Fibre composite repair of

cracked metallic aircraft

components-- practical and
basic aspects
A A BAKER
(Aeronautical Research Laboratories, Australia)
Crack patching, the use of advanced fibre composite patches (such as
boron/epoxy or graphite/epoxy) bonded with structural film adhesives
to repair cracks in metallic aircraft components, is a significant
development in aircraft maintenance technology, offering many
advantages over conventional repair procedures based on metallic
patches and mechanical fasteners. This paper reviews selected
theoretical and experimental aspects of Australian work on this topic
and describes a preliminary design approach for estimating the
minimum thickness patch that could be employed in a given repair
situation. Finally, the paper provides a case study on our repair to the
wing skin of Mirage III aircraft. Aspects discussed include evaluation of
minimum cure and surface treatment conditions for adhesive bonding
in repair situations, potential thermal and residual stress problems,
resulting from patching, studies on overlap joints representing repairs
and crack propagation behaviour in patched panels.

Key words: fibre composite; repair; aircraft components; crack

patching; structural film adhesives; overlap joints; boron/epoxy
composite; graphite/epoxy composite

NOMENCLATURE or0 Stress in metallic component under the

patch, assuming no crack to be present
Ep Young's modulus metallic component (panel) aT Residual stress in metallic component due
ER Young's modulus reinforcement (patch) to thermal expansion mismatch with patch
Gt, Shear modulus metallic component Vy Effective shear yield stress in adhesive
GR Shear modulus reinforcement VA Shear stress in adhesive
GA Shear modulus adhesive eR Strain in the patch material
vp Poisson's ratio metallic component e~ Allowable strain in the patch material
tp Thickness or half thickness metallic component "fA Shear strain in the adhesive
tR Thickness patch "/~, Allowable shear strain in the adhesive
TR Thickness composite material per ply 8 Relative displacement in adhesive
tA Thickness of adhesive /3 Exponent of elastic shear distribution in the
WR Width or half width of patch adhesive
L* Allowable overlap length q~ Inclusion factor
LR Half patch length a crack size or half crack size
Lp Half length to restraint in region to be Ka Stress intensity factor prior to patching
patched K= Upper bound estimate for stress intensity
c~o Coefficient of thermal expansion metallic factor following patching
component K" Upper bound estimate for stress intensity
o4, Effective coefficient of thermal expansion factor, allowing for yielding in the adhesive
metallic component KR Effective or measured value of stress
OtR Coefficient of thermal expansion patch intensity factor following patching
AT Temperature change [T o p e r a t i n g - T cure] KT Stress intensity resulting from residual stress
¢r= Remote stress in metallic component CrT
normal to the crack, in absence of patch F Applied load per unit width (load intensity)

0010-4361/87/040293-16 $3.00 © 1987 Butterworth 8-Co (Publishers) Ltd

COMPOSITES. VOLUME 18 . NO 4. SEPTEMBER 1987 293
 l minimises fretting;
a ~
• no damage to hidden structure;
ttttttt • minimises undesirable structural changes by
tailoring patch to suit stress field;
• simple NDI (b/ep is non-conducting and allows use
of eddy current techniques);
• minimises balance problems on control surfaces;
• minimises foreign object damage to engines - - no
rivets to be lost.
Many of the advantages of b/ep and gr/ep follow from
their mechanical and physical properties listed in
Table 1. On the basis of its superior mechanical
properties, higher expansion coefficient and low
electrical conductivity we usually choose b/ep.
5- However, much greater availability, lower cost and
better formability favour gr/ep in many applications.
Some of our major practical applications are listed
below:
¢~ V "~Patch ~ Patch
?LI • Hercules wing planks;
Crack size, a a -- 7075 T6, 3 m m thick;
Tip reinforcement
-- stress corrosion cracking;
Total reinforcement
-- over 400 repairs since 1975, no crack growth.
• Macchi landing wheels;
Fig. 1 Schematicillustration of crack-patching technology -- Mg alloy MSR, 10 m m thick;
-- low cycle fatigue;
- - at least one lifetime extension.
Adhesively bonded patches of advanced fibre
composites, such as boron/epoxy (b/ep) or graphite/epoxy • F I l l fuselage console truss;
(gr/ep), can provide highly efficient and cost-effective -- 7075 T6, 19 m m thick:
-- stress corrosion cracking;
repairs for metallic aircraft components which have
-- no crack growth since 1980.
suffered cracking due to fatigue or to stress corrosion.
This approach, known in Australia as 'crack-patching', • Mirage III fin;
was pioneered by Aeronautical Research Laboratories -- AU4SG, 1 m m thick;
and (since about 1973) has been applied successfully in - - fatigue cracking;
a range of practical applications. Fig. 1 provides a -- no crack growth since 1978.
schematic illustration and summary of our approach
• Mirage III wing skin;
when repairing thin sheet or thin plate components.
-- AU4SG, 3.5 m m thick;
Reference 1 discusses the scope of the reinforcement -- fatigue cracking;
process for repair of defective metallic components, -- over 180 repairs since 1978.
including applications in situations other than those
• F I l l wing:
involving cracking.
-- D6AC, 6 m m thick;
Some of the advantages of crack-patching with bonded - - overstrain leading to fatigue;
composites over conventional repair procedures based -- reinforcement under development.
on mechanically fastened metallic patches can be
The above repairs are based on b/ep patches and
briefly summarized as follows:
epoxy-nitrile structural film adhesives curing at about
• thin patch: 1/2 to 1/3 thickness of aluminium alloy: 120°C. However, repairs have also been made u s i n g
• minimises stress concentrations; gr/ep (cloth) patches with epoxy paste adhesives curing
• patches easily formed to complex contours; at ambient temperature, and with b/ep patches and
• minimises corrosion: modified acrylic adhesives curing at ambient
Table 1. Relevant material, mechanical and physical properties
Material Modulus Shear Critical Specific Thermal expansion
(GPa) modulus strain gravity coefficient
x 10 -3 (°C-1) x 10 -6

b/ep 208 max 7 7.3 2.0 4.5 min

(unidirectional) 20 min 23 max
gr/ep 130 max 5 13-15 1.6 0.4 rain
(unidirectional) 12 min 28 max
Aluminium alloy
(7075 T6) 72 27 6.5 2.8 23
Aluminium alloy
(2024 T3) 72 27 4.5 2.8 23
Titanium alloy
(6 AI/4V) 110 41 8.0 4.5 9

294 COMPOSITES . SEPTEMBER 1987

temperatures. Reference 1 provides background on Table 3. Nominal tensile shear strength of single
several of our major repairs and a short case study on overlap joints prepared from AF126 or F M 7 3 M
repairs to the wing skin of Mirage III aircraft is cured under various conditions
provided at the end of this paper.
Shear strength (MPa)
In this paper, several aspects of our work on crack
Cure schedule
patching are outlined, including adhesive system and 200C 800C
bonding process selection, effect of thermal and
residual stress, design considerations and crack AF126
propagation behaviour. Greater details on the aspects
covered here are given in Reference 2. 1 h/1210C 28.4 21.3
8 h/80°C 24.6 14.0
ADHESIVE AND BONDING PROCESS SELECTION 15 h/80°C 27.4 21.3
The desirable properties of the adhesive system for FM73M
crack patching include the following:
1 h/1 21 "C 39.2 30.4
Adhesive
8 h/80°C 37.3 27.2
• Fatigue resistance - - under shear and peel stressing.
15 h/80°C 38.1 30.0
• Stress relaxation resistance.
• Minimum temperature cure, close to ambient ideal.
• Durable bond after simple surface treatment.

Pre-bonding surface treatment for metallic component

• Simple to apply in situ. Because of the high performance capability of film
• No noxious chemicals. adhesives (including temperature capability from over
• Ambient temperature process. + 80°C to below -50°C) further work was performed to
• Will not encourage stress-corrosion cracking. evaluate minimum satisfactory cure temperature and
• Will not encourage corrosion. minimum surface treatments for high durability.

