Professional Documents
Culture Documents
Presidents’ Choice
Learning from rotating machinery
failures around the world
Michael Neale OBE BSc(Eng) WhSch FREng FIMechE FCGI Chairman of Neale
Consulting Engineers Ltd Past President of IMechE
Accepted: 17 August 2006 Date Received: 1 June 2006
ISSN: 17518471
LEARNING FROM ROTATING MACHINERY FAILURES
AROUND THE WORLD
MICHAEL NEALE OBE. BSc(Eng). WhSch. FREng. FIMechE. FCGI.
Chairman of Neale Consulting Engineers Ltd. Past President of IMechE.
Introduction
There are a very large number of rotating machines in use around the world, installed in power
stations, process plants and factories, and as the power units in all the transport systems.
Unfortunately they are not totally reliable, and when they do fail, it is essential to determine the
causes, as a guide to the repair programme and to the prevention of further failures in the future.
Many of the failures occur at the components of the machines where there is relative movement
between loaded surfaces, such as bearings, gears, pistons, seals and couplings. This is because
the combination of load and movement, in these components, provides a potential source of
energy to generate noticeable damage. Engineers who understand tribology, which is the
technology of relatively moving surfaces, and their friction, lubrication and wear, can therefore
provide useful guidance on machine failure investigation and design modifications. This paper
summarises the experience of a group of consulting mechanical engineers with particular
experience of tribological problems, who have been investigating machinery failures all over the
world for over 50 years.
One important aspect of this experience is that it can be applied not only to the solution of the
immediate problems, but also to recognising patterns in the failure of all kinds of machines, and
this can be used to provide general guidance on risks and on improvements that are needed in the
methods of machine design. The experience discussed in this paper is based on 160 cases of
major failure which have been investigated over the last 15 years. In common with many failures
and accidents, each event usually has a number of causes which interact and result in the event.
These causes have been analysed, and can be divided into those which are design related and
therefore of particular relevance to this paper, and those which arise from the general working
environment of machines. The causes which are design related are listed in Table 1 in the order
of their occurrence, and Table 2 gives a similar listing of causes related to the working
environment.
Description of the cause of failure Number of cases
Unexpected interaction between components 20
Errors in detail design 19
Loss of operating clearance from thermal instability 17
Errors in design layout 15
Errors in material choice 12
Errors in lubrication system design 9
Errors in lubricant selection 8
Unexpected system resonances 7
Table 1. 107 Design related causes of failure
Description of the cause of failure Number of cases
Manufacturing errors 18
Installation errors 17
Insufficient lubrication 12
Lubricant contamination 11
Machine overload 7
Maintenance and monitoring errors 7
Operating errors 6
Environmental effects 5
Table 2. 83 Working related causes of failure
The following sections of this paper look at examples from various of the categories of causes
listed above, in order to provide guidance on how to avoid similar problems in the future. The
examples are intentionally made anonymous to avoid disclosing the organisations involved, and
to encourage others to add further examples in the future.
Interactions between components
A cause of problems on some electrical machines is an interaction between the electrical and
mechanical systems, with electric currents passing through the mechanical parts. This is not
critical when the current flows through a single component, but when it flows between two
relatively moving components, separated by an oil film, higher voltages can arise leading to
electrical discharges between them. This can produce spark erosion damage, which on plain
bearing surfaces, usually starts at the minimum oil film thickness position, and on rolling
bearings takes the form of rippling of the surface of the races in the loaded region, which can
develop into early fatigue failure. If the current flow cannot be diverted the problem can often be
solved by electrically insulating the bearings from the shaft or housing.
In gearboxes, problems can arise if the shafts are too flexible in bending and as a result cause the
gear mesh to be out of alignment with a resulting local overload of the ends of the teeth.
A similar problem can arise if the sun gear in an epicyclic gear box is driven by a gear coupling,
as shown in Figure 1. When gear couplings are subject to lateral misalignment they can generate
substantial lateral loads on their support shafts, due to a combination of the tilting moment on
their misaligned teeth and the friction moment from the sliding teeth. This can generate lateral
moments on the coupling sleeve of up to one third of the transmitted torque, which gives rise to
lateral loads as shown in Figure 2. In the case of the epicyclic gear box this resulted in a failure
of the sun gear due to misalignment. It was cured by replacing the gear coupling by a quill shaft
drive.
The lateral loads from gear couplings can also cause vibration problems on machines with their
rotors supported on plain bearings. If the misalignment pattern produces an upwards load on the
rotor, the reduced bearing load can allow half speed vibration of the rotor. This can usually be
overcome by realigning the machines to change the direction of the generated load. The problem
can also be overcome by replacing the gear coupling with a diaphragm coupling which does not
generate lateral loads. They are however generally heavier than gear couplings and the increased
overhung mass can reduce the critical speed of the rotor, which may also cause vibration
problems.
