You are on page 1of 117


Gulf Publishing Company
@ 1974
This reference manual has
manuals in the series are:
been reprinted from the
regular monthly
handbooks and
D r
Table of Contents
Page No.
Centrifugal Compressor Performance Testing
Better mechanical testing can improve compressor reliability
Test compressor performance-in the shop
Test compressor performance-in the field
Maintenance Techniques .......
How to improve compressor and maintenance
lmprove machinery mainten
Why clean turbomachines
A closer look at turbomach
Hot alignment too complicated
Turbines using too much steanl"? ...
Better pump grouting
Centrifugal Pumps ..
How to improve pump
How to control purnp vibration
How to prevent pul
Turboexpanders ..1
New develop hot gas
Critical. Related
Are couplings
GasS histories of
Coniprq$sor prol
and C
: causes
Itnik in
new A
t' '
Performance Testing
Better mechanical testing can
improve compressor reliability
Vendor shop testing of critical
process compressors can reveal
mechanical problems prior to
field installation. Here are
some points to consider when
pertorming compressor mec hanical
Douglos F. Neole, Union Carbide Corp.-Chemicals
and Plastics, South Charleston, W. Va.
Conrpnrsson MEoHANTcAL tests have proven effective
in identifying design and fabrication deficiencies. If these
deficiencies are identified before shipment of the com-
pressor corrections can be made and verified under opti-
mum shop conditions.
The value of compressor mechanical tests is limited by
the absence of significant gasJoad and the low energy
involved. The tests are most effective if API test standards
and the customer's supplementary test specifications are
strictly enforced as a minimum requirement, Revised and
new API Standards relating to compressors and oil sys-
tems will significantly strengthen test requirements. The
standards will also increase the obligations of vendors with
regard to test facilities and data acquisition.
Test obiectives. Table 1 is a brief list of specific objec-
tives of compressor mechanical tests. Note that only
partial verification of the considerations listed on the
table is claimed. Even the most sophisticated mechanical
test falls short of complete and prolonged simulation of
actual operation under full gas-load and power.
TABLE l-Objectives of centrifugal compressor
Partial verification of:
oThe quality of over-all unit assembly
o Fieedom from internal rubs
r Bea ring fit, alignment and adequacy of lubrication
o Rot or-bearing system dynamic stability and calculated critical speed
oVibration levels
o Cor rectness of assembly and tightness of shaft oil seals
o Drive coupling fit-up and balance
o Lub rication system cleanliness and performance
oTrain component compatibility (optional)
o Noise (optional)
API test slqndqrds. The Second Edition of API Stand-
ard. 617, "Centrifugal Compressors for General Refinery
Services,"'has provided an accepted basis for mechanical
spin tests since 1963. Most manufacturers and purchasers
added to the requirements of API 617 as new technology
developed. Many of these additional requirements will be
reflected in the Third Edition of API 617, scheduled for
publication this year.2
Table 2 summarizes key differences between the old and
new API Standard 617. The new standard provides a
significant step toward beter defined, more uniform and
more rigorously verified centrifugal compressors, but there
are areas where the purchaser must define and exercise
his rights 'andf or preferences.
Righf to wilness. The purchaser reseryes the right to
observe testing, dismantling, inspection and reassembly of
"when specified." The vendor must provide
"sufficient notice" when shop inspection by the purchaser
is required. Sufficient notice is to be defined by mutual
consent of the purchaser and vendor. We have found
that vendors frequently cannot be counted on to maintain
TABLE 2-APl Standard 617lor centrifugal
compressor mechanical tests
Second Edition (1963) Third Edition (Planned issue)
Essentially same
Purchaser reserves the right to
observe tests
lncrease speed f rom 0-100[ MCS
in 10ft increments
Run at MGS for 4 hours
Use contract shaft seals and bear
ings during test
Not required (except lube)
The actual first critical shall be
Essentially same
Perform and record vibration
Use purchased vibration
Run spare rotors ordered with
Essentially same but include ad-
ditional data on lube and seal
oil systems, and rotor dynamics
Mandatory hydro tests
lnquiry specifications and order
shall specify tests requiring witness
lncrease speed in undefined in-
crements from 0-110[ MCS
Run at MCS for 4 hours,
lnstall oil seals after test and
manually turn
Demonstrate control systems
(E.G.; inlet guide vanes) to the
extent practical
Make every effort to determine
first critical of flexible shafts, short
of opening case and unbalancing
Record throughout operating
speed range
Not required
Not required
Not required
Vendor to provide certified detailed
logs including vibration and oil
temperature data
records of all required inspections or to provide adequate
advance notice of inspections. Notification is frequently
too late to permit the most knowledgeable machinery
specialist to be assigned and made familiar with his
assignment. The alternative is for the purchaser to con-
tinually stress during negotiations and order coordination
meetings, the requirement for responsible and timely ven-
dor notification, and to follow up with frequent checks
by his own inspection/expediting personnel.
Tesl speeds
durqtion. The Third Edition of API
617 is strengthened by the inclusion of a definition of the
increments of increasing speed during mechanical tests.
The four-hour maximum continuous speed run apparently
is less rigorous, as an uninterrupted run is no longer
required. However, we specify successful completion of
an uninterrupted four-hour run, and a repeat uninter-
rupted run if bearings, seals or balance require modifica-
tion of the test.
Verificqtion of shoft seqls. A significant improvement
in the Third Edition of API 617 is the requirement that
contract bearings and shaft end seals be installed during
the mechanical test and that the oil leakage from each
seal be measured with approximate design differential
pressures across the seals. Although not required by the
Second Edition, many have imposed this requirement for
sometime. More widespread testing with seals in place
will occur following the issue of the Third Edition. It will
be easier for purchasers to gain confidence in the me-
chanical adequacy of seals.
Conlrol syslem checks. As indicated in Table 2, the
Second Edition of API 617 required control systems to be
demonstrated "to the extent practical." For centrifugal
compressors, we have interpreted this to relate to lube oil
systems and inlet guide vane actuators. The Third Edition
of API 617 does not contain a comparable requirement.
The new API oil system requires that "the completed
oil system shall be shop run to test operations," etc., but
it is only by rather liberal interpre_t21i6n that this covers
inlet guide vane actuators.3 Within the past two years,
we have experienced several startup difficulties from
miscalibrated or mechanically unsound guide vane link-
ages. Guide vanes will receive more, not less, attention.
Delerminqtion of criticol speeds. The new edition of
API Standard 617 requires that "for flexible-shaft com-
pressors, the actual first critical speed shall be deter-
mingd-" whereas the Second Edition required "every
sff611-16 determine the actual first critical speed-short
of opening the casing to create rotor unbalance." Our
specifications have independently paralleled the evolution
of the API Standard 617 and have avoided forcing criti-
cals. Present thinking is toward mandatory identification
of criticals as required by the Third Edition.
beqring temperqture meqsuremenl
qnd qnqlysis.
The revised issue of API Standard 617
reflects the progress made during the past 10 years on
vibration measurement and analysis. For example, a
sweep of vibration amplitudes at frequencies covering a
minimum range of 25 percent of synchronous to twice
vane passing frequency is now required. This is inter-
preted to mean multiples of number of vanes times sta-
tionary parts such as discharge nozzles or diffuser vanes.
Additionally, purchased vibration probes and detectors
shall be used during the mechanical running test. We
have preferred this practice for several years.
Requirements or recommended practices relating to use
of bearing metal temperature sensors during test are not
stated in the API Standard. These detectors have become
standard for our critical rotating machinery and should
be better covered by APL
Spore rotors. Because of the need for reliable
process compressors, a mechanical run test of spare rotors
ordered with the compressor is now mandatory. The prac-
tice has been valuable by not only ensuring a properly
balanced rotor, but by avoiding situations where the spare,
although properly balanced, just wouldn't fit in the ma-
chine without extensive diaphragm or thrust bearing
Vendor tesl dqto. The Third Edition of API 617 re-
quires submittal of more extensive test data to the pur-
chaser. The more stringent data requirements reflect the
improved scope of the tests as previously discussed. Addi-
tional data includes seal oil temperatures, pressures and
leakage rates; rotor balancing and critical speeds, and
more extensive vibration data.
Optionol tesls. Optional tests which may now be speii-
fied in the purchaser's inquiry or order include hydraulic
performance, complete unit tests, tandem tests, gear tests,
helium leak tests, sound level tests and post-test inspec-
tions. We consider performance tests on an individual
basis as a function of the extent to which the design has
been proven by analogous machines, the degree to which
TABLE 3-AP! Standard 614* for lubrication,
shaft-sealing, and control oil systems-
summary of inspection and test requirements
o Purchaser has right of inspection
oPurchaser's inspector furnished specifications, material certifications, and
running test data
oPurchaser shall specify whether purchased oil system shall be used during
main equipment shop test
oComponent hydrostatic tests comparable to previous API Standard-617,
Second Edition
o Four-hour shop test under normal system operating conditions
o System checked for leaks and proper functioning of controls and alarms
.System capable of riding out filter-cooler and oil pump changeovers
. System cleanliness criteria defined
Planned issue
TABLE 4-Vendor test plans
Test installation
Shop and contract equipment included in the test train
Xnown limitations in driver power
and/or speed
Measuring instruments
Pressure and/or temperature control systems
Seal and lube oil console limitations
Test procedure
Duration of each speed run
Shop data sheets
Special procedures to avoid excessive temperatures or other limitations
Post-test inspections
Report outline
Scheduled issue date
Data to be reported
Disbussion subject headings
final operating conditions can be synthesized, the criti-
cality of the machine, the time schedule and the cost.
Gears are also subjected to running tests. When duplicate
gears are ordered, they are subjected to full speed, full
torque load and back-to-back tests. Following the test, a
full bearing and tooth-contact inspection is performed.
The gear undergoes a standard four-hour shop running
test on reassembly. Complete unit, tandem, or shop sound
level tests are seldom
Lube, seol ond conlrol oi! systems. A properly de-
signed and functioning oil system is a prerequisite for a
reliable compressor train. Thus a discussion of compressor
mechanical tests should include consideration of the oil
system that will support the train. This consideration is
facilitated by the new API Standard 614 entitled "Lubri-
cation, Shaft Sealing, and Control Systems for Special-
Purpose Applications."s
Table 3 briefly summarizes key features of oil system
tests as specified by API 614. The right of inspection and
receipt of finished specifications, material certifications
and running test da[p are established. It is the purchaser's
responsibility to specify whether the purchased oil system
shall be used during the compressor test. Our normal
practice has been to require cleaning, flushing and ship-
ment preservation of oil systems. Operation of the con-
tract oil console during mechanical testing of the com-
pressor has been required only infrequently. Ifowever,
some purchasers are requiring tests to be run with the
contract oil console and this appears to have some advan-
tages. The design and fabrication of oil systems seem to
be a sideline with many compressor manufacturers, and
startup problems are not uncommon. Furthermore, the
mandatory use of the system for test will provide a signifi-
cant incentive to thoroughly clean the systems in the
vendor shops.
Additional points in API Standard 614 bear considera-
tion. Filter and cooler changeovers shall be accomplished
without the system's delivery pressure dropping to the
automatic start setting of the standby pump. The capa-
bility of the control valve shall be demonstrated by start-
ing, running and stopping a second pump (main or
standby) without the delivery pressure dropping below 75
percent of the differential between normal and shutdown
pressures. The valve shall also be capable of holding oil
pressure at minimum oil flow (normal bearing and seal oil
plus steady state control oil flows).
Problems with new systems could have been avoided by
the foregoing types of checks. For instance, we frequently
use shaft-driven main oil pumps. After starting a unit on
the electric motor-driven auxiliary pump, the main pump
is operated briefly in parallel with the auxiliary, and
the recycle control valve is wide open. When the auxiliary
oil pump is tripped off, the valve response was too slow to
prevent a drop in header pressure and subsequent trip of
the compressor train. Some vendors have resolved this
problem through more responsive and better tuned con-
trols. Other vendors have provided a manually controlled
recycle line from the auxiliary pump which can be opened
to effectively remove the pump from the system prior to
its shutdown. We have made similar last minute modifica-
tions in the field. Proper exercising of oil systems during
test can avoid these costly and time-consuming nuisances.
Test plans ond preporolions. The latest API Stand-
ards provide good general definitions of test requirements.
It is only in limited instances, however, that the standards
require vendor submittal of a detailed plan for each spe-
cific compressor system test. We have found that this is a
serious omission. Some vendors stop short of full compli-
ance with API requirements. For several years we have
required the vendor to submit detailed test plans for
review prior to testing of critical machinery.
The contents of an ideal test plan are listed in Table 4.
Advance review of the test setup will avoid surprises to
inspectors andf or machinery specialists when they report
to the vendor's shop. These setups normally have suffi-
cient drive turbine power available to attain specified
system speeds. It is common, however, to encounter limi-
tations with electric motor drives. The unavailability of a
motor with properly rated speed and the inability to vary
speed result in less than adequate tests. The development
of excessive casing gas temperatures during tests have
also compromised tests. Prior knowledge of these circum-
stances can permit corrections or alternatives to be
implemented in a timely manner. Situations where the
purchaser is forced to compromise his test specifications or
lose his scheduled time on the test stand can be avoided.
The proposed test procedures should define the basis for
determining when the compressor undergoing test has
stabilized and when the test operator is free to move on to
another point. Normally, the temperature of oil leaving
the bearings is the last indicator to stabilize. The pur-
chaser's insistence on stabilization of oil temperatures at
each point seldom extends the test significantly. It is
desirable, however, for the vendor to know in advance
that this requirement will be made.
The planned shop data sheet (or typical automatic data
center printout) provides practical insights into the ex-
tent of data to be taken. A quick scan of these sheets can
identify unacceptable omissions. For example, it is com-
mon to find that seal oil leakages are not to be recorded,
or are to be recorded at static conditions or at reduced
speeds and oil temperatures. Every effort should be made
to simulate design speeds and operating temperatures to
get the best possible insight into actual operating per-
The poor legibility of vendor test report forms is a
source of frequent frustration. Specifications call for leg-
ible data sheets, but it seems some will have to be rejected
before satisfactory quality can be attained.
The lack of interpretation of test results is also of con-
cern. Most vendors will provide a formal test report with
bare data. Ordinarily, interpreted d.ata are supplied only
if required by the purchase order. This is a worthwhile
requirement; prior consideration of the characteristics of
a compressor can be valuable when diagnosing a real or
apparent problem on the midnight shift!
Purchaser emphasis on the reliability of compressors has
led to a general upgrading of vendor test installations. It
has become the exception when major vendors are unable
to comply with the older API test requirements.
Vibrqtion ond temperqlure monitoring. Significant
progress has been made in using the purchaser's own
proximity probe holders and probes during mechanical
testing" In the past, vendors were slow to accept the
feasibility of shop installation of the probes----especially
two per
that could be adjusted and replaced while
the machine was in operation. The contract probes were
frequently not available for the test so temporary shop
probes and holders were used. Part of the value of shop
mechanical tests as a reference for startup problem diag-
nosis was lost because surfaces observed by the shop
probes differed from surfaces observed by permanently
installed probes,
It is still a challenging task to develop mutually accept-
able proximity probe installations. Vendor's ability to
utilize probes during test has improved, however. Many
vendors have installed variable frequency filters, real time
About lhe
DoucLAs F. NsAr,E is mo,nager---ltrocess
macldnerg appl;ication
bide Corp.--4hemicals and Plastics,
South Charleston, W.Va. He i,s respon-
the sTtecifi,cation antl process
desi,gn of large machinerg systems
capital erpansions and deaelopment of
new technology in this area. P,r'i,or er-
perience with Union Carbide Corp. in-
cludes plant process design and a oari,ety
of engineet'ing nxanag enxent as sign-
ments. Mt'. Neale holds a B,S. degree i,n mechanical eng,ineer-
Rensselaer Polgtechnic Institute and, a M.S. degree
i,n meclu,nical engineering
Massachusetts Institute of
Technology. He is a registered prof essional engineer in West
Virgina and, a member of A.S.M.E.
spectrum ama)yzerc, X-Y plotters and oscilloscopes to per-
mit any probe or set of probes to be read, analyzed and
displayed. Multichannel tape recorders are becoming in-
creasingly cornmon for simultaneous recording and subse-
quent display of many data channels. Some vendors have
even installed "run-out subtractors" to electronically mask
vibration or excessive rotor or probe surface runout. One
major rnanufacturer recently commissioned a minicom-
puter system to simultaneously analyze multiple data
inputs and print results within minutes of test completion.
Progress has not been so great with regard to bearing
metal temperatures. We are strongly committed to using
bearing metal temperature sensors on critical machines
but have found that even more extensive vendor liaison
is required than with proximity probes. Test installations
for reading out the sensors are frequently vendor impro'
vised and may not be available during the test unless
requested in advance.
Couplings. Common test stand practice is to use shop
couplings except for possibly one hub, during compressor
mechanical tests. We have accepted this practice in the
past. However, recent experiences indicate that couplings
are one of the more troublesome machinery train com-
ponents. Many coupling manufacturers apparently do not
enforce standards for residual unbalance and vibration
comparable to those of compressor manufacturers. Addi-
tionally, coupling manufacturer's use, during balancing, of
coupling spool and sleeve surfaces that were eccentric to
the gear tooth pitch circle has led to
excessive vibration.
The use of the contract coupling is being increasingly
specified for tests of critical machines for several reasons:
The contract coupling will permit the closest practical
simulation of the final field installation with respect to
rotor dynamics.
The mechanical test can confirm coupling balance
and provide an indication of machining quality.
The use of the contract coupling reduces the tempta-
tion for vendor shop personnel to "touch ,p" the shop
coupling to compensate for minor rotor unbalance.
The added expense and time required to adapt the
contract coupling to test service is well
The strengthened API standards supplemented by the
preparations discussed in this paper will contribute to
effective mechanical testing of centrifugal compressors.
The following additional recommendations are made,
A competent machinery engineer experienced in startup
and maintenance problems should witness the tests. On-
the-spot interpretations and decisions will materially influ-
ence the effectiveness of the test and assure satisfactory
performance of the compressor.
Rotor residual unbalance data sheets and run-out maps
should be reviewed prior to the test. The knowledge
gained can alert the test engineer to the possibility of rubs
or abnormal noises.
The vendor's calculations of expected casing tempera-
tures during tests should be reviewed. There are too many
instances where failure to plan a vacuum test has resulted
in termination of the run before planned speeds are
The use of electronic rotor run-out subtractors in con-
with proximity vibration probes should not be
permitted. Vendors should machine probe surfaces well
enough to avoid the need for these corrections.
Insist that actual rotor critical speeds be identified. If
it is necessary to "force" the critical, the removal of. a
coupling bolt will normally suffice. If the coupling is too
close to a bearing, it may be necessary to open the case
and add internal weights.
Compressor design and manufacturing deficiencies be-
come apparent with the passage of time. Therefore, dq-
mand an uninterrupted four-hour maximum continuous
speed run regardless of the test stand problems encoun-
tered. If it is a hardship to keep the machine operating
for four hours on the test stand, it is likely that the same
condition will prevail in the field.
Be certain that the contract or shop drive coupling is
removed prior to rotor rebalance. It can be a costly error
to permit rotor or coupling unbalance to be camouflaged
by correction of an adjacent removable part.
A previously stored or idle rotor should be run at a
speed sufficient for self-correction of bow or "set" before
being rebalanced. Recently one of our machinery engi-
neers refused to accept the vendor's suggestion to balance
and then test. During the initial run, apparent unbalance
disappeared. The vendor's proposal to save time by prior
balancing would have resulted in recurring unbalance
after the "set"of the rotor was spun out.
Our vendor inspectors and test engineers always check
the compressor bearings and shaft end seals following a
test. The case is not opened unless unexplained noise or
vibrations are encountered.
Value of tesls. ft was stated earlier that compressor
mechanical tests only partially accomplish the specific
objectives listed in Table 1. The degree of accomplish-
ment is limited by the inherent shortcomings of a minimal
gas-load, low energy operation. It is influenced by the
extent to which operating temperatures, pressures and
flows are simulated and by the quality and quantity of
recorded measurements.
Our experience indicates that mechanical tests are
effective in confirming the quality of assembly of com-
pressors primarily with regard to:
Adequacy of clearances and alignment of shaft and
wheel labyrinth seals
The absence of Ieaks in lubrication and seal oil passages
Bearing fit, alignment and lubrication
Rotor balance
Rotor-bearing stability and avoidance of criticals.
Several instances of labyrinth seal rubs, bearing mis-
alignment and damaged thrust bearing shoes were iden-
tified. In another instance, a wiped journal
bearing was
discovered. The wipe was explained by the vendor as the
result of a "somewhat tight" bearing, and scraping was
proposed. Our engineer insisted that a new bearing be
installed and the test rerun. The same situation occurred.
It was found that the bearing housing was undersized and
caused excessive bearing crush. If scraping of the bearing
had been permitted, the plant would have inherited a
chronic problem that would arise each time new bearings
were installed.
Not unexpectedly, there have been instances of unde-
tected deficiencies in assembly. In one instance, a casing
support plate was improperly aligned relative to the
machine's centerline. At mechanical spin test tempera-
tures, casing expansion was insuficient to permit detec-
tion. The problem became apparent when the machine
was placed in normal operation. A misbored thrust bear-
ing housing had gone undetected until placed in opera-
tion. The axial forces placed on the rotor during the
mechanical test were insufficient to
the rotor thrust
collar and cause the shaft deflection and vibration ob-
served later in full-load operation.
An error in the pattern drawing used for the upper half
of the diaphragms of one machine went undetected. This
defect permitted hot discharge gas to bypass the balance
piston and recycle to suction through the balance piston
vent line. A performance test would have provided two
indicators of this deficiency: (1) reduced capacity and
(2) higher-than-planned gas temperatures. A simple cal-
culation of the predicted mechanical test air discharge
temperature might have identified the problem.
