DESIGN AND OPERATION OF
PUMPS FOR HOT STANDBY SERVIC!
by
Charles C, Heald
Chief Engineer/Produet Engineering Manager
Ingersoll-Rand Company
Engineered Purmp Division
Phillipshurg, New Tersey
and
Da
G. Penry
Manager, Engineering,
Ingersoll-Rand Company, Limited
Gateshead, England
Charles C. Heald is Chief Engineer
and Product Engineering Manager for
Process and Industrial Pumps, Engi-
neored Pump Division, Ingersoll-Rand
Company, Phillipsburg, New Jersey. He
4s responsible for design, development,
product engineering and contrects
Mr. Heald received « B.S. degree in
Mechanical Engineering from the Univer.
sity of Maine in 1960. Since that time, he
hhas held various positions with Ingersoll
Rand, including Development Engineer, Senior Hydraulic De-
sign Engineer, Manoger of Pump Design and Chief Engineer.
He is a member of ASME and the API Centrifugal Pump Man-
anfacturers Subcommittee, and has recently serced on the API
(610 7th Edition Task Force. He has authored several articles and
technical publications on centrifugal pumps.
David G. Penrys Product Manager for
Process and Industrial Pumps, Ingersoll
Rand Company Limited, Gateshead,
«England, UK. In this function, he main
tains close contact with Ingersoll-Rand
Pump Group headquarters in the United
States in the design and application of
puraps for the European served area
rior to joining Ingersoll-Rand in 973,
Mr. Penry wasa Seetion Engineer for AEI
Company, Limited, where he was directly
concerned with rotating equipment, particularly CO2 circula-
tors, for gas-cooled nuclear reactors. He subsequently spent
three years atecturer in thermodynamics and fluid mechanics
Defore joining C.A. Parsons Company, Limited, as « Group En
gineer in the Compressor Department, where he was Fesponst-
Bie forthe design of «helium cteuator for the gas-coaled mu
Mr. Fenry graduated witha B.Se. degree in Mechanical Eng
neering in 1853 and seas elected to membership inthe Institute of
Mechanical Engineers in 1965. While at C. A. Parsons Company,
Limited, he presented a paper at an I. Mech, E. Conference en-
titled "Cireilator Design for High lemperature Reactors,”
ABSTRACT
Pump thermal distortion problems routinely arise in refinery
applications where hot fluids mast be handled. This is partieu-
lavly the ease when warming up standby pumps. The single
stage, double suetion configuration which is best suited for sich
services, is analyzed in terms of the mechanical interferences
that arise from thermal distortion. This js done withthe aid of
finite element modelling, whieh eot only led to the solution of
the interference problems via design improvements, but also
led to recommendations for application and operation of pumps
tunder hot-fluid cosditions
APPL{CABILITY OF SINGLE STAGE
DOUBLE SUCTION PUMPS
In a variety of eefinery pumplog services, high flowrate and
developed head requirements dictate the type of pump re
aired. In addition, these services can also have high pumping
temperatures (above 400°F/200°C) and liquid spelie graves
below 0.70.
‘ap design alternatives for these applications include the
following I}
+ single stage, double suction, vertically spit
+ singe stage, overhung, vertically split
+ culistage vertically spit (bare)
+ two stage vetialy or horizontally spt (single case
Advantages of thesinglestage, double uetion, vertically split
design (Figore I} over the other alternatives listed makes ithe
fbvious selection
+A double suction design equites Tes suction pressure than
a single suction design (typically at least 30 percent less) atthe
Same speed. This aso meas that sinaler, higher speed putnp
car be selected, which may also cost less and be more ecient
than a siege scbion exit
4 single stage unitcan more easly be arranged with radial
(circular) tain easing joint with a confined, controlled compres-
sion gasket to,more positively contain hazardous Gammmable,
toxic) fighids than a multistage unt
* A double suction symmetrical design can be constructed
with a heavy shaft extending through the impeller eves (inlets)
kun supported with beariggsat both ends ofthe shaft (an API
10 requirement for double suetion Impellets). This redvces
Shaft deflection at mechanical seal faces and weassvgs al ve-no PROCEEDINGS OF THE FIFTH INTERNATIONAL PUMP USERS SYMPOSIUM
Figure I. Typical single Stage, Double Suction, Vertically Split
Process Pump Design
sults ina stable rotor design that operates well below first lateral
critical speed levels.