Early work4 (using a novel test approach) identified Minimum adhesive cure temperature
the epoxy-nitrile structural film adhesive AFI26 Studies 5 were made to assess the minimum temperature
(supplied by 3M) as providing excellent fatigue and at which satisfactory cure of adhesives AFI26 and
stress relaxation resistance and adequate bond FM73M (by Cyanamid) could be achieved: both
durability, with alumina grit blasting as the only adhesives have a recommended cure temperature of
surface treatment. Despite this adhesive's drawback of about 120°C. Adhesive FM73M is a state-of-the-art
requiring an elevated temperature cure (120°C epoxy-nitrile adhesive with improved storage life and
recommended) it was chosen for most of our practical moisture resistance, compared with the older types
applications and has given excellent service. Efforts such as AF126.
have also been made to find suitable adhesives curing
To evaluate cure behaviour, various experimental
at, or close to, ambient temperature. Several epoxy
techniques were employed such as infra-red spectro-
paste adhesives and modified acrylic adhesives have
graphy, differential thermal analysis and scanning
been found suitable, however, these are limited to less
calorimetry. Table 2 lists the minimum effective cure
stringent repair applications than possible with the
times, based on observations of the disappearance of
film adhesives. Recent repair applications 3 were based
the epoxy band in the IR spectra for the adhesive. It is
on the Boeing PANTA (phosphoric acid non-tank
evident from these results that FM73M cures in a
anodizing) surface treatment process which provides
usefully short time at temperatures as low as 80°C. The
excellent bond durability for depot or field repairs.
results of tension tests performed on standard lap shear
specimens (Table 3) show that only a small penalty is
incurred by use of the low cure temperature. Tests 6
Table 2. Times for consumption of epoxide for to measure glass transition temperature (Tg) showed
adhesives A F 1 2 6 and F M 7 3 M "~ o - o •
that, for FM7_ M cured at 76 C, Tg is 103 C while for
the standard 120°C cure, Tg is 98°C. Adhesive AFI26
Temperature (°C) Time for consumption of epoxide behaves more normally where the 76°C cure gives a Tg
of 94°C and the standard 120°C cure gives a Tg of 128°C
AF126
Minimum surface treatments for high bond durability
Recent work 7 has been aimed at evaluating the use of a
80 >12h silane coupling agent to improve bond durability. The
90 5.5 h silane used to promote bonding of epoxy adhesives to
110 70 min aluminium is usually 7-glycidoxypropyltrimethoxy
120 35 min ,silane ('y-GPS). In our study Union Carbide silane
131 23 min A-187 ('y-GPS) was evaluated on clad 2024 T3, using
either alumina grit blasting or the PANTA process 8 as
FM73M the method of surface treatment. The treatment was as
follows:
• abrade to remove scratches:
81 6h • methyl ethyl ketone wipe
• grit blast (50 ttm alumina):

COMPOSITES . SEPTEMBER 1987 295

 1.0 1.0

0.9 120 ° C 100 °C 80 °C 120°C 100 °C 80 °C

0.9

0.8 0.8

0.7 0.7

0.6 95% 0.6

/
o 0.5 / o 0.5
8s%/ 65%
L3 0.4 / / / 85% 0.4
/ /
0.3 / / 0% /
/
7 0.3 100% 95%
7"
/ / / 0% 35%
7- 7 ~00% 95%
0.2 / / / / 7 100%
/ / / 0.2 7-
/ / / / /
77 15% / / / /
95%
0.1 / / / /
/ / / /
0.1 l~ / / 0% / /
0% /
.dl r'0% o% / //
0
/ /. ./. ./. ,/, , /,
0 -'2'3 ./ / .d .
An An GB GB An An An An An An GB GB An An GB An An GB
+ + + ÷ + ~- + + + +
S s S s S S S S S S

a b
Fig. 2 Histogram of wedge test data for (a) adhesive AF126, and (b) FM73M, following surface treatment by various techniques; An refers to the PANTA
process. GB to alumina grit-blast and S to the use of 7-GPS silane. The percent figures refer to the approximate degree of cohesive failure (failure within the
adhesive layer) for each treatment

~. (~hesive /2024 T3 (3.18 mm thick) THERMAL AND RESIDUAL STRESS PROBLEMS

Despite the numerous advantages of b/ep and gr/ep for
repair patches, the potential disadvantages resulting
from residual and thermal stresses must be considered.
Residual stresses (tension in the metal, compression in
the patch) result from the difference in thermal
expansion coefficient between the metal and the
composite patch (Table I) when adhesives curing at

• phosphoric acid gel anodize (PANTA), in some cases

only; Thermal
• immerse in 1% by volume silane solution in distilled
environment
water, in some cases only;
• air dry and apply adhesive. P
Curecycle P
Aircraft
Adhesives AFI26 or FM73M were used to form
standard Boeing wedge-test specimens 8 and were cured
at 80°C, 100°C or at the standard temperature of 120°C.
Fig. 2 shows the results of the wedge test evaluation
I
Thermal
I ResiI dual Cyclic
based on the ratio G48/Gowhere G48 is the apparent
stress stress temperature
energy release rate (based on crack length a) after 48 h i
I I
exposure at 50°C and 95% relative humidity, and Go is Adhesi
ve Crack
bond region
the apparent energy release rate a few minutes after
driving in the wedge; the closer this ratio is to unity the
higher is the durability. Use of the ratio (rather than
the usual approach of simply measuring crack growth
a) allows in part for variations in the initial crack
length ao.
From the above results it was concluded that adhesive
+ Adhesive Crack
~ bond
'
Skin
@region + I + :
FM73M (although generally not performing in the
wedge tests as well as AFI26) provides excellent
durability with the highly simplified surface treatment
of grit blasting and silane priming - - even when the cure
temperature is as low as 80°C. These m i n i m u m surface
treatment and cure conditions should provide a wide
range of repair options for structural film adhesive Fig. 3 Potential thermal-stress and residual-stress problem areas when
FM73M. making aircraft skin repairs with advanced fibre composite patches

296 COMPOSITES. SEPTEMBER1987

above ambient temperatures are employed. Thermal Both models predict that, as expected, o~], has a
stresses result when a local area of a large component minimum of zero when L R / L P -'~ 1. The one-
is heated during application of the patch because dimensional case predicts that dp has a m a x i m u m of
expansion of the heated area is restrained by the ap/2 when L R / L P -* O. However, the two-dimensional
surrounding cold structure. However, the restraint is model predicts that dp is (a~,/2)(l + Vp) which is about
highly beneficial in reducing the effective expansion 30% larger.
coefficient of the metal in the hot zone and thus
The effective KT in restraint situations is found by
reducing the level of residual stresses. Fig. 3 categorizes
substituting o~, for ap in Equation (2) and using the
some of the difficulties which may arise as a result of
resulting value of t~T to find KT from Equation (1).
residual and thermal stresses.
On the basis of experimental work ~° it was shown that,
Our results on two aspects of major concern will be
for the conditions studied, KT would be reduced from
briefly presented here.
approximately 60 MPa (rra)'/" in an unrestrained
situation to approximately 24 MPa (n'a)'j~ in a
The magnitude of the stress intensity factor KT
caused by patching restrained situation.
Residual stresses can be minimised by
It is found 9'1° that, for small cracks in a centre-cracked
panel, KT is approximately given by (i) minimising the heated area,
(ii) curing the adhesive at the lowest possible
K T ~ O'T(Tra) '~ ( 1) temperature,
(iii) pre-curing the adhesive at a low temperature and
where 2a is the crack length and ~V the residual stress then post-curing at a higher temperature,
in the metal (in the absence of the crack) is which (iv) pre-stressing the repair region in compression, this
given by approach is feasible when repairs are made to a
wing, and
tyT = t R E R E p A T ( o t P -- OtR)/(tpEp q- E R t R ) (2) (v) employing the m i n i m u m thickness reinforcement.