Mechanical tests provide significant insight into rotor-
bearing dynamic stability, especially in Iight of the
limited gas dampening provided by the unloaded condi-
tions established for mechanical tests. Excessive vibration
was noted on one of our machines. Touch-up balancing
was performed on the rotor before it and a duplicated
coupling were reinstalled and the test rerun. When the
vibration persisted, the rotor was stripped for total rein-
spection, a potentially disastrous flaw was found, and
the shaft was
Mechanical tests of compressors with seals in place have
also been effective. fn one instance, excessive seal leakage
was evident. The seal O-rings appeared to fit but were not
as specified. In addition, an out-of-round outer seal bush-
ing (floating ring) and incorrect spacing of parts at
assembly were found. The vendor replaced the defective
parts and reworked the seals to obtain a satisfactorily
performing system.
Lube oil consoles have not been tested with the associ-
ated compressor, and pump noise and vibration have been
encountered in several instances during startup. These
faults are usually corrected by field realignment of the
pump and piping. Control system deficiencies hav.e been
more common and commissioning cleanup is always a
headache. Specification of a more extensive workout in
the vendor's shop is increasingly justified. Testing of the
oil console and compressor together is becoming a neces-
sary requirement.
API 617-Ccntrifugal Compressors for General Refinery Service, Seond
Edition, 1963, Amerien Petroleum Institute, Washington, D. C.
617-Third Edition, planned issue.
API 61,!-Lubrication, Sbaft-Sealing, and Control OiI System for Special-
Pupose Applications, planned issue.
Test compressor
the shop
Compression equipment must pertorm
expected. Failure to do so can result
in many startup and operation
difticulties. Here's what can be done
to avoid problems
betore the machinery
is installed
Royce N. Brown, Dow Chemical U.S.A., Houstol
RncrNrr,v there has been much written about thc
mechanical reliability of compressor trains. There is no
question as to the importance of a high degree of rnechau-
ical reliability since it generally results in a "go" or
"no go" situation. Good performance attainability is gen-
erally not as consequential, because partial load operation
may be possible with a deficient compressor. A fact not
generally realized is that the subject of performance is a
requirement for reliability! For a machine to be truly reli-
able it rnust not only run well mechanically but must
perform at 100 percent of its capacity whenever called
upon. It is then very desirable to test prove performance
prior to receiving a machine, o.r as an alternative, to field
test shortly after installation prior to enterins prodrrction.
The basis for code testing is the ASME Powcr Test
Code PTC 10-1965 Compressors and Exhausters.l Several
specific points made in the Code were intended as guiding
principles, yet are often misunderstood. The following
facts must be considered:
The Code establishes the rules for a test, including
the definitions of Code or Non-Code.
The Code is not a textbook on testing.
The Code has to assume that the gas properties oi
the gases involved are known. It recosnizes that this is
Fig,_1-Allowable departure from specified design parameters
for Class ll and Class lll tests.
TEST COMPRESSOR PERFORMANCE-IN allowable departure from specified design parameters for
Class II and III tests.
Another facet of the Code is the establishment of instru-
mentation for the compressor test. The type, number of
points and locations on the comp.ressor are defined and
apply to all classes of Code tests. The Code is specific
about the number of readings per point and the minimum
duration of the run. Calculation methods are also sup-
plied using perfect gas thermodynamic relations for Class
II tests and real gas relations for Class III. Class I inher-
ently requires little correction; but when corrections are
needed, either perfect or real gas relations can be used
depending on the nature of the gas. The Code draws
heavily on work presented by Schultz6 using methods
directly from his paper. Calculation methods will be fur-
ther discussed later in the paper.
In actual practice very few tests are run as "True
Code," or by Code definition, as Code tests. Each vendor
uses some form of deviation to either speed up the test or
to fit his facilities. Most of these shortcuts are not seri-
ous-in fact they contribute to keeping the cost of testing
reasonable. The user must understand fully the Code
requirements and where these deviations occur. An ex-
ample of a nonserious deviation would be the waiver of
the time limit per point, i.e., the time required for settling
out. This may be based on stability of the data only, rather
than data stability and a minimum time. Another devia-
tion generally taken, which the user must evaluate, is the
gage calibration procedure immediately prior to and after
the test. The use of manometers keeps this requirement to
a minimum. There are quite a number of the above ex-
amples, too numerous to mention, none of which generally
is serious. The user should realize, however, in permitting
these deviations he may not have a true code test.
Testing lhe unleslqbles. Compressors, which fall outside
of Code limits due to the nature of the gas, vendor shop
test limits, or machine speed limits, may still be tested
and useful information obtained for the user.
The first and most obvious question is how and to what
degree one evaluates or judges
the test. The most obvious
reply is "it depends." On this seemingly vague note, let us
explore some of the possibilities.
On a simple "once through" compressor, the gas prop-
erties may dictate a test speed higher than the physical
capability of the unit. For a Class II or III test, the Code
requires an equivalent test speed higher than rated speed
so that the wheels are operated at the design volume ratio.
When this is not possible the machine cannot be Code
tested. In the case of the "once through" compressor with
two or maybe three impellers, a predicted air curve may be
developed. The air curve is derived by using the basic
wheel characteristics and calculating a wheel-to-wheel
rematch of the compressor on air. The shape of the curve
will not be truly maintained. However, the design flow
area can be reasonably explored to check capacity and
head. It is reasonable to assume if the predicted air curve
is reproduced, that the wheels did perform as designed
and the compressor should rematch on design gas and
produce expected flow and head. Power requirements are
somewhat less reliable, but a reasonable measure of effi-
ciency may be made.
Fig. 2-Typical multi-stage centrifugal compressor.
not always the case, but must place the burden of knowl-
edge of gas on the test's participating parties.
The Code establishes a basis on which to agree or
disagree. Ultimately the final test procedure and methods
must be agreed on by the purchaser and vendor.
The Code attempts to categorize testing and thereby
establish an inherent degree of accuracy. These categories
are based on methods of test and methods of analysis.
The Code establishes three classes of tests. Class I in-
cludes all tests made on the specified gas (whether treated
as perfect or real) at the speed, inlet pressurg inlet tem-
Class III basically differ only in method of analysis of
data and computation of results. The Class II tesi may
use perfect gas laws in the calculation while Class IIi
must use the more complex
gas,, equations. An
example of a Class II test might be a suction throttled air
comp.ressor. An example of Class III test might be a COz
loop test of a hydrocarbon compressor. Fig. I shows code
The Code establishes a basis
on which to agree or disagree.
The final test procedure
methods must he agreed on by
the purchaser
and vendor.
For more complex compressor configurations with mul-
tiple inlets andf or outlets (Fig. 2), even more judgments
must be made. The unit may have to be tested one section
at a time. A different test speed may be required for each
section. Another method of testing requires the removal of
some impellers. While tedious and expensive, it might
well be justified.
In some cases, particularly if more than
one section is tested at one time, additional instruments
may be required. Instruments may have to be installed
internally within the compressor. The internal instruments
are required to sepa.rate the individual section perform-
ance from the whole. This test normally uses air as the
test gas so the vendor must prepare air-expected curves
for use in test evaluation.
When testing requires significant interpretation a judg-
ment must be made as to whether or not the nature of the
test will develop good temperature rise data. lt may be
necessary to install total temperature probes at the OD of
the last impeller prior to the flow leaving the casing.
When properly installed this type measurement provides a
much more reliable reading of true temperature rise and
therefore gives a more accurate basis for compressor effi-
ciency and input power required. Generally the vendors
resist the specification of internal instruments. Therefore,
good understanding of users' requirements and objectives
is essential. If costs are realistically evaluated, air equiva-
lent testing will cost less than loop testing.
Loop fesfing. A closedJoop test may be necessary to
performance test within the Code. There is more inherent
accuracy in the loop test than in the ai.r test as volume
ratio matching can be more closely achieved. The loop
test has several limitations that may not be obvious. Loop
tests are generally expensive and time consuming. They
may become so complicated with complex compressor con-
fieurations that they become impractical to set up. Finally,
the number of gases available for sh'op-loop testing is
quite limited. Air which is quite available for normal test-
ing is not particularly suitable for loop testing. Fig. 3
shows a typical shop test loop arrangement.
In a loop test, air and oil come into contact causing the
danger of an explosion. Extreme caution must be used.
All combustible, toxic and other gases where safety is
involved are disqualified. This narrows the field consider-
Aboul lhe
Rovcn BRowN is or? Associate Cottstdt-
ing Engineer with Dow Chemical,
U.S.A,, Engineering and Constru,ction
Seroices, Houston. He is responsible
or rotating ma,chinery specifi,cations,
bid eaaluation, and technical asgistance
new equipment. Dut:ies also include
d:iagnostic assistance
equipment and analEtical assistance
eaaluations and rebuilds ol rotat-
ing machinerg. Prior erperience
Fig. 3-Shop test loop arrangement.
ably. The gas cost factor makes the problem even nlore
dificult. The problem of known gas properties adds the
final cornplication. The limitations just
covered, together
with the cost factor, help to provide the incentive to get a
useful test from an open air test even if the instrumenta-
tion is more complex.
fesl correlqtion. Earlier mention was made of gases and
their correlation calculation. A point was made that the
Code could not assume the role of being the final author-
ity on gases. Do not expect the vendor to be all-knowing
in the areas of gas properties. The problem of gas proper-
ties is the responsibility of the user. Much data is pub-
lished on gases which together with high-speed computers
make the job
of defining gas properties somewhat easier
today than it was at the timethe Code was written. This
is not to say that all properties of all gases are as well
defined as they might be. Gas mixtures in particular are
always a problem.
A few suggestions come to mind. Also some useful ref-
erences are listed at the end of the article. The BWR2,3
equation works well with m,any of the hydrocarbons. The
Martin Ho3'a equation works well with refrigerants and
chlorine. Once gas properties are established, correlation
methods suggested by the Code provide results that are
meaningful to the ultimate user of the compressor.
Originally presentcd at the Third Sympmium on Compressor Train Re-
Irability, Manufrcturing Chemists Association, April 24, tg7c, Chrcago.
-- ---
clud,es u;ork in th,e control, pump, steam turbine and com-
Test compressor
the field
Compressors occasionally tail to meet
pertormance exqectations atter
instatlation and startup. When this
occurs, tield testing is required-
Here's how to get the
efticiently and economicallY
H. M. Dovis, Delaval Turbine, Inc., Trenton' N.J.
Cr,Nrnrruc,tl- conlPressors are rugged designs that will
perform satisfactorily fot years at a time betrveen sched-
uled rr-raintenance shutdowns. Flolvever, there are titnes
rr'hen the contltressor's perfortnance mal' be in question
due to excessive power consumption or simply that it will
not provide the compression needed- Compressor perform-
a.rc" is expressed in terms of brake horsepower required
to compress a specified amount of gas flow {rom inlet
pressure to discharge pressure and can be calculated from
measurements of the following items: gas composition,
inlet pressure, discharge pressure, inlet temperature, dis-
charge temperature,
flow and speed. The type of
instruments and method used to obtain this data have a
direct effect upon the accuracy of the results. These
requirements will be discussed in detail later.
The basic cornpressor characteristic
.curve is described
in terms of inlet flow, head rise and brake horsepower for
each operating speed. A sample of a centrifugal perform-
ance curve is shown in Fig. 1. From each set of data,
which is comprised of the seven items listed above, one
compressor operating point can be calculated and plotted
in terms of head, flow and power on the basic compressor
characteristic curye. Normally five or six measurements
are needed, with a flow adjustment for each, to completely
describe the compressor curye for each speed. Usually
field testing is not conducted on this large a scale, but
rather the normal operating point is measured and the
calculated results are compared to a manufacturer's sup-
plied performance curve. These curves are either based on
the results o{ a shop test or if a test was not conducted
prior to shipment, an estimated curve can be used.
The scope of the field tests depends on the extent of
the problem. For example, if the compressor performance
is suspected to be drastically diflerent than it should be,
the manufacturer will need as much information as pos-
sible to try to determine from these tests what corrective
action is recommended. Many times the reason for the
operating difficulty can be determined from the test results
u"a if new parts are needed, they can be manufactured
ahead of time before the compressor is opened. There are
applications where the comP,ressor operating condition
cannot be varied because of the process. The user and
manufacturer must work together to establish the most
meaningful test agenda before the test begins. This helps
to eliminate misunderstandings and has proven to be the
quickest way to get the
job done.
Another reason for checking the compressor's perform-
ance is to determine if it has changed from the initial
startup. If there is a change,
may indicate that the comPr
foreign material or the intern
and a rnainterlance shutdorr'll
particularly true if nothing significant has changed in the
pro..rr, but the comPressor has become the limiting factor
in the plant's production.
0f,Er cAPACfi. 1,O0 CFI{
Fig. 1-Compressor
performance curve
The acquisition of field test data, if meaningful and
accurate results are to be achieved, requi.res planning.
These field tests are time consuming inconvenient, and
can be quite expensive. The extent of the testing agenda
depends on the purpose of the test. If the purpose of the
test is to prove or disprove manufacturer's guarantees,
then by all means the manufacturer should be consulted
on the test procedure and will probably want a rePre-
sentative present during the data acquisition. It is im-
perative that mutual agreement be reached between the
user and the manufacturer concerning the testing accu-
racy, number of test points, method of recqrding the read-
ings, etc., prior to the actual test in order to have the final
results fulfill the intent.
When the purpose of the test is more of a routine
nature and is only to provide information for the user's
benefit, it is still recommended that the manufacturer be
consulted. The manufacturer will usually be willing to
provide a test procedure that outlines the minimum re-
quirements, including the type of instrumentation, neces-
sary to achieve a useful test result and a procedure for
calculating the basic compressor characteristic curve-
The data that will be recorded during the test consists
of readings of pressure, temperature, flow, speed, and
gas properties. Measurements of.
consumption are
normally not available through direct readings, except
when the compressor is driven by an electric motor, but
rather are calculated from the other data.
Following is a table of instruments (per ASME
PTC-10)1 that may be used to test a compressor:
Pressure: Bourdon tube gages
Deadu'eight gages
Liquid rnanometers
Temperature.' Mercury-in-glass thermometers
Resistance thermometers
Flou: Orifice plates
Venturi tubes
Flow nozzles
S peed: \{echanical tachometers
Electrical tachometers
Digital electricity frequency countcrs
Gas properties: Gas sample bottles
The accuracy of the test instrurrents luust be verified
before the test. Some suppliers specify the range of thcir
instrurnents; these should be checked against an appro-
priate standard. Those instruments subject to changes in
calibration during use should be checked before and after
the test. The ASME PTC-10 also specifies that bourdon
tube pressure gages be deadu'eight calibrated at approxi-
mately 5 percent intervals over the anticipated working
range and that thermocouples and mercurl,-in-glass ther-
morneters be certified at 20 percent intervals over thc
u,orking range.
FIow measurements are obtained from either perlDa-
nently or temporarilv installed plant flow meters. lVhen
Fig, 2-Field test instrumentation diagram
these metels are installed properly they provide flow
readings that are suffciently accurate for testing purposes.
Commercially available flow meters state on the name-
plate the accuracy of the instrument and these limits can
be included when determining the over-all accuracy of
the compressor test.
When testing compressors that handle gases othet than
atmospheric air, a gas sample must be taken during the
testing to determine the volumetric analyses of the gas
mixture. When the gas composition varies during the test,
it may be necessary to obtain several gas samples to evalu-
ate the composition of the gas mixture. The analysis of
the gas samples is obtained from an independent labora-
tory after the test has been completed.
The recommended location of the pressure and tern-
perature instrumentation is shown by the diagrarn in
Fig. 2. The cornpressor performance is to be evaluated
from inlet to discharge flange and therefore the pressures
and temperatures should be measured as close to these
connections as possible. This diagram locates four con-
nections on both the suction and discharge of the corn-
pressor where readings are to be taken. Due to the lirni-
tation of some compressor installations and the accessi-
bility of the piping, it may not be possible to obtain four
readings of pressure and temperature at each measuring
plane in the pipe. Under these circumstances, the instru-
mentation is placed in the pipe in the best arrangement
possible and the potential errors in.the readings are con-
sidered when evaluating the results.
During the testing, the operating conditions of the
compressor must be maintained as steady as possible.
However, some small fluctuations can be tolerated with-
out aflecting the accuracy of the test. The table in Fig. 3
is taken from the ASME Power Test Code-l0 and gives
allowable fluctuations of test readings during a test .run.
The ASME PTC-10 has also established allowable devia-
tions for the compressor operating parameters that can
{ r )
0,5 96
05 %
0 259t
4 - tE $nli 3TA110N6
Fig. 3-Allowable fluctuation of test reaCings during a test
Ab,out the outhor
Hucn M. DAVIS is the mnnager of Cen-
trifugal Com,pressor Engineering De-
partment, DeLaoal Ttnbine Dioisiort of
DeLaaal Turbine, Inc,, Trenton, NJ.
He is
the organizat'ion
and direction of all engineering and
related, to the prod-
uct. Mq', Dquis recei,ued, a B.S. d,egree
meclnnical engineering
Rose Polg-
techrri,c Institute, Teme Hau,te, Ind,., in
1956 and a,ssumed lui,s present pos,i,tinn
measured test data in accordance with well established
thermodynamic methods. Section 5, of the ASME PTC-10
for compressors and exhausters, covers comPutation of
the test results.
The method that is used to calculate the test results
depends on the properties of the gas being comPressed.
The calculation is simple when the
gas laws apply-
When dealing with a real gas, the deviation from the
perfect gas laws must be considered.
API Data Book, Second Edition 1970,2 is an excellent
reference for the physical and thermodynamic
of gases and gas mixtures. Procedures are given for desk
calculations with recommendations that some procedures
be computerized. In addition to the above reference, the
ASME PTC-10 lists a bibliography of 87 different author-
ities who have published literature on the subject of ther-
modynamic properties of gases and gas mixtures. The
Code also cautions the engineer that considerable varia-
tion in thermodynamic properties can be found among
the various published papers on certain pure, commonly
encountered gases. For this reason, it should be agreed
upon prior to the test what thermodynamic data is to be
used to evaluate the test data.
The procedures discussed in this article, although cor-
rect, have been simplified and are meant to be an intro-
duction to the subject of compressor field testing. For
those who are actually faced with this problem, it is
recommended that the refelences be studied.
Powcr firmpreesoro and .Exhausters,"
coowirht 1965 miel Enrioerr.
"fdch"niot Dar AmerieriPetrolem Institute,
copyright 1971. I
This paper was originally presented at the
Second Turbomachinery Symposium, Texas
A&M University, College Station, Texas, Octo-
ber 1973.
The Gas Turbine Laboratories at Texas A&M
University have announced their Third Turbo-
machinery Symposium to be held on the Texas
A&M University campus from Oct. 15-17, 1974.
The Symposium will consist of lectures, dis-
cussion groups and tutorials covering all
aspects of design, application, troubleshooting
and maintenance of turbomachinery and re-
lated components.
The object of the Symposium is to provide
interested persons with the opportunity to
learn the application and principles of various
types of turbomachinery, to enable them to
keep abreast of the latest developmen s in this
field and to provide a forum wherein those who
attend can exchange ideas.
Enrollment will be limited to 700 participants
so early registration is suggested. For more
information contact:
Dr. M. P. Boyce
Gas Turbine Laboratories
Mechanical Engineering Department
Texas A&M University
College Station, Texas 77843
vrnrlBrE uiltT DEPAf,IURE i NI
2 l2l
e) TIE oorBllED SFECT OF llEt{S (+ (D) AilO (c} $lr'Lt t{0T Pf,il)t G }xn
- 'I}lAfl
Fig. 4-41|o*"ble departure from specified operating conditions.
esist during a test and still yield valid results- These limits
apply to compressors that are operating with conditions
that'are different than the original comPressor design
conditions. These allowable deviations are shown in Fig'
4. The stated departure allowances from specified oper-
ating conditions apply when the object of the test is to
establish if the compressor meets the manufacturer's per-
forrnance guarantees. This table of allowances can also
be useful to the user. When the operating conditions o'f
the compressor exceed these limits, it can be expected that
the shapi of the compressor characteristic curve, including
stable ringe and efficiency may be different than what will
be obtained when operating with the specified conditions'
The most important objective of field testing is to
obtain accurate measurements in order to calculate the
true performance of the complessor. The following is a
list of rules of good practice for testing.
Plan the test ahead of time.
Prepare a test agenda that will accomplish the
required objective.
Consult with the compressor manufacturer.
Use the best quality of calibrated instruments.
Take measurements simultaneously for each test run.
Record several sets of data for each test run.
Observe the data for consistency during the test.
Allow the opefating conditions to stabilize before
recording data.
Do not rush the test.
The compressor's performance is calculated from the
in Januwg 1970.
Maintenance Techniques
How to improve compressor
operation and maintenance
Centritugal compressors can be
designed and built with operation
and ease of maintenance in mind.
Here's what
specitying a
to consider when
new compfessor
Hugh M. Dovis, Delaval Turbine Division, Trenton, N.J.
TNDUSTRv has experienced trenrendous
glowth during the past trvo decades. The grorvth of
sinele line and continuous process plants and the in-
creasing use of automation have demonstratccl thc in-r-
portancc of component reliabilitv. \{achinerv uscrs are
norv demanding dependabie performance, simplicitl, of
operation and ease of rlaintenance. The sr-rppliers of
centlifugal compressors have been lorced to rcvier,r' tlrcir
designs and sometimes to design nerv equiplt-rent to satisfv
these users' demands. This paper discusses the wa,vs in
rvhich a centrifugal compressor can be built and used to
satisfy a customer's operating and maintenance reqLrire-
Typicol rnuhistoge compressor design. A tlpical
multistage centrifugal compressor, designed to meet a
particular custorner's needs, is slrown in Figure 1. This
machine consists of 9 inpellers in scries and is designcd
to compress 4000 cfm of gas from an inlet pressure of
25 psi to a discharge pressure ol 425 psi. Each inpelier
imparts velocity (kinetic) energy to the gas being ccm-
pressed. This velocity energy is converted into increased
pressure in the difluser passage. The cross-ovcr passage
and the return guide vanes lead the gas to the next
impeller where the compression is continued. The volume
of the gas stream is reduced as it is compressed and each
stage is designed to accept a successively smaller florv.