+ The nominal axial balance ofthe single stage double suetion
design minimizes thrust bearing loads when suction pressures
fare high. Overhung pump desigis must accommodate suetion
[pressure thrust because ofthe single stuffing bovseal chamber
‘design. Multistage pumps are usually arranged to balance the
pressure on each seal chamber, but this s usually accomplished
In conjunction with close fiting throttling devices, such that
axial balance can change (significautls} with normal pump wear
+ Suction and discharge nozales can usualy be arranged for
either top or side locations, depending on which is more conve
nient for the specific application, Usually top suction and dis-
charge arrangements are preferred for hot services with over.
hhead piping to improve piping arrangements and provide more
resistance to nozzle forces and moments created by thermal
srowth.
+ A double suction impeller can have # lower inlet tip speed
than a single suction impeller. Since refinery Muids may contain
suspended solids, this means impellers with lower tip speed will
tend to suffer less abrasive erosio
‘Thereare, ofcourse, some disadvantages tothe single stage dow
De suction desiga as well
‘A two bearing design requires two mechanical seals
whereas a singlestage overhung design requires on
this means that for low flows, and where suetion p
adequate for NPSH requirements, the overhung design may be
preferred
+ Thehigherenersy-per-stage ina single stage doublesuction
‘pump may dictate higher minimuin Noss than a multistage unt.
Also, for higher head applications, the specific speed of a multi-
stage pump may be more nearly optimum, and result in higher
operating eficieney. Units of two or more stages, however, are
move susceptible to interual alignment problems when high
‘operating temperatures impose severe thermal forces and tem
perature gradients throughout the pamp,
* Double suction pumps-whether single stage or the frst
stage of multistage pumps-require symmetrical low tothe s
tion nozzle. If piping leading to the pump creates unequal flow
to-each impeller eye (inlet), an axial unbalance ean oceur, lead-
ing to high thrust bearing loads. This also results in unequal
flows within the impeller, which may cause one side of the im-
peller to cavitate. Unequal flows can als result in rotor oscila-
tons, which eause nasteady bearing and seal loads —and prema
ture fedares, Suction piping arrangements are less critical on
single suction pumps.
This presentation will outline experiences with large single
stage double suction sefinery pumps applied ta severe hot bot
touts service. It will discuss the application and operational
problems encountered and how they ae being adesed and
‘The pumps described inthis application are single stage, dou
bie suction, vertically split with 220 in discharge, 16 in suction
and 31 in noming! anpeller diameter. The pumps are rated for
5,000 US gpm, 2954, operating at 1780 rpm, The service is main
fractionator bottoms, 640°F (338°C) with a specific gravity of
0.552 at pumping temperature. The pump was designed for 720,
psi working pressure at the pumping temperature, Material of
‘construction i carbon steel (ASMT-A2I6 WCC). The pump is
equipped with sleeve radial bearings and ball thrust bearings
(Duplex, back-to-back) with ol ring lubrication and purge mist
The pump was built in accordance with API-610, 5th Fdition
‘The mechanical seals were specified tobe of single, balanced ear
tridge design, with esternal evo oi sh,
‘This pump represents ane ofthe lagest single stage machines
applied to this type of service anywhere in the world, Not only
's the pumping temperature high, but the ambient temperate
(Middle East is also very high
INSTALLATION AND OPERATING PROBLEMS
‘The operational problems diseussed herein will concentrate
‘on those associated with high temperatures
Each pumping wnit was arranged (as is typical) with two
pamps—one operational and ane spare (hot standby). Suetion
and discharge piping Was oriented at the top of the pump (top.
nozzles) with hangers and supports to reduce nozzle loads. The
suction pipe, at about 10 pipe diameters from the pump nozele
(ange) tured 90 degrees and ran parallel to axis of the pump
shaft. In this particular instance, the distance of the 90 degree
turn from the suction nozzle was probably suiicient to at least
rninimize distortion ofthe flow into the suction nozzle~and thus
tothe impellerinlets. However, normally, itis critical that atten-
tion be given to suction piping (and flow distebution) on double
suction pumps (impellers) to ensure that flow is equal to both
sides ofthe impeller to prevent poor suction performance,
totor unbalance and axial rotor shutling-especially at flow less
than design flow.
