Engineering Failure Analysis 35 (2013) 387396

Contents lists available at SciVerse ScienceDirect

Engineering Failure Analysis

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l oc a t e / e n g f a i l a n a l

Investigation of a Boeing 747 wing main landing gear trunnion failure

S.A. Barter
Defence Science & Technology Organisation, Air Vehicles Division, 506 Lorimer St., Fishermens Bend, Victoria 3207, Australia

article info abstract

Article history: A loud noise was heard from the vicinity of the port wing landing gear during pushback of a Boeing 747-300
Available online 28 April 2013 from the terminal at Sydney (Australia) airport. Inspection showed that one of the wing landing gear trunnion
fork assemblies had failed. Detailed investigation revealed that the trunnion had failed by fatigue cracking.
Keywords: Deep machining grooves were found at the root of an internal radius that had not been shot-peened as
Ultra high strength steel required, and a chemical surface process during manufacture had resulted in shallow intergranular attack at
Landing gear the bottom of these grooves. It is probable that the critical cracking started from some of these grooves. In
Fatigue
addition, the wall thickness at the failure location was significantly less than the minimum required in the
Shot peening
drawings.
Non-destructive inspection
Quantitative fractography
Since the deep machining grooves, the lack of peening and the intergranular attack were all consequences
of manufacturing, the fatigue cracking probably started shortly after the component entered service. This
implies that fatigue cracking was present during all the trunnion overhauls, but was not detected by non-
destructive inspections during the over-hauls. Quantitative fractography was used to produce a crack growth
curve based on frac-ture surface markings thought to represent the overhaul timings. The crack growth curve
suggested that the fatigue cracking was large enough to be detected by inspection during the last overhaul, if
not the one before. However, it was probably not easy to detect the cracking. This investigation therefore
highlights the difficulties that can be encountered when inspection is the last (or only) line of defence against
failure owing to unexpected manufacturing deficiencies.

Crown Copyright 2013 Published by Elsevier Ltd. All rights reserved.

1. Introduction

A loud noise was heard from the vicinity of the port wing landing gear during pushback of a Boeing 747-300 from the terminal at Sydney
(Australia) airport. Inspection showed that the Trunnion Fork Assembly Wing Landing Gear, hereafter called the trunnion had failed. The
Australian Transportation Safety Bureau (ATSB) requested the Defence Science and Tech-nology Organisation (DSTO) to carry out a
metallurgical investigation of the failed trunnion. The essential findings of this investigation were used in the final ATSB report [1].

This paper describes the investigation, which established that the trunnion failed by fatigue cracking, and includes Quan-titative Fractography
(QF) to establish a fatigue crack growth curve from overhaul schedules. Although the validity of such a crack growth curve may be questioned,
the results strongly suggest that the cracking was large enough to be detected by inspection during the last overhaul before failure.

Tel.: +61 3 96267396; fax: +61 3 96267264. E-mail

address: simon.barter@dsto.defence.gov.au

1350-6307/$ - see front matter Crown Copyright 2013 Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engfailanal.2013.03.026
388 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396

Fig. 1. A schematic of the part is shown, with the internal bore marked by the grey lines and a thick black line indicating the approximate path of the failure.
An expanded view of a section of the wall is shown in the dashed box. This indicates several features of interest noted in the investigation.

1.1. Trunnion background information

The trunnion was reported to have been manufactured from the ultra-high strength steel 4340M. The manufacturers drawings indicated that
1
the trunnion was forged, machined to size, steel-shot-peened all over, and electroplated with a Tita-niumCadmium (TiCad) coating before
painting. The trunnion consisted of a round tube with various protrusions on the outer surface, including two lugs for attaching the landing gear
drag brace. One end of the tube had an integral fork end fitting that attached to the main strut; the other end had an outer ball shape that fitted
into a bearing attached to the wing. The engineering drawing of the trunnion indicated that the inner surface of the tube had a diameter change
2
about halfway along the bore from a diameter of 135.64 mm (+/ 0.25 mm) to 99.06 mm (+/ 0.25 mm). These diameters were connected by a
shallow internal ramp with a re-entrant radius to the ramp of 25.40 mm. The outside diameter directly over this radius was 146.56 mm, giving
minimum wall thicknesses of 5.33 mm for a concentric bore. The drawing allows a minimum wall thickness of 4.57 mm to allow for some non-
concentricity in the bore diameter with the outer surface diameter in this region. This is probably required due to the complicated form of the
outer surface. A schematic of the trunnion is shown in Fig. 1 with several of these features indicated.

