Composites: Part A 40 (2009) 644–652

Contents lists available at ScienceDirect

Composites: Part A
journal homepage: www.elsevier.com/locate/compositesa

Stress and failure analysis of crimped metal–composite joints used

in electrical insulators subjected to bending
Alain Prenleloup a, Thomas Gmür a,*, John Botsis a, Konstantin O. Papailiou b, Kurt Obrist b
a
Ecole polytechnique fédérale de Lausanne (EPFL), Laboratoire de mécanique appliquée et d’analyse de fiabilité, Bâtiment ME, Station No. 9, 1015 Lausanne, Switzerland
b
Pfisterer-Sefag AG, Elektrotechnische Spezialartikel, Werkstraße 7, 6102 Malters, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history: Experimental and numerical investigations are carried out on metal/fibreglass-reinforced-plastic joints
Received 17 July 2008 integrated in electrical insulators subject to bending. Numerical stress and strain distributions through
Received in revised form 20 February 2009 the bond are calculated with a solid 3D finite element model and the damage initiation in the composite
Accepted 24 February 2009
is highlighted. The simulations are compared to experimental data obtained from several joint specimens
tested under bending on an experimental setup equipped with strain gauges and a six-channel acoustic
emission system. Good correlation between the finite element predictions and the test results is found.
Keywords:
The investigations have identified the stress concentrations in the rod, the onset of damage when the
B. Stress concentrations
C. Finite element analysis (FEA)
load–displacement curve characterizing the bending test deviates from linearity, and the different failure
D. Acoustic emission mechanisms.
E. Joints/joining Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction In the last decade, several experimental and/or numerical stud-

Metal–composite joints are increasingly used in industrial
equipments such as automotive, aircraft, aerospace or electrical
components. As a specific application of mixed joints encountered
in high-voltage networks, silicon composite insulators with
crimped metal end-fittings manufactured for high-voltage trans-
mission lines or substations are now replacing conventional porce-
lain elements due to the hydrophobic nature of their silicon
surface, their insensitiveness to dynamic loads such as short-cir-
cuits or earthquakes, and their light weight compared to porcelain
which facilitates their handling, transport and installation. These
structures, however, require two metal end-fittings in order to
transfer loads from the high-voltage conductor to the tower or
the substation equipment. Two different designs are adopted
according to whether the insulator is loaded primarily in bending
or in traction. In the first case, the core of the insulator is chosen
in the form of a large cylinder with aluminium fittings bonded to
the composite tube with an epoxy adhesive layer. In the second sit-
uation, a thin rod-type insulator is preferred where steel end-fit-
tings are crimped to the composite rod with a radial compression
by means of a hydraulic press until a sufficient plastic deformation
is initiated in the metal. Due to the lower assembly cost of the lat-
ter solution, the crimping strategy is nowadays also applied to
thick rod-type insulators (with generally a diameter above
50 mm) subject to bending.
* Corresponding author. Tel.: +41 21 6932924; fax: +41 21 6937340.
E-mail address: thomas.gmuer@epfl.ch (T. Gmür).
ies have been reported on the stress, fatigue and failure analysis of
metal–composite single, double or tubular lap joints under trac-
tion, torsion, bending or multi-axial loading. However, all these
works are devoted to adhesively bonded joints which do not expe-
rience crimping prior to external mechanical loading. Only a few
papers [1–9] have been published on the stress and failure analysis
of composite insulators consisting of a glass-reinforced polymer
rod with two metal end-fittings radially crimped onto the compos-
ite during assembly. In these works, numerical and experimental
investigations are thoroughly conducted on non-adhesively
bonded joints, but they are restricted to insulators subjected to ax-
ial tensile loads. So far, only two studies [10–12] have been carried
out on the bending behaviour of crimped mixed metal–composite
insulators. According to the limited current research and develop-
ment on this type of joints, a systematic and integrated approach
that would account for different types of configurations and lead
to realistic design methodology for crimped metal–composite
structural joints subject to bending is desirable.
In comparison to single-material components, mixed structural
elements exhibit differences in the constitutive law, thermal or
hydrothermal response and failure characteristics between the
metal, the composite and the possible bonding layer. This intro-
duces zones of high stresses that induce damage initiation and
growth, leading to fracture onset and ultimately failure. Conse-
quently, before such composite–metal joints can be used with con-
fidence in structural applications, a thorough characterization of
their mechanical and failure behaviour under static mechanical
loads is required. The present paper is aimed at a critical evaluation

