Advances in the Bonded Composite Repair o f Metallic Aircraft Structure phần 8 ppsx

390 Advances in the bonded composite repair of metallic aircraft structure

(assumed to apply for this specimen configuration):

(13.11)

Since, from Figure 13.12, omax is around 160 MPa and a is 33 mm, Kcdt is estimated

to be about 56 MPam'/2. Similar results for were obtained from several other

unpatched panels. These values for I& are in reasonable agreement with published

values for 2024T3 panels of this thickness.

For the patched panel, patching theory suggests that K, is approximately

53MPam1/*. Although Ko0 is fairly close to Grit, the former is an upper-bound

estimate of stress intensity so it is tentatively concluded that crack propagation in

the metal was not the cause of the failure.

Strain capacity analysis

A direct estimate, using joint theory, of net strain in the patch over the crack

indicates a value of 7100 microstrain. However, if the extra load attracted to the

patch (as a result of the inclusion effect) is considered, the strain could be as high as

9500 microstrain. Since strain capacity of the boron/epoxy is measured to be about

7300 microstrain, the conclusion is that failure was probably a result of initial

failure of the patch.

Furthermore, as discussed in reference [ 11 for the patch configuration employed,

the ratio (inner-surface strain)/(outer-surface strain) in the patch is significantly

greater than unity. In this case it is estimated to be about 2.5. On this basis the inner

strain could have exceeded 12 000 microstrain; however, the strain elevation would

be very localised.

The conclusion is thus reached that failure in the patched panels resulted from

initial failure of the patch, possibly associated with the strain concentration at its

inner surface.

This failure mode may change where significant disbond growth occurs during

fatigue cycling for two reasons:

0 Stress intensity K, may exceed Lt allowing the crack to grow catastrophically

0 The strain concentration in the patch over the crack will be reduced if even minor

Thus, for a small disbond, say a fewmm, residual strength is likely to increase

because of the reduced stress concentration in the patch.

Increasing the thickness of the patch, say to nine layers (the current patch is

seven layers), should provide some increase in residual strength. However, at higher

stress levels, plastic yielding of the metal around the patch (exacerbated by stress

concentrations at the ends of the patch) will limit this increase. The failure mode is

then expected to change from patch failure to disbonding from at the ends of the

patch.

under the patch.

disbonding occurs.

Chapter 13. Boronlepoxy patching efficiency studies 391

450

400

350

I

5 300

250

b

m

-

m

; 200

2

u)

150

100

50

0

...............................................................................

........

I.,

-

.... ...................

:onstant Amplitude a=% FALSTAFF a=39 mm Fllla=38 mm No Fatigue a=30 mm No Fatigue a=33 mm Standard Boron Standard Boron Standard Boron Unpatched I I Standard Boron

Fig. 13.13. Histogram showing residual strengths for patched panels with or without prior fatigue

testing and for an unpatched panel. The results for the panels with no prior fatigue are plotted in

Figure 13.12.

Residual strength following fatigue testing

Tests were also conducted on panels after fatigue testing under (a) constant

amplitude, (b) F-1 11 spectrum loading-representative of the F-1 11 lower wing skin

or (c) FALSTAFF spectrum, representative of a standard fighter lower wing skin.

Figure 13.13 depicts the results together with those patched after fatigue

cracking. Thermographic NDI was used in an attempt to detect disbond damage

over the crack region in the fatigue-tested specimen; however, damage could only

be detected in the FALSTAFF specimen as a relatively small -2mm ellipse

centred on the crack. This does not imply that the other specimens had not suffered

damage, only that the disbonds were probably smaller or for some reason less

detectable by thermography.

The first conclusion is that the residual strength has not been reduced by cyclic

loading for cracks in the 30-40mm range. Indeed the strength may have actually

increased due to the reduction of stress concentration around the crack caused by

any local disbonding. In the case of the 56-mm crack residual strength was clearly

reduced compared to the others. Since this crack is approaching the boundary of

the patch, it is possible that in this case the critical stress intensity for the crack in

the panel was exceeded, rather than the failure stress of the boron/epoxy. In all test

panels the strength equalled or exceeded oy - although, with no margin in the case

of the panel with the 56-mm crack.

As discussed later, there is a case for equating oJ, with DUL. If this case is

accepted it can be concluded that the patched panels had adequate residual strength

to satisfy most certification requirements.

392 Advances in the bonded composite repair of metallic aircraft structure

13.5. An approach to b/ep patch design

13.5.1. Cyclic loading

Assuming that environmental degradation of the adhesive is not an issue

(through good quality control), the margin of safety, efficiency and durability of a

repair to a cracked component can be assessed from estimates of the following:

(a) The stress intensity range AK and R in the repaired region. This determines

patching efficiency through the crack-growth parameters AR and nR.