This finding for K T goes against the prediction u that

KT should reach a constant value at quite small crack DESIGN CONSIDERATIONS
sizes. Thus, further work is required to resolve this
aspect. However, it is at least conservative to assume The viability and effectiveness of a patch repair can be
that for fairly small cracks, KT is as given assessed from estimates of the following:
approximately by Equation (1). (i) The stress intensity range in the repaired
component.
Influence of constraint by surrounding structure on
the effective expansion coefficient of the metallic (ii) The shear strain or range of shear strain in the
component adhesive layer - - assuming strong and durable
adhesive bonding.
The effective expansion coefficient of the metallic (iii) The tensile strain or range of strain in the
component under conditions of restraint was analysed n reinforcing patch.
using a one-dimensional model, Fig. 4a, and using a
two-dimensional model22 Fig. 4b. For the one- The stress intensity range provides an indication of the
dimensional case it is shown that the effective likely rate of crack propagation following repair, while
expansion coefficient a~p is given by the other two parameters (depending on the materiaFs
allowables) are taken to be an indicator of the likely
a~, = (t~p/2)(1 - - L R / L p ) (3) durability of the repair.
Two complementary design approaches have been
and for the two-dimensional case
developed in Australia. The finite element procedure
developed by Jones u has proved highly valuable in the
0'~ = O~p(l + Vp) I/2 -- L]~/2L~, + (L~/.2L~,) In LR/L p
design of several of our practical repairs. The analy-
tical procedure developed by Rose n is highly valuable
+ (1 -- L~/L~,)/4 In LR/L P (4)
in (a) indicating simply the significance of the
material's properties and geometrical variables, and
(b) providing an approach for developing a rapid
Temperature drop feasibility estimate for a repair - - described later.

Analytical approach overview

In the analytical procedure u predictions for the various
parameters in the patching problems (ignoring bending
effects) can be obtained from a two-stage analysis, as
\ -. illustrated in Fig. 5 for a centre-cracked panel.
Briefly, the first stage of the analysis assumes that the
/ / / / / / / / / patch, thickness t R thickness tR, is bonded to one side on
an uncracked plate, thickness ta, at the prospective
a b location of the crack (Fig. 5b). The stress tx0 in the
Fig. 4 Schematic of (a) one-dimensional model and (b) two-dimensional
metal under the patch is found by modelling the region
model for estimating the influence of restraint to expansion during heating as an inclusion in the plate which is subjected to a
of a metal plate remote stress tr =.

COMPOSITES . SEPTEMBER 1987 297

 O F
ttttt t ttttt
'\\ ~.~\\\ ~\\ / t// ///1 I//iii
@
\ \ \\ \ i~.~/'. / /

O0~ oop
//I/, II lli,,, J
~,1~1 i ; , 1 1 ] x
/ "x.,'~z !! IJ ( \ \
~\

ii 1~ ii II ~1 \ \\ \ \ \ \ \
/ iI i/ / I ~ \~ \~

a b c

t ~.ff K~

I
A

d
Fig. 5 Schematic illustration of the analytical approach to crack-patching

Generally, we have that K= = a0(~rA)v2 (6)

~o = q ~ = (5) The characteristic crack length A is given by

where ¢ is a factor depending on the shape and elastic rrA = (tA/G A)Eptpfl
properties of the patch. Some relationships for q~ are
noted in the Appendix.
which, as can be seen, is independent of crack size a. A
The second stage of the analysis assumes that the relationship for an upper bound for the stress intensity
situation in the cracked region is as represented by factor, allowing for yield in the adhesive (K " ) can also
Fig. 5c where the reinforcement completely covers its be obtained) with/3 is as given in Equation (9).
surface and the applied force intensity (force per unit
Using the above approach, it is shown that conserva-
width) is F.
tive values for the key design parameters can be
It is shown (in the absence of significant yielding in obtained by considering an overlap joint of geometry
the adhesive) that the stress inten~sity factor has an as depicted in Fig. 6a, which is subjected to a force
upper bound K= (Fig. 5d) where intensity 2F. The overlap joint is a much simpler

LR
j ~

62 tA
,=
t
/ / ~ G~an
1"y

l
Shear strain ")'A

a b
Fig. 6 (a) Schematic illustration of joint specimen indicating notation used for thickness and displacements, and (b) elastic/perfect plastic stress strain
behaviour assumed for the adhesive

298 COMPOSITES . SEPTEMBER 1987

system to study theoretically or experimentally than a (v) Strain in the reinforcement
patched panel.
The peak strain in the reinforcement eR occurs over the
The parameter which is of greatest relevance to the gap h, Fig. 6a, and is given by
theory is the relative displacement & because this is
used to (a) determine K= and (b) to determine ")'A the eR = F/(ERtR) (15)
adhesive shear strain or (under cyclic loading from the
displacement range AS) to determine A y A . The shear It is shown by finite element analysis 2 that the strain in
strain range A y A is expected to be an important the reinforcement varies through its thickness and that
parameter in assessing the fatigue life of the adhesive eR, as predicted from Equation (15), is only an estimate
system. The other important parameter influencing of the average strain. The m a x i m u m strain occurs in
repair durability is the tensile strain in the the reinforcement on the surface close to the inner
reinforcement CR. adherend and (for the thicknesses of reinforcement
considered here) could be up to twice eR. The inability
Theoretical results for the overlap joint to estimate the peak eR is a major limitation of the
one-dimensional analysis.
This section lists the important analytical relationships
lbr the overlap joint (based on a one-dimensional Experimental studies on an overlap joint
analysis) used later in the preliminary design
approach. Our experimental studies ~4,15 had the main objectives
of (i) assessing the validity and accuracy of the
(i) Relationship for 6 and ")CA analytical theory based on the one-dimensional
analysis of the overlap joint, and (ii) assessing the
Examination of Fig. 6a shows that fatigue properties of the adhesive system in terms of its

YA = 8/2tA (7)
For elastic behaviour of the adhesive v < ry (Fig. 6b)
---I. I I I I
= 2F/(/~ ERtR) (8) ~ - ~ / C l i p gauge

where the elastic strain exponent [3 is given by a

[3 = {(GA/tA)(I/Eptp + 1/ERtR)}':" (9)
g
E
(ii) Relationship for ~ and ")tA, z" >- K¢
~,0.1 peR
If the force intensity (2F) exceeds the force intensity at
which the adhesive yields (2Fy) it can be shown that =E
= ~y(IA/GA)[I + (F/Fy) 2] (10)
:5
where o 0

F~, = (Vy//3)(l + S) 0 1000 2000 3000

N (cycles)
and
b

/
S = E R I R / E p I P.

(iii) Relationship for F and F' in terms of stresses 3.0

I f ~ is the applied stress at the ends of the overlap (2F) = 1557 kN m -1 y
joint then ¢

F = tpO"= = tpOo( 1 + S) (11) ~ 2.0

E
If the metallic adherend is also subject to a residual
stress aT then the effective force intensity F' is given by
1.0
F' = tp[O "~ + (rT(l + S)] (12) / " (2F) = 1335 k N r n - '
. 7 ~ A"TA ~ 0.17
(iv) Relationships for A ~ and AyA
Under tension/zero cyclic loading to m a x i m u m force I I I I
0 1000 2000 3000 4000
intensity 2F
N (cycles)
A3=3--ST (13) C
where the displacement 3T is due to stress CT Fig. 7 (a) Joint specimen showing position of clip-gauge used to
measure the gap displacement A, (b) plot of gap displacement A versus
cycles N(2F equals 1335 kN m-l), (c) plot of apparent crack size a in the
A"/A = TA -- 8T/2tA (14) adhesive layer versus N

