AE load-cycle dependence applied to monitoring fatigue crack growth under complex loading conditions

AE load-cycle dependence applied to monitoring fatigue crack growth under complex loading conditions

AE load-cycle dependence applied to monitoring fatigue crack growth under complex loading conditions S. J. Bowles Acoustic emission (AE) has been used...

650KB Sizes 0 Downloads 63 Views

AE load-cycle dependence applied to monitoring fatigue crack growth under complex loading conditions S. J. Bowles Acoustic emission (AE) has been used to continuously monitor fatigue critical zones in a Mirage aircraft undergoing full-scale fatigue testing, in order to establish the feasibility of detecting AE due to crack growth during flight. The major problem was to distinguish crack growth from the many spurious AE sources. Thus, to discriminate between AE due to different sources, the activity in each zone was analysed in terms of its load-cycle dependence. Regions on the load cycle where AE due to fatigue crack growth was most likely to occur were identified, and a correlation was found between crack growth and relatively high AE activity in this region. Active cracks deeper than 0.3 0.6 mm were detected, although the size of the AE indication was not proportional to the cracking activity and zoning calibration studies had to be taken into account. No significant cracks were missed.

Keywords: acoustic emission, fatigue crack growth, load-cycle dependence

Acoustic emission (AE) is well suited to the in-flight monitoring of crack growth in safety-critical aircraft components because sensors can be permanently installed near a crack and continuous monitoring is feasible. However, the practical application of AE in this situation requires a technique that can distinguish between AE due to crack growth and AlE due to other sources (noise). Experiments conducted during constant-amplitude fatigue tests have shown that AE due to crack growth occurs at or near the peak load tl 31. This load-cycle dependence of AlE could be used to distinguish AlE due to crack growth from 1. AE due to crack-face interference effects, which mainly occurs for minimum loads and positive load gradients below the mean load level t2'31 and 2. AE due to fretting of a loading pin, which was mainly observed on the negative load gradient t2~. The load-cycle dependence of noise sources, such as fretting and rubbing, has not been determined for typical in-service loading sequences. However, the load dependence of AE due to fatigue crack growth in aluminium alloy 2024 cycled between fixed load limits, subject to a large load (overload) and subsequently cycled between the original load limits, has been studied f'q. Here AE associated with inclusion fracture in the plastic zone ahead of the crack tip (the dominant source of AlE during fatigue crack growth propagation in this alloy) occurred on the positive load gradients of the overload above the constant amplitude level and on the peaks of the load

cycles. After the overload this AE activity was reduced for many load cycles due to a decrease in crack growth rate and a material quietening in the large plastic zone ahead of the crack tip. The AE activity gradually returned to the level observed before the overload as the crack grew through the overload plastic zone. These results have been extrapolated to model the probable load-cycle dependence of AlE due to cracking during the more random loading that is typical of aircraft applications. This model also uses well-established results on the amplitude and frequency effects of loading on fatigue crack propagation [5 8]. The load dependence of AE monitored from the main wing spar of a Mirage aircraft during full-scale testing is analysed with respect to the expected load-cycle dependence of AE due to fatigue crack growth. (The wing spar was made of aluminium alloy AU4SG, which has a higher inclusion content r91 than alloys tested by Scala and Cousland t1'41 and was expected to emit AE due to inclusion fracture during crack propagation.) Finally, the load dependence of AE is compared to crack growth data corresponding to the same zones. The model Loading conditions for AE associated w i t h crack g r o w t h - region 1

Analogous with the observed load-cycle dependence of AE during the application of a single overload, it was

0308-9126/89/010007-07/$3.00 © 1989 Butterworth & Co (Publishers) Ltd NDT International Volume 22 Number 1 February 1989

7

assumed that, on any cycle during random loading, AE due to crack growth could occur on the peak or positive slope of the load cycle for loads above a minimum level, M. The steps used to calculate M, which could vary from cycle to cycle depending on the load history, are indicated below: 1. M was set equal to the average peak load level, Ate. 2. The first load cycle with an amplitude A above this level was treated as a dominating overload, and AE associated with fatigue crack growth was assumed to be possible on the positive slope and peak loads for loads levels >_M (see Figure lat. 3. After this overload, M was set equal to the overload amplitude, A, but was assumed to decrease exponentially during subsequent cycling to a lower limit of 0.6A (re above the crack-opening range 0.5A++O,1A) according to the relationship: A exp (

M

NAv~ At }

for

( exp.,

t

~

for exp

o

~

Region ,

\

ql - . . . . . . . . . . . . . . . . . . . . .