The manufacturers design philosophy with regard to fatigue was not known. However, discussions with the aircraft oper-ator suggested that
the trunnion life was to be determined from a retirement-for-cause or on-condition approach, whereby repairs (which were allowed in some
areas) or retirement depended on the trunnions condition at overhaul. This approach allows an effectively indefinite life as long as unrepairable
damage is not found.
Overhaul was specified to be done regularly with a maximum allowable interval of 12,000 landing cycles or 8 years of service, whichever
came first. However, details of the life and period between overhauls in Table 1 indicate that other oper-ational considerations dictated the
overhaul schedules. In fact, the trunnion had been fitted to five different aircraft, with overhauls each time. Failure occurred after 25,095 landing
cycles, which would have been 6509 cycles, or another 4 years, before the next normally scheduled overhaul.

Table 2 lists the expected manufacturing and overhaul procedures relevant to the present investigation. Note that shot-peening is not part of
the standard overhaul procedure. Peening would be required only if an area was blended to repair damage, and then only on the blended area.

2. Investigation

The failure occurred about halfway along the trunnion, severing the entire cross-section. Fig. 1 shows that the failure was between the drag
brace attachment lugs and the fork end. Fig. 2 shows the two parts of the trunnion after fracture. The frac-ture is clearly seen to run through the
entire cross-section of the tube at about the middle of the component. The circles in Fig. 2 encompass an area of the fracture that was corroded
and appeared to be significantly older than the rest of the fracture.

1
A cadmium plating process used to provide ambient temperature corrosion resistance for high strength steels in fracture critical applications such as aircraft landing gear and
naval arrestor hooks. A typical standard for this process is SAE AMS2419C (2003-05-07).
2
Since the part was designed in inches and the drawings were only marked in inches, measurements reported here have been converted from the inch values and rounded to two
decimal places.
 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396 389
Table 1

History of component under investigation.

Date Total landing cycles since install Incremental landing cycles Overhaul
10/1979 3517 3517 1
12/1986 14,069 10,552 2
02/1992 16,508 2439 3
10/2001 22,604 5996 4
05/2005 25,095 2491 Failed

Table 2

The expected procedures relevant to the investigation used to manufacture and overhaul the component.

Manufacturing finishing procedure (final; machining, Overhaul procedure

heat treatment and dimension checks)
Degrease Degrease and paint strip
Magnetic particle inspect Strip Ti cadmium (chemical process commonly referred to as TiCad)
Steel shot peen as required Visual and magnetic particle inspect (min size of detection for the magnetic particle
method is about a 1 mm surface length crack although this will depend on a number of
factors including the surface finish and access)
Chemically clean Repair any deficiencies found that are repairable
Plate with TiCad Chemically clean
Bake Plate with TiCad
Paint Bake
Repaint

View looking aft View looking forward

Region cut from trunnion

150mm Sections used for analysis and examination of

the microstructure

Fig. 2. The two parts of the trunnion showing each face of the fracture surface. The left view shows the aft section and includes the bearing housing into which the forward ball
end is fitted and the lugs to which the drag brace is attached. The right view shows the forward section that attaches to the main landing gear strut. The circles indicate the region
of the fracture that was old and corroded. The region cut from the trunnion for further investigation and the areas used for microsections and analysis are also marked.

2.1. Examination methods

On receipt at the DSTO and a preliminary examination, the aft section of the trunnion (R.H. part in Fig. 2) was selected for detailed
investigation. The primary cracking region was in the entrance to the re-entrant radius adjacent to the thinner sec-tion of the trunnion wall, and
acetate replicas were used to carry out initial cleaning of the fracture surface. This allowed the removed deposits to be retained and analysed
later. The primary cracking region was then excised by water jet cutting. Three through-thickness samples were taken remote from the fracture
surface. These samples were for chemical and metallo-graphic analysis.

