International Journal of Adhesion and Adhesives 75 (2017) 101107

Contents lists available at ScienceDirect

International Journal of Adhesion and Adhesives

journal homepage: www.elsevier.com/locate/ijadhadh

The analysis of kissing bonds in adhesive joints

,2
MARK
1
C. Jeenjitkaew , F.J. Guild
School of Engineering and Materials Science, Queen Mary, University of London Department of Materials, Mile End Road, London E1 4NS, United Kingdom

A R T I C L E I N F O A B S T R A C T

Keywords: Kissing bonds pose a serious threat and an obstacle to the application of adhesive bonding in engineering
C: Finite element stress analysis structures. This paper describes an investigation of the formation of kissing bonds using various surface
D: Cohesive zone model contaminants. Double-lap joints have been prepared using several contaminants, and tested in quasi-static
Kissing bonds tension. The values of shear strength obtained from these tests are successfully compared with observations of
Double-lap joint
the interfaces and fracture surfaces. Finite element analysis in combination with cohesive elements was used to
simulate the adhesive joint tests. Fracture properties used in the analysis were obtained from independent
material tests. The results indicate very good agreement between experimental and predicted shear strengths for
both control and contaminated joints. The results of the validated nite element model are explored to nd a
strain signature that could be used to detect kissing bonds. The results indicate that measurement of local
lateral strain may form the basis of a technique for monitoring or detection of kissing bonds.

1. Introduction reliable method to detect kissing bonds but this method may cause
damage both around any kissing bond or to composite adherends
Kissing bonds are interfacial weak bonds present when the substrate [12,13].
surface and adhesive are in intimate contact but only form weak Recent advances in Digital Image Correlation (DIC), including three-
adhesion. The possible occurrence of kissing bonds poses a serious dimensional techniques, may be able to detect damage events in joints.
threat in the application of adhesive bonding in modern aircraft Both 3-dimensional and high magnication 2-dimensional DIC mea-
construction. The requirement for additional riveting of adhesive joints surements have been validated using strain gauges for both pre-preg
in structural joints, for example in aircraft, adds to weight and weakens and non-crimp-fabric carbon bre reinforced composite single lap
the surrounding material. Kissing bonds can arise from three main joints. The authors suggested that measurement of out-of-plane dis-
categories of fault: lack of full curing of the adhesive, surface placements using 3-dimensional DIC could be used to detect kissing
contamination, or physical damage of the surface. Surface contamina- bonds [14]. Kissing bonds, prepared using release lm, in unidirec-
tion can occur either during joint preparation or via subsequent ingress tional glass bre reinforced single lap joints have been examined using
by, for example, water. This paper only considers surface contamina- 2-dimensional DIC [15]. The contours of axial strain showed localisa-
tion introduced via various contaminants. tion around the kissing bond at around 50% failure load. The values of
Non-Destructive Testing (NDT) of kissing joints remains a subject of strain were reduced to negative values in these areas; the contours for
interest in the literature, and various methods have been developed for the healthy joints showed the expected variability for single lap joints.
their detection. These techniques include neutron radiography [1], Finite element analysis (FEA) predicted similar values of negative axial
infrared thermography [2] and holographic interferometry [3,4]. The strain above the kissing bond. The authors suggest that further work
most promising technique is probably ultrasonic testing, which can be may establish monitoring of axial strain as a means of kissing bond
divided into two groups depending on whether it interrogates the out- detection [15].
of-plane or in-plane stress-strain characteristics [5]. In-plane techniques This paper describes an investigation of the formation of kissing
include shear-wave resonance and guided wave methods, which may be bonds using various surface contaminants. The overall objective of this
sensitive to kissing bonds since it may detect change of stiness on the paper is to demonstrate that the strength of joints containing kissing
interface [5,6]. High powered out-of-plane non-linear ultrasonic tech- bonds can be predicted using FEA, and that local measurement of strain
nique may be able to detect kissing bonds, particularly at low contact around kissing bonds could form a basis for their monitoring or
pressures [711]. The laser shockwave technique may be the most detection. The joints have been tested and simulated using FEA. The

