Analysis of cracked steel members reinforced by pre-stress composite

patch
P. COLOMB I 1 , A . B A SS E T TI 2 a n d A . N U S S B A U M E R 2
1
Department of Structural Engineering, Milan Technical University, Piazza L. da Vinci, 32, 20133 Milan, Italy, 2Institute of Steel Construction, Swiss
Federal Institute of Technology, EPFL Lausanne, Switzerland
Received in final form 6 August 2002

ABSTRACT Pre-stress bonded composite patch is a promising technique to reinforce steel member
damaged by fatigue. The effectiveness of this technique was verified by fatigue tests on
notched steel plates. Results showed that the application of carbon fibre reinforced
plastic (CFRP) strips and, eventually, the introduction of a compressive stress by preten-
sion of the CFRP strips prior to bonding produced a significant increment of the
remaining fatigue life. In this paper, the stress intensity factor in the notched plates is
computed by a two-dimensional finite element model in connection with the three-layer
technique in order to reduce the computational effort. Due to high stress concentration
at the plate crack tip, debond is assumed at the adhesive±plate interface. The goal is to
illustrate the influence of some reinforcement parameters such as the composite strip
stiffness, the pre-stress level, the adhesive layer thickness and the size of the debonded
region on the effectiveness of the composite patch reinforcement.
Keywords composite patch; fatigue crack reinforcement; parametric analysis.

In Bassetti5 a novel technique was proposed to reinforce effectiveness of pre-stress CFRP-strips to stop fatigue
steel member damaged by fatigue. It consists in the crack emanating from the rivet hole or to prevent further
application of carbon fibre reinforced plastic (CFRP) cracking at other locations.
strips and, eventually, the introduction of a compressive The carbon fibre reinforced plastic laminates have
stress by pretension of CFRP strips prior to bonding. physical and mechanical properties particularly interest-
Note that this new methodology applied to riveted steel ing for reinforcement of fatigue damaged steel elements.
members avoids the drawbacks of standard reinforce- The high fatigue resistance of CFRP avoids crack propa-
ment techniques such as hole drilled at crack front, gation from the cracked steel section into the patch. The
cover plates application, replacement of rivets by high high stiffness of CFRP reduces the stress range in the
strength bolts, cold expansion of the rivet hole and cracked steel section and promotes crack bridging. The
welding of the detected cracks.1 As a step towards the high tensile strength of CFRP allows also the application
validation of this technique, fatigue tests on notched steel of a pretension to composite strips in order to increase
plate reinforced by CFRP strips were performed in Bas- the effectiveness of the bonded patch on thicker steel
setti et al.4 This technique is particularly appealing for section. Finally, the low self-weight of CFRP plates
the fatigue life extension of old riveted steel bridges built limits the dead load increment and simplifies the
up at the beginning of the last century by non-weldable strengthening operations. Composite patch reinforce-
steel. Experimental tests were then performed in Bassetti ment technique is a standard and reliable procedure in
et al.4 on a 91-year-old-cross girder in order to prove the different engineering branches to reinforce structural
elements subjected to extreme actions (high fatigue
loads, high temperature ranges and exposure to aggres-
sive agents) and nowadays, is also a standard reinforce-
Correspondence: P. Colombi department of Structural Engineering,
Milan Technical University, Piazza L. da Vinci, 32, 20133 Milan, Italy. ment methodology in aircraft industries.7 In fact the
E-mail: [email protected] CFRP laminates is becoming a familiar procedure for

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66 59
60 P. COLOMBI et al.

p
the strengthening and rehabilitation of the reinforced
K ˆ
sy
Y …a†
p
a and the stress ratio R. A lower
concrete bridges (Meier, 1995).8 The CFRP-laminates, stress ratio in fact promotes crack closure that reduces
eventually, could be pre-stressed9 in order to increase the crack growth rate.12 The crack length and the crack
reinforcement effectiveness. Some attempts were also growth measurements were carried out using the
done10 to increase the load carrying capacity of steel electric potential drop method. Results of the fatigue
bridges. tests are shown in the crack length to number of cycles
In this paper, reference is made to the notched steel diagram of Fig. 2 where the reference tests are given
plates tested in Bassetti et al.5 The goal is to investigate by the un-reinforced steel plates. Composite patch acts
numerically the effect of CFRP-strips' stiffness, adhesive on the notched plate in three different ways. First, it
thickness, pre-stress level and debond size region on lessens the stress range Dsy at crack tip, second it reduces
the stress intensity factor, the parameter governing the corrector factor Y(a) by bridging the crack lips
crack propagation rate and then fatigue lifetime. To this
end, a finite element model is developed and is based
on the Mindlin plate theory in order to reduce the com-
putational effort connected to a full three-dimensional Detail B: notch
analysis.11

