Upload
kpostulart
View
225
Download
2
Embed Size (px)
Citation preview
7/31/2019 Pump Failure Article
1/9
Feature3434
www.worldpumps.com
WORLD PUMPS October 2012
Before embarking on a pump failure
reduction programme it is impor-
tant to know how well a particular
facilitys pumps perform when compared
to those installed at a competitors plants.
There probably is room for improvement
if a plant does not measure up to above-
average pump failure statistics.
MTBFs and failure analysis
An easy comparison among pump users
is feasible. It consists of adding up all
process pumps installed at a plant and
then dividing by the number of pump
repairs per year. For a well-managed and
reasonably reliability-focused US refinery
with 1,200 installed pumps and 156
repair incidents in one year, the mean
time between failures (MTBF) would be
(1,200/156) = 7.7 years.
The refinery would count as a repair
incident the replacement of parts any
maintenance and better pump monitoring
may have contributed to reduced failure
severity on the typical pump, although the
ultimate consequences of some pump fail-
ures are grave indeed.
The MTBFs (installed life before fa ilure)
shown in Table 1 have been estimated in2004. Published data and observations
made in the course of performing main-
tenance-effectiveness studies and relia-
bility audits in the late 1990s and early
2000s were used in these estimates.
Seal-life statistics were estimated in the early
2000s (Table 2). These led to an estimation
of reasonable goals (Table 3). Note that the
target is less than best actually achieved.
Many plants achieve the months of
installed lives indicated in Tables 2 and 3.
Note that the actual operating life of a
component would thus be about one-
half of its installed life. To reach these
pump life expectancies, the pump
components themselves must be oper-
ating at the highest levels of reliability.
An unsuitable seal with a lifetime of one
month or less would have a serious
negative effect on pump MTBF, as would
an under-performing coupling or bearing.
Calculating MTBF projectionsSimplified calculations2 will give an indi-
cation of the extent to which improving
one or two key pump components can
improve overall pump MTBF.
parts regardless of cost. In this case, a
drain plug worth two dollars or an alloy
impeller costing $5,000 would show up
in the same way on the MTBF statistics.
Only the replacement or change of lube
oil would not be counted as a repair.
A best-practices plant counts in itstotal pump repair cost all direct labour,
materials, indirect labour and overheads,
administration cost and the cost of labour
to procure parts. It assigns a value to
failure avoidance, even the pro-rated
value of avoiding pump-related fire inci-
dents. Likewise, it assigns a monetary
value to a workforce relieved of pump
repair burdens and assignable to proac-
tive asset failure avoidance tasks.
Typical published pump repair costs have
averaged $10,287 in 1984 and $12,000 in
2008. After inflation is factored into a repair,
an actual cost reduction trend is indicated
over this 24-year time span. Predictive
Widely accepted statistics point to around 7% of a plants process
pumps consuming c. 60% of the maintenance funds allocated to
its entire pump population. With repeat failures largely responsible,Heinz Bloch outlines successful failure avoidance actions that can
be employed to extend the mean time between failures.
Problem pumps:a thing of the past
Operating
0262 1762/12 2012 Elsevier Ltd. All rights reserved
Table 1. Pump mean times between failures1
Pump type and location MTBF (years)
ANSI pumps, average, USA 2.5
ANSI/ISO pumps average, Scandinavian pulp & paper plants 3.5
API pumps, average, USA 5.5
API pumps, average, Western Europe 6.1
API pumps, repair-ocused refnery, developing country 1.6
API pumps, Caribbean region 3.9
API pumps, best-o-class, US refnery, Caliornia 9.2
All pumps, best-o-class petrochemical plant, USA (Texas) 10.1
All pumps, major petrochemical company, USA (Texas) 7.5
7/31/2019 Pump Failure Article
2/9
Feature 35
www.worldpumps.com
WORLD PUMPS October 2012
Say, for example, that theres agreement
that the mechanical seal is the pump
component with the shortest life,
followed by the bearings, coupling, shaft
and sometimes impeller, in that order.