Corresponding author.
1
Now at: Element Hitchin, Wilbury Way, Hitchin, Hertfordshire, UK, SG4 0TW.
2
Now at: Department of Mechanical Engineering, Imperial College London, London SW7 2AZ.

http://dx.doi.org/10.1016/j.ijadhadh.2017.02.019
Accepted 12 February 2017
Available online 24 February 2017
0143-7496/ 2017 Elsevier Ltd. All rights reserved.
C. Jeenjitkaew, F.J. Guild International Journal of Adhesion and Adhesives 75 (2017) 101107

Table 1
Elastic properties of adherends and Redux319 adhesive.
Articial Sweat (Pickering Laboratories). This articial sweat is used
by many industries to test products such as textiles and jewellery. It
contains around 19 naturally occurring amino acids, 10 minerals
Material Young's modulus (GPa) Yield stress (MPa) Poisson's ratio
and four metabolites.
Hardened Steel
Al2014T6
214
76.1
1039
396
0.29
0.33
Cutting Oil lubricant So-Cool 12 S (Solent Oil Ltd.). This contami-
nant has been found to be detectable in ultrasonic studies [7] using
Redux319 3.76 21.8 0.33
measurement of ultrasonic nonlinearity. It may be suitable for
producing kissing bonds.
results of the validated FEA are explored to nd a strain signature that
could be used to detect kissing bonds. The results indicate that The geometry of the double-lap joints is illustrated in [17]. The
measurement of local lateral strain may form the basis of a technique adhesive thickness was 0.2 mm. The adherend thickness was 1.6 mm
for monitoring or detection of kissing bonds. Local strain around a for the HDS adherends and 2 mm for the Al2014-T6 adherends. The
kissing bond has been previously suggested as a means of their width was 25.4 mm with an overlap length of 12.7 mm. The
detection, as discussed above, but no consideration of the monitoring contaminant covered 25% of the bonding area, and was placed on the
of local lateral (Poisson's ratio) strain has been found. outer adherends of the double-lap joints since this is the failure path
observed for the control joints. Observation during the contamination
2. Materials and kissing bonds and examination of the failure surfaces (see section 4.2.5) indicated
that the contaminants did not migrate outside the contaminated region.
Double-lap joints were prepared using two metal adherends: Most testing was carried out using the high temperature adhesive
hardened steel (HDS) and aluminium alloy (Al 2014-T6). The hardened without the carrier (Redux319) since the carrier lm interfered with
steel, as a benchmark, was prepared in the laboratory from gauge steel the failure processes, as described above. The full range of tests
(MSC/J & L Industrial Supply) using heating and quenching [16]. The reported here is shown in Table 2. In addition, control specimens
aluminium alloy (Metalfast, UK), which is widely used in high strength without contaminant were tested for each case. At least 5 replicates
and high temperature applications, was also used in this study. The were tested for all conditions except the release lm.
elastic properties of the adherends are shown in Table 1. The same
aluminium alloy joints were tested in a previous study which describes 3. Experimental techniques
the surface chemistry of the contaminated joints [17].
The HDS adherends were cleaned with acetone and then treated The double-lap joints were tested using an Instron 6025 universal
with P320 silicon carbide paper and cleaned with acetone. Pressurised testing machine according to the standard, ASTM D3528 [18]. Tests
air was blown across the surface to remove any loose particles. Finally, were carried out at a displacement rate of 2 mm/min under standard
the surfaces were cleaned with acetone. The Al2014-T6 adherends were laboratory conditions (23 2 C and 55 10% relative humidity).
prepared according to Defence Standard 03-2/1 as described in [17]. Local strain was measured using 1 mm strain gauges on the outer
The same two types of adhesive were used as in the previous study adherends centred on the bond. Strain across the whole bond was
[17]. The rst adhesive is a high temperature cure epoxy lm adhesive measured using a 25 mm extensometer placed symmetrically around
with and without the carrier, Redux319 A and Redux319 (Hexcel, the bonded area.
UK). It was found (see section 4) that the carrier lm could interfere
with the failure processes, so the analysis was carried out for the 4. Experimental results
adhesive without the carrier. The mechanical properties of this
adhesive are shown in Table 1. The second adhesive is a room- 4.1. Interface analysis
temperature cure two-part epoxy paste adhesive, E3348 (Georgia-
Pacic Corp., USA). This was used to investigate whether migration The interfaces from the Al2014-T6 adherends using the two
of contaminants diered between high temperature and room tempera- adhesives have been compared and found to be similar for both
ture adhesives, and was used for the Al2014-T6 adherends only. adhesives [17]. The Frekote 700NC remains at the interface. Both the
The same contaminants were used as in the previous study [17]: articial sweat and the cutting oil lubricant were found to migrate from
the interface into the adhesive, although there may be some salts from
PTFE lm (Aerovac Ltd.). The lm thickness was 25 m. the articial sweat at the interface. The cutting oil lubricant appears to
PTFE spray (Screw Fix Direct). This is widely used as a lubricant cause morphological irregularities at the interface and within the
adhesive. This migration of these contaminants does not depend on
spray for rotor cutters and saw blades. It consists of micron-size
PTFE particles dispersed in a volatile organic solvent. the temperature of the cure of the adhesive. These observations are
Frekote 700-NC mould release agent (Loctite Corp.). This is a semi- conrmed in Section 4.2 below in the comparison of failure loads.
Using the same techniques described previously [17] the interfaces
permanent mould release agent, containing synthetic isoparanic
hydrocarbon, light aliphatic solvent naphtha, dibutyl ether and from the HDS adherends, with the Redux 319 high temperature
polydimethylsiloxane (PDMS). PDMS is one of the most common adhesive, contaminated with PTFE spray and Frekote 700NC were
contaminants found in organic samples. examined and are shown in Fig. 1. The PTFE spray has migrated into
the adhesive; the micron sized particles of PTFE can be detected. In