0.1 mm
EXPERIMENTAL EVIDENCE

500
The effectiveness of pre-stressed CFRP-plates to arrest
the crack propagation was investigated (Fig. 1) on 10 mm
notched steel plates.4 The central notch consists in a B 5 mm
hole and two initial traversing cracks produced by elec-
troerosion. The plates were reinforced on both sides by a0
two unidirectional pultruded carbon fibre sheets with
width of 50 mm as shown in Fig. 1. A two-component
viscous epoxy was used for bonding the laminate to the
CFRP - laminates
500
steel plate. All the specimens were tested under constant
amplitude tensile loading (no thermal stresses are then 75 50 50 50 75
considered) using a stress range Dsy ˆ80 MPa and a
stress ratio R ˆ sy, min/sy, max ˆ 0.4 in the nominal section
of the un-reinforced specimen. Tests were performed on
a hydraulic testing machine with a dynamic capacity of 150 150
800 kN. As is well known, the fatigue crack propagation
is controlled by the stress intensity factor range Fig. 1 Notched steel plate specimen.

120

100
Crack length a [mm]

80
Pre-stress [kN] E [GPa] t [mm]
60 0 174 1.2
CFRP

41.2 174 1.2

41.2 174 1.4
40 41.2 216 1.4
Steel-only specimens (unreinforced)

20
a0 ∆s = 80 MPa; R = 0.4
0
0 106 2·106 3·106 4·106 5·106 6·106 7·106 Fig. 2 Crack length vs. number of cycles
Number of cycles N for different patch configuration.

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66
 COMPOSITE PATCH REPAIR OF CRACKED STEEL MEMBERS 61

and third the compressive stress field produced by technique was proposed to model a composite bonded
the pretension of composite patch reduces the stress reinforcement of a cracked plate. This technique uses
ratio R and then promotes crack closure. two-dimensional finite element analysis, consisting of
By reinforcing the steel plates with non-pre-stress three layers, to model the cracked plate, adhesive and
CFRP strips with mean Young's modulus in the fibres composite patch. It is not required to replace the adhe-
direction E1 ˆ 174 GPa and thickness hc ˆ 1.2 mm, fa- sive layer by shear spring elements (noncontinuum
tigue life of specimens is increased by a factor of about body) since the adhesive is modelled as an elastic con-
three. If the composite strips are pre-stress prior to tinuum medium. In this way the characteristics of
bonding by applying a tensile stress equal to 632 MPa the adhesive required to model nonlinear material behav-
in the CFRP cross-section, the fatigue life is increased iour are also captured. Constraints are used to enforce
by a factor of about five. Even longer fatigue life was the compatibility along the plate-adhesive and the
achieved by stiffer CFRP-strips (E1 ˆ 216 GPa and adhesive±patch interface based on Mindlin assumptions2
hc ˆ 1.4 mm) with the same pre-stress level as before. (Fig. 3).
With this reinforcement, configuration of the fatigue The Mindlin plate theory assumes linear displacement
life is increased by a factor of about 16. As the crack field in the plate thickness. In the three layers technique,
approaches the interior boundary of the patch, high all three layers, cracked plate, adhesive and composite
stress concentration at the crack tip produces debond patch, are assumed to have a linear displacement field
at the plate-adhesive or adhesive±patch interface. along the thickness and they satisfy the relations:16
From the inspection of the failed specimens, it was
uc ˆ uP ‡ 'Py
zc ua ˆ uJ ‡ 'yJ
za us ˆ uA ‡ 'yA
zs
concluded that debond was present between the steel J
plate and the adhesive layer. Debond was also in- vc ˆ vP 'Px
zc va ˆ vJ
za
'x vs ˆ vA 'xA
zs
wc ˆ wP wa ˆ wJ ws ˆ wA
vestigated by Optical Speckle Interferometry tech-
nique.13,6 It was found in Bassetti et al.6 that an …1†
elliptical debond with an aspect ratio c/b equal to 1/5 where the superscript s, a and c are used to denote the
produced a displacement field, which fitted well in the steel plate, adhesive layer and composite patch, respect-
experimental one. ively. The co-ordinates zc, za and zs are measured from
the midplane of each layer. At the plate±adhesive inter-
FINITE ELEMENT MODEL face, where the z co-ordinates for the cracked plate and
the adhesive are equal, and at the adhesive±composite
One of the main aspects of bonded composite reinforce- patch interface, where the z co-ordinates for the adhesive
ment technology is the stress analysis of the reinforced and the composite patch are equal, the constraint equa-
structure and the consequent stress intensity factor tions reduce to:2
evaluation. Three-dimensional finite element analyses11
of composite patch repair have been conducted in the ha hs
ua ˆ us ) uJ uA 'yJ
'yA
ˆ 0
literature. Since the thickness of the adhesive is much 2 2
a s
smaller than the plate and composite patch one, a three J h A h …2†
va ˆ vs ) vJ vA ‡ 'x
‡ 'x
ˆ 0
dimensional model becomes very expensive due to the 2 2
large number of elements required across the thickness wa ˆ ws
to get acceptable aspect ratios in the adhesive layer. at the plate±adhesive interface and:
A number of two-dimensional finite element models
were then employed in the literature to reduce the com-
putational effort.11
In Ratwani et al.14 analyses were conducted by plain P hc
stress two-dimensional elements to model the cracked
plate and composite patch. Shear spring elements were ϕ
used to represent the adhesive. In Sun et al.15 two- J ha
dimensional Mindlin plate elements with transverse
shear deformation capability were used to model both z, w
y, v
the cracked plate and the composite patch. Again shear A hs
spring elements were introduced to represent the
adhesive. The shear springs were connected to the x, u
cracked plate and the composite patch through displace-
ment constraint equations, which satisfy the Mindlin Fig. 3 Modelling of bonded reinforcement and necessary
plate assumptions. In Naboulsi et al.16 the three-layer constraints at the interfaces.