The paint was chemically stripped from the excised region, enabling the presence of a TiCad plating to be confirmed. Then the fracture and
adjacent surfaces were examined optically up to 1000 magnification. The plating was then stripped in an ammonium nitrate solution, and the
fracture was cleaned of the remaining corrosion products in an ammonium citrate solu-tion. The origin area of the fracture was subjected to
detailed examination by optical and electron microscopy. Finally, this half of the fracture was longitudinally sectioned through the origin of one
of the main cracks, and at several other locations close by, to look for any inconsistencies that might have contributed to the failure.
390 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396

Metrological measurements were carried out on parts of the trunnion both before and after paint stripping.
The re-entrant radius was Non Destructively Inspected (NDI) with magnetic rubber (a magnetic particle technique) to check for any
secondary cracking unassociated with the primary cracking region.

2.2. Material properties

One sample from the excised primary cracking region was stripped of its surface coating and analysed by an arc spec-trometer. The analysis
indicated that the material met the required chemical specification for 4340M alloy steel per BMS7-26. Vickers Hardness (HV) measurements
using a 30 kg load were taken near the failure origin. The hardness values were within the specified range (5156Rc). Energy dispersive X-ray
analysis of the deposits removed from the primary re-gion of cracking showed the presence only of elements from the steel and TiCad plating.

2.3. Microstructure

The remaining two samples from the excised primary cracking region were prepared for metallographic examination by polishing and
etching longitudinal cross-sections through the wall thickness. The microstructure was consistent with a fine-grained lightly tempered martensite.
No stringers, retained austenite or significant oxide inclusions were found. The surface of the component did not show any significant
decarburization.

2.4. Metrology

Measurements of the inner diameter of the trunnion in the large bore section were taken after paint removal using a point-to-point method.
These measurements gave an average diameter of 136.12 mm for single point-to-points, and 136.06 mm for three-point averaging. Both
diameters were within the specified dimensions.
The wall thickness was measured at around the diameter bore at the run-out of the re-entrant radius at the base of the ramp see Fig. 1.
These measurements revealed that the wall thickness was between 5.23 mm and 4.88 mm except at approximately the centre of the fatigue crack:
this region had a minimum wall thickness of 3.48 mm. This is about 1.09 mm below the minimum allowable wall thickness deduced from the
engineering drawings.

2.5. Surface and fracture examination

2.5.1. Before cleaning

The fracture was found to consist of several stages of fatigue crack growth originated at the re-entrant radius run-out between the thinner wall
section (as marked in Fig. 3) and the ramp in the inner bore of the trunnion this location is marked in Fig. 1 as the inner wall re-entrant radius
at the base of the ramp. The cracking was covered with varying amounts of corrosion product over a distance of about 50 mm from end-to-end,
as shown in Fig. 3. This corrosion was thick-est over the central 25 mm (approx.) of this region.

Fig. 3. Two views of the origin area showing the corrosion product over this region of the fracture, prior to cleaning.
 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396 391

Two major fatigue cracks designated as C1 and C2 , see Fig. 3, initiated from regions about 11 mm apart and then joined to form a single
crack at about the time that C1 neared the outer surface of the trunnion. After breaking through to the outer surface, cracking then grew around
the trunnion wall as a through crack until final fast fracture and failure of the trunnion. Most of the latter stages of fatigue cracking were
interspersed with bands of fast fracture, such that the total extent of stable fatigue crack growth was only about 50 mm, see the solid lines drawn
on the lower fractograph in Fig. 3. The fast fatigue (stable fatigue with fast fracture bands) extended about 20 mm further: one approximate end
of the fast fatigue region is marked as a dashed line on the lower fractograph in Fig. 3.

The inner surface of the trunnion adjacent to the major fatigue cracks revealed that C1 lay along machining grooves in the radius, while C2
appeared to be at an angle to these grooves. Also noted were cracks in the paint that appeared to indicate independent secondary cracks near the
C1 crack. These features are shown in Fig. 4.

2.5.2. After cleaning

After removal of the paint, TiCad and rust, it was found that C1 had initiated from multiple closely spaced origins at the root of a machining
groove and over a distance of several millimetres. C2 had also developed from several initial fatigue cracks. These were close to each other over
a short distance within the root of a machining groove. The initial fracture sur-faces of C1 and C2 were both parallel to the machining grooves,
despite the fact that the C2 fracture plane soon tilted about 12L away from the plane of the machining grooves, as mentioned in Section 2.5.1.

