P • R • O • C • E • S • S • E • S by G.

Goodwin

The Role of Thorough Failure Investigation in Enhancing

the Reliability of Machinery
Introduction
This article examines some aspects
of typical failure investigation pro-
cesses. Some such investigations be-
gin with a client sending one or more
broken parts to a laboratory or to a
chosen investigator, often a metal-
lurgist, seeking an explanation of the
reason for the failure. Generally, this
is a poor way to begin an investiga-
tion. This paper seeks to explore why
this process may give inconclusive or
misleading results, and how the inves-
tigator may best recover from such a
start and reach a sound conclusion.
The article will also examine some
common design and development
processes that lead to a high risk of
failure, including changes initiated by
users and operators, illustrated with
examples from the diesel engine in-
dustry. This should not be taken to
imply that this industry is necessarily
worse in this respect than others; it is
simply the source of the machinery
that is most familiar to the author and
that is the subject of the majority of
his investigative experience.

How Do Design Changes

Occur without Adequate
Analysis?
Evolving Design
The class of medium- and high-
speed marine diesel engines common-
ly used for propulsion and power
generation in Naval vessels typically power output and the addition of with very different auxiliary equip-
has a design life of 30 years or more. many new features. Commonly, at the ment and control systems. It is likely
Designs for marine diesel engines end of the development of an engine there will be no or few components
evolve over the years. At the end of type, the engine bears a close visual common to the original and final
the design life of the engine type, resemblance to the original model but designs, and several decades will have
there may have been significant devel- may be producing several times the passed during the life of the type. The
opment, with very great increases in power per cylinder of the original, structure may have been modified,

Practical Failure
Failure Analysis
Analysis Volume
Volume 3(2) April 2003 21
 Thorough Failure Investigation in Enhancing the Reliability of Machinery (continued)

component dimensions and materials

changed, and auxiliary equipment
and even key internal moving parts
may have been procured from a
variety of suppliers. The original
design calculations for the strength
of components are likely to have
been performed before the days of
computer-aided design and numerical
stress analysis. Indeed, stresses in
some major components may never
have been accurately calculated or
measured, unless a series of failures
has led to a reassessment of the origi- soon as possible. When an engine is When the new engine is installed
nal design. As a result, in a modern introduced into service, the prototype in the vessel and sea trials are carried
high-performance machine, there is unlikely to have served the thous- out, the engine has to cope with a new
may be some components whose ands of hours the user will expect from system of inlet and exhaust trunking,
design is based on the thinking of a his purchase. Experience may have new fuel supply and cooling systems,
previous generation. Wong, in his shown that, at least in the shorter term, and often a lubricating oil system,
recent text on safety and reliability, upgrading an existing machine produces designed by the shipbuilder. Pressures
has stated that “a modification or a a more reliable engine at lower cost than and flow rates should be about right,
change in the use of a system, or designing a new one from scratch, but but may not be exactly as expected,
existing design, can lead to a higher there are pitfalls, especially when sometimes with quite serious results.
risk of failure and a complete re- components are used that were origi- The power required from a ship
assessment must be carried out.”[1] nally designed long ago. propulsion system varies with speed
in accordance with a cube law. To go,
The evolving design syndrome may
Modifications to New Engines for example, 5% faster requires ap-
affect subsystems. For example, fuel
at the Shipyard proximately 15% more power. If the
injection systems have undergone
In the case of a warship, there is a ship is required by the contract to
great changes in recent years, with
tendency to choose engines with the achieve, for example, 20 knots, and
major changes in the equipment and
maximum possible rating for the type. only achieves 19 knots, the marine
substantial increases in pressure, in the
The operating regime may not be well engineer is faced with a significant
quest for improved efficiency and
defined when the vessel is built and problem. The engines have been
emissions performance. Over the
may change several times over the life specified and installed and are now
same time period lubrication systems
of the vessel. In addition, it is in the required to produce significantly more
have seen far less change, but bearings
nature of the application that engines power that, while difficult to achieve,
have to sustain higher firing pressures,
may be driven particularly hard, or, may be easier than reducing the drag
and auxiliary items such as injection
alternatively, may be made to operate of the ship. The engine manufacturer
pumps and turbochargers handle
at very low load for long periods then may be asked to upgrade the power
higher flow rates and may need to
required to produce full power on output and is often able to do so by
discharge substantially more heat into
demand. This can lead to effects on changing turbochargers and fuel in-
circulating fuel, coolant, and lub-
component loads and temperatures jectors. However, this requires greater
ricating oil.
not foreseen by the designer. Engines air flow, which may create greater
Generally, upgrades to design are and their subsystems may thus be pressure losses in ductwork already
tested by the engine builder to ensure stressed beyond design expectations, built into the ship. The engine manu-
that reliability is not compromised, despite the best attempts of the facturer may simply not know of these
but there is always pressure to bring ordering authorities to define the limitations. It is at this stage that some
the upgraded machine into service as requirements in the specifications. engines or their cooling systems have