IA_v

~.~ _

_

_ #iY"

..... /it } -< 0.6

Time

! L .

ibi

Time

Fig. 1 Regions on the load curve where AE due to crack growth propagation is likely to occur (shown as heavy lines). The minimum level, M, on or above which crack growth AE may occur on a positive slope or peak load is shown as a dashed line for (a) a single overload, ( b ) a second overload in the wake of the first and (c) typical i n flight random loading

8

[;{.......... Averag-e -i F'q

Region 2

'"i

'~,.i

bL ...................

!,'. . . . . . . . . . . . Time

I

{~

-~L C

~

b'

Regi°,, ~

I~

""/

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

[

f",-

~------71- . . . . .

--

/ \

t;

~

i

:d . . . . . . . . . . . . . . . . . . . . . . . .

Time

*- . . . . . . . . . . . . . . . . . . . .

,, A v e r a g e

;-

it

,', l i

-A ~F~D-~-

peak l o a d

Time

7. . . . . . . . . . . . . . . . . . . . . .

Fig. 2 Segments of a load sequence encompassed in each of the four load regions. Data sets of events occurring in each of these regions were compared to determine the load-cycle dependeni characteristics of AE events occurring in each zone

Figure lb). The amplitude ol this overload was set equal to A, and M was reset to decay from this ne~ amplitude, ie this new overload was then used to determine future minimum load levels for probable AE due to crack growth until another overload occurred (see Figure lc). This model does not attempt to detect the low level ol AE associated with the crack growth that may occur immediately after the overload. Instead it attempts to define regions of the load curve where a high proportion of the events would be expected to be due to crack growth• These regions are collectively called region 1 ! sec Figure 2 ).

Other load regions

..,x "M

C

,

,,

~,~M

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

a



S:7 .........

Time & ........

~

Aver--~----i

NAv~> 0.6 )it ]

where N is the number of cycles since the overload and t is the exponential decay constant (in cycles)• Thus, for each cycle after the overload, AE associated with crack propagation was assumed to be possible on a positive load gradient and the peak load for load levels _> M. This models the gradual increase in the probability of detecting AE due to crack growth after the overload as the AE activity returns to its prooverload level. (The exponential decay was made proportional to the ratio A/At, to account for the larger plastic zones created by larger overloads and the subsequently longer 'quiet' AE period after the overload.) 4. The next load cycle with a peak load >_ M and >_ A~, was treated as the new (dominating) overload (see

tA

.... i~ . . . . . . . .

. . . . . . . . . . .

dL ...............................

M = 0.6A

e

AE occurring in region 1 of the load curve may not be exclusively due to fatigue crack growth. AE sources, such as crack-face rubbing, fretting and crack-face unsticking, may also occur under these loading conditions, although they would also be expected to occur at other positions on the load curve where crack growth is not expected In order that AE in different regions of the load curve could be compared and the load-cycle dependency of various spurious sources could be identified, three other regions on the load curve were defined where AE duc tt~ crack growth was not expected. These regions (sec Figure 2) were defined in terms of the load gradient and whether the crack faces were apart or together: • region 2 > 0.6A • region 3 < 0.4A • region 4 < 0.4A

negative load gradients for load levels negative load gradients for load levels positive load gradients for load levels

NDT International February 1989

i

procedure

Experimental

the lower surface of the spar in the middle of the array, was recorded, together with the zone number to which the event was allocated and the time at which the event occurred. The above measurements could be recorded at a rate of up to 300 events per second.

AE monitoring The equipment used to monitor AE from fastener holes in the lower surface of the bottom flange of the main wing spar (see Figure 3) was developed and installed by Battelle Pacific Northwest Laboratories of Richland, USA, under contract to the Aeronautical Research Laboratory. This lower surface included fastener holes that had been reamed and bushed or cold worked to retard cracking and to extend the fatigue life as part of the fatigue test. Close-tolerance fasteners were used in the rear flange holes, with holes 1-5 also having interference-fit stainless steel bush inserts (see Figure 3). Interference-fit fasteners were used in the forward flange holes.

Loading sequence During the fatigue test the aircraft was subjected to a loading sequence representative of that experienced by Mirage in service. The complete loading sequence was constructed from 24 individual test flights (TF) which represented different aspects of flying (see Figure 4). These

.