‘The mechanical performance of the first unit, which was
started cold and brought up to design temperature slowly with
the process, was satisfactory. Major probleuns arose, however,
when the “hot” standby pump rotor would not gen by hand
without experiencing internal rabbing (hard contact) after ap-
proximately 180 degrees of twin. After an elapsed period of
abi ur minutes, the rotor could undertake a futher rotation
(0 180 degrees before contact Was experienced again. Thus, hot
standby startup could not be achieved
Tavestigation revealed several possible factors contributing to
this inability to rotate the pump rotor freely on hot standby:
* The warmup flow to the pump case (Figure 2) cousisted of
4 singular pipe to the discharge piping of the standby wnit suel.
that uniform distribution of the warmup liquid through the
pomp easing and around the rotor was not being achieved. In
view of the large physical size of the pump and the thermal
strabifeation eects, the location of the Warmup line, andl prob-
ably the quantity ofwarm-up liquid aswell, was inadequate [2].
* Coupling alignment was changed significantly from the
“coldta the "hot standby” condition (as muich a 0.020 in- 0.080
in) It was suspected that nozzle loads were moving the pump
relative to the drive. It was also suspected that the pump was
tuying to expand axially and transversely and was being re:
stricted, and that the shaft was moving independently of the
pump casing.DESIGN AND OFERAHION OF PUMPS FOR HOT STANDBY SERVICE m
DISCHARGE
ume,
SUCTION
Figure 2. Warmup Piping Showing Top Warmup Connection
INVESTIGATIONS
io evaluate the influence ofthe original warmup fxs to the
pump easing, a finite element model (Figure 3) of a portion of
the nearly'symmetrial pressure containment vessel was created
such that thermal gradients could be studied and distortions
{movements} could be evaluated relative to known measure
‘ments (Figure 4), When we imposed known deflections on the
madel, the total displaced geometry could be viewed and
studied.
Te vas found that the otiginal warmup low atvangement was
doing very little good, Ie appeared tobe short-crcuiting (Figure
5} and simply going in and out ofthe top of the pump ease. Tt
‘as, in fact, adding tothe distortion (Figure 6)
‘The configuration of the easing was also reviewed to deter
‘mine if changes in construction ceinforeing ribs. ete.) could add
significantly to strvctural rigidity and add resistance to thermal
distortion, Aftec reviewing the finite element model, this was
jected in favor of maintaining a sinoath casing profile. Thermal
oP
‘COUPLING,
END,
BOTTOM
Figure 3. Finite Element Model ofa Porton ofthe Nearly Sym-
‘metrical Pressure Containment Vessel
EXTERNAL ViEW
uNpispLaceo
‘GEOMETRY
pisPLaceo
GEome TRY
(EXAGGERATED)
Figure 4. Model Showing deflection resulting from Original
Warmup Flow into Top Connection
Figure 5. Top Warmup Flow Connection Besults in Short C
‘cult. Warmup Flow Simple Goes in and out of the Top of the
Fump Case, Resulting in Severe Thermal Gradients.
gradients in major reinforeing ribs could actually ereate higher
levels of distortion, and stress, ifnot precisely located and sized,
Originally the pump was dnwelled securely to its mounting
pedestals withont provision for expansion in any disection. This
's routinely done on smaller unis ofthis type to maintain pump
position on its bedplate and maintain alignment with the driver
With such a large pump, however, it appeared that “planned”
‘expansion, partculatly inthe radial direction, eould help elim-
nate distortion and shaft movement at the coupling
Because ofthe volatile nature of the pumpage, the single ba-
Janiced mechanical seals were arranged with cold gas oll Mush to
each seal chamber from an extemal source API-610 Seal Piping
lan 32/63) (Figure 5), This flush was injected into the seal ata
single point. Since the lush has to he on the ile standby pump
asall times (for safety purposes), it has the ellect of cooling the
shaft and seal chamlter area while the restafthe pump was being
heated, Actually, the casing was being heated nonuniformly
with the single point warmup flow in the discharge pipe and the
Shad aod seal chamber area were being cooled sonuniformly12 PROCEEDINGS OF THE PIFTH INTERNATIONAL PUMP USERS SYMPOSIUM
VERTICAL
UNbISPLAcED
‘eomerer
oor L,
1982 INCH
pispuacen.
