steel.
.Whereas in the slip state [48], the cleavage front ad- Xu et al. [111] have also investigated the effect of
vanced at or near the interface, and at a rate faster than primer on the failure mechanisms. In the moisture en-
the cross-head speed. Moreover, in the slip stage the vironment, adhesion at both the adhesive/PP composite
angle of peel was much smaller [48], e.g., 45◦ , and interface and adhesive/primer interface degrades more
the force applied to rubber adherend varied consider- slowly than at the primer/galvanized steel interface.
ably and continuously as the rubber was peeled. This Therefore, the durability of their joints is mainly de-
test showed that there were two possible failure mech- pendent on the retention of adhesion at the interface
anisms: (a) cohesive, and (b) mixed modes. The primer between the primer and the galvanized steel adherend.
had reacted with the surface of the nylon creating a They observed defects such as cavities and pinholes in
third phase, which in peel tests had been the weakest the primer layers. The primer layer on the galvanized
layer and failed cohesively, or the failure was mixed steel was blistered as a result of Cataplasma aging. The

4744
blisters were intitiated at internal cavities and at the 3.1. Bondline defects
edges. The centers of the blisters were corroded first As a result of discussion presented above, bondline de-
and the disbondment of the zinc/primer interface oc- fects in adhesively bonded joints can be generalized
curred around the corroded area. Intergranular corro- [80, 107, 108] as either a (a) debond or (b) a weak
sion took place on the surface of the galvanized steel. It bond, as seen in Fig. 15 . In the structural integrity of an
was found that oxygen significantly accelerated the pro- adhesively bonded joint there are two main problems.
cess of blistering. In contrast, there were many blisters These are the effect of debonds and weak bondlines on
on the primer coating of the galvanized steel soaked in the load transfer and durability of the joints. Therefore,
an open bottle under same conditions. The increase in the integrity of the bondline in the adhesively bonded
the number of blisters was attributed to easier access of joint is quite an important phenomenon. The bondline is
oxygen. Xu et al. [111] concluded that the connection schematically defined in Fig. 16. In adhesively bonded
between the presence of oxygen and blistering indi- joints a debond is simply characterized and identified
cates that the debonding is dependent on a cathodic re- as a separation between the two adherends. As a re-
action leading to reduction of oxygen. Therefore, they sult two traction-free surfaces are created. The gross
suggested that the blistering is essentially a result of form of a debond is illustrated in Fig. 15a, however,
cathodic debonding of the primer/galvanized steel sur- other traction-free microscopic forms of separation in-
face. A schematic of the blistering process is shown in clude voids, porosity and micr-cracking in the adhesive.
Fig. 14. During in-service operations, debonds are generally

Figure 14 A schematic of blistering process from the edge of a specimen of primer-coated Galvanized steel [111, 112]: (a) corrosion of galvanized
steel at its edge, (b) blistering of the primer, and (c) delamination of the primer.

Figure 15 Generalized bondline defects in adhesively bonded joints [80, 108]: (a) debond and (b) weak bond.

4745
Figure 16 Adhesively bonded joint bondline [80].

associated with moisture penetration at panel edges

and bolt holes, bolt inserts in honeycomb panels, poor
surface preparation, impact and/or local over-heating.
Debonds are typically identified using common NDI
methods, such as ultrasonic inspection and acoustic
emission. Figure 17 Types of failure in adhesively bonded joints [2]: (a) adhesive
Although the composites are manufactured correctly failure and (b) cohesive failure.
according to the short-term quality control coupons
tested, they separate in service. The issue is not one
of structural overloading. As indicated by Knox and 3.2. Modes of micromechanical damage:
Cowling [29], the interface between the adhesive and Failure characters
the composite surfaces simply debonds without the System adhesive-adherend was characterized not only
slightest trace of either material ever having adherend from the point of shear strength of adhesive-bonded
to the other. In some cases, the details have been known joints but also on the basis of the character of their
to separate before manufacturing had been completed. failure. On the basis of the discussion in the previous
What appears to be a common element in many of these sections and the past works, we can distinguish three
problems is contamination introduced by the use of “re- kinds of failure [i.e., 2, 48, 109, 113]: (a) adhesive (or
lease” peel ply without subsequent through abrasion of quasiinterfacial) failure (see Fig. 17a) (rupture of the
the bonding surfaces [29]. adhesive bond, such that the separation appears to oc-
In contrast to a debond, weakly bonded joints show cur along one of the two adhesive-adherend interfaces)
no sign of separation in the bondline. A weakly bonded [2, 64], (b) cohesive failure (see Fig. 17b) (rupture of the
joint is still effectively bonding the two adherends to- adhesive bond, such that the separation appears to be
gether. This is shown in Fig. 15b where there is a weak- within the thickness of the adhesive layer) [2, 64, 109],
ening of either the adhesive or the bondline interface. and (c) mixed failure (crack goes partly between adhe-
Currently available NDI methods can not reliably iden- sive and adherend and partly through adhesive and/or
tify weak bonds due to the absence of traction-free sur- adherend) [64]. For a given bond, transitions between
faces [80]. The identification of weak bonds in joint adhesive and cohesive modes of debonding are often
production as a quality assurance measure is conducted observed when changing the geometry of the assem-
through witness coupon destructive testing [113]. In bly, the loading configuration, or the rate of debonding
comparison with the bulk, the structure of the bound- extension. Correspondingly, the strength of the bond,
ary layer is less organized [114]. This agrees with the expressed in terms of a failure stress or in terms of a
conception of a weak boundary layer offered long ago steady state debonding toughness, can differ widely de-
by Bikerman [115]. As pointed out by Hart-Smith [10], pending on the debond extention mode. It is commonly
the cause of weak bonds should be acknowledged as observed that the transition from a cohesive mode at
processing errors- and they usually would be, except low rate to an adhesive mode at high rate which is com-
for one further problem. As noted above, such weak- monly reported in the literature for rubbery copoly-
ness are invariably undetected by conventional (or even mer adhesives, adhesive tapes, or pressure sensitive
exotic) ultrosonic inspections, suggesting to some that adhesives.
the parts might not be defective after all. Only when The difference between a design decifiency and a
there is zero tolerance of interfacial failures in accep- processing decificiency can often be determined by in-
tance criteria, regardless of any measured strength, can spection of the failure surface [2]. A design deficiency
standard lap-shear testing of coupons be relied upon to is characterised by fracture of the adhesive (cohesion
distinguish between durable and suspect bonds. Again, failure) with residues on both surfaces (Fig. 17b). This
the fact that most of these coupons (and parts) do not is the typical failure surface observed from adhesive
simply fall apart as soon as they have been made has strength tests such as lap-shear or peel tests. The ex-
created the illusion that there is nothing wrong with perimental data surveyed [i.e., 22–24] show that the
any of the bonds. And, even when the majority of such cohesive characteristic of the boundary layer are also
bonds separate in service, the fact that some of them affected by the contact thermooxidation reactions of
do not is often sufficient to convince some people that the polymer catalyzed by iron surface compounds. The
those which did must have been overloaded and that adhesive would also contain minimal voids and obvi-
the failures had nothing to do with the processes used ous defects. Processing deficiencies are usually char-
to make the bonds or the laminates [10]. acterized by interfacial failure (adhesion failure) (see

4746
Fig. 17a) of the bond such that there will be areas where strength, it is not necessarily stable against the effects
the adhesive remains on only one of the adherend sur- of water [118]. However, whether the failure path is at
faces, with the matching surface being free of adhesive, the interface or whether it is in the oxide, in a weak
or by the presence of voids or other bondline defects boundary layer or within the primer (if used) is a mat-
in the adhesive. Although interfacial failures occasion- ter of some controversy [32]. The exact failure path
ally occur due to contamination, the most usual cause probably depends upon the particular joint under ex-
of in-service adhesion failures is poor environmental amination [119]. As a consequence of the difference in
durability due to the inadequacy of the surface prepa- adherend thickness, the stress distribution developed in
ration used when the bond was formed. a relatively thin steel adherend used in the automotive
The cohesive failure within the adhesive layer or in industry and the the stress distribution developed in a
the surface layer of the adherend matrix may occur by thick steel adherend for a marine application may pro-
brittle fracture or by a rubbery tearing, depending on duce apparently different failure processes before and
the type of adhesive used [60]. This results in cracks after environmental aging [29].
perpendicular to the load and causes a reduction in the
load-transferring capability of the joint. Cohesive fail-
ure are analyzed using two general approaches [106]: 4. Quality control tests: Assessment
(a) the continuum mechanics (CM) approach, and (b) of bonding quality
the fracture mechanics (FM) approach. CM assesses A successful test would be one in which there was vir-
the nature and magnitude of the stresses and strains tually no disbond growth at all, and one in which there
within the adhesive joint, often predicted by finite ele- were absolutely no interfacial failures [10]. The most
ment analysis (FEA). The criterion that is used varies important thing to note about durability testing of ad-
between (a) maximum values of stress [103, 116], (b) hesively bonded joints is that the mode of failure is
strain [116] or (c) plastic energy density [103]. How- more important than the failure load. The measured
ever, to reduce FE model complexity, sharp corners are strength might be of significance in comparing differ-
often modelled at bi-material interfaces, e.g., at the ends ent cohesive failure modes, but the presence of any
of the adherends which introduce singularities ( a point adhesion failure at the interface would transcend all
of theoretically infinite stress) (Ref. [106]). Clearly, the other considerations. As Hart-Smith pointed out [10],
maximum value failure criteria become meaningless as the great danger with any bonded durability test is that,
these values are a function of how well the singular field if a load can be measured, there will be some who base
is modelled, i.e., the refinement of the mesh. the accept/reject decision on the magnitude of the load,
Balkova et al. [64] have investigated the failure thereby totally undermining the purpose of the test.
modes of the glass fiber/polyester composites joints
(the adhesives used for joining are all two-part epoxy
based adhesives: EPOXY 371, P11, LETOXIT LX 012, 4.1. Standard test methods
LEPOX UNIVERSAL 11) under the various environ- Assessment of bonding structures are usually made by
mental conditions (i.e., exposed to air at room temper- carrying out routine tensile lap shear (pr EN 2243-1)
ature and 60◦ C, and cool water) using the single- and and peel tests (BS EN 2243-2) for metal to metal and
double- lap shear tests. They have observed mixed fail- BS EN 2243-3 for metal to metal honeycomb) to de-
ure in all the joints studied . The separation appears to termine the strength. In addition, early bond loads of
be at the adhesive-adherend interface and partly within particular assembly are assessed for a long period of
the adhesive. Cohesive failure reached nearly 60% of time by tensile lap shear and pull test pieces based on
the whole bonded area during aging processes at room the component adherend form [12].
and elevated temperatures. The longer time the joints The following standard mechanical tests methods
were stored in water the weaker the bonds become, es- have been used widely for quality control and durability
pecially at the joints stuck with LEPOX UNIVERSAL performance (i.e., mechanical or environmental stabil-
11, and the adhesive failure prevailed. Only in the case ity) of adhesive joints in metallic and composite ma-
of LETOXIT LX 012 the adhesive as well as cohesive terials adherends: (a) single-lap shear test, (b) double-
bond of the adhesive was stronger than cohesive bond lap shear test (c) Thick-adherend shear test (TAST),
of the adherend and that is why this was torn up. Dam- (d) floating-roller peel test, (e) Double-cantilaver beam
aged adherend area increased from 50 to 80% of the test, (f) Wedge-crack test, (g) Floating-roller test, (h)
whole bonded area for joints stored at room and ele- Short-beam shear test.
vated temperature and decreased up to 10% for joints
stored in water. The delamination in the composite was
not observed for any type of adhesive joint. 4.1.1. Thick-adherend shear test (TAST)
Whilst the locus of failure of unaged structural joints Applications for structural adhesive in a marine envi-
usually involves a cohesive-type fracture process in the ronment will generally involve much thicker adherends
adhesive layer where the adhesion at the interface re- than used in some other industries such automobile
mains high, after environmental attack failure often oc- industry. As a consequence of the difference in ad-
curs at or very close to the interface between the ad- herend thickness, the stress distribution developed in
hesive and the metallic adherend [29, 117]. Generally, a relatively thin steel adherend used in the automo-
it is known that whilst the bonding in the boundary tive industry and the stress distribution developed in a
zone between a polymer and metal surface is of high thick adherend for a marine application may produce

4747
 maximum strains and stresses, as pointed out by Maheri
and Adams [65].
On the other hand, although structural adhesive test
specimens such as the lap-shear and TAST specimens
are almost always used for testing under static condi-
tions, the adhesive material can show a degree of vis-
Figure 18 The standard thick adherend shear test (TAST) specimen.
coelastic behavior such that its mechanical properties
under dynamic (or impact) loading can be significantly
different from those under static loading. Since many
apparently different failure processes before and after
structural adhesives are used in transport applications
environmental aging [29]. Therefore, in order to sim-
where the bonded joints are subject to dynamic as well
plify the stress distributions present in the tensile lap
as static loading, the adhesive dynamic properties can
shear specimens, the thick-adherend shear test (TAST)
be of significance [65]. An early work [122] on deter-
specimen (see Fig. 18) has been proposed by Althof
mining the dynamic moduli of thin, constrained adhe-
[120] and Krieger [121]. In these specimens, by us-
sive films showed that dynamic moduli could be slightly
ing a thick adherend and a short overlap, the specimen
higher than those obtained in static tests, the difference
rotation and the adherend compliance under load are
decreasing for more brittle, structural epoxies.
largely eliminated, rendering a state of stress in the ad-
hesive almost entirely dominated by uniform shear [65].
However, since the complience of the adherend under 4.1.2. Double-cantilaver beam (DCB)
load is not totally eliminated [70], this makes measur- and wedge-crack tests
ing the shear modulus difficult, because it necessitates Literature survey and a review of standard test meth-
accurate measurement of small displacements and cor- ods revealed [123] that the double cantilaver beam
recting for adherend deformations over a thin adhesive (DCB) and wedge-crack tests (schematic illustration
layer. Even so, the method is adequate for determining of this test method shown in Fig. 19a are well suited to

(a)

(b)

(c)

Figure 19 Double-cantilaver beam (DCB) and wedge-crack tests: (a) principle of the wedge-opened double cantilaver beam (DCB) test with definition
of the notations [124], (b) Tapered double cantilaver beam (TDCB) test specimen [134], and (c) Boeing wedge crack durability test for bonded metal
structures [10].

4748
evaluating the short- and long-term durability adhesive Unfortunately, there is a lack of standard test pro-
bonds [125–128]. In the DCB test, a bonded sample is cedures for measuring the bond strength of adhesively
pulled apart, at a constant test machine crosshead ve- bonded composite materials. ASTM test methods re-
locity, by fixture (hinges or pinned blocks) at the end searched cover either adhesively bonded metals or in-
of the beams. The specimen is loaded and unloaded terlaminar failures in composites. The specimen used
until the crack has propagated entirely through the for the DCB and wedge crack tests is based on those of
sample. ASTM D3433, ASTM D3762, and ASTM D5528-94a
The wedge-opened double cantilaver beam (DCB) Standard Test Method for Mode I Interlaminar Fracture
test is one of the simplest methods for the application Toughness of Unidirectional Fiber-Reinforced Polymer
of fracture mechanics to the testing of crack extension in Matrix Composites.
adhesive joints or composite laminates [129, 130]. The The man responsible for the first reliable and sim-
simplicity of the test rests on the facts that the debond ple specimen with which to predict in the short term
extension is intrinsically stable and that no load cell whether or not a bond would fail interfacially in service
is needed [124]. The principle of the test is sketched was Bert Bethune, of Boeing (Ref. [10]). His wedge-
in Fig. 19a. Two plates, taken in the present case as crack durability test for bonded metal structures (see
having the same thickness, ηs , are bonded face to face Fig. 19c) is now an integral part of virtually all manu-
with an adhesive layer of thickness ηa . The width, ws , facturing programs. The Boeing wedge test for bonded
of the plates can be larger than the width, wa , of the metal structures can be performed with the same speci-
bonded area. A wedge thickness ηw inserted between men as in DCB test, but an angled wedge is driven into
the plates induces the extension of a decohesion crack the crack opening to stimulate crack growth (see Fig.
up to a distance “a” from the wedge. If the plates remain 19c for schematic illustration). Then, the sample is ob-
purely elastic, the “debonding toughness ” of the bond served (often in an environmental exposure chamber)
is equal [124] to the strain energy release rate G. If we and the crack tip propagation is recorded. As in the case
assume that the assembly is free of residual stresses of DCB test, wedge-crack test is also used to evaluat-
and that the contribution of the adhesive layer to the ing the short-and long-term durability adhesive bonds
strain energy of the system can be neglected, G can be [123]. The reason to develop the Boeing-wedge crack
expressed on the basis of a simple beam bending model test (Fig. 19c) was to apply a sustained peel load in a
as [124] humid environment. The bonded metal coupon would
split apart rapidly whenever the surface preparation was
one of those which had failed in service, but there would
3 l 2 E s ηs3 ws
G= (17) be little or no crack growth for the durable processes
16a 4 wa and materials, and whatever fracture did occur would
be cohesive. As Hart-Smith [10] points out, any sign
where E s is the plates Young’s modulus and the of interfacial failure should be interpretated as having
displacement at the wedge is l = ηw − ηa . failed the test.
Several more refined expressions of G have been A wedge is driven into the end of the bond at room
proposed. temperature, and the coupon and wedge are then in-
The application of fracture mechanics to adhesive serted into a hot-wet environment, to see whether the
joints dates back to the 1960s when Ripling et al. [131] initial disbond will grow under the peel load and, if
and later Mostovoy et al. [132] studied experimental so, in what manner. It could grow cohesively (leaving
methods to determine the plane-strain fracture tough- roughly half the adhesive on each adherend) or interfa-
ness (K Ic ) of bonded metallic joints. This work led to cially (showing no sign of adhesion between the adhe-
an ASTM standard [133] which used a simple shear- sive and the adherend) [10]. A small amount of cohesive
corrected beam theory (SBT) to deduce the values of growth may be permitted. The mechanism of this ac-
adhesive fracture energy G Ic from either adhesively celerated failure, if it occurs, is a sustained peel load in
bonded double cantilaver beam (DCB) (see Fig. 19a) or the presence of moisture. In the wedge –crack test, the
tapered double cantilaver beam (TDCB) (see Fig. 19b) specimens are subjected to elevated temperature and
test specimens [134]. The double-cantilever-beam peel humidity, which is a reliable short-term method to pre-
test coupon suffers from the limitations that each such dict bond integrity of a joint over long periods of time
coupon can be tested only at a time, tying up a com- in service, as detailed in the literature [123, 125, 126,
bination of test machine and environmental chamber 135–139].
for an appreciable time per coupon. As suggested by As noted by Hart-Smith [10], the need for a bonded
Hart-Smith [10], this might explain why a major air- composite durability test is just as great as it was for
craft manufacturer relied on this test to choose between metal bonding 20 years ago. So to devise a durability
different surface treatments by through pre-production test-coupon for composite structures as reliable as the
testing, but has not implemented such a test in wedge-crack coupon has been for metal bonding. When
production. Perhaps the greatest drawback of the used with composite laminates, the wedge-crack spec-
double-cantilever beam coupon for tests like this is that imen must be made from unidirectional (0◦ ) tape plies.
the opportunity to measure and vary the prying force, The reason for using all-0◦ adherends is that transverse
raises the temptation of assessing the quality of the surface fibers (at 90◦ to the length of the specimen) are
bond in terms of the prying force rather than the mode so weak in the interlaminar tension and shear that the
of failure [10]. tip of the delamination would be diverted away from

4749
 past experience with metal-to-metal bonding indicates
that this will not occur if the aluminum strips are prop-
erly anodized and primed. (Using aluminum for this
purpose would even reduce the specimen cost.) [10].
Fig. 20b illustrates the adoptation of floating-roller peel
(a) test for evaluating bonded composite structures [10].

4.1.4. Short-beam shear test

The short-beam shear test specified in ASTM D2344
uses three-point bending where failure occurs in shear
mode parallel to the fibers in the midplane as shown
in Fig. 21a. The maximum shear stress τmax (normally
called the interlaminar shear strength, ILSS) is related
[74] to the applied maximum load Pmax and specimen
dimensions, according to the simple elastic beam the-
ory: τmax = 3Pmax /(4wη), where w and η are width
and thickness of specimen. Because of the simplicity
of the test method and minimum complication in spec-
imen preparation, it has become one of the most widely
used methods to determine the quality of both polymer
and metal matrices composites [140]. Unfortunately
(b) this test has an inherent problem in that meaningful re-
sults can not be obtained for very thin specimens [140].
Figure 20 Floating-roller peel test: (a) typical and ideal floating roller
test [123] and (b) adoptation of floating-roller peel test for evaluating The high stress concentration and damage by crushing
bonded composite Structures [10]. in severe cases at the loading points with a very short
span (normally five times the specimen thickness) may
induce premature failure and complicates the interpre-
the bond surface into the laminate at a far lower peel tation of test results. To avoid this problem four-point
stress than should be withstood by a structurally sound bending of a long beam may be used.
bond [10]. In order to solve a manufacturing problem caused by
the use of released nylon peel plies, the short-beam test
was adopted by Fischer, in Austria [10] (see Fig. 21b
for this adoptation). The short-beam shear coupon was
4.1.3. Floating-roller peel test made by bonding together two thinner pieces of lami-
Fig. 20a shows the schematic illustration of floating- nate, using various surface preparations. One skin was
roller peel test method for typical and ideal cases [123]. precured, and the other co-cured with the adhesive
As this test method is usually used to test for durability
of bonded metal structures, it is mounted on the test ma-
chine in ambient environment and a drop of water is ap-
plied to the tip of the initial disbond. Poor surface prepa-
rations result in an instantaneous switch to interfacial
failures, which spread rapidly [10]. One can not use this
same test for composite adherends, because they will
not bend enough. The initial approach utilized floating
roller tests on woven fiberglass and carbon fiberglass
(a)
samples [123]. The ASTM D3167 test is designed for
a thick adherend bonded to a thin metal adherend that
bends arounds the roller during peeling. Because they
bend at too tight a radius of curvature and fractured be-
fore the bond could be broken (see Fig. 20a), this test
method was abandoned in favor of other bond strength
tests that do not require such extreme strains on the
adherends to the fracture the bond.
The concept of this test might be adapted by trying
to peel off a thinner strip of aluminum alloy, properly
treated and bonded to a thicker all-0◦ carbon-epoxy
laminate [10]. The coupon will bow, because of the
thermal unbalance, but this should not affect the use- (b)
fullness of the test unless the bowing were so great that Figure 21 Short-beam shear test: (a) Schematic specimen geometry and
the specimen could not be installed in a test rig. The test loading configuration for short-beam test (140) and (b) Fisher modifica-
would be invalidated by having the adhesive separate tion of short-beam shear coupon to test durability of composite bonded
from the aluminum instead of from the composite, but structures [10].

4750
layer. Even with short-term testing, the released nylon hesive failure in the elastomer [110]. It appears that
peel-ply surface resulted in a loss of 50% of the strength, cavitation like processes might also occur in the highly
with respect to the equivalent coupon in which the peel- stressed peel front region during bond failure, and it
ply surface had been thoroughly sanded before bonding was argued [110] that if such processes could occur
[10]. The failures to the unabraded nylon peel-ply sur- in a time-dependent manner, this could account for
faces were clearly the result of poor adhesion between the occurrance of time-dependent failure in a strain-
the precured laminate and the adhesive layer, occuring crysllazing elastomer. There were indications [48, 49,
at a stress of 30–40 N/mm2 . The failures were interlam- 110] that rates of crack-growth from different test-piece
inar outside the adhesive layer for both a non-released geometries could be related to the strain energy release
polyester peel ply, without sanding, and the thoroughly rate. This suggested that the relationship between the
sanded nylon peel-ply surface, occuring at a shear stress rate of failure and the strain energy release rate might
of 50–60 N/mm2 . Only short-term tests were run, so the be characteristic, geometry-independent property for a
relative durability of the two successful techniques was particular rubber-metal bond system. Using this new
not established, although it was recognised that this test method, it was observed [48] that the rate of failure
specimen could be used for this purpose. increased rapidly with energy level showing about a
sixth-power dependence, a very strong dependence for
the system studied. As was claimed by Ansarifer and
co-workers [48, 49, 110, 141] this new test method,
4.2. Novel test methods therefore, offers a more consistent and accurate means
Last ten years the new test methods have been devel- of assessing integrity of the bond and measuring its
oped (see for example, Refs. [33, 48, 90, 110, 141, strength than the existing constant rate peel tests.
142]) in order to better quantify the mechanical prop- Knox and Cowling [33] have developed a novel
erties of adhesive joints under static and dynamic load- method, called scrap test method in order to investi-
ings. These new test methods have been developed to gate the influence of surface treatment on durability
study the effects of various factors including time de- performance. The aims of designing this test method
pendance, surface treatment, aging, characterisation of were twofolds: (a) the effect of aging on the adhesion
an interface between two pieces of the same polymer strength of an adhesive at the interface, and (b) distin-
material, and dynamic loading on the durability perfor- guishing amongst the durability performance of various
mance of the adhesive joints. surface pretreatments. In this test, a relatively short pe-
Existing laboratory constant peel rate tests [48, 143– riod of time is used. To investigate the effect of aging
145] evaluate mainly mechanical load endurance of the on the adhesion strength of an adhesive at the inter-
bonds and are suitable for quickly performed overload face, a jig was designed (see Fig. 22a), to strip a thick
excursions. In these tests, the bonds are stressed over
short times and researchers have little or no time to ob-
serve, study or document the initiation or progression of
failure [143]. The tests are used for the design, quality
control and inspection purposes of the bonded com-
ponents and are unable to detect slow time-dependent
failure of the bond which may occur over long times
in service [144]. Furthermore, in these tests, failure of-
ten takes place within the adhesive if a strong bond is
formed, and away from them immediate vicinity of the
bonded interface [145]. In service where some rubber-
to-metal bonded components fail unexpectedly soon,
sometimes failure takes place at or near the bonded
interfaces. The inability of current tests to detect time-
dependent failure can produce misleading results for
products intended for long service life. Therefore, be-
cause of these problems faced in practice, Ansarifar
and co-workers [48, 49, 110, 141] developed a new
constant-load peel test method, which was utilized to
(a)
produce failure with the rubber adjacent to the bond,
similar to that occurring in service. Fig. 4a illustrates the
test piece used to assess rubber-to-metal bond failure.
Preliminary studies [110, 141], using peel test pieces
(Fig. 4a) prepared by bonding unfilled and carbon black
filled natural rubber compounds to mild steel with com-
mercial solvent-based bonding agents, indicated that if
a constant force of sufficient magnitude was applied to
the peel leg, time-dependent mechanical failure near to (b)
the bond was observed. The failure locus appeared to
lie slightly within the rubber (typically at a depth of Figure 22 A novel method called scrap test method [33]: (a) design of
a few tenths of a milimeter) and was ascribed to co- scrap jig (not to scale) and (b) scrap test specimen (not to scale).