5

7

9

y

[

/"

I

~

(Z~)



[

4 Z01 Zone

0

\ SLAN

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

..... S.0g . . . . . . . . . . . . . . . . . . . ...... ~ TF6 3,0g

~,-v. . . .

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

TF15 Fig. 4 Four of the 24 flight profiles or test flights (TF) that formed the loading sequence applied to the Mirage during fatigue testing The average load level of 38 g (Av) used to calculate region 1 is also shown

11

13

I

holes

I I

/

|

15

17

C2 /

/

l i D ~| ~ ~ l

19

/

Blind anchor nut hole

/

--

"~--~-----r~----r--t--lf----f----~--f----Ik-F---~-----l~---

~

(

Mirage rnainframe

0

L

"..---~-----J A

/

.

.JA

~ Guard

~"

.

TF11

\ £1/

IRivet .I

.

• ~ . /'~M /#--'~- Av peak __ ~ o¢r - ~ - ~ - - ' \ - - ] ~ ....~ load- -~-t~q "~/ "9 V , TF2 ,-- ~ / ~" &/ 60 s Time E 22-- -Z-_5L5-gc_-_ -2LLLL--L---_-Z----......LZ. . . . Av

The signal characteristics of each AE event were measured at sensor A and only data for AE signals having a pulse height > 1.0 V after 86 dB amplification and a duration < 60 #s were accepted. For each such AE event the spar-loading information, measured by strain gauges on

Zone

.

l~l

The AE sensor array used to monitor fatigue crack growth emanating from the fastener holes is also shown in Figure 3. Sensors A, B, C1, C2, D1 and D2 (all having an air-backed disc with aperture radius 1.7 mm, length 0.9 mm and inductive tuning to 600 kHz) formed a zoning array for the spatial location of each AE event into one of 16 zones by time-of-flight analysis. Nine guard sensors (four on the web of the spar, three on the rear wing skin and two at the ends of the array) were included to reject signals arising from extraneous sources close to, but outside, the area of interest on the spar.

¢ 102 /

D2 D

.O 1 1

i o/,o,Ol

'o /,0,ol

4

8

6

Guard



10

IO106/

O \1100~

,0,o '0' 12

14

~)~12 111 h 1 1 3

16

1Fastener JnumDers

18

Fastener holes AE Sensors



Strain gauge

Fig 3 The AE sensor array (A, B, C1, C2, D1 and D2), the guard sensors and the zones 4-19 used on the lower surface of the bottom flange of the Mirage RH56 wing spar

N D T International February 1989

9

test flights were arranged, depending on the order and number of times each manoeuvre would be typically performed during a flight, to make a test block of 200 flights. The test block was applied repetitively during fatigue loading. (One test block is equivalent to 133.3 h of flying. )

r

.> 3

Procedure The Mirage loading sequence was reconstructed from the strain and time data recorded with each AE event. The loading sequence was then modelled as a sequence of overloads using an average load level of 1.0 V (3.8 g), with t equal to I0 cycles. (This corresponds to M being set at about 90% of the peak load for constant-amplitude cycling, which agrees with the results of background studies.) Only peak load levels of significant amplitude. ie > 0.3 V (1.1 g), were used to count the number of cycles since the last overload. Each AE event was thus sorted in terms of the region on the load curve where it occurred. Those AE events that did not fall into any of these regions were not considered.

Zone

5

9

11

9

11

13 13

_-

t

0

123q Region Zo[]e

:i

fI

8

10

i ~:

Fig. 5 Relative activity between the four ioad regions for each zone just before the end of the test. The shaded areas indicate the lewd of 'corrected' region 1 activity

3---

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

x

,



x

qD

AE data

10

O/

t

w

X

Zones 4, 5, 7, 9, 10 and 14 generally showed an increase in AE event rate with time; zone 9 reached approximately 120 events/TF at the end of the test and zones 4, 5 and 7 reached 60-80 events/TF, All other zones had event rates of less than 20 events/TF'.

For each zone, the activity in region 1 above the lowest activity level observed in the four load regions, at the end of the test, is also indicated in Figure 5. The variation in this corrected AE activity in region l as the test progressed is shown in Figure 6, where the activity level (in events' TF) above the 'lowest' level is plotted for each zone for the five time intervals.