(EEACCERATED)
Figure 6, Model Shows Gasing Displacements Resulting from
Top Warmup Flow Which Actually Added to Distortion.
‘vith the single point injection ofthe cold gas oil lush, Further
nonuniform cooling took place asthe spent cold gas oil stratified
inthe bottom ofthe pump casing adjacent to the seal chambers.
‘The net result was severe “hogging” of the shat, bowing ofthe
casing, and binding ofthe rotor.
CORRECTIVE ACTIONS
‘The finite element model was constructed such tat field mea-
surements of pump cising surface temperatuies top to bottom)
could be used as input. When this was done, resulting distortions
\were in close approximation to those values actually experi
enced on the hot standby pump. The distorted model with ac-
tual measurements of deflections imposed is shown in Figure 6.
The finite element analysis showed that multiple point warm
up injection would be required to uniformly heat the easing
With single point injection, the bearing housing brackets were
actually moving and twisting relative tothe close running clear
ance diameters in the easing, thus esasing rotor contact as well
asbeaving misalignment, By replacing single 34 inch injection
Figure Z.lmproved Warmup Flow Piping into the Bottom of
Each Casing Passage
point at the top ofthe pump with three one-inch connections on
the bottom of the easing fone in the volute and one in each of
the suction passages, adjacent to the seal chambers and bearing
brackets Figure 7), it was predicted thatthe flow of warm-up
liquid would be substantilly increased and would flow more
aniformly throughout the pump casing, resulting in uniform eas
ing growth (Figure 8). Infact, by this means, the original approx
imately 250°C differential across the easing was veduced to an
acceptable 25°C. The ways easing temperatures were distri
ibuted and inputted to nite element program are shown in Fig-
ture 9. Another displacement output to evaluate effect of 360
degree bearing bracket is shown in Figure 10. In arranging for
injection into the suction passages, it was determined that eon
veetive upward flow should be controlled so as not to impinge
‘on the pup shat. The concern sas that such impingement con-
tributed to additional bowing of the shaft.
Agnin studying casing thermal growth, itwas decided that the
pump, because ofits physical size and near eube-tike overall di
‘mensions, should not be dowelled, in order to avoid any possible
internal distortion. Instead, keys were fitted at the centerline of
Doth pump feet to allow free radial expansion and axial expan-
sion toward the drive (Figuve 11)
VERTICAL
—feanne
UNoisPLAceo
cEoMETRY
+
o201NcH
piseuaceo
GeoMerRY
(EXAGGERATED)
Figure 8. Model Shows Improvement in Casing Displacements
tuithInproced Bottom Warmup Connections
650
\ | jpeooe
500
500
1400:
500 bor
.600
Figure 9. Distribution of Casing TemperaturesDESIGN AND OPERATION OF PUMPS FOR HOT STANDBY SERVICE us.
VERTICAL
ARIAL
UNDISPLACED
SEOMETAY
939 INCH
DispLaceo.
GEOMETRY
(EXAGGERATED)
Figure 10. Effect of 960 Degree Bearing Support on Distortion.
Compare to Figure 8.
Figure 11 Keys Fitted o Allow Free Radial and Axial Expansion
tchile Mininizing Distortions Caused by Restricted Movement
In addition, a key block was wed to the lower centerline of
the pump easing to mate with a block on the baseplate. By pro-
viding these sliding features to allow the pamp to expand both
axially and transversely, alignment with the driver could be as-
sured while avoiding any restrictions whieh could add to the dis-
tortion problem.