Ronge of opplicotion. Fig. 2 is a curve rvhich shows
the limits of applications for centrifugal compressors in
terms of flow and speed. The speed is limited by thc
stresses in the impellers. The small f1611,, high speed
compressors have the same working stress levels as the
Fig. 1-Multi-stage fabricated case compressor.
oAs TlJfEfiE
00FRE$$08 sPEm, 1000
Fig. 2-Application chart for centrifugal compressors.
Fig. 3-Four-stage barrel compressor.
large florr, lorv speed nachines.
compressor applica-
tions in the lorv flou' range arc alnost entirel,v clri,,.en
by motors ancl speed increasing
The colrpressors in
the mid-r'angc oi flor.r s are driven by rlotor-gears. steanr
turbines, and sone ges turbines. The large, high florv
compres-cors are practically all dril'en b1, steam turbines.
The size ancl olterating spcccl oI :r centriflrgal com-
pressor har e a dilect effect on the o1;cration arrd the
maintaining of rhe conrpressor.
A smal1 nrachine such es the onc sho.,r,'n in Fie. 3,
besides having high rotating speeils, is usually high pres_
sLlre as lvell. Shaft alignrnent is more critical since the
al pipe forces rnust re_
of the
ut the clear:lnces of ihe
internal seals and bearings rnust be
more closely
ciue to their small ph),sical sizc.
.\ large com the
4, is more diffic e to
The clearances
less critical but er)t
obtain, although more liberal toleranccs are acceptable,
because the components are irot easv to nor.,e anci ipeciat
lifting facilities ire required. The foundatior. for these
lalge rnachines is also of special concern" Unless the
slrpports are designed, constructed. and maintained prop-
erlv the ntacllinery nra), never aclrieve trouble-ft-ee opera-
Rofor dynomics. I,Ioclern process colrpressors are built
in accordance rvith the API specification 617., One irn-
porterlt iten defined by this specification is the natural
fi'equencies of the rotor. These natural frequencies must
lrot occlu in the variabie speed range of the cornpressor.
The dvnamics of a rotor can be studied r,vith the help
ot' the computer and the effect on the rotor of operative
Lrnbalence ch-re to build-up or ntisalignnrent c:1n be eval-
LLateci. Tliese rotor r-rnbalanccs r,vill load the bearings.
CiompuLr:r analr'sis allorrs the engineer to predict
beering loadings encl to desiqn a dependable m.rintenance
free rnacirinc-
Iig. 5 shous the c.-Llculated and neasured rotor re-
Fig. 4-Large turbine-driven centrifugal compressor.
sponse curves for an eight-stage compressor rotor. The
measured values were obtained first during the mechani-
cal test of the compressor. Although the vibration level
was less than 0.7 mils and the bearing forces were below
the design dynamic load limit, the steepness of the vibra-
tion curve near the maximum operating speed was under-
standably cause for concern. The rotor was modified and
These results are shor,vn in Fig. 6. The shop test shows
quite low vibration amplitudes and low bearing loading.
The compressor, once in acual service, u,ill become un-
balanced due to build-up on the rotor. The diflerence
betrveen the calculated and measured response curyes
shows that this compressor will be tolerant to considerable
rotor deposts before it r'vill have to be cleaned. Lightly
Ioaded journal
bearings, such as those used in compres-
sors, can be unstable at high speeds, and a number of
solutions to this problem have been used. The tilting pad
bearing is widely used in compressors. Each shoe tilts
independently to maintain its load carrying hvdrodynamic
pressure wedge. Extensive service in many types of com-
pressors have proven the dependability of this bearing.
Off-design operqfion. Most compressor users take
trouble-free mechanical operation for granted, but the1,
are concerned with compressor performance. Figure 7
shows a typical compressor performance curve. Uncom-
plicated and trouble-free operation can be expected in the
stable performance region to the right of the surge line.
Surging, or unstable operation, can occur in any cen-
trifugal compressor when
,the inlet florv is reduced to
approximately 6O/o of the design inlet flou, or lower.
Compressors that produce large pressure ratios, ratio of
inlet pressure to discharge pressure tend to have more
violent surges. When the compressor is operated re-
peatedly or for prolonged periods of time ir.i surge the
pressure forces can damage the internals of the machine.
For those applications where frequent surge operation
can be expected the compressor internals should be made
of steel, instead of the more common cast iron n-raterial.
When the comp.ressor is operated in surge continuously
it absorbs approximate\ a)/o of the rated horsepower,
however the flow thru-put is greatly reduced and under
some conditions stops completely. The power required to
drive the compressor in surge is therefore largel1, converted
to heat. This causes excessive tempe.rature build-up inside
the compressor and in a matter of seconds, if the condi-
tions are severe enough, can melt the soft labyrinth seals
which control the intemal leakage. The compressor per-
formance suflers once the seals are damaged and the
machine must be opened and the seals replaced to restore
it to the original condition. Excessive temperatures in a
compressor having a balance drum labyrinth seal made
o[ a soft material with a Iow melting temperature can
melt the seal. This will upset the rotor thrust balance
and overload the thrust bearing. When the thrust bearing
fails the rotor will shift axially and the impellers will
rub against the stationary parts causing further damage.
To help avoid these operating p.roblems the compressor
can be provided with a high temperature balance drum
seal made from compressed metal fibers that will with-
stand several times the normal operating discharge tem-
Another safety feature that can be employed is a
sfEEo. r(tro nPil
Fig. s-Eight-stage rotor /esponse curve.
0 f 2945 7I910
Fig. 6-Rotor response curve for modified eight-stage rotor.
110% SPEI
ll{LEI vtLLmE +
ufilTtil,t 0f st EGE
Flg. 7-{ompressor performance
high temperature switch located in the balance drum
leakage pipe. It is wrong to locate this switch in the
discharge pipe. In this location the switch does not pro-
tect the compressor since there is not sufficient discharge
flow to carry the heat to the switch when the compressor
is operated in surge. The balance drum leakage pipe is
the correct location for this protective device. There is
leakage flow in this pipe even when the compressor is
being operated completely shut-off.
Normq! mqintenqnce ilems. It is reasonable to expect
trouble-free operation for periods up to tw,o or three
years between internal inspections for heavy duty com-
1.4 r
,.0 l-
o.N I
0.6 i
t.6 r
0.,1 r
+ MEASURED wrrr ruon-comect H
250 yr
mercial machinery. I{owever, some applications where
the gas stream being compressed is extremely dirty and
internal washing cannot be used it may be necessary to
shut down in order to clean the compressor internals to
restore full flow capacity.
The internal seals that prevent leakage around the
impellers are normally of the labyrinth type. They con-
sist of a series of circumferential knife points that are
positioned closely to the rotating impeller. In order for
the compressor to maintain the design performance these
knife points must not be damaged by rubbing, erosion,
corrosion, or plugged-up with foreign matter. These seals
are normal u.earing parts and spares should be main-
tained in anticipation that they will need to be replaced
after an extended operating period.
Labyrinth seals are used in compressors with various
design features that will extend their life and make
Fig. 8-High-speed balancing machine.
them less susceptible to damage. One popular version
is to machine the knives in the rotating part and to posi-
tion a stationary sleeve of a soft material around the
points to obtain a seal. The clearances between the rotat-
ing and stationary parts of the seal can be reduced be-
cause it is intended that the rotating points cut grooves
in the adjacent material under normal operation. This
type seal allows the compressor to operate with higher
efficiency due to reduced leakage and longer life because
the thin knife points are made of steel and resist erosion.
Also in this design considerable radial rotor motion can
be tolerated without altering'the effectiveness of the seal.
The stationary part of these seals have been manufacured
in babbitt-lined steel, aluminum, non-metallic compounds,
and compressed steel fiber materials. Each has its own
advantages when considering the particular operating
environment of the compressor.
In many compressors labyrinths are also used for the
shaft seals. Where the small leakage allowed by labyrinth
shaft seals can not be tolerated, either the oil fiIm or the
mechanical contact shaft seals can be used.
Compressor shaft seals of the oil barrier type present
potential maintenance and operational problems. A recent
study of a large cross-section of cornpressor users showed
that approximately
of. all compressor failures and loss
of production was due to malfunctioning oil film type
seals ! Regardless of the cause, whether it was design,
operation, or maintenance it points out a particular com-
ponent of the compressor that commands respect.
Oil film seals consist of basically two stationary bush-
ings which surround the rotating shaft with a few thou-
sands of an inch clearance. Seal oil is introduced between
the bushings and leaks in both directions along the shaft.
"O" rings in the seal housing prevent leakage around
the outside of the bushings. The seal oil is maintained at
some pressure higher than the gas pressure inside the
compressor. The differential pressure across the inner
bushing is usually only a few pounds per square inch to
limit the amount of inward oil leakage. This leakage is
collected in a leakage chamber that is separated from
the gas stream by a labyrinth seal and is drained away
through a drain trap. If the gas being compressed con-
taminates the seal leakage, the leakage is discarded" The
outer bushing takes the total pressure drop from seal oil
pressure to the atmospheric drain.
Oil film seals can cause many diflerent types of
operating and maintenance problems. The most common
is excessive inward leakage due to increased clearances
between the bushing and the shaft. When these clearances
become large enough the leakage chamber becomes
flooded and oil spills over through the labyrinth and
enters the compressor. This malfunction can cause many
operating problems ranging from the nuisance of having
to continuously add oil to the reservoir to contaminating
the main process gas. The bushing clearances can in-
crease in time due to corrosion or erosion of the inner
surfaces of the bushing, wiping due to radial shaft vibra-
tion, or dirt in the seal oil.
Another shaft seal, which is not as common in process
compressors, is the mechanical contact type. This seal
has the advantages of being able to maintain low inner
leakage rates with higher oil-to-gas differential pressures
and therefore makes the seal oil pressure control system
simpler. The seal has a spring loaded carbon face that
runs against a face on the seal collar. The complete shal't
seal either consists of two mechanical contact seals in a
back to back arrangement or a combination of the me-
chanical contact seal on the gas side and a bushing seal
on the atmospheric side. These mechanical contact seals
have been used for many years to seal against pressure
up to 1200 psi in natural gas service. They presently are
being tested for pressures as high as 2500 psi differential
across a single sealing face. The mechanical contact seal
has an added advantage when applied to high pressure
applications in that the radial sealing faces have a mini-
mum effect on the rotor dynamics, whereas oil film bush-
ing seals can lose their free floating featrue and cait
upset the stability of the rotor when operating at high
High speed compression machinery must be properly
balanced, especially when it is designed to operate be-
tween the Ist and 2nd lateral critical speeds. The correct
method is to first dynamically balance each impeller
on an arbor, and to check the rotor balance as each
impeller is installed on the shaft. The runout of the shaft
must be watched closely as each of the impellers are
shrunk on the shaft. Abnormal change in shaft runout in-
dicates that the impeller is not square on the shaft and
must be adjusted before checking the balance. \\'hen the
balance check indicates that an unbalance exists, the correc-
tion is lnade to the last impeller that rvas mounted. This
procedure when strictly follolved can ptoduce a rvell bal-
anced rotor even tvhen the balancing is clone in a lou--
speed balance machine. Some users routinely check the
balance of complete rotors and make corrections on the
first and last impellers. This practice is lvrong for flerible
rotors that have more than three impellers. The onl1, sure
rvay to balance a completely assembled high spced rotor
is to nse a high speed balance machine. The rotor un-
balance rnust be checked throughout the operating speed
range and any corrections that are made to the rotor must
be macle at the plane of unbalance. N'Iost balance shops
in this country have balancing equipment that operates
below 1500 rpm. The photograph in Figure B shorvs a
balance n'rachine that will balance a 1000 lb. rotor at
speeds up to 7000 rpm. The machine r,vill handle most
high speed flexible compressor rotors.
Experience has sholvn that high speed compressor
rotors can be successfully balanced in a lorv speed bal-
ancing machine when the progressive impeller by in-r-
pcller method is followed. The compressor is normailr'
subjected to a mechanical test in the shop before ship-
ping. During these tests the vibration amplitudes and
frequencies are measured. When the compressor vibrates
excessivelv the rotor is rebalanced in the high speed
balancing l'iachine, hor'vever this is normally not required.
When a spare rotor is purchased or repaired after the
compressor has been installed the rotor should be given
a high speed balance check in order to prevent delavs
later in the users plant.
Thc alignment betrveen adjacent shaft ends of all
rotatinq equipment in the compressol train is very inr-
portant to the operation and maintenance of the equip-
nent. -\11 compressors, gears, motors, turbines, etc. havc
some toielance for misalignn-rent. IIow-ever, except for
rotor balance, misalignment is the most frequent cause
of running problems. Excessive misalignment can force
the rotating equipument to vibrate and shorten the life
of the bearings and gear tvpe couplings. When misaligned
shafts are rotated, the gear teetli in the coupling must
slide back and forth on each other. This causes bending
lxoments and forces to be imposed upon the shaft ends.
The shaft end rvill fail if the conditions are severe enouglr
or the coupling hub rvill come loose on the shaft.
\,{ost couplings are mountcd on a tapered shaft end.
It is con.imon that a key and keyway be provided to
transmil tite torqne through the connection. The coupling
hub is prrshed upon the shaft taper by a nut r,vhich also
prevents it from coming ofl the taper during operation.
This trpc of coupling hub mounting is satisfacton, for
lorv speed-1or,v horsepor'ver applications but can be a
limitin,r factor for modern high speed machinery. A much
better rnethod is to shrink the coupline hub upon the
shaft taper- using hvdraulic pressure and jacking
as shorr-ir ir.r Fignre 9. This type of mounting allows for
at least trrice as much torque to be transmitted through
the sarne size shafts with the same stress levels. This is
possible since the key is no longer needed and the key'way
st.ress concentration is eliminated. The misalignment
forces frorn the coupling teeth rvill not loosen the cou-
pling hub on the shaft because it is fitted on the shaft
taper rrith considerable interference. To mount a colr-
fouilflilO TOOLT
Fig. 9-Hydraulically-mounted coupling and tools.
Fig. 10-Compressor case with sliding mounts.
pling hydraulicallv, the shaft end is drillccl to pr.ovidc ar.r
oil passage for pun-rping high pressure oil betwccn tlrcr
shalt taper and the coupline hub borr:. A liyclrarrlic hanrl
pump is connected to the shaft enr.l throrrsh an aclapter'.
hLrb is strctcl-ied u,itli hyclraulic pressurr: ancl the
scre\vs (Figure 9^\), are used to push thc i:orrpling hrrlr
t1p on the taper a specifiecl distance.
jnr:k scre\\'s
hold the hub in the correct position r.vhile thc oil prcs-
sure is released. After the oil has drained the nrounting
plate is remor,ed, a nr-rt is rnourted on thc shalt end.
(Figurc 98) to protect thc thrcads. To rcnrove thc
coupling hub the procedure is reversed and the hub is
"popped-off" against the tools with hyclrarrlic plesstrrc,
type sf s6upling n-rounting pro.,,icles Ior lrr casill'
mounted hub that will transn-rit as nrur:h torclLrcr ns tlrc
sl.raft material u.,iIl allow.
Good shaft alignment mr.rst be mairtairccl uncler dr'-
namic conditions if trouble-free operation is to be aclrievetl.
To accomplish this the thermal srowth of the incliviclrral
elements must be taken into account when aligning com-
pressor equipment in the cold conditions. |or instance,
the diameter and length of a compressor case will increase
clue to the heat of compression.
Normallv the casing is supported at the ]rorizontal
centerline to allorv the case to
grow irr cliameter rvithor.rt
changing the position of the shaft.
mountins feet
on one end of the case are bolted and doweled to the
foundation, as shown in Figure 10. The opposite cnd of
the case must be allowed to move axially as the casing
rloul?aD couPLlro
Fig. 1l-Compressor case with centerline support.
Aboul the
Hucs M. D-q.vts is tlLe manager of Cen-
trifugal Cotnpressor Engineering De-
prL,rtnrcnt, DELAVAL Turbine Diaision
of DELAVAL Turbine Inc., Trenton,
N.J. He is responsible
or th,e organiza-
tion and direction of all engineering and
related, to tlte prod-
uct. Mr. Datsis receiaed a B.S. in ME
Rose Polytechnic Institute,
Terre trIaute, Ind., in 1956 and, assumed
ltis present posi,tion iru January 1970,
provicles centerline sllppolt for thc case, has four mount-
ing pads that are all boltecl and dor,veled to the founda-
tion, and requires no maintenance. The ruggeciness of
tiris support n'ill allou, for consideraltle extelnal forces
to be cxcrtcd upon thc casing rvithout changing tlie shaft
lnternql configurqtion. The process market places very
clenranding arLd ever changing requilernents upon the
selcction and alrangcrlent of the compressor inteJnals
end erternirJ casing nozzle configuration. Centrifugal
coiuplcssol' selections should be made rvith the ease of
o1-relation ancl rlaintenance aspects in mincl as well as
tire cornpressing requirerlcnts. Figure 12 shows the most
allangernents of tlie cornpressor internals and
ouLcr casirg in thc folrn of simplified diagrams.
f ire opelai-ion and mailrtenance aclr,antages are listed
fol e:rch as follows:
i. Sing-le corrl)ressor- bocly instead oi tlvo or more
result,s in a sinrpler s)'stem.
2. Hot dischage at center of case to reduce lubri- :Lnc1 oil scal problems.
3. Reduceci po\ver- r'equired to cornprcss gas results
in a srnailcr ch'iver.
4. Back to bacli irnpellcrs rcduce natural rotor thrust
and nllorrs iol mole internal seal \\,ear befole overload-
ing thrust bearing. Increase tirr-re betrveen overhauls.
5. Cold inlet at center of case to reduce iubrication
and oil scal problerns.
[i. Srnaller colnprcssor anci higher specd to do the
sarDe compression
job. Rcduced foundation ancl train-
tenancc problems.
7. Singlc inlct to better suit ertcrnal piping arrange-
B. Single discharge to better suit external piping
9. No external balance piston leakage pipe. Com-
pressor can tolerate increased balance seal rvear r'vithout
upsctting thrust bal:rncing s1'stern and over'loacling thrust
10. Hot or colc1 sections oI case arc adjacent to reduces
thermal gradients and distortion of the cese. Makes
alignment easier to achieve.
Origiually prcsented at the lst Tcxas A&lV{ University Turbomachincry
Slmposium, October 1972, College Station, Texas.
API 617-Centrilugal Compressors for General Refincry Service, Second
Edition. 1963, Amcrican Petroleum Institute, Washington, D.C.
oNE 0(n.lrs
@ ll
$Enoil lll0ss
sl,cf,fll $r
fil0 r::,LtNG
r 1)lYtfai !Y'o-l
8Eftlsl R.Oil
Wrll OOU8(!
frov{ rIlLEf
Ar.r0 sr0l sTREril
Fig. 12-Eight-compressor case and nozzle configurations.
expands. This has been accomplished in the traditional
compressor mount with holddown shoulder bolts that
allow the casc to slide axially upon lubricated shims but
limit the vertical movement of the case. With this ar-
rangement a ve.rtical key and keyway are located between
the case and foundation on the vertical centerline to
prevent transverse movement of the case. This mounting
requires a special founda.tion to support the vertical key
and also regular lubrication of the shims.
The compressor case mounting shown in Figure 11
allows for thermal growth of the case in all directions,
lmprove machinery mai tenance
Reliable operation ol modern
turbomachinery requires up-to-date
mai ntenance technrgues.
Here ate some ideas tor improving
your maintenance procedures
W. E. Nelson, Amoco Oil Co., Texas City, Texas
MaNtrrecruRE AND MATNTENANoB of turbomachinery
are completely different. The first involves shaping and
assembling of various parts to required tolerances while
the second involves restoration of these tolerances through
a series of intelligent compromises. This is the crux of
maintenance techniques-keeping the compromises intelli-
gent. The
industry has pushed for "bigger and
better" turbomachinery until
problems have
become tremendous. We are literally "snowed under" by
these problems. The failure to provide adequate feedback
of reler-ant operational troubles into the design phase is
the greatest problem facing the turbomachinery industry
This lack of
results in much turbo-
machinery designed with too little regard for the operating
and maintenance complexity created. Many of our main-
tenance techniques are, in a word, inadequate to cope
with the troubles we encounter. The mechanics in our
various installations are not, in general, technically com-
petent enough on the complexities of our equipment to
adequately maintain it under unit operational pressures.
The growth of contract maintenance firms has not been
sufficient to fill this void. The ability of the original equip-
ment manufacturer (OEM) to provide technical service
has also frequently been inadequate. Perhaps some of the
pnocedures used at Amoco to combine the best of our
capabilities, those of the OEM, and those of specialty ser-
vice organizations to meet our problem will be of interest
to others facing similar situations.
The Amoco refinery at Texas City is forrltlr lareest in
the nation in crude capacity. Ilt:causc of the sevelitl' o[
some of our reforming plocesscs,
of 60 nrcgl-
u'atts of power, and large ammonia production it larrks
higher in complexity of cqrripment. Much oI our equip-
ment is less than ten years old. It is sophisticatecl, lalse
arrd complex. Since modern design trends tolvlrd sin.-le
train equipment, we have some of the largest cqrripnrerrt
made. The rapid growth of tliis refinery has createrl tle-
mendous pressures on our nrainLenance pcrsonnel and
experience has become extrcrnely scarce. At the
time over two-thirds of the macliinist hourly
har-e under six years experience in the rnachinery ficld.
Having problems, we have bccn forced to find solutiotrs.
These solutions can be diviclccl into forrr brrsic categorit's:
1. Training of personnel
2. Tools and equiprnerrt
3. Replacement parts
4. Reliability improvement projects
\Ve believe that training must be the central themc to
the solutions of our problems. The days of the rucchatric
arned with a ball-peen harnrner, a screwclrir,'cr, ancl a
crescent wrench are gone.
More and more conrplicated
maintenance tools must be placed in the hancls o[ tlre
mechanic and he must bc trained to utilize thern.