22 Volume 3(2) April 2003 Practical Failure Analysis

become overloaded, producing symp- supplier to supply an engine even in a applications has led to design changes
toms of distress that can be difficult non-ideal installation. The alternative that enhance reliability, and the bene-
to correct. Most such problems are is to lose the sale to a competitor. fits may be fed back to improve the
apparent during the testing of the reliability of similar machines in in-
vessel, though not all can readily be Modifications to Correct Problems dustrial applications. The cases are
corrected. Sometimes more dangerous in Service given as examples of failures involving
are the smaller problems that cause When the ship has entered service, a complex series of factors and to show
marginal overloading, which may only problems may be experienced with the effects of evolving design in indus-
manifest itself in reliability issues well reliability or with maintenance try, not to highlight the problems
after the vessel is accepted into service. actions. Sometimes difficult mainten- of any specific company or supplier.
Upgrading for Increased ance actions, intended to be con- The way to avoid such failures may
Performance ducted rarely, are found to be needed appear to be that we start every system
A design change may be demanded much more frequently. This creates with a “clean sheet of paper,” or per-
because of a need for increased per- pressure for the staff of the ship to be haps, these days, a blank computer
formance during the life of a vessel. able to carry out repairs at sea, when disk. However, most engineers would
Once again, this may be a more com- the maintenance plan foresaw that somewhat ruefully admit that the
mon problem in military vessels than these jobs would only be done at the most reliable systems are those con-
in commercial vessels. Most warships shipyard or by contractor staff with taining the maximum number of
have mid-life upgrades, and some the ship alongside the wharf. “proven” components! This view is
experience significant changes in role reinforced by Wong, in his book
Some modifications are instigated
during the typical 30-year lifespan of quoted earlier, when he says, “In the
just to make such maintenance
a vessel. Engines are likely to be design of plant and equipment, it is
actions easier for the crew. Such chan-
required to provide either more pro- usual to use proven components of
ges might be designed and carried out
pulsive power or more electrical load, mature design with a constant failure
by a shipyard or a visiting mainten-
and sometimes both. If the engine can rate.”[2] This approach is valid pro-
ance contractor but without sufficient
be upgraded using the existing base vided the operating conditions for
time or preparation to allow for con-
type, the more powerful variant will the component are also proven and
sultation with the engine supplier, or
need more air, more cooling, and so unchanged.
without a need being perceived for
forth, and existing systems may not such consultations. A need to reassess Pulsation in Diesel Engine
cope as well as expected, compro- some design parameters may be ap- Fuel Systems
mising engine reliability. parent with the benefit of hindsight,
In the first three examples described
In cases where the change is great, but this is not always appreciated in
subsequently, the loads responsible
a complete re-engine, using a new advance.
for failure were created by dynamic
manufacturer, is not unusual. Once pressure in the fuel system. Diesel
again, some of the problems noted Some Failure Cases Related to engines require a precise amount of
previously occur. The air and exhaust Development and Modification
fuel to be injected over a short period
ducting may be impossible to change This article describes examples of time, using extremely high fluid
dramatically, and fuel and cooling sys- from the investigation of failures attri- pressures. Pounder’s Marine Diesel
tems have to cope without the benefit butable to some of the issues and Engines,[3] current edition by Wood-
of designing the hull around the actions described in the preceding
yard, gives a reasonably detailed
systems, which is to some extent poss- paragraphs. In principle, similar issues
description of the operation of fuel
ible in a new vessel. Compromises, in can arise in many industries, because
injection systems.
which the requirements of the new evolving design is the most practicable
engine cannot quite be met, are likely. way for most machinery manufac- Injection pressures have risen in the
Whether this is a re-engine or a per- turers to keep up with technology ad- last few decades from approximately
formance upgrade of an existing type, vances at acceptable cost. The analysis 100 MPa (1000 bar) to close to 200
there is still pressure on the engine of failures in machinery in military MPa (2000 bar). This was necessary

Practical Failure Analysis Volume 3(2) April 2003 23

 Thorough Failure Investigation in Enhancing the Reliability of Machinery (continued)

to enable the injection rate to be

controlled precisely, with a fine atom-
ized spray being injected, even with
heavy fuel oil. Such pressures are of a
similar order of magnitude to the yield
stress of some steels, and the design
of injectors, injection pipes, and
pumps is an ongoing effort by the
suppliers of fuel injection equipment.
Larger marine engines have individual
injection pumps for each cylinder,
with short pipes leading to the fuel
injectors. These pipes are typically
constructed with double walls and
leak detection, and, increasingly, they
are integrated into major components
such as cylinder heads. It is clear that
these pumps and pipes have been the
focus of much careful design effort.
However, the fuel is supplied to the
injection pumps by a relatively low-
pressure system that is typically
treated with much less respect. In the
following case studies, failures occur-
red in these low-pressure systems.
In the type of system described here,
parts of the low-pressure system
supplying the injection pumps are
normally left to the shipbuilder.
Typically, the on-engine supply and
return pipes (referred to as fuel rails)
are regarded as engine components
and connected to “shipbuilder supply”
pipes at the end of the engine. A fuel
boost pump may be mounted some
distance away, supplying fuel at
pressures in the range of 400 to 1000
kPa (4 to 10 bar). Recent experience
across both merchant and Naval fleets
indicates that these “low-pressure pipes comprising the rest of the fuel The existence of pulsating flow in
systems” are responsible for a high system of the ship may be less well fuel systems is well understood in the
proportion of engine room fires.[4] designed. The three cases described engine industry, to the extent that
The demand for fuel in short bursts subsequently are all different types of some manufacturers have proprietary
by injection pumps leads to signifi- failure resulting essentially from devices installed in their systems to
cant pulsation in the low-pressure insufficient attention to the design or mitigate its effects. Others simply rely
system. While it is common to use maintenance of the low-pressure on strong, well-restrained pipework.
strong pipework on the engine, the system. Complete removal of pulsation is