4751
film of adhesive from an adherend (see Fig. 22b). The stress at a large interface portion. Recently, these ba-
force required to remove the adhesive after the sample sic ideas of this arrangement were applied successfully
has undergone aging is recorded. In order to test this to determine the adhesion strength between a single
method, the specimens were prepared by hot curing a fiber and a matrix material [146]. It was claimed [142]
single part paste adhesive (AV119) on a mild steel sub- the main advantages of the test are the simple prepa-
strate to form a thick film. Knox and Cowling [33] have ration of the samples, the simple experimental set-up
used the following procedure to fabricate the scrape as well as the data-reduction scheme. The test applies
test specimens: (a) the required area on the substrate a nearly uniform transverse tensile stress to the inter-
is prepared for bonding, including shot blasting of the face. Therefore, it allows the application of the strength
surface, solvent degrease and followed by pretreatment concept on interfaces without simplifications forced by
with a primer if needed, (b) after applying adhesive to complex stress fields and without introduction of ad-
the adherend surface, heat resistant masking tape is used ditional parameters like a characteristic length. The
to give a clean working edge and control the film size, method is not constrained to glued interfaces, even
(c) then the specimen is laid in a jig (see Fig. 22b) with other preparation methods are possible, for instance,
an adjacent surface, sprayed with PTFE, placed against two-component injection molding or welding. The test
the adhesive film to maintain a constant thickness.The provides the possibility of measuring a real material
specimens were subjected to accelerated environmen- property, which is (nearly) independent of the specific
tal aging (i.e., at 30◦ C and 100% relative humidity), testing arrangement. The main disadvantage of testing
and then tested immediately wet at half-week intervals bimaterial joints [142] is the creation of a well-known
at laboratory ambient temperatures. It was suggested stress distribution along the interface. The strength con-
[33] that the scrap test method appears to be very use- cept also requires a uniform transverse stress at the joint.
ful for rapidly detecting changes in interfacial strength Finally, the interfacial stress must not have concentra-
of an adhesive/adherend system, and for distinguishing tions at the edges of the specimens to avoid surface ef-
amongst the durability performance of various surface fects. The stress analysis revealed [142] that the stress
pretreatments. state in the center of the specimen is biaxial, it has an
Schüller and Lauke [142] have proposed another in-plane tensile component and a transverse tensile one.
novel method for characterization of an interface be- All other components are nearly zero. The ratio of the
tween two pieces of the same polymer material by the two remaining components depends on the specific ge-
interfacial transverse strength. However, they suggested ometry. The use of the maximum stress criterion for
that the basic idea of this test method is not limited to this situation seems reasonable as a first step [142].
such situations but can be applied also to bimaterial The toughness of an adhesive joint may decrease
composites. Fig. 23 shows this new test arrangement considerably under dynamic- or impact-loading [147].
proposed by Schüller and Lauke [142]. The specimen This arises because adhesives are polymeric materi-
consists of two halves, each with a single-edge notch. als that exhibit plastic and viscoelastic deformations,
They are glued together to form a double-edge notched and thus their fracture behavior may be very dependent
specimen. The notches are semicircular. Therefore, the upon the rate of loading and temperature [90]. Hence,
production of such samples is quite simple. The speci- for applications such as in the automotive industry, for
men is then tested in a tensile testing machine. Due to example, when adhesives are being used increasingly
the notched shape, the stress state is not purely tensile in safety-critical areas, it is necessary to evaluate any
in the plane of the interface. It has a transverse com- possible decrease in performance that may occur when
ponent, which is generated by the notched shape. This the adhesively-bonded joints are subjected to impact
allows the optimazation of the interfacial stress state loading. A new test method for the dynamic loading
by changing the geometry of the specimen. Optimaza- has been developed by the automotive industry [90,
tion means, here, avoidance of stress components in the 148, 149] to evaluate the fracture behavior of adhesive
thickness direction and generation of a uniform tensile joints when subjected to relatively high rates of test

Figure 23 Schematic view of the new test arrangement and expected load-displacement Curves proposed by Schüller and Lauke [142].

4752
 strength of the structures. As pointed out by Li and
Lee-Sullivan [46], the use of adhesive bonding in pri-
mary structural applications has been somewhat limited
because of the difficulty in (a) defining and predicting
joint strength, and (b) designing the joint geometry to
optimize strength and reliability. Since these problems
are mechanical in nature, stress analysis is required to
understand how the force loads are distributed along
the adherends and adhesive layer. The stress analysis
of adhesively bonded joints is complex because the
Figure 24 The impact wedge-peel (IWP) test specimen [90]. joints rarely see simple or pure loading of any type (e.g.
shear, tension, or compression) but, rather, are usually
at various test temperatures. This is the impact wedge- subjected to some bending. Moreover, the flexibility
peel (IWP) test, for which an International Standard and geometry of the adhesive layer affects distortion,
(ISO 11343) test method [150] was recently adopted. and thus, the overall stress distribution [46]. As a re-
A schematic of the IWP test is presented in Fig. 24. The sult, rather than rigorous analysis of stress distributions
specimen is shaped like a tunning fork, and a wedge (of for complex geometries in the initial design stage, sim-
a specified shape) is drawan through the bonded portion ple geometries such as the lap and scarf joints are first
of the specimen, as indicated in Fig. 24. The Interna- analyzed numerically and tested experimentally. The
tional Standard [150] specifies that specimens should results from these simple tests provide insight and can
be 90 mm long and 20 mm wide, and made using sheet- also be extrapolated for designing more complex and
metal adherends of between 0.6 and 1.7 mm thick. The practical geometries with the aid of sophisticated com-
adherends should be bonded over a length of 30 mm, putational tools such as finite element analysis (FEA)
the unbonded arms being formed to give the “tuning method.
fork” profile. No starter crack or notch is used with Rather limited work can be found to date in the use
these specimens. The wedge velocities reccommended of finite element analysis (FEA) optimisation on adhe-
by the ISO are 2 m/s for steel adherends, and 3 m/s for sively bonded joints [i.e., 3, 153, 154]. Groth and Nord-
aluminum-alloy adherends. The more detail about the lund [154] studied optimum shapes of adherends with
application of this test method is given in Ref. [90]. the intention of minimising the Tsai-Hill factor (objec-
tive function) along the adhesive overlap. The Tsai-Hill
yield criterion is a general orthotropic form of the Von
5. Stress concentrations in adhesively Mises isotropic yield hypothesis. Joint types included
bonded joints in this analysis were: (a) single-lap, (b) double-lap, (c)
In bonded joints, the adhesive is critical to the trans- double-strap and (d) a console bonded to a rigid wall.
fer of load from one adherend to another. The adhesive The adherends for both the single- and double-lap joints
joint is characterized by stress peaks (both shear and had to be pre-profiled so that the mesh distortions would
peeling components) due to the peculiar structure of not make the analysis unfeasible.
the joint: geometrical and stiffness factors are concur- Adams and Peppiatt [155] have studied the effect
rent in determining the entity of these peaks [61]. Joints of spew fillet size. The spew can be considered as the
are sources of stress concentrations which diminish the result of the adhesive squeezed out of the lap region
overall efficiency of a structure [3]. In strength criti- at the moment of the joint manufacture. The effect of
cal components it becomes imperative to reduce these the fillet size was studied by assuming 45◦ fillet and
stress concentration factors so as to increase structural changing its size. The magnitude of principal stresses
efficiency. Stress concentrations in bonded joints arise obtained from such analysis showed a large difference
from abrupt changes in adherend/adhesive thicknesses when compared to results obtained from analysis which
and from differences in elastic moduli [70, 151] un- assume the adhesive to have a square edge. The highest
der static [152] and dynamic [63] loadings. The actual principal stress was found to lie on the corner of the
magnitude of the stresses depends on many geometric unloaded adherend. This correlated well with observed
and material property parameters. Single-lap joints are failure characteristics during testing. The inclusion of
widely used in most studies due to their simplicity and a triangle spew (45◦ ) decreased the magnitude of the
practical applications. However, they have stress con- maximum principal stress by 40% when compared to a
centration on both edges [72]. Thus, it is necessary to square end adhesive fillet.
reduce the stress concentration by means of changing Crocombe and Adams [152] included the effect of
the adherend configurations. For example, tapered lap additional parameters such as material and geometric
joints (see Fig. 12h) and scarf joints (see Fig. 12c) are properties. In this analysis, the stress distribution across
useful to reduce stress concentration for static loading the adhesive thickness was not assumed constant. Their
[63]. However, it is not known whether these configu- results for a lap joint with a spew fillet indicate that the
rations are effective or not under dynamic condition. averaged maximum adhesive stress increased with de-
The primary considerations in the design of struc- creasing modulus ratio (increasing adhesive stiffness)
tures are the stiffness and strength (particularly with up to a limiting value, after which they decreased again.
respect to weight). Without proper joints, it is not pos- Variations of overlap length, adhesive thickness were
sible to take full advantage of the high stiffness and also included in their analysis.

4753
 Dorn and Weiping [156] performed a number of FEM resulting high stresses often intiate joint failure. There
analysis and some experimental measurements in order are many useful studies on the analysis of bonded joints
to study the effect of spew fillets for joints with different (see for example, Ref. [125–128, 135–139]). Analyses
couples of plastic and metal parts. In that work the spew have been carried out for various joint configurations
angles are not considered design variables, their values and for different properties of the adherends and ad-
are assumed at the beginning and maintained fixed for hesives using closed form or numerical analysis meth-
the whole article. Hildebrand [157] studied different ods. As mentioned previously, the primary function of
shapes (obtained by tapering, rounding or denting) of a joint is to transfer load from one structural member to
the extremities of the adhesive layer in joints of metal another. In most bonded joints the load transfer takes
with fiber-reinforced plastics parts; also in this case, the place through interfacial shear. For example, as shown
analysis was performed by the FEA method. He con- in Fig. 25a, in composite materials, the interfacial shear
sidered the shape of the extremities as design variables gives rise to high interlaminar normal and shear stresses
and developed an optimazationn study. A comparative for a single-lap joint. As pointed out in the previous
analysis of the effect of the spew geometry considering section, the actual magnitude of the stresses depends
triangular and rounded profiles, has been developed by on many geometric and material property parameters,
Lang and Mallick [158] by the FEA method. Detailed such as the thickness and length of the adhesive layer
analysis has been devoted to study the effect of the ad- compared to the corresponding values for the adherend
hesive thickness, while the effect of the spew angle is material, flexibility of the adhesive and type of load to
only addressed briefly and their conclusions on the sub- be transferred (i.e., simple loading of shear, compres-
ject do not explain the reported maximum peel stress sion; complex loading such as bending). As seen in Fig.
dependency. Recently (159) they also investigated the 25a both interlaminar normal and shear stresses have a
effect of a recess (i.e., lack of adhesive in the central large stress concentration near the end of the joint. In the
portion of the lap) in the joint. In both studies the joint rest of the joint they are uniformly distributed. Because
concerns adherends equal in material and thickness. of this high stress concentration in the adhesive layer,
In recent years, the strength of the adhesive joints high stresses are produced in the adjacent plies of the
under impact (i.e., dynamic) loading has become im- adherend laminates [60]. Therefore, failure may initiate
portant because their use expands to the aircraft and in these plies. It was suggested [168] that an effective
automobile industries [63]. For these applications, the way of reducing the local high stresses in the plies ad-
crash worthiness of the structures should be proved, jacent to the adhesive layer is to interleave the plies of
so that the dynamic strength of adhesive joints used the adherend laminates so that adhesion takes place in
in the structures should be predicted analytically. For many layers and, consequently, stresses are distributed
example, the dynamic strength of the adhesive bond in many plies. As pointed out by Agarwal and Brout-
was investigated intensively by Adams and Harris [160, man [60], interleaving is particularly desirable when
161], Kinloch et al. [162], Latailade and Cayssials [163, the number of plies in the laminate is large.
164], Yokoyama [165], Bezemer et al. [166], Blackman The influence of load transfer depends on the stiff-
et al. [90]. However, a few investigations on the stress ness of the bondline. Heslehurst [80] has recently in-
distribution and its variation with time in joints have vestigated the load transfer along the adhesive bondline.
been performed by Zachary and Burger [167], and Sato The elastic distribution of shear stress over the bondline
and Ikegami [63]. Sato and Ikegami [63] have recently of a double-lap joint is now recognized as being non-
investigated the dynamic deformation and the stress uniform, with peaked ends and a shallow trough [80]
concentration in single lap joints, tapered lap joints (smilar to Fig. 25a), The distribution can be modelled
and scarf joints using analytical and experimental tech- as a hyperbolic function [169]:
niques. They simulated the deformation and stress dis-
tribution in these joints under impact loading using the τ (x) = A sin h(λ x) + B cos h(λx) (18)
finite element method.

τav (λ la /2) λx la la
Therefore, based on above discussion and the litera- τ (x) = cos h for − ≤ x ≤ +
ture, we can conclude that the various factors have been sin h(λ la /2) 2 2 2
taken into consideration to analyze and optimise the (19)
adhesively bonded joints. These factors include the (a)
thickness and length of the adhesive layer, (b) adhesive where la is the joint length,
and adherend elastic properties( such as elastic moduli,
flexibility of the adhesive) (c) plastic yielding behavior 2G a
of adhesives [14], (d) geometries of joints [such as ta- λ2 = (20)
pering, rounding, denting, scarf, spew fillet, recessing E s ηs(o) ηa
[62] etc], (e) type of static loading to be transferred (i.e.,
shear, comprerssion, tensile bending), and (f) dynamic E s is the adherend Young’s modulus, G a the adhesive
or impact loading [63, 90, 160–165]. shear modulus for the elastic/plastic shear stress/strain
model, ηs(o) is the outer adherend thickness, ηa is the
adhesive thickness, τav is the average shear stress over
5.1. Load transfer in adhesively the joint length (=P/2la ) and P is the axial load per
bonded joints unit width, la is the joint length. Heslehurst [80] studied
The joint must be carefully analyzed because it rep- the effect on the elastic shear stress distribution with a
resents a discontinuity in the material, and therefore variations in adhesive shear modulus. Using the data
4754
 (a)

(b)

(c)

Figure 25 Stress distributions (or concentration) along the overlap length: (a) variation of interlaminar normal and shear stresses for single-lap joint
[60], (b) effect of adhesive stiffness on the adhesive shear stress distribution [80], and (c) effect of adhesive shear modulus on the shear stress (load
transfer) rate [80].

for an elastically balanced, thermally matched double- length [80]. This is clearly shown in Fig. 25b. Note, the
lap joint, a plot of the shear stress distribution for an average shear stress (τav ) and the areas under each of
overlap length la of 100 mm and applied load/unit width the curves remains the same.
(P) of 1313.5 kN/m), shows the classical stress distribu- The rate of change of the elastic shear stress distribu-
tion, as seen in Fig. 25b. This shear stress distribution is tion (or the rate of load transfer) is thus given as [170]:
based on the adhesive elastic-plastic shear stress/strain
model [169]. The adhesive is a ductile type with shear ∂τ (x)
= Bλ sin h(λx) (21)
modulus of 414 MPa. If the adhesive mechanical prop- ∂x
erties are reduced (i.e., shear modulus) the elastic stress
where
distribution is modified by significant reduction in peak
stress at the overlap ends, but only marginal increase in τav (λ la /2)
the shear stresses over the central portion of the overlap B= (22)
sin h(λ la /2)
4755
A comparative study for a double-lap adhesively theory principles [182]. It would therefore be useful
bonded joint, using the same data as in Fig. 25b, gives to compare the differences between plane strain and
the shear stress rate of transfer over variations in adhe- plane stress conditions for simply supported (rotation)
sive shear modulus. This is shown in Fig. 25c. Thus, and fully fixed (non-rotational) boundary conditions for
from this figure it was concluded [80] that with a higher joints with relatively thick adherends in comparison to
shear modulus the load transfer rate is significantly the adhesive layer [46]. As compared to the theoreti-
greater. Therefore, with a higher rate of load transfer cal and FEM studies, there is even less published ex-
there is a concentration of load at the joint end that in- perimental work in this area. The experiments usually
duces problems. Such problems will ultimately lead to consist of strain gaging joint specimens and/or the mea-
a failure at the joint end. However, stiffness losses in surement of adhesive strains using the full-field Moire
the adhesive do have an advantage, but a corresponding interferometry technique [173–176].
loss of strength will also be evident [170]. In the FEA model quadrilateral (i.e., a flat shape with
four sides such as a square or a rectangle) elements
require a very fine mesh for adequate accuracy [91].
6. Stress analysis using 2-D and 3-D finite For this reason efforts have been made to reduce the
element analysis methods in adhesive size of the models. The first approach was made by
joints: Geometricly-linear and nonlinear Yadigari et al. [183] who modeled the adhesive layer
adhesive design using only one row of six-node plane elements. Car-
Stress analysis provides the connection between the ap- penter and Barsoum [184] modeled the adherends us-
plied load and the created interfacial stress as well as ing two-node beam elements and the adhesive layer by
the optimal specimen geometry. The determination of a linear plane element with offset nodes. The number
the best geometry is very complex problem. Structural of degrees-of-freedom is reduced appreciably by this
optimazation of a joint tries to modify the geometry approach because the adherends and the adhesive use
of the adhesive layer and also of the adherend parts, the same nodes. Previously, Amijima and Fujii [185]
with the scope of reducing the intensity of the stress developed a similar model but they required that the
peaks and therefore increase the strength of the joint displacements in the adhesive-adherend interfaces be
[61]. This approach has been made possible by the nu- compatible. Taylor [186] used a similar approach to
merical analysis to treat cases with geometrical features model three-dimensional (3-D) adhesive joints; in this
that are beyond the analytical approach, such as fillets case, the adhesive was modeled using solid 3-D brick
introduced at the ends of the overlap zone. This line of elements and the adherends were represented by plates.
research has been explored over the last two decades Naboulsi and Mall [187] modeled the adhesive layer
[i.e., 70, 156–159, 171, 172], accounting also for the and adherends using shell elements; this approach con-
non-linearities occurring during collapse (material in- siders the contribution of the bending stiffness of the
elasticity, large displacements). The effects occurring adhesive layer.
at the ends of the overlap are extensively discussed in Andruet et al. [91] have recently developed a spe-
detail in Ref. [70], on the basis of previous paper of cial 2- and 3-D finite element analysis (FEA) methods
Adams and co-workers. for stress and displacement analyses in bonded joints.
Since the 1970s, finite element analysis (FEA) meth- Both the 2-D and 3-D elements were used to model the
ods have increasingly been used to analyze the various whole adhesive system; adherends and adhesive layer.
adhesively bonded joints. Wooley and Carver [98] made In their two-dimensional finite element analysis (2-D
one of the first finite element analyses of a single lap FEA), Andruet et al. [91] have used Bernoulli beam el-
joint. They used plane stress elements, and their re- ements to represent adherends, with axial deformation
sults were comparable with those from the Goland and and the adhesive layer by plane-stress or plane-strain
Reissner [72] solution, the seminal work in the field. elements. The nodes of the plane stress-strain elements
Countless studies of various joints have followed this that lie at the adherend-adhesive interface are rigidly
first work. The computational approach is now consid- linked with the nodes of the beam elements, resulting
ered a necessary and cost effective tool for obtaining in offset nodes which coincide with the midplanes of
simulations and also for validating analytical solutions. the adherends. On the other hand, the 3-D elements
The relatively large amount of work in this field can be consist of shell elements that represent the adherends
found (see for example, Refs. [98, 105, 116, 152, 155, and solid brick elements to model the adhesive. The re-
173–181]). However, in many of the work, the bound- sulting mesh can represent arbitrary beam- or plate-like
ary conditions at the two unbonded adherend ends are geometries, which are a larger part of adhesive joint de-
assumed to be hinged at one joint end and a roller on signs. This model can include debonds as well as cracks
the other. The horizontal tensile force is then applied within the adhesive, therefore it can be used for dura-
at the roller end. Moreover, a two-dimensional (2D) bility analysis of adhesive joints. Their models (the 2-D
plane strain condition is usually employed in the FEM and 3-dimensional formulations) included (a) geomet-
simulations as the adherend width is much larger than ric nonlinearities, and (b) the temperature and moisture
its thickness. The applied boundary conditions are con- effects. In their formulations the reason to include the
sidered as rotation boundary conditions since the two geometric nonlinearities was to account for large dis-
unbonded adherend ends can rotate during tensile load- plaments in the 2-D and 3-D adhesive elements.
ing [46]. Mostly, analytical solutions also assume plane On the basis of above discussion and literature,
strain conditions since the analyses are based on beam we can therefore clarify and summarize the various

4756
conditions used in the FEA methods: (a) 2- or 3-D for- the beam-truss elements. This element converges faster
mulations, (b) plane-strain or plane-stress condition, (c) than those with non-compatible displacements in the
geometrically linear and non-linear analyses, (d) types adhesive-adherend interface, as shown by Taylor [186].
of loading (i.e., static or dynamic loading), (e) tempera- Adherends were modelled using the large-displacement
ture and moisture effects, (f) boundary conditions (i.e., Bernoulli-Euler beam elements. The formulation of this
rotational or non-rotational), (g) geometric shapes of element can be found in the literature [188]. Using this
joints (i.e., single- or double-lap joints, tapering, scarf, formulation Andruet et al. [91] obtained the folowing
fillet, spew, chamfer, recessing etc), (h) linear elastic be- finite element model for the adherends as
havior of the joint components, (i) non-linear or plastic
yielding behavior of joint components (i.e., large rota-
[K Adher ({})]{} = {F} (23)
tion, plasticity).
As mentioned above, in designing 2-D geometric
non-linear adhesive, Andruet et al. (2001) [91] have where {} is the nodal displacement vector, {u 1 , u 2 ,
modelled the adherends as Bernoulli/Euler beam ele- s1 , s2 , s3 , s4 }, K ({}) the stiffness matrix and {F} the
ments with axial and bending deformation modes. The nodal load vector.
adhesive is a standard plane-strain quadrilateral ele- Andruet and his co-workers [91] have formulated
ment except that the nodes are offset to coincide with the the adhesive layer as a plane-stress/strain solid element
midplane of the corresponding adherend. The beams with offset nodes. In order to make the displacements
and special adhesive elements are shown in Fig. 26a. in the adhesive-adherend interfaces compatible, special
(In this figure, u and v are axial and transverse dis- interpolation functions were used [186]. The finite el-
placements, respectively). The formulation is similar ement discretization, coordinates of any point in the
to that of Carpenter and Barsoum [184], except that adhesive can be expressed as a function of the nodal
the interpolation functions for the adhesive elements coordinates of the element
were chosen to be compatible with the deformation of

{x} = Ni x i (24)

where {x} = {x, y} are the coordinates of a generic

point in the adhesive, {xi } = {xi , yi } the nodal coordi-
nates and Ni the linear shape functions of local coordi-
nates ψ and ϕ (see Fig. 26a). The displacements of the
adherends were expressed as a summation of polyno-
mial functions. The expression for the displacement of
the bottom adherend is

{u adhes } = [Ni ]adhes {u i }adhes (25)

where
(a) 
u
{u adhes } =
v adhes
{u i }adhes = {u 1 , v1 , u 2 , v2 , u 3 , v3 , u 4 , v4 }adhes

and [Ni ]adhes are interpolation functions in the adhesive

layer.
In order to have displacement continuity in the
adhesive-adherend interfaces, appropriate interpolation
functions for the adhesive layer must be found. For this,
[Ni ]adhes are obtained by imposition of equal displace-
ment of the adhesive and the adherends at both inter-
faces ab and cd (Fig. 26a),

{u adhes }ab = [Niab ]adhes {u i }adhes = {u adher }b

= [Ni ]adher {u i }adher b (26)
{u adhes }cd = [Ni cd ]adhes {u i }adhes = {u adher }u
(b) = [Ni ]adher {u i }adher u (27)
Figure 26 Finite element configuration using the geometrically-
nonlinear adhesive [91]: (a) 2-D adhesive finite element and (b) 3-D where [Ni ab ]adhes and [Ni cd ]adhes are the interpolation
sandwich-type configuration of the adhesive element. functions for the adhesive evaluated at the ab and cd
4757
interfaces, respectively. From these expressions the dis- In their study, Andruet et al. [91] have also taken into
placements in the adhesive layer were obtained as func- account the effects of thermal and moisture by defining
tion of the nodal displacement of the adherends an initial strain εo , which is a function of the thermal and
moisture coefficients and the temperature and moisture
{u adhes } = [Ni ]{u}adher (28) variations relative to the initial times. This initial strain
does not produce stress, accordingly
where [Ni ] are the interpolation functions of the dis-
placements in the adhesive layer with respect to the σ = D(ε − εo ) (34)
adherend nodal displacements.
From these equations Andruet et al. [91] obtained Andruet et al. [40] have also formulated a 3-D finite
cubic and quadratic interpolation functions for the ver- element model for geometric nonlinear design, which
tical and horizontal displacements, respectively, in the is an extension of the adhesive element developed by
direction of the axis of the joint, the ψaxis in Fig. 26a. Taylor [186]. This model of the adhesively bonded joint
In the direction transverse to the axis of the joint, axis has two main components: (a) the adhesive layer and
ϕ in Fig. 23a, linear interpolation functions are used. (b) the adherends. These components are combined in a
The constitutive law for the adhesive is sandwich-type configuration as shown in Fig. 26b. The
  model consists of two adherends represented by gen-
1 νa 0 eral shell elements and the adhesive layer modeled as a
Ea 
t C ijrs = D = νa 1 0   solid element with offset nodes in the mid-planes of the
1 − νa2
1−νa adherends. In the work by Taylor [186], the adherends
0 0
  2
 were modeled by Mindlin plate elements. Since cuved
Plane Stress structures are modeled better with shell than with plate
elements, the newly formulated element enhances the
or rate of convergence for practical joint configurations
  [91]. Using the modelling of the adherends as shell ele-
1 − νa νa 0
 ν ments, formulation given by Bathe [189], Andruet et al.
0 
Ea
D=  1 − νa  [91] have formulated the displacement vector, and that
(1 + va )(1 − 2va ) 1−2νa
0 0 they presented the constitutive relation for the 3-D ge-
  2
 ometric nonlinear design as follows:
Plane Stress
 
(29) E 11 E 12 0 0 0 0
E 
 12 E 22 0 0 0 0 
where E a is the elastic modulus and va the Poisson ratio.  
0 0 0 0 0 0 
Substituting Equations 28 and 29 into Equation 1 in C =
0
 (35)

Ref. [91], they obtained the finite element model of the  0 0 G 12 0 0 
 5 
adhesive 0 0 0 0 6 G 23 0 
5
0 0 0 0 0 6 G 21
[K Adhes ({})]{} = {F} (30)
The third row and colomn are empty because σ33 is
and the model of the adherend-adhesive system is the negligible for plates and shells.
summation of the stiffness matrices of the adherend and They modelled the adhesive layer as 3-D solid brick
adhesive: elements, with nodes offset to the shell midplanes [91].
This is equivalent to condensing out the degress of free-
[K Adher + K Adhes ]{} = [K ]{}{F} (31) dom of the continuum elements into the degrees of free-
dom of the adjacent shells that model the adherends.
Equation 31 can be expressed as follows: Using the displacement continuity at the adhesive-
adherend interfaces improves the accuracy of relatively
[K L + K NL ]{U } = {F} (32) coarse meshes. Quadratic interpolation functions were
used in the planes of the element parallel to the ad-
where K L and K NL are matrices representing the linear herend mid-planes in order to have compatibility of the
and nonlinear stiffness of the model, respectively, and displacements in the adhesive/adherend interfaces [91].
U is the displacement vector. The nonlinear Equation In the thickness direction, linear interpolation functions
32 is solved by means of the modified Newton-Raphson were used. Quadratic and linear Lagrange polynomials
method: are used. Analogous to the two-dimensional element,
t  t + t
the coefficients of the interpolation functions for the ad-
t KL +tt K NL U (i) = f (i−1) (33) hesive were obtained by imposition of equal displace-
ments of adherends and adhesive in their interfaces.
where U (i) is the displacement vector for the ith it- Isotropic and orthotropic constitutive laws were used.
eration and R-F is the unbalanced force vector for the Similar to the 2-D case, in 3-D geometric nonlin-
t + t step. The solution of Equation 33, U (i) , was ear adhesive design, thermal and moisture effects were
used to obtain an improved prediction of the equilib- taken into account by definition of an initial strain εo ,
rium configuration. which is a function of thermal and moisture coefficients
4758
and the temperature and moisture variations relative to terion based on a stress intensity factor derived from
the initial times. Using the D parameter in Equation 35, bi-material singularities at the ends of bonded joints.
Equation 34 can be used in the present case. This initial In another approach, Crocombe et al. [191] ignored the
strain does not produce stress, accordingly. final element incorporating the singularity point and
Gonçalves et al. [92] have developed a new 3-D FEA extrapolated to the corner to obtain the peak stress val-
method of adhesive joints. The model considers geo- ues. Adams and Harris [192] reasoned that, in prac-
metric and material non-linerarities as in the case of tice, sharp corners at the end of the lap joint are always
Andruet et al.’s work [91] and used solid brick elements rounded slightly during manufacture. Therefore, to deal
as well as specially developed interface elements. The with singularity behavior they used rounded corners,
main objective was to calculate the stresses at the in- which produced a uniform stress field.
terfaces between adherends and adhesives, which are
considered critical regions in these structures. They for-
mulated an equilibrium equations of a structural prob- 7.1. Stress and strain analysis under static
lem involving contact as minimization of the poten- and impact loadings
tial energy subjected to certain kinematic constraints. Schüller and Lauke [142] have analyzed the stress state
Contrary to continuum elements where stress-strain re- of a newly developed test specimen with a semicircular
lationships are used, interface elements are governed notch (see Fig. 23) using the 3-D finite-element anlysis
by stress-relative displacements relationships. The vec- model. As seen in Fig. 23, the two halves in the speci-
tor of relative displacements between two homologous men were treated as homogeneous solids connected by
points were obtained from the displacements fields as- an ideal interface. It was assumed [142] that this sim-
sociated with the element faces (top and bottom). Then plification of the complex structure of a real interphase
the stresses at the interfaces were obtained from the is reasonable as long as the interphase is thin com-
relative displacements (δ) as follows pared with the whole specimen and does not influence
strongly, the stress field. The average tensile stress in
τsn = ks δs the smallest cross-section is σav = P/(2w r η), where P,
wr and η denote the applied load, the remaining width of
τtn = kt δt (36) one notched specimen half and the specimen thickness,
τnn = kn δn respectively. Due to the symmetry of the specimen, all
shear components vanish in the center of the specimen.
where s and t represent the tangential directions and n Finite-element results for a necked specimen published
the normal direction. or Equation 36 can be written in in [193] suggest that plane-stress condition is a good
matrix form approximation for the stress field leading to the assump-
tion that the stress state in the center of the specimen
σ = Dδ (37) is approximately biaxial. Assuming the linear-elastic
material, all non-zero stress components are propor-
where tional to the applied load and also to σav . As stated by
  Schüller and Lauke [142], then, the normalized interfa-
ks 0 0 cial transverse stress σit /σav and the combined loading
  ratio σip /σit completely describe the biaxial stress field,
D = 0 kt 0 (38)
0 0 kn where σip denotes the in-plane component. These ra-
tios depend on the specific geometry. The notch also
The parameters ks and kt are the shear interface stiff- causes a stress concentration at its ground. It was sug-
ness and kn is the normal interface stiffness. Their units gested [142] that if this stress concentration exceeds the
are N/m2 and they represent the penalty parameter in- linear-elastic limit of the polymer material during ex-
troduced by the user. Note that these parameters need periment, this analysis would not be valid. Therefore,
to be carefully chosen in order to obtain a good perfor- even the stress ratio σit /σeqv is an important parameter
mance of the model; small values induce large inter- describing a specific geometry. σeqv is the maximum
pretations incompatible with the physical reality while equivalent stress at the neck ground. The ratio expresses
large values produce numerical problems. the highest attainable interfacial stress within the limit
of linear-elastic theory. The variation of the notch ra-
dius RN and each shape is uniquely characterized by
7. Application of 2-D and 3-D finite element the shape parameter RN /a.
analyses methods on optimizing The three-dimensional model finite element analy-
the adhesive joints sis allows evaluating even the variation of the inter-
At the adhesive joint ends, the analytical solution (i.e., facial stress in the direction of thickness. The model
finite element analysis) of stresses can display singular is parametric, which makes the evaluation of differ-
behavior (i.e., approach infinite values) [62]. An infinite ent specimen shape very convenient. The interfacial
stress is of course a mathematical anomaly, and in prac- transverse stress σit is equal to the stress component
tice does not exist. To avoid any adverse effects from the σx at the interface (x = 0). The interfacial transverse
singularity point several approaches have been under- stress has a wide maximum between the notches that
taken. One of these approaches is based on the fracture reaches σit /σav = 0.24 for this geometry (RN /wr = 1,
mechanics principles of energy release rate and stress η/wr = 1). It is slightly higher at the surface of the
intensity factors. Groth [190] formulated a fracture cri- specimen but does not show any concentration or even
4759
Figure 27 Normalized interfacial tranverse stress σit /σav (above) and
stress ratio σit /σeqv (below) versus shape parameter RN /wr [142]. Figure 28 Variation of stress in the adhesive layer of single lap joint
under step loading (L = 100 mm) (Ref. [63]).