Fig. 5 gives an overview of the cleaned fatigue fracture surfaces of C1 and C2. The dashed inner line on C2 marks the boundary between a
heavily corroded fatigue region and a much less corroded region. The dashed outer lines on C1 and C2 mark the boundaries where less corroded
fatigue regions with clear progression markings were succeeded by rougher fracture surfaces with poorly defined progression markings. Fig. 6
illustrates these features in more detail for C2. The rougher fatigue regions and poorly defined progression markings may have resulted from
post-fracture corrosion or possibly corro-sion-assisted cracking.

2.5.3. Examination of the inner and outer surfaces of the trunnion

After paint removal the inner surface of the trunnion was found to have well-defined machining grooves that extended up to the radius in the
larger diameter section. The fatigue cracks had initiated from the roots of these grooves. Less well-defined grooves were present on the smaller
diameter surface, the bore taper between the two inner diameters, and the outer surface.

Further examination of the longitudinal cross-sections taken through the wall, including the origin region of C1 and the nearby region of the
radius, revealed that the outer surface of the trunnion was reasonably flat. There was evidence of cold work at the surface, in the form of a thin
layer of noticeably deformed material. This is consistent with shot peening as specified on the manufacturers drawings. A small region of this
surface adjacent to the failure is shown in Fig. 7. While this surface was slightly

Fig. 4. The first view shows the inner surface of the trunnion and the profile of the two main fatigue cracks. The second view shows cracks in the paint (arrows) that appear to
indicate further cracking. Note that the C1 crack and the paint indications are aligned with machining grooves (direction indicated by white dashed line) while C2 is at an angle to
them.
392 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396

C2
C1

Fig. 5. View of the fatigue crack region of the fracture, after cleaning.

Outer surface of trunnion

Rougher fatigue region - possibly corrosion assisted cracking

Band of relatively clean fatigue

Heavily corroded
fatigue region

Inner wall of trunnion

Fig. 6. View of C2 showing the features noted in the text. It should be noted that the fracture as shown here had been cleaned of most of the corrosion product so that features such
as some progression markings and the significant boundaries may be more easily seen.

Fig. 7. A micrograph of the outer surface of the trunnion showing a surface without discrete machining grooves. (Etchant: Nital 2%).

Fig. 8. Micrographs of the surface of the radius showing the undulating surface produced by machining, with small intergranular penetrations in the valley of each groove.
(Etchant: Nital 2% for the first view only).

Fig. 9. Micrographs of the surface of the radius showing some other fatigue cracks revealed by the polished sections. All of these were found to be growing from the
valleys of the machining grooves. (Etchant:Nital 2%).
 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396 393

uneven, it was very different from the wavy internal surface at the re-entrant radius, Fig. 8. This wavy profile is due to machining grooves, and
Fig. 8 also shows an example of small intergranular penetrations at the root of a groove.
The longitudinal cross-section through the origin of C1 and the nearby material revealed several secondary and indepen-dent fatigue cracks.
Each of these cracks initiated at the root of a machining groove and was associated with shallow inter-granular penetrations (attack). Examples
of these secondary cracks are shown in Fig. 9. Apart from the shallow intergranular attack there were no other defects associated with any of
these cracks.

2.5.4. Quantitative Fractography (QF)

Several prominent boundaries were noted on the fatigue fracture surfaces as well as crack growth progression markings, particularly for
crack C2. The trunnion history included a number of overhauls during which it would have been exposed to chemicals, including water, used to
remove the paint and plating, see Table 1. Furthermore, these chemicals could have had access to the fracture surfaces of the fatigue cracks when
they were present, resulting in fracture surface corrosion. This pos-sibility or likelihood could explain some of the prominent fracture surface
boundaries.
Assuming that the overhauls contributed to the fracture surface corrosion, then the last overhaul could have been the final contribution to the
heavily corroded fatigue region bounded by the dashed line in Fig. 6. In other words, before the last overhaul the fatigue cracking may have
progressed to the depth indicated by the dashed line in Fig. 6. This interpretation is supported by the condition of the following region, which
was less corroded, and it places the C2 crack size at the last over-haul at about 1.6 mm deep, with a surface length of about 4 mm.