24 Volume 3(2) April 2003 Practical Failure Analysis

extremely difficult, due to the nature The effect of the release of injection the fuel rails, giving rise to pulsating
of the demand of the engine for fuel. pressure via the spill port, known as pressure and vibrations induced by
In the course of a Board of Inquiry[5] the spill pulse, is understood to be akin dynamic mass flow.
conducted by the Royal Australian to a shock wave that propagates
It is worth noting that the pressure
Navy (RAN) in 1998 as a result of the through the system. After spill, the
in a “low-pressure” fuel system is
accident involving the HMAS plunger continues to the top of its
similar to that found in workshop
Westralia (described later in this stroke, expelling further fuel into the
compressed-air lines. Engineers and
article), the phenomenon was the sub- low-pressure fuel rails. It then de-
technicians treat these with consid-
ject of much discussion, and a search scends, refilling the pump from the
erable care and respect; if they were
of the literature was undertaken. A same fuel rails. This action causes
filled with fuel, even more caution
research project conducted on behalf rapid fluctuations of flow velocity in
might be expected. Yet the descrip-
of the Maritime Safety Authority in
tion “low-pressure” seems to lead to
the United Kingdom[6] and published
fuel supply systems of the ship being
earlier that year cited little published
treated in a very routine fashion.
data. The report included the results
of practical tests showing that, in a
typical engine, the injection pumps A Failure Driven by
emitted pulses into the low-pressure Pulse Pressures
system with peak pressures an order
Figure 2 shows HMAS Westralia, a
of magnitude greater than the rated
replenishment tanker of 40,000 ton
pressure of the system. In a later,
gross in service with the RAN. She
unpublished study for the RAN fol-
was originally ordered in the United
lowing some injection pump fail-
Kingdom in the mid-1970s as one of
ures,[7] Cannell cited a 1990 paper by
four commercial product tankers. The
Gatjens and Rulfs[8] that described
original purchaser met financial
the pulsation mechanism, explained return difficulties, and the four ships were
subsequently. A schematic cross
bought during construction by the
section of an injection pump is shown Fig. 1 Schematic cross section of the pumping British Royal Fleet Auxiliary and
in Fig. 1. element of a fuel injection pump converted to allow underway refuel-
The low-pressure system ensures ing of Royal Navy vessels. By the late
that the space above the plunger is 1980s, the ship was surplus to Royal
completely filled with fuel. When fuel Navy requirements and was leased by
is required at the injector, the plunger the RAN, then purchased outright.
is driven upwards by the camshaft. D uring the early years of the
Initially fuel is expelled from both Westralia’s RAN service, one of the
ports, which are usually connected by problems experienced in maintenance
a gallery within the pump housing, was a propensity for joints to leak in
until the ports are closed by the the on-engine pipework of the low-
plunger. Fuel trapped above the pressure fuel system. Fixing these
plunger is then compressed to the leaks was difficult because of the way
injection pressure and delivered to the the rigid pipework was assembled. In
injector. At the end of injection the 1998, in order to simplify repairs, rigid
spill port is uncovered and the steel pipes connecting the injection
injection pressure is released into the pumps to the fuel supply and return
low-pressure system. The delivery rails were replaced with flexible hoses.
valve closes to prevent reverse flow After only approximately 50 h of
from the injector line and injector. Fig. 2 HMAS Westralia operation, while the laden tanker was

Practical Failure Analysis Volume 3(2) April 2003 25

 Thorough Failure Investigation in Enhancing the Reliability of Machinery (continued)

sailing up the shipping channel off the When the failed hoses were exam- report,[6] which was published just a
port of Fremantle, one of the hoses ined after the event, there was clear few weeks before the accident to
failed, causing a substantial leak of evidence of high-cycle fatigue failure. HMAS Westralia, that pulses with
fuel oil. The engine was shut down, The two burst hoses had failed in pressures at least an order of mag-
but a few minutes later a second hose fatigue, and further inspection reveal- nitude greater than the static pressure
failed, this time on the other engine. ed broken wires in the steel braiding were emitted by injection pumps,
Before the second engine was shut of several of the hoses. Some of the and that the phenomenon was not
down, leaving the ship without unfailed hoses from the engine were widely understood when this engine
propulsion, a fierce fire broke out. also tested to destruction and failed modification was made in 1998.
Tragically, four young sailors perished at a wide range of pressures. One with
in the engine room fire, and the vessel over 50 broken wires failed at 6000
came close to running aground before kPa. Figure 4 is a close-up of the
a tow was able to be secured. braiding showing the way the wires
were laid up. (Crossed wires such as
The accident, investigated by a
those seen in the center of the image
Board of Inquiry, was found to be a
were found in a number of places and
direct result of the modifications to
were judged not to be a significant
the low-pressure fuel system using
problem.) The individual wires are 0.3
flexible hose. This was intended to
mm in diameter. Figure 5 shows a
simplify a problematical maintenance
micrograph of the end of a wire from
activity, and, on the face of it, appeared
one of the failed hoses. Close inspec- Fig. 3 The failed hose found to be responsible
logical. A photograph of the failed for the engine room fire on HMAS
tion of the fracture surface revealed
hose appears as Fig. 3, which is re- Westralia in 1998
the characteristic progression marks
produced from the report of the Board
of a high-cycle fatigue failure, and the
of Inquiry.[5]
lack of necking also pointed to a fati-
The initial investigation centered gue failure. By contrast, when a new
on a mechanical and metallurgical hose was tested to destruction, the
study of a burst hose that had been broken wires were examined micro-
found to leak fuel onto a hot spot, scopically and found to be severely
leading to a serious fire. The main necked before failure. The report of
reinforcing material of the hose, a this work[9] was tendered in evidence
hard drawn type 304 stainless steel to the Board of Inquiry.
wire, was found to conform to the Fig. 4 Close-up of braiding of reinforcing
specified standard, as given in BS It was concluded by the Board of
wire around the type of hose shown in
1554:1990, both for chemical analysis Inquiry that the system was subject Fig. 3
and hardness. The hose supplied had to pulsations that had not been under-
a specified minimum burst pressure stood when the engine modification
of 33,000 kPa. Some spare unused was made. Calculation showed that it
hoses were tested to destruction and would take approximately 66 h for
failed at pressures in excess of 40,000 each injection pump to produce one
kPa, significantly higher than their million spill pulses. The first failure
rated strength. The static pressure in occurred after approximately 50 h.
the system was approximately 600 Assuming each hose was affected only
kPa, so it was clear that the hoses were by the pump to which it was attached,
rated to over 50 times the static pres- this places the number of stress cycles
sure in the system and were actually in the hose at the lower end of the
Fig. 5 Micrograph of a broken wire from a
stronger than their rated burst pres- regime known as high-cycle fatigue. It braided hose of the type shown in Fig.
sure. So why did they fail? was clear from the MSA research 3, with progression marks

