Composite Structures

Volume 57, Issues 1–4, July 2002, Pages 323–330

Fatigue crack growth behavior of cracked

aluminum plate repaired with composite
patch
 Dae-Cheol Seo,
 Jung-Ju Lee ,

 Department of Mechanical Engineering, Korea Advanced Institute of Science

and Technology, 373-1, Kusong-dong, Yusung-gu, Taejon, 305-701, South
Korea
 Available online 28 May 2002.
 http://dx.doi.org.www.sndl1.arn.dz/10.1016/S0263-8223(02)00095-8, How to
Cite or Link Using DOI
 Cited by in Scopus (29)
 Permissions & Reprints
Abstract
In this study, we investigated the fatigue crack growth behavior
of cracked aluminum plate repaired with bonded composite
patch especially in thick plate. Adhesively bonded composite
patch repair technique has been successfully applied to military
aircraft repair and expanded its application to commercial
aircraft industry recently. Also this technique has been expanded
its application to the repair of load bearing primary structure
from secondary structure repair. Therefore, a through
understanding of crack growth behavior of thick panel repaired
with bonded composite patch is needed. We investigated the
fatigue crack growth behavior of thick panel repaired with
bonded composite patch using the stress intensity factor range
(ΔK) and fatigue crack growth rate (da/dN). The stress intensity
factor of patched crack was determined from experimental result
by comparing the crack growth behavior of specimens with and
without repair. Also, by considering the three-dimensional (3D)
stress state of patch crack, 3D finite element analyses were
performed to obtain the stress intensity factor of crack repaired
by bonded composite patch. Two types of crack front modeling,
i.e. uniform crack front model and skew crack front model, were
used. The stress intensity factor calculated using FEM was
compared with the experimentally determined values.
 Keywords
 Composite patch repair;
 Thick plate patch;
 Fatigue

1. Introduction
The traditional repair method for cracked aircraft structure is
fastening metal reinforcement using bolt and rivet. This method
makes worse the stress concentration problem due to drilling of
additional fastener hole. Also the application of metal
reinforcement changes the stress distribution of repair area and
causes the stress concentration along the neighborhood of
repair. But, the adhesively bonded composite repairs cause
minimum stress concentration and alter the load path that
induce efficient load transfer from cracked structure to
reinforcement. Thus, the reduction of stress intensity factor
caused by bonded patch repair prevents or retard crack re-
initiation or further growth. Improvements in durability and
damage tolerance have been also demonstrated through this
technology.

Several studies involving both experimental and analytical

techniques have investigated the mechanics of bonded
composite repairs on cracked thin metallic structures of about 3
mm or less thickness [1], [2],[3] and [4]. Repairs of thick plates
have also been investigated, both numerically and
experimentally, but in a limited amount [5] and [6]. A few
experiments, performed in this area, showed the difference in
crack growth rates between the repaired thin and thick
components [7]. Further, experiments with thick panels have
neglected bending effects due to asymmetric repair or
attempted to restrict bending by applying stiffeners to the
specimens [5] and [6]. Recently, Schubbe and
Mall [8] investigated the fatigue crack growth rate of thick panels
under different patch configuration. They also developed finite
element analysis involving three layers of two-dimensional
Mindlin pate elements to characterize fatigue crack growth
behavior of a thick metallic panel repaired with an adhesively
bonded composite patch [9]. Jones and Chiu [10] investigated a
series of experimental and numerical studies on the repair of
cracks in thick structural components.
In this study, we investigated the fatigue crack growth behavior
of cracked aluminum plate repaired with bonded composite
patch especially in thick plate. Adhesively bonded composite
patch repair technique has been expanded its application to the
repair of load bearing primary structure from secondary
structure repair. Therefore, a through understanding of crack
growth behavior of thick panel repaired with bonded composite
patch is needed. In this study, we investigated the fatigue crack
growth behavior of thick panel repaired with bonded composite
patch using the stress intensity factor range and fatigue crack
growth rate. Also, by considering the 3D stress state of patch
crack, 3D finite element analyses were performed to obtain the
stress intensity factor of crack repaired by bonded composite
patch. Two types of crack front modeling, i.e. uniform crack front
model and skew crack front model, were used. The stress
intensity factor calculated using FEM was compared with the
experimentally determined values.