The anticipated MTBF (operating MTBF)
of a complete pump assembly can beapproximated by summing the individual
MTBF rates of the individual components,
using the following expression:
1/MTBF = [(1/L1)2 + (1/L2)
2 + (1/L3)2
+ (1/L4)2]0.5 (1)
In a 1980s study, the problem of mechan-
ical seal life was investigated. An assess-
ment was made of probable failure
avoidance that would result if shaft
deflections could be reduced. It was
decided that limiting shaft deflection atthe seal face to a maximum of 0.001 in
(0.025 mm) probably would increase seal
life by 10%. It was thought that increasing
seal housing dimensions to accommodate
modern seal configurations would more
than double seal MTBF. All components
that could be upgraded were examined.
The life estimates were collected and then
used in MTBF calculations.
In Equation 1, L1, L2, L3 and L4 represent
the life, in years, of the component
subject to failure. Using applicable datacollected by a large petrochemical
company in the 1980s, MTBFs and esti-
mated values for a reliability-upgraded
pump were calculated. The results are
presented in Table 4. As an example, a
standard construction ANSI B73.1 pump
with a mechanical seal MTBF of 1.2,
bearing MTBF of 3.0, coupling MTBF of 4.0
and shaft MTBF value of 15.0, resulted in
a total pump MTBF of 1.07 years (actual
operating hours). By upgrading the seal
and bearings, the estimated achievable
pump MTBF (operating hours) can be
improved by 80%, to 1.93 years.
Table 4 shows the influence of selectively
upgrading either bearings or seals or both
on the overall pump MTBF of relatively
small, inexpensive pumps. Choosing a
2.4-year MTBF seal and a six-year MTBF
bearing (considered achievable by
preventing lube oil contamination via
superior bearing housing protector seals)
had a major impact on increasing the
pump MTBF. Assuming the upgrade cost is
reasonable, better seals are the best choice.
Based on year 2002 reports, a typical ANSI
pump repair costs $5,000. This average
cost includes material, parts, labour and
overheads. Assume that the MTBF for a
particular pump is 12 months and that it
could be extended to 18 months. Thiswould result in a cost avoidance of
$2,500/year which is greater than the
premium one would pay for the relia-
bility-upgraded centrifugal pump.
Reduced power demand would, in many
cases, further improve the payback.
Selecting advantageous pump hydraulics
benefits both pump life and operating effi -
ciency. Audits of two large US plants iden-
tified seemingly small pump and pumping
system effi ciency gains that resulted in
power-cost savings of many hundreds of
thousands of dollars per year. Thus, the
primary advantages of reliability-upgraded
process pumps are extended operating life,
higher operating effi ciency and lower
operating and maintenance costs.
Table 4 provides a quick means of
approximating the annual pump repair
frequency based on the total (installed
life) MTBF. Equation 1 and Table 4 also
can be used to determine potential
savings from upgrades and should shape
the pump users strategies.
An experience-based observation assumes
that every missed upgrade item reduces
pump life by 10% to a new life factor of
only 0.9 years. If we miss six such upgrade
items we will have reduced the antici-
pated life or MTBF to (0.9)6 = 54% ofwhat it might otherwise have been.
Older pumps versus newer pumps
After 50 or 60 years of service and many
maintenance actions, a large number of
standard ANSI and ISO-compliant pumps
are still operating. When they were
designed in the 1950s and 1960s time
frame, frequent repairs were accepted.
Also, plant maintenance departments
were staffed with more personnel than in
later decades.
Unless selectively upgraded, a decades-
old standard pump population will not
allow 21st Century facilities to reach
their true reliability and profitability
potentials. An older pump will generally
fail more often than a newer pump. Like-
wise, a standard process pump will fail
more often than an upgraded process
pump. It should also be realized that
even the latest industry standards tend
to list minimum component require-
ments only. So, consider upgrading
beyond the standard.
Buried in Table 1 is a plant with more
than 2,000 installed pumps; their average
Table 2. Suggested refnery seal target MTBFs1
Target or seal MTBF in oil refneries
Excellent >90 months
Very good 70/90 months
Average 70 months
Fair 62/70 months
Poor 60 months
Bearings All plants
Continuous operation 60 months
Spared operation 120 months
Pumps General types
Based on series system
calculation
48 months
7/31/2019 Pump Failure Article
3/9
Feature3636
www.worldpumps.com
WORLD PUMPS October 2012
Time
Temperature.
It is extremely important to accept the
basic premise that components will only
fail due to one, or perhaps a combination
of several, of these four failure agents. We
use the acronym FRETT to recall thesefour agents.