1359-835X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesa.2009.02.020
 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652
of the bending behaviour of specific steel–glass/epoxy crimped
joints used in thick-rod composite insulators connecting trans-
formers and substations to overhead lines up to the highest voltage
levels. Numerical simulations of the stress distributions through-
out the joints are performed by means of full 3D finite element
models. The different damage mechanisms observed experimen-
tally are identified and the main parameters influencing the tough-
ness of the joints and the distributions of the internal stresses are
investigated. An experimental setup based upon acoustic emission
and strain gauges is also developed in order to validate the numer-
ical models and to follow, up to failure, the damage progress in the
joints. This work should serve as a basic step towards the optimi-
zation of the mechanical bearing capacity of thick-rod insulators
subject to bending.
2. Description of the steel–glass/epoxy crimped joint under
investigation
Manufactured by Pfisterer-SefagÒ (Switzerland), the thick rod-
type composite insulators used in this study are made of an epoxy
rod reinforced with ECR-glass fibres and of two steel fittings
strongly crimped to both ends of the rod (Fig. 1). For preventing
degradation in corrosive environments and electrical leakage cur-
rents, the composite is covered with a silicone rubber housing or
shed which is not considered in the analysis.
The composite solid core, obtained by a pultrusion technique, is
a unidirectional glass-reinforced polymer (GRP) rod. The matrix is
made of epoxy resin and the volume fraction of the ECR-glass fibres
Fig. 1. Schematic configuration of a silicon composite insulator.
645
is measured as 56%. Geometrically, the pultruded rod considered
has a nominal diameter between 51 and 88 mm and a length,
adjustable to customer’s need, ranging from 0.4 to 2.0 m. Tem-
pered unalloyed carbon steel is chosen for the metal end-fittings
fabricated directly by the manufacturer of the insulators. Achieved
by the insulator manufacturer, the crimping of the end-fittings on
the composite rod is performed by means of an eight-jaw hydraulic
press. A high internal pressure, corresponding to a crimping load of
up to 600 kN and inducing a radial shortening of the steel adher-
ends, is required in order to reach a prescribed value, depending
on the diameter of the rod, after relaxing the crimping pressure.
It should however be stressed that the crimping phase is not cru-
cial for rods subjected to bending in contrast with insulators under
tensile loading where crimping has to prevent the rod from sliding
out of the end-fitting.
The values of the constitutive parameters for the epoxy resin
and the ECR-glass fibres have been chosen in the literature. By
using the classical rule of mixtures applied to the whole homoge-
nized GRP core and knowing the volume fraction of the fibres, the
elastic properties of the unidirectional composite rod could be eas-
ily estimated. Some of these evaluated properties have been cor-
roborated with the values provided by the supplier of the
composite rod and have been confirmed with a normalized tensile
test performed on an in-house axial/torsion testing system. The
latter has also been used for determining the elastic and plastic
constitutive parameters of the steel end-fittings. No heat or surface
treatment is applied to the fittings. The properties found for the
different materials are listed in Table 1 and the stress–strain graph
derived from the load–displacement curve measured for the steel
is shown in Fig. 2.
3. Numerical stress and failure analysis
3.1. Finite element models
Full 3D solid finite element models of the different joints have
been elaborated using the commercially available ABAQUSÒ soft-
ware. Non-linear simulations are performed successively on the
51- to 88-mm-thick composite insulators for the three following
steps: the crimping of the end-fitting onto the composite rod, the
relaxation of the crimping pressure and the bending of the entire
joint.
The numerical models are based upon hexahedral quadratic 20-
node 3D solid elements with a reduced integration scheme for the
computation of the structural components (C3D20R in the ABA-
QUSÒ element nomenclature). By symmetry, only a half of the
complete joint is modelled. The mesh, represented in Fig. 3 for a
63-mm-thick insulator, is composed of 112 finite elements per
hardened steel jaw or blade compressing the end-fitting on the
composite rod, 4915 elements for the steel half-fitting and 6912
elements for the composite half-cylinder. As illustrated in Fig. 3,
the grid for the composite rod is refined at the outlet of the end-fit-
ting since this zone is subject to stress concentrations during the
bending phase as shown in the next subsections.
For the crimping process, a Coulomb contact law is chosen for
symbolizing the contact at the interfaces between the press and
the end-fitting and between the latter and the composite rod. A
friction coefficient of 0.35, corresponding to the mean value of data
measured on several specimens, is adopted for both interfaces. This
value has been measured through a pull-out test applied on vari-
ous slices of a rod constrained by an end-fitting and also by means
of a tensile test on flat GRP specimens cut out of a rod and com-
pressed between two steel blocks with the same surface roughness
as on the inner surface of the end-fitting. For the bending step, a
full damage tensor criterion (similar to the Tsai–Wu criterion) is
646 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652