2. Experimental
Specimens were machined from aluminum sheet (Al 7075T6).
Specimens were 220 mm long × 70 mm wide × 10 mm thick
center cracked tension (CCT) specimen as shown in Fig. 1. It has
a 6 mm diameter hole with 1 mm long crack starter notches on
both sides. Before the composite patch repair, fatigue loading
for precracking was applied until the crack propagated around 1
mm. Thus, the initial crack length, 2a, was 10 mm. The cracked
specimen was repaired with a graphite/epoxy composite patch.
The composite patch was made using eight layers of
unidirectional prepreg and had 70×70×1 mm dimension. The
composite patch was cured using an autoclave and bonded to
the cracked specimen using epoxy film adhesive.
 Fig. 1. Patched CCT specimen (dimensions: mm).

Fatigue tests were performed using a servo-hydraulic testing

machine. The test frequency was 11 Hz and it is reduced to 2 Hz
when measuring the crack length. A stress ratio of R=0.1 and
maximum stress of 45 MPa sinusoidal waveform was used to
fatigue specimens. Load and displacement signal were obtained
using a PC with data acquisition system. The fatigue crack length
was measured using a traveling microscope and a crack
propagation gage during the fatigue test.

3. Finite element analysis

In the patch repair of thin plate, there is a small crack front
growth variation in thickness direction. But, in the thick plate
repaired with bonded patch, there is a considerable crack front
growth variation between patched and unpatched side as shown
in Fig. 2. This is because the stress intensity factor variation of
thick plate in the thickness direction is more severe than that of
the thin plate. In the analysis of composite patch repair of the
thin plate, many researchers ignored the crack growth difference
between patched and unpatched side. But in the analysis of
composite patch repair of thick plate, the crack growth
difference between patched and unpatched side should be
regarded. From the observation of the cross section of fractured
specimen, the crack growth difference between patched and
unpatched side is about 10 mm at 30 mm crack length. The
crack growth difference at other crack length is assumed as
linearly varied from 0 to 10 mm. The experimental crack front is
an elliptic shape, but in the FEM modeling, the crack front is
modeled as a linear shape for simplicity.

Fig. 2. The cross section of fractured specimen.

Fig. 3 shows the model used in the analysis. Two types of crack
front modeling were used. The one is uniform crack front model
in which the crack growth rates of patched and unpatched side
are the same. Thus, the crack front is parallel to the thickness
direction. The other is skew crack front model in which the crack
growth rates of patched and unpatched side are different. Thus,
the crack front is skewed to the thickness direction. The 20 node
iso-parametric brick elements were used and the quarter-point
crack tip singular elements were used to model crack tip area.
The plate was modeled with five layers of elements and the
adhesive layer was modeled with one layer of element and the
repair patch was modeled with two layers of elements. Material
properties of aluminum plate, graphite/epoxy composite patch
and adhesive used in this analysis were listed in Table 1. The
analysis were performed at 5, 10, 15, 20, 25, 30 mm crack length
to calculate the stress intensity factor as a function of crack
length. Due to symmetry, only a half area of the plate was
analyzed.

Fig. 3. Finite element model showing the crack front configuration: (a)
crack tip model (b) uniform crack front (c) skew crack front.

Table 1. Material properties and dimensions used in FEM

Layer Length L(m Width W(m Thickness h( Material properties

m) m) mm)
Alumin 220 70 10 E=71.02
um plate GPa, ν=0.32
Adhesiv 70 70 0.1016 G=0.965
e GPa, ν=0.32
Compos 70 70 0.9 E1=134
ite patch GPa, E2=E3=10.3
GPa, G23=3.2
GPa, G12=G13=5.5
GPa,ν12=ν13=0.33, ν23
 Layer Length L(m Width W(m Thickness h( Material properties
m) m) mm)
=0.53
Finite element analysis was performed using a commercial FEM
package (ABAQUS) and *COUNTOUR INTEGRAL (J integral)
values were calculated. In linear elastic fracture mechanics,
the J integral value is equal to the energy release rate, G, and
have the relation with the stress intensity factor KI as follows,
(1)