Because there are no failure agents
beyond these four, the troubleshooter
must remain fully focused on these four
agents. To re-emphasize by an example, a
bearing can only fail if it has been
subjected to a deviation (or deviations) in
allowable force (F), or has been exposed
to a reactive environment (RE), has been
in service beyond its design life (T), or
was subjected to temperatures outside
the permissible range (T).
The need for knowledge must not be
overlooked. For instance, bearings can fail
(overheat) when they are too lightly
loaded. The rolling elements will then
skid. But there we go again: skidding is
traceable to an inadequate force (F) and
will manifest itself as a temperature excur-
sion (T). Two of the four FRETT agents
are at work.
Each failure, and indeed each problem
incident, is the effect of a causal event. In
other words, for every effect there is a
cause; or, there is a reason for every
failure. Heres an example:
[Man injured] because man fell
[Man fell] because man slipped
[Man slipped] because there was
oil on the floor
[Oil on floor] because a gasket
leaked.
By arriving at the word gasket, the cause-
and-effect chain is focused at the compo-
nent level. Once we have narrowed issues
down to the component level, we know
that one or sometimes two troublesome
or unexpected or overlooked FRETT
Maintenance deficiencies, including
neglect/procedures
Improper operation.
Searching for additional cause categories
will not add value because anything
uncovered will, at best, be a subset of
these seven. However, if one systemati-cally concentrates on eliminating five or
six of the seven categories in succession,
one will arrive at the category where a
deviation exists. That will make it possible
to concentrate on understanding what
led to the deviation.
The pump failure analyst must pay close
attention to the under-appreciated,
generally non-glamorous basics and do
so before opting for the often costly
and sometimes unnecessary high-tech
solution. Pumps obey the laws ofphysics and there is always a cause-
and-effect relationship. It follows that
even seemingly elusive and generally
costly repeat problems can usually be
eliminated without spending much
money.
An integrated, comprehensive approach
to failure analysis starts out by either
describing the deviation, or by stating the
problem. Next, such an approach encour-
ages, or even mandates, careful observa-
tion and definition of failure modes. The
approach should employ pre-existing or
developed-as-you-go checklists and trou-
bleshooting tables1,3. Already-existing
checklists are supplied by pump manufac-
turers and can also be found in a very
large body of literature.
The FRETT approach
From observation and examination of a
failed part one identifies failure agent(s),
realizing that there are only four
possibilities3:
Force
Reactive environment
size is close to 30 hp. In 2010 this pump
population had an MTBF of slightly more
than nine years. Its owner-operators
prided themselves in cultivating effective
interaction between the mechanical and
process-technical workforce members.
The reliability professionals at this plant
fully understood that pumps are part of
a system and that the system must becorrectly designed, installed and oper-
ated if high reliability is to be achieved
with consistency. It should also be
pointed out that this plant (and others in
its peer group) conducted periodic
pump reliability reviews.
Structured analysis solves problems
Repeat pump failures are an indication
that the root cause of a problem has not
been found. Alternatively, and if the
problem cause is known, someone musthave decided not to do anything about it.
Pursuing a structured failure analysis
approach is necessary to solve problems.
Guessing or going by feel will never do.
Structured analysis means a repeatable
approach that can be learned and
employed by more than one person3.
Once an accurate analysis is documented,
remedial steps can be agreed upon and
can be implemented. Also, whenever it can
be established that a pump at location A
suffers more failures than an identical
pump at location B, we can be sure that
an explanation exists. The explanation is
found in deviations from best practices in
one or more of the following seven cause
categories:
Faulty design
Material defects
Fabrication and/or processing
(machining) errors
Assembly or installation defects
Off-design or unintended service
conditions
Table 4. How selective component upgrading infuences MTBF2
ANSI pump upgrade
measure
Seal MTBF
(years)
Bearing
MTBF (years)
Coupling
MTBF (years)
Shat MTBF
(years)
Composite
pump MTBF
(years)
None, i.e. standard 1.2 3.0 4.0 15.0 1.07
Seal and bearings 2.4 6.0 4.0 15.0 1.93
Seal housing only 2.4 3.0 4.0 15.0 1.69
Bearing environment 1.2 6.0 4.0 15.0 1.13
7/31/2019 Pump Failure Article
4/9
Feature 37
www.worldpumps.com
WORLD PUMPS October 2012
Head
% Flow110%80%
Reliability Curve
Best Efficiency Point
Hightemperature
rise
Low flowcavitation
Cavitation
Lower bearing &seal life
Lower impeller life
Lower bearing &low seal life
Dischargerecirculation
Suctionrecirculation
Reactive environment: none found;
normal chemical plant location and
ambient environment.