People must be trained, motivatcd, and directed so tlrat
thei' gain experience and develop, not into meclranics, but
into highly capable technicians. Good training is cxpensivt:
but it yields great returns. Machinery has srown more
complex, requiring more knowledge in many areas.
olcl traditional craft lines must yield to the rrraintenance
needs of complicated equipment. A joint eft'olt by crafts-
men is necessary to accomplish this.
Bqsic mqchinist trqining. To solve otrr ploblenrs rve
have embarked upon ambitious training proglrnrs. Since
Fig. 1-Models used to teach reverse indicator alignment.
1967, over 100 nerv machinists have completed an intensi-
fied training prosram involving 800 classroom hours of
instmction in machinery principles and concepts. Ntlost of
this training has been developed and conducted by in-plant
personnel. It has been expresslv tailorecl to our equiptrlent
and is highly detailed. Training rnust be carefully planned
and adrnirristered to fit the requiren'rents of each situation.
At present u,e have a full tiurc training Staff of two people
in our maintcnance division. The responsibilities of the
group range across all crafts. These machinerv related sub-
have beren covered recently:
Rcvelsc indicator :rlignnrent seminars (see Fig. 1)
Gas turbine overhaul
Mechanical seal ltaintenante
'a rideo-tape n'as de-
velopcd in orrr corporate train;ng grouP.
Prqcticql froining. Aftcr the machinist receives his funda-
mental training, his on-thc-job expcrience should continue
his training and test his skills. For this reason \ve attempt
to clo as much repair work u'ith our own people as pos-
sible. Each man is rotated among various
jobs to accelerate
his learning and development and is as familiar with a
Iarge cornpressor as he is with a small pump. Over half of
our shop personnel can operate a balancing machine ex-
pertly. The remainder havc worked on the machine and
arc farrriliar with its o1;eration. The spreadins around of
the hardest jobs clevclops nrore competent people. If rve
restrict a man to onc type of rvork, he r,vill probabl,v be-
come expert in tliat arca but his curiosity, u.hich rve feel
is a prime nrotivator, will er.'entually fadc.
\Ve cncouragc the participation of the machinist in the
solr.rtion of difficult problerns. This policy often causes the
machinist to seek out information on his orvn. References
to API spccifications are not uncommon on our shop
floor. To encourage this type of participation we have a
library adjacent to the shop floor with filed drawings,
written histories of thc equipment, catalogs, and other
literature pcrtinent to thc machine maintenance field.
The "librarian" and "editor" of the weekly summarv of
repair histories is an hourly-paid machinist. The job is
rotated periodically to develcp interest. FIe also alters
equipment drawings and service manuals as revisions and
changes are made.
Updote lroining. In order to pass on knowledge and to
update the mechanic, unusual experiences or special tech-
niques are rvritten up and distributed to all machinists
frequently. The ground rules of this "nervspaper" limit it
to two typervritten pages and use of reduced size sketches
andf or drawings is encouraged. In addition, the machinists
are given copies of all machinist-oriented issues of a
similar training newsletter issued by our Engineering and
Technical Division. We feel that knowledge breeds more
Mqnufqclurer trqining. Our personnel are sent to
manufacturer-conducted training sessions. As an example,
a number of our people, including first-line foremetr, have
attended the very excellent maintenance school conducted
by the manufacturer of the six gas turbines we oPerate.
This is a very effective effort on the part of the OEM to
provide maintenance instruction on rnachinery to the cus-
tomer. We would like to see more along these lines.
trqiners. \Ve look upon nanufacturers'
sen,icemen as trainers. Their primary reason for being in
our plant is to train our people. Because of thesc t,iervs,
our standards for this type of u'ork arc very high. \Ve
attempt to get lasting value frotn their services. Lrnfortu-
nateJv, the equipment manufacturer frcquently suffers
from the san-re problems that 11's 121'g-2 shortrrge of
capable people. For this reason rve train our own
as much as possible. We havc one loreman rvho is ottr
in-plant serviceman. He supervises no hor-rrly personnel,
but acts as a consultant to maintenance
jobs. A three-man
technical group devoted to turbornachinery rvas establishcd
in 1971 to further our internal capabilities.
!nslruclionql books. Manufacturers' instruction books
are often inadequate. \\Ie have resorted to writing main-
tenance rnanuals for the mechanic on such subjects as
rnechanical seals, vertical pumps. hot-tapping machines,
and more recently, gas turbines. The gas turbinc overhaul
manual consists of :
Step-b)'-step overhaul procedures devcloped largely
from the manufacturer's training school mentioned pre-
Almost a hundred photoeraphs illustrating the step-
by-step proccdures using one of our gas turbines
An arror'v diagram showing the sequences of the pro-
This r'vas bound into a book form and used as thc basis
for a four-hour training session to our supervisors and
mechanics that combined the best of the experiences of all
our people.
Detailed drawings are developed to aid in nrainterrance.
Our first large scale effort involved a contact seal assembly
which was dcvelopcd after rve held up thc operation of
our giant catalytic cracking unit because the "t1'pica1"
dimensionless drau,ing supplied bv the OEM rvas not ade-
quate to correctly assemble thc compressor seals. Many
other assembly drarvings have been der,eloped since that
first effort.
Many maintenance tools should be made available to
the mechanic. We attempt to do as much specialized ma-
chine work as economicalll,feasible. Fig. 2 shorvs a cooling
water pump being machined to replace a suction flange
that r.r''as broken. Doing jobs like this keeps us technically
compctent enough to tackle other jobs. The catalytic
cracking unit air blower turbine nozzle blocks shown in
Fig. 3 r^"'ere fabricated in our shops in eight shifts because
a ne\v one could not be delivered in under three months.
The "homemade" nozzles have been in service for several
months with no loss of efficiencl'.
In-plant balancing capability is one of the most valu-
able tools in a petrochemicals plant. Our 4,000-pound
rotor capacity balancing machine has had a very signifi-
cant effect in development of know-how. Until about six
years ago, we relied entirely upon outside balancing fa-
cilities and only balanced the more critical equipment.
After purchasing balancing equipment, we began to prac-
tice on equipment that l,ve once felt did not require dy-
namic balancing. Vast improvement of pump seal lile as
rvell as bearing life on some of the smaller equipment was
achier-ed. The static balance of pump impellers is woefully
inadequate for long runs and reliable service. In brief,
the balancing machine paid for itself in balancing pumps,
motors, and small turbines that r,r,,e once thousht did not
need dvnamic balancing. The d,vnamic balancing of im-
pellers of ventical pumps is cspecially advantagcous in re-
ducing maintenance. The long'. slender shafts are highly
susceptible to any vibration induced by imbalance. Thc
man familiar rvith the plant equipment tends to balance
closer if the machine's operational history has proven to
have a balance sensitive rotor', another advantage. This
sparks ihe development of that much sousht-after knou,-
fechniques to check balance gear-t)'pe couplings for the
large high-speed con-rpressor and turbine drives as a unit
were der-eloped. This led to the solving of many vibration
problems. High speed couplings are routinely check-
balanced norv.
Spare parts problems are inherent in the machinery
maintenance business. Cost of replacement parts, long
deliverv times, and quality are problems everyone has
faced. Amoco, at Texas City, undsylook to combat this
problcnr by forn'ring a spare parts group to consolidate the
spare parts activities for its expandins rgfinsly on a com-
panv-rvide basis. We believe in stocking spare parts: our
present inventory is over 16,000 items, including 75 com-
plete rotors. The field of spare
arts is changing very
rapidlv and is much more complex rhan ir the past. \{any
pieces of equipment on process units are made up of
unitized components from several different vendors. The
traditional attitude has been to look to the packaging
vendor as a source of supply. \{any vendors are refusing
to handle requests for replacement parts on equipment
not directly manufactured by them. More and more spe-
cialty companies are entering the equipment parts busi-
ness. Sorne are supplying parts directly to OEM companies
for resale as their "own" brand. Others are supplying part5
directll to the end user. The primary function of our spare
parts group is to develop multiple sources of supply for as
manv parts as possible. Gaskets, turbine carbon packing,
and mechanical seal parts are purchased from loca1
sources almost entirely. Shafts, sleeves, and cast parts such
are becoming increasingly available from local
We have found some OEN{'s are altering their spare
Fig. 2-Versatile tools develop skills.
Fig. 3-Shop-made nozzle block in semi{inished condition.
parts system to improve servicc clue to this local compcti-
tion, definitely a bright spot in thc
High naintenance costs and lorv operatinlr reliabilitv
go harrd in hand. Usually thc lorv rcliability is a (rceter
economic factor than the high rnaintenanr.e r:osts. Ahnost
one-third of the unschcduled (ancl costly) shutdor,vns irr
our refinery are caused by machincrv failures. Machine
"revamps" and alterations h:rve been necessary to inrpror-c
reliability. A brief discussion of somr: of this tvpe of rnain-
tenance r,r'ill give a feel for some of thc benefits of our
training, spare parts prosrarn, ancl our major esset-
Governor redesigns. Due to a long history of failures
and poor performance u,ith our four process sas turbine
governor control systents, rve undertook to redesisn these
systems using state-of-the-art electronics and "plug-in"
concepts for ease of maintenance. f'he first s1,51s111, dubbcd
"Turbotronic," was installed in 1969. Since tliat time two
other gas turbines have been converted. All of these iu-
stallations have been highly successful in that rnaintenauce
has been minimal and is usually accomplished on-stream.
Also, turbine performance, speed control, and flexibility
are greatly improved. The original design has been supple-
mented to include a self-contained alarm system, a sen-ri-
automatic sequential start system, and a con'rplete trip and
protection system as r,vell as the elcctronic controls. Our
eost oI this svstcrn is considcrably less than the cost of a
sinrilar device olTered by thc OEM on nerv ruachines. The
controls for the fotrrth and final conversion are due for
installation in laic 1973. The original installation r.vill be
dismantlecl ancl an trpdated version installed at the same
timc. Toial horscpower involvcd is alnrost 80,000.
Witlr thesc successful installations, rve turned to stcam
turbines. A conrprcssor clrivc on oLrr most vital process unit
lracl been plaaucd bv sovernor dri"'e reduction gcar prob-
lerns. The tLrrbine speed (approximatcly 9,000 rpm) ivas
rcduccd fc.rr
drive via a three-shaft gear train.
\/ibration induced by the poor design limited the maxi-
rnurn rlrn to about thrcc nionths. The \\roodrvard gover-
nor and the
were rcmovcd and our electronic
sovclnor installcd. This governor n'as built, installed, and
test run in foLrr days. Since the installation in May 1971,
vibration Jras been minimal and governor performance
has be<:n cxcellent.
To date, eiqht more nrulti-stage steam turbines driving
pro( ess conlprcssors with a total horseporver of 58,000
havr: been converted. These electronic govcrnor installa-
tions agair-r are less cxpensive and rnore effectivc (in or-rr
opinion) tlran any conrmcrcially available product.
Beoring conversion. Rcaring failurcs rank high as callscs
oI rotaling equipment failrrres but bearing design has
Iaggccl belrincl as rotational speecls hare risen. Parameters
fol bearing design are u'ell defined by Herbage' and
Abrarnovitzr. It has becn dernonstrated that the reliabiliti,
ol rotatine ccluipn'rcnt bc increased through the use
oI bcaring dcsien irnprover)rent. Dcsign modifications fa1I
irrl,r tirret' nre.j,,r' 11 p65,
1 . Tlier clir ngins of taPcrcd Iancl thrust bearings to
tiltirre pacl thrust bcarings rvith leveling links: Experi-
encc orrrs ancl othtrrs
(:t, 'tl-in
dicates that for equal
an'as. irnprovcments in loecl-carrying capabilitv can be
ac hievccl. Srrclclen lo:rd surses and liquid slugs rvill occur.
Thnrst lrcarinss r.r.liich do not have the ability to absorb
such lo:rcls r.r,ill Iail; tht:rclore, rvc fcel rve should have
thc nrarinrrrtrr possiblc thrrrst c.apacit1,.
2. Changine radial bearinss fr:orl cvlindrical and/or
prcssrrre prcl babbittcd slccve t,vpe to tiltina pad L,earings.
1-his irrrProves rotor stabilitr', elirninatcs chances of oi1
uhill rnd irrproves srrrlivll oI jor-rrnal bearings under
lrtor: inrb:rlance and dtrring <:r'itical speed drvclls. Present
state-of-the-art indicatcs thrrt the tilting pad bearing is
scnerallv tlrc lrest in bcaring clesign because of self-stabil-
izins charac tr:r'istics.
3. A matcrial chause of tilting pad
and thrust
Aboui lhe oulhor
W. E. Npr,soN is Manager of Mainte-
nance Seraices at the Amoco Oil Com-
pnny refinery in Teuas CitE, Teras. He
has been uith Amoco
or nineteen gears
i,n oarious enginecring, purchasing, a,nd
tnai,ntena,ncc assignments. " E d" rece'ia ed
o, B.S" degree in Mechanical Engineet'-
Teras A&M Uniuersitg in
1951 and, is a Registered Professi,onal
Engi,neer in Teuas.
bearings. These type bearings are generally supplied with
steel-backed babbitted pads. Changing to copper alIo,v-
backed, thin-babbitted pads conducts heat away from the
babbitted surface at a faster rate than the steel'backed
shoes, improving load-carrving capacit,v. Experience indi-
cates that a 50 to 100!t load-carrl,ing improvetrtent can
be obtained.
Over 150,000 horseporvcr has been ccnverted since
late 1969. Two dozen more machines have been recom-
mendcd for conversion. We do not believe we should \vait
for a failure before deciding to change because of the
process penalties resulting from a bearing failure. The
conversions have been done using the services of some of
the best bearing specialists in the country. Amoco partici-
pates in the fundamental design by supplying basic dimen-
sions and some machine peculiarities. One equiltment
manufacturer has begun to offer bearing upgrading "kits"
for the machines alreadv in serr,ice. Again a bright spot
in the over-all picture.
Rotor redesigns. \\/e havc found that many steam tur-
bines of stacked rotor design are built by the manufacturer
r.r,ith substantial imbalance couples. It has proved advan-
tageolrs in manv instances to unstack a rotor, indil idually
balance each lvheel. and then proeressivell, balance the
rotor as it is reassembled, the rvay most exPerts agree
it should be done. Manr machines, once highly sensitive
to vibration, have been t'trade lttore reliablc by tliis tech-
nique. In addition, the shrink of the rvheels on tlie shaft
pror,ided by thc manufacturer olten proves to be inade-
cluatc for the steam operating temllcratllres. This allows
the r'r"heerls to move on the shaft. Shafts have been rede-
signed to pror-ide locliinq rings for each rvheel.
Six rnachincs of one make havc been rerlanufactuled
to date. Continual impt'ovcrnent and ulldating of our ma-
c.hiner,v is necessary to maintain long runs. We can no
longer "fix like r'r,e did that last time" but must colr-
tinually strive for newer. bctter r'va1's to make repaiLs. T'he
rnachinist is often called upon more frequently to rl-rake
revisions than he is to "fix" a machine.
are no simple solr-rtions to the maintenance dif-
ficulties we face. The principal ingredients of strccessful
rnninlgnance are training, eqtripment,
parts, and people (both hoLrrli' and technical) . T'hese
factors must be couplecl ',vith a determined managenent
if rve are to move ahead and find solutions to thcsc
Iems. Our experience has been that each of the foLLt' areas
contributes to our solutions. A11 are nccessary.
Today's process
cannot be dependcnt ul)on anv
one source for solutions to its troubles. Original equipment
manufacturcts, contract service shops, contract trtainte-
nance companies, and specialtv companies can only sup-
plcment the basic skills that tuust be developed rr-ithin
orlr o\\ n organizations.
prcscnted et the 2nd Annual Texas ,A.&tr{ Turllornaclrinery
Sympo.irm, October 23-25. 1973. Col)ege Stati.n, Texa..
Why clean turbornachines onstream?
Cleaning turbomachinery while in
service is tricky but the benefits are
signiticant. Here's how to clean your
machinery sately and eftectively
Briqn Turner, Imperial Oil Enterprises, Ltd., Sarnia, Ont.
"ON-srneau CleeNrNc" is defined as the periodic
removal of accumulated deposits while the equipment
continues in service. This definition excludes from the
discussiorr the various techniques of preventing the de-
posit accurnulation. It also excludes those cleaning meth-
ods rvhich do not require disrnar.rtling, but do require that
the machine is either idling or stopped.
The paper cliscussed the rcasons for cleaning, fouling
indicators and abrasive and solvent cleaning techniques.
Details of turbine washin_q are used to illustrate the appli-
cetion r,I solverrt cleaning.
Why sleon? There are at least three reasons for "on-
streanr" cleaning.
The first is to restore the system capability. If the unit
is a driver, it's naximum horsepower wiil probably drop
as it becomes dirty. Cleaning will restore this limit. If
the machine is a dynamic corlipressor, the fouling may
have reduced its head, and therefore, the maximum gas
flcw rate. Cleaning r,v'iil restore the capacity limit.
The second ieason is to increase the machine's effi-
ciencv" fn most, but not all cases, fouling rviil increase
the frrel or power required to do a certain task. The
mecha:-rism is tharl the deposit changes the flow contours,
R.emoval of the deposits u'iil restore the originai nrcfiles
and thr efficiencv.
The tl:ird leason for cleaning is to prevent failures
due to abnormai operating modes. Fouling on the rotor
blacies of stearl turbines can cause ttrrust bearing failures.
Deposil-s on steam turbine governor valves and trip and
throttle valves are suspected of causing overspeed failures.
Foulins in balance piston labyrinths and in balance lines
has caused thrust bearine failures in centrifueal rnachines.
Any rotor deposit can cause vibration due to unbalance
if it is not laid down uniformly or if it sluffs off non-
uniformly. There could be other similar effects which
will cause failure of the unit.
Fouling indicqtors. A prerequisite of a cleaning pro-
gram is some kind of fouling detection system. Naturally,
thrs system must cover the prime reason for cleaning, If
the machine is a gas turbine, then the prime reason mav
be horsepo'rver capability or it may be efficiency. On
the other hand, on a back-pressure steam turbine, the
prime reason may be either horsepower capability or
thrust bearing protection. On a centrifugal compressor.
the prime reason for cleaning may be to restore capacitl-,
to improve efficiency or reduce thrust loading.
The selection of a fouling detection system will be
strongly influenced by the safety and complexity of clean-
ing procedure. For example, the procedure may be to
throw 10 pounds of rice into the suction of a gas turbine.
Or, it mav involve injecting a quart of water into a
mechanical-drive turbine, with a 30oF super-
heated inlet. In either case, ihe risk of damage and the
lnanpower required is so low that no monitoring can be
justified. The cieaning should be frequent and routine.
On the other hand, the cleaning may invoive renoving
3000 of superheat from 200,000 pounds/hour oi sieam
entering an eight-stage turbine. This is a much more
compiex case. If the turbine really is dirty, and is not
washed, you may end up u,ith a nreck. Alternativelr..
to wash, especiaily the firit time, ma1, result in misa.iign-
ment due to piping stress, water slugging, ioss of clear-
ance due to differentiai contraction, vibration due to non-
uniform cieposit removal, or thrust trearing failure. This
case obviolslv calls .trr a reiiable indicator.
Irouling indicators ir"hich have been
Gas turbine exhaust temperature
Steam iurbine stearn cirest Dressitre
used inch.rcie:
(singie governor
Steam turbine P1 pressure (multi-valve)
b and c above modified to correct for steam flow
The exponent n-1/n on a compressor or gas turbine
where k is either known or is relativell constant
The exponent n-1/n in one section of a machine rel-
ative to another section handling the same gas
The pressure ratio in one section of a machine rel-
ative to another
Thrust loading or thrust bearing metal temperature
Balance line to suction differential pressure.
Cleoning techniques. There are two basic approaches
to cleaning. These are abrasion and solvent cleaning.
Abrasion is the simplest of the trvo methods, but is
usually the least eflective. The more common abrasives
are nut shells, sized about
inch or rice. The abrasive
must have sufficient mass to achieve the momentum re-
quired to dislodge the dirt. However, high mass particles
do not follow the gas stream. Also, they are hit by the
ieading edge of the moving wheels and blades. Conse-
quently, the traiting edges are not abraded. The closer
the dirt is to the point of injection, the less significant
the asymmetrical distribution.
The abrasive must also be sufficiently tough to resist
breakage on impact. This is a problem r'vith rice, since
it shatters readily. Again, the closer the injection to the
deposit, the less significant the toughness.
Another problem with abrasives is what happens to
them after they have done the cleaning. In a simple
cycle gas turbine, they will probably be burnt. However,
on a regenerative unit they can deposit in the regener-
ator. Some regenerator burnouts have been attributed to
this. In steam svstem, they would probablv plug up traps
throughout the system.
Duling discussions about abrasive cleaning, the pos-
sibilit;r of causing labyrinth damage is alrva;-s laised. fn
fa-ct, these apprehensions have proven groundless. We
do not i,now why this is so. It couid be that the parti-
clcs are too big to enter the clearance space. On a cen-
trii,-lgai compressor, a iypical radial clearance on the
inl,erstage shaft iabyrinth is 0.008 inches, as compareci
a particle size oi 0.060 inches. The e1'e labvrinth
iLas a ;ruch iarger clearance, but a particle rvould have
to rnake an unguided 1B0o turn to reach it' it is unlikely
tilac it would do so.
ilor''r are the abrasives rntrociuced into ihe machine?
iTith air compressors, the abrasive can be thrown into
rhe open suction. If the suction or
point of injection is
pressurizcd, the abrasives can be introduceci using a blow
pot. An eductor should be used to put the abrasive
ieaving the blor'v pot into a fluidized state before intro-
clucing it to the main sas stream. A good starting point
for ilre injection rate is 0. 1 weight
cent of
Sclvellt cieaning is a :nuch :nore delicaie lecitnique
tha:r ihe blgte force of abrasic:e" Iit :'e;:iiti,-, there ,vill
:-iinost z;i'.';ays be solae ahrasive a":ti':.: i:rvc1'zed. Tle
idea is ic rjLissoLve ;.he dep*sit
r- s*iteai. The soirrticl
fi EFH ! liT ES FR0fl"4 ll';'il RCIeA REOI{ Pi10C h$*q; i',1 G
must then be removed from the system before the solute
is re-deposited. Each solvent cleaning application pre-
sents different problems.
