VIBRATION MONITORING OF VERTICALLY MOUNTED PUMPS—
TOOLS FOR MANUFACTURERS AND USERS
by
‘Thomas J. Walter
Manager, Mechanical Engineering
‘Matthew M. Marchione
Project Engineer
Mechanical Technology; Incorporated
Shaker Research Divi
Latham, New York
and
Gordon Shugars
Project Manager
Electric Power Research Institute
Palo Alto, California
‘Thomas J. Walter is Manager of Mechan
ical Engineering for the Shaker Research
Division of Mechanical Technology, Incor-
porated (MTD. He joined MTP in 1977
‘and coordinates and maintains marketing,
financial, and technical sponsibility for
‘efforts inclving fied troubleshooting and
periodic vibration monitoring of machinery
‘gslems, including onsite acquisition, analy
Ss, and diagnosis of data to determine ma
chinery problems,
{In previous positions at MT1, Mr: Waller was Manager of various
‘engineering sections here efforts ined development and appli-
Cation of high speed balancing techniques, including automatic
removal of material using pulied lasers. As Proect Engineer, he
‘maintained project responsibility for contracts involving vibration
‘and balancing. Prior to joining MTF, he sas a test engineer at
Newport News Shipbuilding, where he was involved in nove and
vibration reduction efforts velating to US. Navy submarines and
surface ships.
‘Mr. Waller received his B.S. degree in Mechanical Engineering
from the University of Dayton in 1971 and an MBA. from the
College of William and Mary in 1976. He has authored many
technical reports and has published several papers for seminars and.
workshops.
Maithew M. Maychione is a Projet En.
agineer with the Shaker Research Dinision of
Mechanical Technology, Incorporated
(MT0). He is ingotved wih vibration based
‘analss and field troubleshooting of rola
ing machinery problems. He also develops
computer based rotordynamie models to as-
Sis 0 identifying root causes of such prob-
lems. His areas of involvement have in-
cluded technical and financial management
of several projets involving monitoring and
diagnosis of vertically mounted pumps
Prior (0 joining MTU in 1985, Mr. Marchione was a Field
Engineer with Vibration Specialty Corporation, where he performed.
%
vibration based diagnosis and troubleshooting of machinery. He also
performed predictive maintenance surveys for a wide variety of
tlietsthyoughont the country
‘Mr. Marchione obtained his B.S. degree in General Science form
‘illancoa University in 1980, He has attended many machinery
oriented seminars and has lectured om several topics.
H. Gordon Shugarsisa Project Manager
in the Plan Availity Progam, Nuclear
Poser Division, a the Elecrc Power Re
Seaveh Insite (EPRI), in Palo A,
Catfrnia.He joined he ltt n 1972
‘Ms Sugars responsible for projets dea.
ng wi equipment monoring nd dag
sates ond poser plant insttomendation
td conta
efor joing the nti, Mr: Sugars
sped thee year ih the Bachil Poser
Corporation. He was & Planning and Scheduling Engineer for
ower plnt design in Son Franca: eotracion i Beri
Pennslania; and ting in Cobyp, Montana. From 1969 to
1973, Mr Shugars wat a Projet Manager Nucear Engineer sth
HeUS. Nar S. Atomic Energy Conaiicion. He was responsible
for the design, caniration, teing, maintenance and operating
‘ricedues fr mice pore sibarines and surface Ships
Mr Shigars recived his BS. dere in Elected! Engineering
fom the University of Hints, Urbana-Chanpaign in 196%. He
esa Baching Asstt in the Department of Electrical Enginer-
ing athe Unters of Tinos ote aching hs MS. degree in
letra Engineering in 1974. Mr Shugars he thor of many
publications dealing sath poner plant sntramentation, and has
freinied and published seeral popers for EPRI seminars and
Soros. He's a negidered electra! engineerin the Sate of
California.76 PROCEEDINGS OF THE FOURTH INTERNATIONAL PUMP SYMPOSIUM
ABSTRACT.
Recent efforts to develop and implement techniques to
‘monitor and diagnose problems in submerged vertical shaft
pumps are summarized. Projects based on iterative analy
‘al and experimental efforts were implemented on several
different types of vertical pumps typically found in power
ants
MAlthough modelling aspects are briefly described, the
major focus is on the restlts of the experimental aspects,
including the “lessons learned” in instrumenting and
monitoring six vertically mounted pumps in operational
power plants.