Time: ascertained that run length was
not excessive; the hub failed after just a
few weeks of operation.
Temperature: suppose the coupling was
heated to facilitate installation. How
was the heat applied? What tells us
that the temperature was within limits?
The temperature could have been too
high (causing overstretch) or too low
(not allowing dilation to result in suffi -
cient axial advance).
In both of these examples, the pump
failure analyst has to determine in which
cause category there is a deviation fromthe norm, which item needs to be modi-
fied and how this modification must be
implemented so as to prevent a repeat
failure. Data will be required to support
any conclusions. With data one can
define the root causes of a problem.
Without data one can, at best, determine
a probable cause.
Change analysis
Change analysis parallels and supple-
ments the structured, comprehensiveapproach. It seeks to identify what is
different in the defective item as
compared to an identical but unaffected
item. The analyst probes into when,
where and why the change occurred.
The analyst then outlines a number of
contributors must now be found. In this
case, a gasket leaked. A gasket is clearly a
component. So:
[Gasket leaked] must be due to: force?
reactive environment? time? temperature?
We must check it out on the basis of data.Without data we would be guessing, and
guessing does not lead to repeatable results.
Force: too much why do we rule it in
or rule it out? Not enough why do
we rule it in or rule it out?
Reactive environment: wrong material
selected for the medium transported in
the pipe? Why do we rule it in or rule
it out?
Time: was the same gasket left in placefor many years? Why do we rule it in or
rule it out?
Temperature: too high? too low? Which
one of these two (or perhaps both)
might be ruled in or can confidently
be ruled out in a particular instance?
The pump failure analyst must take a very
similar approach with pumps and other
machinery. For every effect there is a
cause; there is a reason for every failure
and we have to find it:
[Pump is down] because the shaft broke
[Shaft broke] our failure mode inven-
tory was consulted; lets assume we
found the surface has fretting damage.
That is a deviation from the norm.
[Surface damaged] because the
coupling hub was loose. That would
explain the fretting damage.
An analyst can now try to get to the root
cause by remembering that all pump
failure events fit into one or more of the
seven cause categories listed above. If the
coupling hub was found to be loose, what
cause categories are likely and which ones
can we eliminate?
Design error? Unlikely, since other
couplings are designed the same way
and we have verified that they are
holding quite well.
Material defects? No, since a thorough
metallurgical exam checks OK.
Fabrication error? No, because the
hardness checked OK; dimensional
correctness was verified and had been
recorded upon installation three years
ago.
Assembly/installation defect? Suppose
we have no data and defer it for
possible consideration later.
Off-design or unintended service condi-
tions? No; we rule it out.
Maintenance deficiencies (neglect/
procedures)? No, since no preventative
maintenance (PM) is required on a
coupling hub.
Improper operation? No, because we
have ascertained operator activities
were in accordance with our estab-
lished standards.
At this stage the analyst would get
back to what needs to be investigated
further or requires follow-up examination.
This might be a good time to start
compiling:
(a) A checklist of possible assembly errors.
From discussions with maintenance
personnel we might conclude that
none apply in this instance.
(b) A checklist of possible installationerrors:
Force: could have overstretched hub,
or could have had insuffi cient axial
advance on taper (insuffi cient inter fer-
ence fit).
Figure 1. Staying near the centre of this reliability curve also known as a BarringerNelson curve is a wise
course of action4. (Source: Paul Barringer, www.barringer1.com, by permission.)
7/31/2019 Pump Failure Article
5/9
Feature3838
www.worldpumps.com
WORLD PUMPS October 2012
remedial action steps and will have to
choose the steps that best meet defined
objectives. These objectives must achieve
highest safety and the analyst may pick
from a list that includes lowest life-cycle
cost (LCC), present value, highest initial
quality, meeting a certain industrystandard, a deadline, etc.