Wqfer wqshing steqm furbines. The cleaning of rvater-
soluble deposits from steam turbines is a classic example
of solvent cleaning and the complex auxiliary considera-
tions. Since the solvent is water, the intent must be to
bring water into contact with the deposits. In this case,
the working fluid is steam and water will only remain
in equilibrium rvith it if the temperature is at saturation
for the pressure at that point. Water must be added not
only to provide the solvent, but to cool the fluid to sat-
uration. We would assume sufficient water must be added
to make the inlet steam l/o wet. Therefore, the rvater
injection rate could be as high as 25/o of the initial
steam mass flow. It is the injection of such larse quan-
tities of liquid that creates the potential problems.
The potential problems of water washing steam tur-
bines are:
Misalignment due to piping stress as the tenpera-
ture is reduced.
Water slugging.
Loss of clearance due to diflerential contrzrction be-
tween rotor and stator.
Vibration due to non-uniform deposit rernoval.
Thrust failure due to almost conplete plugging of
one stage.
Damage to blading if it hits water retained in the
exhaust casing.
On most machines, the misalisnment due to pipe stress
will not be significant. After all, we are only disposing
of the superheat, whereas during run up) the machine
is exposed to a temperature change at least tu-o times
as Iarge. However, if piping strains are a problem on
startup, one must make sure all sliding supports are free
before atiempting to wash. We have had no problems
other than an increase (doubling) of axial vibration due
to misaiisnrnent.
trVe trlr to avoid rvater slugging by tr,vo rleasures. "rVe
alwavs use a venturi nozzie lor desuperheating. -\lso.
insist that the piping fall continuouslv betr'r,eeu ri're de-
superheater and Lhe rnachine ialet. Er.en if the rvater is
:iot broken up into cirooiets in the ciesuperheater, it rviil
pass inoccuously through the turbine as a constarlt stream.
To prevent loss of clearance.
alu,ays limi'c the rate
of temperature chanse to 180oF per hour. The
':tazard wouid be failure of the injection pumps rvhen at
maximum injection rate. Such a iailure
;!. very high rate of change of temperature ald vrould
raost likely resuit ur an axiai rub. To guard :his. rue iiry
tc use boiler feeciwater, since these p'"rmns are the ncsi
retiable in the plant.
tr'Ye a-ttempt te reiuce rhe cha-nces of r,li'l-lr:lfcrrn
di:posit reinoval i;i, ha'.tirg riie ilcrease oI i'rjet:iicl rate
u'h.e;r:.i.:i depcsi: is actr-rel1y bsilg r-::l:.o.,rec. T':ris cot-
dition is detected by neasuring the conductivity of the
exhaust condellsate.
It is alleged that thrust bearing failures have occurred
bccause oI stage pluggin-q n'hen an upstream wheel has
sluffed its deposit before a dorynstream one. If this is
so. the halting of injection increases tvhen material is remor,'ed rvould prevent it.
One of the criterion \re use to check a turbine design
before purchase is "Can all the condensate be removed
from tlre exhaust?" Some turbine designs are such that
the l:laclir.rg is 'nvithin abour 1 inch of the bottom of the
casir,g. Others don't have a casing drain at the lor,r.est
point. Others have a
inch or
inch casing drain.
All thcse designs are suspect unless the condensate can
drain frcely out of the exhaust.
\\Ie consider that a stage is washed adequately when
the conclensate falls to half its peak level.
\\'lien this point is reached, the rvater injection rate is
agair.i increased. The rvash is considered completed when
the inlet stearn is saturated and the exhaust conductivity
is dou,n to 200 micromhos.
After the rvash is completed, the inlet temperature is
raised to norrnal at a maxinum rate of change of 180"
per hour.
Initiation cf the normal steam flow path, bypassing the
desuperheater'. is the last hazard. We have found that
rrater builds up in the line upstream of the valve, even
rvhen the br'pass is left open. We norv al'"vays leave the
main r.alve cracked open to pre\/ent the water buildup.
Over tlie past 13 years, \\'e have successfully com-
pleted about 30 turbine \\'ashings. These involved six
diflerent machines, Iocated irr four plants. Based on this,
u,e concludc- that on-load \\,ashing is safe provided rea-
sonable cale is exerciscd. \\'e have, howe\/er, observed
that deposit solubilities vary considerablv between sub-
scqLrent washes on the sanre machine. This same vari-
abiliry has been observed on tr,vo machines supplied by
steanr fronr the same source for the same time period.
Some machines can be successfully cleaned rvithout mak-
ine the inlet saturated but most have required a r,vet inlet.
Durins our earliest washes, rve believed that condensate
n,as essential as the superheating medium. We reasoned
that any other water u'ould leave salts behind during
the remperature increasing phase. Five of the machines
have now been washed using boiler leed water,
ani, sb...r^tle problems or deterioration of the cleaning.
AIso during our earliest u,ashcs we noted an apparent
accelerated louling rate durins tlie first fe,,v dz.ys after
a wash. We ha,,-e no reasonable explanation of this phe_
nomenon. It levels out quicklv and does not appear to
affect eithel the maximum mass flow or the efficiencr,.
Currently, r,r.e merely rvarn the operators to disregard it
if it is obserr.ed.
In spite oI the abor.e hazards, steam turbine rvashinq
is an idcal application of solvent cleanins. If the inlei
steam is unclei 1400 psig, is initially superh"eated, and the
turbine is under load, the exhaust n'ill al,,r,ays be r,r,etter
than the inlet. Therefore, the water is injected.
the rvashing uill occur first on the dou'nstream blades.
Further, as the u-ashing proceeds, the solution passing
through parts of the turbine alreadl' cleaned is alr,var.s
more dilute than r.vhere the deposit is going into solutior-r.
There is no chence rhat dissolred nrrierial uill be redc-
posited. The water soluble deposits in these machir.res
are usually highly soluble and rvashing is rapid. Al1 the
deposit removed remains in solution in the condensate
and can be removed from the discharge end by dr-ainir-rg.
Other solvent cleoning opplicotions. Other nrachines
are not as ideally suited to solvent rvashinq as turbines.
fn compressors, for example, the tendency is for the so1-
vent to e\/aporate as it passes through tlre machine. In
addition, the solvency of n-rost solvents is much belo,,r
that of rvater for sodium chloride. The problem is to
find a solvent that rvill attack the deposit, and
port it out of the machine. In addition, the solution must
either be r.vithdrau,able from the discharge s)/stem or must
be acceptable in the product. The solvent ma,v be a t\\.o-
component svstem if the deposits are at tlie cold end-a
light one to clo the actual dissolving and a heavy one to
pre\-ent dcposition as tlie forrner er.aporates. \\/hen the
deposits are at the dischara.e, a light sol'u'ent can be used.
prorided it is inj.cterl rs close as possible to the doposits.
in large droplets, and a high rate. These three criteria
help to prevent premature solvent evaporation.
\A'c hacl an example of premature evaporation recentlv.
Dilute caustic soda \\,as inadvertently admitted into a
centrifugal cornpressor. Thc rvater evaporated leaving a
solid deposit betrveen the sccond and third wheels. \\'e
did not \vant to use water as the soivent for tu'o reasons.
The machine rvas equipped rvitli alurninum labyrinths
and rre u'ere afraid ihe wet caustic rvould dcstroy this
metal. Further, since the discharge temperature u,as 300o.
it seenred unlikely that enough water could be added
to carn' the solution out of the discharge. We found
that a liquid hydrocarbon with a final boiling point of
about 4500 would move the deposit. Using this as a
rvash, the deposit was moved to betrveen the fifth and
sixth rvheels. Hou,ever, even r,vith injection into the stage
drain before the fifth wheel, it r,vas not possible to move
it out of the clischarge. Finall,v, \ve gave in,,
and rvashed "off line" w.ith mildly acidic lvater.
As rvith turbine washing, there must be a u,ay of re-
mo\-ing the solution from the discharge casing. If the
discharge flange points down, this will not present a
problem. Ilowever, if it points up, and espccially if the
discharge velocitv is lou,, an adequate sized casing drain
is essential.
Originally presented at the 2nd Anoual Texas A&M Turbomachinerv
Symposium, Oct. 23-25, 1973, College Station, Tex.
Aboul the
BnrnN TTnNr,r" is an engineering asso-
ci,ate in Imperial Oi,l'a Dttginaering
Diaision at Sarruia, OnL Hie
include prepat ation of maclvinerg speci-
nd consu
ce and,
has been in the Engineet ing Diui,sion
76 util:ities, pro-
cess en hinety. Before
ttso,e emmloued,
i,n- a rat"i,ety of
ou)et, rL_
gland.. He usas e
lectri n_
prof e er
in Adwrio.
A closer look at
turbomachinery alignment
Successful shatt alignment
requires consideration of all
factors that influence movement
of the machinery. Here's what to
look tor before and during the
actual alignment
Jock N. Essinger, Shell Chemical Co., Houston
Gooo runsoMACHrNERy ALTcNMENT consists of three
interrelated parts. Each part is an equally important step
in realizing the goal of an equipment train capable of
operating safely for long periods without the many prob-
lems resulting from severe misalignment.
l. The sysfem survey. Many steps vital to good tur-
bomachinery alignment are taken well ahead of the actual
"cold alignment." IJnless the person responsible for align-
ment has first-hand knowledge that pre-alignment prepa-
rations have been properly carried out and completed,
verification is essential. Some of the items which warrant
specific attention are:
1. Piping. A visual piping inspection by those respon-
sible for alignment is vital. This inspection will assure
that the piping is installed in apparent agreement with
design criteria, is complete and in its functional state.
Obvious items to look for are proper placement and ad-
justment of guides, anchors and supports; proper adjust-
ment of tie-bolts on expansion joints; correct positioning
of spring hangers; complete make-up of flanges with gas-
kets in place and bolts tightened; absence of slip-blinds
which may have been installed for hydrotesting of pipe
lines, and proper orientation of check valves. In short, it
must be verified that the system is in order so post-align-
ment piping modifications will not nullify the alignment
2. Grouting should be checked to be sure it is com-
plete and apparently well done.
3. Foundation bolts should be checked for tightness.
4. Check all shimpacks. Shims are the vital link be-
tween the machine and the foundation and are essential
to maintaining alignment over long periods. It is my ex-
perience that the only reasonable assurance of proper
shimpacks is to physically remove and inspect the shims
at every machine support just prior to final alignment.
Obvious problems with shims include rust, improperly
cut shims, folds and wrinkles, burrs, hammer marks and
dirt. I once found a turbine installed on shimpacks
wrapped in masking tape. It is good practice to use as few
shims as possible and replace many thin shims with ferver
shims of greater thickness. Stainless steel shims will pay for
themselves many times over by minimizing alignment
problems associated with shim deterioration. It is equally
important to see that surfaces of equipment supports and
soleplate/baseplate are clean and in good condition.
5. Check for misalignment of machine supports relative
to the soleplate. A relatively simple test for problems in
this area can be made when shimpacks are checked.
Mount a dial indicator on the machine supPort with the
indicator stem resting on the soleplate. Watch the indi-
cator as the hold-down bolts are loosened' If movement
of the indicator is more than 0.001 to 0.002 inches, it is
an indication of a problem that must be defined and
eliminated. Remove the shimpack and check with feeler
gages to be certain the machine suPPort is parallel with
the soleplate. If not, re-grout, re-machine the support or
prepare tapered shims.
Be suspicious if one support moves less than others. If
three of the supports move 0.001 to 0.002 inches, and
the fourth shows no movementJ it may indicate that the
support is carrying more than its share of the load.
6. Check the casing for distortion. Despite the massive
appearance of turbomachinery casings, they are flexible
and easily distorted. This relatively simple test for gross
distortion can be made when shimpacks are checked:
a. With three supports tightened d6rt'n, remove the
shimpack from the fourth support.
b. Determine the total thickness of the shimpack and
record the dimension.
c. Let this corner of the machine down (the machine
is now supported only at three points) . Using feeler gages,
determine the distance from the soleplate to the machine
support. Record the dimension.
d. Subtract the feeler gage dimension from the shim-
pack thickness. This is the total deflection of the machine
casing with no support at the corner being checked.
e. Repeat the procedure at each of the four suPPorts
and compare the deflection of each. Gross differences in
deflections at arly of the four supports is an indication
of probable casing distortion.
Fag. l-Typical arrangement for "Face-OD', readings to de-
termine cold alignment of shafts.
7. Check for piping strain. Piping strain is seldom de-
tectable by visual observation. Ilowever, gross problems
can be detected by a rather simple test. Following the
check for casing distortion, place dial indicators on the
machine to monitor both vertical and horizontal move-
ment of the casing or shaft. Now loosen all the hold-
down bolts. If the machine moves more than the average
observed when checking individual supports, it is obviously
the result of an external force-probably the piping.'
B. See that bearings are properly installed in the ma-
chines, that they are lubricated and that the bearing
covers are properly tightened. These are rudimentary pre-
cautions made on the basis of expelience through at-
tempted alignment of machines r+'ith the bearings removed.
9. For turbomachinery trains with high-speed gearing,
extraordinary precautions should be taken to set and align
the gear itself. High-speed gearing is normally the most
precisely made equipment in the train; it is also the most
sensitive and most vulnerable to catastrophic failure as
a result of poor installation. IJnfortunately, installation
instructions for rnost gears are grossly inadequate. One
fundamental precaution is offered: if a gear is involved,
be very, very careful.
Planning. Since turbomachinery alignment normally
involves the participation of many people in various
functions, it is imperative that an over-all plan be de-
veloped to assure that the goals of the program are
known and understood and that the method of attack
is consistent with the desired results. Moreover, such
planning is valuable from the standpoint of minimizing
the time involved (hence, the cost) to achieve satisfactory
alignment. Plans should include:
1. Determine the desired placement of the shafts (cold
settings) considering anticipated thermal growth of the
various components. Define movement of shafts within
bearing clearances (rising-pinion gears, for example),
hydraulic loading and any other factors expected to pro-
duce relative movement of shaft centerlines when the
machines are operated.
2. The sequence of alignment must be determined for
multi-unit trains. For two-component trains, determine
which of the two machines is to be moved.
3. Select a specific method for determining relative
shaft positions.
4. Decide on tolerances for theoretical cold alignment
settings. Allowable deviations which are unrealistically
strict cost time and money, and may not be attainable.
Tolerances which are too loose are an invitation to
5. Make a list of tools and instruments required for
the alignment and make certain they are available and
in good working condition. These include such items as
alignment brackets, dial indicators, tools required for
hot alignment checks, pre-cut shims of various thicknesses,
tools for lifting the machines during shim changes,
wrenches of the proper size for hold-down bolts, coupling
tools and a device for rotating the shafts.
6. Provisions should be made for permanent recording
of alignment data. Many good sets of alignment readings,
written upon the side of a compressor, have been cov-
ered by the painter's gun.
2. Cold olignment. The term
"cold alignment" refers
to the position of a turbomachine's shaft centerline rela-
tive to the shaft centerline of a connected machine, with
both machines in a non-operating or "cold" condition.
Offset and angularity are both implied by the term. Cold
alignment is important because it is normally the only
check made to directly determine the relative position of
the two shafts. Results of the check form the basis for
determining shaft alignment during operation.
The only cold alignment techniques discussed here are
those using dial indicators. Procedures ale old and well
established, but do have problems and pitfalls.
Fig. 1 illustrates the most widely used of the traditional
alignment methods, commonly referred to as the
OD" method. A bracket is attached to one shaft and ex-
tends near the coupling hub on the adjacent shaft. Dial
indicators are attached to the.bracket as shown, with the
stem of one indicator resting on the face of the coupling
hub. The stem of the other indicator is resting on ttri
OD of the same hub. Oflset of the shafts is determined
OD readings. Angularity is determined by
The method has and continues to serve well in the vast
majority of industrial alignment problems. Indeed, it is
the specific method outlined in virtually every mainte-
nance manual for industrial rotating equipment.
For high-speed, high-hp turbomachinery, where re-
quirements for precise
alignment are more stringent,
shortcomings of the method become increasin$ly impor-
tant. Consider the following points:
The indicator readings from this method reflect not
only misalignment of the shafts of the two machines, but
inaccuracies in geometry of the coupling hub upon which
readings are made. Typicalty, the measurements should
have an accuracy approaching 0.001 inches. It is not
uncommon to find runout of the face or the OD of a
coupling hub in excess of this amount, even on god
quality, high-speed couplings. This error can be eiimi-
When making face measurements it is absolutely
necessary that the axial floats in both shafts be accounted
for. This is of little concern with small equipment where
shafts are fixed axially by ball bearings. It becomes a
problem, however, on machines equipped with hydrody-
namic thrust bearings (or no thrust bearings at all) . The
normal procedure is to separate the shafts axially each
time a reading is taken. The axial position of the shafts
as they come to rest depends on how hard the craftsman
pushes. Thus, consistent readings are difficult to obtain.
The practice also leads to the use of makeshift tools to
pry the shafts apart, and a coupling seldom comes through
the alignment procedure unscathed.
The face diameter upon which readings are taken
to determine angularity of the two machines is relatively
small, especially in high-speed equipment-sometimes no
larger than 3 inches. For example, a reading of 0.002
inches (normally termed
"acceptable") on the f.ace of. a
4-inch diameter coupling hub represents an angularity of
0.006 inches per foot or an error of 0.030 inches at the
opposite end of a machine 5 feet long. Add to this a 0.001
inch error in the face of the coupling and the outboard
end of the machine is 0.045 inches from the desired po.
sition. In most instances such an error is not catastrophic,
but it is an error that should not exist.
The tool normally used to hold the dial indicators
is either a "universal" bracket, or makeshift device con-
trived on the spur of the moment. Brackets specifically
designed for the machines being aligned are not normally
considered necessary and are, therefore, not used. The as-
sumption is usually made that the bracket being used is
stiff; i.e., it does not deflect under its own weight. Nor-
mally there is no attempt to verify the assumption, which
is a very poor one, especially for couplings with relatively
long spacers. I have found many brackets which deflect
0.005 to 0.010 inches over a l2-inch span. In an extreme
case (a machine in trouble)
it was found that the indi-
cator bracket which had been used deflected about 0.040
The method requires that the coupling spacer be re-
moved to make readings. For routine alignment checks
this means added ef{ort and the ever-present risk of coup-
ling damage during disassembly and re-assembly. For new
installations, it normally means that the coupling is left
open until after cold alignment is complete. The interval
is usually several weeks from the time equipment is first
set upon the foundation until the alignment is completed,
during which time the coupling is subjected to corrosion,
dirt and physical abuse. Some couplings do not survive
this initial treatment.
Reverse indicqtor method. fn contrast 1o d1s "p2ss-
OD" method, indicator readings can be taken by the
reverse indicator method on the OD of the coupling hubs
(or the shaft) only. Two brackets are used simultaneously,
which is normally preferred.
Equally acceptable but less convenient is a method
that uses a single bracket switched back and fourth for
each set of readings.t Analysis of information gathered
by this method yields suflicient data for determination of
relative shaft positions. Advantages of this method over
1. By proper design of the alignment brackets, the re-
quired removal of the coupling spacer is eliminated in
Alignment hints
o Assign a responsible, technically astute individual to
follow details of the alignment. It is too important to
receive second-class attention.
o Spend adequate time in preparation. It is money well
o Make sure the craftsmen understand both the method
and the intent of the alignment procedures.
o Check the tools. A sticky indicator has no place
among the alignment tools.
o Assure yourself that the indicator brackets are sturdy.
The best test is a healthy shake; if tJre indicator read-
ings do not return to the original position, the bracket
is inadequate. The use of magnetic indicator brackets
is never permitted for shaft alignment.
o Find a good way to turn the shafts. If you are lucky,
the person who specified the machine had the manufac-
turer put wrench flats on the shaft extension at the out-
board end of the machine. If not, find another means,
such as a strap wrench or a clamp-on fixture made
specifically for the purpose.
o When placing the indicator, be sure it starts out at
mid-range, and be sure it rests squarely on the shaft. An
indicator which sets askew gives bad readings. Likewise,
be sure that the range of the indicator is not exceeded
when readings are taken.
o Before making any moves (and before final accep-
tance), get at least two sets of identical readings. If the
readings cannot be repeated, they are not acceptable.
o Check the data before making a shim change or
horizontal move of the machine. It takes very Iittle time
to check the data-it takes a long time to move a large
o \4ake one move at a time, then check the work. Don't
worry about horizontal movements until the elevation is
o Be alert to peculiarities of the specific machines being
aligned. Tilting-pad bearings, for example, sometimes
give trouble by permitting the shaft to rock back and
forth on the lower pad, making repeatable measure-
ments difficult. It is sometimes necessary to put tem-
porary shims in such bearings to assu,re that the shaft
is stabilized for alignment purposes.
o Stop the shaft at precise 90-degree increments for
measurements. Turn the shafts in one direction only.
If you pass the 9O-degree mark, go around another com-
plete turn. Extremely handy for tJre purpose of de-
termining the quarter turn is a tool with four spirit
levels mounted at 9O-degree increments.
o Be wary of planned changes that do not turn out
right. Alignment changes are calculable and predictable;
if they consistently go wrong, look for something amiss
in the system.
most cases (shafts must be sufliciently in alignment to
prevent binding of the coupling, of course) . This feature
provides three distinct advantages:
a. Wear and tear on the coupling is reduced.
b. Since both shafts turn as a unit with the spacer in-
stalled, errors caused by coupling hub runout are entirely
c. By spanning the entire coupling, angular misalign-
ment is greatly magnified and more precisely diagnosed.
For example, a span of 16 inches gives angular misalign-
ment readings eight times as great as if measured on the
face of a 4-inch-diameter hub.
2. Since face readings have been eliminated, there is
no concern about axial float of shafts.
3. This method is not successful if makeshift tools are
used-special tools must be constructed. Alignment results
are simply better when proper tools are used. I wonder
why alignment tools for high-speed couplings are not
available in standard designs from the coupling manu-
facturer ?