singularity. The variation of the interfacial stress along static loads. For instance, the distributions of dynamic
the thickness indicates the stress state is really three- stress were not symmetric although those of the static
dimensional and that the specimen thickness influences stress were symmetric. Stress distributions in single lap
the stress field. The shear components σxy , σxz , σyz and joints under static loading depended on the distance be-
also σy are small compared with σx . The ratio of bi- tween the joints and the load applying points, i.e., on
axiality is σip /σit = 3.5. The influence of the shape the adherend length without the joint length [63]. If the
parameter RN /a on the normalized interfacial trans- distance increased, the bending deformations and the
verse stress σit /σav and stress ratio σit /σeqv was shown stress concentration also increased. Similar phenom-
in Fig. 27. The highest attainable interfacial transverse ena occurred during the impact test. As the flexural
stress is reached for a shape parameter RN /wr ≈ 1.5. waves propagated outward in the adherends, the region
Nevertheless, Schüller and Lauke [142] recommended of bending deformations was spread out. It was equiv-
to use RN /wr = 1 because the interfacial transverse alent to the increasing of the adherend length at static
stress σit /σav has a maximum there and the stress ra- tests. Then the stresses at the edges increased gradually
tio σit /σeqv is nearly the same. The wide maximum of as shown in Fig. 28.
σit /σav prevents impression of sample preparation (es- They also studied the stress distributions and the time
pecially imprecise shape parameter RN /wr ) from hav- variation of the stresses in the adhesive layer of a tapered
ing large effects on σit /σav . lap joint, which has the same lap length of 100 mm as
The strength of the adhesive joints under impact (or the single lap in the joint. These joints showed similar
dynamic) loading has become important because of the phenomena to the case of the single lap joint, i.e., there
important industrial applications such as the aircraft and were transient variation and gradual increase. The stress
automobile industries. The dynamic strength of adhe- values at both the edges of the tapered lap joint were
sive bond was investigated by many researchers (i.e., less than half of those of the single lap joint which has
for example, Refs. [160–167]). Sato and Ikegami [63] the same lap length. Therefore, tapering of adherends in
have recently investigated the stress distribution and lap joints was effective to reduce stress concentration.
the time variation of stress and strain in the joints un- Finally, in the scarf joints investigated (i.e., which
der tensile impact loading using the 2-D finite element had the same lap length of 100 mm as the single lap
method with plane strain condidions, considering the joint and the tapered lap joint), the stress values and
viscoelastic properties of the adhesive in the single lap the stress concentration were smaller than those in the
joints, tapered joints and scarf joints. The joints stud- other joint configurations discussed in this work. As
ied consisted of aluminum alloy and epoxy resin. The pointed out by Sato and Ikegami [63], scarf joints do
element thickness of the adherends was changed grad- not have a geometrical offset of the adherends. Thus,
ually in the tapered part so that the acute angle of the a bending moment and a flexural wave, which often
adherend could be described. The adhesive layer was cause stress concentration in lap joints, are not induced
treated as a single element layer because the adhesive because stress waves propagate smoothly in the joints.
layer was very thin. Calculated results of the dynamic deformation of the
As shown in Fig. 28, there was stress concentration at joints in the present FEM analysis showed good agree-
both edges of the adhesive layer in the single lap joint. ment with the experimental results [63] (see Section 8
At the initial stage of loading, the stresses increased for more detail).
and showed peaks at both the edges in the adhesive
layer. The peak value of a shear stress at the left edge
(stress-wave-imput side) was the greatest in the stress 7.2. Effects of geometric linear and
components (see Fig. 28). The maximum value of the non-linearity, T-joints, fillets, spews,
shear stress was over 0.4 MPa when the applied stress chamfer size, and recessing on the
was 1 MPa. At the right edge, there was also a great strength of single-lap joints
peak of a normal stress in the y-direction, and the value 7.2.1. Geometrically linear and non-linear
was over 0.4 MPa. After the peaks, the stresses in the analysis in 2-D finite element method
single lap joint decreased. The stress distributions dur- Li et al. [73] have studied the effects of processing
ing the initial stage were greatly different from those of variables such as bonded length, adhesive thickness,
4760
and adherend thickness on the static stiffness (i.e., lin- As shown in Fig. 30a, the linear stiffness is affected
ear and rotational stiffnesses) of the adhesive bonded by the bonded length for the case of loading Px , and the
T-joints using the linear elastic 2-D FEA method. The effect becomes less as the bonded length is increased. In
adhesive joint consists of a right angle plate bonded to comparison between a bonded length of 10 and 16 mm
a rigid plate using adhesive (i.e., T-joint) [see for exam- the linear stiffness increases by only 0.1%, but the in-
ple, Fig. 4a (θ = 90◦ ). Both adherends bonded are mild crease is 0.7% between length of 6 and 10 mm. How-
steel with a modulus of elasticity of 210 GPa and Pois- ever, for the case of the loading direction Py , when the
son’s ratio of 0.29. An epoxy resin adhesive was used bonded length increases from 6 to 10 mm, the linear
having a modulus of elasticity of 3.48 GPa. Because stiffness increases by 1.8%.The rate of increase contin-
of the discontinuous nature of the joint geometry and uously gets less, and the linear stiffness increases only
material properties in the interface between adherend 0.19% as the bonded length increases from 10 mm up
and adhesive, large stresses and strain may exist around to 16 mm. It is very clear that the absolute stiffness in
these regions. For this region, a gradual mesh refine- the y direction is considerably higher than in the x di-
ment was used to concentrate elements around te adhe- rection. The effect of the different bonded lengths on
sive layer.The joint model with plane strain condition the linear stiffness is plotted in Fig. 30a. For the case
was subjected to two linear loads and one bending mo- of effect of bonded length on the rotational stiffness,
ment. It was assumed that the adhesive and adherends between bonded length of 6 and 10 mm, increasing the
had linear elastic properties.The analysis of such joints bonded length results in increase of 0.4% in rotational
is complicated by the geometric complexity and by the stiffness and between a bonded length of 10 and 16 mm,
discontinuous nature of the adherend/adhesive inter- the increase is 0.2% [73]. Therefore, a bonded length
face. The linear FEA element method has been shown of 10 mm can be used without significant loss of the
to give good agreement with experimental results for joint stiffness compared with using any higher values.
T-joints [70] and was therefore used.The deflections Li et al. [73] have also analyzed the effect of adhesive
along the Px and Py load directions in the T-joints with thickness on linear and rotational stiffnesses. Fig. 30b
dimensions are shown in Fig. 29. The maximum de- indicates the effect of adhesive thickness on the stiff-
flections generally occur in the directions in which the ness of the T-joint (rotational stiffness in the bending
loads are applied and they decreased towards to the joint moment M), stiffness increasing by 6.14 between ad-
region.The linear stiffness (in Nmm−1 ) of the joint was hesive thickness of 0.05 and 0.5 mm. The variation of
calculated from the slope of the load/deflection curves linear stiffness in the y-direction shows similar trends,
in the direction of application of Px and Py . The rota- the stiffness decreasing by 37.64% between adhesive
tional stiffness (in Nmm) was calculated from the ratio thickness of 0.05 and 0.5 mm. It can be noted that the
of the applied moment M and the corresponding rota- stiffness in the y-direction is much higher than in the x
tional angle at the top point of the central axes of the direction. Therefore, it can be concluded that the joint
vertical and the horizontal plates. stiffness was reduced when the adhesive thickness was
increased.
Fig. 30c shows the effect of the adherend thickness
on the stiffness in the x direction lineraly increases by
88% between adherend thickness of 2–4 mm. The varia-
tion of linear stiffness in the y-direction shows different
trends, the stiffness increasing 57% between adherend
thickness of 2 and 4 mm. It can be found that the stiff-
ness in the y direction is also much higher than in the
x-direction. The variation of rotational stiffness shows
similar trends with that in the x direction, and the stiff-
ness is increasing by 88% between adherend thickness
of 2 and 4 mm. Therefore, It is clear that the joint stiff-
ness was increased when the adherend thickness was
increased.
In summary, the bonded length, adhesive thickness
and adherend thickness of the joint have marked effects
on the T-joint stiffness. It was also found that the linear
stiffness in the y-direction and rotational stiffness un-
der bending moment loading are much higher than the
linear stiffness in the x direction for all conditions.
Andruet et al. [91] have analyzed a single lap joint
using the 2-D finite element analysis methods. The
principals of these methods were presented in Sec-
tion 6. Note that the single lap joint geometry used
was popular in industrial applications as a benchmark
test and the bibliography for this joint is very extensive;
Figure 29 The deflections of the joints [73]: (a) loading in the x- there are many analytical solutions and test results. Peel
direction Px , (b) loading in the y-direction, and (c) loading in the moment and shear stresses were obtained performing a linear
M.

4761
 (a)

(a)

(b)

Figure 31 Stress distributions for a single lap joint with linear analysis
[91]: (a) peel stress and (b) shear stress.

analysis with a ADH2D, and that they were compared

(b) with those from a finite element analysis with ABAQUS
[ref. 194], using 2-D plane-stress finite elements (see
Fig. 31a and b).
The same specimen was also analyzed with geomet-
ric nonlinearities. This time the whole structure was
modeled, so the moments at the beginning of the over-
lap are no longer required. The specimen was subjected
to different levels of load. σav is the average adherend
stress, or the total load divided by the thickness and
the width of the adherends. Fig. 32a and b show the
effect of the nonlinearity in the peel and shear stress
distributions. Also, these figures show that stress con-
centrations decrease as the load increases. This is the
result of the reduction of the moment arm of the two
eccentric loads as their magnitude increases [91]. There
is a physical limit for that reduction of the moment arm,
which is when the two loads are colinear. Therefore, it
was suggested [91] that the change in the stress concen-
tration between two load levels becomes less important
as the load increases. The results for σav = 400 MPa
(c) from the ADH2D model are in good agreement with
those from the ABAQUS [ref. 194] model as shown in
Figure 30 Effects of bonded length, and adhesive and adherend thick-
Fig. 32a and b.
nesses on the stiffness of the T -joint [73]: (a) the linear stiffness versus
bonded length, (b) the rotational stiffness (in the bending moment M) Li and Lee-Sullivan [46] have studied an adhesive
versus adhesive thickness, and (c) the linear stiffness (in the x-direction) joint with relatively thick adherends in tension un-
versus adherend thickness. der both plane strain and plane stress conditions using
4762
 Figure 33 A comparison between FEM results and Oplinger’s solution
for bending moment factor (kop ) values for an unfilleted joint bonded
with the flexible adhesive [46]: (◦): Oplinger (plane strain), (): 2D
(a) plane strain FEM—No rotation, (): 2D plane stress FEM—No rotation,
(): 2D plane strain FEM—Rotation, and (∇): 2D plane stress FEM—
Rotation.

at the center node at the right roller adherend end. Most

of the analytical work [e.g. 72, 103, 104] applies bound-
ary conditions which allow for rotation. In the present
study for non-rotation boundary condition, the equiva-
lent nodal load was applied to every surface node based
on the iso-longitudinal strain condion, i.e., all the nodes
at the unbounded adherend end have the same longitu-
dinal strain, and the surface load or force can be deter-
mined. The corresponding nodal nodes can therefore be
determined based on the equivalent principle of static
forces from the finite element theory. Rotation is re-
(b)
stricted by not allowing vertical deflection at the last
two nodes along the adherend centerline.
Figure 32 Stress distributions for a single lap joint [91]. Geometric non- Fig. 33 compares the FEM results on the variation
linearities included. Dimensions of stresses are MPa: (a) peel stress and in the bending moment factor, k, for unfilleted joints
(b) shear stress.
with the theoretical solution proposed by Oplinger (see
Equation 10 in Section 2.3 for more details). As this
figure shows, the 2D plane strain FEM simulations for
2-D FEA method with the inclusion of a geometrically rotation boundary conditions give the upper-bound val-
non-linearity. Furthermore, the additional influence of ues of k while plane stress and no-rotation boundary
other factors, such as boundary conditions, overlap end conditions give the lower-bound values. The Oplinger’s
geometries, and adhesive modulus, on the bending mo- solution [102] (see Equation 16) using plane strain con-
ment factor at the overlap edge, k, and adhesive stresses dition agrees very closely with the k values predicted
(see Equation 1 in Section 1.2) has also been investi- by both plane strain and plane stress FEM simulations
gated. The bending moment factor, k, relates the bend- for rotation boundary conditions. The good agreement
ing moment on the adherend at the end of the overlap to should be expected since the analytical model implic-
the in-plane loading (see Section 2.4 for more detail). In itly assumes end rotation. These results suggest [46]
order to assess the accuracy of the FEM simulations, ex- that the boundary conditions at the adherend ends have
perimental testing which included measuring the strain a greater effect on k than the stress state of the joint,
distributions along the overlap region of the adherend i.e., whether plane strain and stress conditions. How-
was also performed. The strain variation between the ever, for same boundary conditions, the plane strain
upper and lower surfaces of the unbonded adherend condition would give slightly larger predictions of k
section adjacent to the overlap was measured. In the than plane stress.
FEM studies, filleted and unfilleted joints bonded us- Fig. 34 compares the effect of fillets on the variation
ing two different types of epoxy adhesives (i.e., rigid of k as obtained by FEM simulations. It is seen that the
and flexible) were simulated. The higher modulus resin k values of the unfilleted joint results are slightly higher
was referred to as the rigid adhesive [46]. The adherend than those for the filleted joints under the same bound-
material was aluminum 6061-T6. In this study, two dif- ary conditions. Some of the difference is probably due
ferent boundary conditions [i.e., as rotation (R) and to the slightly different planes of analysis chosen. It is
no-rotation (NR)] and FEM meshes were used for both seen in both Figs 33 and 34 that the k predictions be-
the filleted and unfilleted joints. In setting the rotation tween the filleted and unfilleted joints become increas-
boundary condition, only concentrated force is applied ingly closer with increasing tensile force. Additionally,
4763
 continuous single-lap joint in order to provide a base-
line for comparing the stresses in recessed joints.The
continuous joint had a lap length of 63.5 mm and a
spew radius of 3.3 mm. This makes the total distance
from the left spew (point O) to the right spew (point
E) along the interface to be approximately 70 mm. The
2-D FEA of the continuous joint shows that the shear
stress, peel stress and axial stress distribution at the ad-
hesive mid-thickness is symmetric about the vertical
centerline of the overlap. Across the mid-section of the
overlap, the stresses are constant and have much lower
values as compared to the peak stresses, which occur
near spew surfaces. At the spew ends shear and axial
stresses approach zero satisfying free edge boundary
Figure 34 The variation of bending moment factor k as a function of conditions, while peel stress decreases to a smaller non-
the tensile force for the filleted and unfilleted joints with the flexible
zero value. Away from the adhesive mid-thickness the
adhesive. N and R refer to no-rotation and rotation boundary conditions
used in FEM simulations [46]: (◦): plane strain FEM—R (with fillet), stress distributions are highly non-symmetric and the
(): plane stress FEM—R (with fillet), ( ): plane stress FEM–N (with maximum stress values are highly non-symmetric and
fillet), (•): plane strain FEM—R (no fillet), (): plane stress FEM—R the maximum stress values are much larger at the inter-
(no fillet), (
): plane strain FEM—N (no fillet), and ( ): plane stress face than at the mid-thickness [62]. The stresses along
FEM—N (no fillet).
the interface have approximately the same magnitude as
the stresses at the adhesive mid-thickness. However, at
the differences in predictions due to different boundary the loaded end of the spew, steep stress gradients along
conditions, i.e., rotation and no-rotation, also decrease the inteface appear with stresses approaching extremely
with increasing tensile force. It is expected that the pre- high values at the interface corner. At the unloaded end,
dictions will ultimately converge at increasingly higher the stresses are much smaller and are not of much con-
tensile forces [46]. cern. As an example, Fig. 35a shows the peel variation
Lang and Mallick [62] have investigated the bonded against the distance along the overlap including spews
single-lap joints using the 2-D FEM method (in plane for a continuous single-lap joint.
strain condition) under tensile loading by extending the The stress distributions for shear, peel and axial
work done by Mazumdar and Mallick [195, 196]. How- stresses at the adhesive mid-thickness for a recessed
ever, they produced the single-lap joints by removing joint with an effective lap length of 25.4 mm are shown
portions of the adhesive from the interior of the overlap in Fig. 35b. As this figure shows, the stress distribu-
(i.e., recessed joints). The gap or recess in the recessed tions at the mid-thickness are the same on each side
joint is centrally located in the overlap region. Full- of the recess, and near the lap ends they are similar to
rounded spew was selected as the spew shape, since it those observed at the adhesive midd-thickness of the
best represents the spew geometry [62]. The total lap continuous joint (see Fig. 35a). For instance, all three
length is defined as the length over which the two sub- stresses show maximum values close to the spew ends
strates are overlapped, while the effective or true lap and immediately reduce to lower values away from the
length is the actual length over which the adhesive is spew as the mid-section of the overlap is approached.
used between the substrates. Note that the effective lap However, near the ends of the recess, steep stress gra-
length decreases as the gap or recess in the adhesive dients appear for shear, peel and aqxial stresses. The
is increased. In the limit, the effective lap length ap- shear stress near the recess end increases to 50% of the
proaches zero when no adhesive is used in the overlap; peak stress and then decreases to zero at the free edge
spews are present at each end, which transfer the load of the recess. The peel stress has a steep compressive
from one substrate to the other.Both the adhesive and value, while the axial stress decreases to zero from the
the substrate were assumed to behave as linear elastic interior of the overlap to the free edge of the recess.
materials. The loading and boundary conditions used Lang and Mallick (1999) [62] have also investigated
simulate those encountered in a tensile test. The stress the effect of recessing on maximum stresses, as shown
distributions at the adhesive mid-thickness and inter- in Fig. 35c. This figure displays the stress distribution
face (i.e., at the interface between the adhesive and the of shear, peel, and axial stresses along the interface for
substrate) were determined for joints with various levels various effective lap lengths. As can be seen in this fig-
of recessing, and compared to the stress distributions in ure the peak stresses at the loaded end, near the interface
continuous single-lap joints. In the Lang and Mallick’s corner, do not increase significantly as the effective lap
study, an approach similar to that of Crocombe et al. length decreases. This is seen more clearly in Fig. 35d,
[191] was undertaken by considering stress values very which depicts the variation of shear, peel and axial
close to the singularity point. By using an an extremely stresses found at the interface corner (point E) as the
fine mesh around the singularity point, stress values effective lap length is reduced. Fig. 35d shows that all
as close as 0.04 mm away from the interface were three stresses stay constant, and increase only slightly
obtained. as the zero effective lap length is reached. One should
Lang and Mallick [62] first investigated the stresses note that the stresses used in Fig. 35d were taken at the
at the adhesive mid-thickness and at the interface in a interface corner, which is a singularity point. Therefore,

4764
to avoid portraying unrealistically high stress values, structure was modeled, including the two lateral shells,
interface stresses one and two nodes away from the to avoid any kind of assumption or simplification in the
singularity point were determined [62]. model. The thickness of the adherends is 1.6 mm, the
adhesive layer thickness is 0.102 mm, and the speci-
men’s width is 25.4 mm. The aluminum adherends and
7.2.2. Geometrically linear and non-linear epoxy adhesive layer were used in this study. In this
analysis in 3-D finite element method work, the finite element model of the single lap joint
Andruet et al. [91] have also studied the single lap joint was taken to be composed of 1190 nodes and 252 el-
specimen using the 3-dimensional version of the adhe- ements. The mesh is finer in the overlap than in the
sive finite element (ADH3D) (see Section 6 for the basis lateral shells. Since the most important effects occur at
of this analysis). The model of the adhesively bonded the overlap ends, refinement of the mesh in this region
joint has two main components: The adhesive layer and would improve the accuracy of the model.
the adherends. These components are combined in a Peel stress distributions at the mid-plane of the over-
sandwich-type configuration as shown in Fig. 26b. Even lap region are shown in Fig. 36. 3-D effects can be
though symmetry conditions could be used, the whole observed; specifically, the peel stresses are largest at

(a)

(b)

Figure 35 Peel (σyy ), shear (τxy ), and axial (σxx ) stresses for continuous and recessed (including spews) single-lap joint [62]. Load = 6.90 MPa;
Adhesive thickness = 0.762 mm; (Total) lap length = 63.5 mm: (a) variation of peel (σyy ) stress against the distance along the overlap including spews
for a continuous single-lap joint at the adhesive mid-thickness. (b) Peel stress (σyy ) variation for recessed single-lap joint at the adhesive mid-thickness.
Effective lap length = 25.4 mm. (c) Shear stress (τxy ) variation at the interface for a single-lap joint with various lap lengths. (d) Variation of shear
(τxy ), peel (σyy ), and axial (σxx ) stresses at the interface corner as the effective lap length is reduced. (Continued)

4765
 (c)

(d)

Figure 35 (Continued).

the center of the joint and decrease at the edges. This of the adherends, on the stress field of a steel/fiber-
behavior is in qualitative agreement with experimental reinforced plastic single-lap joints in order to opti-
results of Adams et al. [197]. This is due to the Poisson mise the design, identifying the geometrical configu-
effects and the anti-elastic bending of the adherends. ration that minimizes the stress peaks. The study has
The shear stress distribution in the yz direction (τyz ) been performed numerically, by means of the 3-D FEA
shows strong stress concentrations at the two corners method, assuming plane strain and linear elastic be-
in the free edge of the overlap, but the maximum ab- haviors of all the joint components. The latter limiting
solute values are on the order of the maximum values assumption was justified by the argument that, even if
of the other components of the stress tensor [91]. This it is well known that the ultimate strength of the joint
stress distribution clearly shows 3-D effects. largely depends on non-linear phenomena (e.g., large
To optimize the adhesive lap joints, Belingardi et al. rotation, plasticity), the reduction of the elastic stress
[61] have investigated the effects played by the joint peaks is often beneficial in case of low ductility and es-
geometry, namely the spews (see Fig. 12d) (i.e., shoul- sential in case of fatigue loading [61].The plates used
ders of adhesive connecting the unloaded ends of the as adherends are made of different materials, namely a
adherends) and of chamfers (see Fig. 12e) at the ends steel for plate 1 (i.e., adherend 1) and fiber-reinforced

4766
Figure 36 Distribution of peel stresses (σyy ) in the midplane of the adhesive layer of the single lap joint [91].

plastics for plate 2 (i.e., adherend 2), and have different Equations 39 and 40 hold exactly if the adherends are
thickness. The adhesive is an epoxy resin. The tensile identical, in the present case they are reasonably ap-
load applied to the joint corresponds to 10 MPa in the proximate since the stiffness values of the adherends
steel plate. The four-noded element type 42 has been are not too different. In order to differentiate the effects
used for all zones. The total number of finite elements is of fillet and spew on the adhesive joints, peel and shear
about 10.000, the adhesive layer contains six elements stresses were also evaluted for the basic configuration
along its thickness (0.1 mm). For comparison, mainly to of the single lap joint. Fig. 37 shows the behavior of
assess whether the level of mesh refinement was enough the peel and shear stresses in the continuous single-lap
to give correct results, also the simplified solution of joints. The stress fields were shown on two levels: (a)
Bigwood and Crocombe [97] has been calculated in the interface between steel and adhesive, (b) the inter-
this case. In order to accomplish this, Belingardi et al. face between adhesive and composite. In general terms,
[61] have used the following two equations: the stress field is affected by the typical peaks at the ends
of the overlap zone, both for the peel and for the shear
d4 σy stresses. On each interface (adhesive-steel, adhesive-
+ 4K 54 σy = 0 (39) composite) the highest peak occurs at the overlap end
dx 4
corresponding to the loaded adherend. This is particu-
d3 τxy larly true for the peel stress.
− K 62 τxy = 0 (40)
dx 3 Starting from the basic configuration, both solutions
in which a fillet made of adhesive covers the end of
where σyy and τxy are the peel (i.e., σa ) and shear stress the adhesive layer only (see Fig. 12f), and solutions in
(i.e., τa ) components, K 5 and K 6 are constants related which a spew made of adhesive covers the end of the
to the material properties and to the thickness values. adhesive layer and of the adherend (see Fig. 12d) have

Figure 37 Basic configuration [61]: peel and shear stresses in the joint and comparison with the simplified theoretical solution.