Closer examination of the C2 fracture surface showed corrosion-enhanced crack front markings in the heavily corroded region. These
markings are indicated in Fig. 10, where it is suggested that they relate to chemical attack during previous overhauls. The set of (assumed)
overhaul-related progression markings could be used to construct a crack growth curve pro-vided that estimates could be made of (a) the size of
the discontinuity, most probably intergranular penetration (attack), that initiated the fatigue cracking, (b) the time at which cracking began, and
(c) the crack size at failure:

(a) The intergranular attack at the roots of machining grooves would have acted like (very small) cracks, and the machin-ing grooves would
also act as stress concentrations. This makes it reasonable to assume that the initial discontinuity size was approximately the depth of the
machining groove at the main origin of the C2 fatigue crack.
(b) Since the machining grooves and intergranular attack would have been present from manufacturing, and also the trunnion wall thickness
at the re-entrant radius was well below the specified minimum, it seems likely that fatigue cracking began shortly after the trunnion
entered service.
(c) Failure probably occurred soon after the fatigue cracking penetrated the wall of the trunnion. The wall thickness is therefore a first-order
estimate of the final crack size.

Fig. 11 presents a crack growth curve constructed on the basis of the foregoing assumptions. The curve is shown as the log of the crack depth
versus the assumed life of the trunnion. In the light of much experience with the growth of small cracks in high strength aerospace alloys [2], it is
considered that the crack growth curve is very reasonable.

2.6. NDI

The remainder of the re-entrant radius was inspected with magnetic rubber. (a magnetic particle inspection technique). No significant
cracks other than those associated with the primary cracking region and shown in Fig. 4 were found, although many very small cracks, such as
those shown in Fig. 9 may have existed near the main cracking.

Direction of crack Overhaul 3

measurement
Overhaul 2

Overhaul 1
Origin used in crack growth
Fig. 10. A close-up view of the C2 surface showing marks thought to be an indication of the crack size at the indicated overhauls.
394 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396

10

Failure 25090 cycles

Overhaul 4, 22604 cycles

Overhaul 3, 16506 cycles

Overhaul 2, 14069 cycles

0.1

Overhaul 1, 3517 cycles

0.01
0 5000 10000 15000 20000 25000 30000
Total cycles

Fig. 11. A crack growth (depth) curve for the C2 crack produced by making the assumptions noted in the text. The surface lengths at each of these points would have been
considerably longer.

3. Discussion

3.1. Discussion and summary of the trunnion investigation

The trunnion was manufactured from 4340M ultra-high strength steel as specified. The hardness was within the specified levels and the
microstructure was that of a good quality aircraft grade steel having very few and only very small inclusions and a consistent fine grained
tempered martensite matrix. The surface did not show any significant evidence of decarburization.

The final fracture initiated from a large fatigue crack that had penetrated the wall of the trunnion prior to failure. The fatigue crack appeared
to have been present for a considerable period of the life of the trunnion. This crack initiated in the root of the re-entrant radius between the thin
wall section and the ramp up to the thick wall section of the trunnion, and at a particularly thin section of the wall. This section was only about
3.48 mm thick, much less than the specification minimum of 4.57 mm. This thin-wall discrepancy would have considerably increased the
operating stresses in this area.
Each of the main fatigue cracks had initiated from many origins at the root of a machining groove to form two main fa-tigue cracks (C1 and
C2), which later joined. There were several secondary fatigue cracks unassociated with the main cracks. These cracks all began from deep
machining grooves with small intergranular penetrations. This intergranular attack prob-ably occurred during chemical cleaning before applying
the original TiCad plating during manufacture of the trunnion. This attack should not have occurred, but the re-entrant radius had not undergone
a specified shot-peening treatment prior to cleaning and plating: shot-peening would have introduced compressive stresses at the radius surface
and hence prevented the intergranular attack. Since the machining grooves and intergranular attack would have been present from the time the
trunnion was manufactured, as also the thin-wall discrepancy at the re-entrant radius, the fatigue cracking probably started when the trunnion
entered service, or soon after.