The single problem with the
method that
is not solved by the "reverse indicator" method is deflec-
tion in the alignment bracket. This is a matter of real
concern, and one which must not be ignored if proper
cold alignment is to be achieved. The problem is readily
handled, however, by determining the deflection in the
alignment fixture and making appropriate corrections
in alignment data. Fig. 2 indicates one method for de-
termining the error. As shown, barstock of appropriate
size is chucked in a lathe, and the end is turned to the
proper diameter to accept the alignment bracket.
Without removing the barstock from the lathe, the
alignment bracket is attached. With the bracket on toP
of the barstock, the indicator is set to zero. The entire
assembly is then rotated 180 degrees, so the indicator now
reads directly on the bottom of the barstock. The indica-
tor reading thus obtained is reflective of the deflection in
the alignment bracket.
Once determined, deflection should be stamped promi-
nently upon the bracket, as should the designation of the
equipment for which the bracket was specifically made.
The alignment bracket becomes a special tool for that
machine, and should be cared for as such.
3. The hot olignmenl check. It is now generally ac-
cepted that the traditional "hot check" for turboma-
chinery alignment is'of little value. Not only is it costly
and time consuming to bring a machine up to tempera-
ture, stop it, break couplings and attempt to determine
alignment before it cools off, but results are highly ques-
tionable. In some instances a hot check is dangerous
because it creates an unwarranted sense of security.
It is not possible to make the check quickly enough to
accurately determine thermal growth of the equipment.
For machines operatine at temperatures well away from
About the
JAcK N. ESSINGER is a staff engineer
with Sh,ell Chemical Co., Houston. His
duties inclucle specifi,cation, eaaluati,on,
insta,llation, startup, troubleshooting ond
maintenance of rotating mo,clnnery in
chemical plants. Mr. Essinger holds a
B. S. degree in mecho,nical eng'ineering
Arizona State Uni,uersitg. He is a
member of the Amet"ican Societg of
Mechanical Engineers anil the American
Fig. 2-Setup for determining deflection in alignment brackets.
Fig. 3-Basic elements of mechanical device for checking hot
ambient, it is not uncommon to see a thermal change
of 0.001 inch per minute when the machine is first shut
down. At this rate, the effectiveness of the hot check is
certainly lost before the alignment data can be obtained.
Misalignment resulting from hydraulic forces and torque
reactions, which can be significant, are never revealed
by the traditional methods because the forces disappear
when the machine is stopped.
It is generally accepted today that a superior alternative
is to use the cold position of the shafts as a benchmark
and deduce the hot alignment by monitoring the move-
ment of the machine casings, or shafts, from the cold po-
sition to the hot position. Several methods have been used
for this monitoring, with a variety of techniques applied
to the actual determination of casing or shaft movements.
The most widely known methods use optical or electronic
Details of these techniques have been well publicized.2
Flowever, a purely mechanical method for measuring hot
alignment is also being used successfully. This method uses
permanently mounted tooling balls for measurement lo-
cation references" A spring-loaded mechanical gaging de-
vice measures relative movement between the reference
tooling balls in increments of 0.001 inches. Fig. 3 shows
the basic elements of this method as applied to a typical
turbomachine. Again, the cold position is used as a bench-
mark for alignment corrections.
Abstracted from the paper
of Turbomachinery," oiginally pre
sented to the Scond Simposium on Compressor Train Reliability, Manu-
facturing Chemists Association, April 4, 1972.
C.. "Hw to align barrel-t?e centrifugal compresom," .flydro-
iarbon Proeessins, 50 No. 9 p. 189 September 1971.
Jacksoo, C., ':Succesful
ihatt hot aligment," Hydrocarbot Processitg.
48 No. t p. 100 January
1969. t
Hot alignment too complicated?
Existing measurement technigues tor hot
alignment of rotating machinery are
complex and dilticult to use, Here is a
simplilied way to obtain the necessary
Jock N. Essinger, Shell Chen-rica1 Co., Ilouston
Ixroart.lrroN FoR TrrE hot alignment check is nolmallr'
obtained b1' measuring vertical arrd horizontal movements
of shafts or bearing ho-.rsings of turbomachinerl compo-
nents relative to fixed references. When optical alignment
techniques are emp1o1'ed, the references are vertical and
horizortal planes of sight established by the optical instru-
ment,i. Por electronic or mechanical techniques, the fixed
reference is norn-rally the machine foundation. For the
typical installation, horvever, the foundation is not in the
proxir.nitv of the bearing housing, and it is necessarl' to
install pedestals extendins lrom the foundation to a point
near the bcarins housing, these pedestals bccomins the
refet'euce fronr uhich measurelrents are taken. To asstrre
the ecculacv of tl're alignnrent data, the pedestals must be
substantial, and thev mr-rst be maintaincd at constant terrl-
per:rtlu'e to minimize ineasurement inaccuracies resulting
fron thelnral grorrth of the pedestals themseh'es. The
perlestals arc nornally rvater cooled.
The alternate method presented here also uses the nra-
chine forrndation as a reference, but climinates the require-
ment ior alignment pedestals b,v using permanent refer-
cncc points rvhich are affixecl directly to the foundation
ancl to the bearing housings (or the machine case)
shoun in Fig. 1. All four refercnce points lie in a planc
to the ccnterline of the rnachinc shaft.
Similar relcrence points are mor-rnted at each bear:ing
hoLrsins in the tr:rin.
Follorving cold alignment of the compressor train, ref-
efence dimensions A and B, and angles O and
are deter-
n'rinecl et cach bealing housing, and are recorcled. \\'hen
the rllrchine is brotrght on line, din-rensions A'and B'are
measulecl at each position in the hot runnins condition.
Thc clata thus obtained is adequate for determining the
vertjcal and horizontal movement of each bearing housing
in the nrachine train relative to the foundation. See Fig. 2.
Using c.ommon grid paper (4 x 4 grid is usuallv a con-
venient size), lay out reference vectors A and B at angles
O ancl
having these vectors closs at one of thc grid
intersections. The intersection of these vectors represents
the centerline of the machine shaft in the cold position.
Nolr- refer to the cold ancl l'rot measln'ements previousll'
n'rade (A, A', R, ancl B'), arrd dctcrrrrine the rrrovernent
of the bearing housing along vec.tors A ancl 13 by taking
the differences between c:old ancl hot rneasulemerrts (1A
and -tB) for each location. Lrry out these nrovemcnts
along vectors A and B using any convenient scale. sa1'
inch equals 0.001 inch, to cstablish points n and b. Nou'
Fig. l-Typical placement
bearing housing.
of benchmarks on foundation
Fig. 2-Graphical determination of shaft in hot position ielative
to cold position.
FiE.3-Stainless steel
wirh protective cover.
Fig.4-Precision tool set for hot alignment checks.
drarr ljrrcs Llrrorrgh a ancl b
to \-ectors
A arrrl l:i.
lines rcprcsent arcs of racliuscs A'and B'
dr:rwrr ironr tlrc foundal.ior-r Jrr:ncJrmalks. The intersectiort
oI tlrr:sr: linc-q de[incs t]re
oi tire machine shaft
t:orto'lirt' irL thr' hot position relativc to the cold positiorr.
f'o dr:lr:rrrrinc tlrr: nrovurnent in vertical and holizontal
dirc'ctiors, it is necr:ssary c-,nli, to scalc off tile dirur:ttsious
reicllecl to as AII:Lnd AV, using the same scaie zrs used in
plotlirru A.A. and Ati. A sirnilar piot for the ciata secured at
crr:h bcarirr.g ]roLrsing afforcls srrfficiert infonnation for
plottin.g tlri' hot a'lisnmcnt of the entire trrrbomachinerv
tr:rir r.
Tlrc Lcrrchnrarks for this techniquc are
inch diam-
c[cr' prccision ba]ls. llecr.rse t]resc benchmarks becon-re an
ol the installation, arrd because the accuracv
oI alignnrent rci:ords, over thc long tern, are depcndent
uyrol thesc Lel{:renr;cs, it is reconrrrrended t}rat they be
rrraclc of stainlcss sti:cl to prc\cr)t corrosion end that thel'
br: nrounted strbstantially to avoicl inadvertant movement.
Fig. 3 shorvs a bcnchmark and protcctive cover specificall-r'
clesigncd for ttrrbomach ineri. rvork.
shows :r set of precision tools specifically n-rade
for the
of turbornachincrl' alignment by this tech-
alisnmcnt gage cmploys a long range dial
indicatol set into :r si;ring-loaded telescoping column u'ith
inch dianreter spherical seats at either end. B), t sing
six extensions, thc tooi covers a measurement range of
25-60 inches, in incrcments of 0.001 inch.
range is
adequate for nearly all turbomachinery in use today. The
star.rdard pror,idcd for calihration of tlie basic tool is con-
structed of Invar', as are all of the extensions, thereby
Fig.5-Calibration of the alignment gage with the lnvar
stand ard.
concern Iol tirermal srorvth oi the align-
ment gaLlge. Thc inclinornetel is used lor cietermining
angles at rrhich the measLrrLrilents arc made.
.\ steprrise procedut'e for alisntuent bv this :letl-rod is
as loliorr s:
1. Aflir bcnchmarks to thc rnachinc anC lo ihe [ounda-
2. Aiign the tur:bomachinelv train in thc nolrnal mAnnef-
lecording coupling alignment dat:r.r,r
3. Using thc standard, calil:rate thr alisnrrent
gauge 3s
rorr'n in Fig. 5.
4" Usine the appropriate gauqe extension, clcttt'tnine ani
record colcl measlrremcnts at cach locatior-r. a: shorrn in
Fig. 6.
5. Deternrine the anq-les at rvhich measLrrcmelrr; arc being
made bt' usine the inclinorneter as shou't'L in Fig. 7.
Recorci the data.
6. \\Ihen the train is brought on line. obtain end record
hot aiignment data by repeating Steps 3 and I at cach
reference station.
7. Plot the data as indicated in Fig. 2 to detenline ver-
tical and horizontal movement of each bearing housing.
B. Using combined data from the individua[ locations-
plot the hot alignment of the entire equiprnellt train.
9. Make appropriate alignment changes as dictated by
the above data.
Fig. 7-lnclinometer for determining angle at which rneasure-
ments are taken.
Fig. 6-Reference measurement between bearing housing and
machine foundation.
Fig. B shous t1.pical placcment of ber.ichmarks at the
various bearing housing locations in a turbine-contplcssor
This alignment technique has been used successfi-rlly on
severai turbomachinery trains over the
three vears.
Some of the apparent advantages of the methoci are:
Thc tcchnique is simple, and no speciiic technical skills
are required.
\rcrtical and horizontal movernents of the rnachinc are
obtaincd r,vith equal ease.
Once the bencl'rmarks arc established, setup tirnc for hot
checks, or le-checks. is minirnal.
The benchmarks are permancnt, and are readiiv main-
tained for long-term monitoring of aiignment.
The method is highly resistant to accidental loss of
reference; i.e., to being "bun.ipcd" during use.
It is readilv adaptable to crarnDed quarters.
About lhe
JAcK EssINCER is a stalj engineer taith
Shell Clrcmical Co., Houston. His duties
irtclude speci,fi,cct tion, euoluation, installa-
tion, startuTt, troubieshooting and main-
tenance of rotating m,achiner1l in clrcmi-
cal plants. Mr. Essinger holds a B.S.
degree in mechctnicctl engineet"ing
Aq'izona Skttc Utzirersity. He is a mem-
ber of the American Society of Mechani-
cal Engineers and the American Society
or Metals.
Fig. 8-Typical benchmark placement on a turbocompressor
Datataking is a one-n'ran operation.
Initial investment in harclr,r,arc is modest. Rc-use is high.
The equipment is highll'portable.
The hardrr.are is simple, reliable, and rerquires rnirimal
C..."llo_w r-o- ali^en barrel-t1pp cenrrifuqal ,unrlircs\r,,\. ftr tlrurrr-
Lon Procc.sinS. 50. No. a. p. lBr), Septcmber Ir;1.
Ioo-k ar turhnmaclrinery rlignment." H,drmrrbnn
Procc..irrq. 5:, N,'. 9, p. 1BJ, Seprcmber 1973.'
using too
much steam?
Steam tlow in single-stage turbines
is relatively easy to calculate.
Multistage turbines
calculation uncertainties to which
must be added losses in couPlings
and compressor llow uncertainties
H. Steen-Johnsen, Consultant, East Hampton, N.Y.
A nBrrNBn was concerned that therr single stage tur-
bines were using too much steam. And so they were. Some
of their turbines were rated at 100 hp. The manufacturer
wanted to make sure the
was there, so he nozzled
control valve. So, of course, they used too much steam at
100 hp.
Couta the operators have determined how safe they
turbine is still on governor control. When the ring pressure
is 75 percent of throttle pressure, close the hand valve.
Nozzle flow meter. The first-stage nozzle is an excellent
flow meter.. For a single valve turbine with critical drop
over the nozzles, calcuiate the steanr florv rvitlr the fornrrrla
lb.lhr.lin.2, where p is the ring
in psia
and K is from Fig. 1.
Example. Assume a 600 psia steam throttle pressure at
7500 F. The ring pressure is 550 psia. The enthalpy at the
throttle is 1381.-From Fig. l, K
43.4 and the flow is:
SSO (43.1)
23,900 lb.lhr.linj' Then, if the tur-
flow per sq. in. of nozzle area for steam' (Flow
Abs. press, x Factor K, lb.,/hr. in.')
bine has a nozzle area of 0.23 inJz, the turbine is passing,
23,900 (0.23)
5,500Ib'/hr. which is within 2 or 3
percent of being corect.
When the first-stage pressure is greater than critical, as
it frequently is for a multistage turbine, the critical flow
must be corrected, using a correction curve obtained from
the manufacturer.
Multistage turbines. For larger multistage turbines with
multiple inlet valves, the problem is more complicated.
Total flow is the sum of the flow through the several
valves that are open, and requires a multiplicity of gages.
A turbine of this type should be equipped with a flow
nozzle in the throttle line to obtain a reasonable measure-
ment. And here we turn to the PTC 6 Report,l Fig. 4.2'
pages 16-18.
first item noticed is that "liquid" measurement is
rfr rr00
Fig. 2-Location of the metering nozzle
here produces a flow measurement un-
certainty of
18 percent.
Fig. 3-Location of the metering nozzle here produces a flow
measurement uncertainty of
+ 1.75 percent.
Fig. 4-y;n;rum tlow nozzle arrangement
for compressors.
rated better than superheated steam measurement. This
"liquid" procedure, of course, requires a unit boiler-
turbine-condenser system, which is rare in a process plant.
The liquid measurement allows accurate determination of
the condensate and the feedwater.
The next item noted is the requirement for inspection.
A flow nozzle in a steam line is subject to considerable
abuse when a plant is started. Welding beads, welding
rod and other debris go through on initial blowdown. The
resultant turbulence increases A,p, and a steam nozzle
reading 5 percent higher than the feedwater is not un-
Location of the metering nozzle is important, as noted
by the diflerence between Figs. 2 and 3.
Flow measurement uncertainty. If the arrangement in
Fig. 2 is used without inspection before perrnanent in-
stallation, the PTC 6 Report rates an uncertainty as t 18
percent. If the arrangement in Fig. 3 is used with satis-
factory results of an inspection immediately before test,
the report rates the uncertainty as
For the primary element, PTC 6 recommends tfre use
of the throat tap nozzle with a coeflicient of discharge, C,
given in Fig. 4.4, page 23 of PTC 6.1
As background for comparing uncertainty of turbine
and compressor flow measurement, the expression for tur-
bine flow as recorlmended in PTC 6, page 21, is from
PTC 19.5;4, page 57 Equation 6:
zo is rate of flow, lb./hr.
C is coefficient of discharge
F is the velocity of approach factor
IVF-- P'
When the flow section is calibrated, the product of. C F
is established most accurately.
d is the nozzle throat diameter.
Fo is the thermal expansion factor for ttre nozzle.
Yo is the adiabatic expansion factor to allow for the
change in ur. It is a function of the pressure ratio over the
nozzle prf pr, which
600-pound steam pressure might
be (607)
l$l+) =
0.987. The curve Fig.43 A, gives Fo
as 0.984. Therefore, an error here is a minor uncertainty,
and, it k: 1.3 or 1.4 it makes little diflerence.
lr. is the differential pressure, inches water at 680 F
(not Hg).
ur is the specific volume, cu. ft./lb., at the inlet of the
primary element.
The diflerential pressure, /r-, should be read on a man-
ometer for best results. The use of a recorder for direct
flow reading results in an uncertainty. The flow formula
is programed in the recorder, and the linkage must be
adjusted accordingly. The proper installation of a man-
ometer is discussed in the code.1
Power to coupling. With the flow established with an
acceptable degree of uncertainty, the next question is how
much power is delivered at the coupling. In other words,
how efficiently is that flow used?
In a back-pressure turbine with more than 25o F super-
heat in the exhaust, the heat drop in the turbine can be
established, and thus the internal power may be com-
puted. Using
reasonable allowance for mechanical
losses, the power at the coupling may then be calculated.
For a condensing turbine, the expansion line may be
available, and there may be some interstage taps. Pressure
readings at these points are of value only to establish
gradual fouling.
The interstage readings should only be used as a source
for internal efficiency determination, when steam is ex-
tracted for feed-water heating, and reliable readings with-
out turbulence can be obtained from the piping to the
heater. And look out for gland leakoff returns that may
upset the reading.
After the flow is established, there are two means for
determining the coupling power. The manufacturer's effi-
ciency data in conjunction with first stage pressure, or a
dynamometer with reasonable uncertainty. For a mechani-
cal drive, the dynamometer may be either an absorption
type or a transmission type. For a field test with the driven
machine in place, the coupling power is best determined
by a transmission dynamometer. This device measures the
torsional twist in an extended type coupling, and there are
several good ones on the market. Properly installed and
calibrated, it will quickly resolve whether or not the prob-
lem is in the turbine or the driven comPressor or
gos composiiion. For those situa-
tions where the turbine is 98 percent efficient, and the
compressor 60 percent, the instrumentation should be
checked first, and of equal importance, the gas composi-
tion must be carefully evaluated and matched to the com-
pressor design.
Gompressor flow meqsurement. While the turbine
flow is being established by a Fig. 3 flow measuring sec-
tion, take a look at the means for compressor flow mea-
surement. A flow nozzle in the suction to a process com-
pressor should, as a minimum, conform to Fig' 4, which
corresponds to PTC 10 requirements.l
No specific guidance is available from PTC l0 for the
uncertainty asiciated with this auangement or deviations
from it. This meas.rring section is shorter than Fig' 3,
if the PTC 6 Report for uncertainty is applied to this
installation, we arrive at-+4-?5
and without flow
+8 percent (disregarding deviations in gas
For a further look at the uncertainties of compressor
testing, take a look at the expressions
fo-r flow and shaft
ho.r"il*", in PTC 10. The expression for flow is:- W
31.501 Kd' Yo
p,) lb./min.
This is the same as the lb./hr. expression for turbines,
as can be seen by substituting
- P) =h*127'7^
multiplying by 60, where 27'T inches of water at 680 F
I psi.
C F is the flow coefficient, and represents an item
of potential controversy, particularly
for an orifice' C is
the coefficient of discharge, and F
llt/ 1-p'
velocity of approach faci-or. For a nozzle with high Rp
and a iow
iatio, K is close to 0.99 with a small uncer-
tainty that can become large, if the straight pipe and flow
straiihtener are not specified. For an orifice with Ra
K isiqual to 0'6205 for
F =
0'5' an{
to 0'699 for
B --
O.i. That is an increase in K of 12.7 percent, and the
O Report assigns an uncertainty of
K when it is increased from 0.5 to 0.7.
Air compressors are frequently equipped
with Venturi
tubes in th-e suction. The p essure loss is small, and K is
well established.
remember that the humidity must
be included in the computation of p.
F, is the thermal expansion
factor of the metering ele-
ment. It is of minor eflect.
Y is the expansion factor that corrects for the volume
increase, I" fbr a nozz'le, Yo fot an orifice, as the com-
pressible fluid expands from pt
pe in the measuring
The expansion factor for a nozzle is given in
PTC 10 as:
" " (L-
l;.i, ^
where: r
and nr is the isentropic exponent'
The expansion factor for an orifice is given in PTC 10
Yo= |
(0.41 + 0.35
9) lQ,- P)/@,
and originates in the Flow Measurement Report by the
Fluid Meters Committee. For
p ratios in the 0.5-0'6 range,
and (fu-
in the range of 0.03 pr or less, this factor
is in'ihe trigtr O.Ss, and the uncertainty may be on the
order of one percent.
p' is the density in the lb./ft.s at pr For some gasses it
is a reliable number, for some
streams it is subject
to conjecture. The conjecture stems from the deviation
from a perfect gas as expressed bY
04+ P) IQRT)
PM) l(ts+s
About the
H. Strox-Jonwsow is a consulting engi-
neer, East Ham,pton, N.Y, He recently
as cluief staff engineer wi,th
Elliott Co. After graduating
lunbia Unwe,r'sity with a B.S. ond, M.S.
degree in mecha,nical engineering, Mr.
S te en-J ohnsen joi,ned
Corp,, where he did prod,uct d,eoelop-
ment in turbi,ne control and design. Wi,th
Dlliott, he also was ch:ief turbine eng!-
neeq' and, chief product engineer super-
uising turbine and com,pressor deti,gn.
In this expression the compressibility Z is not
always known within
several percentage points for
certain gasses. And if p is not what the compressor is de-
signed for, the mass flow (zo) delivered will be an un-
pleasant surprise. In a centrifugal compressor, a heavy gas
picks up more pressure in a wheel than a light one, and
with more pressure goes less volume, and less volume
means a different area schedule for the wheel from hub
to tip. Thus uncertainty in p does not only 'affect w di-
rectly, but it also upsets the flow geometry of the wheel.