INTRODUCTION
For many years, uiliies and other industries have
snonitored vibration to detect failures and to plan mainte:
tance for horizontal achinery, Inmany instanees, however,
this technique has not been succesfully applied to vertical
pumps for several reasons, including lack of experience in
Ineasuring vibration of submerged portions of the pumps
hind lack of understanding of how sh factors unbaanee
fand shatt alignment affect machinery health. In addition,
imathematically modelling vertical pumps often involves
{sumptions that differ from those involved in modeling
horizontal pumps thus making it ifieult to predict wea oF
damage from typical sibration meastrements. As‘ reslt,
lures often oveur without warn
‘Atempis to more thoroughly measure vibration on the
submerged portion of vertical pumps have often rested
lata that cannot be readily interpreted
When compared to horizontally mounted pumps, vertical
pumps can exhibit high vibration levels and yet contin to
Sperate satisfactorily. The pumpsotten operate wit signs ot
Aistable operation large misiligoments nd other chara
teristics et would dicate immediate shutdown in mo
thachinery. Many of these phenomena go sndetected at
‘raditional vibration sensor lations (ege on drive motor
Housings). Experiments have ko shown that significant
thanges to the mechanical properties of the pump (e.g
lange changesin (unbalance) donot avaysresultin expected
Changes in bration level, Experimental data, have also
Shown that some changes in the wbration level may be
associated with the pump hydraulic load {1}
“Toaldress these problems, projects have been undertak
en to develop analytical techniques to interpret vibration
Ieasurements, verify these techatiguesby in-plant test, and
develop recommendations for application of submergible
vibration sensors
“The projects combined both analytical and experimental
efforts Following an nial survey of available iterature and
inquires with puinp manufactarers and users regerding
modeling and vibration analysis of vertical pumps, three
{ypesof vertical pumps were selected for stacy ata grsctred
power plant on the Pacific Coast in northern Cahiforna
Eibsequenty additional circulating pumps have Been st
ied atan oibfired plant on the Headson River near Pough-
teepaic, New York
DISCUSSION
Project Methodology
Each of the selected pumps was studied under analytical
and experimental efforts through iterative and interactive
processes. In general, an intial mathematical moclel of the
pump was made, and a computer program was used to
predict natural frequencies and mode shapes. Thisinforma-
tion was used to selec the types and lotions of tensors.
Each’ pump wis hen opened and inspected, and’ precise
mmeaaurementsof fis and dlearances were made and ted to
Update the model. Following reinstallation, vibration and
performance parameters from the instrumented pimp
Mere continuously monkored. Resulting information was
tsed to tine the model to reflect actual operation. ‘The
tmodel was then used to predict the effect oft given change
ine pump.
All pumps were to be removed from service, inspected,
and alodified ie, changes made in unbalance, malign:
ment, hydraulic loading). ‘The pumps were then to be
feinwalled and run for several additional months andthe
esting test data used io une and simply the mel Test
data were used to refine the sensor types and. locations
needed to adequately monitor pump vibration
The inital intent was to implement on subsequent pump
types those modelling and sensor techniquet that were
Successful onthe fist pump. Availability ofthese pumpunits
for sty was limited however, because all pups are in
inthe planned schedule were necessary namely the need to
instrument and monitor several pumps, simultaneously
rather than ‘sequentially. Plant power requirements ad
tverhaul schedles also limited the number af pup remov-
Ste secomplihed
Analytical Efforts
A basicobjective was to evaluate the ability of rotordynam-
ic analysis to predict the vibration performance of vertically
mounted submerged shaft pumps. However, the overall goal
was not 10 prodiice a complex moclel that duplicates the
actual performance of the pump being analyzed), but rather
to develop a modelling technique that ineludes only those
clements pertinent to predicting the general vibration per
formance. Although the techniques may not be sufficient 10
accurately simulate precise operation of the units, a primary
“objective was to develop guidelines, so an interested vertical
pump manufacturer or user will beable o adequately model
vertically mounted pump to reliably predict its operational
‘bration characteristics,
‘perimental Efforts
‘The analytical model initially provided important direc-
mn for determining optimum sensor locations on the
pump. However, experimentally acquired data were re-
quited co both verify and, when necessary, tune the model
The tuned model, in turn, provided further direction t0
‘modify the experimental seuup,
An important objective was to develop recommendations
for proper application of submergible vibration sensors,
‘The work was not intended to develop new types of sensors
but to evaluate the ability of existing sensor types to measure
vibration reliably while they were submerged
The experimental efforts were directed toward establish
ing guidelines forthe selection and use of sensors to monitor