The objective of aiming for lowest LCC
usually makes considerable sense. Calcu-
lating this parameter would include the
cost of staffi ng a pump selection or reli-
ability review with dedicated, knowl-
edgeable individuals. LCC analyses must
also include the value of downtime
avoidance and MTBF extensions, as well
as the value of avoided fire and safety
incidents.
Recall that fewer pump failures translate
into fewer fires and decreased insurancepremiums. Failure avoidance creates
goodwill and enhances a companys repu-
tation. Also, having to cope with fewer
failures encourages a safety culture and
frees up personnel whose proactive activi-
ties avoid other failures, etc.
Over the decades, we have come to
realize that pump failure statistics are
rarely very scientific. Still, they are experi-
ence-based and should not be disre-
garded. If your MTBF hovers around
average, identify the repeat offenders and
subject them to an uncompromising
improvement programme. In the hydro-
carbon processing industry, about 7% of
the pump population consumes 60% of
the money spent on pump maintenance
and repair. Getting at the root causes of
failures on these 7% will save a lot of
money.
A strategy that involves rational thinking
is solidly supported by a minutes worth
of looking up vendor documentation. A
sound strategy also mandates respect for
the simple laws of physics. Its a strategy
that results in failure cause identification;
it will lead to future failure avoidance and
will extend pump MTBF.
ContactHeinz P. Bloch
Process Machinery Consulting
3163 W. 111th Drive
Westminster, CO 80031, USA
Tel: +1 515 225 0 668
E-mail: [email protected]
www.heinzbloch.com
It is fitting, then, to conclude or recap this
article by pointing to a very simple illus-
tration (Figure 1). This illustration tries to
convey that many parameters interact to
cause repeat failures in pumps. Many of
these are classified as hydraulic issues and
much work has been done to improvepump hydraulics. However, the majority of
what we choose to call elusive failure
causes are linked to mechanical issues.
We have become accustomed to mainte-
nance routines that rarely question the
adequacy of a vendors design. Failure
causes have become elusive because we
overlook or forget (and even disregard)
the laws of physics.
It should also be pointed out that process
pump vendors often merely furnish a
barely adequate design5. Vendors are left
with the impression that users are unwilling
to pay for a superior design. Moreover,vendors and pump manufacturers benefit
from the sale of replacement parts and are
in business to generate income.
We must also not forget that pump
manufacturers have right-sized, down-
sized and economized the way they do
business. Few (if any) of these organiza-
tional realignments benefit the user and a
preponderance of repeat failures attests to
this lack of benefit. Some vendors and
manufacturers no longer employ process
pump experts and diligent craftsmen. The
user-purchaser may belatedly come to
realize that he has become the manufac-
turers quality control inspector. Many
must experience failures before they
accept this fact. When they learn the hard
way, they must allocate money to ward
off this eventuality by suitable pre-
delivery inspections.
Timely and competent up-front action by
the owner-purchaser is one of the keys to
failure avoidance. This up-front action
includes development of detailed specifi-
cations for process pumps and some of
the key components that go into good
process pumps. Once a process pump
arrives in the field, it must be properly
installed and maintained. To be effective,
the facility must adopt work processes
and procedures that harmonize with best-
of-class thinking5.
To avoid repeat failures, pump owner-oper-
ators must deliberately push certain routinemaintenance actions into the superior
maintenance category. Superior mainte-
nance efforts will lead to (or are synony-
mous with) pump reliability upgrading.
In essence, the course of wisdom
demands that we move away from
business as usual. Before one can apply
practical wisdom5, one must acquire
knowledge and understanding. I hope
that this article has helped the reader
in this regard.
References
[1] H.P. Bloch and A.R. Budris, Pump Users
Handbook, 2nd Edition, Fairmont Press,
Lilburn, GA, USA, ISBN 0-88173-517-5,
(2006).
[2] H.P. Bloch and D. Johnson, Downtime
Prompts Upgrading of Centrifugal Pumps,
Chemical Engineering, November 25, (1985).
[3] H.P. Bloch and F.K. Geitner, Machinery
Failure Analysis and Troubleshooting, 4thEdition, Butterworth-Heinemann
Publishing, Stoneham, MA, USA, ISBN
978-0123-860453, (2012).
