A method of improving mechanical seal reliability
TJ DAVID, BSc, CEng, FiMechE
Floxibox Linlled, Manchester, UK.
sores,
Mechanical seals cane into O11 Refinery service some 40 years ago replacing packed glands on
Centeifugal pumps. O12 Companies have tended to move with technology on purchases of now equipment
Dut have show a aarked reluctance to up-grade existing equipment.
Sone Refinery's Process Units have long operating tives, seals supplied in the 1950°s and 60°s are
Stila in service with wear Componants being replaced to the original material spect fication:
For the past two years we have carried out a munber of Reliobility Audite for Customers, with the aim
Of ineredsing "aechanical seal life giving a resultant recuction in seiatenance costs. and improving
Fe1tability and safety.
‘These surveys nave shown vast ditterences petween the soa) 1218 for similar piancs, dependent on tne
Gogree of up-arading end correction of problem seals. "the majority of the ep-arading involves only
‘the replacement of ut dated face saterials within the existing seal at the next failure, with Little
of no cost impact. Only approrinately 108 of seals on the worst sites need complete repiacenent.
‘Te result of up-grading could provide aaintenance savings of hundreds of thousands of pounds per
‘annum fo the industry.
2.0 mrRooucTIoN
Prior to writing this paper (and st is 7 years ‘ais paper is based on my experience as a
fines my last contribution), I'took tine ou £0 Radsnery Ehginecr and the last 2 years as a
Zead” some tecent. papers” on Mechanical Seals. Plant Reliability Enginear with a sajor seal
iat struck me moet’ was sone of the conclusions company, where I” have had the opporcunity to
rg, "Seal" technology has progressed at a review "the perforeance and advise cn up-grading
Phatononal rate" and "eal Lives have increase:
‘Srenetically in recent years".
af 'a'nunber of locations.
During the period of 20 years working in
Seats have been in use for 40 yeare, about the Oil Industry at a Machinery Engineer, ay
‘the sage” tine scale a6 the gas turbine, unlike working Life seened to revolve around keeping
‘enn early ges turbines ate now museun the product in’ the punps. There. were little
seo the process as 2 slew evolution, excursions and” flurries of excitenent” when one
lon ew Savelopaant sowing us forward. mut of the Major machines. such as 9 covoressor oF
fo\tingie change hes produced the quantum leap i turbine, "Alccupped but’ the baseload was. always
Serizes. the universal panacea to'seal failures there “Mechanical Seal Leakage". It seemed to
{a'atiti to be discovered. be classed. ae an inevitable event and’ not the
failure of “engineers to produce a iong 1if6
Undoubtedly todays products have «mechanical component fora centrifugal pomp.
significantly. iaproved and thts paper aime to
Lifysteate the opportunities thet exist” to seal tatiuce analysis nas not nad the
improve the performance of seals on “existing glamour of Vibration signature. Analyste and has
plants. Bot had the high profile, but, ineidiously, seal
fepaies eat away “at the Maintenance Sudget’ and
fre the major cause of pump fa:lureNechanical seale have now be
tase for over 10 years and the life
the early seals" was low by today's standards,
Dot compared fevourably with the packed glands
they replaces.
‘Today, seal purchasers expect & seal to
last a umber of years when ft aitisately
failes st must do "so in a safe manner” with the
loss bf toxic. or inflammable prodect minintsed
ost O11 Companies now have very stringent
standards on ail new equipment but seldom are
these’ Standards nade” cetroupective to. existing
Sith’ the nexr “the” opportunities for, improved
Performance "are slow fo. be taken up. Tass
Shortie 1s costing the Oil Industry in the UK
hundreds of thousands of pounds per, annus
Reintenance costs on mechanical seals at one Ux
Refinery size runs at over £1,000,000 per year
All data used in this report coltectea
acing the last to years, relates to” our avn
products, the sites involved had & range of
Peon 35% to S04 of them. I do not believe this
found the overall performance of the major Seal
EGnpanies involved is very esmilar.
2. SEAL LIFE EXPEcrANCY
seals? Anyone invoivea' in the subject. will
Gonfire lifespan to be very variable, dependent
on the sealing duty, "seal design’ selected.