Table 1
Material properties (engineering constants).

Component Young’s modulus Shear modulus Poisson’s ratio

E1a (GPa) E2, E3 (GPa) G23 (GPa) G31, G12 (GPa) m23 ( ) m31, m12 ( )
End-fitting 200 – – – 0.29 –
Rod 44 10.3 4.3 5.1 0.5 0.32
a
For the composite, subscripts 1, 2 and 3 refer to the fibre and both transverse directions.

Table 2
Ultimate strengths for the composite rod.

Direction Value (MPa)

Longitudinal (tension) 1038
Longitudinal (compression) 794
Transverse (tension) 32
Transverse (compression) 141
Longitudinal–transverse 54

crimping process is, as aforementioned, not critical for thick insu-

Fig. 2. Stress–strain curve for the carbon steel used for manufacturing the end-
fittings.
applied to the elastic transversely isotropic composite. Derived
from in-house performed axial tensile and compression, in-plane
shear and torsion, and biaxial compression strength tests on spec-
imens cut in rods of various diameters, the ultimate strengths mea-
sured are shown in Table 2. These values have been obtained with
cylindrical longitudinal specimens (axial tension and compres-
sion), flat transverse coupons cut in thick rods (transverse tension
and compression), cylindrical samples with two half-section
notches (longitudinal–transverse shear) and cubic specimens
(biaxial compression).
3.2. Numerical stress analysis
It can be shown [11] that, contrary to the situation inherent in
thin composite insulators subject to tensile loading [8,9], the
lators. Since these structures are primarily designed for resisting to
bending moments caused by loads mainly orthogonal to the longi-
tudinal or glass fibre axis, the crimping pressure is chosen suffi-
ciently low to avoid any damage of the composite rod. The stress
analysis is thus focused here on the behaviour of the structures
during the bending phase. Furthermore, the results presented are
limited to the 51- and 63-mm-thick insulators since they are in es-
sence similar for all the range of diameters.
A mechanical loading of 9 and 10 kN is applied at the free end of
the 51- and 63-mm-thick models, respectively. Located in Fig. 4,
the distributions of the longitudinal and radial normal stresses
rzz and rrr along the outer compressive fibre of the rod are shown
in Fig. 5. Situated at the outlet of the end-fitting, a maximum lon-
gitudinal stress rzz of 910 MPa and a highest radial stress rrr of
280 MPa are then reported in the composite for the thinner insu-
lator, while these values are reduced to 810 and 260 MPa for
the thicker composite rod. The distributions of the radial–axial
shear stresses srz on the axis of the rod and on the quasi outer com-
pressive fibre of the composite (Fig. 4) are illustrated in Fig. 6. It
should be noted that according to the hertzian contact law the
highest shear stresses within the built-in length appear in the com-
posite at a distance from the interface between the rod and the
end-fitting corresponding to 3% of the radius of the rod. Maximum
shear stresses srz of 70 MPa and 100 MPa are predicted on the
axis of the 51-mm-thick rod in the middle of the end-fitting and

Fig. 3. Modelling of the composite joint with 3D finite elements (63-mm-thick composite insulator).
 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652 647

Fig. 4. Loading description and location of reported stress distributions.