(b) The tensile strain eR in the b/ep patch which allows estimation of the margin of

safety for failure of the patch. It is assumed for a composite patch that fatigue

is not an issue; if it were then the range of strain AeR and R ratio would have to

be considered.

(c) A (validated) damage parameter in the adhesive system (including the

composite interface). Possible parameters are the shear strain range Ay or

Mode I1 energy release rate AGII. This allows estimation of the fatigue

durability of the adhesive system. It is best, if feasible, to design the repair so

that the damage threshold of the adhesive system over the crack is not

exceeded; however, if it is not feasible the disbond growth rate, db/dN (Section

13.2.3) must be included in the analysis, using Eq. (4). Limited disbond growth

over the crack is acceptable, however, and within limits will not dramatically

reduce patching efficiency.

Another important factor needed for design of the repair system is the length L*

available for the patch between obstructions (Figure 13.14), since this can limit the

allowable patch thickness. The length LR required for efficient load transfer

depends on the patch and adhesive parameters (Figure 13.3) including patch

thickness tp and the taper rate at the outer ends of the patch.

Assuming largely elastic conditions in the adhesive (as required to avoid patch

system fatigue), a conservative estimate of the patch length [l] is given by:

6

LR = - + length of the taper , D (13.12)

where /3 is given by Eq. (Id), The taper rate for b/ep we use is around 3 mm per ply.

Finally, the residual stress oT, resulting from patch and component thermal

expansion mismatch, must be included in the analysis, since this influences Ay, eR

and RR. Residual stress CT depends on AT= (Toperating temperature - Tcure temperature),

typically 100 "C for a 120 "C curing adhesive and, Aa = (@pat& - acomponent). The

length between thermal expansion constraints in the component structure (see

Figure 13.13) influences acomponent which for full constraint is 0.5 aP.

Based on Rose's analysis described earlier, the author [l] developed a simple

algorithm for estimating the minimum thickness patch that could be applied within

the installation constraints that would survive the external cyclic loading.

It is generally desired to use the thinnest patch feasible for several reasons,

including (a) to minimise the residual stress problems, (b) to maintain aerodynamic

Chapter 13. Boronlepoxy patching eficiency studies 393

Patch

Craack

PARAMETERS

FIRST CYCLE FOR MIN THICKNESS

PATCH

Fig. 13.14. Outline of algorithm for designing the minimum thickness patch.

acceptability, for example to minimise disturbance to the airflow when repairs are

made to an external surface, (c) to minimise balance problems; for example, when

repairs are made to a control surface, and (d) to comply with installation restraints,

for example, not to exceed available fastener lengths when fasteners must pass

through the patch for system requirements, or to maintain clearance between

moving surfaces.

The logic for the design approach is shown in flow chart form in Figure 13.14,

which is based on comparison of the following, as the patch is increased in

thickness one ply at a time:

0 The computed patch length LR with the allowable (available) length L*

0 The computed styin in the patch compared with the experimentally determined

allowable strain e,; a value of 5000 microstrain was found to be reasonable for b/

eP.

0 The computed shear-strain range compared with experimentally determined

allowable A?* = 0.18 was originally used for FM73, but current work suggests

that 0.10 may be more appropriate for long life repairs.

These patch and adhesive allowables were obtained from tests on representative

bonded joints. Increasing patch thickness increases LR but reduces eR and Ay.

Assuming constant amplitude fatigue at Bo, and R, Figure 13.15, shows the

outcome of a calculation based on the parameters listed.

394 Advances in the bonded composite repair of metallic aircraft structure

~~138, Rz0.1

2024T3

AT=IOO”C

FREE EDGES

25 mm L* = 80 mm

EXAMPLE

A?*. ~0.18

e*R= 5x1 O3

t. = 0.1 9 mm

3 mm -

7 plies blep

eR = 3x1 o3

ATA ~0.16

A K, = 12.5 MNm’”

A K, = 40 MNm’”

uT=67 MPa

LR= 57 mm

Fig. 13.15. Outcome of an analysis for the minimum patch thickness, AKa is the stress intensity for the

unpatched case.

Once AK, is estimated the inspection interval N can be determined from

Eq. (13.2) and (la) or (if disbonding is a consideration) from Eq. (13.4) as:

(13.13)

where ai is the initial crack size and ax is the size chosen for inspection. Typically ax

would be less than one third patch width to provide at least three chances of finding

the crack before it grows out from under the patch.

As shown in Figure 13.14, if the inspection interval is too short, (the AK

reduction is inadequate) there is an option to increase the thickness of the patch

providing it can still fit within the allowable length.

13.5.2. Spectrum loading

Crack-growth analysis is significantly more complex under spectrum loading. It

is feasible to assess crack growth for the cracked component and damage growth in

the adhesive system on a cycle-by-cycle basis for the various values of effective

AKo, and R.