26 Volume 3(2) April 2003 Practical Failure Analysis

Vibration of a Fuel inspection and scheduled
replacement. The engine
Hose Containing manufacturer recommended
Pulsed Flow replacement every two years
Later in 1998 an engine for flexible hoses connected to
room fire disabled another their engines. It was under-
Australian vessel, this time the stood that this hose was a
icebreaking Antarctic supply modification, and therefore
vessel RV Aurora Australis (Fig. not mentioned in the original
6). The author was asked by documentation of the ship-
the Marine Incident Inves- Fig. 6 Aurora Australis in Hobart, Tasmania builder. Because it was also in
tigations Unit (MIIU), now a hidden location, it was
part of the Australian Transport always likely to be overlooked.
Safety Bureau, to assist with the forcement was completely embedded It was concluded that this failure
investigation. in the layers of synthetic rubber that was mainly a result of vibration rather
In this case, once again, a failed formed the bulk of the hose. No fault than pulsation, though it is probable
flexible hose was found to be the was found in the materials, but the that pulsation contributed, because
immediate cause of the accident.[10] hose had aged to a point where the the main forcing mechanism driving
Figure 7 shows the hose as it was rubber had begun to break down, and vibration was pulsating flow in the
found when the deck plates covering the wires had been chafing against system. The real reason for the failure
it were removed. In the dirt and debris each other. Hoses of this type in any was that the particular hose, part of
the hose is difficult to see, but it runs safety-critical location require regular the “shipbuilder pipework,” had been
from left to right across the middle in a hidden location and had not been
of the frame. The leak came from a replaced when other hoses in the
split approximately 20 mm long, system had become due for scheduled
indicated at the center of the replacement. In this case, the hose was
photograph. of adequate strength, but inadequate
For the event to occur as it did, the records and maintenance documen-
wire reinforcement must have failed tation led to the hose being used
over most of the length of the leak beyond its service life.
before the layers of the rubber cover-
ing parted. A Failure of Steel Pipework
Because this event occurred only a
Due to Pulsation Induced
few months after the accident to Fig. 7 The split hose that led to the July 1998 Vibration
engine room fire aboard RV Aurora Since the above incident, another
HMAS Westralia, the investigators
Australis
looked for similarities in the two case involving fatigue failure of low-
failures. However, microscopic exami- pressure shipbuilder-fitted fuel pipes
nation of the hose structure found has come to the attention of the
quite a different failure mechanism. author. This time the failure was in
While the final failure was in fatigue, the joint arrangements for rigid steel
some of the wires providing the main pipes. Once again, the mechanism was
strength of the hose had mechanically vibration driven by pulsating flow in
worn more than halfway through the fuel system, originating in the
before a fatigue crack was initiated, injection pumps. However, in this
as can be seen in Fig. 8.[11] The wire case, the failure occurred repeatedly
was laid up in a braided form similar Fig. 8 Micrograph of one failed wire from the at a joint where a fillet weld around
to that in Fig. 4, and the wire rein- hose depicted in Fig. 7 the pipe provided a ready-made crack

Practical Failure Analysis Volume 3(2) April 2003 27

 Thorough Failure Investigation in Enhancing the Reliability of Machinery (continued)

initiation site. A diagram of the design was modified to mount them ces due to the flow and the motion of
arrangement is shown in Fig. 9. on the bedplate. A downpipe ap- the pipe. This was also exciting other
proximately 1.5 m long was inter- vibration modes in the pipes. Such
Pulsation dampers (lower left of
posed between the rails and dampers. vibration imposed an unexpected
Fig. 9) were fitted to each fuel rail to
An unforeseen result was that the strength requirement for the joint,
mitigate the pulsating flow that was
oscillating mass flow now had a verti- which simply could not be met. While
causing vibration elsewhere in the
cal component in the downpipes. it has been possible to stiffen the
system. The mass of fluid in the off-
These were relatively poorly re- pipework to reduce the failure rate, it
engine pipe system was such that
strained in the vertical direction, most is clear that mitigation of the
considerable pressure pulses were re-
of the restraint being supplied by the pulsating flow closer to the source
corded. The original plan was to fit
flanged coupling at the damper. Fig- would provide a more satisfactory
the pulsation dampers directly to the
ure 10 shows the welded coupling solution. The solution proposed by
end of the fuel rails, to isolate pul-
used to join the pipe to the damper. DSTO is still in the course of devel-
sating flow along the rails from the
The design was obviously never in- opment, but the addition of smaller
off-engine pipework.
tended to cope with large bending accumulators directly to the fuel rails
The pulsation damper is essentially loads and failed adjacent to the weld. is showing promise.
a hydraulic accumulator, with its Figure 11 shows a cross section of the In all the cases cited, the failure
connecting manifold configured to failed joint, with a clean straight crack mechanism was clearly metal fatigue.
prevent pulses bypassing the accumu- through the pipe outside the weld. However, the diagnosis of the cause
lator. Adding this device would trans- The initiation site in this case appears of failure led to the identification of a
mit much smaller pressure fluc- to have been in the heat-affected zone design issue in the injection pumps.
tuations into the rest of the system but at the toe of the weld, though the exact Fatigue failures, especially in fluid
give rise to increased oscillating mass location is of little importance in this systems, are sometimes initially iden-
flow along the rails, which were case. The quality of the materials and tified as the result of material inade-
well restrained in the longitudinal work was not in doubt here. Any stress quacies, while a wider study of these
direction. concentration at the toe was small. systems revealed that a design change
When the installation was made, it The problem arose because of the in the fluid process could substantially
was found that the specified pulsation rearrangement of the design of the reduce cyclic loads.
dampers were too big and heavy to fit pipework during the refit.
directly to the end of the rails, so the A Diesel Engine Auxiliary
Measurement of the vibration and
Gearbox Failure
of pressure at the flanged connection
Marine diesel engines have numer-
to the damper[12] showed a close cor-
ous auxiliary machines and devices,
respondence between the inertial for-