[4] P. Barringer, API Pump Curve Practices
and Effects on Pump Life from Variability
About BEP, Weibull Analysis Course [avail-
able from: www.barringer1.com] (see also
Ref. 2, p. 621).
[5] H.P. Bloch, Pump Wisdom, John Wiley &
Sons, Hoboken, NJ, USA, ISBN 978-1-118-
04123-9, (2011).
"Repeat pump failures are an indicationthat the root cause of a problem has not
been found..."
This article is excerpted, by permission,
from Heinz P. Blochs textbook Pump
Wisdom, published by John Wiley &
Sons, ISBN 978-1-118-04123-9 (2011).
7/31/2019 Pump Failure Article
6/9
Feature4040
www.worldpumps.com
WORLD PUMPS October 2012
Any defect in the rotating assembly
of a pump (particularly when a
component becomes loose or
disconnected) could result in an excessive
unbalance and high vibration levels, which
in turn can lead to an emergency trip. In
case of any serious malfunction, the expec-
tation is that the pump will come to a safe
stop (through an emergency trip) with the
minimum possible damage.
Pump failure mechanisms
Rotating pump components (such asimpellers or blade-rows) may exhibit reso-
nance with any excitations generated by
the pump package. Resonances with the
first and second natural frequencies can
be particularly dangerous. Generally, there
could be numerous cases of resonance in
a given pump system; some of them
could even be unexpected. For example,
in one pump, the second natural
frequency of a rotating assembly proved
to be almost exactly an integer multiple
of the first natural frequency, which
resulted in excitation and some damage.
Liquid-induced vibration, oscillatory
changes in liquid pressure and turbulent
flow can be causes of high vibration levels
or even a failure in some pumps. The stress
amplitudes should be analysed. The
stresses should clearly be within accept-
able limits, for instance, fatigue limits such
as a high cycle fatigue (HCF) failure or a
low cycle fatigue (LCF) failure. A form of
fatigue is usually involved in most rotating
part failures and such failures can only take
place after a number of operating hours. In
other words, a few hours of a pump shop
performance test (in the manufacturers
shop) usually cannot show signs of a
future fatigue failure.
practice, some pump rotor assembly
designs are unbalance-insensitive while
other designs are sensitive to any unbal-
ance. Generally, an accurate rotordynamics
study can identify the sensitivity of the
rotating assembly to unbalance. In the
The sudden failure of a pump component can have serious and
costly consequences. Amin Almasi considers the failure modes in
a pumps rotating assembly, including resonance issues, fatiguefailures, vibration, the chain of events and secondary damage to
critical auxiliaries. The discussion is illustrated with case studies.
Practical notes oncomponent failure
Operating
0262 1762/12 2012 Elsevier Ltd. All rights reserved
Figure 1. An example of a damaged pump drive shaft.
7/31/2019 Pump Failure Article
7/9
Feature 41
www.worldpumps.com
WORLD PUMPS October 2012
As a result of high vibration levels or
extreme dynamic forces, some auxiliary
piping may be damaged. The most critical
such element is the lubrication oil piping
system. Any serious damage to the piping
feeding the lubrication oil to the bearings
(particularly to the hydrodynamic bearings
in large pumps) could lead to loss of the
oil flow and destruction of the Babbitt
metal in the journal bearing liner in avery short time (say around 13 seconds).
A short time of operation (sometimes as
little as 38 seconds) without any lubrica-
tion oil for some bearings can result in
their total destruction and other serious
damage (even explosion or fire). Failure of
a rotating part inside a pumps electric
motor driver could result in a serious
short-circuit incident and probably an
explosion.
In the case of an explosion or a serious
incident, the pump (or motor) casing
should fulfil the function of containment.
Structural damage to auxiliaries (particu-
larly critical elements such as the lubrica-
tion oil system) is also of concern, and
the design should be such as to ensure
that the auxiliary piping, flammable liquid/
hydrocarbon systems and electrical facili-
ties will safely sustain any fault conditions.
Pump shaft failure
The major reasons for pump shaft failure
can be broken down into the following
categories:
Mechanical: such as overhung load,1.
bending load, torsional load and axial
load.
Dynamic: vibration, cyclic, shock, etc.2.
Residual: manufacturing processes,3.
repair processes, etc.
Thermal: temperature gradients, rotor4.
bowing, etc.