!9

0 ,'

I

o~ 3p

The load-cycle dependence of the AE detected in each zone at the end of the test, ie AE activity versus load region, is shown in Figure 5. A large proportion of the activity in zone 11 occurred in region 1 of the load cycle, with relatively little activity occurring over the rest of the load cycle. The activity in zones 9 and t0 also occurred mainly in region 1, although both of these zones also had relatively high levels of activity in other regions of the load cycle. In contrast, a high proportion of the activity in zone 14 occurred in region 2 on the load curve.

17

O/

~f

~Y:r

Results

15 [

,,/ / / o

The Mirage wing spar was monitored for 47 test periods. equivalent to 6265 flight hours, and the AE data were analysed at five intervals. The analysis concentrated on the larger amplitude test flights having maximum loads of 5.0 g or more, where the overall AE level was higher and the load curve was most accurately reconstructed. For each zone, the variation of AE as a function of load-cycle position, or region on the load cycle, was determined. Here AE that occurred in each region of the load curve, for a particular zone, was expressed as a percentage of the total activity (for all the zones) that occurred in each of those regions. As the test progressed the relative AE across the four load regions was observed and compared to the fractographic and metallographic data obtained at the end of the test.

R J

)~

. - :&t

> Zone 5

7

9

Time 13

II

i5

.... ]9

l/

o o1:1o o, o, o,,.J, !

/

,'"

1.0ol "[3 .-

X

c~ ~-

L

~

~ 7 - o ~ r ' - ~_ . . . .

U

~-'-it*--~

rV,•

--~

Time Zone

~

6

8

I 0

! 3

! q

16

L

Fig. 6 Variation in the corrected AE activity in region ! for each of the five time intervals analysed

NDT International February 1989

Fractographic results

Discussion

The depths and growth curves for the largest cracks found in the wing spar [1°] are shown in Figure 7. Many cracks of different sizes were found in most of the holes monitored.

If it was assumed that a high AE activity indicated crack growth, then, from the distribution of total AE events across the wing spar monitored, it could be concluded that zones 4, 5, 7, 9, 10 and 14 contained the most active cracks.

The greatest cracking activity occurred in hole 1 and the SLAN rivet holes (zone 5), where a maximum crack depth of 5.5 mm was observed with a maximum crack growth rate of 0.4 mm/test block. Strong crack growth was also observed in hole 3 (zone 9), where three cracks were growing, the biggest of which was 5.5 mm deep and had a final growth rate of 0.3 ram/test block. The next deepest active crack occurred in hole 104 (zone 8). Here, multiple cracks grew up to a depth of 0.7 mm at rates of up to 0.04 mm/test block. All other cracks were less than 0.6 mm deep at the end of the test.

Metallographic results Metallographic inspection of the wing spar also provided evidence of fretting of the wing pin, fretting of bushes in fastener holes 1-5 and fretting of crack faces for the large cracks.

The load dependence of the AE occurring in these zones at the end of the test (see Figure 5) shows that much of the AE activity occurred in regions of the load curve where AE due to crack growth is not expected, ie regions 2, 3 and 4. Therefore, it can be assumed that most of the activity in these zones is not due to crack growth. The converse does not apply. Although there is strong evidence to suggest that AE due to crack growth will occur mainly in region 1 of the load curve, there is no evidence to suggest that all the AE activity in this region is due to crack growth. However, it is likely that most other AE sources will also be active in other regions of the load curve. In order to identify AE indications of crack growth among the noise, the level of AE in region 1 of the load cycle is analysed with respect to the AE in the rest of the load curve on the following basis: • AE activity for a particular zone occurring evenly over the four load regions (see Figure 8a), ie exhibiting no

6.0

Hole number f=

3

5.0

41[

4.0

E c-

,4

~- 3.0

Slan

L

2.0

1.o

I

2000

q000

I

I

6000 8000 10000 Simulated flights

,,~ 103 ;-~________~101 6 12000

AE Data analysed Fig. 7 Crack g r o w t h data for the largest cracks in the RH56 wing spar. Hole numbers are indicated. Cracks less than 0.2 mm deep at the end of the test are not s h o w n