Fig. 9 Arrangement of pipework between the

on-engine fuel rails and a pulsation Fig. 11 Cross section through the failed pipe at
damper Fig. 10 The welded coupling that failed the location indicated in Fig. 10

28 Volume 3(2) April 2003 Practical Failure Analysis

and it is common to drive these from case-hardened steel with considerable under starting torque. The maximum
the same gear train as the camshafts. QA requirements. Figure 15 shows stress was at the hub fillet radius,
Items driven in this way can include the calculated stresses in the gear because the helical teeth of the gear
the governor, a compressor for sealing exerted axial thrust and the overall 414
air in turbochargers and other com- mm diameter was very large compared
pressed air requirements, oil and fuel with the thickness of the web. In the
pumps, and for the cooling system, starting sequence, maximum stress
freshwater and seawater pumps. One was estimated at over 530 MPa, well
engine recently investigated by the beyond any reasonable estimate of
author also used this gear train to carry the fatigue endurance limit for the
starter motor torque. The manufac- material.
turer had provided two locations for
Inquiries indicated that the tandem
the starter motor—on the front and
starter motor configuration had not
rear of the gearbox. Various starter
previously been used in the long life
motors were available (electric and air-
of the engine type, and it was possible
driven). The particular application
that the gear design had never been
required a rapid and positive start, and
Fig. 12 A failed gear from the auxiliary gear assessed for stresses under starter
both locations had been used in train in a high-speed marine diesel motor loads. The manufacturer quick-
parallel, with two powerful air-driven engine
ly proposed a design change, using a
motors. The internal arrangement of
greater web thickness, integrating the
the gearbox meant that the starter
hub and the web as one component,
motor torque was transferred to the
and increasing the fillet radius join-
crankshaft via an intermediate gear.
ing these surfaces. Further FEA
After a time one of these gears failed,
calculation showed that the peak
losing a large chunk with quite a few
stress was reduced by more than 50%,
teeth, fortunately with a surprisingly
and replacement across the class is
small amount of collateral damage.
Fig. 13 The fracture surface showed fatigue underway. This is a case where the
The failed gear is depicted in Fig. 12.
cracking initiated at the hub/web combination of mechanical engineer-
transition ing and FEA to explain the findings
Failures of gear teeth are an occa-
from a materials investigation led to
sional hazard in this industry, but
a design improvement, when the fail-
metallurgical examination revealed
ure might otherwise have been treat-
that this failure had begun not from
ed as an isolated incident.
the teeth but from the hub. Figure 13
shows the region of the fracture
surface where several initiation points Camshaft Bearings in a Marinized
were found. Figure 14 is a higher reso- Locomotive Engine with a
lution micrograph, identifying some Fabricated Structure
of the larger progression marks. These In order to meet cost and schedule
occurred near the end of the crack constraints associated with building
propagation process. Load calcula- to order, the Royal Australian Navy
tions and finite-element analysis has purchased some of its ships
(FEA) at the DSTO laboratory second-hand. One example was a pair
revealed that the bending load at the of landing ships purchased from the
hub during the start sequence caused U.S. Navy and refurbished in Australia
local stress above the fatigue endur- to serve as amphibious landing plat-
ance limit of the material, despite the Fig. 14 Detail of the fracture surface showing forms. One of these, HMAS Kan-
gear being made from a high-quality major progression marks imbla, is shown in Fig. 16.

Practical Failure Analysis Volume 3(2) April 2003 29

 Thorough Failure Investigation in Enhancing the Reliability of Machinery (continued)

the 251C engines alongside the camshaft, enclosed by a

were upgraded to the cover as seen on the right of the
“F” specification in drawing in Fig. 18, and operated by a
many respects. rocker, which loads the camshaft and
its bearings in the upward direction.
During the refit of
Thus, the camshaft bearings are
these vessels, a num-
subjected to alternating vertical load
ber of cracked cam-
at a frequency of half engine speed,
shaft bearing sup-
giving rise to a cycle rate of the order
ports were found,
of a million cycles per day of opera-
such as that in Fig.
tion. Because the failure was in fati-
17. The crack can be
gue, and presumably had not been
seen running around
manifest in prototype testing, it was
the top of the welded
concluded that the stress in the weld-
joint and well down
ed joints was just above their fatigue
the near side.
endurance limit. A small increase in
The mechanical the strength of the joints could be
arrangement is expected to reduce the failure rate to
shown in Fig. 18, near zero.
with the camshaft
bearing shaded in Most of the cracks discovered
gray. The housing appeared to be fairly small, though in
Fig. 15 Calculated stress distribution in the broken gear under some cases, it appeared that the bear-
conditions of starting torque is welded to the in-
ner crankcase wall,
These ships were fitted with ALCO which forms part of the water jacket
engines, of quite an old design known for the cylinders. The camshaft suffers
as Type 251C, whose crankcase is a repeated loading downwards from the
welded steel fabrication. Each ship valve lifters, each valve opening on
had six V-16 propulsion engines and every second crankshaft revolution.
three similar V-8 engines for electrical While both valves are shut, the fuel
power generation. During the Aus- injection pump is operated by an
tralian refit, each ship was fitted with adjacent cam. These three cams are
a fourth generator to allow for in- on one span, between two camshaft
creased “hotel loads” (power require- bearings. The injection pump is
ments other than propulsion) in the mounted outside the crankcase,
new role. The additional engines were
of a later type, known as 251F, and