Table 2
Test matrix.

Adherend Adhesive Contaminants

PTFE lm PTFE spray Frekote 700NC Articial sweat Cutting oil lubricant

HDS Redux319 A x x x x
HDS Redux319 x x
Al2014-T6 Redux319 x x
Al2014-T6 E3348 x x

102
C. Jeenjitkaew, F.J. Guild International Journal of Adhesion and Adhesives 75 (2017) 101107

Frekote
PTFE spray
Stteel

Redux3
319

Contaminate interface Fig. 3. Eect of adherend metal on shear strength.

Fig. 1. Interface analysis of HDS/Redux 319 adhesive joints.

contrast, the Frekote 700NC remains at the interface.

4.2. Experimentally measured values of shear strength

The experimental results are presented as values of shear strength.

This is calculated from the value of maximum load in tension divided by
the bonded area. All experimental results are plotted with error bars;
the bars show one standard deviation. Thus if error bars overlap, results
from dierent tests are not signicantly dierent.

4.2.1. Eect of carrier

The values of shear strength for the control specimens and the
Fig. 4. Eect of adhesive and contaminant on shear strength for Al2014-T6 adherends.
specimens including PTFE lm are compared for the Redux319 A
adhesive, including the carrier, and the Redux319 adhesive, without
control joint. However, for both adhesives, the values for the articial
the carrier in Fig. 2. The adherends were hardened steel. It is clear that
sweat and cutting oil lubricant contaminants are the same values,
the carrier reduces the shear strength both for the control and the
within experimental error, as the control joints.
weakened bonds. Observation of the failure surface revealed that the
These results conrm the observations of the interface analysis [17].
carrier cloth was interfering with the failure path. All further testing
The Frekote 700NC remains at the interface. The articial sweat and the
was carried out using Redux319 adhesive, without the carrier.
cutting oil lubricant migrate into the adhesive, and do not cause a
reduction in shear strength. The migration of contaminants is not
4.2.2. Eect of adherends aected by the temperature of adhesive cure.
The shear strength for the control specimens and the specimens
contaminated with Frekote 700NC with the steel or aluminium
4.2.4. Eect of Frekote 700NC
adherends are compared in Fig. 3. The values of shear strength are
The results reported in Section 4.2.3 show that Frekote 700NC
reduced for the contaminated joints. The values for both the control and
causes reduction of shear strength for both adhesives. Fig. 5 compares
contaminated joints are the same for the two adherends, within the
the values of shear strength for HDS adherends with PTFE release lm,
experimental error. These results show that this joint geometry is a
PTFE spray and Frekote 700NC. The PTFE spray does cause some
good test for kissing joints.
reduction in shear strength, probably arising from some PTFE micron-
sized particles remaining at the interface or causing a weak failure path,
4.2.3. Eect of adhesives and contaminants
see Fig. 1. The value obtained from the release lm is the same, within
The results from the two adhesives are compared for the Al2014-T6
experimental error, as the value obtained from Frekote 700NC. This
adherends in Fig. 4. All values of shear strength are lower for the low
implies that Frekote 700NC causes zero adhesion between the surfaces,
temperature cure adhesive. For both adhesives, the value for the
and as a consequence this result has been used in the modelling of the
Frekote 700NC contaminated joint is signicantly lower than for the
joints, as described in Section 5 below.

Fig. 2. Eect of carrier cloth on shear strength. Fig. 5. Eect of contaminant on shear strength for HDS adherends.

103
C. Jeenjitkaew, F.J. Guild International Journal of Adhesion and Adhesives 75 (2017) 101107

Fig. 6. Failure surfaces from control and contaminated joints.

4.2.5. Failure surfaces thickness symmetry of the double-lap joint. The geometry and global
The results deduced in Section 4.2.4 are conrmed by observation boundary conditions of the models are shown in Fig. 7. The mesh was
of the failure surfaces, shown in Fig. 6; these results are with HDS created using ABAQUS CAE. Detail of the mesh around the bonded area
adherends and Redux319 adhesive. The control specimen shows is shown in Fig. 8. Two solid elements were used across the adhesive
complete cohesive failure. The joint contaminated with PTFE spray thickness for the Al 2014-T6 simulations and three solid elements for
shows mixed mode adhesive/cohesive failure. The failure surfaces from the HDS simulations. In addition to these solid elements, zero thickness
the PTFE lm and Frekote 700NC are very similar, with adhesive failure 8-noded zero thickness cohesive elements were inserted along the
at or around the contaminant. upper interface. The models contained around 78,000 elements. The
elements used for the adherends and adhesive were 8-noded solid brick
5. Finite element model incompatible mode elements. These elements were used since they are
capable of capturing the eect of local bending, which must occur
Finite element models were developed for the control and Frekote during failure of these joints, and thus allow convergence of the
700NC contaminated specimens with HDS or Al2014-T6 adherends. analysis. The use of cohesive elements restricted the choice of elements
The models were three-dimensional and were drawn using the through- to 8-noded elements. Identical results using a dense mesh of solid brick

Fig. 7. Geometry of the nite element models.

104
C. Jeenjitkaew, F.J. Guild International Journal of Adhesion and Adhesives 75 (2017) 101107

Fig. 8. The nite element mesh around the bonded area (dimensions in mm).