where E′ represent E′=E/(1− ν2) in plain strain and E′=E in plain

stress condition.
The stress distribution (σy) obtained by finite element analysis
was shown in Fig. 4. The stress near the crack was considerably
reduced by patch effect as shown in Fig. 4. The reduced stress
also caused the reduction of stress intensity factor. As the single
side patching was applied in this analysis, there exists out-of-
plane bending effect and the stress intensity factor variation in
the thickness direction. Fig. 5 shows the stress intensity factor
variation of no patch specimen and patched specimen with
uniform crack front and skew crack front in the thickness
direction. The stress intensity factor of unpatched side is much
higher than that of patched side. In the skew crack front model,
the difference of stress intensity factor between patched and
unpatched side is reduced than that of the uniform crack front
model. This is because the length of crack at patched side is
short in the skew crack front model. Therefore, the large portion
of remained section restrains the crack opening and it is resulted
in the reduced stress intensity factor.

Fig. 4. Stress (σy) distribution of (a) unpatched (b) patched cracked

aluminum plate.
 Fig. 5. Stress intensity factor variation along the thickness direction: (a)
no patch (b) patched uniform crack front (c) patched skew crack front.

As the stress intensity factor of unpatched side is much higher

than that of patched side, it is needed to define an average
value of the stress intensity factor to represent the effective
stress intensity factor at a given crack length. Several average
values, i.e. average all values in the thickness direction (avg_all),
average all values except the values of both side element in the
thickness direction (avg_28), root mean square value of all values
(avg_rms), average all values from unpatched side to mid-point
(avg_half), value at mid-point (avg_mid), value at unpatched side
(avg_unpatch), were calculate and compared. Fig. 6 shows the
average stress intensity factor of patched specimen as a function
of crack length. In the uniform crack front model, avg_all,
avg_28, avg_mid show similar value and show the lowest value.
Avg_half value shows a similar value to no patch value and
avg_unpatch show even higher value than that of no patch
specimen. Therefore, it may be guessed that the avg_unpatch
and avg_half over estimate the stress intensity factor and the
effective stress intensity factor locates between avg_all and
avg_half. In the skew crack front model, all average values except
avg_unpatch at small crack length show smaller value than that
of no patch specimen. Therefore, the skew crack front model
predict the effective stress intensity factor more precisely than
the uniform crack front model. This is because the skew crack
front model is more close to the real crack front shape than the
uniform crack front.
 Fig. 6. Stress intensity factor of patched specimen as a function of
crack length from FEM results with various averaging method: (a)
uniform crack front (b) skew crack front.

4. Experimental results and discussion

4.1. Fatigue test results

To obtain the fatigue crack growth behavior of aluminum plate

with and without composite patch repair, fatigue tests of CCT
specimen with and without bonded composite patch were
performed. The crack length-cycle (a–N) curves of the base
specimen and patched specimen were shown in Fig. 7. The crack
growth rate was calculated using seven-point incremental
polynomial method suggested in ASTM Standard E-647. Stress
intensity factor, K, was calculated using the equation for a
through crack of CCT specimen shown as follows:
(2)

Fig. 7. Fatigue crack growth of patched specimen as a function of

cycle.

4.2. Calculation of SIF from experimental results

As the effect of composite patch acts as a reduction of stress

intensity factor, the fatigue crack growth behavior of aluminum
plate itself is the same in specimens with and without composite
patch repair. Therefore, the stress intensity factor of patched
crack can be determined from experimental results by
comparing the fatigue crack growth behavior of the base
specimen and composite patch repaired specimen. The Paris law
of crack propagation, i.e.

(3)

where N is the number of the fatigue cycle; C, m are material

constants; and ΔK is the stress intensity factor range, was used to
study the fatigue behaviors of cracked aluminum specimens with
and without repairs. It would be assumed that the value of
material constants C and m remains unchanged for the fatigue
analysis of the cracked specimens with or without repairs as the
effect of patch repair acts as a reduction of stress intensity
factor. Therefore, the ΔK of patch repaired specimen can be
determined experimentally from the crack growth rate, da/dN,
data and the stress intensity factor of cracked specimen without
repair. The fatigue crack growth rate–stress intensity factor range
(da/dN–ΔK) curves of base specimen and patched specimens
were shown in Fig. 8. Here, ΔK of patched specimen was
calculated using base CCT specimen stress intensity factor
relation (Eq. (2)) for comparison. The experimentally determined
stress intensity factor of patch repaired specimen was calculated
using this figure.
 Fig. 8. Fatigue crack growth rate of patched specimen as a function
of ΔK calculated from no patch CCT specimen (Eq. (2)).