dictions given by the strength-of-materials theory for a cantilever

beam subject to bending with shearing. It should be observed that
within the bond all the normal and shear peak stresses overrun the
corresponding ultimate strengths (rlim or slim) given in Table 2 for
the longitudinal, transverse and longitudinal–transverse direc-
tions. Since at the loading level chosen the integrity of the insula-
tors is still preserved as it will be seen later, a hidden damage has
probably been initiated at the current stress state in the composite
rod, which will be confirmed in the section devoted to the experi-
mental analysis of the joints. As an additional information, the
intensity of the normal and shear stresses in the mid-plane of
the rod for the 63-mm-thick insulator is illustrated in Fig. 7, where
the stress concentrations are highlighted with a circle.

Fig. 5. Numerical distribution of the longitudinal and radial normal stresses for the
51- and 63-mm-thick composite rods loaded at 9 and 10 kN, respectively.

Fig. 6. Numerical distribution of the radial–axial shear stresses on the neutral axis
and on the nearly outer compressive fibre for the 51- and 63-mm-thick composite
rods loaded at 9 and 10 kN, respectively.

at the outlet of the fitting on the nearly outer compressive fibre,

respectively. For the 63-mm-thick insulators, the corresponding
stresses are equal to 65 and 90 MPa.
It can be seen in Figs. 5 and 6 that the stress distributions are
very similar for both insulators and that the peak stresses appear
at the same relative locations. Outside the bond, the normal and
Fig. 7. Numerical stress distribution in the joint for the 63-mm-thick composite rod
shear stresses follow exactly the linear and constant analytical pre- loaded at 10 kN (a circle indicates maximum stress location).
648 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652
3.3. Numerical damage analysis
In order to verify the rod integrity, a Tsai–Wu-type damage ten-
sor criterion is applied to the unidirectional composite. The results
are expressed in the form of a normalized index (stress ratio) cor-
responding to the inverse of the factor by which the applied bend-
ing load can be linearly magnified before failure [13], an index of 1
indicating the onset of damage. The predictions found for the dam-
age criterion when subjecting the 51- and 63-mm-thick insulators
to a bending load of 9 and 10 kN, respectively, are shown in Fig. 8.
It can be observed that the unit limit value for the damage index is
exceeded slightly but very locally. The critical zones, marked with
circles in the figure, are the middle of the 63-mm-thick composite
rod, where the shear stresses are dominant, and for both insulators
the vicinity of the two contact points with the end-fitting, sub-
jected to high radial stresses and located at the beginning of the
rod for the part of the composite under tension and at the bond
end for the part in compression. At this loading intensity, only local
damage is however initiated, which explains that the mechanical
bearing capacity of these insulators is actually much higher, as it
will be shown later. It should also be added that the zone appar-
ently critical on the end face of the composite is not due to the con-
tact of the rod with the bottom of the end-fitting (a sufficiently
large gap is present between the rod and the bottom of the fitting),
but is in fact to be attributed to the warping of the end face, induc-
ing normal radial and hoop tensile stresses which overrun in this
region the relatively low ultimate transverse strength in tension
(Table 2); the resulting damage index reported on the end face of
the rod is however not critical for the bending phase of the insula-
tors, since the corresponding local stress–strain state does not af-
fect the overall bending stiffness of the structure.
3.4. Numerical sensitivity analysis
The numerical stress and damage study is here completed by a
brief sensitivity analysis of the main parameters influencing the
load bearing capacity of the joints when subjected to bending.
Although for the crimping step several parameters such as the fric-
tion coefficient between the GRP rod and the end-fitting, the size of
the initial gap between the rod and the fitting, the thickness of the
end-fitting, the radial shortening of the rod and the length of the
Fig. 8. Normalized damage index for the (a) 51- and (b) 63-mm-thick composite
rods loaded at 9 and 10 kN, respectively (a circle indicates highest index location).
bond, play an important role in the optimization of the joints [8],
only the last two characteristics are crucial for the bending phase,
so that the sensitivity analysis is limited here to these parameters.
The sensitivity of the stress–strain state captured during the
bending step to the radial shortening applied to the composite rod
during the crimping phase is analysed for the 51-mm-thick insulator
loaded at 9 kN when ranging the parameter between 70% and 140%
of its nominal value. The resulting maximum normal and shear
stresses, referred to the corresponding ultimate strengths (Table
2), are illustrated in Fig. 9a, where it can be observed that the longi-
tudinal stress rzz on the outer fibre of the composite is only slightly
affected by a change in radial shortening, but the shear stress srz on
the neutral axis is drastically reduced by almost 40% with a 40% in-
crease of this parameter. For the same increase, the radial normal
stress rrr on the outer fibre and the shear stress srz on the nearly out-
er fibre are however increased by up to 40% since the crimping effect
is more pronounced with a larger radial shortening.
The sensitivity of the stresses to the length of the joint is next
studied, again for the 51-mm-thick insulator under a bending load
of 9 kN. The parameter is varied from 75% to 150% of the nominal
value chosen as reference. Presented in Fig. 9b, the results show
that all the maximum normal and shear stresses, referred to the
corresponding ultimate strengths (Table 2), are decreasing with
an enlargement of that parameter, especially the shear stress srz
predicted on the neutral axis which is reduced by more than 40%
for an increase by 50% of the built-in distance. It is thus seen that
Fig. 9. Sensitivity of the maximum stresses to the radial shortening and the bond
length in the 51-mm-thick composite rod.
 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652 649