elements were obtained up to the onset of failure processes, so the Table 4

meshes used were deemed to render accurate results. The simulations Parameter values used for the exponent Drucker-Prager model [16].
were performed using ABAQUS version 6.9-2.
Dilation Angle, Parameter, a Parameter, b
The material properties for the adherends were elastic-plastic
although simulations showed that some plasticity occurs before failure 24.0 0.053 2.0
for the Al2014-T6 adherends only. The elastic-plastic properties were
measured using tensile testing [16] and are shown in Table 3. The True tensile stress (MPa) True plastic strain
properties for the adhesive were simulated using an exponent Drucker-
Prager model since this allows both the shear and tensile behaviour of 19.850 0
23.244 0.006
the adhesive to be properly simulated. The parameters used in the
25.333 0.007
analysis are shown in Table 4. These parameters were derived from 28.337 0.008
34.001 0.010
39.326 0.012
Table 3 44.581 0.014
Plastic properties used for the modelling of Al 2014-T6 adherends [16]. 49.492 0.016
54.058 0.018
True tensile stress (MPa) True plastic strain 60.348 0.021
64.041 0.023
380.00 0 67.193 0.025
471.894 0.005980 71.545 0.028
436.4325 0.007041 75.554 0.031
443.6586 0.008024 78.724 0.034
447.4121 0.008997 80.844 0.036
450.1938 0.010109 82.393 0.038
451.8948 0.011005 83.678 0.040
453.7353 0.011975 86.173 0.045
455.2046 0.012926 87.076 0.048
457.0433 0.013996 87.801 0.052
458.5767 0.014963
459.9801 0.015985
461.7202 0.017052 shear and tensile tests as described elsewhere [16].
463.0593 0.017923 Failure was simulated using zero thickness cohesive elements
464.6814 0.018951
inserted along the interface between the outer adherend and the
466.0613 0.019904
467.7266 0.020903 adhesive, as indicated in Fig. 7. The presence of the Frekote 700NC
469.0443 0.021826 was modelled using uncoupled surfaces as indicated by the experi-
470.5677 0.022850 mental investigations. The properties of the cohesive elements were
472.1158 0.023809 obtained from separate material tests; the data from these tests were
473.5115 0.024775
475.0313 0.025778
inserted directly into the model for these properties. The value of
476.6566 0.026890 fracture energy in Mode I was obtained from the xed arm peel test and
477.8425 0.027735 the value in Mode II from the four point bend end notch exure test,
479.2036 0.028725 4ENF. The maximum traction loads were taken as the maximum load
480.5959 0.029779
reached in these tests. The parameters used are shown in Table 5. Full
481.8320 0.030721
483.0735 0.031672 details of the derivation of these parameters are described elsewhere
520.0000 0.071000 [16].
700.0000 0.180000

105
C. Jeenjitkaew, F.J. Guild International Journal of Adhesion and Adhesives 75 (2017) 101107

Table 5
Properties of the cohesive elements [16].

1) Elastic traction (penalty stiness), Kp

Normal direction, 1-direction shear, 2-direction shear,
Knn (Pa/m) Kss (Pa/m) Ktt (Pa/m)
1.881019 1.881019 1.881019

2) Maximum stress criterion (Maxs) for damage initiation

Normal mode 1-direction shear 2-direction shear
nominal stress (MPa), nominal stress (MPa), nominal stress (MPa),
1 2
44 53 53

3) Energy mode with BK exponential softening for damage evolution (BK

parameter=3.55)
GIc (kJ/m2) GIIc (kJ/m2) GIIIc (kJ/m2)
0.885 4.457 4.457

Fig. 10. Comparison of predicted and experimental values of strain a) measured by the
extensometer b) measured by the strain gauge.

Fig. 9. Comparison of predicted and experimental values of shear strength.