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66
62 P. COLOMBI et al.

Based on the experimental evidence, the debond crack

hc ha
uc ˆ ua ) uP uJ 'Py
'yJ
ˆ 0 was assumed to lie in the plate±adhesive interface with a
2 2
c a semi-elliptical shape (Fig. 5).
P h J h …3†
vc ˆ v a ) v P vJ ‡ 'x
‡ 'x
ˆ 0 The finite element analysis is performed for a given
2 2
delamination shape parameter c/b. The major semiaxis
wc ˆ wa
b was assumed equal to the crack size a plus the dimen-
at the adhesive±patch interface. In the debonded region sion of the plastic zone at crack front in plate and
between the steel plate and the adhesive layer, the layers was aligned with the crack path. The minor semiaxis c
are independent of each other and the constraint equa- was then located in a direction orthogonal to the crack
tions are no longer valid. Note that due to the symmetry one3 (Fig. 5). No attempts were then done to model
of the reinforcement no overlapping due to translation in the delamination size growing in the finite element
the thickness direction is possible. model. A parametric analysis was performed in order to
investigate the sensitivity of the stress intensity
factor level to variations in the patch stiffness (E1 and
hc), adhesive thickness ha, strips pre-stress level sp. and
STRESS INTENSITY FACTOR EVALUATION
the size of the debonded region. The strain energy
Numerical model release rate, G, was computed by using the standard
virtual crack extension method17 and converted into the
The commercial finite element code ABAQUSÕ was
stress intensity factor, K, by assuming plane stress for
used to perform the analyses using the three-layer tech-
the local stress field near the crack tip and mode
nique. The main geometric parameters and material
I crack opening:3
properties of the notched steel plate in Fig. 1 are sum-
p
marised in Table 1. K ˆ G
E …4†
Since composite patches are located on both side of the
plate, only 1/8 of the specimen was meshed (Fig. 4) as a
continuum medium. Since tensile loading is applied to
the specimens, no variation of the stress intensity factor is
expected over the thickness. Standard eight-noded shell
elements were used for the adhesive layer and the com- CFRP - laminate
posite patch, and plane stress elements were used for the
steel plate.

Patch

Adhesive

c
Z b debonding
Y
Steel plate a
X
Fig. 5 Detail of the finite element model in the debonding region
Fig. 4 1/8 of the three layer finite element model. (a ˆ 50 mm).