The coupling alignment is maintained by the transverse key
fitted at the bottom ofthe pump. The keys were fitted with 0.005
in side clearance {maxintum) to limit angular movement. To
evaluate the effectiveness of corrective actions, the pump as
warmed up with the coupling spacer removed and sufficient
time allowed for the pumnp toreach stabilization. Pump to motor
alignment was determined with the pump in the hot condition
using a “three clock” method. Three dial indicators were at
tached tothe motor coupling hub with one indicator registering
‘pump coupling hub rim runout and tww indicators registering
‘pump coupling hub face runout. Two indicators were necessary
for the face measuteinent to compensate for pump shaft axial
rmovernent dusing tation, Indicator support arins Were suf
ciently rigid to avold sag over the Aste between the to
hhobs. Alignment (to less than 0.009 in) was achieved by adjust
ing the motor position and shimming under the motor feet. A
*oftfoot” check was performed on the motor after aligament.
Pump casing skin temperatures were obtained by the use of
ten thermgoouples attached statically tothe casing by spot
‘welding. Thermocouples were also attached to the warmup line
{0 monitor the warmup liquid temperature and tothe sea of
feed line to monitor the sel ol temperature. allthermocouples
were connected 10a chart recorder which reisteved each tem-
peratute every 60 seconds
“The rate offlow'to each warmup injection point was contelled
by avalve such hata warmup rte not exceeding Cimin and
4 maximum differential temperature at any. point in tae be
tween al pump thermocouples of 30°C could be achieved. Tt
was found thatthe maximum diferential objective was atan
ble, but is was not posible to maintain the 3*C/min vse at all
Poinis inthe ver inital stages of warmup. A steady sate tem:
peratuse was reached after approximately two hous of ntrodce
tog hot liguid. Ultimately, the warmup Tiquié contol valves
weve to be replaced by sized orifices
Shaft to casing relative movement could be monitored only
where the shaft extended from the coupling end bearing hous-
Jing, because of insuificient space between the seal glands and
Yiebearing housings. A single dial indicator was attached tothe
coupling end bearing housing Yo resister the telative vertical
‘movement of the shaft during warmup. Te was shown that this
indicator was a dependable guide tothe extent of bowing of the
Shaft at the coupling, the dtletion at that peint being approx
‘nately half that atthe coupling, Ths deflection couldbe ver
‘ducing shaft rotation alter warmup. A lint of. 005in was placed
on the acceptable shaft unoot or tatsip, knowing that the shaft
‘would fully straighten very shortly aftr startup.
“To minimize any possible etvor arising ftom the growth of
structures in close proximity tothe hot pump easing, al dial
Gicator supports were carefully insulated. Care was taken to en-
Sure that support structures themselves were sufficiently rigid
to withstand reasonable accidental knocks, The axial movement
‘ofthe pump suction line relative toa fxed poi was also mi
tored to establish its posible effect on the pump esing move-
snent during warmup, In adition tothe continuous monitoring
oftempratares al nestor eadigs wer recorded 1 in
“These medications resulted in a sigaticant improvement.
The shaft was now capable of being rotated continuously
through full 360 degrees. However, an unacceptable upward
Shaft bowing atthe coupling prevented automate startup from
the hot standby condition. Coupling malignment vias more
than was desived prior to startup. Further testing clearly ind
sated a telitionship to the old flash ol injetion. Shat move
ment a the coupling could be roodied significantly by altering
the rate of Bow of the cold fash oil injection, obviously aot at
acceptable situation.
Teas determined that the shaft bowing resulted fiom the
cold oil fish impinging onthe shaft overa restricted cicumer
fence due tothe single point sel injection (Fig 12). In con-
fonction with the seal manisactucer, the seals were modiied to
‘simplified multipoit injection. At the same time, an attempt
‘was rade to avoid dvect impingement of the ht warmup” lig
uid on the underside of the shat at the coupling end of the
pump. However, these modifications resulted in ony limited
To abtain further improvement and minimize the effects of
the cold fhish off smpinging on the shalt, arangements were
nade to preheat the as ofl to approsimately 100°C (212"F) ina4 PROGEEDINGS OF THE FIFTH INTERNATIONAL PUMP USBRS SYMPOSIUM
Figure 12, Bottom Warmup Connections in Each Pump Passage
Improves Distribution of Warmup Liquid and Results in More
Uniform Casing Growth
steam heater, By this means, bowing ofthe shaft was veduced
toan acceptable level such that the rotor did not bind and move~
‘ment atthe coupling was virtually eliminated. A suecessfl start-
‘up from the hot standby condition was made.