6. Predictive results and comparison with experimental results

The predicted and experimental values of shear strength are

compared in Fig. 9 for the control and Frekote 700NC contaminated
joints, with HDS and Al2014-T6 adherends. Almost all the results are in
agreement within the experimental error. The agreement is not perfect
for the control Al2014-T6 adherends; this probably arises from the
plasticity in the simulation which may not have been as extensive in the
experiments due to the hydrostatic constraint.
The predicted and experimental values of axial strain measured
using the extensometer across the joint are compared for the control
and Frekote 700NC contaminated joint, with HDS adherends in
Fig. 10a. Good agreement is found for both the control and the
contaminated joint in the linear regime, but there is less good
agreement after the onset on non-linearity which arises from the
eccentricity of these double lap joints. Fig. 10b shows the comparison
of the predicted and experimental values of axial strain measured using
the strain gauge for the same joints and good agreement is found; these
results are almost linear until the onset of failure.
The strain proles of axial and lateral strain along the centre prole
as indicated in Fig. 7 were extracted. The results are shown in Fig. 11; Fig. 11. Comparison of predicted values of strain along the top surface (see Fig. 7) at
these proles were extracted at around 23% failure load. The compar- approximately 23% failure load a) axial strain b) lateral (Poisson's Ratio) strain.
ison of axial strain between the control and contaminated joints
(Fig. 11a) shows some dierence. The comparison of 2-direction, bonds since these contaminants migrated into both the high tempera-
lateral (Poisson's ratio) strain (Fig. 11b) shows a more marked ture and low temperature cure adhesives. PTFE spray did cause some
dierence with a peak in the reduced value of negative strain found reduction in shear strength, but it was unclear whether this was caused
for the contaminated joint over the kissing bond. This result may form by the weakness at the interface or a weak layer within the adhesive
the basis of a detection method for kissing joints. arising from the micron-sized PTFE particles. The double-lap joint was
shown to be a valid joint geometry to test kissing joints since the same
7. Discussion results were obtained for dierent adherend metals. The joints con-
taminated with Frekote 700NC had the same values of shear strength as
Kissing bonds in double-lap joints have been successfully prepared the joints including PTFE lm. It was deduced that the Frekote 700NC
using Frekote 700NC mould release. Other contaminants tested, contamination lead to zero adhesion.
articial sweat and cutting oil lubricant, did not give rise to kissing The failure of control and Frekote 700NC contaminated joints was