COMPOSITES . SEPTEMBER 1987 299

allowable shear strain range Aye. Experiments to study A PRELIMINARY DESIGN APPROACH
aspect (i) were made on the specimen schematically
illustrated in Fig. 7a. The gap displacement A The main aim of the design approach is to provide a
was measured by a clip gauge which provided an feasibility assessment of repairability of a cracked
experimental value for 8 through component and, if repair is feasible, to provide
estimates of:
A=heR+6 (16) • minimum required patch thickness,
• minimum required bond overlap length,
where eR is given by Equation (15). • expected durability of the repair scheme,
In each joint the inner adherend was 2024 T3 (tp • conservative value for the reduction in stress
3.16 mm) and the outer adherend b/ep (tR = 0.51 mm) intensity.
bonded with adhesive AF126 (tA "--0.13 mm). Only the situation where the crack is completely
The main conclusion drawn from these studies was covered by the reinforcement is considered here. The
that the simple one-dimensional joint analysis for 8 reinforcement may cover the crack on one side or both
(and hence for YA and K~) was adequate (taking 2GA sides of the component. However, where reinforcement
for the adhesive as 0.54 GPa) providing corrections is only on one side, it is assumed here that bending
were made for shear lag in the b/ep (due to its effects, as produced by displacement of the neutral axis
relatively low shear modulus, GR) and for residual by the patch, are reacted out by the supporting
stress effects. The shear lag correction ~3 was made by structure.
modifying Equation (9) for 13 by replacing tA/GA with It is generally desired to employ the thinnest feasible
its effective value patch for several reasons, including to (a) minimise
residual stress problems, (b) maintain aerodynamic
(tA/GA)eff = (tA/GA -F tp/2Gp -+- 3tR/8GR) (17) acceptability on external surfaces, (c) minimise balance
problems, for example on control surfaces, and (d)
Preliminary experiments 2 to study the fatigue comply with installation restraints, ldealised patching
performance of the adhesive system were also made on situations are depicted schematically in Fig. 8, where
the specimen depicted schematically in Fig. 7a, the cracked component is modelled as a panel, either
however in this case the thickness of the adherends centre of edged cracked.
was doubled to allow an increase in the strains in the
adhesive without yielding the aluminium adherend. The logic for the design procedure, T M shown in flow
chart form in Fig. 9 and, in simplified form, in Fig. 10,
The expected forms of damage in the adhesive system is based on comparison of the following, as the patch
are (i) gross cracking or disbonding of the adhesive is increased in thickness one ply at a time:
layer, (ii) microcracking or crazing, or (iii) some more
subtle deterioration of the polymer structure. Assuming (i) the computed overlap length LR with the
cracking in the adhesive of length a on each side of the allowable length L*; the method employed to
joint then from Equation (16) obtain LR is described at the end of this section:
(ii) the computed peak strain in the patch eR with the
allowable e~, taken as 5 × 10-3 for b/ep:
A=(h +2a)eR+ A8+6c (18) (iii) the computed shear strain range A'),A in the
adhesive with the allowable ATe,, taken as 0.18
where 8c is the displacement (equal to Ac) due to creep for adhesive AF126.
of the adhesive. Increasing patch thickness (as shown later) increases
A typical plot of A versus cycles N is shown in Fig. 7b. LR but as can be seen from the theoretical results given
Crack growth in the adhesive produces a divergence of in section 4, it also reduces both eR and A'fA. This
the upper and lower bounds of the plot. From this procedure should result in specification of the thinnest
result Fig. 7c plots results of crack growth a versus N
for two levels of Ay A. Of note is the linear relationship
0 "~ , R 0"~, R
between effective a and N which is expected since, in
this specimen design, the total energy release rate G is
expected to be constant. ~ - - I Expansion
Several specimen microscopic examinations revealed 7"_7_Z_Y J J _ ZZZZ_7~ ~ rest rai nts
no gross cracking, even though damage was indicated Panel ! (Ep, tp, O~p)

~Yl
- - as shown in Fig. 7b. Where damage was found it r ~/~C~/Patch L*
occurred as cohesive cracking within the adhesive ( E R , t R , O~a )
layer. Despite the doubt (in some cases) of the nature 7.:
of the cracking, the concept of effective crack size in
the adhesive appears to be most useful for design
purposes, using A),~ for acceptable levels of effective ~_j_~_j_ 2 WR
da/dN as the major design parameter. 7_7-~__ZZZ ~ fJ_TjZ-~Z~22
The experimental value for K= is given to a first
approximation by

K= = (Ep F 6/2(1 + S)tp) '/~ (19) a b

Fig. 8 Schematic of patching situation assumed in the analysis with
which accounts for the influence of stress relaxation (a) centre crack, and {b) edge crack showing the position of restraints to
and fatigue damage in the adhesive. thermal expansion

300 COMPOSITES . SEPTEMBER 1987

 patch system which not exceed the available length
!o"L* EG
m

Ox a T R t A / and will not fail under the service loads.

GR TA / The peak tensile strain eR (rather than the strain range)
is chosen here as a criterion for damage in the
reinforcement since b/ep is relatively insensitive to
damage by cyclic loading: e l is thus taken as the
--I
--I Initialization ] allowable level of tensile strain and, at this stage, no
allowance is made for strain concentration effects. 2 It is
I ..... 1 I= relevant to note that due to residual stresses, the b/ep
I
C$~ = RTR
will usually be under compression which enhances its
(9) with (17) tensile strain capability. 1°
LH (20)

The value chosen for the allowable shear strain Aye, in

Yes +
the adhesive system is assumed to determine its fatigue
K~
life and therefore the durability of the patch system.
s
The value of 0.18 chosen for A-y~, is based on the
n - n-1 I o'o(5) (A2) fatigue test results described earlier, making allowance
ea (11) (15) for the residual stress in the test specimen employed:
I tests 2 on similar joint specimens with aluminium rather
Parameters
for reference Yes than b/ep outer adherends (where (rT should be zero)
confirmed that this AT* value was appropriate for
Repair n o t
feasible fatigue in adhesive AFI26. As indicated in Fig. 9,
o~ (2) (3) or (9)
several increments in A7 ~, above the value initially

+ , i
LX'~A(7) to (12)
taken are allowed, a warning of reduced durability is
Yes accordingly given.
In the design approach no attempt is made to design to
a specified stress intensity because of (i) the
KT 11} (3) or (4) insensitivity of K= to patch thickness, (ii) the difficulty

+ Warningof /
limited durability
of allowing for retardation effects (discussed later), (iii)
the uncertainties in the estimate of AK~, and (iv) the
large errors in prediction of crack growth due to
materials variablility and environmental influences.
Repair feasible l~

l---- I
I- Fig. 10 illustrates a simplified patching situation for the
r~ El"
t
/ '
edge-cracked configuration and lists the results
obtained from a computer program based on the flow
chart given in Fig. 9, using the input data listed in
Fig. 9 Flow diagram for patch analysis Table 4 for case 1. Table 4 also gives results for several
other similar cases.

I I ~ Input
parameters I
Example

Yl
o~,R I Inc I a ~ = 138, R = 0 . 1
thickness II ~

I 2024T3

Pa~ LP f ear L R < L*

A T = 100°C A~ A = 0.18

Free edges

Cra&
JJY L * = 80 mm
e R = 5 x 10 _3

tA = 0.19
25 mm
~3 mm

A~A < A"~A

I °utoutI
parameters @
First cycle for rain thickness
patch

Fig. 10 Schematic of patching situation, also showing simplified flow diagram for patch analysis and results for case 1, Table 4

COMPOSITES . SEPTEMBER 1987 301

Table 4. Input data and output results for minimum patch thickness, based on the design procedure. In
the table F-E signifies free edge during patching and ~ signifies an edge clamped at infinity