To verify the validity of experimentally determined stress

intensity factors of patch repaired specimens, they were
compared with the stress intensity factors calculated using FEM.
As mentioned above, there exists a variation of stress intensity
factor in the thickness direction. Therefore, it is needed to define
an average value of the stress intensity factor to represent the
effective stress intensity factor at a given crack length.Fig.
9 shows the comparison of several averaged stress intensity
factors from Fig. 6 with the experimentally determined stress
intensity factor, ΔKexp. The stress intensity factor calculated using
FEM and experimentally determined value shows a similar trend
of change. ΔKexp located between avg_rms and avg_half values in
the uniform crack front model. In the skew crack front
model, ΔKexp shows a similar value to other averaged values from
FEM except at the large crack length region.

Fig. 9. Comparison of ΔK determined from experimental results (ΔKexp)

and averaged FEM results (ΔKavg): (a) uniform crack front (b) skew crack
front.

4.3. Prediction of fatigue life

The integration of Eq. (3) yields the fatigue life Nf of the cracked
structure
(4)

where a0 and af are initial and final crack lengths, respectively.

The magnitude of ΔK depends on crack length. Thus, a series of
finite element analyses of the cracked specimen with different
crack (as shown in Fig. 3) is performed to trace the history of ΔK.
Once the values of ΔK at different crack lengths are available, Eq.
(4) is then solved via the numerical integration method the
Gauss rule to yield Nf.
Fig. 10 shows the comparison of fatigue life between
experimental results and predicted values using ΔKexp. The
predicted values show short fatigue life than the experimental
results. If ΔKexp were used in prediction of fatigue life, it would be
a conservative design. Fig. 11 shows the comparison of predicted
fatigue life from several averaged FEM results using uniform
crack front model and skew crack front model. In contrast to the
difference of stress intensity factor between two models
especially in unpatched side value, the difference of predicted
fatigue life is not so large. The skew crack front model shows
somewhat better results. In the average values showing short
fatigue life (i.e. avg_unpatch and avg_half), the skew crack front
model shows larger values than the uniform crack front model.
But, in the average values showing long fatigue life (i.e. avg_all,
avg_28 and avg_mid), the skew crack front model shows smaller
values than the uniform crack front model. Therefore, it can be
conclude that the skew crack front model shows better results
than the uniform crack front model.

Fig. 10. Comparison of fatigue life between experimental results and

predicted values from ΔKexp.

Fig. 11. Comparison of predicted fatigue life from averaged FEM results
(ΔKavg) with uniform crack front and skew crack front.

Fig. 12 shows the comparison of fatigue life between

experimental results and several averaged FEM results using the
skew crack front model. Fatigue life using avg_all shows the
longest value and fatigue life using avg_unpatch shows the
shortest value. The experimental results lie between avg_half and
avg_rms. Thus, avg_half and avg_rms values would be good
choices in several average values for representing the effective
stress intensity factor and predicting the fatigue life. Also, this
result indicates that the stress intensity factor at unpatched side
occupy more important position in determining the effective
stress intensity factor than that of patched side. It is because that
avg_half and avg_rms values depend much upon the unpatched
side value than the patched side value. Fig. 13 shows the
comparison of predicted fatigue life from the experimentally
determined value (ΔKexp) and averaged FEM results (ΔKavg). The
predicted fatigue life using experimentally determined value lie
between avg_half and avg_rms values. Therefore, the
experimentally determined stress factor shows good agreement
with the averaged stress intensity factor by FEM.

Fig. 12. Comparison of fatigue life between experimental results and

predicted values using averaged FEM results (ΔKavg).
 Fig. 13. Comparison of predicted fatigue life from experimental results
(ΔKexp) and averaged FEM results (ΔKavg).