106
C. Jeenjitkaew, F.J. Guild International Journal of Adhesion and Adhesives 75 (2017) 101107

modelled using nite element analysis. The Frekote 700NC area was would like to thank our project collaborators: University of Bristol,
modelled as uncoupled surfaces. The areas of adhesion were modelled QinetiQ and Airbus. We gratefully acknowledge valuable discussion
using cohesive elements. The properties of these cohesive elements with Professor Robert Smith, University of Bristol.
were measured in separate material tests and the values obtained from
these tests were used for the properties of these elements without any References
tuning. Good agreement was found for the values of shear strength for
both control and contaminated joints. Values of strain measured either [1] Michaloudaki M, Lehmann E, Kosteas D. Neutron imaging as a tool for the non-
using an extensometer or from a strain gauge were also found to be in destructive evaluation of adhesive joints in aluminium. Int J Adhes Adhes
2005;25:25767.
good agreement over the linear ranges. These comparisons gave [2] Meola C, Carlomagno GM. Recent advances in the use of infrared thermography.
condence in the delity of the nite element models. Meas Sci Technol 2004;15:2758.
With the condence obtained from the comparisons of experimental [3] Heslehurst RB, Baird JP, Williamson HM. The eect on adhesion stiness due to
bonded surface contamination. J Adv Mater 1995;26:115.
and predictive results, the strains along the top surface of the adherend [4] Heslehurst RB. Observation in the structural response of adhesive bondline defects.
above the bond were extracted. It was found that the prole of lateral Int J Adhes Adhes 1999;19:13354.
(Poisson's ratio) strain showed marked features when the kissing bond [5] Smith RA, Weise VL, Dalton RP. Kissing bond and environmental degradation
detection using nonlinear ultrasonics, Qinetiq Ltd. 2002;38.
was present. This observation leads to speculation whether monitoring [6] Vijaya Kumar RL, Bhat MR, Murthy CRL. Some studies on evaluation of degradation
or detection of this strain could lead to a detection method for kissing in composite adhesive joints using ultrasonic techniques. Ultrasonics
bonds. This method would require the joint to be loaded, but not to the 2013;53:115062.
[7] Brotherhood CJ, Drinkwater BW, Dixon S. The detectability of kissing bonds in
extent expected from the monitoring of axial strain (see section 1).
adhesive joints using ultrasonic techniques. Ultrasonics 2003;41:5219.
Loading of the joint is likely to be required for any detection method for [8] Van Den Abeele K, Delrue S, Haupert S and Aleshin V. Modeling nonlinear response
kissing bonds arising from their denition: good contact but little or no from distributed damage and kissing bonds. In: Proceedings of the Meetings on
adhesion [19]. If axial strain is imposed, local Poisson's ratio strains will Acoustics, vol. 16, Acoustical Society of America, 2012. pp. 136.
[9] Rothenfusser M, Mayr M, Baumann J. Acoustic nonlinearities in adhesive joints.
also arise. The monitoring or detection of strain across the joint at Ultrasonics 2000;38:3226.
certain intervals may be more tractable than monitoring or detection of [10] Yan D, Drinkwater BW, Neild SA. Measurement of the ultrasonic nonlinearity of
axial strain. Any signicant reductions in local lateral strain could be kissing bonds in adhesive joints. NDTE Int 2009;42:45966.
[11] Yan D, Drinkwater BW, Neild SA. Modelling and measurement of the nonlinear