Case

Input 1 2 3 4 5 6 7 8 9 10

~= (MPa) 138 138 138 216 138 138 138 138 138 138
R 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1
Ep (MPa) 72 72 72 72 72 72 72 110 110 110
te (mm) 3.14 3.14 3.14 3.14 7 3.14 3.14 3.14 3.14 3.14
ao(°(C-1 ) x 10-6 2.4 2.4 2.4 2.4 2.4 2.4 2.4 9 9 9
Lp (min) F-E oo 100 100 100 100 100 1O0 oo F-E
L* (mm) 80 80 80 80 80 80 40 80 80 80
a (mm) 25 25 25 25 25 25 25 25 25 25
ER (GPa) 200 200 200 200 200 200 200 200 200 200
TR (mm) 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13
WR (mm) 80 80 80 80 80 80 40 80 80 80
GR (GPa) 10 10 10 10 10 10 10 10 10 10
e i~ x 10 -3 5.0 5.0 5.0 5.0 5.0 5.0 5.0 5.0 5.0 5.0
Q'R (°C-1) X 10 -6 4 4 4 4 4 4 4 4 4 4
GA (GPa) 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54
tA (mm) 0.19 0.19 0.19 0.19 0.19 0.19 0.19 0.19 0.19 0.19
Ay~ 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18
ry (MPa) 30 30 30 30 30 30 30 30 30 30
TA (°C) 120 120 120 120 120 20 120 120 120 120

Output 1 2 3 4 5 6 7 8 9 10

n 7 5 5 9 9 5 3 4 4 5
L (mm) 57 48 48 65 76 48 37 44 44 49
A~A 0.16 0.16 0.13 0.17 0.28 0.14 0.22 0.16 0.17 0.16
ea X 10 -3 3 4.2 4.2 4.2 4.2 6.5 4.8 4.7 3.9
~r (MPa) 67 21 -4 -6 No repair 0 No repair -7 1.2 15
AK= (MPa m-y2) 12.5 14 14 18 14 19 19 18
AK" (MPa m-'A) 25 22 18 22 -- 18.7 -- 24 25 27
AKa (MPa m-'~) 38.7 38.7 38.7 60.5 -- 60.5 60.5 60.5 60.5 60.5

Overlap length I R performance of patched specimens and then compares

the results with theoretical predictions.
An estimate for the overlap length LR required for the
patch can be obtained using the approach taken in Effective patching is expected to have two major
Reference 17, again modelling the patched region as an beneficial effects in inhibiting crack growth. It should
overlap joint. If the degree of yielding of the adhesive (i) retard the reinitiation of the crack, and (ii) reduce
is fairly small and the patch relatively thin, a the rate of crack growth once growth resumes. Both
conservative estimate for LR is given by effects result from the reduction in stress intensity
range from zXKa for the unpatched specimen, to AK R,
LR = 6//3 + length of taper (20) the effective value following patching - ideally equal to
or less than ZkK=, the m a x i m u m predicted value.
where/3 is given by Equations O) and (17). The overlap
Essentially retardation ~8 is associated with the
length 6//3 is the m i n i m u m length required to develop
formation of a plastic zone at the crack tip during
a central region of low shear stress at each side of the
loading, where the zone acts as an oversize inclusion,
joint - - the elastic trough. This region serves to (i)
resulting in a compressive residual stress from the
minimise stress relaxation in the patch under static
surrounding elastic material when the loads are
load, (ii) provide allowance for manufacturing defects,
removed. When a patch is applied the peak stress
such as voids and disbonds, and (iii) allow for some in-
intensity is reduced from K~ to KR. However, the
service deterioration. opening of the crack tip will be very much less than
To minimise peel stresses, the outer edges of the patch would normally occur at KR due to the compressive
must be tapered ~7, we generally employ a taper angle of residual stress field. The onset of further plasticity at
about 5° and very conservatively, add this length to the the crack tip will similarly be greatly reduced. Since,
elastic transfer length. during cyclic loading, the damage per cycle is
dependent on the size of the plastic zone formed
FATIGUE CRACK PROPAGATION STUDIES IN during each cycle, the result of patching is a marked
PATCHED SPECIMEN reduction in crack growth rate, until the crack grows
out of the region of the original (pre-patching) zone.
General considerations Even under service (variable) loading, some retardation
This section describes work to evaluate the fatigue effects would be expected, depending on the stress level

302 COMPOSITES . SEPTEMBER 1987

 stripping of the patch (at about 200°C). In a number of
60 specimens there was evidence of limited delamination
Unpatched of the patch on each side of the crack region, most
5O clearly indicated by the thermal-stripping technique.
t4A$
E The locus of failure was generally between the glass
40 -
scrim (used as a carrier for the boron/epoxy tape) and
Pa I' the epoxy matrix. The delamination occurred as a
oa 30 layer }!'} 13oo roughly triangular zone with its base at the edge of the
i " - b/eppatch- Y I ! panel and its apex at the tip of the crack. This mode of
2o
failure of the boron/epoxy was rarely seen ill the joint
lO ' 2024.T3 ~ "''' specimen: in these specimens damage, when it was
3.18 mm thick detected, was found to be cohesive in the adhesive
layer.
0 1.0 2.0 3.0 4.0
N (cycles) x 10s Some of the findings and conclusions from the crack
Fig. 11 Plotsof crack growth versus cycles for patched specimen propagation studies are as follows:
(AF126 adhesive) and unpatched specimen
(i) Patching efficiency is very high, as shown in
imposed on the component immediately prior to Fig. 11 which compares crack growth data lbr
patching. typical patched and unpatched specimens tested
under the same nominal loading conditions.
Following patching, and alter retardation effects on (ii) Significant retardation in crack growth is
growth are exhausted, the reduction in the rate of crack experienced with the longer (25 mm) starting
growth reflects the reduction in stress intensity range. crack, but none with the shorter (5 mm) starting
Further, since (as previously discussed) patching theory crack. This difference in behaviour between the
predicts that the stress intensity K= following two crack sizes (borne out by experience in the
patching should be independent of crack size, the rate Mirage Test Program, described later) can be
of growth da/dN is expected to be constant in explained as tbllows: the degree of retardation of
accordance with the empirical relationship crack growth following patching depends on the
daldN o~ (z~),~
reduction in stress intensity range from zkK, to
Z~fft'R. Since the minimum stress intensity range
where n is an exponent of the order of 3 to 4. AKR is expected to be independent of crack size,
Finally, two potentially important complicating factors the reduction in stress intensity range tor the
must also be considered. The first is the presence large crack and hence the degree of retardation is
following patching of the residual stress (rT, the second much greater than for the small crack.
is the influence of heat treatment following patching (iii) Crack growth da/dN under the patch often
on the behaviour of the cracked component. lbllows a parabolic relationship - see lot example
the short crack in Fig. 11. This implies that zk/t'~
Crack propagation studies lbr the patched specimen was not totally
Edge-notched specimens, as depicted in Fig. 11 (inset), independent of crack length a - contrary to
were subjected to cyclic loading at constant load theoretical expectation. This behaviour may
amplitude. The specimens conformed to the idealised have resulted from some disbonding of the patch.
situation illustrated in Fig. 8 and one of the design (iv) The degree of retardation is significantly reduced
conditions largely conformed to the situation assumed when relatively high temperature treatments are
for case 1, Table 4, in which the patch is assumed to be used. This behaviour is evident in Fig. 12. It
bonded at 120°C to one side of the panel with AF126 appears likely that the heat treatment relaxes the
adhesive - - with no edge restraint. level of elastic constraint acting on the crack,
probably by a process of creep recovery in the
In our fatigue test approach two similar specimens are plastic zone.
simultaneously tested while joined together to form a
honeycomb panel, with the patched sides facing
outwards. The aim of this configuration is firstly to
minimise curvature, caused by the residual stresses 120 °C
for 1 h
following patching: thus patches are usually bonded at 120 °C for 1 h °C
the same time as the panels are bonded to the 60 175 "C °C 120 for 1 h
for l h 175 for1 h
honeycomb core. The second aim of the configuration / f° /P
is to minimise bending of the panels which would 5O // 80 o C for 8 h ,/'
E / -/ /"
otherwise occur during testing. The moments which 40 / / / /
cause bending arise from the displacement of the
neutral axis of the metal panel by the patch. The 30
support provided by the honeycomb panel i
configuration is considered to be a reasonable 20
simulation of support that would be provided in typical (.3
! Unpatched
military aircraft structures. 10
Crack growth in the patched panels was monitored
through the patch using an eddy current procedure. 0 1.0 2.0 3.0 4.0
Following testing, the panels were evaluated for
N (cycles) × 105
disbonding or delamination of the patch, initially by
Fig. 12 Plots of crack growth versus cycles for patched specimen
X-ray radiography (using an X-ray absorbent fluid, subjected to various heat-treatment conditions during patch application,
tetrabromoethane) and finally by destructive thermal The curve labelled 175°C for 1 h is for adhesive FM300