Environmental: corrosion, moisture,5.
erosion, wear, cavitation and similar.
Before the cause of a shaft failure can be
accurately determined, it is necessary to
understand clearly the loading and stress
acting on the shaft. The ability to properly
characterize the microstructure and the
surface topology of a failed shaft are also
critical steps in analysing a failure. The most
common tools available to do this are:
visual inspection; optical microscopy; scan-ning electron microscopy; transmission elec-
tron microscopy; and metallurgical analysis.
With experience and a fundamental knowl-
edge of shaft failure causes, a significant
number of failures can be diagnosed by
visual inspection. Confirmation may then be
sought from a metallurgical laboratory.
Table 1 gives an approximate breakdown
(rule of thumb) of the main causes of
pump shaft failures. Cavitation could be
considered as an impeller/casing failure
mechanism. There are other studies thatsuggest that fatigue-related failures are
more important for a pump shaft (>40%).
The failure mode is also dependent on the
pump type and the pump service/
application.
Table 1. Main causes of pump shaft failure
Cause of shaft failure Percentage (%)
Corrosion (various) 36
Fatigue (various) 31
Brittle fracture 16
Overload 11
Creep, wear, abrasion, erosion, etc. 6
"Before the cause of a shaft failure can beaccurately determined, it is necessary tounderstand clearly the loading and stressacting on the shaft."
7/31/2019 Pump Failure Article
8/9
Feature4242
www.worldpumps.com
WORLD PUMPS October 2012
Residual stresses and fretting
Residual stresses or initial deflections are
usually independent of external loading
on the shaft. A wide variety of manufac-
turing or repair processes can affect the
amount of residual stress or initial deflec-tion including: drawing, bending,
straightening, machining, grinding,
surface rolling, shot blasting and
polishing. All of these operations can
produce residual stresses and initial
deflection by plastic deformation. In
addition, thermal processes that can
introduce residual stress and deflection
include: hot rolling, welding, torch
cutting and heat treatment.
Shaft fretting can cause serious damage
to the shaft and mating part. Typicallocations are points on the shaft where a
press or slip fit exists. The presence of
reddish-brown ferric oxide (rust) between
the mating surfaces is a strong indication
that fretting has occurred. The cause of
this condition is some degree of move-
ment between the two mating parts and
the presence of oxygen. Once fretting
occurs, the shaft may be very sensitive
to fatigue cracking (which could result in
a fatigue failure). Shaft vibration can
worsen this situation.
Case studies
Axial pump blade
This first case study concerns damage to
one of the blades of a large axial pump,
which resulted in extremely high vibra-
tions, damage to the lubrication feed line
piping and serious damage to the bearings
(extensive and costly consequences). The
blade in question failed because of reso-
nance and LCF failure, additionally
damaging neighbouring blades. The blade
loss incident led to an unusually large
unbalance of the pump, causing it to trip
as a result (a high-vibration emergency
trip). The pump rotor came to a stop, and
the pump casing fulfilled its containment
function perfectly. Because of the high
vibration level, the lubrication feed line to
the bearings was also damaged, which led
to the lubrication oil being cut off to the
sensitive hydrodynamic bearings during
the emergency shutdown period. All bear-
ings and many associated systems/compo-
nents were extensively damaged.
The lesson learned here was that the lubri-
cation oil piping (and any critical auxiliary
piping) should be designed to be robust
it is located in the area where the highest
shaft loading occurs. Fatigue cracks
usually start in the fillets or roots of the
keyway. A keyway that ends with sharp
step(s) has a higher level of stress concen-
tration than a keyway that uses a sled-runner type design. In the case of heavy
shaft loading, cracks frequently emanate
from this sharp step. A connection using
any form of key should be avoided to the
maximum extent possible. In special
cases, when other non-key solutions
cannot be used, it is important to obtain
an adequate radius on the edges of the
keyway.
Fatigue failures usually follow the weak-
link theory. The shaft fatigue failure
process usually consists of the following:
The fatigue leads to an initial crack on
the surface of the shaft.
The crack or cracks propagate until the
remaining shaft cross-section is too
weak to carry the load.
Finally, a sudden fracture of the
remaining area occurs.