The expression for shaft horsepower is:
P"n:lu (ha- ha) *
Q,+ Q** Q'i142.408
Note: zu, the mass flow, is subject to some uncertainty,
as discussed above.
haand hr the enthalpy out and in, depends on accurate
pressure and temperature readings and. a good Mollier
diagram or similar gas information. And of course, where
(ha- he) is small, it is doubly important that the instru-
mentation is the best and that the readings are precise for
minimum uncertainty of the difference between them.
is the casing heat loss by radiation, and depending
on air circulation, this quantity rr:oy vary quite a bit. It
must be evaluated as a percentage of the whole.
is the heat equivalent of the mechanical losses
which should be quite well established.
is the heat equivalent of the seal losses. This quan-
tity may be quite debatable for a process gas, and the seal
conditions may have deteriorated. As a percentage of the
total, it may be a sigrrificant portion and affect the total
uncertainty of the power required.
With flow and shaft horsepower computed, the per-
formance of the compressor is established, except for an
evaluation of the uncertainty. The head vs. flow per-
formance is established
by the data for (ha
hr) .
From this discussion of compressor test, the several
origins of uncertainty can be listed, and the parties to the
test can arrive at some
percentage numbers, that rnay
be combined and applied to the firial computation. Con-
siderable help with this evaluation can be obtained from
the PTC 6 Guidance Report, which also outlines a pro-
cedure for combining the individual percentages into a
total uncertaintY.
pump grouting
Deficient grouting is an otten
unrecognized contributor to high
vibration and poor alignment retention.
The result can be
premature tailure of
bearings and seals. Follow these
guidelines to good grouting
Mqlcolm G. Murray, Jr., Murray & Garig Tool Works,
Brownsville, Texas
Evr,n rexB a tapping tour? Try it-you may be sur-
prised at what you find. Take a sma1l hammer and walk
arrrong the pumps in your plant, tapping the tops of their
baseplates at random. You will probably hear more hollorv
"thuds" than crisp "clinks." The thuds indicate voids or
cavities, rather than the even support supposedly provided
by the grout.
Groufing. Let's back up a bit. What is grouting, anyway?
Grouting, as concerned with here, is a procedure r'vhere a
semi-fluid mixture called grout fills the space between
machine foundation and baseplate, including cavities and
irregularities. After curing, the handened grout provides a
strong support, evenly distributing the weight and forces
from the machine through the baseplate and into the
Why is grouting important? The trend for the last 15
years has been to use fabricated steel baseplates and cen-
trifugal pumps with higher rotational speeds. These base-
plates are less rigid than the older cast iron designs, and
more likely to vibrate excessively at the higher speeds,
especially if grouting is deficient. Excessive vibration can
cause early bearing and seal failures. Poor grouting can
also result in poor alignment retention, with further pre-
mature wear. Good grouting generally costs more than
poor grouting, but in improved reliability and reduced
maintenance cost, oflers an attractive return on invest-
Much has been written on
grouting of c'ompressors, turbogenerators, and other heavy
machinery. By contrast, very little has been published on
pump grouting. Pump grouting may lack the appanent
importance or difficulty anticipated with the larger ma-
chines, but this assumption can be deceptive. Pump base-
plates can be trickier to grout than compressors or their
soleplates. Also, a pump often requires several times the
Fig. 1-An example of failed Portland cement-sand-water
Fig. 2-Estimating graph-pump horsepower and rpm vs. cubic
fe'et of grout required. For material cost, figure $40 to $60
cubic foot for epoxy grout and $25
per cubic foot for a good
proprietary inorganic grout. For grouting labor, flgure 1.5 man-
hours per cubic foot, with a minimum of 6 manhours.
Fig. 3-Grouting sequence illustrations. This procedure pro-
vides an easy means of leveling, eliminates shims and wedges
and allows pressurizing the main cavity without overflowing
the form,
Fig. S-Pump baseplate leveled with screws resting on small
bearing plates,
Fig. 7-Using. a pneumatic.diaphragm pump to place the grout
(inorganic only-not epoxy).
Fig. 6-The rapid-cure mixtureo is quickly troweled in place to
close the gap around the baseplate lower perimeter. This mix-
ture will be hard to the touch in 15 minutes.
Fig. 8-The baseplate cavity 1s nearly filled. Weighted covers
are used to prevent overflow while grout is force-pumped into
remaining voids.
Fig. 4-Grouting equipment tool cart used at a major overseas
oil /efinery.
grout volume of a colnPressor hlvinq equal horsepower.
Finally, the typical
plant has about 10 times as
many pumps as all other rtrachines
togeiher, so they
are statistically important eveIl if less noticeable indi-
Great care normally is exercised in grouting a com-
pressor, with engineers, erectors, and top field forces mak-
ing every effort to use the best materials and procedures
to get a good
job. A typical'Pump grout job,'on the other
hand, is usually left to the discretion of the plant or con-
tractor labor crew. Generally a rather soupy Portland
cement-sand-water mix is used, perhaps with some iron
powder to offset shrinkage. The result is often poor-
Fig. 1 shows such a grout job altet a few years.
Pump grouting-improved. Good pump grouting is
not difficult, trut it does require careful preparations and
planning, and attention to numerous details, as follows:
Baseplate and foundation. Certain baseplate design
features, such as leveling screws) grout filling and vent
holes and corrosion protection, are recommended.s These
provisions cost little and make good grouting much
easier. The foundation is also important. It should extend
at least 3 inches outside the leveling screws. Its surface
should be chipped back
to 1 inch to remove the
laitance, or weak top layer, to provide a rough surface
for bonding.
Leveling. With vertical scrervs, the baseplate can be
Ieveled easily without using wedges or shims (Fig. 5).
Such scrervs also permit "single-pour" grouting, thus
saving time and labor. After the grout cures, the screws
can be removed to let the baseplate be supported evenly
on the grout.
Materials selection. No single grout material is best
for all
jobs. The traditional Portland cement-sand-water
mix should not be used, hor'vever. When soupy enough to
flow easily, it gives a rveak grout subject to shrinkage and
cracking. Although much stronger if dry-packed,5 it is
difficult to place in the typical pump baseplate cavity'
Iron-additive grouts also are not recommended. Cases on
record show such grouts have continued expanding for
many yeam after installation, causing repeated machine
The choice of grouts should be made from among the
"good" proprietary mixtures. Such grouts fall into two
categories. These are the "organics," or plastics, generally
epoxies or polyesters, and the non-metallic "inorganics"-
basically Portland cement with controlled silica aggregate
and non-metallic anti-shrink and cure-acceleration addi-
The plastic grouts are the most exPensivg-two to
three times the cost of the inorganics. On a single large
pump, this cost diflerence can easily run
the plastics are subject to creep above 200o F and should
not be used where baseplate temperatures are expected to
exceed 175o F over major surface areas. Ilowever, they
are easy to use, cure adequately in a'bout 36 hours and
have high compressive strength. Plastics are also resistant
to many chemicals. As a general rule, use plastic grout
on small pumps, up to about 6 cubic feet maximum grout
volume. Beyond this, the ease of grouting is more than
offset by the higher material cost.
The proprietary inorganic grouts require mixing with
Fig. 9-Using a simple topping pump to fill the small space
beneath the baseplate top surface. This type of pump can
handle any
grout, including epoxies and coarse aggregates.
Fig. 10-For the inorganic grouts a paddle mixer speeds the
ln this design, the mixing carriage detaches and is wheeled
to the grouting site. The mixture can be pumped directly from
carriage to baseplate.
is compte,";ilt.
,."a to retain
moisture for a three-day cure. A watering manifold keeps them
wet automatically with minimum water waste.
Fig. 12-Sealing the leveling screw hol'es with epoxy.
Grouting equipment. Grouting can be done with
simple and inexpensive equipment. If more than a few
pumps are to be grouted with inorganics, it pays to obtain
additional items such as small vibrator, a diaphragm
pump and a paddle-type mixer (Fig. 10). Fig. 4 shows
a portable grouting equipment tool cart and its contents.
Such a kit is handy for keeping these tools together and
having them on hand for each grout job.
Grouting procedures are most effective if thought
out in advance, step by step, and written down. Copies
then can be distributed to all concerned, and a short
meeting held several days prior to the
job. At this meet-
ing, procedural details are cleared up, and resp,onsibilities
agreed upon for getting necessary personnel, equip-
ment and materials to the job
site. With modern, rapid-
curing grouts, the middle of the job is not the time to
discover that something essential is missing! Literature
citation 2 includes detailed procedures for several grout
In broad terms, the gncuting procedure should include
the following information:
A. Preparatory steps such as foundation cleanup and
oil removal.l
B. Forming instructions. Generally the form top should
be z/ainch
above the baseplate bottom edge.
C. Equipment and materials required.
D. Grout mixing and placement instructions, including
pump, mixer and vibrator usage where appropriate. Some-
times timing is critical and must be emphasized. One in-
organic grout requires a minimum of 2/z minutes mixing,
but becomes too stiff to pump if mixed longer than Z/z
minutes! See Figs. 3,6,7, B, 9 and 10.
E. Curing instructions. See Fig. 11.
F. Leveling screw removal and hole sealing instruc-
tions. See Fig. 12.
poor grout
Sometimes a poor grout
can be salvaged, saving the trouble of completi re-
gncuting. This is done by injection of epoxy or polyester
liquid, without aggregate, to fill the cracks, voidi and
cavities in the old grout. Fig. 13 shows this procedure.
Fig. l3-Pumping epoxy through grease
fittinos to fill voids
baseptate to satvage a poor-grout job
avoro the need for complete /egrouting.
water (sometimes icewater), and are somewhat trickier
to use than the plastic gr-outs.
will, hou,e,",er, rvith-
stand temperatures up to 5000 F (1.0000 F for one mix-
Centrifugal Pumps
_ )
Howto improve
pump performance
lnducers have been used tor some time
to incre;ase centritugal pump IVPSH,
A new design teatures an inducer that
is integral with the impeller.
Results are better suction pertormance
and efticiency. Here are the
details of tfris new development
Sh. Yedidich, Worthington Standard Pump Corp., East
Orange, N.J.
Trrn usnrur, wonr performed by a centrifugal pump
varies as the cube of its speed. For example, when the
operating speed of a pump is doubled, its flow-rate is
doubled and the head increases four times (i.e. the power
output increases eight times).
It would, therefore, be most economical to operate a
centrifugal pump at the highest speed possible.
However, the use of centrifugal pumps at high oper-
ating speeds has been restricted, for many years, by Net
Positive Suction Head (NPSH) limitations. Only in re-
cent years has there been a dramatic breakthrough with
the commercial application of inducers.l Fig. 1 shows a
centrifugal pump equipped with such an inducer. Several
pump manufacturers offer such inducers as an option, and
they have become widely accepted by users. Fig 2 shows
how such an inducer is capable of improving the suction
capability of a pump and, at the same time, of preserving
its head-capacity relations and efficiencies.
As flow rates increase and pumps become larger, the
normal two piece impeller-inducer combination experi-
ences some problems in both hydraulic and mechanical
design. These are ove.rcome by the natural joining of the
two pieces into a one piece "inducerpeller".
Figs. 3, 4, 5 and 6 show different variations of inducer-
pellers. As can be seen from these illustrations, the main
features of such an inducerpeller are as follows:
The blades of the inducerpeller extend far into the
eye of the impeller where they form a geometrical
shape which resembles, to a certain degree, either a
Fig. l-Centrifugal pump with inducer.
Flg. 2-Performance of a stendard 6 X 4 X I pump wlth and
without an inducer.
Fig. 3-Compressor wheel with integral inducer blades. Fig. 4-prrp impeller with axial blade extensions.
Fig. S-Experimental impellers with integral inducer blades.
screw-conveyor or an axial flow impeller.
Downstream of the axial inlet, the orientation of the
blades starts to change gradually as does their shape.
Near the outlet, the shape of the blades assumes the
form of regular impeller vanes.
The basic idea of such a blade configuration is not new.
It has been used for years in the compressor industry (Fig.
3), and is used frequently in pumps (Fig. a). However, in
compressors, the sole eflect of such a blade configuration
is the increase of the pressure-ratio. In pumps, the geom-
etry represented by Fig. 4 is often used for the same pur-
pose as in the modern inducerpeller: to increase the
suction capability. However, the improvements achieved
with such a geometry were usually very small. Even the
most successful designs were much inferior to the results
represented in Fig. 2.
It rvas evident, therefore, that to develop an inducer-
peller which would be capable of matching the perfor-
mance of a pump provided with an extra inducer, an
extensive research program lvas needed.
The problems to be solved were not easy ones. For
example, one of the "secrets" of the success of an impeller
p.orii"j with an extra inducer (Figs. 1 and 2) is the
fact that the inducer is designed for a different flow rate
than the irrpeller. Such a con'rbination is impossible when
the inducer blade and the impeller vane form one con-
tinuous surface. In this case, both the inducer-part and
the impeller-part of the blades have to be designed for
exactly the same flow rates.
Fig. G-lmpeller used for test shown in Fig. 9.
3.500 FPM
TAReET rFFrcrENcy/
7-lnducerpelter with good efficiency but poor
suction performance,
In a- cgrlain way, the awareness of this fact happened to
be useful. It showed us the need for a new appro4ch, with
geometries completely different from that shown in Fig. 4.
However, each of the variations tested furnished useful
information, which was used for improving the results of
the subsequent tests.
After extensive testing and analytical work, we arrived
at a design which was capable of fulfilling all require-
ments. It has shown performance and efliciencies which
were formerly believed to be attainable only with con_
ventional impellers, and suction performance equal to a
Aboul the
Srr. YSDIpHH is o eeialist
yth Wwthingbn
Corp., Eost Orange, iea aie
to det;elop and prouide specifi.c solutions
to problems
relating to centrifugal
pum,ps, to create neut designs and a,ct
a,s cornpanA hgdrauli,c design consul-
tant. Before
Yedidiah serued as t
and, consulting eng
ooerseas marutfacturers. He has pub_
Lished many technical papers and articles ond, holds seieral
patents. He is a registered professional
engineer in an, ouer_
seas countrA and a member of ASME.
Fig, 8-lnducerpeller with good suction performance
but poor efficiency.
s 5,0[ t00[
'ml 8ol
,ol mf
20,000 @
Slneformance of inducerpeller versus impeller with in-
dividual inducer.
pump provided with an extra inducer. Fig. 6 shows a
photogiaph of the final design, and its pJrfo.m.rc" is
given in Fig. 9. For comparison, test data fiom a standard
pump with an individual inducer are shown as the dashed
line on Fig. 9. The successful results of the
design are clearly seen from these graphs.
The successful completion of the research program has
opened up new horizons for the pump industry. Larger
operating at higher speeds can be built thus keip-
ing the size, space requirernents and total investment costs
efliciency, in turn, can bring about signfficant savings in
power consumption.
"Inducer Blades Cut Into Required NPSH,,,
How to control
Many factors are involved in causing
to vibrate. Smooth operation
begins with proper
selection and
design tor a specilic service. lf you
want reliable pertormance,
is what to consider when
matching pump
to seryice
W. P. Honcock, Shell Curacao, Netherlands Antilles
Trra or;rcr of the study was to determine the effect
on vibration levels of those factors inherent in the process
pump type and the severity of service. Therefore, before
selecting the pump sample
,to be studied, it was essential
to eliminate those factors affecting vibration levels that
may be attributed to faulty manufacture, installation, and
The sample of about 100 pumps chosen, therefore,
had been mechanical balanced, well aligned, free from
cavitation, and operating stably near peak efficiency.
The pumps were in an age range of 2 to 25 years.
Vibration levels were then charted weekly for one year
for each pump. Readings were taken on the two bearing
housings in three directions (horizontal, vertical, axial)
using a vibration velocity meter indicating "unfiltered
peak" velocity values. At the end of the year-long
study, an average value of vibration velocity was com-
puted from the 312 (2x3x52) readings taken for each
Furthermore, in order to ascertain the correlation be-
tween pump reliability and vibration level, the main-
Fig. l-Vibration leVel according to pump type and tem-
of service.
TABIE f-gummqry ol process pump designs ond vibrqiion crilerio
mm/eec. lns/eec. mm/oec. lns/sec. mm/sec. lns/sec
An excellent temperature compensated design.
Vibratiou is independent of temperature.
Fig. t
8.0 20.o
capable of withstanding
ins to 500oc (932oF),
Radial Split.
More flexible design.
Uncompensated designs suitable only for cold
Existing compensated designs satisfactory to 340.C
(644'F) need further re6nement to reach specified
425.C (800"F).
Dynamic performance
see above. "Back to back"
impeller combination glves thrust problems.
Front to front is much improved but gives higher
stumng box pressure.
Multi-stage. .
Barrel casing.
A robust design rvell compensated for temperatul
efiect suitable to 450.C.
The design susceptible to internal deformatior
causing sticking of hot pumps q'hen priming. Im
provements required on method of pump suDport
for application above 250'C (4820F).
ing. Im-
tenance histories were studied for the last three years.
Regarding the inherent factors affecting vibration
levels, the following emerged in descending order of
Pump design
Service temperature
Size effect
(u) by increasing size at constant rpm, or
(b) by increasing rpm of a constant size
Head per stage and number of stages
Comparison of packed gland pumps with mechanical
seal conversions.
Pump type. Pump type is of fundamental importance
to vibration behavior. Consequently, samples of each
of the following pump types were studied.
(a) End suction-overhung impeller type
(b) Single and 2 stage - radial split case-volute
(c) Multi-stage radial split-difluser
(d) Vertical shaft-in line branches-direct coupled
The vibration norrns according to pump design and
other factors are shown in Fig. 1 and Table 1.
fempercture effect. Undoubtedly the most signifi-
cant factor affecting vibration levels is the temperature
effect which produces a change in alignment between
pump and driver.
If a pump is designed so that its movements due to
expansion are controlled and equalized, then a change
of alignment with temperature is not possible. The pump
may then be termed "temperature compensated."
Most present-day standard hot pumps are either "un-
compensated" or "pa.rt compensated" and the eflect of
this is clearly illustrated on the vibration vs temperature
norrns (Fig. 1). Vibration level increases rapidly with
temperature. The standard end-suction designs perform
extremely well. They are compensated vertically by sus-
pending the pump about the shaft center-line (in common
with almost all process pump types) but are not com-
pensated in the horizontal plane. The design also has the
advantage of a robust shaft between a short bearing span
and consequently runs below the first critical speed at
3600 rpm.
The standard radial split volute designs having much
longer bodies are more susceptible to horizontal movements
and the uncompensated types vibrate excessively above
l00o C (2120 F). At 3600 rpm they generally operate
above the first critical speed, See Fig. 3(b).
Designs which are part-compensated by using longi-
tudinal and/or transverse casing keys perform much
better to a maximum temperature of 3350 C (6350 F) .
For this type of design and the multistage barrel de-
signs, compensation is essential above 1000 C and is
desirable even for cold service.
The temperature compensated pump. To maintain
alignment under all service temperatures, the pump must
expand freely and equally on its pedestal supports.
Most existing designs are not permitted freedom to
move since they are solidly bolted to the pedestal sup-
ports, thereby subjecting the casing to considerable forces
in the hot state. This is the major reason why many
pumps become stuck when primed with the hot product.
In practice, the only solution to this problem is the use
of excessive running clearances (up to 0.050 in. at
B00o F) which incur significant efficiency loss.
The pump feet must be free to move and yet pre-
vented from lifting by special hold down arrangements
(Fig. a) which are fairly standard in steam turbines and
high-speed boiler feedwater pumps.
PEDESTAL IE|GHT, H, Ct (ti[i.)
alignment correction curves for pumps on hot
the supporting pedestals are maintained at ambient
temperature. In practice the pedestals conduct appreci-
able heat from the pump body (Fig. 2) and the pedestals
can grow up to 0.014 in. (0.036 cm) for a large pump at
750" F (3980 C).
The pedestal growth can be determined using Fig. 3
and a cold alignment correction made.
Vertical displacement of the driver also takes place,
but generally for motor drives, the temperature rise is
so small that the displacement is negligible.
For turbine drives, the vertical displacement must be
taken into account with the pump vertical displacement
in order to ascertain the relative vertical displacement.
The turbine displacement is usually specified on the
manufacturers drawings or can be estimated using Fig.
3 by substituting steam temperature for product tem-
For turbines it is also very desirable to use center-line
Horizontal compensation. Pump movements in the
horizontal plane can be controlled by using longitudinal
and transverse keys. On the bigger designs, some manu-
facturers use two longitudinal keys, while others use
both transverse and longitudinal keys. However, the
fiH XErl@S PftEvBfi Rm h0fit LFflls t,I tu.oynE mExHI urEmtty ]
Fig. 4-Suggest'ed improved radial split case design for hot
biggest fault with existing designs is the solid bolting-
down arangements dealt with previously.
Of the existing key arrangements in use, two lon-
gitudinal keys gave the best results for the radial split
casing and barrel types. Of the end suction types, very
few were fitted with keys, but the use of one transverse
key at one pump foot and a center peg beneath the
pump body proved to be very effective. Although the
uncompensated end suction pumps gave good results up
to 3500 C (660" F), the few compensated pumps
,much less and should be suitable up to 5000 C
(9320 F). The multistage barrel casing types are well
compensated and suitable for temperatures up to 4000
c (7520 F).
It can be seen that the biggest need for improvement
lies with the radial split voluted designs. A suggested
improvement is illustrated in Fig. 4. The revised holding
down arrangements would be necessary on the longer
bodied pumps operating at temperatures in excess of
2500 c (4820 F).
Even better compensation can be obtained by
the pump and its pipework in order to achieve perfect
thermal symmetry.
Several vears ago a compensated high speed feedpump
l0 lrs{to mtT (8) lacfirlGD w SHER
Fig. s-An ideally "temperature compensated" design.
was designed and built with twin suction and twin
discharge branches arranged symmetrically together with
the pipework systems. This thermally compensated de-
signl greatly alleviated problems of distortion and align-
IJnfortunately, the duplicated pipework systems are
more costly and this may be the reason why the design
appears to have lost popularity. However, there are few
advanced refinery duties which would
such a
Ideally compensated designs. The pump designs that
are ideally temperature compensated are those types
mounted integrally with their'driver (usually a motor).