COMPOSITES . SEPTEMBER 1987 303

 AK. (MPa m 1/2) that
10 12 14 16 18 20 24 28 R R = ( R ~ a = + a T ) / ( ~ a = + aT)
-2.8 &o
= (10.4 MPa + 67 MPa)/
R = 0.6 /R = 0.1 (104 MPa + 67 MPa) = 0.45
-3.0 ° ° • A°
The value of aT is obtained from Table 4, case I where
is obtained using Equations (A2) for the conditions
A -3.2 listed in case I. Fig. 14 shows the calculated curve for
I
Z ~ a versus a, the estimated m i n i m u m and m a x i m u m
0
-3.4 values for z3~ R (with or without residual stress) and the
E
theoretical values for Z ~ = and Z ~ ' ~
E -3.6 Although the method employed here for estimating
z~d£R is approximate (particularly since several other

¢-

o
-3.8

-4.0- 0 •
/ factors could influence crack growth - - such as
environmental isolation of the crack by the patch) the
following conclusions are reached.
(i) Agreement between AK= and Z~R is sufficiently
-4.2 MPa good for ~ = to be usefully used for preliminary
7.9-79 R=0.1 design estimation.
= 7.9-79 R=0.1 (ii) The value of z~d£" which allows for plasticity in
-4.4 " 4.0-40 R=0.1
._1 0
• 20 - 60 R = 0.3
the adhesive clearly over-estimates Z~R.
• 40 -120 R=0.3 (iii) The influence of a T on Z ~ R is not great, about
-4.6
o 6 0 - 100 R = 0.6 10%, and may not be a major concern in most
patching applications.
-4.8 (iv) Z ~ R varies (to a small extent) with crack size a in
contrast to theoretical prediction; this variation
-5.0 may result from the development of some
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 delamination damage in the patch.
Log A K a (MPa m 1/2)
Finally, a similar approach to the above was used to
Fig. 13 Plots of log crack growth rate (da/dN) versus log A K a for R provide a preliminary estimate of patching efficiency
values as indicated for a range of candidate repair adhesives. These are (i)
Flexon 241 (from Permabond), a modified acrylic
adhesive suitable for near-ambient temperature repair
(v) Use of an adhesive (FM300 from Cyanamid) applications, (ii) AFI63 (from 3M) and FM73M,
curing at the highest temperature (175°C) leads to a alternatives to AFI26 for intermediate temperature
moderately increased rate of crack growth. This is applications up to about 80°C, and (iii) FM300, for
probably due to the higher level of aT when the application to temperatures up to about 110°C.
patch is applied at 175°C. Thus further reference
to Fig. 12 shows that when the patch is applied The results of the experiments, plotted in Fig. 15
(taking R = 0.1), show that AK R for the above
at 120°C followed by heat treatment at 175°C no
adhesives falls within a fairly narrow band; for
increase in growth rate is evident - - even though
example, at a crack length of 20 m m AK R ranges from
the degree of retardation is similar. The theore-
above 11 M P a m '/2 to under 13 MPa m '/2. Adhesive
tical value of aT following patching with adhesive
AF126 is about 67 MPa (case 1, Table 4) whereas FM300 is the least efficient, as expected from the high
with adhesive FM300 aT is about 102 MPa. level of a T resulting from its high cure temperature.
Acrylic adhesive Flexon 241 shows excellent
performance, considering its ambient temperature cure.
Estimation of stress intensity, K R
In an effort to provide an approximate quantitative
assessment of ZXKR for comparison with zXK=, the A

relationship between ZkKa and da/dN was E 60

experimentally determined for unpatched specimens. #_
Data for various ratios of m i n i m u m to m a x i m u m stress 50
(R values) are plotted in Fig. 13. Using the crack
growth data for the short crack (Fig. 11), the apparent 40,
stress intensity AKR was then read off from Fig. 13 for e-

the two extreme R values 0.1 and 0.6; these results are ,~ 30
plotted in Fig. 14. As shown in the following, these R
AK"
values bracket the effective R value in the patched ~e- 20
specimen RR. / R = 0.1 ]-AKR
N ~AK®~ R = 0.6 J
For the unpatched panel, tested under the same
nominal stress conditions as the patched panel, we N End o f patch

have that 0 I() 20 30 40 50 60 70 80 90

R = amin/a ~ = 13.8 MPa/138 M P a = 0.1 Crack length, a (mm)

Fig. 14 Plots of apparent stress intensity range ~K B (derived from Fig. 13)
While, for the patched panel, the m a x i m u m stress for the specimen with the 5 mm starting crack (plotted in Fig. 1 1 ) versus
under the patch is qba" and the m i n i m u m stress is crack length a. Also shown is the calculated relationship between A K a and
R ~ a ~. Thus, allowing for residual stress aT, we have a and the values for AK® and AK', taken from Table 4

304 COMPOSITES . SEPTEMBER 1987

 Since the wing skin forms part of the main torsion box,
A 20 AKa = 1.1,5o = X ~ it is in a state of shear which results in cracks
propagating normally to the principal tensile stress at
g.E 18 ~ FM300 an angle of 45 ° to the spar as indicated in Fig. 17. The
skins consist of aluminium alloy AU4SG (a French
alloy, similar to 2014 T6) about 3.5 mm thick with
~
<1 16 ~ j A F 1 2 6
integral reinforcing risers. More recently an adjacent
fairing attachment hole region was also found to be
prone to cracking. However, the problem in this area is
substantially less critical than that associated with the
._=gc 1012
~~ e ~e temp(°C) fuel decant hole.
A b/ep crack-patching repair was proposed for repair
of both problems. The advantages over the conven-
tional alternative of a bolted aluminium plate repair
8 F241 Ambient were given as the following: (i) the b/ep repair (due to
End of patch its highly directional properties) should introduce no
$ significant local elevation of strain in the loaded spar,
6
0 10 20 30 40 50 60 70 80 (ii) mechanical damage would not be inflicted on the
Crack length, a (mm) skin 01o fastener holes), (iii) the thickness of the skin
Fig. 1 5 Plots of apparent stress intensity range AK R versus N for a
(including the repair) would impose no system or
series of candidate patching adhesives as indicated, a value of R = 0.1 is fastener clearance difficulties, (iv) eddy-current
assumed procedures could be employed to check lor crack
growth, and (v) implementation of the repair under
field conditions (with fast turn-around time) should be
CONCLUSIONS FROM EXPERIMENTAL WORK possible.
(a) For repair applications, structural film adhesive, In view of the high stresses involved and the
FM73M can be cured at a temperature well below significance of the repair, it was decided to proceed
that recommended by the manufacturer. initially with a feasibility study based mainly on design
(b) Silane coupling agent T-GPS can provide good studies and detail testing. If feasible, the aim was to
durability with structural film adhesive FM73M, apply the repair to all aircraft, to repair cracked wings
when alumina grit blasting is the only surface or to act as a preventative measure for uncracked
treatment of the aluminium alloy component. wings. The final configuration of the repairs is shown
(c) Repair of metals with fibre composite patches in Fig. 17.
bonded with adhesives curing at elevated
temperature, results in undesirable residual The decant hole patch chosen comprised of a seven
stresses. However, in most practical applications layer unidirectional laminate internally stepped, (ie,
the residual stresses will be minimised by
structural restraint.
(d) A simple preliminary design approach is
proposed to allow design of a minimum thickness
patch, based on the allowable overlap length for
the bonded patch and the allowable strain level
in the patch and shear strain range in the ...........
adhesive.
(e) Crack propagation studies on b/ep patched
aluminium alloy 2024 T3 panels under constant
amplitude loading show that (i) marked
retardation of crack growth can occur, however
retardation is most marked for cracks which were
relatively long when patched and when patch
bonding temperature is below IO0°C, and (ii)
prediction of the effective stress intensity
tbllowing patching, based on an analytical model,
is probably good enough lbr preliminary repair
design purposes.