4767
 (a)

(b)

Figure 38 Effects of spew on the peel and shear stresses along the adhesive-composite interface [61]: (a) peel stress and (b) shear stress.

been analyzed similarly to [157, 159, 198]. Belingardi than for the shear component, being about five times
et al. [61] first focused on the spew case, to identify the for the peel stress and two times for the shear stress
spew angle α that minimizes the stress peaks. (see Fig. 38). As a synthetic rule, it can be concluded
Fig. 38a and b show the effects of spew on the peel that for α equal to 45◦ the stress is well reduced and the
and shear stresses along the adhesive-composite inter- additional peaks at the ends of the spew do not exceed
face, As seen in this figure, the effect of the adhesive those at the ends of the overlap.
spew on the stress field can be appreciated. In gen- In order to further reduce the stress peaks, an im-
eral terms, the amplitude of the stresses with α under provement that has been tested in addition to the spew,
high values of α the additional stress peaks that occur Belingardi et al. [61] have also considered the cham-
at the ends of the spews can even exceed the peaks fering in the the unloaded ends of the adherends on the
occurring at the ends of the overlap, when α is lower inner side, as shown in Fig. 12e. This idea, yet consid-
such additional stress peaks tends to disappear. Never- ered in [157], was here further developed by Belingardi
theless, the dependence of the stress on the spew an- et al. [61] to find the best configuration. For the sake
gle α is fairly complicated [61]; Belingardi et al. [61] of simplicity, the same angle value was adopted both
have summarized the effects of spew angle α on the for the spew and for the chamfer, the latter concerns
peel and shear stresses: (a) considering the peel com- two-thirds of the thickness, whilst the remaining third
ponent on the adhesive-steel interface, the highest stress is square. Fig. 39a and b show the effects of the chamfer
peaks always appear (except for the case 15◦ ) at the left and of the spew on the peel and shear stresses along the
overlap end or at the left spew end, i.e., at the loaded adhesive-composite interface, respectively. As seen in
end, as in the basic configuration (see Fig. 37 for com- these figures, each peak is shifted inwards, at the lo-
parison); b) conversely, for the peel component on the cation where the chamfer starts. For the shear compo-
adhesive-composite interface the highest stress peaks nent on the adhesive-composite interface (see Fig. 39b)
(see Fig. 38a) usually appear (except for the case 15◦ ) the highest peaks always appear at the right spew end
at the unloaded end, that is, again at the left end; only (α high) or right overlap (α low). The highest peaks
in the case 75◦ the right spew end causes the highest in peel component appear at the left spew end on the
peak; (c) as far as the shear component is concerned, on adhesive-steel interface [61], and at the right spew end
three interfaces the highest peaks appear always at the on the adhesive-composite interface (see Fig. 39b), that
right end. It can be noticed [126] that, with respect to is, they are always at the loaded end (for low-angle val-
the basic configuration (see Fig. 37), the stress reduc- ues the peaks tend to be equal and very small). With
tion due to the spew is more pronounced for the peel respect to the case of spew only it can be concluded

4768
 (a)

(b)

Figure 39 Effect of the chamfer and of the spew on the peel and shear stresses along the Adhesive-composite interface [61]: (a) peel stress and (b)
shear stress.

[61] that: (a) the stresses are, in general, lower; (b) were evaluated for the various normalized stresses in-
when the angle α assumes values below 45◦ the peaks cluding σxx , σyy , σzz , τxy , τxz , and τyz . Fig. 40a depicts
of the peel stress occurring at the ends of the spews tend the stress distributions for the normalized stresses σxx ,
to zero and those at the ends of the overlap even dis- τxy , τxz , where each graph in this figure represents the
appear. The latter result is of considerable relevance, longitudinal distribution of a normalized stress com-
since these stress peaks are the most harmful for the ponent along the overlap length in several longitudi-
joint. nal planes. The average shear stress at the adhesive
Gonçalves et al. [92] have investigated a new model (τav ) was used to normalize the stress components. In
for 3-D FEA of adhesive joints considering geomet- Fig. 40a, the σxx normal stress component is plotted. Its
ric and material nonlinearities and used solid brick ele- maximum value occurs close to the ends of the overlap,
ments as well as specially developed interface elements at the longitudinal plane that is nearest to the symmetry
using a single-lap joint (see Section 6.1.2 for the de- plane (z = 0.176 mm). The σxx distribution is almost
scription of this new FEA method). They used the fol- constant across the joint width but decreases signifi-
lowing two different analyses approches for both the cantly near the edge (z = 12.324 mm). That effect is
adhesive and adherends in this model: (a) linear elas- also verified for the σyy , and σzz components. The σyy
tic, and (b) elasto-plastic material models. Geometric stresses at the ends of the overlap and in the plane closer
nonlinearity was considered in all analyses. Gonçalves to the longididunal symmetry plane are higher than the
et al. [199, 200] have previously developd an interface other stress components. The σyy , and σzz stresses are
FEA model for 3-D analysis. It consists of 18 nodes dis- of the same order of magnitude. Fig. 40b shows the
tributed in two faces. Contrary to continuum elements distribution of the shear stress τxy . This stress distribu-
where stress-strain relationships are used, interface el- tion does not change significantly across the joint width,
ements are governed by stress-relative displacements which means that this stress component is less sensitive
relationships [92]. The main objective of this work was to the 3-D effects. Whereas the stress τxz is zero at the
to calculate the stresses at the interfaces between ad- longitudinal symmetry plane. For other planes, there
herends and adhesive, which are considered critical re- are peaks near the ends of the overlap. These peaks are
gions in these structures [92]. increasingly higher for longitudinal planes closer to the
Using the linear-elastic materials model, the stress joint edges. The stress τyz is also zero at the longitudi-
distributions at the middle of the adhesive thickness nal symmetry plane and the peak increases with the z

4769
 (a)
(a)

(b)

Figure 40 The longitudinal distribution of a normalized stress compo- (b)

nent along the overlap length in several longitudinal planes at the middle
of the adhesive thickness using the linear elastic material [92]: (a) nor- Figure 41 Distribution of normalized stresses at the interface using the
malized stress (σxx ) and (b) Normalized stress τxy . linear-elastic material [92]: (a) normalized stress σyy (y = 0.125 mm)
and (b) normalized stress τyz (y = 0.125 mm).

coordinate. It should be noted that the 3-D effect for the materials’ behavior.The von Mises yield criterion and
two last stress components (i.e., τxz and τyz ) is present Raghava et al. [201] parabolic criteria were used for the
across all the joint’s width. This is contrary to what adherends and adhesive, respectively. In the Raghava
happens with the normal stress components where the et al. criterion a ratio of 1.3 between the compression
3-D effect is only present near the edges. and tensile strengths was assumed. From this analysis,
Gonçalves et al. [92] have also studied the stresses at the maximum normalized stresses σyy and τxy at the up-
the interfaces for linear elastic materials (see Fig. 41a per interface (y = 0.125 mm) for different load levels
and b). It should be noted that the stresses were plot- were determined, and are presented in Fig. 42. It can be
ted along the longitudinal planes that are different from stated that the peak normalized stresses decrease as the
the ones in Fig. 40. As can be seen in Fig. 41 the peak
stresses at the interfaces are much higher than at the
middle of the adhesive layer (see Fig. 40). This ex-
plains why the interfaces are critical regions regarding
adhesive joints failure [92]. As observed for the normal
stress components in the middle of the adhesive layer,
the normal stress σyy at the interface is almost constant
across the joint width, but decreases near the edges. In
Fig. 41a it can be seen that the maximum stress σyy oc-
curs at the end of the overlap and at the symmetry plane
(z = 0 mm). The shear stress τxy has a similar behav-
ior. It must be noted that these two stress components
have peak values that are much higher than the ones in
the middle of the adhesive layer. The shear stress τyz
at the interface (see Fig. 41b) and in the middle of the
adhesive layers have similar distributions.
Gonçalves et al. [92] have also performed a material Figure 42 Change of the maximum values of normalized stresses σyy
and geometric nonlinear FEA using the elasto-plastic and τxy at the interface using the elasto-plastic material behavior [92].

4770
applied load increases. Also, for the higher load levels and shear stresses are influenced by the modulus of
the peak normalized stresses shift from the symmetry the bonding adhesive-the stiffer adhesive, the higher
plane to the free edge. the stresses; (d) the analytical solution for peel stresses
agrees quite well with FEM simulations but is higher
for shear stresses; (e) the main effect of having fillets
8. Mechanical joint analysis using is the decrease in the adhesive maximum peel stresses
the experimental methods while the maximum shear stresses slightly increase.
8.1. Peel and shear stresses, yielding Wang and Chalkley [14] have studied experimen-
behavior and bending under static tally the plastic yielding behavior of a rubber-toughened
loading film adhesive (FM73) (i.e., widely employed in the
Li and Lee-Sullivan [46] have studied adhesive stresses aerospace industries) subjected to triaxial stresses. The
(i.e., peel and shear) and compared them with the FEA adherends used were made of Aluminum alloy 2024-
results. The adhesive peel and shear stress distributions T3. The specimen used to evaluate the yield behav-
along the centerline of the bond layer using Equation ior in the presence of combined stresses is the modi-
1 in Section 1.2 are shown in Fig. 43a and b for ten- fied Iosipescu specimen [202, 203]. Note that Iosipescu
sile loads up to 5 kN for the filleted and unfilleted rigid specimens were originally designed for use with fiber
joints, respectively. Li and Lee-Sullivan [46] concluded composite materials. Experiments were conducted un-
from these plots that: (a) for the same boundary con- der various combination of tension, compression and
ditions, plane strain and plane stress FEM simulations shear-loading to determine the yield locus of the film
give quite similar adhesive stress didtributions; (b) ad- adhesive over the full range of hydrostatic stress.
hesive stresses are not significantly affected by the na- The Iosipescu specimen was modified by Broughton
ture of the adherend end boundary conditions, i.e., ro- [202] to allow a compressive load Pβ to be applied
tation or no-rotation boundary conditions; (c) the peel at an angle β to the longitudinal axis of the bond-
line. A biaxial stress state subsequently develops in
the adhesive bondline with the stresses σx and τxy be-
ing given by, referring to Fig. 44 for coordinate and
notations.
−Pβ sin(β)
σx = (41)
Sb
−Pβα cos(β)
τxy = (42)
Sb

where Pβ is the compressive load, Sb denotes the bond-

line area, σx is the through-thickness tensile stress and
τxy the shear stress.
Using the experimental results, the applicability of
several yield criteria (i.e., the modified von Mises
yield criterion [15, 16], the modified Tresca yield cri-
(a) terion [15, 16] and the Druker-Prager criterion [16,
17]) commonly employed to characterize neat adhe-
sives was first critically assessed. However, according
to Wang and Chalkley [14], these three yield criteria
suffered the same deficiency: Over-predicting the ben-
eficial effect of compressive hydrostatic stress. To over-
come this difficulty, Wang and Chalkley have adopted
the following modified Drucker-Prager/Cap plasticity
model [14, 194], developed mainly for determining the

(b)

Figure 43 The variation of adhesive stresses along the bond centerline

when the applied load is 5 kN for [46]. (a) The unfilleted rigid joint: (
):
Oplinger, (): plane-strain FEM—No-rotation, (): plane-stress—No-
rotation, (◦): plane strain FEM—Rotation, and (∇): plane stress FEM—
Rotation. (b) The filleted ridid joint: (
): Plane strain FEM—Rotation,
(): Plane stress FEM—Rotation, (◦): Plane strain FEM—No-rotation, Figure 44 The modified Iosipescu specimen under combined loading
and (∇): Plane stress FEM—No-rotation. [14].

4771
pressure dependent yield failure of cohesive geological and  can be expressed in terms of the ratio of uniaxial
materials, to model the pressure-sensitive yielding be- compressive yield stress to uniaxial tensile yield stress
havior of adhesives. There are three parts to this model: ϑ(ϑ = σeo /σio , see Chiang and Chai [17],
(a) The first part uses the Drucker-Prager model [14, 16,
17] to describe the yield surface corresponding to pre- 2+ϑ
dominantly shearing behavior. (b) The second part is a Y = (53)
2ϑ + 1
transition yield surface that has a constant radius in the
meridonial plane, ensuring the continuity of the overall 3(ϑ − 1)
tan  = (54)
yield locus. (c) The third part is a “cap” yield surface ϑ +2
which has an elliptical shape with constant eccentricity
in the meridonial plane and also includes dependence In the present case, since ϑ ≈ 1.64, we have Y = 0.85
on the third stress invariant, through the deviatoric stress and  = 27.8◦ . This friction angle β is significantly
σd in the deviatoric plane. Therefore, the yield surface lower than what the experimental results suggest [14].
consists of three surfaces given by This is primarily due to the error in assuming the yield
points corresponding to the uniaxial tension and uni-
Fs = 0 (43) axial compression would lie on the same straight line.
The present experimental results clearly demonstrate
Ft = 0 (44)
that this is not the case for the structural adhesive being
Fe = 0 (45) considered [14].
The parameters β and  control respectively the ra-
where Fs , Ft and Fe are dius of the transition yield surface and the cap eccentric-
ity. Generally parameter β is a small number, with val-
Fs = σd − p tan  − σc = 0 (46) ues typically between 0.01, and 0.05, where the value
 of  must be greater than 0 and is generally less than
" #2
cos  − a 1.0. The transition surface is a less important segment
Ft = ( p − pa ) + t −
2 (σc + pa tan ) of the yield locus. It was developed in order to create
cos 
a smooth transition between the shear failure surface
− β[σc + pa tan()] (47) and the cap surface, but is not significant in the initial

" #2 stages of fitting the yield function to the experimental
σd
Fe = ( p − pa ) +
2 results, as seen in Fig. 45.
1 + β − β/ cos() As can be seen from Equations 46, 47 and 48 there
− [σc + pa tan()] (48) are total of six material constants that need to be identi-
fied to fully specify the entire yield envelope: Y , , σc ,
where β, , pa . In the present study, the parameter Y is set to
be 0.778. Then by inspecting the shape of the yield con-
"

3 # tour, the parameters  and σc can be obtained by curve
q 1 1 r
σd = 1+ − 1− (49) fitting the experimental data corresponding negative hy-
2 Y Y q drostatic pressure [14]. The remaining three constants
$
q = 3J2 (50) can be determined by curve fitting the experimental data
27
r3 =
2
(2σ1 + σ2 + σ3 )(2σ2 + σ3 + σ1 )(2σ3 + σ1 + σ2 )
=
2
(51)

where σ1 and σ3 are the maximum and minimum

principal stresses.


1
σc = 1 − tan  σeo (52)
3

Here q is the von Mises equivalent stress and r is the

third invariant of the deviatoric stress. The use of the
deviatoric stress measure σd is to allow the model to
match different yield-stress values in tension and com-
pression in the deviatoric plane. The constant  is the
material angle of friction, σc is the material cohesion
stress, and the parameter Y controls the shape of the
yield surface in the deviatoric plane. The value of Y is
equal to the ratio of the flow stress in triaxial tension to
the flow stress in triaxial compression. For example, Y Figure 45 Modified Drucker-Prager curve fit [14].

4772
 ence probably being that of Gent and Petrich [43], later
examples being given in [51, 207–209].
Ansarifar et al. [48] have studied the bonding prop-
erties of rubber to steel, aluminum and nylon 6,6 sub-
strates by means of peel tests. Fig. 4a shows the peel
test piece (i.e., θ < 90◦ ) used to assess rubber-to-metal
bond failure. This new peel test method was described
in Section 4.2. Peel tests were performed at an angle of
90◦ at 23◦ C, either at constant rate of grip separation
Figure 46 Peel geometry in which the peel arm corresponds to (a nominal peel rate of 0.05 to ∼16.7 mm/s), when the
HDPE/PE* with separation occurring at the EVOH/PE* interface [36]. peeling force was recorded as a function of cross-head
separation (see Fig. 5) or under constant load, where the
corresponding to positive hydrostatic pressure using the load applied to the peeling leg increased from 216 N
least-squares method. to 483 N. The actual peel rate was was determined
Amongst modern polymeric materials, the use of by recording the distance the peel advanced along the
multi-layer structures is becoming popular. Guiu and bond and dividing it with the time it took. In Fig. 5, the
Shanahan [36] have studied the adhesion between two first peaks corresponding to the onset of crack growth,
semi-crystalline polymers (i.e., EVOH and PE) us- where the peeling force was still rising, and the last
ing peel tests at various peel angles. This study deals peak corresponding to when test stopped or leg break
specifically with a five-layer system composed of two occurred, were ignored . The remaining peaks on the
outer strata of high-density polyethylene (HDPE) with trace were utilized for calculating an average peeling
an inner layer of an ethylene/vinyl alcohol copolymer force for the test piece (see Fig. 5).
(EVOH) (see Fig. 46). The further two layers consist of After the peel experiments were completed, force
an “adhesive” between the HDPE and EVOH phases: a values were placed in Equation 31 to calculate peel-
grafted polyethylene (PE). Both temperature and peel ing energies for the test pieces. From constant rate peel
speed have been used as experimental variables. tests, the nominal peel rate was calculated from the
Both so-called “L-peel” and “T-peel” tests have been cros-head speed and the test-piece geometry. The peel-
used in this investigation, but they studied the “L-peel” ing rate variation as a function of the peeling energy was
tests in details. As shown in Fig. 46, seperation is presented in Fig. 6. The peeling energies measured for
initiated by a blade at the EVOH/PE interface and a rubber/nylon combination (∼24 kJ/m2 ), were notice-
peel arm consisting of HDPE/PE* is peeled away from ably higher than those calculated for the rubber/steel
the remaining HDPE/PE*/EVOH structure which re- and rubber/aluminum combinations (∼13 kJ/m2 ). Ev-
mains essentially flat, being (adhesively) attached to idently, the bond strength of rubber to steel and alu-
aluminum backing plate (not shown in Fig. 46. Peel minum was almost identical, and noticeably lower
angle, θ , is variable; the term “L-peel” referes to the than the values measured for the rubber/nylon joint
observed configuration for θ = 90◦ . failure.
Using Equation 3 the overall peel energy was cal- Ansarifar et al. [48] used a second alternative pro-
culted at temperatures in the range of 0–70◦ C and for cedure for producing bond failure carried out on 100
peel speeds in the range of 0.1–100 mm min−1 as repre- mm long test pieces involved a newly developed con-
sented by G p versus log (vp ) (see Fig. 8). Various points stant load peel test method [110, 141]. In these exper-
are worth noting [192]: (a) Peel energy, G p , increases iments, load was applied to the peeled leg, and bond
with peel rate, vp , irrespective of test temperature, (b) failure proceededing a time-dependent manner at or
Two regimes of behavior are present. For temperatures near the interface. The locus of failure was almost par-
above 21◦ C, G p , is relatively low, of the order of 5 allel to the bond, but visual examination of the peel
kJm−2 (although this is still a high value for peel en- front zone showed large cavities to be present over
ergy) whereas for lower temperatures, larger values of an extensive region with unbroken strands of rubber
G p are obtained, typically of the order of 20 kJm−2 . (c) in between. The presence of cavities was attributed
At 21◦ C, a transition between the two types of behavior to dilatational stresses created in the peel front [48].
can be observed where a jump in peel energy occurs in Moreover, the fracture surfaces produced in these tests
the vicinity of v = 10 mm/min. This jump is of the were either rough, where slow time-dependent failure
order of 10 kJm−2 . For the higher energy domain, mi- along the bond was recorded at rates approximately
croscopic observation of the peeled arm revealed the 10−6 –10−3 mm/s, or smooth, where peel propagated
existence of small crazelike regions at intervals of ca. along the bond likewise in a time-dependent manner
200 µm in PE* [204]. It was suggested [205] that these at rates reaching 240 mm/s (see Fig. 47). Slope of
highly damaged polymer regions require large energy the best fit to the results in Fig. 47 showed a power
expenditure, thus explaining the considerable jump in law index of about 25. It is interesting that rapid time-
peel energy. Such crazing behavior was previously re- dependent failure along the bond was present in all the
ported by Reynolds in polyethelene [206] and in amor- constant load peel tests, irrespective of the substrates
phous thermoplastics by Cho et al. [37]. Some sort of tested.
stick-slip fracture behavior seems likely. Similar tran- Silva and Adams [74] have studied experimentally
sitions in various different systems have been reported the failure mode of loads of a variety of 90◦ T-joints
over the long history of the peel test, the first refer- made of mild steel/adhesive/mild steel material for a

4773
 (a)

Figure 47 Time-dependent peel rate versus peeling energy relationship

from constant peel tests [48]: (
): Nylon, (•): Steel, (◦): Aluminum,
and ( ): Aluminum (Ra ∼ 1 µm).

range of reinforcing methods, and ascertain the main

parameters that influence the strength. In this study, a (b)
standard toughened epoxy paste adhesive was used. The
mild steel adherend used was typical of that used in the Figure 48 T -joints designs and failure loads (74): (a) Schematic dia-
gram of T -joint test. Where P is the loading in the plane of the sheet, Pp
manufacture of a car bodyshell. The geometry of the is the tensile transverse (or peel) loading, and ηs is the adherend thick-
joints is illustrated in Fig. 48a. The supports consists ness. All dimensions in mm and (b) T-joint designs and failure loads for
of 90◦ gussets. The joints were loaded by the force P 10 an 1.5 mm base thickness.
as shown in Fig. 48a, and supported by clamping some
distance away from the joint area. Fig. 48b shows the
failure loads of a series of T-joints on a 1.5 mm base. Then, for the complete beam, the bending moment is
Joint (a) failed at the base of the vertical member by given by
peeling of the adhesive. Joint (b) is a similar design but
with a 90◦ gusset. The thickened vertical portion re- P
duced the rotation and hence reduced the peel stress in Mx = M − x (56)
2
the adhesive; this increased the joint strength consider-
ably. Further improvement was obtained by combining For a T-joint without support, the critical region is the
(a) with a 90◦ gusset to give (c), while the strongest middle of the beam where the bending moment is the
joint used two gussets (d). In all of these joints studied, highest. For a T-joint with supports, such as the gussets
yielding was observed in the steel base sheet, creat- shown in Fig. 48b, the middle part of the joint where
ing a plastic hinge. The designs that used two supports the supports are located is not critical anymore since
gussets reduced the bending moment and increased the the stresses are reduced by a factor of 8 (I2 /I1 = 8).
failure load. For a simple beam fixed at both ends, the The transition between the base and the base plus the
bending moment was calculated from tabulated formu- support is the critical part. The joint failure is dictated
lae [210]. The maximum bending moment is Pl/4 at by yielding of the adherend base plate at this point. The
the middle and at the ends. Here P is the force and bending moment is reduced in comparison with a T-
l is the half joint length. For the case of design (d), joint without support and the joint will fail at a higher
the bending moment is more complicated because the load.
beam has a variable second moment of area along the Silva and Adams [74] have applied a FEA method
span (moment of inertia I1 and I3 ). with a plane-strain condition to understand the plas-
tic behavior of the T-joint. In addition, the non-linear
" # geometry was accounted for. The steel was allowed to
P l − (1 − n)a 2 Pl 25
M= = (55) yield with subsequent hardening and the adhesive was
4 l − (1 − n)a 4 18 linear elastic. From the FEA results, the magnitude of
the plastic deformation of the steel was determined with
where n = I2 /I1 = 1/8 and l = 2a = 25 mm. increasing load. The base starts to yield at 200 N near
4774
the clamping zone. The plastic zone spreads through-
out the section and eventually forms a hinge. At 575 N,
the base yields near the support and an additional hinge
is formed when the section is fully plastic at 1000 N.
The experimental failure load for this configuration is
approximately 1000 N, as seen in Fig. 48b. The failure
prediction is probably dictated by yielding of the steel.
Thus, the failure prediction should be around the region
where the adhesive actually is. This would explain why
Figure 50 Rectangular area used in the averaging of FEM results ob-
the joint fails at a higher load. tained for the three nodes a, b, and c. The nodes represent the centers of
Li and Lee-Sullivan [46] have investigated the fil- the triaxial strain gages [46].
leted single-lap joints bonded (tested in tension) using
a 2-D geometrically nonlinear FEA method under both
plane strain and plane stress conditions, respectively,
and verified either experimentally or theoretically. The with the strains mesured off the rectangular rosette
6061–T651 aluminum adherends were bonded using a gages bonded along overlap edge, an averaging tech-
flexible epoxy resin. The corners of the overlap were nique was applied. A rectangular area with dimen-
milled to a radius r = 3.18 mm to obtain a consistent sions 4.62 × 1.58 mm was considered in Fig. 50
fillet shape.The aluminum-to-aluminum joint specimen in which the finite element strain at the centerline
was strain gaged in two ways. In this study, gage strains was obtained by averaging the strains at the cen-
(at 135◦ , 90◦ , 45◦ ), bending moment (k), peel and shear ter locations of each gage. The raw FEM data were
stresses of adhesives along the bondline were deter- transformed into their appropriate orientations before
mined. The first set of strain gages was mounted along averaging.
the overlap region of the adherend very close to ad- Accordingly, ε135◦ , ε90◦ , and ε45◦
hesive bondline. These triaxial resettes were mounted
on the specimen at 135, 90 and 45◦ . The second set of ε135◦ |averaged
gages was mounted onto the upper and lower surfaces ε135◦ | at “a” point + ε135◦ | at “b” point + ε135 | at “c” point
of the adherend at about 7.2 mm from the corner of the =
overlap end. The gages were referred to as the longitu- 3
dinal gages since they align with the centroidal axis of (57a)
the adherend. ε90◦ |averaged
The filleted aluminum-to-aluminum single-lap joint
was tested three times and the results averaged. The ε90◦ |at “a” point + ε90◦ |at “b” point + ε90◦ |at “c” point
=
FEM results for longitudinal strains at the upper (US) 3
and lower (LS) adherend surfaces located at 7.2 mm (57b)
from the overlap end are compared with experimental
longitudinal gage results, as shown in Fig. 49. The ε45◦ |averaged
experimental results clearly fall within the range of ε45◦ |at “a” point + ε45◦ |at “b” point + ε45◦ |at “c” point
the results generated by the FEM simulations. In or- =
3
der to compare the strain values predicted by FEM
(57c)

The strain distributions along the overlap region for

the different gage orientations 135, 90 and 45◦ were
compared with the FEM simulations. The averaged
strains from FEM analyses were plotted along with
the maximum and minimum values using the same
markers for each position along the overlap. It can
be concluded [46] that the experimental results for all
three gage directions (135, 90 and 45◦ ) agree well with
FEM simulations along the overlap but there is poor
agreement at the overlap end corners. See for exam-
ple Fig. 51 for the 90◦ gage strains for the experimen-
tal and FEM simulationsin in the overlap section. It
is difficult to determine which of the simulated con-
ditions, e.g., plane strain/stress or rotation/no-rotation,
give the best predictions since the error bars for the
experimental results are relatively large. It is interest-
Figure 49 The variation of longitudinal strains and FEM simulations ing, however, that the plane stress values are always
at the lower (LS) and upper (US) adherend surfaces at 7.2 mm from
the overlap edge [46]. Note that the US and LS denote for the upper
higher than the plane strain values. Also, the averaged
and lower substrates, respectively: (
): Plane strain FEM—No rotation, 90◦ gage results (see Fig. 51) which had the smallest
(•): Plane strain FEM—Rotation, (): Experiment, (): Plane stress error range, lie between the plane stress and plane strain
FEM—No rotation, and (◦): Plane stress FEM—Rotation. results.
4775
 waves could be regarded as waves measured at different
points of an adhesive rod with an infinite length.
The viscoelastic parameters of the cured epoxy resin
adhesive were determined using the Voight models
which has three elements and five elements approaches.
The relations between the complex compliance and the
viscoelastic parametres of Voight model, which has
three elements approach, are as follows:

1 1
J1 (ω) = + 2
Eo E 1 + ω2 φ22
(58)
ω φ1
J2 (ω) = 2
E 1 + ω2 φ12
Figure 51 Comparison between the measured 90◦ gage strains and FEM
simulations in the overlap section when applied load is 5 kN (Ref. [46]):
The relations of Voight model which has five elements
(
): Plane strain FEM—No-rotation, (•): Plane strain—Rotation, (): approach are as follows:
Experimental (average), (): Plane stress FEM—No-rotation, and (◦):
Plane stress FEM—Rotation. 1 E1 E2
J1 (ω) = + 2 + 2
Eo E 1 + ω φ1
2 2
E 2 + ω2 φ22
(59)
ω φ1 ωφ2
8.2. Dynamic deformation of adhesives and J2 (ω) = 2 + 2
joints under impact (or dynamic) E 1 + ω2 φ12 E 2 + ω2 φ22
loading
Since many structural adhesives are used in transport J1 and J2 indicate real and imaginary parts of com-
applications where the bonded joints are subject to plex compliance. ω indicates angular frequency. The
dynamic as well as static loading, the adhesive dy- real part of the complex compliance of the adhesive de-
namic properties (i.e., dynamic stress and strain, dy- creases with respect to the frequency of the load spec-
namic modulus) can be of significance. Although struc- trum. The imginary part of the complex compliance
tural adhesive test specimens such as the lap-shear and increases with respect to the frequency. Here, E o , E 1
TAST (thick-adherend shear test) specimens are almost and E 2 indicate elastic constants of spring elements, φ1
always used for testing under static conditions, the ad- and φ2 indicate viscosities of dashpot elements in the
hesive material can show a degree of viscoelastic be- Voight model.
havior such that its mechanical properties under dy- The frequency spectrum of strain variations (i.e.,
namic loading can be significantly different from those ε̄(x, ω)) can be obtained
under static loading [65]. Therefore, the durability for "
 #
dynamic loading is the important subject in designing ω
ε̄(x, ω) = ε̄(0, ω) exp −rd + J x (60)
the adhesive joints. vs
Sato and Ikegami [63] have experimentally investi-
gated the propagation of stress waves and the concen- where x indicates position at which the strain measured,
tration of dynamic stress under tensile impact loading rd and vs indicate the decrement ratio, and velocity of
(i.e., dynamic deformation) in single lap joints, tapered the stress wave, respectively.
lap joints and scarf joints, which were bonded adhe- The calculated results of the model with five elements
sively. The joints consisted of aluminum alloy adherend showed good agreement with the complex compliance
and an epoxy resin adhesive. Viscoelastic properties of obtained experimentally. The calculated results of the
epoxy adhesive resin were also obtained from variation model with three elements deviated from the experi-
of strain waves which propagated in a rod of cured resin mental results. Thus, Sato and Ikegami [63] concluded
specimen. A compressive stress wave was generated by that the viscoelastic model of the five elements was
the collision, and propagated in the specimen. The com- sufficient to describe the viscoelastic properties of the
pressive stress wave was reflected at the other end of cured adhesive.
specimen, and changed to a tensile wave. This reflec- Sato and Ikegami [63] have also carried out impact
tion of the stress wave was repeated at both the edges tests to verify the analytical method described in the
of the specimen. Thus, the stress wave were propagated Section 7.1. The length of the adherends was made
in the specimen repeatedly. The variation of the stress to be 1300 mm to prolong the period until the stress
waves was measured using strain gages adherend on waves, which were reflected at the end of specimen,
the surface of the specimen. The stress distribution and eached the bonded part. The results for the strain vari-
the time variation of stress and strain in the joints under ations with respect to time are shown in Fig. 52a–c
tensile impact loading were calculated using the finite for the single-lap, tapered lap, and scarf joints, respec-
element method, considering the viscoelastic proper- tively. The strain variation on the single lap joint and
ties of the adhesive. In this study, 2-D analysis was per- the tapered lap joints exhibited complicated changes
formed assuming in-plane strain conditions. The strain because of the bending deformation. However, in the
wave propagation in the axial direction of epoxy adhe- scarf joint, the smooth propagation of the stress wave
sive specimen in the experiment were mesured. These occurred because the bending moment could be gnored.
4776
 smaller than that of the single lap joints with the same
lap length. Thus, the use of tapered lap joints instead of
single lap joint is effective in reducing the stress con-
centration not only under, static loading but also under
dynamic loading. Scarf joint is the most effective shape
to reduce the stress concentration in the adhesive layer
not only under static loading but also under dynamic
loading.
Maheri and Adams [65] have investigated experi-
mentally the dynamic shear modulus of three struc-
(a)
tural adhesives (i.e., Araldite AV119, Araldite A420,
Hysol EA956; all epoxy-based, two-part adhesives, ex-
cept AV119) using the the thick-adherend shear test
(TAST) specimens. In this study they used dynamic
models of mass-spring and standing waves. Using the
mass-spring model they determined the following ex-
pressions for dynamic shear modulus of adhesive layer
for the axial and wide-wise lateral vibrations:
the axial vibration of the mass-spring model:

2π 2 f 2 ηa m s
Ga = (61)
Sa
(b)
the wide-wise lateral vibration:

2π 2 f 2 Jo ηa
Ga = (62)
JA
where f is the resonance frequency, ηa is the bondline
thickness, ms is the mass of one adherend, Sa is the
surface area of the adhesive layer, Jo is the polar mass
moment of inertia about its center o, and JA is the polar
second moment of adhesive layer.
The dynamic test results on bulk adhesives are tab-
(c)
ulated in Table I. Also shown in this table, are typi-
Figure 52 Comparison between experimental and the calculated (i.e., cal values of shear modulus of each adhesives type,
FEM) predictions of strain waves [63] for: (a) the single lap joint (over as obtained by using both axial and width-wise lateral
lap length = 100 mm), (b) the tapered lap joint (over lap length = 100 dynamic tests. For each adhesive type, the resonance
mm; taper length = 50 mm), and (c) the scarf joint (lap length = 100
frequencies f of the same TAST sample was used to
mm). Note that calculated predictions were made for the three and five
elements in Voight model, respectively [(i.e., cal. (1), cal. (2)]. predict the shear modulus according to different vibra-
tion models. Nominal shear modulus values have also
been tabulated for comparison. From Table I, it is clear
In these figures, the experimental results are compared that both the axial and the width-wise flexural results
with calculated results by the analytical method (i.e., fall well short of the value of shear modulus found from
FEA) in the described above. The analytical results of tests on the bulk adhesive. The bulk adhesive dynamic
the tapered lap joint and the scarf joint showed good test results, on the other hand, are seen to correlate well
agreement with the experimental results. The analyti- with the expected (nominal) values, and can be assumed
cal results of the single lap joint almost agreed with the to represent the correct value of the adhesive dynamic
experimental results too. Therefore, the actual strain shear modulus. Also, if there were to be any viscoelastic
and stress in the joints seemed to be predicted by using effect, then the static tests should give a lower modulus
the analytical method. However, in the case of the sin- [65]. It is therefore appeared that, for a standard TAST
gle lap joint as shown in Fig. 52a, the analytical results specimen where the bondline thickness is of the order
at the point [2] on the adherend of the stress-input side 0.5 mm, the implicit assumption in the vibratory mod-
of deviated from the experimental results. The reason els concerning the relative dynamic compliance of the
seemed to be due to the lack of uniformity in the adhe- adherend was not valid. The shear stiffness ka of a block
sive thickness [63]. The joints had long adherends, so of material of area, Sa , thickness ηa and shear modulus
it was not easy to bond the adherends with geometrical G a [65] is
precision. Therefore, in the joints, the thickness of the
adhesive layer might have variations, which caused the Sa G a
deviation between the analytical results and the exper- ka = (63)
ηa
imental result.
As Sato and Ikegami [63] summarized, the stress oc- For a typical adhesive, the shear modulus is of the or-
curring in the adhesive layers of the tapered lap joint is der of 1 GPa, while that of the steel adherend is of the
4777
T A B L E I Results of dynamic tests on bulk adhesives and TAST specimens, using three different adhesives [65]

Computed shear modulus (GPa)

Axial mode Lateral mode

Nominal Bulk dynamic Bondline Axial Lateral Mass- Stand Mass- Stand
shear modulus shear modulus thickness freq. freq. spring wave spring wave
Adhesive (GPa) (GPa) (mm) (Hz) (Hz) model model model model

AV119 1.1a 1.14 0.460 8820 3980 0.656 0.830 0.780 0.795
A420 0.65 0.443 7450 2987 0.450 0.549 0.423 0.403
EA956 0.91b 1.00 0.550 8330 3452 0.699 0.873 0.701 0.690

a Manufacturer (Ciba-Geigy) data sheet.

b Determined from the quoted value of extensional modulus in the manufacturer’s (Dexter Aerospace) sheet using a Poisson’s ratio of 0.38.

ness of the adhesive layer, is the only variable factor in

a TAST specimen of standard dimensions. The follow-
ing two expressions are cubic polynomial based on the
experimental points shown in Fig. 53.

Axial mode:
Ga
= 4.8222 × 10−11 × f 3 − 7.3454 × 10−7 × f 2
ηa
+3.0045 × 10−3 × f (64)

Wide-wide Lateral Mode:

Ga
= 2.7166 × 10−10 × f 3 − 1.7967 × 10−6 × f 2
Figure 53 Resonance frequency f as a function of the adhesive layer ηa
parameter G a /ηa (Ref. [65]).
+ 3.4513 × 10−3 × f (65)

order of 82 GPa. In the TAST joint, the adhesive and in which the units for G a , ηa and f are (GPa), (mm) and
adherend (steps) thickness are 0.5 and 5.75 mm respec- (Hz) respectively.These empirical expressions give the
tively. Fig. 53 shows the variation of the square of reso- dynamic shear modulus of the adhesive in a standard
nance frequency f with G a /ηa for both axial as well as TAST specimen from the measured thickness of the
width-wise lateral modes of vibration for AV119 adhe- adhesive layer, and the resonance frequency of either
sive (earlier test results on standard TAST specimens the width-wise lateral, or axial modes of vibration, pro-
with AV119 were also plotted). The points are based vided that the same specimen configuration used in the
on the tests carried out on individual TAST specimens present study and the same adherend material (steel) is
and show, according to Equations 61 and 62, the varia- used. The equations were applied to the test data listed
tion of the square of resonance frequency with the ratio in Table I, and the results are listed below in Table
G a /ηa in which G a is the true dynamic shear modu- II. The results of dynamic tests on the bulk specimens
lus of the adhesive as found from the bulk specimens are also shown for comparison. Therefore, the results
tests (Table I). The solid lines are simply the graphi- based on the empirical relations correlate reasonably
cal representation of Equations 61 and 62 for axial and well with the bulk adhesive dynamic test results. This
lateral modes of vibration respectively, both of which is to be expected for AV119 as this was used to produce
predict a linear variation of f 2 with G a /ηa . It is noted the original master curves. The “unknown” adhesives
that the ratio G a /ηa , which is proportional to the stiff- A420 and EA956 were then assessed using the master

T A B L E I I Comparison of results based on empirical relations and bulk adhesive test results [65]

Predicted shear
modulus (GPa)

Bulk dynamic Bondline Axial Lateral Axial Lateral

shear modulus thickness frequency frequency mode mode
Adhesive (GPa) (mm) (Hz) (Hz) (Eq. 34) (Eq. 35)

AV119 1.14 0.460 8820 3980 1.12 1.11

A420 0.65 0.443 7450 2987 0.69 0.67
EA956 1.00 0.550 8330 3452 1.06 0.92

4778
curves and their measured natural frequencies. The re-
sults are within 8% of the previously measured bulk
value, and validate the test methodology.
Sato and Ikegami [63] concluded that the use of these
empirical relationships for measuring the shear modu-
lus, G a , of an adhesive material could also be applied
to other TAST configurations. However, it would then
be necessary to establish the master curves appropriate
to the new parameters.

9. Effects of hot/humid environment

on the durability of adhesive joints
A major concern is that the mechanical performance Figure 54 Effect of joint interface on water flux [28].
of adhesive joints involving metallic or nonmetallic ad-
herends may deteriorate upon being exposed to aqueous
environment [32, 211, 212]. It is the interphase of the
joint, i.e., the region adjacent to the interface between the experimental results, which were used to study the
the adherend and the polymeric adhesive, which is suc- durability of adhesively bonded joints aged in hot/wet
ceptible to such attack and on which attention must be environment. Two bonded joints have been considered
focused. Various factors can affect the durability and in this study, namely, single lap joint and butt joint.
performance of adhesive joints under the aging condi- The joints were immersed in water at 60◦ C for up to 60
tions (see for example recent studies in references [22, weeks. The aim of this investigation was to understand
28, 29, 33, 213]). Some of these important factors [29, and explain the degradation in the joint strength due to
213] are: (a) various joint geometric factors, (b) the water uptake in terms of the change mechanical prop-
application of stress, (c) the joint orientation, (d) ad- erties and swelling using FE coupled stress-diffusion
hesives and adherend materials, (e) adhesive/adherend analysis [28]. Both dry and wet specimens were sub-
interfacial zone, (f) environmental effects, (g) the sur- jected to tensile loading and the stress/strain diagrams
face pretreatment method, and (h) elastic modulus of were obtained. Abdel Wahab et al. [28] have investi-
adherends and adhesives. gated the effect of the joint interface experimentally
Water is the substance that presents the greatest dura- on the water uptake using a laminated “diffusion disc”
bility problems to adhesive bond [32]. Water also has consisting of cast adhesive and perforated aluminum
a weakening effect on the adhesives. Water may affect foil (only on the upper side of the disc). Fig. 54 shows a
the adhesive properties by plasticisation and cracking. It comparison between the water uptake in this laminated
may affect the adhesive/adherend interface by displac- discs with that in a disc of bulk adhesive having the
ing the adhesive or by hydrating the metal or metal ox- same exposed adhesive area. It can be seen that during
ide surface of the adherend [213]. Many workers were the early stages, the flux rate in the laminated disc is
devoted to study the durability performance of bonded about 50% higher than that in the bulk epoxy disc. It
joints (see for example, Refs. [29, 33, 83, 214–216]). was found that the water flux rate is higher when the
Wylde and Spelt [83] used open-faced specimens in or- interface is present during the early stages of exposure,
der to accelerate aging in an aluminum/epoxy double as seen in Fig. 54. As the laminated disc has a larger
cantilaver beam. Bistact et al. [214] studied the durabil- diameter, it is possible that the enhanced uptake result
ity of steel/polymer immersed in an aqueous environ- from increased two-dimensional diffusion and not in-
ment (liquids such as water, salt, acid and basic solu- terfacial transport.
tions). Brewis et al. [215] proposed to use cryoblasting Abdel Wahab et al. [28] have analyzed the mois-
as a pretreatment for aluminum in order to enhance ture distribution in the adhesive layer with coupling the
joint strength. Rider and Arnott [216] immersed alu- diffusion-stress simulations using the transient finite el-
minum alloys in boiling water followed by soaking ement diffusion for single lap joint and butt joint. As the
in 1% 3-glycidoxytrimethoxysilane as a part of pre- present finite element analysis is two-dimensional (as-
treatment. They found that the durability of bonding suming constant moisture distribution across the width
was notably improved in comparison with conventional of the joint), it does not consider three-dimensional
pre-treatment. Knox and Cowling [29] investigated the effect. Additionally any moisture transport along the
durability performance of thick adherend steel lap shear interface as assumed to be negligle. Therefore, it was
joints and bulk adhesive using accelerated aging tech- suggested that in practise the moisture diffusion into the
niques. They also demonstrated [33] the benefits from joint may be more rapid [28]. The aim of this analysis
using a surface pre-treatment prior to bonding in order is to investigate modelling of this by the moisture trans-
to enhance the durability performance. port process into both configurations, namely adhesive
disc and laminated disc with holes. It was observed
that the moisture is dissipated under the aluminum foil
9.1. Effects of diffusion and oxidation at the whole edge. The rate of mass uptake per quare
Abdel Wahab et al. [28] have investigated coupled mm (or water flux, microgram/sq. mm/h) was calcu-
stress-diffusion finite element analyses to compare with lated from the finite element results using the following

4779
Figure 55 Comparison of water uptake for experimental and FEA results
[28].

expression:
Figure 56 Peel strength Ap of PE steel and EVAC steel adhesive joints
& % ' versus the contact time t at a constant temperature of 433 K in air at a
%n n Cm
Vi free access of oxygen through the polymer melt layer [22].
(m)i i=1 C∞
ρ
Water flux = i=1
=
Ss t Ss t
(66) polymer joints were prepared in the form of laminate
panels.The samples were then held at a fixed tempera-
where m is the mass uptake, Cm /C∞ is the moisture ture in free air for a definite time (to avoid free access
concentration, V is the volume, Ss is the exposed area, of oxygen through the polymer melt layer). The change
n is the number of adhesive element, ρ is the water den- in peel strengt Ap of adhesive joints of PE or EVAC
sity and t is the time. Using this relation, Abdel Wahab (copolymer of ethylene with vinylacetate) with steel,
et al. [28] have plotted the variation of water flux as a the thickness of a residual adhesive layer ηa on a metal
function of time for both models, as shown in Fig, 55. substrate after a cohesive fracture of the adhesive joints
In this figure, the finite element results were compared were investigated as a function of a contact thermooxi-
to the experimental ones. Considering the FEA results, dation time t under conditions of catalytic contact oxi-
during the early stages, the water flux in the adhesive dation. Fig. 56 shows the variations in the peel strength
disc model calculated from FEA is slightly lower than Ap with the contact time t for PE and EVAC. The char-
that in the laminated disc model due to the difference acteristic of the Ap (t) curves is similar. Both curves pass
in volume between the two models. However, in com- through the A maximum (Ap(max) = Ap (t)|(d Ap /dt = 0) .
paring the FEA data with the experimental results is The greatest growth rate of peel strength for both curves
evident that the simple diffusion model, which ignores corresponds to t = 0 ( Ap = lim dAp /dt)|t→0 ). As was
moisture diffusion through the interface gives a signif- experimentally proven [25], the shape of A(t) function
icantly lower prediction of water flux in the laminated for the PE-steel adhesive joints is determined by the
disc. Thus, observation suggests that the interface diffu- rate ratio of the two main competing contact thermoox-
sion does in fact play an important role in the moisture idation processes which occur in the boundary layer
transport process. and affect its cohesive characteristics. The processes
During oxidation in contact with steel, the iron com- which lead to a growth in strength of the boundary layer
pounds dissolve and penetrate into the bulk of a polyeth- and hence to an increase in Ap (the accumulation of
lene layer [217]. The intensity of dissolution correlates oxygen-containing groups and oxidative cross-linking),
with the rate of contact oxidation. Judging from the predominate before reaching the point of Ap(max) . After
value of an efficient diffusion coefficient of the iron reaching the maximum, the oxidative destruction pre-
compounds, the process of transfer of the iron com- vails.The coincidence of the Ap (t) function for both ad-
pounds was suggested to be diffusion-controlled [22]. hesives may indicate the evidence that the contact ther-
The results of several model experiments show that mooxidation process in the boundary layer are similar.
the most presumable reaction of the iron compounds The peel strength of EVAC adhesive joints grows more
generation is the formation of iron carboxylates as a rapidly (Ap is higher) and reaches much greater values
result of the interaction of iron oxide with the prod- of Ap(max) later than PE. These differences together with
ucts of contact oxidation of the polymer containing car- a much lower ratio of Ap(max) /Ap(max) |t=3.6 ks indicate
boxyl groups [217]. Kalnis and Ozolins [22] have stud- that, in the case of EVAC, the oxidative destruction pro-
ied the contact thermooxidation on the structure and cesses may affect the shape of the Ap (t) function less
characteristics of the boundary layer of the polyolefin than that in the case of PE.
adhesive (i.e., PE and EVAC adhesives) which forms The fracture of adhesive joints of both polymers in
the adhesion bond with steel. Note that the theoreti- the course of peeling is of cohesive character. As ob-
cal background about the contact thermooxidation was served by Kalnins and Ozolins [22], the thickness ηa of
presented in Section 1.2. Samples of adhesive metal- the residual adhesive layers grows with contact time t.
4780
This is an indication of the progress in the contact oxi-
dation processes in the adhesive near the interface. The
chemical changes caused by contact oxidation steadily
penetrate deeper into the bulk of adhesive. The structure
of the residual adhesive layer significantly differs from
the bulk dhesive. Kalnins and Ozolins [22] observed
that the content of a cross-linked part of the polymer
gelfraction), cgel , in the adhesive boundary layer grows,
the content of low-molecular part, clam , increases, and
the molecular weight of the polymer in the layer de-
creases with the contact time. This means that the com-
peting contact oxidtion processes (i.e., cross-linking
and destruction) proceed simultaneously in the adhe-
sive boundry layer. Kalnis and Ozolins [22] have intro-
duced an expression, which can somewhat estimate the
ratio of contribution of two competing contact oxidtion
processes. They suggested that this simple empirical
Figure 58 Ap /η versus the contact time t (Ref. [22]).
quantity K is useful for this purpose.

1 + cgel
K = (67) (Fig. 58) and Ap (Fig. 56) with the contact time is
(1 + clam )[m i=0 /m i ]
quite different. The values of AV increase just at a very
early stage of contact and, after reaching the maximum
where m i is the molecular weight of the dhesive in the
clearly decrease with the contact time t. It was sug-
boundar layer at contact time ti and m i=0 is the ini-
gested [22] that the growth in Ap with contact time oc-
tial molecular weight. In the absence of macromolec-
curs primarily due to the increase in the residual layer
ular changes, K = 1 (cgel , clam = 0, m i=0 = m i ). The
thickness η [22]. The curves K (t) (Fig. 57) and AV (t)
growth in K indicates that the cross-linking processes
(Fig. 58) are quite similar. As observed by Kalnins and
prevail, while the decay in K shows that the oxida-
Ozolins, there is a certain correlation between the val-
tive destruction predominates. The varition of K with
ues of AV and K for an equal contact time t. Hence, the
contact time t is shown in Fig. 57. The value of K di-
peel energy per unit volume of the polymer adhesive
minishes with the contact time due to the increased con-
involved in the deformation and consequent fracture,
tribution of the process of contct oxidation destruction.
AV , really depends only on the strength-deformation
Comparison between Figs 56 and 57 shows that there is
chacaracteristics of the adhesive (σB , E a ). It is of inter-
no expected similarity between K (t) and Ap (t) curves.
est to consider the relation Av = Ap /η in the case where
As suggested by Kalnins and Ozolins [22], probable
η depends not on the contact oxidation but on some
reason for this may be the fact that the peel strength
other factor, for instance, on the peel rate (or cross-
by itself does not probably reflect the state of strength-
head rate). As seen from Table II, the value of η varies
deformation charcteristics of the boundary layer.
significantly with the peel rate. Simultaneously, at the
Using the Kaelble’s approach [57] (see Section 1.2
same cross-head rates, some tensile characteristics of
for detail) Kalnins and Ozolins have calculated the
the adhesive (EVAC) were measured: tensile strength
peel energy characteristic, AV = Ap /η. This energy
σB , elongation at break εB , and specific fracture energy
is independent of the thickness η (see Equation 4 in
AB . Fig. 59 shows a good correlation between AV and
Section 1.1). The character of the variation in AV
AB measured at equal cross-head rates. This means that
AV is a quantity depending on the strength-deformation
characteristics of the adhesive boundary layer, which
breaks cohesively during the peel tests.

9.2. Effects of bulk adhesive, fillet, stress,

surface pretreatment, and primer
Knox and Cowling [29] have studied the durability per-
formance of thick-adherend steel lap joints and the bulk
adhesive (i.e., AV119: a hot cured, one part, epoxy
paste, toughened) using accelerated aging techniques
(30◦ C, 100% relative humidity). In some cases the spew
was removed by filling after cure. Fig. 60 shows the ef-
fects of accelerated aging, fillet and preload on the dura-
bility performance of AV119 bonded steel lap shear
joints, which show a similar trend. The ultimate fail-
ure load decreased as the environmental aging time in-
Figure 57 K versus the contact time t for EVAC steel and PE steel creased, but there is some scatter in the results. This
adhesives [22]. trend is as might be expected because in the past, many
4781
 Figure 61 Accelerated durability performance of AV119 bulk tensile
samples with and without a preload [29]. (The preload is 20% tensile
strength 1800 N.)
Figure 59 Correlation of AV and AB measured at different cross-head
rates (AV , AB measured at a certain rate and AV300 , AB3000 at the greatest
rate of 3000 mm/s) (Ref. [22]).
to the interface. This is a classic indication of en-
vironmental attack [220]. On fracture, water is seen
glistering on the adherend surface “shadows” and this
area soon corrodes after exposure to air. Gledhill and
Kinloch [220] established that substrate corrosion was
not an operative degradation mechanism but rather a
post-operative phenomenon.
The performance of the joints declined with expo-
sure time, therefore some degradation must have oc-
curred. Knox and Cowling [29] suggested that the loss
of strength was largely due to degradation of the adhe-
sive itself since the properties of unstressed bulk adhe-
sive specimens also degraded over the 12 week period
(see Fig. 61). However, the reduction in adhesive bulk
properties over the same total time span were less (20%
decrease in tensile stress and 10% decrease in Young’s
modulus) than in the lap shear joint (28% decrease in
strength) [29]. While it can not be assumed that the bulk
adhesive and the adhesive in the joint are saturated by
Figure 60 Accelerated durability performance of AV119-bonded steel water to the same extent, the durability performance of
lap shear joints showing the effect of a fillet and preload [29]. (The the joint clearly can not be attributed solely to the degra-
preload level is approximately 15% of the initial failure load.) dation of the bulk adhesive. It can also be demonstrated
[221] that the durability performance of this type of
specimen can be improved by using a surface primer.
studies have been carried out showing the decrease of This indicates that the durability performance can be
residual strength of bonded metallic substrates with improved by the modification of the interfacial region
epoxy adhesives after aging, in both natural environ- and suggests that in the current study the same degra-
ments and accelerated aging in wet or humid surround- dation is occurring in this region. Gledhill and Kin-
ings [32, 217–219]. This would indicate that removing loch [220] observations also suggest that losses in joint
the spew fillet from the joint and/or applying a load to a strength are caused by adverse effects of water on the
joint has a detrimental effect on durability performance. interface rather than the bulk adhesive.
Others [32] also have found that a hot wet environment Knox and Cowling [33] have investigated the use
was by far the more hostile environment and that the of a novel experimental technique (see Section 4.2 for
presence of an applied load increased the rate of loss more detail about this noval test) to measure the resid-
of strength. Know and Cowling has also observed [29] ual interfacial strength of an adhesive (i.e., AV119) to
that there is a decrease in the static failure strength over the adherend (i.e., a mild steel) after aging in a “re-
a 12 week time span for all specimen types. The largest alistic” environment (100% relative humidity, 30◦ C)
reduction in durability performance was obtained af- in order to determine the effects of various surface
ter 12 weeks from loaded specimens when the residual pretreatment methods and various silanes. The resid-
static failure strength had fallen to approximately 65% ual adhesive/adherend interfacial strength was quanti-
of the intial value [29]. The ingress of water degrades fied by recording the required force to remove a strip
the interfacial zone and failure occurs at or very close of adhesive from the adherend surface using this test
4782
 for a primer such as SiP, the loss in strength is negligi-
ble over the 12 week exposure period. Similar results
have been found by Fay and Maddison [222], who re-
port that stressed samples showed excellent durability
performance when treated with Accomet C or a silane.
The visual examination of the present aged samples
by Know and Cowling showed [29] that the failure
was similar in visual appearence to the unaged sam-
ple [221] demonstrating good adhesion. The effect of
aging on the failure surface has very little influence.
Thus, the use of a silane primer is to maintain the in-
tegrity of the interfacial zone over the test period in
question [33]. The effectiveness of silane treatments is
well established and it is thought to involve the forma-
tion of primary bonds at the substrate/silane interface.
However, the actual mechanism by which the primer
Figure 62 Accelerated durability performance of bonded steel lap shear improves performance is not entirely understood and
joints showing the effect of an adequate surface pretreatment prior to the
application of a primer (A 187) (Ref. [33]).
whilst silane primers do improve the water stability of
adhesion on metals they do not necessarily increase re-
sistance to bondline corrosion. When a silane is used to
method. The geometry of the specimens enables rapid improve durability it has been demonstrated that they
water ingress into the adhesive and interfacial zone produce polysiloxane coatings on steel [223]. This may
[33]. Fig. 62 shows the effect of various surface pre- lead to a more durable interface. Throughout the work
treatments on the accelerated durability performance on primer selection, no consideration was given to the
of bonded steel lap shear joints. A preliminary set of curing procedure, application method, layer thickness,
experiments reviewed the potential for a silane primer etc. Studies [224] have shown that many processing pa-
on a shot blasted surface and on an unprepared surface, rameters can affect the performance of the primer and,
i.e., with the mill scale still intact [33]. As this figure therefore, the performance of the adhesive joint.
shows, for a primer to be effective, it must first be ap- Bhuniye and Maiti [225] have also studied the ef-
plied to a clean prepared surface. Fay and Maddison fect of the environmental conditions on the durability
[222] found similar results when they investigated the of the three different adherend interfaces: (a) metal-
use of Accomet C and a silane primer on thin steel lap metal (i.e., Al-Al), (b) teak wood-metal (i.e., W-Al),
shear joints. They concluded that these processes re- and (c) teak wood-teak wood (W-W). The durability of
quire application to clean surfaces, which may be dif- adhesive materials under various harsh environmental
ficult to attain in volume manufacturing. conditions like boiling water, 5% salt water, 5% aque-
Knox and Cowling [33] has also studied the effects ous HCl and chemical mixture (i.e., toluene 30%, cy-
of four primers on the accelerated durability perfor- clohexane 60%, isooctane 9%, n-butyldisulfide 0.9%;
mance of bonded steel lap shear joints (see Fig. 63). As n-butylmercaptan 0.1%) has been tested. Under these
seen in this figure, all four primers, on a shot-blasted severe conditions only the Ciba-Geigy’s hardener cured
surface, improve joint durability compared to a stan- samples were treated. The retention values of the adhe-
dard shot-blasted surface. Fig. 63 also shows that the sive strength of polymer P-2 and standard epoxy resin
silane-treated surfaces are superior to the standard shot- on various surfaces such as wood-wood, metal-wood,
blasted samples and also the corrosion inhibitors. And, metal-metal surfaces are presented in Table III. It is
clear that stability of the adhesive bond of polymer P-2
under different harsh environments is inferior to that
of the standard epoxy resin. After 5 days of 5% HCl
treatment, it was found that the percentage of adhe-
sive strength retention in case of polymer P-2 less than
5%. This may be due to the higher acid-susceptibility
of the phospho-ester linkage of polymer which causes
hydrolysis of the P-O linkage of the resin [225].
To investigate the effect of various additives on the
lap-shear adhesive strength of the polymer, a mixture
of 10.0% Al2 O3 and 10.0% silica (precipitated type)
together and 10% Al2 O3 and 10% vinyltrichlorosilane
were mixed with polymer P-2 and the standard epoxy
resin before curing. Bhuniye and Maiti [225] have
tested the lap-shear strength of these mixtures. In case
of polymer P-2, the enhancement of adhesive strength
for metal-metal interfaces is higher than wood-wood
Figure 63 Accelerated durability performance of bonded steel lap shear and wood-metal interfaces. Even percentage of en-
joints showing the effect of various primers (Ref. [33]). hancement of adhesive strength is higher than that of

