Sequence of events
On 17 November 2003, at 1630 UTC, as Boeing Company 747-438
(747), registered VH-OJI, rotated during takeoff from Changi
International Airport, Singapore, the crew heard a loud bang and
the aircraft yawed left. The aircraft's instrument indications were
consistent with a failure of the number-2 (left inboard) engine.
The crew shut down the engine and, after dumping fuel, returned to
Changi.
An initial inspection found that there had been a failure within
the engine's first-stage high-pressure compressor assembly (HPC 1).
Following removal from the aircraft and return to Australia, the
engine was disassembled at the operator's maintenance facility,
where the failure was confirmed as the loss of one blade and the
extensive thermal and mechanical damage of the remaining blades
from the HPC 1 disc. The liberated blade stub was recovered from
the engine confines.
Engine
The engine was a Rolls-Royce RB211-524G2-T-19 model, serial
number 13211 (see figures 1 and 2). The designation 'T' in the
engine model number indicated that the original engine had been
upgraded with power-train components from the larger Rolls Royce
Trent 700 series engine. That enabled greater fuel efficiency and
lower exhaust gas temperatures. At the time of the failure, the
engine had operated for a total of 50,847 hours and through 6,659
cycles. Module 41, the engine's high-pressure compressor and
turbine system, was fitted at the last engine refurbishment on 18
August 2000 and had operated for 14,166 hours and through 1,456
cycles since installation.
Components received
The HPC 1 blades (including the liberated blade) were removed
from the engine by the operator and forwarded to the ATSB for
examination. For the purposes of the investigation, the blades were
identified by numbering in a counter-clockwise direction (looking
rearward), commencing from the blade adjacent to the position of
the liberated blade. The blade numbered 21 was not received. All
examinable blades carried the part number FK28595 H124, embossed on
the underside of the root section.
Blade condition - retained blades
Preliminary examination of the HPC 1 blades that remained on the
compressor disc during the failure showed that all sustained
extensive damage to the tips and edges such that the blade profiles
resembled that of an arrowhead. The melting and loss of the leading
and trailing edge corners of the blades (figures 3 and 4) is
characteristic of a titanium fire within the assembly. Evidence of
metallisation was noted on the blade aerofoil and platform
surfaces. A close visual inspection of the blade root section
revealed no evidence of cracking around the dovetail corners, nor
was there any evidence of isolated or general mechanical damage
such as may be sustained during handling or installation.
A study of the condition of the dovetail bedding surfaces found
that all exhibited varying levels of surface galling damage across
the full width (figure 5), with the damage appearing most
pronounced toward the rear (outflow) end of the blade root. Low
power stereomicroscopy study of the galling showed a predominantly
axial orientation to the damage (transverse to the axes of the
dovetail faces).
Several areas showed heavy localised galling, producing a
gouging effect with notable metal loss (figure 6). Evidence of
frictional heating with associated surface tinting (blueing)
surrounded the heavily galled area.
Review of the distribution of dovetail galling found that some
blades showed light galling up to and encompassing the transition
radius between the upper dovetail edge and the root body (figure
7).
Failed (liberated) blade
The remnant stub of the single blade liberated from the
compressor disk, despite extensive damage, showed the 'arrowhead'
form produced by the loss of the leading and trailing edge corners
(figure 8). Although the blade root fracture features were damaged
beyond recognition, the similarity in form to previous failures
indicated a high-cycle fatigue cracking mechanism.
The blade root showed the apparent fracture of the complete
length of the dovetail toe along the leading (concave) side of the
blade and roughly one-half of the toe length on the trailing side
(figure 9). Damage and abrasion sustained following the blade loss
prevented any useful examination of the fracture surfaces. However,
the fracture profiles did not exhibit any associated plastic
deformation or bending of the blade root body as could be expected
if an external load had been applied to the blade, through impact
with a foreign object or other internally liberated component.
Cracking and subsequent failure of the blade's dovetail root
corners allowed the blade to move radially outward from its slot,
under the influence of centrifugal operating loads. As the blade
contacted the compressor housing, the resulting friction initiated
the titanium fire that melted the blade corners, before the blade
completely released from the slot.
The aerofoil section of the blade stub showed uniform transverse
bending in a direction opposing the rotation of the compressor disc
(figure 10). The layer of metallisation and melted debris on the
inside of the bend showed a degree of cracking and fissuring that
suggests deposition prior to the bending of the blade.
Service information
The engine manufacturer was aware of seven similar HPC 1 blade
release failures in RB211-524G/H-T series engines worldwide. The
ATSB had previously investigated one of those occurrences (see ATSB
report BO/200205895). The engine manufacturer was also aware of a
failure involving a Trent 700 series engine.
On 6 August 2003, in response to the first five failures, the
engine manufacturer published Service Bulletin RB211-72-E181,
applicable to the occurrence engine, that introduced a revised dry
film lubricant on the stage-1 high pressure compressor blade root.
The reason for issue of the service bulletin was the inadequacy of
the earlier lubricant in preventing incomplete blade bedding and
uneven wear, leading to subsequent high-cycle fatigue cracking and
potential failure of the blade root. Accomplishment of the Service
Bulletin was required to be denoted by a change in the blade part
number, from FK28595 to FW26617. Compliance with the Service
Bulletin was listed as 'Recommended', with suggested accomplishment
when the engine or engine module was disassembled for refurbishment
or overhaul.
The engine manufacturer also indicated that scoring and sharp
edge damage in the blade root area, leading to local stress
concentrations, may have contributed to HPC 1 blade failures. To
prevent that damage, the manufacturer revised the manufacturing
process, changing the blade root machining process from broaching
to milling.