Guality of installation ané process operation.
nore the seat design saeaity meets tne
application, the seai wiil last the Life of the
equipment. a¢ 20 years or nore. whereas on sone
Spplications ‘mare’ an incompatible design for
the duty tas buen installed, Iife cen be counted
in aaye
the peime factor is cost, requiring to
niniaise’ “the” initial expenditore against
neenance and operating costs, but also giving
Consideration to reliability and. the safety
Sequiresent. A. simple low cost. seal protected
from the product punped by injecting an external
Clean cool flugh’ can give. an acceptably. Long
iife, but at a high cost for the down grading
and ices, of injected product. Setter to nave
paid more for a see! compatible with the product
Pumped.
ig ts aitticuis so give | mininun
Service but I. believe any seal, that nas. an
Gvorage Life of less than one gear, car be snoun
tov be economically” unacceptable and should be
Changed out for 2 Higher performance seal
‘you aay consider this figure to be low but
1 nave yet to finda Uk Refinery which does not
have a tunper of” seals which have consistently
Tess than one Year, and. T have found
‘been repaired four oF
A reasonable target for new or up-graded
als anovid bea minimus of 3 years. life, but
Enis’ i not econoatcally viable in every case,
Technically it it achievable but tho sealing
Configuration would be at too high 3 cost.
2
in evaluating seal_ performance,
collective Site or Plant Indices are the best
tool for comperison. The one I” use is Mean
‘ine Between” Fatiure’ (cost date has
Ristorieally not been available }, te.
Nean Time Between Failure
‘otal Nusber of Punos
Total No, of Failures in Honth
= Mean Tina Between Failure (Henths)
A moving average of 3 to 6 months should
bee's ‘multiplying fector of typically 0.5 to
0:75, the rationale being that alt the pumps are
not" running. together, ‘commonly process units
having an 'A’ and 'B' pump for each service with
only one running at any tise. “It” is generally
Sgrted that miltiple “sterte cet into’ component
Tite" sacter" than continuous running, 30" the
logic of the running factor 18 not clear
Using this index, what is the present
achievable’ ¥.1.B.. for @ “Refinery Site? From
fhe data availabie, the top quartile re “now
Greeeding an NTBF of 6 Years (no sunning factor]
Gna this “sovid ‘be achievable st” any. site
Mere are also locations claiming 8 years. Ik
Teva moving target upwards!
AL the other end of the scale, there are
etiil major UK siter with sn MTBF of loss. than
fo yeara. and the Dest audited UK site to-date
B57 Wyityears, "A “rar “east” hetsnery wnicn 1
Fecently visited has an MIBF of close to? years
whereas an East European site reportea less than
one year.
An too often, maintenance is carried out
ina vacuun and the publication of MTBF data
should” allow companies to compare their
performance with their peere and i hope to
Sncourage thie.
3. costs
‘Te repair of mechanical sesle tnatatiea in
Centrifugal puspe forms the major part of pump
Guintenance "corte in the oll end. many other
Process industries." Analysi¢ of data’ at a
Rusber of locations hae shown that 50¢ to 708 of
posp aeintenance work is" initiated by the
Failure of @ mechanical seal.
The cost of replacing the vorn components
of a mechanical seal ia in general three t0 our
Ginas'the ‘cost of the seal comoonents, aoe t2
the"“wrk “involved and” the "replacement, of
secondary itaag such as bearings. A” figure
quoted as @ typical cost for a seal repair of &
fetion size” beck pull-out pump ls £1,200 wnica
Could in "Sone instences be the sane "as a” new
fate of the ert’ seal. with 800 to 2,000
pumps on a Refinery coaplez, the return’ on
{nvestaent’ for improving seal iif ie seit
evidentTue capital costs of a Failure Reduction
Programe ‘are relatively low, "as only” a snl
perentege of the installed geal will require
Gomplete replacement. When it" is possible to
Upgrade the existing Seal re-using the non
Nearing components of the “seal, then the
“ieferentisl costs are very lov. These cases
Seualiy involve the up-grading of the herd face
‘one argusent against change which has been
used fe" "What about the old. faces have
refurbished end I noid in stock?". In general
Ht ie not cost effective to install these asthe
Inpravenant in seal life with the new seal face
gives an overall reduction in cost’
me rationalisation of components also
provides a return on investment, there are cases
Shere usere have’ five rotating seal rings with
Gttferent face, materiel Dut of the “sane oize
held in stores. spanning the evolutionary history
of mechanical seals, all could be replaced with
today's standard item.