a longer end-fitting contributes to reduce the failure index and,

consequently, to delay the onset of damage.

4. Experimental stress and failure analysis

4.1. Description of the experimental setup

In order to validate the numerical stress state and damage index

generated during the bending phase of the insulators, a specific
experimental setup has been developed for the analysis of the
composite–metal joint under flexure.
Applying a transverse load at the free end of a built-in insulator,
a servo-hydraulic test frame, provided by the manufacturer of the
insulators, has been constructed for subjecting joint specimens to
bending loads up to 50 kN (Fig. 10). The experimental system is
equipped with a hydraulic jack and a load cell provided by Vibrom-
eterÒ (Switzerland), and instrumented with an LVDT (linear vari-
able differential transformer) displacement transducer. The
regulation loop for the control of the hydraulic group and the data
transfer are performed with a multi-channel acquisition card
implemented on a PC. While the end-fitting is firmly bolted to Fig. 11. Location of the strain gauges (a) and the acoustic emission microphones (b)
(51-mm-thick composite insulator).
the rigid frame in order to obtain perfect clamping conditions,
the free end is coupled to the hydraulic jack through a ring placed
in a circular groove manufactured on the composite and by means 4.2. Experimental stress analysis
of a universal joint for avoiding any parasitic axial loading.
For the stress analysis, six HBMÒ 3/350LY13 resistive strain Several 51-, 63- and 88-mm-thick insulators have been assem-
gauges are pasted at different locations on the composite rod and bled by the manufacturer of the joints by using the nominal pres-
the outer surface of the fixed end-fitting (Fig. 11a). For the failure sure [8] for crimping the end-fittings on the composite rod. These
study, a six-channel acoustic emission (AE) system (Mistras joints are next mounted in the aforementioned test frame and a
2001) provided by Euro Physical AcousticsÒ is added to the test transverse load is progressively applied with a controlled loading
equipment to shed light in the process of damage activity on the velocity (protocol fixed by IEEE standards) at the free end of the
joint as a function of the bending moment (Fig. 11b). specimens clamped at the other side.

Fig. 10. Experimental setup used for the joint specimens subject to bending (63-mm-thick composite insulator).
650 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652