Table 1 Material properties and dimensions

Layer Length (mm) Width (mm) Thickness (mm) Material properties

Steel plate 1000 300 10 Es ˆ 210 GPa, ns ˆ 0.3

Adhesive 500 50 0.3 Ea ˆ 714.14 MPa, na ˆ 0.32
CFRP patch 500 50 1.2 E1 ˆ 174 GPa, E2 ˆ 9 GPa
n12 ˆ 0.28, G12 ˆ G13 ˆ 4.5 GPa
G13 ˆ 4.4 GPa

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66
 COMPOSITE PATCH REPAIR OF CRACKED STEEL MEMBERS 63

In particular for short cracks the reduction of the stress

RESULTS AND DISCUSSION
intensity factor is marginal and a pre-stress must be
Figures 6±12 illustrate the results of the parametric an- introduced in order to prevent crack propagation. In
alysis. The reference values of the model parameters are fact, in this case the patch reinforcement simply reduces
E1 ˆ 174 GPa, ha ˆ 0.3 mm, hc ˆ 1.2 mm, an elliptical the stress range in the steel plate. Since the stiffness of
debonded region with c/b equal to 1/5 and no pre-stress the steel plate is much larger than the composite strips
(sp. ˆ0) in the composite strips. The corresponding one, a marginal reduction of the stress range is achieved.
stress intensity factor values are normalised with respect The effectiveness of the patch reinforcement is maxi-
p
to sy; max
p
a (sy, max ˆ 133 MPa is the maximum mized for a long crack close to the external boundary of
stress in the nominal section of the un-reinforced steel the strip. In this case, in fact, the patch covers a large
plate) and reported as function of the crack size a (the amount of the crack and its plastic zone and then an
crack size is measured from the central axis of the plate additional significant reduction of the correction factor
and then includes the hole radius). Results show a differ- Y(a) is achieved. pre-stress of the CFRP strips does not
ent behaviour for short cracks, which are outside the produce a reduction of the stress intensity factor range
strip area, compared to long cracks, which are partially but a reduction of the stress ratio R that promotes crack
covered by the patch. Figure 6 shows the stress intensity closure. As an example the application of a pre-stress
factor for notched plates reinforced by CFRP strips with sp ˆ 632 MPa reduces the stress ratio to 0.03 (Fig. 6).
and without pre-stress. Since the patch stiffness is proportional both to the
Without pre-stress, the effectiveness of the reinforce- composite patch Young's modulus E1 and the patch
ment is higher for long cracks compared to short cracks. thickness hc, variations of these two parameters produce

1.4 No patch
R = 0.4
Patch s p = 0
1.2
Patch s p = 632 MPa
1.0
K/(s (pa)1/2) [−]

R = 0.4
0.8

0.6
R = 0.03
0.4

0.2
CFRP strip
Fig. 6 Normalised stress intensity factor 0.0
vs. crack length for composite strips with 0 20 40 60 80 100
and without pre-stress. a (mm)

1.4
Young’s modulus E1
1.2 100 GPa
174 GPa
1.0 216 GPa
300 GPa
K/(s (pa)1/2) [−]

0.8 400 GPa

Steel only
0.6

0.4

0.2
CFRP strip
Fig. 7 Normalised stress intensity factor 0.0
vs. crack length as function of the patch 0 20 40 60 80 100
Young's modulus. a (mm)

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66
64 P. COLOMBI et al.

1.4

1.2
Thickness h c
1.0 1.0 mm
K/(s(pa)1/2) [−]

1.2 mm
0.8
1.4 mm
1.7 mm
0.6 2.0 mm
Steel only
0.4

0.2
CFRP strip
0.0 Fig. 8 Normalised stress intensity factor
0 20 40 60 80 100 vs. crack length as function of the patch
a [mm] thickness.

1.4

1.2
Thickness h a
1.0 0.15 mm
K/(s (πa)1/2) [−]

0.30 mm
0.8 0.50 mm
0.80 mm
0.6 1.20 mm
Steel only
0.4

0.2
CFRP strip
0.0 Fig. 9 Normalised stress intensity factor
0 20 40 60 80 100 vs. crack length as function of the adhesive
a [mm] thickness.