The fracture surfaces of both main fatigue cracks C1 and C2 showed several corrosion-enhanced progression markings. These were assumed
to indicate the crack depths at individual overhauls. QF based on this assumption and additional assumptions about the initial discontinuity size,
final crack size, and the beginning of fatigue cracking soon after the trunnion entered service resulted in a credible crack growth curve for C2,
shown in Fig. 11. From this curve it is seen that at the time of the last overhaul the C2 crack depth should have been more than 1.5 mm. This is
normally a detectable size for magnetic particle NDI, which was the expected method used during overhaul, see Table 2. However, it is possible
that the presence of deep machining grooves, and the locations of the fatigue cracks at the root of some of these grooves, could have made crack
detection more difficult: such grooves can create false indications that lead an inspector to reject valid indications.

The results may be summarised as follows. Failure of the trunnion was the result of fatigue cracking that may be attrib-uted to four main
factors:

(1) Fatigue cracking began from intergranular attack at the roots of deep machining grooves on an internal radius of the trunnion cylinder.
 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396 395

(2) During finishing, the inner surface of the trunnion cylinder did not undergo the specified shot-peening. This omission allowed the
machining groove profiles to remain and also allowed chemical attack at the roots of these grooves.
(3) The wall thickness of the trunnion at the fatigue cracking location was well below the minimum specification. This means that besides
stress concentrations due to the machining grooves and intergranular attack the local stresses would have been higher than expected. This
suggests that cracking began soon after the trunnion entered service and eventually led to premature failure.

(4) The NDI during overhaul failed to detect the fatigue cracking prior to failure, even though at least one and most prob-ably both of the
main fatigue cracks was of detectable size at the last overhaul.

3.2. Implications from the failure

3.2.1. Combinations of problems

The failure was unusual in that premature fatigue cracking was facilitated by at least two (apparently) unassociated prob-lems, the omission
of shot-peening and a thin-wall discrepancy that both came from the trunnions manufacture. If the wall thickness had been within the
specifications, then omission of shot-peening may still have resulted in fatigue cracking, but it would have taken longer to reach a critical size.
This would have allowed more opportunities for NDI during overhaul to de-tect cracking, although detection would still have been challenging
owing to the crack locations at the roots of machining grooves.

Despite the unusual combination of initial circumstances, i.e. the omission of shot-peening and a thin wall discrepancy that acted
synergistically to cause premature failure of the trunnion, it is possible that failures could occur in other trun-nions. In the authors experience
many unexpected fatigue failures have been due to mis-applied or missing fatigue life extension treatments combined with poorly specified or
challenging NDI procedures. Examples are: failure to shot-peen as required [3], excessive shot-peening, incorrectly cold expanded holes [4],
poorly fitted interference bushes [5], and poorly fitted reinforcement straps [6]. All of these deficiencies were associated with highly stressed
airframe or engine locations, and they were often associated with manufacturing problems.

3.2.2. Surface treatments for fatigue improvement

The use of surface treatments for fatigue improvement needs to be carefully controlled. The inherent difficulty in determining if such
treatments have been applied and whether the applications were effective, particularly for an NDI inspector checking many parts at overhaul, has
not been given as much attention as might be expected. Non-destructive instruments are only now becoming available to measure surface
residual stresses in difficult areas of large components [7]. Such measurements may be included in future overhaul schedules, but until then extra
care needs to be taken in selecting and using surface treatments for fatigue improvement and choosing the appropriate NDI techniques.

3.2.3. Retirement-for-cause
The trunnion was probably designed to have infinite life in the region that failed. The stresses would have been designed to be low, probably
being selected to be below the fatigue limit of a StressLife master curve for this configuration (the re-entrant radius). There would also be the
added benefit of shot-peening.
For service use a safe-life approach may originally have been used, whereby the region in question survived a long time without failure
during a test, and the test life was used to specify a finite service safe-life even though the part had not failed. However, this traditional safe-life
approach is an expensive option when fatigue calculations suggest unlimited life. A more attractive option is continued service with periodic
inspections. This is commonly called retirement-for-cause or life-on-condition.

To allow continued service and prevent failure from unforeseen damage, the manufacturer instituted periodic inspec-tions, probably at a
conservative interval, in the present case a maximum allowable interval of 12,000 landing cycles or 8 years of service. These inspection intervals
may have been determined from an overall safe-life estimate, or else chosen to avoid typical service damage such as corrosion and wear. In other
words, the manufacturer did not necessarily have spe-cific trunnion locations to be targeted for fatigue-related NDI. This possibility (or
probability) is a severe weakness in the retirement-for-cause approach.