4783
T A B L E I I I Standard retention of the adhesive strength of polymers, P-2, and standard resin after various harsh environment [225]

Retention of adhesive strength (%)

5% HCl treatment 5% Salt water treatment Chemical treatment

Boiling water
Polymer code Interfaces treatment (1 h) 3 days 5 days 3 days 5 days 3 days 5 days

P-2 W-W 23.0 38.0 5.0 42.0 13.0 60.0 32.0

W-Al 20.6 22.0 2.0 37.0 7.0 51.0 21.0
Al-Al 9.7 21.0 Nil 29.0 5.0 48.0 18.0
Standard resin W-W 71.3 80.0 62.0 90.0 78.0 99.2 97.3
W-Al 53.9 90.0 73.0 84.7 73.0 98.9 93.0
Al-Al 54.0 87.3 65.0 77.6 77.0 99.0 96.0

the standard resin. It is observed that vinyltrichlorosi- There is typically tradeoff between quick cure and long
lane is a powerful additive for enhancement of adhesive pot life.
strength in each case by forming an effective bonding Adhesive joints are natural to consider for polymeric
between epoxy polymers and metal or wood surfaces, matric composite materials because many matrix resins
i.e., it acts as a good coupling agent [226]. In the case, are also good adhesives. For example, epoxies are used
Si O bond is generated from the hydrolysis of Si Cl as adhesives for fiber-reinforced epoxy laminate as well
bond of vinyltrichlorosilane which strongly interacts as for many other materials.When the matrix material
with the metal substrate surfaces forming Si O bond of the laminates in composites is also used as the adhe-
[225]. On the silane/adhesive interface, a coupling in- sive in the joint, excellent adhesion might result [60].
teraction arises from the polymerisation of the vinyl However, even with the excellent adhesion, the joint
group of vinyltrichlorosilane under curing conditions, does represent a discontinuity in the material, and re-
and the polymer thus formed interacts with the base sulting high stresses often initiate joint failure, which
resin. Thus using vinyltrichlorosilane as an additive the was discussed in Section 3. Complex composite ma-
bonding between the adherends is enhanced more than terials found their way into high-tech industries like
in other cases. space travelling and aircraft construction a long time
ago. However, as production methods have developed
and raw material prices have been decreasing, the usage
10. Selection of adhesives in civil engineering has increased. The pultration pro-
10.1. Process optimisation cess offers a production method that enables polymer-
Tremendous advances have been made in the devel- based composite profiles with a sufficiently high ma-
opment of adhesives that are stronger, easier to use, terial quality at an affordable price to be used in var-
less costly and more reliable than alternative methods ious construction applications [64]. The applications
of joining. From early applications, such as plywood are tightly bound with joining of composite profiles
and World War II aircraft component assembly, the use using appropriate adhesives.Therefore, the task is to
of structural adhesives (where the adhesive is a load- select right adhesives for bonding of composites with
transmitting part of the product) has grown rapidly [11]. respect to various environmental conditions. Naturally,
Whenever bonding is planned, adhesive selection is of successful adhesive testing depends on the number of
primary concern. The chemical industry has provided factors [64]: (a) proper surface preparation of compos-
the user with a multitude of adhesives to choose from ite adherends, (b) mixing and application of the tested
for any given application. Two-part, medium-viscosity, adhesive, (c) clamping of the joint and curing of the
nonslumping, room-temperature-curing adhesives are adhesive, (d) test specimen preparation, and (e) final
the most successful because of their user-friendly prop- test execution. The shear properties of adhesive-bonded
erties [8]. The desirable properties [8, 227] are (a) a 1:1 joints should be evaluated properly under the realistic
mix ratio of two different-color components that com- environmetal conditions since many adhesive-bonded
bine to give a third distinct color, signifying a complete composite assemblies are designed to shear stresses.
mix, (b) a uniform bond layer thickness as well as a uni- Therefore, single- and double-lap tensile shear tests
form spread of the adhesive over the contact area are were used for composite specimens [64]. The tested ad-
required with possibly no or minimum excess adhesive hesives were applied to one or both bonded adherends
protruding out at the edges of the bond area. Therefore, and the joint results are compared.
once the adhesive has been selected the crucial task in A uniform bond layer thickness as well as a uniform
designing the production technology remains how and spread of the adhesive over the joint contact are re-
where to deposit predefined quantities of this adhesive quired with possibly no or minimum excess adhesive
in order to achieve a bond layer of the desired quality protruding out at the edges of the bond area. Therefore,
and (c) a quick cure for rapid handling strength. There- once the adhesive has been selected the crucial task in
fore, in cases where fast-setting adhesives are used, for designing the production technology remains how and
example, cyanoacrylates, the spreading of the adhesive where to deposite predefined quantities of this adhesive
should be completed in the shortest time possible. Yet, in order to achieve a bond layer of the desired quantitiy.
the adhesive must possess a pot life sufficient enough to In order to solve this problem, Babic [227] studied the-
allow time to complete the application without rushing. oretically the problem of bonding rectangular plates to

4784
substrates with a low viscosity, fast-setting adhesive. T A B L E I V Comparison of physical and mechanical properties be-
The main problem in this case is how to assure that the tween FM R
300-2 and FM
R
300 adhesive film aluminum adherends
[6]
adhesive will (a) evenly spread over the whole rectangu-
lar contact area including the problematic four corners FM 300-2 K FM 300 K
of the plate and (b) complete this task in the shortest cured 90 min cured 60 min
time possible. Babic [227] first found an analytical so- Property at 121◦ C at 177◦ C
lution for the case, where a drop of liquid adhesive is Lap shear strength (MPa)
deposited between two large rectangular plates with a 24◦ C 38.6 37.9
plane surface. The adhesive between the two plates ini- 121◦ C 26.9 27.6
tially occupies a cylinder-like volume with a radius Ro . 150◦ C 15.9 18.6
Floating roller peel (kN/m)
 24◦ C 6.4 6.2
mo
Ro = (68) Honeycomb sandwich peel (Nm/m)
πρa do 24◦ C 74 67
Flatwise tensile (MPa)
where m o and ρa are the mass and density of the adhe- 24◦ C 7.6 6.9
150◦ C 2.8 3.2
sive deposited and do the initial distance between the Flow (%) 450–550 450–550
plates. If the two plates are squeezed together, that is, if Tg ◦ C (TMA) 144 148
the distance d is time dependent, the radius R increases.
We obtain

mo
R(t) = (69) that this construction method has a lot to offer for fu-
πρa d(t)
ture design programmes.
Kohli [6] has reported on the development of a new
and consequently
adhesive system for aeroengine bonding applications.
 This adhesive film designated FM R 300–2 is a 121◦ C

R(t) do ◦ 
= (70) cure version of 177 C curing FM 300 adhesive film
R

Ro d(t) and provides similar stress-strain and mechanical per-

formance to the FM R 300 system [228, 229]. In this

with the distance d reduced work, he investigated the performance of FM R 300-
√ for example by a factor 2,
the radius R increases by 2. 2, epoxy-based adhesive, for various metal and com-
Evidently, an equivalent effect can be achieved by posite bonding applications including applying this ad-
keeping the distance d constant, that is d = do , and hesive to aluminum adherends. This adhesive can be
increasing the mass m o , by the same factor. In this case, cured at 121 to 177◦ C and provides similar physical
and mechanical properties irrespective of the cure tem-
 perature. Due to its low moisture absorption and high
R(t) m(t) glass transition temperature, this shows excellent per-
= (71)
Ro mo formance when bonding to wet Normex honeycomb
and in bonding precured thermoset and thermoplas-
This simple conclusion means that the spreading of the tic composite substrates. Comparison of physical and
adhesive between two parallel plates being squeezed mechanical property data between these adhesives is
together resembles the process of injection molding shown in Table IV. It is notable that the 121◦ C curing
where the region between the plates represents the mold adhesive has a glass transition temperature of 144◦ C
cavity and the adhesive the molten polymer, respec- as compared to 148◦ C for the 177◦ C cured. Because
tively [227]. As the distance d between the two plates of this high glass transition temperature, this film ad-
is very small compared to the lateral dimensions L, hesive is capable of significant retention of its strength
that is d L, in many practical cases the Hele-Shaw up to 150◦ C. The toughness properties of FM R 300-2

assumptions for the liquid flow can be adopted. adhesive as indicated by peel strength are quite sim-
ilar to FM R 300 . Both these systems have metal to

metal peel strength of 6.4 kN/m. Because of their simi-

10.2. Progress in aerospace structural lar toughness and flow properties, both adhesives have
adhesives similar sandwich properties. One of the key objectives
Adhesive bonding has been used in the manufacture of in the development of this new adhesive was that it
primary aircraft fuselage and wing structures for over should have stress-strain properties similar to the ad-
50 years. As such, it is a direct competitor process to riv- hesive. The 121◦ C cured FM R 300-2 adhesive has the

eted structures but not as dominant. Adhesive bonding same stress and strain properties as FM R
300 adhe-
is used mainly for attaching stringers to fuselage and sive, not only under dry but also under hot/wet con-
wing skins to stiffen the structures against buckling. It ditions up to 104◦ C. Therefore, the lower temperature
is also used to manufacture stiff lightweight structures curing adhesive can be used to minimize stresses caused
of metal honeycomb cores inside metal skins for the by differential coefficients of expansion when bonding
flight control component structures (elevators, ailerons, dissimilar substrates [6].
spoilers, etc.). As stated by Higgins [12], the adhesive Higgins [12] has recently reported the adhesive ma-
bonded aircraft structures are stable and durable and terials used in the aerospace industry. In this work,
4785
the bonding of the major aluminum alloys used as ad- T A B L E V I I Comparison of lap shear strengths at service temperature
herends in commercial aircraft structure (i.e., 2014A, (i.e., −55◦ C, +80◦ C) for the various adhesives used in the aerospace
application [12]
2024, 7075 and 7150) were presented. According to
Higgins, the adhesive materials used for these pur- Lap shear (MPa min) at
poses fall into three distinct groups.These are: (a) Adhesive (−55◦ C) (+80◦ C)
Metal/metal-hot cure, (b) metal/metal honeycomb-hot
Redux film 775 20 10
cure, and (c) metal to metal-cold cure. The hot cure FM47 20 9
adhesive materials used by the aircraft industry are AF31 16 14
mainly based on either phenolic or epoxy resin sys- Redux308A/NA 30 25
tems. Phenolic-based hot-bonded systems cure by con- FM73 30 25
densation reactions which means that water vapor is AF163-2 30 25
EA9330.1 12 6
generated during the cure. To prevent the bond line be- SW9323B/A-150 12 16
ing pushed apart by the vapor release, high pressures
need to be applied across the joint during the reaction
period. This in turn effectively limits the bond widths
to approximately 300 mm if optimum glue line thick-
Subsonic aircrafts are designed to perform within
ness is to be achieved [12]. Large-area bonding will be
a service temperature range of +80 −55◦ C and the
required to be designed to give “windows” in the struc-
properties of the adhesives at these temperature have
ture to allow the expanding surplus resin to spew out.
to be considered by designers. Typically the lap shear
The main structural phenolic and epoxy adhesives used
strength of the adhesives at these extremes of operating
on aircraft components, with the bonding conditions
temperature are presented [12] in Table VII. Faced with
required are given [12] in Table V. The most important
these figures it is tempting to conclude that the epoxy
category of structural adhesives is that of the metal to
systems are superior to the phenolic systems. However,
metal hot cure types and a comparison of their room
structural engineers considering adhesive bonding for
temperature mechanical properties (i.e., lap shear and
primary structures need also to concern themselves with
peel strengths) are shown in Table VI (Ref. [12]).
the durability of the bonded joints in service, the ef-
fects of fluids contacted in service (i.e., water, fuel, oil,
Skaydrol, de-icing fluids, etc.) as well as the effects of
T A B L E V The hot cured commercial adhesive materials used by the high humidity.
metal to metal or metal to metal honeycomb bonding (i.e., mainly used by
aircraft industry) (Ref. [12]). This table shows the bonding temperatures
and bonding pressures
10.3. Process design and optimisation
Bonding Bonding
Adhesives temperature (◦ C) Pressure (kPa)
on the basis of effects of humid/hot
environments
Metal to metal bonds-hot cure In recent years there have been much efforts to find
Phenolic systems suitable adhesives on their long term durability and
Redux film 775 145–156 480–760 performance of rubber to metal bonding. The manufac-
CYTEC FM47 150–175 480–1380 turers of of rubber to metal bonded components con-
3M Co. AF31 170–180 310–620
Epoxy systems
sume large quantities of volatile organic solvent-based
Hexel redux 308A 170–180 310–380 materials to degrease the metal parts and bond rubber
3M Co. 163-2 115–125 200–276 to metal [230]. This may cause considerable hazards
Metal to metal honeycomb bonds-hot cure to health such as risk of fire, explosion and harmful
Epoxy systems vapors and inevitably pollutes the working environ-
Hexel redux 308 180 345 ment [49]. Therefore, there has been urgent need to
3M Co. AF3109-24 180 345 promote a wider use of commercially important water-
based cleaning materials and bonding agents. There are
numerous aqueous bonding systems in the market at the
T A B L E V I Comparison of room temperature mechanical properties present time. A more extensive use of these products has
(i.e., lap shear and peel strengths) of the various hot-cured adhesives been hampered by lack of sufficient data on their long
bonding to metal-to-metal adherends used in the aircraft industry [12]a term durability and performance. Bond [231] studied
Lap shear Peel strength the characteristics and performance of some waterborne
Adhesive strength (MPa) (N/25.4 mm) adhesives and highlighted some of their advantages
over the organic solvent-ones which are still being used
Redux 775 27–35 180–270 for bonding metal to rubber adherends. Advantages of
Redux 308A/NA 40–45 200–310
FM73 35–40 245–350
the water-based adhesives can be summarized on the
AF163-2 35–40 245–330 basis of Bond’s work [231]: (a) adherends for aque-
EA9330.1 27–35 120–155 ous adhesives can be cleaned with warm water or soap
SW9323B/A150 27–35 120–155 solution, therefore eliminating the need for additional
a National
solvent usage. (b) comparatively lowering the costs of
specifications covering these adhesives are: DTD 5577 (UK),
MMM-A-132A and MIL-A-25463 (USA). waste disposal, ventilation, and heated air. (c) In labora-
Typical room temperature values for cold cure epoxy adhesives are: Lap tory testing, waterborne adhesives achieve good results
shear 27–35 MPa and Peel 130 N/25.4 mm. in bonding a variety of elastomers, including natural
4786
rubber, to various substrates during the curing process. air at room temperature and 60◦ C, and cool water).
A recent work by Ansarifar et al. [49] also showed that LETOXIT LX 012 exhibited large strengthening dur-
using aqueous cleaning materials and bonding agents ing 1 year stored at room temperature. After that time
in rubber to nylon 6,6 bonding offers excellent bonding the strength reached a value of about 13 MPa. Produc-
properties as well as a cleaner and safer working envi- ers say [64] that adhesives should attain 90% of final
ronment. Furthermore, costs can be reduced by elimi- strength within 5 days. This is valid for all adhesives ex-
nating the primer and using water-based covercoat. Ny- cept LETOXIT LX 012, which reached only about 70%
lon 6,6 is a suitable replacement for steel in rubber to of the strength at that time. The strength rose as curing
metal bonding applications. process continued. The strength of joints immersed in
Balkova et al. [64] have investigated the effects of water decreased after 10 days of storage except EPOXY
atmospheric condition and cool water at different tem- 371. LEPOX UNIVERSAL 11 recorded the greatest
peratures (i.e., room temperature and 60◦ C) to select deterioration of the shear strength, from 8 to 3 MPa.
the most suitable epoxy-based adhesive for bonding EPOXY 371, also with other type of primer (P11),
glas fiber/polymer composites material systems. Lap exhibited good water-resistance. Composite materials
shear specimens were tested within 1 year. In this study, may be applied at elevated temperature. They have cho-
three types of two-part epoxy-based adhesives (i.e., sen 60◦ C for testing shear properties (tested adhesives
EPOXY 371, LETOXITE LX 012, LEPOX UNIVER- are applicable up to 80◦ C). They showed that the shear
SAL 11) were used for bonding fiber-reinforced plastic strength of all adhesive joints improved rapidly after 1
pultrusions. Selected adhesives are representatives of day in the oven against the same time at room tempera-
moisture-curable adhesives. Glass fiber/polyester com- ture. LEPOX UNIVERSAL 11 stored at 60◦ C showed
posites, produced by pultrusion technique, were used as better shear properties compared to room temperature.
adherends. The rough area of the composite adherend Balkova et al. concluded that the most commonly used
was covered by thin film of fluid adhesive in order to wet adhesive EPOXY 371 has maintained the shear prop-
the surface completely and leave no voids. Quality of erties without great change in all environments tested,
adhesive bond was tested on single- and double-overlap also with other type of primer (P11), and exhibited the
shear specimens prepared according to ASTM stan- best shear strength in cool water. It was therefore rec-
dards D-1002 (single-lap), and D-3528 (double-lap). ommended for universal usage. The elongation at break
Four overlap lengths (10, 20, 30, and 40 mm) were was also evaluated for all adhesive joints stored in wa-
tested to determine the optimal length of the lap [109, ter within 1 year. Elongation reached values from 1.3
232, 233]. Primer in EPOXY 371 was changed for an- to 2.5% in water, from 1 to 2.5% at 60◦ C, and from 1.3
other one “P11” on the basis of assumption that shear to 3.0% at room temperature.
properties could improve. Kohli [6] studied the effect of prebond humidity
It was observed [64] that the shear strength of double- by exposing the adhesive film (i.e., FM 300-2K) and
lap shear specimens was always lower that of the single- the metal substrates (i.e., aluminum adherends) to two
lap ones. This was predominantly made by fracture of weeks at 80% relative humidity at 24◦ C. Whereas for
one weaker adhesive joint. Therefore, it was concluded postbond humidity studies, individually cut coupons
that the results from single-lap shear specimens are were exposed to 71◦ C and 100% relative humidity for
more representative. It was also observed that appli- 30 days. The effects of these humidity on the dura-
cation of adhesive on both bonded adherends improved bility of the bonded structures were evaluated using
the shear strength. From the variation of the overlap the various mechanical test methods [i.e., the lap shear
length, the optimal overlap length was determined to (i.e., ASTM D1002); floating roller peel (i.e., ASTM
be as 10 mm using the dependence of stress at frac- D3167), (c) Honetcomb Sandwich Peel (i.e., ASTM
ture on length of single-lap, as seen in Fig. 64. Balkova D1781); flatwise Tensile (i.e., ASTM C297)]. The float-
et al. [64] have tested the the glass-fiber/polyester com- ing roller peel strength data is shown in Table VIII. The
posites joints in various environments (i.e., exposed to FM R 300-2 adhesive shows excellent retention of its

strength up to 121◦ C after prebond and postbond hu-

midity exposures.The outstanding moisture resistance
of this adhesive is also demonstrated by its ability to

T A B L E V I I I Effect of prebond and postbond humidity exposures on

peel strength FM 300-2 K adhesive film BR 127 adhesive primer [6]

Floating roller
peel (kN/m)
Prebond film Bonded coupon
exposure exposure 24◦ C 121◦ C

None None 6.3 6.3

30 days at 6.2 6.9
71◦ C/100% RH
15 days at 80% RH None 7.0 6.5
at 24◦ C 30 days at 5.6 6.3
71◦ C/100% RH
Figure 64 Strength of joint as a function of the overlap length [64].

4787
T A B L E I X Wet Nomex bonding with FM 300-2 K adhesive film [6]

Bonded coupon Flatwise tensile (MPa)

(i.e., postbond)
Nomex core exposure exposure 24◦ C 82◦ C 120◦ C

None 5.4 5.3 4.3

(core) (core) (core)
3 weeks at 65% 30 days at 71◦ C 4.8 3.9
RH at 24◦ C 100% RH (core) (core)
Immersed in water None 5.0 5.1 4.2
for 24 h at 60◦ C (core) (core) (core) (a)
then bonded within 2 h
30 days at 71◦ C 5.0 3.7 –
100% RH (core) (core) (core)

Note: Failure mode: 100% core. Adhesive weight: 392 gm/cm2 .

bond to wet Nomex and maintain its strength up to

121◦ C even after humidity exposures. The wet Nomex
honeycomb bonding data is shown in Table IX.
Kohli [6] has also studied the effect of humidity on (b)
the composite substrate (i.e., carbon-carbon substrates) Figure 65 Strength retantion of the lap joints after Cataplasma and cyclic
using the same adhesive as mentioned previously. The agings [111]: (a) Cataplasma aging and (b) Cyclic aging.
bonded specimens were tested before and after humid-
ity exposures. The wet conditioning of the bonded spec-
imens involved exposure to 71◦ C/100% RH for 30 days Fig. 65a). Without primer, the lap-shear joints made
for the Ciba 6376 precured substrates. The lap-shear with HDPE-based adhesive retained about 77% of their
strength data indicate that the performance of this adhe- initial strength after five weeks of Cataplasma aging
sive system is not significantly affected by the exposure (see Fig. 65a). The poor strength retention of adhesive
to moisture.This is a remarkable result and indicates joints with primer showed that the primer coating could
that there is no need to dry the composite laminates not stand the severe Cataplasma aging conditions. How-
prior to bonding with the FM 300-2 adhesive film.The ever, the five weeks of cyclic moisture/temperature did
post-bond humidity exposures data (82◦ C/wet) shows not have any significant negative influence on the joint
that in the case of precured laminates exposed to the strength (see Fig. 65b) and failure mode of adhesive
dry or ambient conditions prior to bonding there is no joints.
degradation in performance. However, when the pre- When new adhesive systems are being assessed the
cured laminates were exposed to the hot/wet conditions static properties are not the main criteria. For aircraft
prior to bonding, there was some degradation in proper- structure, durability is of greater importance and warm,
ties after the post-bond humidity exposures. This shift wet, testing by soaking in water at 35◦ C for long peri-
in properties is attributed to the lowering of the stiff- ods or high humidity testing (at 95% RH and 70◦ C) is
ness of the composite and does not indicate a problem used to determine this [12]. Usually the testing is car-
related to the adhesive system [6]. The failure mode in ried out back to back with a proven existing adhesive
most cases was thin cohesive within the adhesive layer for comparison. Higgins [12] has recently studied the
[6]. durability aspects of the adhesive bonding of aircraft
Xu et al. [111] have investigated the durability per- structures using the main adhesives applied in aircraft
formance of galvanized steel to laminated glass-fiber industry. As Higgins pointed out, both epoxy and phe-
PP/thermoplastic adhesive joints using two aging meth- nolic adhesive bonded joints have good resistance to
ods: (a) the Cataplasma aging tests, and (b) cyclic mois- fluid immersion tests but the phenolics perform bet-
ture/temperature tests. For the Cataplasma aging test, ter in high temperature/high humidity or warm (35◦ C)
lap shear specimens were aged in an oven at 70◦ C for wet aging testing and this is born out by performance
5 weeks. The aged lap shear specimens were then tested. in service. Long-term warm, wet, aging studies [234]
The cyclic moisture/temperature aging test was done for up to two years immersion has shown that Redux
by hanging the lap-shear samples in the different envi- Film 775 bonding has the best performance followed
ronmental conditions listed (i.e., 50◦ C, 100% humidity; by AF 163-2 adhesive bonds. FM73 was slightly worse
22◦ C; 50% humidity; −20◦ C). The complete cycle took than AF 163-2 while Redux 308A was significantly
one week, and the specimen was cycled for five weeks. worse than these. The two-part cold cure epoxy ad-
For lap-shear joints made with HDPE-based thermo- hesive are least able to withstand warm, wet, aging.
plastic adhesive, Cataplasma exposure and cyclic mois- A past study [235] concluded that “Results from ac-
ture/temperature exposure, were performed to study celerated environmental tests on CAA (chromic acid
the joint durability. When the galvanized steel surface anodised) aluminum adherends with Redux 775 (as
was treated with primer, the lap shear joints made with well as AF 163-2) has shown that in the absence of
HDPE-based adhesive retained only in 11% of their ini- primary bond line corrosion, that good stability of the
tial strength after five weeks of Cataplasma aging (see interface between oxide and adhesive in the presence
4788
of water at 35◦ C is obtained for periods of up to, and ysis that behaves like a spring with normal and shear
probably more than 4 years”. This confirms the various stiffnesses, which are dependent on the thickness of the
studies carried over the past 20 years.This experience adhesive layer. Carpenter [240] developed a simple el-
has shown that Redux 775 was even more durable on ement that considered the adhesive with no thickness.
aluminum surfaces that had been prepared by alumina Later, the same author [241] developed another element
blasting and chromic/sulphiric acid pickling (with no that considers the adhesive thickness and a linear vari-
primer [12].The high durability of Redux 775 is leg- ation of the displacement field across it. Carpenter and
endary. This is shown by the successful disbond free Barsoum [242] developed two specific elements that
performance of primary structure on aircraft that have incorporate some assumptions used in several analyti-
been in service for over 25 years [12]. cal studies. These elements can be used with 2-D plane
stress or plane strain elements, as well as with shell or
beam elements. Rao et al. [243] developed a six-node
11. Concluding remarks: Mechanical isoparametric element similar to the one presented in
and environmental durabilty [239]. Yadigari et al. [244] modified that element to
Adhesive bonding is being increasingly used in struc- include longitudinal normal stresses and viscoelastic
tural applications recent years. The characteristics of analyses. Lin and Lin [245] introduced an element that
adhesive joints make them attractive in industries such represents both the adhesive layer and the adherends
as aeronautics, automotive and civil engineering. When and that is based on Timoshenko’s beam theory. Reddy
compared to mechanically fastened joints, adhesive and Roy [246, 247] developed a special element for
joints have the advantages of having less sources of geometric and material nonlinear analyses.
stress concentrations, more uniform distribution of All the finite element analyses referred to above are
load, and better fatigue properties [92]. One of the ma- two-dimensional . However, even for a simple adhe-
jor advantages of adhesive bonding is its ability to join sive joint such as the single-lap joint, stresses are of
dissimilar materials [212, 236]. Adhesive bonding pro- 3-D nature ([i.e., 91, 179, 248, 249]). This was inves-
vides a more synergistic use of dissimilar materials in tigated by Tsai and Morton [179] who performed 3-D
design. The potential of fast, in situ processses makes linear elastic FEA of a single-lap joint in which bound-
the thermoplastic bonding an attractive option for future ary conditions account for the geometric nonlinear ef-
manufacturing. For example, laminated glass-fiber— fects.These authors showed that 3-D regions exist in
reinforced polypropylene (PP) provides fast and clean the specimen. They showed that the adherend and ad-
processing and lower costs than most thermoplastic hesive stress distributions in the overlap, near the free
composite. surface, are quite different from those occurring in the
The increased application of adhesive joints was ac- interior. Pandey and Narasimhan [248] also contributed
companied by the development of mathematical models to the identification of the 3-D nature of stresses in
to analyze the behavior of those joints. Both analytical adhesively bonded single-lap joints by performing 3-
and numerical models (i.e., finite element model) have D viscoelastic analysis considering material and geo-
been developed [92]. However, its use in truly struc- metric nonlinearities. Special elements that model the
tural applications is still often limited. This is mainly adhesive layer were also developed for 3-D analysis.
due to a lack of confidence in the performance of ad- Edlund and Klarbring [249] developed an element for
hesive joints, since the mechanical performance of the geometric nonlinear analysis. They assumed that the
joints may deteriorate upon being subjected to cyclic- adhesive layer is of negligible thickness and has linear
fatigue loading, especially if the joints are also exposed elastic behavior. Andruet et al. [91] developed a model
to a moist environment [211]. As Gonçalves et al. [92] for 3-D analysis of adhesive joints based on shell and
pointed out, the ability to quantitatively describe this solid elements. The shell elements are used to model
reduction in performance and to predict the lifetime the adherends and the adhesive layer is modeled as a
of bonded joints would be a powerful tool, enabling solid element with offset nodes in the mid-planes of the
manufacturers to make wider and more efficient use adherends.The element formulation includes geometric
of adhesive bonding. The 2- and 3-D FEA approaches nonlinearities, and thermal and moisture effects.
have been extensively applied by many workers (see In addition to considering the 3-D nature of adhe-
for example references [92, 116, 237–247] to anal- sive joints, the calculation of accurate stresses at the
yse the adhesive joints considering the linear and ge- adherend-adhesive interfaces is of primary concern.
ometric nonlinearities. The 2-D FEA models that in- The failure of adhesive joints usually occurs in the ad-
corporate nonlinearities were also developed. Cooper hesive (cohesive failure) or at the adherend-adhesive
and Sawyer [237] considered geometric nonlinearities interfaces (adhesive failure). Moreover, cohesive fail-
in their model. A similar analysis was performed by ure usually occurs near the adherend-adhesive inter-
Tsai and Morton [238]. Harris and Adams [116] incor- face. However, conventional finite elements based on
porated material and geometric nonlinearities in their displacements do not satisfy the continuity of the stress
analysis. They used triangular and quadrangular plane vector at the interfaces. Cheikh et al. [250] proposed a
strain elements and modeled the spew fillet at the end technique, based on the finite element method formu-
of the adhesive layer. Several special elements to model lated in terms of displacements that enables stress con-
the behavior of adhesive layer were developed. Barker ditions to be imposed. They used 2-D FEA and applied
and Hatt [239] published one of the earlier works in this their technique in the study of the stress distributions at
area. They developed a four-node element for 2-D anal- the adherend-adhesive interfaces.