Fig. 18 Arrangement of camshaft and support

Fig. 17 Camshaft bearing housing cut from bearing, tappets and rockers in ALCO
Fig. 16 HMAS Kanimbla ALCO 251C engine 251C engines

30 Volume 3(2) April 2003 Practical Failure Analysis

ing housing was close to separating be modified in situ to stiffen the cools. In the situation of these
from the structure completely. In one structure to a similar specification. camshaft bearings, inspection of the
engine, the cracks had penetrated the However, close inspection showed root run of the weld is clearly im-
water jacket and done considerable that the camshaft bearing supports possible after completion of the work,
damage, so a decision was taken to were far from the last items fitted in because the joint forms a closed
write off this engine and replace it the construction procedure, and a perimeter.
with a 251F engine. Several of the retrofit of additional webs would be a
cracked bearings were cut out of the The proposed procedure [13] was
very difficult procedure. The third tested on some of the cracked housing
cylinder block, and the cracked welds option was to carry out a weld repair
were examined at the Melbourne joints cut from the scrapped engine,
to the original specification but seek and welders were qualified by section-
laboratory of DSTO (Fig. 17). The
to ensure that the new joints were ing their test articles. It was necessary
investigation showed that these were
slightly stronger than the original. to grind out the damaged areas well
fatigue cracks that had begun at the
inside of the joint and spread slowly Given the apparently marginal beyond the ends of the crack. The area
out to the visible perimeter. original strength, this last option was then thoroughly washed and
appeared feasible. dried to remove, as far as possible,
An important finding was that traces of oil and solvents. The welded
There was a weld repair procedure,
cracks discovered at the outside of the joint was rebuilt a little deeper than
but advice from an experienced
weld were considerably longer be- the original weld to increase the
Western Australian company indi-
neath the surface. Given the age of strength. During this process thermal
cated that it was quite difficult. The
the engines, it was concluded that distortion was accepted. On comple-
main problem was to maintain the
there were almost certainly some more tion, the bearing was bored oversize,
geometry of the bearing, which could
slowly growing subsurface cracks. then an undersize bush inserted.
be subject to misalignment due to
These could not be reliably detected Finally, the bush was rebored in line
thermal distortion during welding.
but could be discovered by appropriate to complete the procedure, which was
Instead, a new weld repair procedure
periodic inspection after the engines successfully adopted by the shipyard.
was proposed by DSTO, which could
were returned to service. Cracking
be applied with confidence under the
through the wall was deemed to be It is expected that the repaired
local conditions at the Australian
unlikely if the cracks were discovered bearings will be at least as strong as
shipyard.
early. the originals, ideally slightly stronger
Once cracks are present in the struc- so that if any failures of the repaired
Among the changes between 251C ture, lubricating oil tends to seep into bearings do occur, they will be at very
and 251F, the camshaft bearing and them. For at least three reasons it is long intervals. More likely is the
its support structure had been modi- necessary before welding to remove as discovery of newly surfacing failures
fied, the later engines having much much of this oil as possible. One in other bearings, and an inspection
deeper webs supporting the camshaft reason is to reduce the effects of fumes interval has been chosen to bring
bearings. Clearly, the 251C design had on the welder, because a manual these to light at an early stage. In this
been found wanting in this respect as welding technique is the only practi- way, the overall reliability of the
the engines aged, which is not an cable approach. A second is to reduce repaired engines, at least in respect of
unusual situation in engines given the the risk of a minor fire when the weld camshaft bearing cracking, is expected
usual philosophy of evolving design. is carried out. Perhaps most impor- to be at least as great as the original.
However, the project team was now tantly, it is also necessary to avoid the A guaranteed significant increase in
faced with three options. One would problem of the weld being weakened reliability is difficult to achieve with-
be to replace the cylinder blocks with by hydrogen-assisted cracking. This out replacing the cylinder block, the
new ones to 251F or later specifi- phenomenon occurs when welding in main structural component of the
cation. The cost of removing engines the presence of water or hydrocarbons engine. The main unknown in the
from the bowels of a ship is consid- causes contamination of the weld with procedure is the effectiveness of the
erable, and this would be an expensive hydrogen. The presence of hydrogen precautions against hydrogen-assisted
operation. Secondly, the engines could leads to local cracking as the weld (continued on page 40)