Figure 1. Lateral loads from the gear coupling caused failure of the teeth of the sunwheel
Figure 2. The lateral loads generated by a misaligned gear coupling
Problems can also arise if it is not fully appreciated that a shaft supported in fluid film lubricated
plain journal bearings will operate in an eccentric position in order to generate the necessary
hydrodynamic pressures. An example was in the drive system for a rotating machine, with the
drive shaft supported on plain bearings as shown in Figure 3. It was also required for the driven
machine to be barred over at a lower speed under some operating conditions, and to allow this
the shaft was fitted with clutch teeth, which could be meshed with an internally toothed drive
ring, from a low speed drive. When this was operated the driving teeth suffered severe fretting
and wear due to their eccentric relative operating positions. In this case the load on the shaft
support bearings arose just from the shaft weight which was constant. It was therefore possible to
set the support bearings for the toothed ring in an eccentric position so that it became concentric
with the operating position of the drive shaft, and avoided the problem.
Component interactions can also cause problems in the rolling supports for large heavy moving
structures, such as moving doors or the support rails for astronomical telescopes. These are
commonly mounted on load carrying wheels, often in conjunction with lateral guide wheels. If
the load carrying wheels have cylindrical or conical running surfaces and are supported on
cylindrical bearings, they will follow a rolling direction exactly at right angles to the wheel
contact line, as shown in Figure 4. If this does not correspond exactly to the track set by the
lateral guide wheels, then these can be seriously overloaded by the main wheels determining the
track of the actual movement. This can result in major structural distortions to the point of
failure. This problem can be avoided by making the main wheels steerable, to match their
required lateral location. This requires the wheel outside diameters to be of a spherical profile,
and for the wheels to be mounted on self aligning bearings which allow the wheels to tilt
laterally, so that they can self steer to follow the track dictated by the lateral guide system.
Clearance loss from thermal instability
This is a problem, which arises mainly during the starting of machines which operate in a cold
environment. After starting, the moving parts warm up more rapidly than the machine casing,
and the resulting differences in thermal expansion lead to a loss of the operating clearance
between them.
An example of where this could arise would be on ram air turbines on aircraft which provide
emergency hydraulic power in the event of engine failure. They consist of a hydraulic pump
which is driven by an air fan turbine and is pushed out into the airstream when an emergency
occurs. The unit is stored just under the aircraft skin, where the temperature is around –10 o C and
when extended into the airstream, it speeds up to a few thousand RPM in about 5 seconds. To
overcome any potential problems it would be essential to use bearings with the maximum
operating clearance, and to position them in housings of maximum flexibility, that is, housings
with a radial split or a low radial wall thickness. The problem would also be made worse if the
drive shafts are made tubular to give reduced weight, which will also result in a reduced warm up
time.
The problem has occurred in the gearboxes of wind driven generators, which are in an exposed
position at the top of a high support column in cold air. When these are started from rest, the
moving parts warm up more rapidly than their more substantial housings, and operating
clearances can be lost. These gearboxes are also required to provide a high speed up ratio
between the rotor and the generator, which requires a number of gear trains, as shown in Figure 5
Figure 3. Clutch teeth failed by fretting due to eccentricity of shaft operating in plain
bearings
Figure 4. Wheels roll at right angles to their contact line. Guide systems to correct this get
overloaded
Figure 5. Wind generator gear box with bearing inside the 3 rd stage gear wheel, which fails on
cold starting
generator bearing ; simulated thermal transient
90
80
70
60
Temp (degC)
50 Journal
Housing & bearing
40 Difference
30
20
10
0
0.0 50.0 100.0 150.0 200.0
Time (mins)
Figure 6. Temperature differences across the bearing of a large alternator when started from cold
.
In the design of most machines such as this, it is usually considered good practice to mount the
bearings inside a relatively strong and rigid support housing. In this case one of the bearings was
mounted inside the robust third stage gearwheel and was supported on a tubular shaft. This
bearing failed on start up from loss of operating clearance. In this case the problem was solved
by the use of bearings with the maximum possible radial clearance, but could also have required
the repositioning of the bearing to an area with a more flexible housing.
Another type of unit which has experienced this problem is marine thrusters, which are mounted
below a ships hull and carry a directional propeller to provide lateral or other directional thrusts.
Once again these are in a cold environment, provided by the sea water, and are brought up to
speed from rest in a short time when required to provide positioning of the ship. The combination
of a cold housing and rapid shaft warm up again requires the maximum radial clearance in the
bearings, and the maximum local flexibility in the bearing housings.