The experimental load–displacement curves (deflection at the

LVDT location) and the corresponding strain–displacement curves
for the six strain gauges pasted on the rod and the end-fitting
(Fig. 11a) are recorded for each specimen. Since the graphs are
nearly coincident for all the joints, only the results obtained with
one of the 51-mm-thick insulators are considered here. The data
are reported in Fig. 12, where it can be observed that up to 9 kN
the load and strain curves are linear with the transverse displace-
ment and that the metal fitting even experiences none deformation
as highlighted by the response of the strain gauge pasted on the
adherend (gauge #1). By increasing the load up to 13.5 kN, all
the curves deviate from linearity. The signals of the strain gauges
on the rod (gauges #3, #4 and #6) show that the composite is sub-
jected to high deformations and the gauge #1 confirms that the
steel end-fitting undergoes a plastic deformation at this load level.
When the transverse load is still increased from 13.5 kN up to the
ultimate load of 16 kN, the signals of the strain gauges are lost one
by one and the joint fails as described in the next subsection. Fig. 13. Numerical and experimental load–displacement curves for three 51- and
The simulated load–displacement curves for both the 51- and 63-mm-thick insulators.
63-mm-thick insulators are compared in Fig. 13 with the corre-
sponding experimental data (three specimens have been tested
for each diameter). For the sake of clarity, the curves are not fully severe AE activity appears at the onset of non-linearity (10 kN) in
plotted after reaching the maximum mechanical load. An excellent the load–displacement curve and that the intensity of the activity
agreement between the two sets of results can be observed in the increases gradually up to the ultimate mechanical loading. A slight
nearly linear behaviour range of the experiments. It could also be AE activity can however already be observed at load levels below
shown that in this range the numerical strain predictions fully the limit of linearity, indicating a possible damage initiation in
agree with the measurements derived from the strain gauges. the composite rod.
However, the onset of non-linearity in the load–displacement The damage mechanism for the 51-mm-thick composite rod is
curves suggests that a local damage is initiated in the composite shown in Fig. 15, where it is seen that failure is caused by a trans-
rod at 50–60% of the maximum mechanical loading, which corre- verse crack propagation due to the normal stresses in the compres-
sponds to a bending load of 9–10 kN. This confirms the suspicion sive part of the rod near the outlet of the end-fitting. This
of a damage initiation as already mentioned in Sections 3.2 and 3.3. phenomenon, due mainly to the bending of the rod and not the
crimping of the end-fitting, is to be related with the damage index
4.3. Experimental failure analysis shown in Fig. 8a which reaches its highest value precisely in the
zone near the initiation of the transverse cracks. The observation
In parallel to the experimental stress analysis described in the of the damage area with a microscope reveals that near the outer
previous subsection, the test frame is also used to highlight the ini- surface of the composite rod the damage corresponds to a fracture
tiation of the local damage in the different insulators instrumented of the polymer matrix and the glass fibres (Fig. 16a), whereas in-
with the AE sensors. While the metal–composite joints are loaded side the composite the damage is governed by a micro-buckling
up to failure, the progress of the damage activity is measured with
the AE equipment. The maximum amplitude (highest peak volt-
age), expressed on a logarithmic scale (1 lV = 0 dB), of the AE
waveforms and the so-called AE energy of the measured acoustic
activity [14] in function of the applied bending load is illustrated
in Fig. 14 for a 51-mm-thick insulator. The LVDT deflection in func-
tion of the applied load is also given. It is clearly apparent that a