Crack formation and fatigue
The origin of cracks caused by fatigue is
usually the presence of surface disconti-
nuities, which are commonly referred to
as stress raisers. Examples of this on
pump shafts are keyways, steps, shoulders,collars, threads, holes, snap ring grooves,
shaft damage or flaws that would
produce a stress raiser (wherever there is
a surface discontinuity a stress raiser will
exist; Figure 3). Corrosion can also create
stress raisers. For typical pumps, the two
most common problematic areas are at
the shoulder on the bearing or in the
coupling keyway region.
In the case of fatigue caused by axial
loads, the thrust bearing carrying the axial
load would most often show fatigue
(often contact fatigue) before the shaft.
However, there are numerous examples
where the shaft is damaged before thepump is stopped.
Keyways are commonly used to secure
rotating components, rotor cores and
couplings to the shaft. The keyway on the
take-off end or drive/driven end of the
shaft is the one of most concern because
Figure 2. A degraded pump after some years of operation.
"[For custom-engineered equipment]accurate modelling techniques and proper
analysis methods should be employed forall parts and components."
7/31/2019 Pump Failure Article
9/9
Feature 43
www worldpumps com
WORLD PUMPS October 2012
Fit and ForgetThe BoWex GT offers the following benefits:
3 x higher misalignments capability compared to market standard
easy assembly
maintenace free operationsAs a service KTR will undertake a Risk Analysis following the European
Machine Guideline.
www.ktr.com
phenomenon responsible for this fracture
was fatigue.
The clearance between the impeller and the
static part was relatively tight, which is often
seen in non-API small pumps. As a result of
operational forces (unbalance, liquid, thermaland dynamic forces), rotor rubbing occurred,
which caused extremely high bending
moments and stresses on the shaft. Proper
condition monitoring was not provided for
this small pump and the rubbing was unde-
tected. The main cause of fracture was the
rubbing of the rotor assembly. This then led
to an excessive fatigue cycle, resulting in
fatigue failure of the shaft.
enough to be able to feed vital lubrication
oil to the bearings at all times of emer-
gency (such as extreme conditions of high
vibration, extraordinary forces, etc.) to bring
the pump to a safe stop.
Motor fan
The second study relates to a catastrophic
failure in the electric motor driver of a
pump. An axial fan with 13 blades was
used as part of the electric motor cooling
system to dissipate the generated heat.
Fracture of the cooling fan blades occurred
just 260 hours after start-up. The fracture
caused a serious short-circuit incident
between the rotor and stator and conse-
quently an internal explosion in the motor
with extensive damage. An accurate finite
element analysis (FEA) modelling and thor-ough investigations showed that the first
natural frequency of the fan blade at ~659
Hz (accurately calculated after the failure)
was very close (almost within a 1.5%
margin) to the frequency of the exciting
force caused by the shaft rotation (13
blades 50 Hz = 650 Hz). This resonance
was the initial cause of the failure.
Further investigations indicated that when
the fan blades were excited by the 650 Hz
frequency, the blades vibrated intensely
and the stresses exceeded 300 MPa, whichcould be suffi cient for a fatigue failure. This
resonance and associated strong vibration
led to crack initiation/propagation, blade
failure, the short-circuit incident and the
electric motor internal explosion. The
explosion was contained, but the damage
to the electric motor was extensive and
costly. The lesson learned: a relatively large
(or medium-size) electric motor driver for a
pump will most probably be custom-engi-
neered equipment, and accurate modelling
techniques and proper analysis methodsshould be employed for all parts and
components.
Rotor failure
In this third case study, a rotor failure
occurred in a low-speed, small, manufac-
turer-standard pump after only approxi-
mately 2,000 hours of operation due to a
shaft fracture. The fracture was exactly at
a diameter-change section (a step) of the
shaft. The fracture surface was smooth
and perpendicular to the shaft axis.When the crack propagated to around
60% of the shaft diameter, the remaining
shaft section could not tolerate the
applied stress and a fast fracture
occurred. Experiments showed that the
Contact
Amin AlmasiLead rotating equipment engineer
WorleyParsons Services Pty Ltd
Level 10, 151 Roma Street (East Tower)
Brisbane, QLD 4000, Australia
Tel: +61 7 3319 3902
Email: [email protected]
Figure 3. An example of a pump shaft with obvious surface discontinuities (stress raisers).