4. Conclusions

Investigation of the failure of a Boeing 747 wing main landing gear trunnion showed that:

(1) The trunnion failed by fatigue cracking.

(2) Fatigue cracks initiated at the roots of machining grooves in a radius between the inner larger bore of the trunnion and a taper, which ran
up to a narrower bore.
(3) The radius and the larger bore of the trunnion showed no evidence of having been shot-peened, a requirement indi-cated on the drawings.
396 S.A. Barter / Engineering Failure Analysis 35 (2013) 387396

(4) The wall thickness in the region of crack initiation was significantly less than the minimum specified for this location.

(5) The lack of shot-peening allowed the machining groove profiles to be retained and also meant that residual compres-sive stresses were
not introduced on the groove surfaces. The lack of compressive residual stresses allowed chemical attack during manufacture to cause
small intergranular penetrations at the roots of the machining grooves. The inter-granular attack most probably occurred during a wet
cleaning process prior to TiCad plating the trunnion.
(6) The combination of stress concentrations due to the machining grooves and the intergranular attack, and the local increase in service
stresses owing to a sub-size thin-walled section, resulted in early fatigue cracking in the radius, most probably soon after the trunnion
entered service.
(7) Cracking was probably present at each overhaul. A crack growth curve obtained for one of the main fatigue cracks indi-cated that it
would have been detectable admittedly under favourable conditions by magnetic particle NDI during the last overhaul before failure.

(8) Failure to detect the cracking may be attributed partly to a masking effect of false indications from the deep machining grooves and the
chemical attack in many of them; the limited access inside the trunnion; and possibly a lack of expec-tation of cracking in this location.

The failure has also shown that the use of surface treatments for fatigue improvement needs to be very carefully con-trolled, and the
implications of faults and omissions in the treatment processes need to be considered. In the present case the failure to shot-peen the trunnion
bore should have been detected (as should the inadequate wall thickness) during man-ufacture. The inherent difficulty in determining if such
surface treatments have been applied, and whether the application was effective, particularly for an NDI inspector checking many parts during
overhaul, has not been given as much attention as might be expected.

Finally, this investigation has served to highlight weaknesses in the retirement-for-cause philosophy. Combinations of manufacturing errors,
coupled with difficulties in in-service inspections, can act to make failures harder to anticipate. This should be taken into account when critical
(single load path) structures are assessed for their fatigue life. If retirement-for-cause is to be used to maintain structural integrity, then the
inspection methods need to be carefully considered and vali-dated on the target components, especially taking account of its effectiveness under
adverse conditions.

References

[1] Anon. Fractured trunnion, Boeing B747-300 aircraft, JA 8184 at Sydney Aero., Aviation Safety Investigation Report. Australian Transport Safety Board, Occurrence No.
200502400; June 2006.
[2] Molent L, Barter SA, Wanhill RJH. The lead rack fatigue lifing framework. Int J Fatigue 2011;33:32331.
[3] Dixon B, Molent L, Barter SA, Mau V. Flaw identification through the application of loads: teardown of the F/A-18 centre barrel from USN F/A D buno No. 163507, DSTO-
TR-1740, Defence Science and Technology Organisation, Air Vehicles Division, Melbourne, Australia; July 2004.
[4] Barter S, Dixon B. Investigation using quantitative fractography of an unexpected failure in an F/A-18 centre fuselage bulkhead in the FINAL teardown program. Eng Fail
Anal 2009;16:83348.
[5] Lewis FG, Glanvill DW, Bishop BC. Accident to Viscount aircraft VH-RMQ metallurgical examination of components from the wing structure, ARL Applied Report 66,
Aeronautical Research Laboratory, Melbourne, Australia; August 1969.
[6] Swift SJ. The Aero Commander Chronicle, ICAF 95. In: Grandage JM, Jost GS, editors. Estimation, enhancement and control of aircraft fatigue performance, vol. 1.
Proceedings of the 18th symposium on the international committee of aeronautical fatigue, 35 May, 1995, Melbourne, Australia, EMAS, Warley, West Midlands, UK; 1995.
p. 50730.
[7] Withers PJ, Bhadeshia HKDH. Residual stress Part 1 measurement techniques. Mater Sci Technol 2001;17:35565.