A typical Refinery will old over £500,000
of mechanical seal" spares and the cost’ of
‘Ouership ia rually given as 20% per annum.
Improvenents in seai_reliapiiity also
reflects in the spares maximun/ainimin stocking
Taveia based on sage, so” that inprovenent in
Seal life further reduces spares levels
‘Tee aim of an Audit is to provide a plant with
Performance appraisal of the mechanical seals
appropriate, ‘recomendations for’ the up-grading
of the seals fo today's. achievable standards of
Performance, Up-grading should be based on an
‘ind’ improved reliability and safety,
An essential pre-requisite of an audit is
documented failure’ date, I have found that
enory 1s poor over the tine acale involved, we
Stly fenenber the things that ‘ceuse us. pain. “T
have “asked” nusarous Engineera over the years
‘inet are your probleme?* and never yot received
ago answer. "A seal that fails once every two
gears but” requires immediate attention for
Process reasons will ‘be Aighlighted, but the
Stel ne “sends bie technician to” repair every.
Ensee months as a routine work item ie ignored.
ALL Refineries 1 have visited nave Kept
fatiure data to various degrees of accuracy, the
Efend to" computerised job control, is. providing
an improved data bank. Sone Refineries have
Jory comprenenaive computerised failure record
Gystens but sil. tend to be. Gamping grounds for
Gata with no feed-back programmes to implement a
Failure Reduction Programe.
‘Te same coment applies to cost data,
with the increased use of computers this is now
Deconing "avaiianie. “but needs suitable
Progreaming “to present the data in a usable
ing Refinery 1 know has a
computerised” record, ayaten which provides
fallure data with the details of each repair
in addition, each failure wes analysed by a
Wachinezy Engineer and” reported even if the
Conclusion “war "No changes requires, wear out
with acceptable “iife'. The resulta of thelr
Failure Reduction Programe (Refinery "A" is
Given in Figure 1. the MBF is for’ all pomp
Yatlures not only mechanical. seals, but it Goes
denonstrate the scope for Inprovesent and that a
Yong term commftment le required. The line for
Refineries 'B' and. 'C’ are for plants oudited
tea is the MBF for mechanical esis only. The
faprovenent of Refinery 'C’ are primariiy the
reauive of a hard face up-grading From stoliive
Returning to what canbe achieved with
existing. failure data, witha large pump
Population a. ressonably accurate” MTBF can be
Sbtained with only one month's data. A nore
Securate MTBF Sa achieved by averaging over a
three or six month period. This will indicate
the overall performance of the site of unit and
wila"sllow comparison with (t9 peers.
Moan tine between ¢ailuee range
under 2 years Very poor - will require
extensive up-arading
2 to 4 years Poor - analysis data to decide on
4 to 6 years A good base - scope for
improvenent
Still provide improvenent
The MTBF allows evaluation of the site
performence But “does aot provide you with the
Gata to initiate change. For this You require a
Hinioun ‘of one year's. failure data, aagitional
Years improve sccurscy but account should be
Toren of any geal modifications nade in the
audit perioa.
From this failere data you can plot, +
faiiureagainat ¥ pup population (Figure 2),
the form “of this curve establishes the scope of
work. A steeply” fising curve indicates « small
Pereantage of problen seals whereas a 45 degree
total enape of the “curve 4a infivenced by the
unber of years of “data, but normally Aas a
‘Aininal effect on the “slope up to the 10% punp
population. point, the "aren ‘Gt masn” anceresr’
Far curves shown’ in’ Figure 2. ere for” sctual
audits, and show the range of cata collected to-
Gate 208 to 708” failures associated with 108 of
‘the’ pump population.
In deciding the course of action tobe taxen,
the “failure ‘curve must "be coneidered 1a
Conjunction with the MTBF.
‘The following hypothetical examples illustrate
the appropriate approach fortwo different
scenarios.