4789
11.1. Effect of adherend surface This limits the water penetration, by physically reduc-
pretreatment on the durability ing the water flux to the interface and to the center of the
performance of joints joint. The most critical aspect of environmental dura-
Many research works have been carried out in order bility testing is the assessment of the test results and
to evaluate the effects of various surface pretreatment the extrapolation of those results to predicted service
methods on the durability performance of adhesive life. The new test (i.e., scrap test) described in Section
joints (see for example, references [10, 33, 39, 213, 4.2 was suggested to be ideal for rapidly discriminating
251]. It took almost a decade to acknowledge the crit- amongst the improved structural adhesives and surface
icality of surface pretreatment for metal-bonded struc- treatments, if not for predicting service life. The only
tures. As pointed out by Kinloch et al. [251] the fracture major drawback seen for this new test method is that it
mechanics approach provides an excellent method for can only be used on relatively smooth adherends [33].
evaluating the effects of the different surface pretreat- It was demonstrated [33] that the use of silane primers
ments on the durability of the adhesively bonded joints. increased the durability performance of mild steel ad-
In Particular, the combination loading such cyclic- herend/epoxy adhesive joints. Wheras the use of silane
fatigue and the presence of an aqueous environment leads to improving the durability of butt and lap joint
leads to an assessment of the environmental resistance types [28]. Lap joints without silane showed a strength
of the bonded joint within a matter of weeks, as opposed loss during the first three weeks aging, then a strength
to the more typical accelerated aging tests which in- recovery in the next 3–6 weeks, before a further strength
volve exposing the joint, unstressed, in water for many loss [28]. Moisture degraded butt joints are faster than
months [251]. Also, in unstressed tests, the water tem- lap joints [28].
perature is often relatively high, well above any likely Molitor and Young [213] have investigated the effects
service temperature, in order to try to produce a large of various surface pretreatments including Excimer
accelerated–aging factor. This frequently leads to un- laser cleaning and SHA (sodium hydroxide anodisa-
representative failure mechanisms being observed in tion) on the adhesive bonding of a titanium alloy to a
these tests, and very misleading results being obtained. glass-fiber reinforced composite material in a hot/wet
Knox and Cowling [33] have demonstrated that a environment(i.e., in hot water at 70◦ C for 30 days) us-
surface pretreatment of adherends prior to bonding in- ing the 90◦ peel tests. Based upon the 90◦ peel test re-
creased the durability performance using standard ac- sults, SHA and surface cleaning by Excimer laser were
celerated environmental testing techniques. For exam- selected. Both thermosetting and thermoplastic com-
ple, to ensure, initially, reasonably, strong joints, in the posites were evaluated using two aerospace industry
case of mild steel, it is generally sufficient to remove standard film adhesives. An investigation of the dura-
surface contamination and weak oxides [118] which bility of a titanium/glass fiber composite lap joints in
may act as weak boundary layers. However, to pro- a hot/wet environment using a single lap shear testing
duce durable joints, it is necessary to form stable ox- has shown that surface treatment by Excimer laser re-
ides which are receptive to the adhesive/primer and to sulted in a five-fold increase in peel strength (dry), but
establish strong, stable, intrinsic forces at the adhesive the environmental durability was shown to be poor. In
or the primer and oxide surface. comparison, the surface preparation of the titanium ad-
While other experimental techniques may be used herend by SHA method resulted in relatively high peel
to quantify the effects of various surface pretreatments strength and provided a durable composite/metal bond.
on durability performance they will generally not be as It is important to understand how water has affected
simple or give results as quickly. The Boeing wedge test the titanium/adhesive interface. Two of the most impor-
is another technique which has been employed [33], tant parameters governing the durability of titanium ad-
as described in Section 4.1.2. As Knox and Cowling herends are providing a surface with significant micro-
[33] claimed the results from this type of test are only roughness and providing a stable oxide layer [252]. The
qualitative in nature and it is not possible to distinguish significant drop-off in shear strength following envi-
quantitatively between different grades of interface sta- ronmental aging indicates a requirement to protect the
bility [32]. Knox and Cowling [33] have developed a titanium oxide, possibly by a primer.
novel test method (i.e., scrap test) to rapidly discrim-
inate amongst the durability performance of adherend
surface conditions. This scrap test was applied to de- 11.2. Environmental durability
termine the effect of a silane primer on the durability of adhesively bonded joints
performance of a steel adherend shot-blast surface. The Environmental resistance is fundamental to the dura-
Boeing wedge test would generally require a longer test bility of a bonded joint or repair. Most in-service fail-
duration than the scrap test experiments. It was pointed ures are caused by environmental degradation of the
out [33] that the results are obtained so rapidly from interface between the bonding surface and the adhe-
the new scrape test as the test area, i.e., the bulk adhe- sive. These failures are characterized by adhesion fail-
sive and the adhesive/adherend interface become very ures at the interface. The environmental resistance of
quickly saturated with water, due to the test geome- an adhesive bond is determined by the chemical bonds
try providing a larger surface area for water ingress. formed during cure of the adhesive and the resistance
Whereas, in standard test specimens, e.g., lap shear of the chemical bonds to environmental degradation.
joints, only a relatively small surface bond area is ex- As discussed in the previous section, surface prepara-
posed to the environment, i.e., the bondline thickness. tion is the most significant factor in determining the

4790
environmental performance of a bonded joint, and the hesively bonded joints by one or a combination of the
only method for producing a durable bond is to en- following processes [32, 255]: (a) diffusion through the
sure that the surface is prepared using a process which adhesive [256], (b) transport along the interface, and (c)
has been correctly validated. There is no design tech- capillary action through cracks and crazes in the adhe-
nique to manage bond environmental degradation and sive and/or diffusion through the adherend, if it is per-
bond strength will eventually decay to zero, if compo- meable. Water entry may cause weakening by one or a
nent failure does not occur first. It is well known that combination of the following actions [32, 213, 255]: (a)
the most common and most important factors influenc- altering the adhesive properties in a reversible manner,
ing the long-term behavior of unprotected adhesively such as by platicisation; (b) altering the adhesive prop-
bonded metal joints is the presence of high humidity or erties in an irreversible manner either by causing it to
liquid water [32]. This has been a subject of concern for hydrolyse, crack, or craze; (c) attacking the interface,
many years. Whereas the initial strength of a structural either by displacing the adhesive or hydrating the metal
adhesive is fairly high and can be predicted from the vast or metal oxide.
amount of data available, subsequent time-dependent Zanni-Deffarges and Shanahan [257] proposed that
deterioration, particularly in humid conditions, may be the phenomenon of capillary diffusion excarbates wa-
observed. This type of failure is less predictable due ter ingress. Surface tension effects near the metal ox-
to limited real-time, real-environment test data [253] ide/polymer interfacial regions increase the effective
and the relatively short time that synthetic structural driving force of water penetration. Such mechanism has
adhesives have been in common use. been reported in the degradation of composite materials
There are several physical mechanisms, which are [258] where water enters along the interface between
responsible for the degradation in metallic adhesively the matrix and fiber. Interfacial transportation can not
bonded joints due to water uptake. These include [32]: be excluded as a possible mechanism of water entry into
(a) rupture of secondary bonds at the interface, which the joint. It was shown [259] that a full depth spew fillet
causes displacement of adhesive by water, (b) mechan- has a distinct influence on stress distribution within the
ical weakening due to the change in the oxide structure joint and joint strength. The benefits through leaving the
of adherend and (c) hydrolisis in the boundary of the ad- spew fillet intact may be due to a number of contribut-
hesive layer. As pointed out by Abdel Wahab et al. [28], ing factors [29]: (a) Firstly, due to simple geometric
regardless which of these mechanisms is dominant, the considerations, a spew fillet will increase the total path
strength of a joint is governed by other parameters, length that the water must travel. However, as seen in
such as diffusion rate, change in mechanical proper- Fig. 66 the water can penetrate a joint with all fillets
ties due to moisture and swelling. The effect of change removed to a depth of 5 mm in 12 weeks. This demon-
in elastic modulus and swelling on joint strength was strates that a specimen with a fillet can be affected by
discussed in Ref. [254]. The reduction in modulus due water in the test time span, as a typical fillet is approx-
to moisture can reduce the stress concentration at the imately 3 mm in length. Therefore, there is a sufficient
joint edges, which leads to beneficial effect on the joint length of time for water to penetrate into the critical
strength. Swelling in the adhesive layer due to water load-bearing zone. (b) Secondly, removal of the spew
uptake has an effect on the stress in the joint and in turn fillet after cure modifies the exposed adhesive surface.
on the joint strength. Joint failure after aging appears When the spew fillet removed, the surface texture of the
to be intimately related to plasticisation of the polymer exposed adhesive is rougher and of a matt finish, ex-
by water and weakening of the steel substrate adhesive posing any voids or cracks in the adhesive layer, which
interfacial zone [255]. may act as entry sites for the water into the adhesive
Water is the substance that causes the greatest dura- [32]. Therefore, water diffuses into the bulk adhesive
bility problems in the environmental stability of adhe- causing aging of the adhesive itself. The diffusion pro-
sive joints [32, 213]. There are two fundamental prob- cess into critical areas is slowed down by leaving the
lems with water, its abundance and the fact that the spew fillet intact due to increased diffusion path length,
polar groups (which is therefore readily attracted to reduced interfacial stress and surface conditions of the
pre-treated surfaces) that confer adhesive properties on exposed adhesive.
a substance are inherently hydrophilic [29, 32]. Water
may affect the adhesive properties by plasticisation and
cracking. It may affect the adhesive/adherend interface
by displacing the adhesive or by hydrating the metal or
metal oxide surface of the adherend. For example, wa-
ter has a weakening effect on epoxide, which is used,
in the matrix of a large number of fibrous composites
[213]. In a marine application the majority of bonded
structural components will be exposed to moist air and
if the relative humidity is high then over a period of
time the strength of the joint will usually decline. Pre-
vious studies in the literature have demonstrated this
by exposing adhesive joints to high humidity, natural
climates or laboratory environments (see for example Figure 66 Water penetration into and AV119-bonded steel lap shear
references [32, 218]). Water may enter and affect ad- joint during accelerating aging (30◦ C, 100 Relative humidity) [29].

4791
11.3. Effects of joint shapes (i.e., fillet, the driving force then the unstressed samples with all
spew, chamfer, recessing) on the joint fillets removed should not be affected by water. This
durability performance is not the case (see Fig. 60). As stress does influence
There are many processing variables (geometrical and joint performance (Fig. 60), a reduction in stress at the
materials variables) in designing a durable adhesive interface must be of added benefit. It may be suggested
joint. These include (a) fillet, (b) spew, (c) chamfer, that the spew fillet acts as a “seal” on the joint edge and
(d) recessing, (e) bond length, (f) adhesive thickness, reduces water penetration into the joint at the interface
(g) adherend thickness, (h) choosing suitable materials due to the contributory factors discussed above.
for adhesives and adherends . Using the finite element analysis method, Belingardi
Knox and Cowling [29] have studied the effect of a et al. [61] have investigated the effect of spew (explor-
spew fillet and preload on the durability performance ing a wide range of spew angles, from 90◦ (i.e., square
of AV119-bonded steel lap joints (see Fig. 60). Consid- edge to 15◦ ) and chamfer size on the stresses (i.e., shear
ering first the results relating to unstressed samples, it and peel) in metal/plastics adhesive joints. It was shown
can be seen that the initial drop in strength (of approx- [61] that the stress reduction can be achieved not only
imately 21%) in the first two week period is followed if the spew type adhesive extremity is adopted but, by
by a much slower decline in joint strength over the next varying properly the geometry of spews, an optimal
10 weeks. Other workers have found similar trend in solution can be devised. Further improvement can be
results [76, 211]. It was suggested [29] that if a line obtained by adopting the spew and chamfer-type adhe-
is fitted through the results after the initial strength re- sive extremity. The magnitude of the stress peaks (for
duction, using regression analysis, the rate of decline in both shear and peeling components) decrease in entity
the samples with fillets is less than that of the samples with the decrease of the spew angle, although the most
with fillets removed and thus this indicates that a fillet is of the advantage is obtained within 45◦ solution. The
beneficial to durability performance. Using a numerical stress distribution in the adhesive, in the cases of spew
analysis Crocombe [260] also came to a similar conclu- solution, shows the presence of a secondary peak whose
sion. He reported that the effect of fillet of adhesive at entity can exceed the primary peak magnitude when the
the end of an overlap region on the kinetics of moisture spew angle is large but tends to disappear as the spew
diffusion is significant over considerable time scales. angle decreases. They pointed out, hovever, that yet
The general trend for the stressed samples is similar to for angles of 45◦ the peak stress reduction is adequate
the unstressed sample, showing first a large reduction and of the order of five times for the peel component
in strength followed by a much slower decline in joint (which is usually the most harmful for the adhesive)
strength [29]. However, the difference between stressed and of two times for the shear component with respect
samples with and without fillets is less distinctive and to the basic solution.The stress distribution in the ad-
there is also a larger amount of scatter in the results. hesive, in the cases of both spew and chamfer solution,
It may therefore be concluded that a stressed sample shows again the presence of a secondary peak whose
has a poorer durability performance than an unstressed magnitue can exceed that of the primary peak when the
sample (Fig. 60), and that the fillet has little influence spew angle is large, but tends to disappear as the spew
on stressed sample performance. angle decreases. Their study has, also indicated that yet
As pointed out by Knox and Cowling [29], the bene- for angles of 45◦ the peak stress reduction is adequate
fits through leaving the spew fillet intact may be due to a and of the order of two times for the peel component
number of contributing factors, none of which are dom- (which is usually the most harmful for the adhesive) and
inant: (a) Firstly, due to simple geometric considera- of two times for the shear component with respect to
tions, a spew fillet will increase the total path length that the previous spew solution. Belingardi et al. [61] con-
the water must travel. However, as this study showed cluded that adopting the spew and chamfer solution, an
water can penetrate a joint with all fillets removed to a advantage of one order of magnitued for the peel stress
depth of 5 mm in 12 weeks (see Fig. 66). This demon- component and of one half order of magnitude for the
strates that a specimen with a fillet can be affected by shear stress component can be gained in the maximum
water in the test time span, as a typical fillet is approx- stress values in the adhesive layer. With respect to other
imately 3 mm in length. Therefore there is a sufficient solutions presented in the literature, those considered
length of time for water to penetrate into the critical here are feasible and geometrically simple. Obviously,
load-bearing zone. (b) Secondly, removal of the spew the obtained result concerns the elastic behavior, thus it
fillet after cure modifies the exposed adhesive surface. can not be directly inferred that the ultimate strength of
Initially, after cure, the adhesive surface has a “glossy” the joint increases as much as the peak stress is reduced.
appearence. When the spew fillet is removed, the sur- Li and Lee-Sullivan [46] investigated the filleted
face texture of the exposed adhesive is rougher and of single-lap balanced joints in tension using the finite
a matt finish, exposing any voids or cracks in the ad- element analysis and experimental techniques. They
hesive layer, which may act as entry sites for the water concluded that there is minimal difference between
into the adhesive [32]. (c) Finally, by leaving the fillet plane strain and plane stress conditions in FEM pre-
intact, the stresses at the interface are reduced and peak dictions of the bending moment factor, k, and the adhe-
stresses occur at approximately 45◦ to the surface of sive stress distributions. The influence of rotation/no-
the adherend near th corner [260]. It may be assumed rotation boundary conditions on the bending moment k
that a stress is the required driving force for water entry is fairly large at low tensile loads but becomes increas-
into a joint at the interface [29]. But if stress alone was ingly smaller as the tensile loads are increased. The

4792
predicted k values using the analytical solution pro- modulus of 2.6 GPa. Thus, a further reduction of ap-
posed by Oplinger are similar to FEM simulations for proximately 8% would be needed to ensure a margin of
rotation boundary conditions within the tensile force safety equal to one.
range studied. With increasing use of adhesive bonding joints in
Steven and Xie [261, 262] have developed a simple industry, the analysis and design of the joints are be-
evolutionary structural optimization (EVOLVE) proce- coming more and more important. Cost and weight are
dure (i.e., a FEA method) for the optimisation of adhe- design criteria that are becoming increasingly impor-
sively bonded joints by shaping the adherend profiles tant, especially in the automotive and aerospace indus-
to reduce peak stresses at the end of the overlaps. The tries [63]. Therefore, it is important to find a suitable
EVOLVE method consists basically of removing the design criteria for adhesive bonding that may provide
low stressed part of the material progressively from the some means of reducing cost and weight. For exam-
structure [3]. Typically, the selection of an adequate ple, one method of doing so is recessing the adhesive
joint requires a number of design and analysis cycles layer by removing portions of the adhesive within the
which rely heavily on the trial-and-error method. The overlap [62]. This technique is termed as RECESSED
EVOLVE method relies on an iterative FEA and pro- BONDING. A recess joint is shown in Fig. 12g. Analy-
gressive removal of elements which takes the guess sis of recessed bonded joints is limited in the literature,
work out of the design loop. Firstly, in this procedure a as most researchers concentrate on continuous single-
FEA of the desired structure is performed and the stress lap joints (see for example, Refs. [72, 237, 263, 264])
distribution found. A plate stress file is then set-up and or on single-lap joints that have the substrate profile
using some criterion for rejection, here called a rejec- altered to reduce peak stresses [265, 266]. Olia and
tion criterion (RC) [3], such as the Von Mises stress, Rosettos [267] performed an analytical study on the ef-
the low stressed material is removed. For example, el- fects of gaps (i.e., recess) in single-lap adhesive joints
ements which have a Von Misses stress in the structure subjected to combined axial and bending loads. They
are deleted. The FEA and rejection cycle are then re- viewed the gap as a defect, such as debond or a void,
peated until a steady state (s.s.) is reached. This pro- that may reduce the joint strength rather than as a pos-
cess is repeated until the desired optimum is reached, sible means of reducing cost and weight. Their results
for example, till all stress levels are 20% of the max- showed that the presence of a gap creates high peel
imum in the whole structure. Therefore, the objective stresses at the free edges created by the gap. However,
is to minimise the maximum stresses in the joints. The if the gap is centrally located, the peel stresses at the
EVOLVE code was found to be effective in reducing the outside edges increase only slightly as compared to the
maximum principal stress in the adhesive for all mod- case without the gap. The shear stresses remain essen-
els studied [3]. Rispler et al. [3] have proposed another tially uneffected at distances far from the gap. In two
method called a stepped optimisation process in order earlier papers [268, 269], Rosettos and his co-workers
to achieve accurate solutions both in terms of stress have shown that the shear stresses are hardly affected
magnitudes and optimum shape of fillets.This method by a central void size of up to 70% of the overlap length.
effectively reduces computation time in reaching an op- If the void is very close to the overlap ends, differences
timum design. of the order of 20% in maximum shear stresses may oc-
Rispler et al. [3] have recently employed the cur. Mazumdar and Mallick [196] have experimentally
EVOLVE method to optimize the shape of adhesive investigated the effects of recessing on the failure load
fillets found in the tabs of tensile test speciments in of single-lap joints. By conducting static tensile tests
which the adherends were titanium and T800 unidirec- on single-lap joints with varied degrees of recessing,
tional tape bonded with a generic paste adhesive. In and comparing the results to a continuous joint, they
this study, the reduction in maximum principal stress showed that the average failure load did not increase
achieved with respect to the baseline model (CFRP up- with increased recessing. However, the lap shear values
per and middle adherends with E adhesive = 3.8 GPa) (i.e., peak load divided by the net bonded area) nearly
ranged between 48 and 64%. The peak peel stresses in doubled by recessing the joint, which indicates that the
the central CFRP adherend and in the adhesive were re- average strength of the bond increases with recessing.
duced by up to 66%. All double lap joint models had a Let us now summarize the above-mentioned results.
CFRP middle adherend. The minimum principal stress Recessing a single-lap joint does not cause a signif-
in the adhesive was achieved by optimising the config- icant increase in the maximum stresses near the ad-
uration with a titanium upper adherend and adhesive hesive spew terminus. Since stress conditions do not
with a Young’s modulus of 2.6 Pa. The minimum peel change significantly with recessing, it appears that re-
stresses in the adhesive and the central CFRP adherend cessed joints should be considered a viable alternative
were obtained with the configuration having titanium for structural applications where weight and cost sav-
as an upper adherend and an adhesive a Young’s mod- ings are highly desirable. A joint with small effective
ulus of 5.0 GPa.The maximum tension allowable of lap length would be similar to a spot-welded joint, since
the epoxy adhesive is 40 MPa [3]. Rispler et al. [3] the adhesive is used only near the ends of the overlap
concluded by comparing the maximum principal stress [62]. From the standpoint of fatigue failure, small ef-
values obtained from the optimised models, that the fective lap length may not be desirable; however, it is
only optimised structure that come close to the maxi- possible to combine spot adhesive layers with other
mum tension allowable is the configuration with tita- types of fasteners to create a fatigue resistant joint. It
nium as upper adherend and an adhesive with a Young’s should be noted that large amounts of recessing creates

4793
localized stress peaks at the recess ends [62]. At the ing of Materials,” edited by R. W. Cahn, P. Haasen and E. J. Kramer
interface, on the loaded side of the joint, σxx and σyy (Wiley-VCH, Weinheim, Germany, 1997) Vol. 18, p. 724.
36. A . G U I U and M . E . R . S H A N A H A N , Int. J. Adhesion Adhe-
become compressive at the recess ends, which may help
sives 22 (2002) 415.
prevent crack propagation in that area of the joint. At 37. K . C H O , H . R . B R O W N and D . C . M I L L E R , J. Polym. Sci.
the unloaded side, large tensile stresses occur, which Polym. Phys. 28 (1990) 1699.
need to be accounted for. 38. E . B O U C H E R , J . P . F O L K E R S , M . H E R V E T and C .
C R E T O N , Macromolecules 29 (1996) 774.
39. A . B A L D A N , J. Materials Sci. 39 (2004) 4729.
References 40. H . H . K A U S C H and M . T I R R E L L , Annu. Rev. Mater. Sci. 19
1. F . L . M A T H E W S , “Handbook of Polymer Composites for En- (1989) 341.
gineeres” (Woodhead Publishing Limited, Cambridge, UK, 1999). 41. R . P . W O O L , B . L . Y U A N and O . M C G A R E L , J. Polym.
2. M . D A V I S and D . B O N D , Int. J. Adhesion Adhesives 19 (1999) Eng. Sci. 29 (1989) 1340.
91. 42. A . N . G E N T and A . A H A G O N , J. Polym. Sci. Phys. 13 (1975)
3. A . R . R I S P L E R , L . T O N G , G . P . S T E V E N and M . R . 1285.
W I N S O M , ibid. 20 (2000) 221. 43. A . N . G E N T and R . P . P E T R I C H , Proc. R. Soc. London A
4. L . J . H A R T - S M I T H , in “Joining Fiber Reinforced Plastics,” 310 (1969) 433.
edited by F. L. Mathews (Elsevier, London, 1987). 44. M . S H A N A H A N and C . B O U N G O S - M O N N I E R , Int. J. Ad-
5. J . F . P . O W E N S and P . L E E - S U L L I V A N , Int. J. Adhesion hesion Adhesives 16 (1996) 129.
Adhesives 20 (2000) 39. 45. A . N . G E N T and G . R . H A M E D , Polym. Eng. Sci. 17 (1977)
6. D . H . K O H L I , ibid. 19 (1999) 231. 462.
7. R . E . P O L I T I , SAMPE International Technical Conference 46. G . L I and P . L E E - S U L L I V A N , Int. J. Adhesion Adhesives 21
(1987) Vol. 19, p. 36. (2001) 211.
8. M E L M . S C H W A R T Z , “Composite Materials Handbook,” 2nd 47. P . B . L I N D L E Y , J. Inst. Rubber Ind. 5 (1971) 243.
ed. (McGraw-Hill, 1992) p. 6.39. 48. M . A . A N S A R I F A R , J . Z H A N G , J . B A K E R , A . B E L L
9. A . B . C A R T E R , SAMPE J. 25(4) (1989) 21. and R . J . E L L I S , Int. J. Adhesion Adhesives 21 (2001) 369.
10. L . J . H A R T - S M I T H , Int. J. Adhesion Adhesives 19 (1999) 181. 49. M . A . A N S A R I F A R , J . Z H A N G , A . B E L L and R . J .
11. E . P A U L D E G A R M O , J . T . B L A C K and R . A . K O H S E R , E L L I S , ibid. 22 (2002) 245.
“Materials and Processes in Manufacturing,” 8th ed. (Prentice-Hall, 50. D . W . A U B R E Y , G . N . W E L D I N G and T . W O N G , J. Appl.
1997) Chap. 8, p. 1051. Polym. Sci. 13 (1969) 2193.
12. A . H I G G I N S , Int. J. Adhesion Adhesives 20 (2000) 367. 51. C . D R A I L , A . A L L A L , G . M A R T I N and P .
13. J . K . J E T H W A and A . J . K I N L O C H , J. Adhesion 61 (1997) T O R D J E M A N , J. Adhesion 61 (1997) 123.
71. 52. C . G ’ S E L L and J . J . J O N E S , J. Mater. Sci. 14 (1977) 583.
14. C . H . W A N G and P . C H A L K L E Y , Int. J. Adhesion Adhesives 53. Idem., ibid. 16 (1981) 1956.
20 (2000) 155. 54. D . W . V A N K R E V E L E N , “Properties of Polymers: Their Cor-
15. R . N . H A W A R D , “The Physics of Glassy Polymers” (Applied relation with Chemical Structure” (Elsevier, Amsterdam, 1990).
Science Publishers, London, 1973). 55. J . V E R D U , Materiaux Polymers, Relations, Structure-Propriétés.
16. P . B . B O W D E N and J . A . J U K E S , J. Mater. Sci. 7 (1972) 52. Lectures ENSAM, Paris, 1992.
17. M . Y . M . C H I A N G and H . C H A I , Int. J. Solids Struct. 31 56. F . H . H A M M O D , ASTM Spec. Tech. Publ. 360 (1964) 123.
(1994) 2477. 57. D . K A L E L B L E , Trans. Soc. Rheol. 4 (1960) 45.
18. C . H . W A N G and L . F . R . R O S E , Int. J. Adhesion Adhesives 58. J . J O H N S O N , Adhesive Age 11 (1968) 20.
17 (1997) 17. 59. T . I G A R A S H I , J. Polym. Sci. Polym. Phys. Phys. Ed. 13 (1975)
19. J . D . F E R R Y , “Viscoelastic Properties of Polymers” (Wiley, 2129.
New York, 1980). 60. B . D . A G A R W A L and L . J . B R O U T M A N , “Analysis and
20. P . M O L I T O R , V . B A R R O N and T . Y O U N G , Int. J. Adhesion Performance of Fiber Composites,” 2nd ed. (John Wiley and Sons,
Adhesives 21 (2001) 129. Inc., 1990) p. 277.
21. L . J . H A R T - S M I T H , Adhesive-Bonded Single Lap Joints, 61. G . B E L I N G A R D I , L . G O G L I O and A . T A R D I T I , Int. J.
NASA CR 112236, January 1973. Adhesion Adhesives 22 (2002) 273.
22. M . K A L N I N S and J . O Z O L I N S , Int. J. Adhesion Adhesives 22 62. T . P . L A N G and P . K . M A L L I C K , ibid. 19 (1999) 257.
(2002) 179. 63. C . S A T O and K . I K E G A M I , ibid. 20 (2000) 17.
23. M . K A L N I N S , “Adhesive Interaction of Polyethyelene with 64. R . B A L K O V A , S . H O L O N E R O V A and V . C E C H , ibid. 22
Steel” (Riga. Latvia, Zinatne, 1990) p. 143. (2002) 291.
24. M . K A L N I N S , J. Adhesion 35 (1991) 173. 65. M . R . M A H E R I and R . D . A D A M S , ibid. 22 (2002) 119.
25. M . K A L N I N S and J . M A L E R S , ibid. 50 (1995) 83. 66. C . H . W A N G and L . R . F . R O S E , ibid. 20 (2000) 145.
26. M . R . B O W D I T C H , Int. J. Adhesion Adhesives 16 (1996) 73. 67. A . R . R I S P L E R and L . T O N G , ibid. 20 (2000) 221.
27. W . B R O C K M A N , The Environmental Resistance of Metal 68. J . C O T T O N , J . W . G R A N T , M . K . J E N S E N and B . J .
Bonds in New Industries and Applications for Advanced Materials L O V E , ibid. 21 (2001) 65.
Technology, 19th SAMPE Symp., Exhibition, Azuza, CA, 1974. 69. G . L I and P . L E E - S U L L I V A N , ibid. 21 (2001) 211.
28. M . M . A B D E L W A H A B , A . D . C R O C O M B E , A . 70. R . D . A D A M S , J . C O M Y N and W . C . W A K E , “Structural
B E E V E R S and K . E B T E H A J , Int. J. Adhesion Adhesives 22 Adhesive Joints in Engineering” (Kluwer Academic Publishers,
(2002) 61. London, 1997).
29. E . M . K N O X and M . J . C O W L I N G , ibid. 20 (2000) 323. 71. O . V O L K E R S E N , Luftfahrtforschung 15 (1938) 41.
30. S . S M I L T N I E K S and G . M . S P I N K S , in Proc. of Structural 72. M . G O L A N D and E . R E I S S N E R , J. Appl. Mech. 11A (1944)
Adhesives in Engineering (V. Bristol, UK, Inst. of Materials, 1998), 17.
p. 47, ISBN 1-86125-066-5. 73. W . L I , L . B L U N T and K . J . S T O U T , Int. J. Adhesion Adhe-
31. D . G . L E E , J . W . K W O N and D . H . C H O , J. Adhesion Sci. sives 19 (1999) 315.
Technol. 12 (1998) 1253. 74. L . F . M . D A S I L V A and R . D . A D A M S , ibid. 22 (2002)
32. A . J . K I N L O C H , in “Durability, of Structural Adhesives,” edited 311.
by A. J. Kinloch (Elsevier, Amsterdam, Applied Science, 1983). 75. L . D . R . G R A N T , The Characterisation of Adhesive Joints
33. E . M . K N O X and M . J . C O W L I N G , Int. J. Adhesion Adhesives Found Typical in the Automotive Industry, Ph.D. Dissertation, Uni-
20 (2000) 201. versity of Bristol, 1994.
34. D . J . A R R O W S M I T H and A . M A D D I S O N , ibid. 7 (1987) 76. R . A . S H E N O I and F . L . M . V I O L E T T E , J. Compos. Mater.
15. 34 (1990) 644.
35. C O N S T A N T I N O . C R E T O N , “Materials Science of Pressure 77. R . A . B A R T H O L O M E U S Z , A . S E A R L , A . A . B A K E R
Sensitive Adhesives, Materials Science and Technology; Process- and R . J . C H E S T E R , Bonded Composite Reinforcement of the