1.4
R = 0.4 Pre-stress s p
1.2 0
400 MPa
1.0 632 MPa
800 MPa
K /(s (p a)1/2) [−]

R = 0.4
0.8 1000 MPa
R = 0.21 Acier seul
0.6
R = 0.03

0.4 R = −0.15
R = −0.5
0.2
CFRP strip
0.0 Fig. 10 Normalised stress intensity factor
0 20 40 60 80 100 vs. crack length as function of the
a [mm] pretension level.

a similar effect on the stress intensity factor. Figure 7 The reduction of the stress intensity factor is more
depicts the sensitivity of the stress intensity factor to important for long cracks compared to short cracks.
the composite patch Young's modulus E1 in the fibre For long cracks, partially covered by the patch, CFRP
direction. strips modify the crack geometry by bridging the crack

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66
 COMPOSITE PATCH REPAIR OF CRACKED STEEL MEMBERS 65

lips and this effect is more pronounced for stiff patch. pre-stress value. Note that in this case the stress intensity
For short cracks, the CFRP strips simply produce a factor range remains unchanged but the stress ratio is
reduction of the stress range. Figure 8 depicts the influ- decreased, and even negative stress ratio is achieved as
ence of the composite strip thickness hc. indicated in Fig. 10. This promotes crack closure effects
Note that, in this case, due to patch shear deformation, that reduces the effective stress intensity factor range and
the stress intensity factor decrement is less pronounced then the corresponding crack growth rate.
compared to the one produced by a variation of the The influence of the size of the debonded region is
composite patch Young's modulus E1. Figure 9 illustrates studied with reference to elliptical debond with different
the effect of the adhesive thickness on the stress intensity aspect ratio c/b (Fig. 11).
factor values. Figure 12 illustrates the effect of the size of the
Clearly, the effectiveness of the reinforcement decreases debonded region on the stress intensity factor values.
as the adhesive thickness increases. This is due to shear The effect is more important for long cracks for which
deformation of the adhesive layer. This effect is more the crack bridging effect is reduced by the increment of
important for long cracks where the capacity of the the debonded area. On the contrary, the size of the
CFRP strips to bridge the crack lips is reduced by shear debonded region has no effect on short cracks since, in
deformation of the adhesive layer. On the contrary, the this case, the debonded region does not influence the
adhesive has no influence on short cracks that are outside stress field at crack tip.
the patch area. The effect of the pre-stress level is
reported in Fig. 10.
CONCLUSIONS
A pronounced reduction, in particular for short cracks,
of the stress intensity factor is achieved by increasing the Experimental result shows that the application of pre-
stress CFRP strip is a promising technique for the re-
inforcement of fatigue damaged steel section. In particu-
lar, it is appealing for riveted bridges built up at the
beginning of the last century by non-weldable steel.
This technique can be used to reduce or to stop crack
c / b = 1/2 propagation of existing cracks. Moreover, pre-stress of
c / b = 1/5 the CFRP strips prior to bonding introduces compres-
c c / b = 1/10 sive stress, which prevent further cracking by promoting
crack closure effect. Crack closure is a very important
aspect since it reduces the effective stress intensity factor
a range and then decreases significantly the fatigue crack
growth rate. Pre-stressing of the CFRP strips is manda-
b
tory in order to prevent further crack propagation of
short cracks in steel elements.
CFRP strip Results of the sensitivity analysis clearly indicate a dif-
ferent effect of the pre-stress reinforcement on short
Fig. 11 Elliptical debond used in the parametric analysis. cracks compared to long cracks. They show that the

1.4
Aspect ratio c / b
1.2 No debond
1/10
1.0 1/5
K/(s ( pa)1/2) [−]

0.8 1/2
Steel only
0.6

0.4

0.2
CFRP strip
Fig. 12 Normalised stress intensity factor
0.0
vs. crack length as function of the 0 20 40 60 80 100
debonded area. a [mm]

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66
66 P. COLOMBI et al.