CASE STUDY

Fatigue cracking in the wing skins of Mirage III

~ircraft
A severe fatigue-cracking problem developed in the
lower wing skins in several Australian Mirage aircraft.
The region of cracking is illustrated in Fig. 16 which
shows that it occurs from a fuel decant hole close to
the intersection of the main spar and the root rib. The
cracking appears to be related to a stress concentration
associated with the proximity of a rivet hole (used to Fig. 1 6 Silhouette of Mirage III aircraft showing where fatigue cracks
develop in some aircraft. Inset: fuel decant hole region showing nature of
anchor an attachmcnt nut) to the fuel decant hole. fatigue cracking

20MPOSITES .SEPTEMBER 1987 305

 the fairing hole region. During application of the
Inboard. Main spar patches, the panels were restrained at their ends to
simulate the restraint to thermal expansion expected in
+ + + - 4-4--z + + ..... 5 the wing and, during testing, the panels were restrained
i rr:.-'--"r-- = = = --~ ~ ' against secondary bending by using a vacuum box
technique? For comparison with the results on cracked
and patched panels, tests were initially performed on
IP.;IIIdoJ;io/" ~--~o~.~t'~)',,',:;] ,,1 I
li!!;l!
Itll I
patch
0
~ ~ IhNein'll~;ll
• I I II
Illl; some unpatched panels. Fig. 18 plots the results of
I@ I
IIIiiii ,,o ~ °/,U"']lilill ,,,,, ', crack growth for cracks initiated from a region of a
li!]il', -\ / I .d')J',!l '
I!l',ille2<...<. o o / %;~,j ,IL,,, ', panel with the fairing hole configuration. This shows
I!li;lt -': \ x
I' I t \ " - ~ , ~ v
i Ill
"-;"////
ill
ill
~k~
) oo o iI1
I
+ i[c<',,7-%2",, ;,,,7;~/ " "k k k . itl It that the panel could not withstand more than one
" "" "~ . . . . . "U-~----=---:~--'. block-loading sequence before crack growth became
catastrophic.
.x-.
Results of crack size versus programs of loading are
Drain hole patch Fairing attachment
hole patch plotted in Fig. 19; growth under the patch was followed
using an eddy-current procedure. In the test represen-
Fig. 1 7 Schematic of (a) fuel decant hole, and (b) fairing attachment ted by the dotted lines, b/ep patch repairs were made
hole regions, showing cracking and outline of b/ep patches only to a simulated decant hole region, while in the
test represented by the solid lines, repairs were made to
with the largest layer on the outside) around its simulated fairing hole and decant hole regions.
periphery. The aim of this construction is to reduce Comparison of the above results with those for the
interlaminar shear stresses and provide good external unrepaired panel, Fig. 18, shows that a major improve-
smoothness. The fairing hole patch was similar in ment in crack propagation performance is obtained in
configuration but only six layers thick. The b/ep all cases. Comparison of the various results in Fig. 19
patches were to be bonded with AF126, employing the shows that they differ in the degree of retardation of
PANTA process as the method for surface treating the crack growth. In particular the short (simulated fairing)
wing skin. On the basis of the more recent work cracks show very little retardation. This result is in
described in this paper we would now probably choose accord with that reported for the comparison between
adhesive FM73M (cured at 80°C-100°C) with the skin long and short cracks in Section 6 and can be
initially surface treated by grit blasting and then explained by the reasons given there - - even though
primed with y-GPS. the tests reported in this section were conducted under
The finite element procedure was employed for the programmed loading.
design studies, as outlined in Reference 3. The On the basis of the above, and other tests, it was
conclusions of this study were that the strains in the decided to proceed with implementation of the repair
patch and (apart from a few small regions) in the program to RAAF aircraft.
adhesive were acceptably low and that the reductions
in stress intensity for the cracks was very significant. Repair implementation: decant hole repair
Repair qualification Patches were manufactured by a modified autoclave
moulding procedure from pre-preg tape. Because of the
Two main tests were performed to qualify the repair,
large number of patches required, a numerically
(i) a strain survey on a Mirage wing with a decant hole
controlled laser was first employed to cut the tape into
repair patch installed to check that the patch did not
plies containing the design holes and cut-outs. The
significantly elevate the local strain in the spar, and
bonding surface of the patches was prepared both by
(ii) fatigue tests on aluminium alloy panels configured
removal of peel ply and by alumina grit blasting and
to simulate the cracked areas to check that the
covered with a film of (uncured AF126) for shipment,
predicted reduction in AK could be achieved, and that
the patch/adhesive system could endure the fatigue
loading. The strain survey ]9 confirmed the expectation
that no significant strain elevation occurred in the spar
under load after patching.
Tension-tension fatigue tests 3,z° were undertaken on 36
panels made from alloy 2024 T3 and later, from
BSL-104 (similar to AU4SG). Only selected results on
BSL-104 panels are described here, since these most
closely represented the repair situation. The tests were
undertaken using a block loading sequence 2° con-
2o
structed from Mirage usage data. Each complete
sequence nominally corresponds to a year of aircraft ¢O
operation. Only uniaxial tension fatigue loads were
applied which were intended to represent the principal 15 'c~
tensile stress component of the shear stress present in
10 . S a w
the wing (compression stresses at 90 ° were thus
9 i - - ~-Hole'B' . .....
neglected). 678987654321123456788765432112345
Cracks were initiated from saw-cuts and propagated by Load levels 1
fatigue under the block loading sequence to the desired L - - One program- -
size. The usual size was up to 30 m m from each side of Fig. 18 Plot of crack length versus cycles for an unpatched specimen
the decant hole and up to 15 m m from each hole for with the cracks initiating from the fairing hole region