However, preheating of the gascil fs not considered an accept
able long term solution for many reasons, among which is the
{temperature ofthe shaft at entry to the beatings (high tempera
ture close to the babbitt material),
‘Other limitations have necessitated further changes. Experi=
ence has shown that diferent rates of warm up flow are desirable
toavoid excessive rates of temperature rise fom cold to that re-
{quired to hold the pump at a sufficiently high temperature to
avoid thermal shack on start up. The use of diferent orifice
plates in the warm up harness to control the rate and direetion
offlow would be desitable, but has been shown tobe impractical
inthis particular installation beeause ofthe danger of plugging
‘Along term solution has been being pursued based an the ex:
peviences gained to date (Figure 13). This will inefude the use
of double mechanical seals, thus limiting the quantity of cold
ffush, and containing itso that it eannot enter the pump suction
passages and eause thermal stratification and subsequent distor
tion. Tangential introduction ofthe hot warmup liquid (Pigure
1) will minimize the inpingement of hot liquid of the lower
Figure 13. Tongential Injection of Warmup Flows Prevented Im-
pingement on Shaft
Figure 14. Warmup Piping Showing Tangential Injection and
Temperature Control of Hot Flush Liquid.
portion ofthe shaft. A natural conveetion cooler may also be in-
troduced into the warmup line o provide eonttol of the pump
warmup rate from ambient conditions
CONCLUSION
thas been enmmon practice to warmup pumps by introdue-
ing hot ligatd into the discharge pipework and permitting the
flow fo petmeate through the pump bad tothe pump suetion.
In general, this has been perfectly satisfactory in achieving 60
vital conditions for immediate hot startup:
+ Controlling the warmup rate ofthe pump casing and hokd-
ing the pump at a sufficient high temperature to avotd thermal
shock on startup.
+ Holding an acceptable iffeentil temperature throughout
the pump body to avotd thermal distortion and rotor binding.
‘As pumps become larger, a simple warmup flow to the dis-
charge piping is not sufiient
‘Thermal gradients within the body ofa lage pump can over
come the natural liguid low reslting instant ateas within
the pump and extensive temperature differentials aross see-
tions of the pump casing. Such thermal differentials can easy
result in enough distortion to cause rotor binding (contact) and
‘The method of pump warmup must be given careful consider
ation fr each pump installation. There is no universal “standard”
solution. At present, i would appear that pump manufacturers
arerarely consulted. twould be advantageous to manufacturers
tnd users alike if more communication in this important area
took place atthe initial stages ofa contract
‘Many factors come into play in the decision to use a single or
double mechanical seal, but seldom is one of those factors the
consideration of thermal distortion of the pump rotor or easing,
Experience has shown, however, that even mechanical seal
selection can be important to satsfictory pump startup fom 2
hot standby condition
Both pumping temperature and pump size must be consi
dered when evaluating. warmup flow requirements for hot
standby pumps. Larger pamps at high pumping temperaturesDESIGN AND OPERATION OF FUMPS FOR HOT STANDBY SERVICE us
demand special atention to many ofthe procedures routinely
agate to males, lover ten pertie machines. Exch istalaton
serves individual consideration, and it iin the best interest
of both the user and the pomp manubicture to communicate on
warmup arrangements early enough to avoid unnecessary start
p problems:
REFERENCES
1. Stepanoll, A. J., Gontrifugal and Asial Flow Pumps, New
York: Jobn Wiley & Sons, Incorporated (957),
2. Nelson, W. E, "Maintenaace and Troubleshooting of Singhe
Stage Pumps Freceedings ofthe First Intermationl Pury
‘Symposium, Turbemachinery Laboratories, Department of
Mechaslal Engineering, Texas A&M University, College
Station, Texas (1984)
ACKNOWLEDGMENTS
‘The suthors wish to thank the management of the Ingersell-
Rand Pump Group for permission to publish these experiences
‘They also extenda special thanks to Mr. Ron Palgrave, Chief Ent
reer and Mr. Colin Ruddick, Engineering, Managet, tagersoll-
Rand Company, Limited, Gateshead, England, for. their
assistance.