NDT International February 1 989

load-cycle dependence, was assumed to be due to source(s) other than crack growth. • Most AE occurring mainly in region 1 (see Figure 8b), with little or no activity in the rest of the load curve, was assumed to be due to crack growth. • Although sources other than crack growth could also contribute to the activity in region 1, they were assumed to have different load-cycle dependence characteristics than AE due to crack growth (see Figure 8c). • Significant AE activity in region 1 and one or two other load regions (see Figure 8d) was assumed to be due to a combination of crack growth and one or more spurious sources. The load dependence of these sources could overlap so that activity in region 1 could be due to a combination of crack growth and spurious sources. If the distribution functions of each possible source were known, it would be a simple process to subtract them and to find the activity due to each source. Unfortunately, the load distribution of spurious sources under complex loading conditions is not known. Instead the lowest level of activity observed over the' whole of the load cycle was used as an approximation to the 'background' activity, caused by spurious sources, which could be contributing to the activity in region 1. Activity above this lowest level was thus assumed to be associated with fatigue crack growth. Therefore, an estimation of the crack growth was determined by the corrected region 1 activity, which is equal to the level of activity in region 1 minus the lowest level of activity observed over the whole of the load cycle, as indicated in Figures 8b and 8d. The load-cycle dependence of AE occurring in each zone at the end of the test shows that only zones 9, 10 and 11 had load-dependent characteristics similar to those expected for AE due to crack growth. These zones were

11

>.

>. > u co

4-,

co cd

1

a

2

3

4

2

b

Region of load c u r v e

>. 4o

~

q

Region of load c u r v e

>, .o

>

>

i

t~

i t

>

F

ry,

'Background'

/

-$ r¢ i

i

....... J . . . . . .

1

c

2

3

q

Region of load c u r v e

2

d

'

L J .......

[ _ i

:;

Region of load ~l~rve

Fig. 8 Simplified analysis of the AE load-cycle dependence, (a) AE activity evenly distributed implies only load-independent source(s), m crack growth is not expected. (b) A predominance of activity in region 1, where AE due to crack growth is most likely to occur, suggests that most of the AE is due to crack growth. (c) Other sources may contribute to the region activity bur were expected to have a different toad dependence to crack growth. (d) High activity in region 1 and in one or two of the other regions could be due to a combination of crack growth and other sources. The region 1 activity minus the lowest level over all the regions, indicated by shading, is assumed to give an indication of the cracking activity

also the only ones in which indications of crack growth. as given by the corrected region 1 activity (shown in Figure 6), were consistently found throughout the test,

crack in zone 9 ( hole 3 ): and the AE indication of cracking in zone 10 coincides with the growth of a 0.7 mm deeF, crack in zone 8 (hole 104).

The AE indications in zones 9, 10 and 11 did not directly correlate with the fractographic results (see Figure 7), which showed that cracks had been propagating in holes in zones 5 and 9 throughout the test and in zone 8 for most of this time. The discrepancy can be explained by the results of calibration studies [11], which showed that inaccuracies could be expected in the zone location of AE events. Correction of the load-dependent AE indication for this mis-zoning factor results in good agreement with fractography data; consistent AE indications > 0.25 e v e n t s / T F correspond to active cracks greater than 0.3-0.6 m m deep. After correction for zone location, the total AE activity still gave several false indications of cracking.

Untbrtunately, the zone 10 indication, similar in magnitude to that observed in zone~ ~) and 11, is overemphasized in relation to the crack sizes, which are an order of magnitude greater for the latter zones. One possible explanation is that AE due to a spurious source, with similar load-cycle dependence characteristics as assumed here for crack growth, has propagated into the spar from the wing fairing, which is directly fastened into hole 104. Such activity could be assigned to zone 8 or 10. The indication in zone 10 may have been reduced if waveform analysis had been performed on each AE event

The calibration studies found that • AE data events assigned to zone 9 could have been due to sources in zones, 5, 7 or 9 • AE events assigned to zone 11 could have been due to sources in zones 9 or 11 • AE events assigned to zone 10 could have been due to sources in zone 8 or 10 Therefore, the strong, load-dependent AE indication in zone 9 corresponds to active cracks up to 5.5 m m deep in zone 5 (holes 1 and SLAN) and zone 9 (hole 3); the strong AE indication in zone 11 corresponds to the large

12

The remaining smaller AE indication in zones 13 iX could correspond to some of the remaining cracks, all less than 0.6 mm deep. However, insufficient information is available on the accuracy of the zoning of these sources and no conclusions can be drawn. These cracks would be of little concern in most components and are at the threshold of detection in other N D I techniques (ey eddy current inspections). In addition to indicating significant crack growth, the load-cycle analysis (see Figures 5 and 6) also successfully rejected the high level of activity observed in zone 4: only cracks of depth less than 0.3 m m could have contributed to activity in this zone. This zone 4 activity was concentrated in regions other than region 1 on the load cycle