Practical Failure Analysis Volume 3(2) April 2003 31

 Thorough Failure Investigation in Enhancing the Reliability of Machinery (continued
(continued)from page 31)

cracking in the repair procedure. to be a major contributor to a failure, as early as possible and in as undis-
Experience so far is encouraging, and especially in welded structures where turbed a state as possible. If a mech-
no new problems have emerged dur- defects are almost inevitable and in anical engineer and a metallurgist (or
ing the first few years back in service. some materials very difficult to detect. other materials expert where non-
A realistic designer uses a factor of metallic materials are involved)
In this case, a comprehensive failure
safety sufficient to allow a credible examine the site together, the differ-
analysis was neither practicable nor
defect (one that might reasonably be ent perspectives are synergistic. In the
regarded as necessary. The main
expected to occur and not be detect- experience of the author, the mech-
loading mechanism, while difficult to
ed) to not lead to a failure under the anical engineer typically looks out-
quantify, was easy to identify. The
anticipated loading. ward, for the source of loading or
upgrades to the engine could be
When a failure has occurred, and foreign bodies, while the materials
shown not to increase the camshaft
particularly when other similar scientist tends to look inward, seeking
loads significantly. Therefore, it was
failures have occurred or a safety- clues in more microscopic details. An
possible to show with confidence that
critical component has failed, the experienced investigator may cover all
a small increase in effective strength
investigator needs to determine such aspects, but it is unusual to find
was sufficient to produce a repair with
whether or not the loading exceeded an investigator with a sufficiently wide
a life at least as long as that of the
that expected by the designer. It can range of skills to do such a job alone.
original joint and with a reasonable
be very difficult to revisit the original One expert might home in on a
prospect of increased life.
design decisions, especially in mach- particular theory and eventually
propose a sequence of events and
Reflections inery whose design has evolved. The
causal factors not supported by all the
In many of these cases, the investi- component designer is not likely to
be available for consultation, and the evidence. A team is more likely to
gation began with a single failed question theories critically. In the end,
component being forwarded in a crate design may have been based on “pencil
and paper” methods that are not the explanation must be supported by
to the laboratory with a request for a all the evidence.
metallurgical investigation. The user amenable to reexamination using
wants to know why the item has failed. modern methods.
In engine work, and in many other
Provided fracture surfaces have not After a failure, it is important to machinery applications, there is a
been too seriously damaged, the determine the sequence of events that tendency for the machine to do much
metallurgical investigation is likely to led up to the failure. Once the first damage to some components in the
reveal whether or not the failure is a component to fail has been identified, vicinity of the failure during the time
sudden overload or fatigue and may the best way to determine the loading in which the machine is shut down
identify stress corrosion or other may be for the investigating team to and comes to rest. Sometimes the
failure modes. It may sometimes find assess the loads independently of the original evidence of the failure is
a material defect at the initiation point manufacturer, then compare the two actually destroyed, and it is then a
of the fracture. Such a defect may not calculations if possible. In this way, difficult job to make any reliable
be the prime cause of the failure; a different assumptions made by each determination of the cause of failure.
normally acceptable defect may party may be brought to light and Many components may be damaged,
simply have provided a starting point their validity challenged. In this and it may not be at all obvious which
for a fracture, driven by a load the process, ideally both the design as- failure occurred first. Furthermore,
designer did not anticipate or by sumptions and the assumptions of the failures sometimes lead to damage to
vibration causing a cyclic component investigators can be tested before surrounding items, such as local fire
of load that was not expected. conclusions and repair or modifi- damage, and so forth. In such situa-
The specified strength of any cation options are finalized. tions, the introduction of physicists
material includes consideration of Where a failure occurs to a major and chemists into the team to explain
assumptions and statistics about the piece of capital equipment, it is ad- the condition of some of the wreckage
size and frequency of defects. Some- vantageous to have as many informed can be invaluable. Coming from a
times a defect can genuinely be said eyes as possible examine the wreckage multi-disciplinary Defence laboratory,

40 Volume 3(2) April 2003 Practical Failure Analysis

the author has on several occasions Conclusions well supported by all the evidence
seen the value of a diverse team of Investigation of machinery failure from the failure.
investigators.
is a difficult task. It requires a range Investigation may require a redeter-
Perhaps the most important lesson of skills rarely found in any one indi- mination of the loading of an item,
is that the entire machine and its vidual. Any organization attempting independently of the original design
operation, not just the broken com- to investigate failures as a significant calculations. Sometimes earlier cal-
ponent, need to be investigated, and part of its activities needs a suitable culations may have been inadequate,
as soon as possible. Fortunately, most team of experts in engineering, but more often, developments in the
machinery failures do not involve ser- materials science, and microscopy, machine have led to changes in
ious injury. However, if the machine equipped with adequate facilities, and loading. These are more likely to be
could be treated as “evidence” and not with freedom to investigate without identified by a new examination of the
disturbed, the investigators would excessively restrictive time constraints. loading so that previous assumptions
have a much easier job. In practice, it There may be a need to achieve quick- do not cloud the thinking of the in-
is unrealistic to expect a failed mach- ly a clear idea of what happened, so vestigators. Modern FEA techniques
ine to be undisturbed, because the that a machine can be put back into allow calculations to be done, some-
operator or a local repairer would service rapidly with minimum risk of times on a personal computer, with an
initially investigate the majority of a repeat occurrence, but this need not expedition and accuracy not available
failures. Expert assistance will be cause the investigation to be aban- to the original designer, so the
sought only when the repair is beyond doned. However, it is rarely possible availability to the investigating team
the locally available resources or the to investigate something to a point of a thorough stress analysis capability
failure is expensive, dangerous, or where certainty is complete. is very valuable. In Naval engines,
repetitive, and seems inexplicable to which may have particularly demand-
the operator. Investigators should therefore be
ing load requirements, a careful and
encouraged to arrive at a preliminary
It is not unusual to find that errors comprehensive failure analysis process
conclusion as quickly as practicable
in maintenance or operation are has been found to produce significant
but should feel able to declare their improvements in reliability, with
“covered up” to avoid blame being uncertainties. On this basis, especially
placed locally. Unfortunately, this concomitant cost and safety benefits.
where safety issues affect similar
situation makes the investigation machines (in a class of vessels or air-
much more difficult, and when such craft, for example), preventive meas- Acknowledgments
an instance is discovered, may even- ures such as safety inspections and The author would like to acknowl-
tually lead to blame being assigned operational limitations can be put in edge the contribution of a number of
locally, when the fault truly lies place early to guard against similar Royal Australian Navy staff and
elsewhere. failures. Should such preliminary DSTO materials scientists, metallur-
decisions have to be made, the inves- gists, and technical staff to the investi-
Failure investigation is necessarily
tigators should have the freedom to gations discussed in this article.
a somewhat open-ended activity.
While costs must be controlled in the follow up uncertainties until satisfied Most of the figures are from reports
modern environment, an investi- that the explanation is robust. Only originating in DSTO or elsewhere in
gation that provides an accurate and then should decisions on redesign be the Department of Defence. Figures
comprehensive determination of the taken. Decisions taken on the basis of 3, 4, and 5 are from the report of the
causes can save a great deal of money an early preliminary conclusion can Board of Inquiry cited in Ref 5 but
in the longer term. Where major be difficult to reverse if further fail- originated at the AMEC Metlabs
capital equipment is concerned, on- ures occur. The more expensive the Laboratory in Western Australia. Fig-
going failures due to inadequate failed machine, and the greater the ures 6, 7, and 8 are from the MIIU
investigations are likely to cost much safety risk involved in failure, the more Report (Ref 8), Figures 6 and 7 be-
more than a thorough failure important it is that the explanation, long to ATSB (formerly MIIU), and
investigation. and therefore the corrective action, is Fig. 8 originated in work performed