Fig. 12. Load–displacement and strain–displacement curves for the 51-mm-thick Fig. 14. Intensity of the acoustic emission activity in function of the bending load
insulator (the location of the strain gauges is shown in Fig. 10a). amplitude for the 51-mm-thick insulator.
 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652
Fig. 15. Failure mode of the 51-mm-thick insulator: transverse cracking.
Fig. 16. Micrographs of the fibre and matrix breaking (a) and of the fibre micro-
buckling (b) for the 51-mm-thick insulator.
of the fibres (Fig. 16b). It should be noted that the delamination
visible on the tensile part of the rod (Fig. 15) is a post-failure dam-
age induced by the relaxation of the bending load.
For the 63- and 88-mm-thick insulators, a longitudinal crack
propagation due to the shear stresses in the centre of the rod oc-
curs in addition to the transverse crack propagation, as shown in
Fig. 17 for the 63-mm-thick insulator. This observation is again
confirmed by the distribution of the damage index in Fig. 8b, where
it is seen that a second critical failure area appears in the middle of
the thicker composites.
5. Conclusions
The stress behaviour and the failure process under bending
loading of an electrical silicon composite insulator made of a
glass/epoxy rod with crimped steel end-fittings have been ana-
lysed in this paper. The metal/composite joint has been studied
numerically with a 3D finite element model and experimentally
with a test equipment including an acoustic emission system for
monitoring the damage progression during the solicitation. The
investigations have shown that stress concentrations appear in
the composite near the interface with the steel fitting at the outlet
of the joint and on the mid-surface of the rod in the middle of the
651
Fig. 17. Failure mode of the 63-mm-thick insulator: transverse and longitudinal
cracking.
end-fitting. They have also highlighted that the damage is initiated
in the composite rod when the load–displacement curve character-
izing the bending test deviates from linearity, the acoustic emis-
sion activity being then drastically increased. Moreover, an
excellent correlation between the numerical simulations and the
experimental results issued from several test specimens have been
observed, at least in the linear range of loading. Micrographs of the
failed joints have allowed identifying the damage mechanisms
which are different according to the thickness of the composite
rod. The analysis has also shown that the some geometrical param-
eters are crucial for the integrity and performance of the joint.
Acknowledgements
The authors wish to thank Philippe Bonhôte (HEIG-VD, Yverdon,
Switzerland) for his careful advices and appreciated help in several
aspects of this study. This work was partially supported by the
Swiss Commission for Technology and Innovation (CTI), Grant
No. 6798.1.
References
[1] Bansal A, Schubert A, Balakrishnan MV, Kumosa M. Finite element analysis of
substation composite insulators. Compos Sci Technol 1995;55(4):375–89.
[2] Lanteigne J, Lalonde S, de Tourreil C. Optimization of stresses in the end-
fittings of composite insulators for distribution and transmission lines. J Reinf
Plast Comp 1996;15(6):467–78.
[3] Bansal A, Kumosa M. Mechanical evaluation of axially loaded composite
insulators with crimped end-fittings. J Compos Mater 1997;31(20):2074–104.
[4] Kumosa M, Han Y, Kumosa L. Analyses of composite insulators with crimped
end-fittings: Part I. Non-linear finite element. Compos Sci Technol
2002;62(9):1191–207.
[5] Kumosa M, Armentrout D, Kumosa L, Han Y, Carpenter SH. Analyses of
composite insulators with crimped end-fittings: Part II. Suitable crimping
conditions. Compos Sci Technol 2002;62(9):1209–21.
[6] Mobasher B, Kingsbury D, Montesinos J, Gorur RS. Mechanical aspects of
crimped glass reinforced plastic (GRP) rods. IEEE Trans Power Deliv
2003;18(3):852–8.
[7] Duriatti D, Béakou A, Levillain R. Optimisation of the crimping process of a
metal end-fitting onto a composite rod. Compos Struct 2006;73(3):278–89.
[8] Prenleloup A, Gmür T, Botsis J, Papailiou KO. Acoustic emission study and
strength analysis of crimped steel–composite joints under traction. Compos
Struct 2006;74(3):370–8.
[9] Prenleloup A, Gmür T, Bonhôte P, Botsis J, Papailiou KO. Experimental and
numerical strength analysis of mixed metal–composite crimped or adhesively
bonded joints. In: Proceedings of CompTest 2006 conference, Porto, April,
2006. p. 25–6.
[10] Prenleloup A, Gmür T, Botsis J, Papailiou KO, Obrist K, Bonhôte P. Acoustic
emission inspection and analysis of crimped metal–composite joints subjected
to bending. In: Proceedings of fourth ICNDT conference, Chania, October, 2007.
Paper 5-2.
[11] Prenleloup A. Analyse de l’état de contrainte et de l’endommagement
d’assemblages sertis en matériau mixte métal–composite sollicités en
652 A. Prenleloup et al. / Composites: Part A 40 (2009) 644–652

traction ou en flexion. Ph.D. Thesis No. 4005. Lausanne: Ecole polytechnique [13] Kollár LP, Springer GS. Mechanics of composite
fédérale de Lausanne; 2008. structures. Cambridge: Cambridge University Press; 2003.
[12] Dumora D, Feldmann D, Gaudry M. Mechanical behavior of flexurally stressed [14] Pollock AA. Acoustic emission inspection. In: Davis JR, editor. Metals
composite insulators. IEEE Trans Power Deliv 1990;5(2): handbook. Nondestructive evaluation and quality control, vol. 17. Metals
1066–73. Park, OH: ASM International; 1989. p. 278–94.