NDT International February 1989

and is most likely due to fretting of the wing attachment pin. Similarly, a large proportion of activity observed in zones 5, 7 and 9 occurred in regions of the load cycle other than region 1. In zone 5 some of this activity could also be due to fretting of the wing attachment pin, while fretting and rubbing of the crack faces of cracks in hole 1 and the SLAN rivet holes could have contributed to activity in zones 5, 7 and 9. The load-cycle dependence analysis also shows that the increase in activity observed in zone 14 was due to an AE source occurring preferentially in region 2 of the load curve and was not due to crack growth. This activity could be due to an extraneous source originating on, or beyond, the forward wing skin as no guard sensors were suitably positioned to eliminate noise from this direction.

Conclusions Analysis of the load-cycle dependence of AE events can be applied, even under irregular loading conditions, to distinguish AE due to significant crack growth from most of the AE noise sources occurring during flight. Active cracks of depth greater than 0.3 0.6 m m were indicated by a relatively high level of AE occurring in the region of the load cycle where AE due to crack growth is most likely to occur, compared to the activity occurring in other regions of the load curve. This correlation was only observed after a consideration of the results of zoning calibration studies and illustrates the need for careful calibration of each monitoring application. The total AE activity detected near a crack is not representative of the cracking activity, even where guard sensors are used and basic screening of AE amplitude and duration are first performed. It is not sufficient to monitor the total activity occurring in the region of the load cycle where AE due to crack growth could occur. The level of AE on the load curve where AE due to crack growth may occur must be considered with reference to the level of AE occurring over the rest of the load curve. Unfortunately, the cracking indications were not always proportional to the degree of cracking, although the degree of cracking was difficult to quantify. Research into

the load-cycle dependence of AE sources in complex fatigue cycle sequences is required before the accuracy of the procedure can be checked or improved.

Acknowledgements The assistance of Dr C.M. Scala, R.A. Coyle, I.G. Scott, Dr J.Q. Clayton and Dr D.R. Arnott is gratefully acknowledged.

References 1 Scala, C.M. and Cousland, S.MeK. 'Acoustic emission during fatigue crack propagation in the aluminium alloys 2024 and 2144' Materials Science and Engineering 61 (1983) pp 211 218 2 Graham, L.J. and Elsey, R.K. 'AE source identification by frequency spectral analysis for an aircraft monitoring application' Journal of Acoustic Emission 2 1/2 (1983) pp 47-55 3 Lindley,T.L., Palmer, I.G. and Richards, C.E. 'Acoustic emission monitoring of fatigue crack growth' Materials Science andEngineering 32 (1978) pp I 15 4 Scala, C.M. and Cousland, S.McK. 'Acoustic emission during fatigue of aluminium alloy 2024: the effect of an overload' Materials Science and Engineering 76 (1985) pp 83 88 5 Corbly,D.M. and Packman, P.F. 'On the influence of single and multiple peak overloads on fatigue crack propagation in 7075T6511 aluminium' Engineering Fracture Mechanics 5 (1973) pp 47%497 6 yon Eaw, E.F.J., Hertzberg, R.W. and Roberts, R. 'Delay effects in fatigue crack propagation' American Society for Testing and Materials' STP 513 (1972) pp 230-259 7 Nelson,D.V. 'Review of fatigue-crack-growthprediction models' Experimental Mechanics February (1977) pp 41~,9 8 Trebules, V.M. Jr., Roberts, R. and Hertzberg, R.W. 'Effect of multiple overloads on fatigue crack propagation in 2024-T3 aluminium alloy' American Society for Testing and Materials STP 536 (1973) pp 115-146 9 Martin, G.G. 'Tripartite Mirage programme report September 1979' ARL-S TR UC T- TECH- MEMO 307 Aeronautical Research Laboratory, Melbourne, Australia (1979) 10 Pell,R.A. 'Defect failure analysis report' M1/86/RAP Aeronautical Research Laboratory, Melbourne, Australia (1986) 11 Scala,C.M. and Coyle,R.A. 'Acousticemission waveformanalysis to identify fatigue crack propagation in a Mirage aircraft" Journal of Acoustic Emission 6 4 (1987) pp 249 256

Author S.J. Bowles is with the Aeronautical Research Laboratory, Melbourne, Australia.

Paper received 6 June 1988

NDT International February 1989

13