Practical Failure Analysis Volume 3(2) April 2003 41

 Thorough Failure Investigation in Enhancing the Reliability of Machinery (continued)

by Herman Research Laboratory on sional Engineering Publishing Ltd., Lon- Pty Ltd, Welshpool WA, Report 8H24/
behalf of the ship owner. The quality don, 2002, p. 13. M1, June 1998.
of the work of these laboratories, and 2. W. Wong: How Did That Happen?– 10. Departmental Investigation into the Engine
where appropriate the permission of Engineering Safety and Reliability, Profes- Room Fire on Board the Australian Antarctic
sional Engineering Publishing Ltd., Lon- Research and Supply Vessel AURORA
the owners to use these images, is duly don, 2002, p. 128. AUSTRALIS at the Antarctic Ice Edge on
acknowledged. Figure 15 is from an 22 July 1998, Incidents at Sea Report 135,
internal DSTO report and was 3. D Woodyard: Pounder’s Marine Diesel
Marine Incident Investigation Unit,
Engines, 7th ed., Butterworth Heinemann,
generated from the excellent FEA 1998, p. 158ff.
Commonwealth Department of Transport
modeling work of Hai Hoang Tran, and Regional Services, Canberra, June
4. M. Jennings: Prevent the Leak; Cut the 1999.
formerly of DSTO’s Maritime Risks, Marine Engineers Review, Sept
Platforms Division. 1998, pp. 13-15. 11. G. Goodwin: Report on the Failure of a
Flexible Hose in the Engine Room of RSV
DSTO staff who have made a 5. Report of the Board of Inquiry into the Fire Aurora Australis, Leading to the Fire of 22
particular contribution to the investi- in HMAS WESTRALIA on 5 May 1998, July 1998, published as an Attachment to
Defence Publishing Services, Canberra, Ref 10.
gations described in this article 1998.
include Dr. Stan Lynch, Dr. Len 12. S. Teo: Vibration and Pressure Pulsation
6. J.R. Galpin and M.E. Davies: Failures of Measurements on Main Engines Fuel Lines
Davidson, Dr. Darren Edwards, Dr.
Low Pressure Fuel Systems on Ships’ Diesel on the HMAS Westralia, AMEC report
Bob Phillips, Stewart Alkemade, Engines, BMT Edon Liddiard Vince Ltd., prepared for Defence Materiel Organi-
Alban Cole, Max Bentley, and Tom MSA Research Project 401, Marine sation, July 2001.
Fraser. A number of invaluable lessons Safety Agency, Southampton, Jan 1997.
13. R Phillips: Weld Repair Procedure for
were learned from interaction with 7. J. Cannell: HMAS Westralia Fire 5/5/98, Camshaft Bearings in ALCO 251C Engines
them. Without the diverse skills and Study of Fuel System Characteristics, Final in RAN Service in LPA Vessels, DSTO
insights brought to the work by these Report, Engineering Dynamics Pty. Ltd., report to LPA Project Director, 1998.
unpublished report for Director of Naval DSTO File No. 495/206/0257.
people and other colleagues, the con- Materiel Certification, Canberra, 12 July
clusions of several investigations 2000.
would not have been reached as
8. H.J. Gatjens and H. Rulfs: Damping of
quickly or reliably, and in some cases Pressure Vibrations in Fuel Systems of Ship
the author’s initial theories might not Diesel Engines, Hansa, Sept 1990, 127(17/
G. Goodwin, Senior Research
have been debunked as promptly. 18), p. 1064ff. Engineer, Marine Propulsion, DSTO
Maritime Platforms Division, Cordite
9. J. Bromley: Metallurgical Examination and
References Testing of Flexible Fuel Hoses, Report for
Avenue, Maribyrnong, Victoria, 3032,
1. W. Wong: How Did That Happen?– the HMAS WESTRALIA Board of Australia. Contact e-mail: geoff.
Engineering Safety and Reliability, Profes- Inquiry by Metlabs, AMEC Engineering [email protected].

What do you want to learn today?

Subscription information on page 2.

42 Volume 3(2) April 2003 Practical Failure Analysis