used as an indicator of a kissing bond. behaviour of kissing bonds in adhesive joints. NDTE Int 2011;47:1825.
The means of this monitoring or detection can be further speculated. [12] Perton M, Blouin A, Monchalin J-P. Adhesive bond testing of carbon-epoxy
For small scale applications, DIC measurements might be possible. For composites by laser shockwave. J Phys D: Appl Phys 2011;44:112.
[13] Ehrhart B, Ecault R, Touchard F, Boustie M, Berthe L, Bockenheimer C, Valeske B.
composite materials, monitoring could be established using smart bres Development of a laser shock adhesion test for the assessment of weak adhesive
[e.g. 20], perhaps nearer the bond line which might amplify the results; bonded CFRP structures. Int J Adhes Adhes 2014;52:5765.
this point needs further investigation. For carbon bre reinforced [14] Comer AJ, Katnam KB, Stanley WF, Young TM. Characterising the behaviour of
single lap joints using digital image correlation. Int J Adhes Adhes 2013;40:21523.
composites, Raman spectroscopy [e.g. 21] is highly eective, but is [15] Vijaya Kumar RL, Bhat MR, Murthy CRL. Evaluation of kissing bond in composite
only available for surface bres. The application of this technique adhesive lap joints using digital image correlation. Int J Adhes Adhes 2013;42:608.
would be dependent on the lay-up with respect to the bond line. Further [16] Jeenjitkaew C. Kissing bonds in adhesive joints: a holistic approach for surface
chemistry and joint mechanics. [Ph.D. Thesis]. Queen Mary: University of London;
complications could arise from the presence of residual stress [e.g. 22]
2011.
or global constraints which may aect the size and form of the lateral [17] Jeenjitkaew C, Luklinska Z, Guild FJ. Morphology and surface chemistry of kissing
strain. bonds in adhesive joints produced by surface contamination. Int J Adhes Adhes
2010;30:64353.
[18] ASTM D. 3528-96 Standard test method for strength properties of double lap shear
8. Conclusions adhesive joints by tension loading. Annual Book of ASTM Standards. American
Society for Testing Material; 2002.
Kissing bonds have been successfully prepared, tested and simu- [19] Marty PN, Desai N, and Andersson J. NDT of kissing bond in aeronautical
structures. In: Proceedings of the 16th World Conference on NDT; 2004.
lated. The agreement between experimental and simulated results is [20] Read IJ, Foote PD. Sea and ight trials of optical bre Bragg grating strain sensing
very good. This delity of the simulations leads to deduction of a systems. Smart Mater Struct 2001;10:1085.
possible detection method for kissing bonds, namely by monitoring or [21] Montes-Moran MA, Young RJ. Raman spectroscopy study of HM carbon bres:
eect of plasma treatment on the interfacial properties of single bre/epoxy
detection of the lateral strain around the kissing bond. Further work is composites. Carbon 2002;40:84555.
required to investigate the sensitivity and applicability of this proposed [22] Edwards L, Fitzpatrick M, Irving P, Sinclair I, Zhang X, Yapp D. An integrated
technique to large scale joints with complications such as constraint approach to the determination and consequences of residual stress on the fatigue
performance of welded aircraft structures. In Residual Stress Eects on Fatigue and
and residual stress.
Fracture Testing and Incorporation of Results into Design. ASTM International;
2007.
Acknowledgements

This project was funded by EPSRC (EP/D055954). The authors

107