application of high stiffness pre-stress CFRP strips

REFERENCES
bonded perpendicular to the crack path modifies the
crack geometry by bridging the crack lips. Pre-stress 1 Bassetti, A., Nussbaumer, A. and Hirt, M. (March 2000) Crack
reduces the corresponding stress ratio and then pro- repair and fatigue life extension of riveted bridge members using
composite materials. In: Bridge Engineering Conference, Sharm,
motes crack closure that reduces the fatigue crack
El-Sheikh, Egypt, 1, pp. 227±238.
growth rate. 2 Bassetti, A., Colombi, P. and Nussbaumer, A. (May 2000) Finite
For long cracks in steel elements, CFRP patches reduce element analysis of steel members repaired by pre-stress
the stress intensity factor range, even without pre-stress, composite patch. In: Proceedings IGF2000, CD-ROM, Bari, Italy.
since the patch covers the plastic zone and a large 3 Bassetti, A., Nussbaumer, A. and Colombi, P. (October 2000)
portion of the crack. For short cracks in steel elements, Repair of riveted bridge members damaged by fatigue using
crack bridging effect by CFRP without pre-stress is CFRP materials. In: Advanced FRP materials for civil structures,
Bologna, Italy, pp. 33±42.
negligible (Figs 7 and 8) since the cracks are outside the
4 Bassetti, A., Nussbaumer, A. and Hirt, A. (September 2000)
strip area. Fatigue life extension of riveted bridge members using pre-stress
Shear deformation of the adhesive layer has a significant carbon fibre composites. In: Steel structures of the 2000s, ECCS,
effect of the effectiveness of the bonded reinforcement. Istanbul, pp. 375±380.
Figure 9 shows that as the adhesive shear deformation 5 Bassetti, A. (2001) Application de lamelles pre-contraintes en
increases with adhesive thickness, the effectiveness of the fibres de carbone pour le renforcement d'elements de pont
bonded reinforcement technique is reduced. Adhesive rivetes endommages par fatigue (in French), PhD Thesis, Swiss
Federal Institute of Technology, EPFL.
thickness effect is more important for long cracks com-
6 Bassetti, A., Facchini, M. and Jacquot, P. (April 2001)
pared to short ones since shear deformation reduces Ameliorament de la tenue en fatigue de structures metalliques:
crack bridging. evaluation par interferometrie speckle du procede de
Pre-stress of the composite strips is introduced to renforcement par lamelles composites (in French). In:
reduce the stress ratio in the cracked plate and then to Photomecanique 2001, Poiters, pp. 351±358.
promote crack closure. This phenomenon is very import- 7 Baker, A. A and Jones, R. (1988) Bonded repair of aircraft struc-
ant for short crack since crack bridging has a negligible tures. Martinus-Nijhoff Publishers.
8 Meier, U. (1995) Strengthening of structures using carbon fibre/
effect in this case.
epoxy composites. Construc. Build. Mater. 9, 341±351.
Debond of the composite patch at the plate±adhesive 9 Deuring, M. (1993) VerstaÈrken Von Stahlbeton Mit Gespannten
interface plays an important role for long cracks Faserverbundwerkstoffen (in German). Report 224, EMPA.
(Fig. 12). It reduces the bridging effect of composite 10 Ammar, N. (1996) Rehabilitation of steel bridge girder with graphite
patch due to the deformation of the debonded patch. pultrasion. CCM report 96±26, University of Delaware.
The corresponding increment of the stress intensity 11 Umamaheswar, T. V. R. S. and Singh, R. (1999) Modelling of
factor in the plate is similar, for long cracks, to the one patch repair to a thin cracked sheet. Engng Fract. Mech. 62,
267±289.
produced by the reduction of the CFRP strips stiffness.
12 Suresh, S. (1991) Fatigue of Materials. Cambridge University
Then, for long cracks, the influence of the debonded Press.
region cannot be discharged. 13 Cloud, G. (1995) Optical Methods in Engineering Analysis.
Cambridge University Press.
14 Ratwani, M. M. (1979) Analysis of cracked adhesively bonded
laminated structures. AIAA J. 17, 988±994.
15 Sun, C. T. and Klug, J. (1996) Analysis of cracked aluminium
Acknowledgements plates repaired by bonded composite patches. AIAA J. 34,
The financial support to the first author by the Italian 369±374.
16 Naboulsi, S. and Mall, S. (1996) Modelling of a cracked metallic
Ministry of Instruction, University and Research
structure with bonded composite patch using the three layer
(MIUR) and to the second and third authors by the technique. Compos. Struct. 35, 295±308.
Swiss National Science Foundation and EPFL Lausanne 17 Parks, D. M. (1977) The virtual crack extension method for
is gratefully acknowledged. Thanks are also expressed to non-linear material behaviour. Comput. Meth. Appl. Mech. Engng
Sika AG ± ZuÈrich for supplying composite materials. 12, 353±364.

ß 2003 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 26, 59±66