306 COMPOSITES . SEPTEMBER 1987

 ACKNOWLEDGEMENT
R.oot ~ . / ~
r'b/~ ,~Spar T h e a u t h o r is p l e a s e d to a c k n o w l e d g e the theoretical
work o f D r F r a n c i s Rose a n d the e x p e r i m e n t a l work o f
E ,A
E 50 / B~,~A
M r Bill Broughton, M r Giles Paul a n d M r J o h n
Roberts, e m p l o y e d in the m a i n b o d y o f this paper. He
40 / PB D is also p l e a s e d to a c k n o w l e d g e the m a n y A R L staff
............ _. ~ 2-"e--~ ¢ - - - ~ involved in the d e v e l o p m e n t o f the M i r a g e skin repair,
30 -~ a n d in p a r t i c u l a r wishes to t h a n k M r Brian H o s k i n for
(.3 his c o n t r i b u t i o n s a n d e n c o u r a g e m e n t .
D This p a p e r was first presented at an A G A R D meeting
on "The R e p a i r o f Aircraft Structures involving
10 i ~ C o m p o s i t e Materials', Oslo, Norway, April 1986, a n d is
% p u b l i s h e d here by kind p e r m i s s i o n o f the A G A R D
5 10 15 20 25 30 35 90 95 100 Structures a n d Materials Panel.
Programs
Fig, 19 Plots of crack growth versus cycles for patched panels. Solid
lines are results from one panel and dotted lines from another REFERENCES
I Baker, A.A. ~Repair of cracked or defective metallic aircraft
u n d e r refrigeration to the R A A F bases where the components with advanced fibre composites -- An overview
repairs were applied. of Australian work" Composite Structures 2 (1984) pp153-181
2 Baker, A.A. and Jones, R. (editors) "Bonded repair of aircraft
To aid i m p l e m e n t a t i o n o f the r e p a i r by ( A R L trained) structure', to be published in the series "Applications of
R A A F technical personnel, a c r a c k - p a t c h i n g g r o u n d fracture mechanics', general editor G.C. Sih by Martinus
s u p p o r t unit was developed. This unit c o n t a i n s all Nijhoff. Chapter 6. "Crack patching, experimental studies
necessary e q u i p m e n t i n c l u d i n g (i) a gritblaster, (ii) an practical applications" by A.A. Baker
3 Baker, A.A., Callinan, R.J., Davis, M.J., Jones, R. and
a n o d i z i n g unit, (iii) t e m p e r a t u r e controllers to control Williams, J.G. 'Repair of Mirage III aircraft using the BI-RP
cure t e m p e r a t u r e at "" 120°C, a n d (iv) a patch crack-patching technique' Theoretical and Applied Fracture
pressurisation system b a s e d on a h y d r a u l i c jack. Mechanics 2 (1984) ppl-15
4 Baker, A.A "Evaluation of adhesives for fibre composite
T h e r e are eight m a j o r steps in the r e p a i r procedure: n reintbrcement of fatigue cracked aluminium alloys" 5NMPE J
(a) Aircraft p r e p a r a t i o n : seal fuel leaks ( i n c l u d i n g the 15 (1979) pp 10-17
5 Davidson, R.G., Ennis, B.C., Mestan, F.C. and Morris, C.E.M.
crack inside the wing skin) a n d detail the crack
Technical Report, OCD, 84/3 (Materials Research
positions. Laboratories. Australia. 1984)
(b) B l a n k i n g off: seal the d e c a n t hole with an 6 Pearce, P.J., Morris, C.E.M., Ennis, B.C. (Materials
a l u r n i n i u m disc b o n d e d with p o l y s u l p h i d e Research Laboratories, Australia) work to be published
rubber. 7 Broughton, W. and Baker, A.A. (Aeronautical Research
(c) P r e h e a t drying: dry out any fuel which has not Laboratories. Australia) work to be published
been removed. D u r i n g this o p e r a t i o n the wing 8 Lock, M.C., Horton, R.E., MeCarty, J.E. "Anodize
optimisation and adhesive evaluation for repair
tanks are p u r g e d with nitrogen to e l i m i n a t e risk applications' AFML-TR-78-104
of ignition o f fuel vapour. 9 Baker, A.A., Hawkes, G.A. and Lumley, E.J. 'Fibre composite
(d) M a s k i n g off: cover regions s u r r o u n d i n g the repair reinforcement ot cracked aircraft structures' Proc Second lnt
with a d h e s i v e tape to prevent c o n t a m i n a t i o n with Cmt/" on Composite Materialx Metallurgical Socic"O' o/'AIME,
p h o s p h o r i c acid gel. Toronto. Canada. 1978
(e) G r i t blasting: surface is first c l e a n e d with 10 Baker, A.A., Davis, M.J. and Hawkes, G.A. "Repair of fatigue
cracked aircraft structures with advanced fibre composites:
detergent followed by solvent until water b r e a k residual stress and thermal fatigue studies" Proc lOth hit
clean, then t h o r o u g h l y grit-blast with fresh Committee on Aeronautical Fatigue (ICAF) Syrup, Brussel,~; May
alumina. 1979 paper 4.3
(f) A n o d i z e the surface: use the p h o s p h o r i c acid gel 11 Rose, L.R.F. 'A cracked plate repaired by bonded
a n o d i z i n g p r o c e d u r e - - c h e c k for a n o d i z e c o n d i t i o n , reinforcements'hit J Fracture 18 (1982) ppl35-144
repeat if required. 12 Jones, R. and Callinan, R.J. 'Thermal considerations in the
patching of metal sheets with composite overlays" J Struct
(g) A d h e s i v e curing: b o n d patch to wing at a b o u t Mech 8 (1980) ppl43-149
120°C with direct pressure. A g a i n nitrogen 13 Jones, R. and Callinan, R.J. 'Finite element analysis of
flushing is e m p l o y e d to e l i m i n a t e fire risk. patched cracks J Struet Mec'h 7 (1979) pp 107-130
(h) F i n a l sealing: seal various regions a n d cover 14 Baker, A.A., Roberts, J.D. and Rose, L.R.F. 'Use of overlap
patch with a protective layer o f a l u m i n i u m to act joint parameters in estimating K reduction due to crack
as a moisture b a r r i e r to m i n i m i s e e n v i r o n m e n t a l patching" Proc b~t Workshop on Dc~'nce Applications c?/"
Advanced Repair Technology. Naval Research Laboratories;
d e g r a d a t i o n , p a r t i c u l a r l y for protection against
Washington DC USA (1981) pp335-345
p a i n t stripper. 15 Baker, A.A., Roberts, J.D. and Rose, L.R.F. 'Experimental
study of overlap .joint parameters relevant to the K reduction
G e n e r a l l y the repairs were a p p l i e d d u r i n g routine due to crack patching' Proc 28th National SAMPE Svmpo,qum.
service o f the aircraft so as to m a x i m i s e aircraft Azusa. Sarape (1983) pp 627-639
availability, a n d required two to three w o r k i n g days. 16 Baker, A.A. and Packer, M.E. (Aeronautical Research
The p a t c h i n g p r o g r a m c o m m e n c e d with e x p e r i m e n t a l Laboratories, Australia) work to be published
trials in D e c e m b e r 1979 a n d was c o m p l e t e d by O c t o b e r 17 Hart-Smith LJ. "Analysis and design of advanced composite
1981 with patches a p p l i e d to a b o u t 180 M i r a g e wings. bonded ,joints" NASA ('R-2218 (1974)
18 Broek, D. 'Elementar3' Engineering Fracture Mechanics'
To date, the r e p a i r a p p e a r s to have been highly (Sijthoff and Noordhoff t978) chapter Ill
successful, a l t h o u g h a few cases o f c r a c k growth 19 Hoskin, B.C. and Baker, A.A. Structures 7?chnical
following p a t c h i n g were r e p o r t e d for wings with large Memorandum 321 (Aeronautical Research Laboratory.
cracks further growth has ceased in all b u t a few cases. Australia. 1980)

COMPOSITES . SEPTEMBER 1 987 307

20 Van-Blaricum, T.J. Structures Technical Memorandum 404 ~ = (1 + S) -1 (A1)
(Aeronautical Research Laboratories, Australia, 1985)
21 Davis, M.J. and Roberts, J.D. Materials Technical
Memorandum 373 (Aeronautical Research Laboratories, For the case of main interest here, the unidirectional
Australia, 1981) patch is modelled as an orthotropic inclusion of
22 Rose, L.R.F. lnt J Solids and Structures 17 ( 1981 ) pp 827-838 elliptical shape and allowance is made for biaxial
stresses
AUTHOR q5 = (1 + S)-'[1 -4- P + 1,'pS(]~ Jr- q)l (A2a)
The author is with the Department of Defense,
Aeronautical Research Laboratories, 506 Lorimer where vp is Poisson's ratio for the metal plate and Y~ is
Street, Fishermen's Bend, PO Box 4331, Melbourne, the ratio o~Rr = and o~ is the stress parallel to the crack
Victoria 3001, Australia. direction. The values of P and q are obtained by
solving the following simultaneous equations.

APPENDIX
I -~- 2(1 -]- S ) / ~ 1 / v p S - - (l + S)/~ 2
Inclusion factor q5 (see References 11, 22) -}i + / 1 + (l + + 2(1 + S)X-lR
The reduced stress in the prospective location of the
patch is given by P 1
- - S ( I - - 1,'p]~";~) ~ (A2b)
q -vp
~ 0 = q~°'~

In the simplest case where the patch totally covers the where X = G R / E R x , ER~ is the transverse modulus lor
panel the situation is similar to a lap joint, then the patch and R = L R / W R.

308 COMPOSITES . SEPTEMBER 1987