Gas turbine driven alternators operating in a cold climate can also experience similar problems,
particularly if the alternator is of a typically robust power station type with heavy end casings,
and the gas turbine is an aircraft type able to accelerate rapidly up to full speed. An estimate of
the shaft and housing temperatures in a typical case is shown in Figure 6. It is interesting that, in
this case, the time taken to achieve the maximum temperature difference, and therefore clearance
loss, is about 40 minutes, unlike some other lighter weight examples where the failure can be
relatively rapid.
Similar problems have also been experienced on power station coal mills, which again are
usually positioned in an unheated area adjacent to the coal pile. These may have a relatively long
horizontal drive shaft from an electric motor kept at a distance from the operating area. The drive
shaft is usually in a tubular casing supported in rolling bearings designed to allow some axial
movement to allow for temperature changes in the shaft. Most designers will tend to position the
bearings in line with any external radial support plates, but the higher radial rigidity and
increased warm up time in these areas, can give rise to clearance loss inhibiting any axial
adjustment and resulting in bearing overloading and failure.
For the same general reasons, it is usually undesirable to use thick walled solid bronze bearings
in larger machines except at low rotational speeds. This is because, if they are in a relatively rigid
housing, they can expand inwards, when warming up, and reduce the operating clearance, and
can also yield circumferentially and lose their fit in the housing. For higher speeds, steel backed
plain bearings should be used with a whitemetal or bronze lining.
Various design errors
When designing a complex machine it is always possible for various detail design errors to
occur such as relatively sharp changes of section or welding on shafts that end up carrying
bending loads and most designers are familiar with these problems. More significant errors are
those which relate to the design arrangement. One of these was where a small steam turbine was
modelled on a successful larger machine and experienced half speed shaft vibration because the
bearing loads were too low. This arose because in the downsizing, the rotor weight decreased in
proportion to D 3 , while the bearing area had only decreased in proportion to D 2 . Another problem
arose when a fan, turbine driven via a gearbox, was duplicated on a plant, and to match the
ducting arrangements was made to rotate in the opposite direction. This changed the direction of
the bearing loads in the gearbox and resulting in bearing failure.
While it is not difficult to learn the lessons from the previous examples, there are some more
significant problems now arising from computer aided design which is now widely applied. An
example of this was in a large tilting pad thrust bearing where the design team used a computer
to design a bearing with the maximum film thickness. The computer program produced a design
with the largest possible pads mounted closely together, which resulted in very high operating
temperatures in service. The reason for the problem was that tilting pad bearings require a clear
space between the pads to allow the hot oil from one pad to escape and be replaced with cold oil
to feed the next pad. The lesson from this is that the design layout of any machine needs to be
determined by an experienced designer, and his initial attempt can then be analysed and
improved by computer analysis. The simple rule is, use computers for their excellent
performance in design analysis, but do not use them to create design layouts.
In contrast to this previous example it should be remembered that previous design experience is
not infallible if there are major changes in some critical operating features. An example of this
was a problem in the design of a high pressure fuel pump designed to operate with its plain
bearings lubricated by the fuel. The bearing dimensions were kept broadly the same as for oil
lubrication, whereas in fact, as the operating viscosity of a lubricant is decreased, the bearing
diameters need to be increased. In examples of this kind, it also needs to be remembered that the
use of lubricants other than oil generally requires a change of bearing material. For example, in
the water lubrication of bearings and gears, the optimum combination is generally steel and
plastic.
It is also possible to make errors in the choice of materials if a design is taken outside the
previous limits of experience. This occurred in the design of a very large cylindrical roller
bearing with an outside diameter of 3 metres. The rollers were made from steel of the same
specification as for smaller bearings, but because of their size were made from bar stock. Since
this was axially rolled any inclusions were in an axial direction, which combined with some
errors in heat treatment, resulted in the rollers breaking in two, lengthwise, in service as shown in
Figure 7. This problem was cured by making the rollers from individual forgings and improving
the heat treatment.
Errors, which are not immediately obvious, can also occur in the design of lubrication systems.
An example of this was in the design of a large oil lubricated hydrogen seal on a 400Mw
alternator. This incorporated a whitemetal lined face seal ring with an adjacent thrust bearing to
control the close operating clearance, as shown in Figure 8. Oil was fed in from the outer side of
the rings, and drained from the inside. Wear occurred in service because, with the oil flow
directions, any dirt fed in with the oil was retained towards the outer side by centrifugal
separation. This problem was cured by reversing the oil flow, putting the drains on the outside, so
that the dirt could easily flow away.