vertical pump vibration. Sensor survivability, practicality,
and ease of use were parameters considered for instrumen-
tation selection,
Pumps Studied
The following pumps were studied:
+ Heater Drip Pump (7 stages, 13-fe long shaft, 40 hp, 8600
‘cpm, 210 gpm, 500-f total head), This pump normally runs
‘continuously with load varying as a function of plant load‘VIBRATION MONITORING OF VERTICALLY MOUNTED PUMPS—TOOLS FOR MANUFACTURERS AND USERS 77
+ Condensate Pumps (2) ( stages, 15-4 long sha, 150 hp,
1770 cpm, 760 gpm, 520-0 total head). These purnps
normally cycle on'a weekly basis with load varying asa
function of plant lad. The pumps ae identical excep one
Contains an inner shaft spider” bearing near misao the
pump sha; the second pump has no such bearing
+ Main Circulator Pumps 2) (1 stage, 194{long shaft, 125
tp, 880 cpm, 12.900 gpro- S0-f tua head). These units
pimp seawater and normaly operate continuously at ely
onstant load. Tidal scion is nominally 5 fc
* Rw Water Pump (1 stage, 41-f¢long shaft, 2250 hp, 240
pi, 160,000 gpm, 30 total head). This unix pumps tidal
influenced freshwater and nominally operates continuously
4 uniform loa. Tidal action is nominally = 3 fe
These pumps ae all located at operational power plans
‘This aspect was important, because many’of the problems
usually associated with laboratory tests the difficlty in
$pplying suck results to actual in-plant pumps were avoided
through onsite testing
Results of Research Efforts
Recent work [2] has resulted in a number of “lessons
earned that may be applicable in future efforts to monitor
aand diagnose problems in submerged vertical pumps. The
following isa summary of some ofthe findingson the pumps
studi
Modelling
Computer simulation of a pump’ rotordynamic charac
teristics has provided valuable insight into understanding
the vibration performance of vertically mounted pumps.
Modelling is especially useful in:
+ Determining initial sensor types and locations.
+ Identifying the source of unknown frequencies in the
pump’ vibration signature.
+ Assessing “what if" scenarios to help determine “sensi-
tive" modes and in correcting pump vibration problems
Modelling guidelines [2] have been developed to assist the
interested user, manufacturer, oF diagnostician assess vert-
cally mounted pumps. Together with an existing computer
program,* these guidelines** have been effective in pre-
dieting the behavior of the pumps studied. The models do
not have to be overly complex to achieve results with
sufficient accuracy to determine natural frequencies, mode
shapes, and approximate damped response of the pump.
‘The work accomplished has demonstrated that only a
reasonable amount of care is needed to develop a pump
‘model that realistically simulates the vibration modes that
are of primary interest (0 users and manufacturers of
variety of submerged shaft vertical pumps
* The Finite Element Analysis Tol for Utility Rotordy
‘namic Evaluation (FEATURE) computer program [3]
performs linear critical speed, damped response, and
Stability calculations fora wide range of rotor geometries,
including multiple, concentric, counter rotating spools
and large, mullspan, rigidly coupled rotors. A faite
«lement formulation gives the program numerical stabil
ty. Acompanion program [4] provides fui film bearing
properties. Zote, etal, (5] further describe the FEA
FURE program.
"= These. guidelines incorporate applicable concepts
from seal and Mid film bearing theory as part of the
techniques for medeling submerged shat vertical pumps
[6.7.8.9]
Iisimportant to note that pumps with shafts exceeding 40
feet and those with many intermediate bearings have not
teen studied at this ime. Therefore, the capability ofthe
computer program and the modelling guidelines to accu
Tate sate response of this type of pump has not been
ested
Monitoring without Modelling
Even though computer modeling can provide valuable
asssacein determining the probable source of many types
af vibration in verical pumps itisnotan essental ingredient
ff monitoring project. AKhough ones prognostic ability
mnay be limited, many vertical pumps can be effectively
monitored for trends and changes in bration without the
Hsitance of computer model, The remainder of the
discussion wil focus on recom experience in inatrumentin
Sid monitoring several types of vertically mounted pumps,
Allof the pumps studied were modelled partofthe overall
‘ffort. Some of the findings are based an the modelling
Sccomplshed; others are independent ofthat work.
Impact Ts
Impacting structures and measuring their response con-
stitutes a good method of identifying the sources of some of |
the frequencies that may be observed in the vibration
spectrum of the operating pump. The opportunity to ac-
quire such data on intermediate tubes, bearing support
structures, assembled pump casings, ete, should not be
missed. However, “hang and bang” responses of a pumps
rotating elements should not be expected to coincide with
those observed in the running pump. The frequencies
measured on the rotating elements of a multistage pump
during impact tests and during operation are shown in
Figure 1. The differences are mainly due to the stiffening
ceffectof the bearings and wear rings (water lubricated) in the
pump,
V a '
Figure 1. Comparsion of Natural Frequencies and Mode Shapes for
Impact and Dustalled Configuration of a Nertical Pump.