4794
 F/A-18 Y470.5 Bulkhead, Applications with Through Thickness 113. A . H . L A N D R O C K , “Adhesive Technology Handbook” (Noyes
Stresses. International Aerospace Congress, Sydney, 24–27 Feb. Publication, Park Ridge, NJ, 1985).
1997. 114. I . K U L E S H O V , S . K A I B I N and M . K A L N I N S , Vysokomol
78. C . H . W A N G and L . R . F . R O S E , Int. J. Adhesion Adhesives Soedin B 25 (1983) 366.
17 (1997) 17. 115. J . B I K E R M A N , J. Adhesion 3 (1972) 333.
79. M . Y . M . C H I A N G and H . C H A I , Int. J. Solids Struct. 31 116. J . A . H A R R I S and R . D . A D A M S , Int. J. Adhesion Adhesives
(1994) 2477. 4 (1984) 65.
80. R . B . H E S L E H U R S T , Int. J. Adhesion Adhesives 19 (1999) 117. A . J . K I N L O C H , J. Mater. Sci. 17 (1980) 617.
133. 118. W . B R O C K M A N , “Durability of Structural Adhesives,” edited
81. W . K . L O H , A . D . C R O C O M B E , M . M . A B D E L by A. J. Kinloch (Amsterdam, Elsevier Applied Science Publishers,
W A H A B and I . A . A S H C R O F T , Eng. Fracture Mechanics 69 1983), p. 306, Chap. 7.
(2002) 2113. 119. S . A . H A S H I M and M . J . C O W L I N G , in Proceedings of
82. A . J . K I N L O C H , J. Adhesion 10 (1979) 193. the Structural Adhesives in Engineering IV (Bristol, UK, Inst. of
83. J . W . W Y L D E and J . K . S P E L T , Int. J. Adhesion Adhesives Materials, 1995) p. 245.
18 (1998) 237. 120. W . A L T H O F , Verfahren zur Ermittlung von Schubspannungs
84. I . A . A S H C R O F T , M . M . A B D E L W A H A B , A . D . Gleitungs Diagrammen von Konstruktions Klebstoffen in dünnen
C R O C O M B E , D . J . H U G H E S and S . J . S H A W , Compos- Klebschichten, DFVLR IB 152-74/18, 1974.
ites, Part A 32 (2001) 45. 121. R . B . K R I E G E R , “Stiffness Characteristics of Structural Ad-
85. M . M . A B D E L W A H A B , I . A . A S H C R O F T , A . D . hesives for Stress Analysis in Hostile Environment” (American
C R O C O M B E , D . J . H U G H E S and S . J . S H A W , ibid. 32 Cyanamid Co., Havre de Grace, MD, 1975).
(2001) 59. 122. R . D . A D A M S and J . O P P E N D A L E , J. Mech. Eng. Sci. 18
86. D . B R E W I S , J . C O M Y N and J . T E G G , Int. J. Adhesion (1976) 149.
Adhesives 1 (1980) 35. 123. J . D . B A R D I S and K . T . K E D W A R D , Department of Me-
87. J . C R A N K and G . P A R K (eds), “Diffusion in Polymers” chanical and Environmental Engineering University of California,
(New York, Academic Press, 1968). Santa Barbara, CA 93106.
88. D . R . L E F E B U R E , D . A . D I L L A R D and T . C . W A R D , 124. J E A N S - Y V E S S E N E R , T . F E R R A C I N , L . C A U S S I N and
J. Adhesion 27 (1989) 1. F . D E L A N N A Y , Int. J. Adhesion Adhesives 22 (2002) 129.
89. A . K A U L , N . S U N G , I . C H I N and C . S U N G , Polym. Eng. 125. P . B . C R O S L E Y and E . J . R I P L I N G , J. Test. Eval. 19 (1991)
Sci. 24 (1984) 493. 24.
90. B . R . K . B L A C K M A N , J . P . D E A R , A . J . K I N L O C H , 126. D . J . C H A N G , R . M U K I and A . W E S T M A N N , Int. J. Solids
H . M A C G I L L I V R A Y , Y . W A N G , J . G . W I L L I A M S and and Structures 12 (1976) 13.
P . Y A Y L A , J. Mater. Sci. 30 (1995) 5885. 127. F . E . P E N A D O , J. Comp. Mater. 27 (1993) 383.
91. R . H . A N D R U E T , D . A . D I L L A R D and S . M . H O L Z E R , 128. A . K . E L - S E N U S S I and J . P . H . W A B B E R , J. Appl. Phys.
Int. J. Adhesion Adhesives 21 (2000) 17. 56 (1984) 885.
92. J . P . M . G O N Ç A L V E S , M . F . S . D E M O U R A and P . 129. J . C O G N A R D , J. Adhesion 57 (1996) 31.
M . S . T . D E C A S T R O , ibid. 22 (2002) 357. 130. S . J . B E N N I S O N , G . A . A N D R E J A C K and L . A .
93. O . V O L K E R S E N , Constr. Met. 4 (1965) 3. S I L V E R M A N , in Proc. Euradh’96 (Cambridge University Press,
94. L . J . S E G E R L I N D , J. Appl. Mech. 35 (1968) 177. 1996) p. 439.
95. W . J . R E N T O N and J . R . V I N S O N , ibid. 44 (1977) 131. E . J . R I P L I N G , S . M O S T O V O Y and R . L . P A T R I C K ,
101. Mater. Res. Standards (ASTM Bulletin) 4(3) (1964) 129.
96. I . U . O J A L V O and H . L . E I D I N O F F , AIAA J. 16 (1978) 132. S . M O S T O V O Y , P . B . C R O S L E Y and E . J . R I P L I N G , J.
204. Mater. 2 (1967) 661.
97. D . A . B I G W O O D and A . D . C R O C O M B E , Int. J. Adhesion 133. ASTM, ASTM D3433, in Annual Book of ASTM Standards: Ad-
Adhesives 9 (1989) 229. hesion Section 15, Philadalphia, 1990.
98. G . R . W O O L E Y and D . R . C A R V E R , J. Aircraft 8 (1971) 134. B . R . K . B L A C K M A N , H . H A D A V I N I A , A . J .
817. K I N L O C H , M . P A R A S C H I and J . G . W I L L I A M S , Eng.
99. T . R . G U E S S , R . E . A L L R E D A and F . P . G E R S T L E , J. Fracture Mech. 70 (2003) 233.
Test. Eval. 5 (1977) 84. 135. J . C O G N A R D , J. Adhesion 22 (1987) 97.
100. N . I . H A R R I S O N and W . J . H A R R I S O N , J. Adhesion 3 136. Idem., ibid. 20 (1986) 1.
(1972) 195. 137. E . J . R I P L I N G , S . M O S T O V O Y and C . B E R S C H , ibid. 3
101. R . D . A D A M S and N . A . P E P P I A T T , J. Strain Anal. 9 (1974) (1971) 145.
185. 138. F . S L O A N , J. Comp. Mater. 27 (1993) 1606.
102. D . W . O P L I N G E R , Int. J. Solids Struct. 31 (1994) 2565. 139. W . S . J O H N S O N and L . M . B U T K U S , Fatigue and Fract.
103. L . J . H A R T - S M I T H , “Adhesive Bonded Single Lap Joints,” Engng. Mater. Struct. 21 (1988) 465.
NASA Contractor Report 112236, 1973. 140. J O N G - K Y O K I M and Y I U - W I N G M A I , “Interfaces in Com-
104. D . W . O P L I N G E R , A Layered Beam Theory For Single Lap posites,” Materials Science and Technology, edited by R. W. Cahn,
Joints. Army Materials Technology Laboratory Report MTL TR P. Haasen and E. J. Kramer, Structure and Properties of Composites
91-23, 1991. (VCH, Weinheim, 1993) Vol. 13, p. 241.
105. M . V . T S A I and J . M O R T O N , Int. J. Solids Struct. 31 (1994) 141. M . A . A N S A R I F A R and G . J . L A K E , Failure of Rubber-to-
2537. Metal Bonded Units, Adhesion 93, The Fifth Int. Conference of
106. A . T O W S E , K . D . P O T T E R , M . R . W I N S O M and R . D . the Adhesives, Section of the Polymer Industry Divisionn of the
A D A M S , Int. J. Adhesion Adhesives 19 (1999) 71. Institute of Materials, York, UK, 1993.
107. R . B . H E S L E H U R S T , Repair of Composite and Bonded Air- 142. T . S C H Ü L L E R and B . L A U K E , Int. J. Adhesion Adhesives 22
craft Structures, Course Notes. Australia Defence Force Academy. (2002) 169.
108. R . B . H E S L E H U R S T , J . P . B A I R D , H . M . 143. British Standard, BS 903, Part A21, 1989.
W I L L I A M S O N and R . K . C L A R K , Can Aging Adhesively 144. ASTM: D429-81. Annual Book of ASTM Standards, vol. 09.01,
Bonded Joints be Found? in Proc. 41st SAMPE Int. Symp. And 1993.
Exhibition, Anaheim, CA, 1996, p. 925. 145. ASTM: D429-64, Method B and ASTM: D429-64, Method C.
109. Locitite Corporation. In: Locitite Worlwide Design Handbook, 146. T . S C H Ü L L E R , W . B E C K E R T , B . L A U K E and K .
Edition 1996/1997. F R I E D R I C H , Comp. Sci. Technol. 60 (2000) 2077.
110. M . A . A N S A R I F A R and G . J . L A K E , J. Adhesion 53 (1995) 147. B . R . K . B L A C K M A N , A . J . K I N L O C H , A . C .
183. T A Y L O R and Y . W A N G , J. Mater. Sci. 35 (2000) 1867.
111. C . X U , K . R A M A N I and G . K U M A R , Int. J. Adhesion Ad- 148. P . A . F A Y , R . E . D A V I S and G . D . S U T H U R S T , in Proc.
hesives 22 (2002) 187. of Impact and Fatigue Testing of Adhesives (Plastics and Rubber
112. W . F U N K E , J. Coat. Technol. 55 (1985) 31. Institute, London, 1990).

4795
149. R . A . D I C K I E and S . M . W A R D , in “Mittal Festschrift,” 188. J . N . R E D D Y , “An Introduction to the Finite Element Method,”
edited by W. J. van Ooij and H. R. Anderson (VSP, Zeist, 1997) 2nd ed. (McGraw-Hill, New York, 1993).
p. 1. 189. K . J . B A T H E , “Finite Element Procedures” (Prentice-Hall,
150. International Standard Organisations, ISO 11343 (ISO, Geneva, Englewood Cliffs, NJ, 1996).
1993). 190. H . L . G R O W T H , Int. J. Adhesion Adhesives 8 (1988) 107.
151. L . T O N G and G . P . S T E V E N S , “Analysis and Design of Struc- 191. A . D . C R O C O M B E , D . A . B I G W O O D and G . R I C H A R D ,
tural Bonded Joints” (Kluwer, Dortrecht, 1999. ibid. 12 (1990) 167.
152. A . D . C R O C O M B E and R . D . A D A M S , J. Adhesion 13 (1981) 192. R . D . A D A M S , J. Adhesion 30 (1989) 219.
141. 193. T . S C H Ü L L E R , W . B E C K E R T , B . L A U K E , C .
153. R . D . A D A M S , R . W . A T K I N S , J . A . H A R R I S and A . A G E O R G E S and K . F R I E D R I C H , Composites A 31 (2001)
J . K I N L O C H , ibid. 20 (1986) 29. 661.
154. H . L . G R O T H , Int. J. Adhesion Adhesives 11 (1991) 204. 194. ABAQUS, General-Purpose Finite Element System, Hibbit,
155. R . D . A D A M S and N . A . P E P P I A T T , J. Strain Anal. 9 (1974) Karlson and Sorensen, Inc., 100 Medway Street, Providence, RI,
185. 1997.
156. L . D O R N and L . W E I P I N G , Int. J. Adhesion Adhesives 13 195. S . K . M A Z U M D E R and P . K . M A L L I C K , Polym. Comp. 19
(1993) 21. (1997) 139.
157. M . H I L D E B R A N D , ibid. 14 (1994) 261. 196. Idem., in Proc. 12th Annual Tech. Conf., American Society for
158. T . P . L A N G and P . K . M A L L I C K , ibid. 18 (1998) 167. Composites, October 1997, “Static and Fatigue Properties of Ad-
159. Idem., ibid. 19 (1999) 257. hesive Joints in SRIM-SRIM Composites.”
160. R . D . A D A M S and J . A . H A R R I S , ibid. 16 (1996) 61. 197. R . D . A D A M S and V . M A L L I C K , J. Adhesion 38 (1992) 199.
161. J . A . H A R R I S and R . D . A D A M S , An Assessment of the 198. L . D O R N and L . W E I P I N G , Int. J. Adhesion Adhesives 13
Impact Performance of Bonded Joints for Use in High Energy (1993) 21.
Absorbing Structures, in Proc. Inst. Mech. Engrs. 199 (1985) 121. 199. M . F . S . F . D E M O U R A , J . P . M . G O N Ç A L V E S , A .
162. A . J . K I N L O C H and G . A . K O D O K I A N , J. Adhesion 24 T . M A R Q U E S and P . M . S . T . D E C A S T R O , J. Compos.
(1987) 109. Mater. 31 (1997) 1462.
163. J . L . L A T A I L A D E and C . K E I S L E R , Charobonnet, ph. Effect 200. J . P . M . G O N Ç A L V E S , M . F . S . F . D E M O U R A , P . M .
of the Substrate on the Adhesion in Adhesive Bonds Loading at S . T . D E C A S T R O and A . T . M A R Q U E S , Eng. Comput.:
High Strain Rate, Preprint EURADH 92 (1992) 585. Int. J. Comput-Aided Eng. Software 17 (2000) 28.
164. F . C A Y S S I A L S and J . L . L A T A I L L A D E , J. Adhesion 58 201. R . S . R A G H A V A , R . M . C A D E L L and G . S . Y E H ,
(1996) 584. J. Mater. Sci. 8 (1973) 225.
165. T . Y O K O Y A M A , “A New Test Technique for Determining the 202. W . R . B R O U G H T O N , Shear Properties of Unidirectional Car-
Impact Shear Strength of Adhesive Joints,” in Proceedings of the bon Fiber Composites. Ph.D. Thesis, Darwin College, Cambridge.
ATEM’ 97, 1997, p. 221. 203. G . W . W Y C H E R L E Y , S . A . M E S T A N and I .
166. A . A . B E Z E M E R , C . B . G U Y T and A . V O L T , Int. J. Adhe- G R A B O V A C , J. Test. Eval. 18 (1990) 203.
sion Adhesives 18 (1998) 255. 204. A . G U I U and M . E . R . S H A N A H A N , J. Polym. Sci. Polym.
167. L . W . Z A C H A R Y and C . P . B U R G E R , Exp. Mech. 20 (1980) Phys. 39 (2001) 2843.
162. 205. E . M . K N O X , Marin Applications for Structural Adhesives,
168. K . R . B E R G , “Problems in the Design of Joints and Attach- Ph.D. Thesis, Glasgow, University, Glasgow, UK, 1996.
ments,” edited by F. W. Wendt, H. L. Liebowitz and N. Perrone 206. P . T . R E Y N O L D S , J. Mater. Sci. Lett. 4 (1985) 1194.
(Mechanics of Composite Materials, Pergamon, New York, 1970). 207. D . M A V G I S and M . B A R Q U I N S , “Adhesion,” edited by K.
169. J . H A R T - S M I T H , Adhesive-Bonded Double Lap Joints, NASA W. Allen (Elsevier, London, 1988) Vol. 12, p. 205.
Contractual Report, NASA CR-112235, January 1973. 208. C . D R A I L , A . A L L A T , G . M A R T I N and P .
170. R . B . H E S L E H U R S T , Ph.D. Dissertation, School of Aerospace T O R D J E M A N , J. Adhesion 68 (1998) 203.
and Mechanical Engineering, University College, UNSW, ADFA, 209. F . X . G I B E R T , A . A L L A T , G . M A R I N and C . D R A I L ,
Canberra, 1998. J. Adhesion Sci. Technol. 13 (1999) 1029.
171. D . A . B I G W O O D and A . D . C R O C O M B E , Int. J. Adhesion 210. R . J . R O A R K , “Roark’s Formulas for Stress and Strain,” 6th ed.
Adhesives 10 (1990) 31. (McGraw-Hill, New York, London, 1989).
172. G . R I C H A R D S O N , A . D . C R O C O M B E and P . A . S M I T H , 211. A . J . K I N L O C H , “Adhesion and Adhesives” (Chapman and
ibid. 13 ( 1993) 193. Hall, London, Glasgow, 1994).
173. M . V . T S A I , J. Morton, Mech. Mater. 20 (1995) 183. 212. A . J . K I N L O C H , Proc. Int. Mech. Eng. 211 (Part G) (1997) 307.
174. M . V . T S A I , J. Morton, Comp. Struct. 32 (1995) 123. 213. P . M O L I T O R and T . Y O U N G , Int. J. Adhesion Adhesives 22
175. M . Y . T S A I and J . M O R T O N , J. Comp. Mater. 29 (1995) 1163. (2002) 101.
176. M . Y . T S A I , D . W . O P L I N G E R and J . M O R T O N , 214. S . B I S T A C , M . F . V A L L A T and J . S C H U L T Z , ibid. 18
Int. J. Solids Struct. 35 (1998) 1163. (1998) 365.
177. W . C . C A R P E N T E R , J. Adhesion 35 (1991) 55. 215. D . M . B R E W I S , G . W . C R I T L O W and C . A . C U R T I S ,
178. W . K . C H I U and R . J O N E S , Int. J. Adhesion Adhesives 12 ibid. 19 (1999) 253.
(1992) 219. 216. A . N . R I D E R and D . R . A R N O T T , ibid. 20 (2000) 209.
179. M . Y . T S A I and J . M O R T O N , J. Strain Anal. 29 (1994) 137. 217. S . U . N I N G , R . I . M A C K I E and W . J . H A R V E Y , ibid.
180. C . H . W A N G and L . R . F . R O S E , J. Adhesion Adhesives 17 12 (1992) 85.
(1997) 17. 218. M . P . Z A N N I - D E F F A R G E S and M . E . R . S H A N A H A N ,
181. G . L I , P . L E E - S U L L I V A N and R . W . T H R I G , Comp. ibid. 13 (1993) 41.
Struct. 46 (1999) 395. 219. R . E . D A V I S and P . A . F A Y , in Proc. of the Adhesion’90,
182. F . P . B E E R and R . J O H N S O N , “Mechanics of Materials,” Cambridge, UK, The Plastics and Rubber Institute, 1990, p. 27/1-
Matric. ed. (McGraw-Hill, England, 1992). 27/6; ISBN 1 871571 219.
183. S . Y A D I G A R I , C . P A P I R E D D Y and T . S A N J E E V A 220. R . A . G L E D H I L L and A . J . K I N L O C H , J. Adhesion 6 (1974)
R E D D Y , Comp. Struct. 27 (1987) 445. 315.
184. W . C . C A R P E N T E R and R . B A R S O U M , J. Adhesion 30 221. E . M . K N O X , Ph.D. Thesis, Glasgow University, Glasgow, UK,
(1989) 25. 1996.
185. S . A M I J I M A and T . F U J I I , Int. J. Adhesion Adhesives 7 (1987) 222. P . A . F A Y and A . M A D D I S O N , in Procs. of Structural Adhe-
199. sives in Enginerring II (Butterworths Scientific, Bristol, UK, 1989)
186. M . V . T A Y L O R , M.Sc. Degree, Engineering Mechanics De- p. 112.
partment of Virgina Polytechnic Institute and State University, 223. M . G E T T I N S and A . J . K I N L O C H , J. Mater. Sci. 12 (1977)
Blacksburg, VA, 1996. 2511.
187. S . N A B O U L S I and S . M A L L , Theoret. Appl. Fracture Mech. 224. D . A . T O D , R . W . A T K I N S and S . J . S H A W , Int. J.
26 (1997) 1. Adhesion Adhesives 12 (1992) 159.

4796
225. S . P . B H U N I Y E and S . M A I T I , Europ. Polym. J. 38 (2002) 250. M . C H E I K H , P . C O O R E V I T S A and A . L O R E D O , Int. J.
195. Adhesion Adhesives 21 (2001) 249.
226. A . J . K I N L O C H , J. Mater. Sci. 12 (1980) 2141. 251. A . J . K I N L O C H , M . S . G . L I T T L E and J . F . W A T T S ,
227. M . B A B I C , Int. J. Adhesion Adhesives 21 (2001) 11. Acta Mater. 48 (2000) 4543.
228. D . K . K O H L I , in Proc. of 10th Int. Eurepean Chapter Conference 252. A . M A H O O N , “Titanium Adherends in Durability of Structural
of the SAMPE 55 (1989) 239. Adhesives,” edited by A. J. Kinloch (1983) p. 259.
229. D . K . K O H L I , 39th International SAMPE Symposium, 11–14 253. A . M A D D I S O N , in Proc. of the Structural Adhesives in Engi-
April 1994. neering V (Inst. of Materials, Bristol, UK, 1998) p. 68.
230. International Standard: ISO 813-1986 (E). 254. Y . W E I T S M A N , J. Comp. Mater. 11 (1977) 378.
231. K . B O N D , Adhesive Age 33(Part 2) (1996) 22. 255. M . R . B O W D I T C H , Int. J. Adhesion Adhesives 16 (1996)
232. L . K O V A V I C , Lepenie Kovov a Plastov, Praha, SNTL, 1980. 73.
233. I . S K E I S T (ed), “Handbook of Adhesives” (Chapman and Hall, 256. R . A . G L E D H I L L and A . J . K I N L O C H , J. Adhesion 6 (1974)
New York, 1990). 315.
234. Various Bae Internal Studies of Long Term Warm/Wet Test Pieces 257. M . P . Z A N N I - D E F F A R G E S and M . E . R . S H A N A H A N ,
of Hexcel Redux 775, Hexcel Redux 308 A, Cytec FM 73 and 3M Int. J. Adhesion Adhesives 15 (1995) 1520.
Co, AF 163-2 Adhesives. 258. C . S . S M I T H , “Design of Marine Structures in Composite Ma-
235. D . D I X O N , Prediction of Lap Shear and Peel Strength of Adhe- terials” (Applied Science, Amsterdam, Elsevier, 1990) Chap. 3,
sive Bonds, and Update on Results of Accelerated Environmental p. 99.
Testing, Bae Sowerby Research Centre, Report No. 3S 13729, May 259. R . D . A D A M S , “Structural Adhesives in Engineering,” C180/86,
1997, p. 3. (I. Mech. E., London, 1986), p. 17.
236. J . C O M Y N , “Adhesion Science” (The Royal Society of Chem- 260. A . D . C R O C O M B E , Int. J. Adhesion Adhesives 17 (1997) 229.
istry, UK, 1997). 261. Y . M . X I E and G . P . S T E V E N , Comput. Struct. 49 (1994)
237. P . A . C O O P E R and J . W . S A W Y E R , Report No. TP-1507, 885.
NASA, 1979. 262. Idem., Eng. Comput. 11 (1994) 295.
238. M . Y . T S A I and J . M O R T O N , Int. J. Solids Struct. 31 (1994) 263. I . U . O J A L V O and H . L . E I D O N O F F , AIAA J. 16 (1978)
2537. 204.
239. R . M . B A R K E R and F . H A T T , AIAA 11 (1973) 1650. 264. W . J . R E N T O N and J . R . V I N S O N , J. Adhesion 7 (1975)
240. W . C . C A R P E N T E R , Int. J. Numer. Methods Eng. 6 (1973) 430. 175.
241. Idem., ibid. 15 (1980) 1659. 265. F . E R D O G A N and M . R A T W A N I , J. Comp. Mater. 5 (1971)
242. W . C . C A R P E N T E R and R . B A R S O U M , J. Adhesion 30 378.
(1989) 25. 266. R . C . G I V L E R and R . B . P I P E S , “Analysis of the ‘Joggle-
243. B . N . R A O , Y . K . R A O and S . Y A D I G A R I , Fibre Sci. Lap’ Joint for Automotive Applications,” edited by K. T. Kedward
Technol. 17 (1982) 77. (Joining of Composite Materials, Vol. ASTM STP 749, ASTM,
244. S . Y A D I G A R I , C . P . R E D D Y and T . S . R E D D Y , Comput. 1985), p. 61.
Struct. 27 (1987) 445. 267. M . O L I A and J . N . R O S S E T T O S , Int. J. Solids Struct. 33
245. C . C . L I N and Y . S . L I N , Int. J. Solids Struct. 30 (1993) 1679. (1996) 2681.
246. J . N . R E D D Y and S . R O Y , Int. J. Non-Linear Mech. 23 (1988) 268. J . N . R O S S E T T O S , P . L I N and H . N A Y E B - H A S H E M I ,
97. J. Engng. Mater. Technol. 116 (1994) 533.
247. S . R O Y and J . N . R E D D Y , Int. J. Numer. Methods Eng. 26 269. J . N . R O S S E T T O S and E . Z A N G , J. Appl. Mech. 60 (1993)
(1988) 2531. 559.
248. P . C . P A N D E Y and S . N A R A S I M H A N , Comput. Struct. 79
(2001) 769.
249. U . E D L U N D and A . K L A R B R I N G , Comput. Methods Appl. Received 10 September 2003
Mech. Eng. 96 (1992) 329. and accepted 1 April 2004

4797