Problems can also occur from unexpected system resonances such as those which occurred on a
large electrically driven water circulating pump as shown in Figure 9. In this case the rolling
bearings in the motor failed from what appeared to be a very high rotating load, and the pump
shaft fractured in fatigue just below the rotor. This had arisen because the critical speed of the
unit had been reduced to just below the running speed, by the combined flexibility of the
unbraced support structure and the overhung mounting of the water pump. The problem was
cured by stiffening the frame, and reducing the length of the drive shaft.
Figure 7. A failed roller, from a very large bearing, made from steel bar stock with axial
inclusions
Figure 8. Hydrogen seals from a large alternator with oil flow radially inwards, failed from dirt
accumulation
Figure 9. Motor bearings and pump shaft failed from critical speed resonance due to flexible
frame and pump mounting
Errors in installation and maintenance
There are many potential errors in the installation and maintenance of machines but two
examples leading to failures of critical moving parts are included here.
The first, as an example of an installation problem, was of a very large offshore mooring buoy
designed to load and unload oil from tankers. This had a rotary top, so that the connecting hoses
and mooring lines could rotate to follow ship movements, as shown in Figure 10. The rotation
required a large diameter rolling bearing to be fitted in a housing on the buoy. The housing was
too large to machine, so the bearing was installed by positioning it accurately and then pouring in
casting resin to fix it, as shown in Figure 11. To do this, the bearing was positioned on four
equally spaced jacks. When it went into service, it was found to have four tight areas during
rotation, which developed into equally spaced fatigued areas after about a year, and required the
buoy to be replaced. The problem was due to sagging of the bearing between its supports during
the resin casting process. Stiffness calculations showed that 16 equally spaced supports were
actually required to keep the sagging ripples less than the operating clearance in the bearing.
An example of a maintenance problem was the failure of a helicopter main gearbox in flight, due
to the complete failure of a rolling element bearing. Other bearings were found to be on the verge
of failure, when the unit was examined after recovery. The gearbox was fitted with condition
monitoring magnetic plugs to pick up any wear debris from incipient bearing failure. The
manufacturers maintenance instructions were to check the plugs after every 50 hours of flying
time, and if any large particles were detected to halve the inspection intervals. They also required
that the particles should be stored, closely on sticky paper, so that the total area collected could
be identified and recorded. When an area of 50 sq.mm, or approximately 7mm x 7mm, was
reached, the gearbox had to be removed and overhauled. The overseas maintenance crew were
very overconfident and considered that 50 sq.mm must mean a square 50mm by 50mm. The
aircraft fell out of the sky when the area had reached about 25mm by 25mm, showing that the
manufacturers had in fact allowed a good safety margin, even allowing for the rapid increase in
the rate of debris generation that would be expected to occur under these conditions.
Figure 10. Large offshore oil transfer buoy with a rotating top turntable
Figure 11. Cross section of the buoy showing turntable bearing mounted in resin
Discussion and Conclusions
The cause of many of these failures may appear almost obvious with hindsight, but they still
occur widely because it is usually only the design team involved who learn the lessons from what
has happened. Their personal view is very often that they are slightly ashamed of what has
happened, and their management do not want the situation revealed to avoid any risk of legal
action against them, or damage to the company reputation. It is interesting to note however that
many of the failures described are repeated across many different companies and applications,
which indicates a real need to have a method of collecting and correlating machine experience of
this kind, so that improved design and operating experience can be achieved to everyone’s
advantage and safety. Also the correlation of operating experience is a major source of an
improved understanding of the technology, which is highly complementary to the conventional
idea that this only comes from the results of academic research. An example of this is shown in
Figure 12 which shows the wear rates of engine cylinders based solely on operating data obtained
in the field. This is relevant and correct information which could never be obtained in a
laboratory. There are many examples of potential studies of this kind such as the collection of
data from around the world of damage at the wheel rail contact at a practical range of wheel
loads, curve radii and rail profiles. This would be highly complementary to the excellent
academic studies of the nature of the stresses at these contacts.
It is interesting that airline pilots have devised a system to collect information on situations that
have come close to causing major accidents. This involves reporting them to their professional
association in an anonymous way, which does not challenge their personal position, but allows
everyone else to learn from their experience. There is clearly a role here for the professional
engineering institutions to set up such a system, to encourage their members to report
engineering incidents in a way that can be distilled into guidance on future design and
technological developments. This provision of leadership in technology would provide a major
role for the institutions to justify their future existence, over and above the qualification and
listing of members, the organisation of a few technical meetings and the publication of papers
and journals.
Diametral wear rate inches/1000 hrs
FIGURE 12. The wear performance of engine cylinder liners,
collected from service experience around the world