5. Conclusions
Fatigue crack growth behaviors of cracked plate with bonded
composite patch were investigated through experimental and
numerical study. Experiments involved fatigue tests of 10 mm
thick specimens with center crack repaired with unidirectional
graphite/epoxy patch. The fatigue crack growth and fatigue life
of thick plate were obtained. The stress intensity factor of
patched crack was determined from experimental result by
comparing the crack growth behavior of specimens with and
without repair. Also, 3D finite element analyses, which model the
non-uniform crack growth through the thickness direction, were
performed. The stress intensity factor calculated using FEM was
compared with the experimentally determined values. The
following conclusions were obtained from this study.

(1) 3D finite element analyses which model the skew crack

front growth behavior of thick specimen repaired with
bonded composite patch were performed. The stress
intensity factor of thick specimen showed a large variation
through thickness direction. Thus, several averaged stress
intensity factor were calculated and compared. The root
mean square value (avg_rms) and averaged value from
unpatched side to mid-plane (avg_half) show a good
agreement in comparing with experimentally determined
values.

(2) The da/dN–ΔK curves of CCT specimens repaired with

bonded composite patch were obtained through fatigue
tests. The stress intensity factors of patched crack were
determined from the experimental result and compared
with the calculated values using FEM.
(3) Fatigue life was predicted and compared using the
experimentally determined stress intensity factor and
several averaged stress intensity factor from FEM. In the
comparison of the skew crack front model and the uniform
crack front model, the skew crack front model showed
somewhat better result. But, the predicted fatigue life is
much more dependent on the selection of stress intensity
factor averaging method. The experimental fatigue life lie
between the predicted values using avg_rms and avg_half.
Therefore, the averaged stress intensity factor (avg_rms and
 avg_half) could be the lower and upper bound of the
effective stress intensity factor of patch repair specimen.

References
1.
o [1]
o A.A Baker, R Jones
o Bonded repair of aircraft structures

o Martinus Nijhoff Publishers, The Hague (1988)

2.
o [2]
o A.A Baker
o Repair efficiency in fatigue cracked aluminum components
reinforced with boron/epoxy patches

o Fatigue Fract. Engng. Mater. Struct., 16 (1993), pp. 753–765

3.
o [3]
o J.J Denney, S Mall
o Characterization of disbond effects on fatigue crack growth
behavior in aluminum plate with bonded composite patch

o Engng. Fract. Mech., 57 (5) (1997), pp. 507–525

 o

4.
o [4]
o Fredell RS. Damage tolerant repair techniques for pressurized
aircraft fuselages. Ph.D. Dissertation, Faculty of Aerospace
Engineering, Delft University of Technology, Delft, The
Netherlands, 1994

5.
o [5]
o Kan HP, Ratwani MM. Composite patch repair of thick aluminum
structures-final report. Airtask no.WF41-400, PE 62241. Report no.
NADC-82139-60. United States Navy-Naval Air Development
Center, Warminster, PA 18974, 1983

6.
o [6]
o R Jones, L Moment, A.A Baker, M.J Davis
o Bonded repair of metallic components: thick sections

o Theoretical and Applied Fracture Mechanics, 9 (1988), pp. 61–70

o Article | PDF (722 K) |

View Record in Scopus
| Cited By in Scopus (7)
7.
o [7]
 o Labor JD, Ratwani MM. Development of bonded composite patch
repairs for cracked metal structure-final report. Report no.
NADC79066-60. vol. 1. United States Navy-Naval Air
Development Center. Warminster, PA 18974,1980
o

8.
o [8]
o J.J Schubbe, S Mall
o Investigation of a cracked thick aluminum panel repaired
with a bonded composite patch

o Engng. Fract. Mech., 63 (1999), pp. 305–323

o Article | PDF (815 K) |

View Record in Scopus
| Cited By in Scopus (33)
9.
o [9]
o J.J Schubbe, S Mall
o Modeling of cracked thick metallic structure with bonded
composite patch repair using three-layer technique

o Compos. Struct., 45 (1999), pp. 185–193

o Article | PDF (941 K) |

View Record in Scopus
| Cited By in Scopus (30)
10.
o [10]
o R Jones, W.K Chui
o Composite repairs to cracks in thick metallic components

o Compos. Struct., 44 (1999), pp. 17–29

o Article | PDF (1441 K) |

View Record in Scopus
|
Full Text via CrossRef
| Cited By in Scopus (35)

Corresponding author. Tel.: +82-42-869-3073; fax: +82-42-

869-3210