The type which is in common use in recently constructed
refineries is the direct coupled in-line branches pump
(Fig. s).
A limited sample of the in-line types was tested and
gave excellent results to 2000 C (3920 F).
Size efiect. The effects of size can be produced in two
1. Increasing size (eeomctricalll'
sintilar designs) at
constant speed. Most process purnps fall rr,ithin a fairll,
narrow range of scale factor for ear:h particular design
and consequently size factor had littlc eflect orr vibratiorr
level. However for general service pumps such as the
single stage horizontal split casing types, it is possible
to have a wide range of sizes from 2 to l2O in. diameter
discharge. For these pumps, size effect can be predicted
to have considerable eflect on vibration level. This is
because the ratio of unbalanced dynamic forces to pump
stiffness iircreases disporportionately with size, thereby in-
creasing the amplitude of vibration and hence vibration
velocity. In practice of course, the eflect is always offset
by progressive speed reduction. It is important however
to realize that for a given design (retaining the same
speed) there is a limit to which the size can be increased
if vibrations are to be maintained at a reasonable level.
2. Increasing speed of a given pump. For process duties
this is by far the most common manner of inducing size
effects which greatly influence vibration levels. Consider-
ing the pump as a simple undampened spring subjected
to imbalance forces, it is possible to predict the amplitude
of vibrations with rpm.
Imbalance forces are proportional to (rpm)2, there-
fore, amplitude is also proportional to (rpm)r. Further-
more, frequency is directly proportional to rpm, hence,
vibration velocity is proportional to (rp-)r.
In practice, the influence of dampening effects and
the pump critical speeds modify the theoretical relation-
ship considerably. A number of pump types were tested
in order to obtain an empirical relationship between
vibration velocity and rpm.
The results are shown in Fig. 6. All the end suction
pumps operated as rigid rotor types below 4,000 rpm.
Moreover, vibration velocity is approximately propor-
tional to (.p-)'. For the flexible rotor types having
the first critical speed below 4,000 rpm, it is not possible
to establish a speed correction factor. It is, however,
enlightening to notice the close proximity of the critical
speeds to the two pole motor speeds of 3,000 rpm (50
cps) and 3,600 rpm (60 cps). This factor certainly
accounts partly for the relatively poor dynamic behavior
of the flexible rotor pumps compared with the stiff rotor
end suction types.
API Standard 610s is very specific with regard to
recommendations on critical speeds and it is hoped
that these may be fully implemented in future process
pump designs.
Most existing standard process designs are rated for
a maximum of 3,600 to 4,000 rpm and most process
duties require a speed of 3,000 or 3,600 rpm for single
stage pumps.
Efrecl of heod per s?oge ond number of sloges. It
is reasonable to expect that head pe.r stage affects vibra-
tion level since all pump casing types give some un-
balanced hydraulic reactions, the magnitude of which
depends on the casing design utilized and the head/stage
For process pumps, most manufacturers use single-
Fig. 6-Effect of rpm on vibration level.
volute casings up to 100 M. of head/stage and double-
volute casings for 100-400 M. of head/stage.
For multi-stage designs, the multi-channel diffuser
inside a barrel casing is in common use.
The diffuser gives the best radial hydraulic balance
over a wide flow range. Single-and two-stage pumps use
single-and double-volute casings because they are
simple and inexpensive. These types provide good
hydraulic balance at best efficiency flow but relatively
poor balance below 25 percent of best efficiency flow.
These designs show signifi'cant increase of vibration level
below 25 percent of best efficiency flow (Fig. 7). Con-
sequently, they should not be run continuously below 25
percent of best efficiency flow even though they may be
suitable for low flow operation on the basis of permissible
temperature rise across the pump. Diffuser designs do
not sufler to the same extent because of better hydraulic
balance at low percentage flows. The study revealed that
for pumps operating near best efficiency point, head/stage
had negligible effect on vibration level. It must be
stressed, however, that the impeller-volute generating
circle gap should be a certain minimum distance.2
In the past, certain pumps were found to be vibrating
excessively and the cause was that this gap had been
reduced to only l/o of impeller diameter by re-welding
new casing tongues. When the tongues were modified to
give a l0/o gap, the vibration disappeared.
Number of stages. ft was shown previousln that for
rigid rotor single stage pumps, a decrease in operating
speed lowers vibration levels and incJeases reliability.
For multistage flexible rotor pumps, it is a diflerent
matter, since vibration levels above the first critical
speed remain fairly constant with increasing speed.
Furthermore, the advantages of less stages and a shorter,
less complicated pump are obvious. High pressure feed
pumps are now built to run at speeds exceeding 8,000
rpm and generating over 100 kg/sq cm (1422 psi) per
Effect of pocked glond in dompening vibrolions.
Contrary to popular belief, the study revealed little
evidence to substantiate the theory that packed gland
pumps (with the advantage of additional support) vi-
brated less than pur4ps converted to mechanical seals.
A sample of 24 identical 4x6 in. end suction pumps
was examined, eight of which had been converted to
mechanical seals. The results (Fig. 8) clearly indicate
that the change had no effect on vibration level.
It is conceded, however, that on the very long flexible
shafts of the multistage pump, the packed gland may
slightly dampen vibration.
New process pumps designed especially to incorporate
seals permit a much shorter rotating element thereby
reducing the maximum static deflection of the shaft
to one-half to one-third of the packed gland pumps. This
pump type is expected to show a clear advantage over
the Ionger packed gland type.
The vibration norrns will enable the maintenance en-
gineer to judge the running condition of each of his
process pumps against the average vibration level ex-
pected for the particular pump design and service. The
other vital factor that now has to be considered is the
Fig. 7-Effect of part load operation on vibration level for a
double volute pump.
lElPInAIURe, r(T'
Fig. 8-Effect of packed gland/mechanical seal conversion
on end suction type pump.
vibration limits that each design can withstand.
In practice it is necessary to define two vibration limits:
vibrqtion level. This should
be the criterion to be used for all vital process pump
(i.e., those that heavily penalize plant production when
they breakdown) also those pumps on highly hazardous
service. It is the vibration level above which production
losses and maintenance costs begin to soar. It can be
shown that for a high penalty pump, a pump availability
of 0.995 has to be achieved, (availability
hrs. in service
- hrs. in repair/hrs. in service) . In order to determine
the vibration level consistent with this order of avail-
ability, maintenance histories were carefully studied for
the last three years. Availabilities were then plotted
against average vibration level during the current year
for each pump.
IJsing this curve (Fig. 9), it is possible to define
approximately the maximum acceptable vibration level
for any required availability. For vital pumps the vibra-
tion limit is therefore 8 mm/sec.
Donger level. Many process pumps fall outside the
category described previously, inctrrring little or no pro-
duction losses when out of service. A much lower avail-
ability can therefore be tolerated, and hence it is per-
missible to allow such pumps to operate above 8 mm/sec'
However, a much coarser vibration limit has still to be
used and this may be termed "danger level." This can
be defined as the vibration level above which there is a
high probability of imminent pump failure. This is the
level which
justifies immediate shutdown since there is
a risk of consequential damages. During the last three
Fig. 9-Relationship of pump
availabitity with vibration levels.
years, several trials have been conducted on pumps sub-
to very high vibration levels. One end suction
pump was run at 50 mm/sec for three weeks without
breakdown, but this was exceptional. Another run at 30
mm/sec lasted only 48 hours. A further twelve pumps
running between 15-25 mm lasted from two months to
one year.
On the basis of the practical findings, the following
tentative danger limits are proposed:
(a) Rigid Rotor pumps (e.g., end suction) 20 mm/sec.
(0.8 inlsec.).
(b) Flexible Rotor pumps (i..., 1st critical speed
(below 3,500, e.g., two-stage volute, multistage) l5
mm/sec. (0.6 in/sec.).
These will serve only as a tentative guide. In practice, it
will be found that for certain designs on severe service,
risk of failure will be imminent at only l0 mm/sec.
These special cases will be known and will account for
only a small percentage of the pumps, the bulk of which
will behave in a predictable manner.
Generc!. Vibration behavior is a sound basis for use
with preventive maintenance or controlled breakdown
maintenance procedure. In the former case, 8 mm/sec.
would be the criterion used for scheduling a pump
repair. Pumps in this category would receive vibration
logging on a routine basis together with other checks
such as noise, bearing temperatures, heads, flows, and
power readings. With a skilled running equipment in-
spector, early diagnosis of future problems is possible in
at least 95/o of all cases. The number of unforeseen
failures is small and the production and maintenance
savings easily justify
the costs of regular inspection.
Vibration trend lines should be charted for each pump.
A sudden rise of 2-to 6 mm/sec. vibration level may
be more indicative of early bearing failure than a pump
running at a constant 15 mm/sec. for the last six months.
In such a case the inspector would check all possible
operational causes such as Iow flow operation or cavita-
tion. The frequency of vibration would be checked since
a high frequency vibration is indicative of wear of anti-
friction bearings and in its more advanced stages produces
the high pitched sound which signifies that the bearings
life is quickly drawing to a close. From early wear in-
dications to advanced wear nornally takes
one week to several months and usually gives sufficient
time to effect a planned repair.
peratures by hand. Any major change in pump behavior
will certainly be noted and will be suffilieni incentive
to call on expert assistance to make a detailed examina-
still be a number of complete
r,, the financial consequences
ere is no financial justification
Hol olignment. For noncompensated pumps, hot align-
ment is very useful, but must be carried out after the
plant has been in operation for sometime and then
executed as quickly as possible. It is not sufficient just
to prime the pump with hot product because this gives
asymmetric heating to the pipework, since the discharge
pipework stays cold.
Ifowever, hot alignment can be both dangerous and
time consuming and is the partial cure rather than the
prevention of the problem. Moreover, hot pipework is
constantly moving the uncompensated pump dependent
on ambient temperature and wind conditions.
Robusiness of the pump design ond its pedestols.
The pump desigrr needs to be sufficiently robust to with-
stand forces and moments from the hot pipework. Most
older designs are extremely solid and withstand quite
large forces without change of alignment. IJnfortunately,
economic pressures are forcing some manufacturers to
produce much lighter designs which perform well on cold
service but vibrate exessively in hot service. In this
respect, it is felt that the section of API 610 dealing with
vibration criteria could be improved, since it applies only
to vibration levels of the pump tested under cold service
in the manufacturers works.
Pedestals should also be fairly rigid. On one end
suction boiler circulating pump with high pressure, high
temperature pipework, the pedestals were so flexible that
the coupling alignment changed 0.150 in. when the pump
was heated. Considerable reinforcement was necessary
to obtain a satisfactory result.
Vibration measurement provides a sound basis for
establishing the running condition of process pumps. The
vibration data evolved by the study will provide an ex-
cellent foundation on which preventive/corrective main-
tenance programs can be built.
For the standard process pump, the principal in-
herent factors affecting vibration level are pump design,
temperature of service, and rpm.
Vibration behavior under ideal cold conditions will
not be sirnulated under actual hot working conditions
unless the pump is temperature compensated. Many
existing process designs are not temperature compensated
and perform poorly at elevated temperatures.
The in-line pump and the barrel casing designs are
well compensated and vibration is virtually independent
of temperature. Most end-suction and single/2 stage
Fig. 10-Process pump selection chart showing limits of ilows, heads and temperatures.
radial split designs are nontemperature compensated and
the temperature effect on the larger bodied radial split
designs is quite considerable.
All standard process designs have solid bolting down
arrangements which do not permit free (yet controlled)
movement of the pump with temperature. Consequently,
the casing distortion duing priming with hot product
often makes the pump rotor stick. The only practical
remedy is the use of excessive running clearances which
considerably reduce pump efficiency. The most suscep-
Aboul fhe
Paur, HaNcocr is head of rotating
equipment engineering
Shell Cura-
cao, Netherlands Antilles. He is respon-
or application, troubleshooting,
aho,lysis and
studies of refinerg rotating equipment
ulhich cotttprises ouer 2,500 pllnlps, conx-
pressors, blowers and turbines. Prior
etperience includes seraiec as a pump
design engineer u;ith, Matlter and Platt
Ltd. and project engineering with the
Central Electricity Genero,ting Bcard. Paul uas born in Eng-
land and receiaed, a B.Sc.Teclt. degree in mecltaruical eng'ineer-
the Unittersity of Manchester. He is a chartered
engineer of the Institute of Mechanical Engineers (London)
and a member of ASME.
tible of all pump types to this phenomenon are the long
multistage designs. For designs of this nature, considerable
improvement could be made by using holding down ar-
rangements incorporating either the "paw foot" design
or the shouldered bolt which allow free lateral and lon-
gitudinal movements but prevent vertical movement.
These more sophisticated designs would be
justified for
pumps operating above 2500 C (482' F).
End-suction and in-line designs give excellent me-
chanical performance. Furthermore, they have low initial
cost, are easy to repair, and have a wide degree of
interchangeability. Consequently, they should be chosen
as widely as possible for process duties. In certain cases,
two-stage and even multistage designs are used for duties
that could be satisfied by end-suction designs.
A complete guide to the limitations of head flow
and temperature are shown in Fig. 10. The vibration
norms according to design and severity of service are
given in Fig. I and Table 1 which summarizes the main
findings of the study.
The author is indebted to the
practical assistance rccciycd from
Jr B"i1:
in the substantial field worked involved with the study. Thir papcr origin4h
3i.",ill.l"3n1l,,$.13_Yf ,De5ign
Eneineering Technial conlcrcncc, cin'
Fig. 1a-By viewing the impeller through
the suction portion
of the pump
the initial stages of cavitation
seen with vapor beginning to form on the
entrance edges of the vanes.
Fig. 1b-Little noise or reduction in
pump p.erformance
is noted in lhis stage
of cavitation. lmpeller vanes may shdw
some pitting
after extended oferation
under these conditions, especially if
there is any tendency toward corroSion
Fig... 1c-As a result of increasing the
suction on the pump,
which decr6ases
the pressure
in the entrance of the im-
vanes, the vapor extends over a
large portion
of the entrance. Noise re-
duction of efficiency should be evident
by this stage of cavitation.
at the impeller vanes
by further increasing
marginal at this point,
Fig. l-Photog{aphs_.91
cavitation in action courtesy of
less Pump Division, FMC Corp.
a "ping" against the vane surface. A pump impeller
exposed to cavitation will have a "shot piening,, on the
su'rf ac e' c avi tati on'$":ii"
;ilf :," .:ffi ,f,'t'iji:;
The threshold pressure required for a vaporized liquid
to implode is called the "suppression preisure.,,
manufacturers perform a suppression test used for NpSH
Fig. 1 illustrates the cavitation process as seen on im-
peller vanes viewed through
a transparent suction line.
How to
Cavitation can destroy process
pump perlormance
and, in time,
will destroy the pump
Caretul attention to fluid
conditions at the pump
will minimize the possibility
of cavitation
Chorles Jockson, Monsanto Polymers and
Petrochemical Co., Texas City, Texas
Cevrrerrorv is a breach of contract! IJsers c,rntract
with a ma
that will perform
at specific
th parties. Some-
thing clran
ching the contract
and result
i." the basic
factors that lead to cavitation and how to avoid them.
Cqvilqfion occurs when the pump suction pressure drops
causing some liquid to vaporize and bubbles to form. As
these bubbles enter the pump vanes, the pressure in-
creases until the bubbles implode or burst inward making
Fig. 2 shows a damaged impeller taken from a 30-inch
cooling tower pump from our Chocolate Bayou, Texas,
plant. Note the picture through the mirror showing cavi-
tation damage progressing through the vane.
Whqt is NPSH? It is the Net (remaining) Positive
(greater than zero) Suction Head (absolute pressure at
the pump suction flange expressed in feet of liquid
pumped above the liquid's vapor pressure at the pump-
ing temperature).
To the system designer, NPSH is the remaining pres-
sure (net) at the suction flange of the intended pump
af.ter all negative forces that restrict liquid from getting
into the pump are subtracted from all the positive forces
that assist liquid in getting ino the pump. NPSH varies
strongly with change of flow. Most pump curves are
plotted with NPSH versus capacity,
in gpm. But
NPSH characteristics allow flows not frequently plotted
by pump manufacturers. It is not uncommon for the
NPSHR curve to increase or turn upward at low flows
(see dashed lines on NPSH curve, Fig.3). Low flow ef-
fects can often be amplified by selecting too large a pump
for the service. This forces recirculation within the im-
pellerl (sirnilar to "surging" in a centrifugal compressor) .
An instability point on the H-Q curve is now indicated
by some pump manufacturers (see dashed lines, Fig. 3).
NPSHR. Purchased pumps should have received a care'
ful suppression test following the original pump design.
This test should accurately predict the pump's suction
performance on some standard fluids. Water is usually
used. The NPSHR (Net Positive Suction lleads"n,1."6)
by the pump manufacturer is confirmedlby this test.
To the pump manufacturer NPSHR is the head of
Iiquid required, at each proposed flow rate, for the liquid
to travel from the suction flange into the pumping vane
area without vaporizing. Vaporization can be thought of
as "cold boiling" from the pressure drop in the liquid as
it flows from Point A to B (see Fig. 4) . If NPSHp colr-
ditions are met and entrained gas is low ( (
1 percent
by volume),' the pump manufacturer states that the
pump should perform according to the H-Q curve pro-
vided for that size, impeller and speed. (See Fig. 5 for
effects of entrained gas on H-Q curves.)
NPSHA. The purchaser or user must calculate the NPSH1
(Net Positive Suction lleade,,rrurr") at the suction flange
of the pump for the range of system Head-Capacity
(H-Q) cu.rves.
All equations for determining NPSHa are simply equa-
tions in fluid flow.
Push fluid to
Pressure over the fluid.
Static head of fluid above center
of suction flange.
Note: Sometimes the reference
point is to the pump centerline
or an agreed reference (datum)
Atmospheric pressure.
Velocity head.
Fig. 2-Cavitation damaged impeller. Note reflection from the
mirror of damage to back side of impeller.
Suppression lesl. The pump manufacturer provides a
suppression test (also called cavitation test). The test
procedure is outlined in ASME PT 8.2 for centrifugal
pumps. A test set-up is shown schematically in Fig. 6.
Measured flow is sometimes provided by tested flow
nozzles changed out consecutively. An accttate, often cali-
brated flow measuring station is also used.
Example. A 200-gpm nozzle is installed and the pump
is throttled on the suction until the purnp cauitates. The
NPSH value is plotted versus 200 gpm. Actually, audible
cavitation is seldom obtained but the beginning point of
cavitation is considered to be the condition at which a
three-percent deviation in total head occurs. Next, the
300-gpm flow nozzle is installed and the procedure re-
peated progressing to maximum pump flow capabilities.
Push fluid back from pump
Friction loss in piping system
(including entrance or exit
loss from vessels).
Vapor pressure of liquid.
Fig. S-Typical H-Q curve. (Courtesy Bingham Pump Co.)
Fig. 4-Cross-section of a typical pum.p
shows the pressure
drop area (A to B) for vaporization (courtesy Bingham
Relative pressures
in the entrance section of a pump are showrr on the ihart airight, Note pressdre drop'betw;en poinis'A an'o
B (courtesy
The Duriron Co., lnc.).
Performqnce. The performance curve (Fig. 3) indicates
that at 180 gpm and a head of 215 feet, the NPSHp is
10 feet. Calculated NPSHA might be 72 feet for this
example. Note: This pump will cavitate at 216 gpm if
NPSHA is 12 feet.
Provided the pump manufacturer made the suppression
test, the pump could fail to perform at operating con-
dition because of:
Increased circulation from wear rings or clea.rance
settings upsetting flow into the impeller,
Deteriorated impeller, e.g., inlets, imperfect vane,
plugged passageway, corrosion/erosion holes, vane exits,
Plugged suction or suction screen,
Entrained gas.
More often, conditions at the suction flange change
causing perforrnance failures to occur because:
Pumping temperature increases,
Suction vessel pressure-over-the-liquid drops, e.g..
blanket gas,
Suction level is extremely low,
Change in iiquid composition increases vapor pressrre,
Flow rate is significantly increased purposely,
System head pressure decreases causing an unexpected
flow rate increase,
Partial suction line restriction,
Gas is drawn into the suction exceeding trvo percent by
volume of gas, blocking flow2 but often causing an inlet
p.ressure drop, forcing cavitation,
Poor suction pipe layout,3
Gas-liquid multicomponent mixtures.a
Suction specific speed, Iike specific speed, is a design
parameter for pumps and is defined by the equation,
Fig. 6-Manufacturer's test loop to make a suppression test
(courtesy Peerless Pump Division, FMC Corp.).
Fig. 7-Approximate minimum limits for required NPSH,
NOTES: l. Caution should be used in applying pumps with NPSH require-
ment. Iess than shown above--
- -
2. NPSH limits apply only to the besr efficiency
3. R^equired NP-SH limits are based on a suction-specific speed of
13,000. ITpellers with lower suction specific speeds williequire
increased NPSH.
4. For double suction impellers, use
oI pump capacity.
S: suction specific speed
rotauve speed, rpm
flow, gpm
NPSHg (required) for satisfactory
operation, feet.
For double-suction pumps, use
total flow, i.e.,
About fhe
CrrARlES Jrcrsow is an engineering
in a Mechanical Technologg Sec-
tion utorking on rlechanical and methoil-
oriented, problems at the Monsa,nto
Polymers ond, Petroleurn Co., Tetas
Citg, Tenas. He has been with the com-
in Dioi,si,on
Engineering, Production, Maintenance
and, Plant Design, before setting up the
present depo,rtment in 7960 later being
appointed eng'ineering specinhst, then
senior engineering specialist. Mr. Jackson holds a B.S. degree
'i.n mechanical engineering
Tenas A*M Uni,aersitg, plus
an AAS degree in electronics technologg
College of the
Mainland, Teras C,itg, Teuas. He is o, registered prof essi,onal
eng'i,neer, a menlber of ASME, API and S,ESA.
Fig. 7 is shown as a guide for lower threshold limits
of NPSH based on suction specific speed limits. Note that
capacity and suction speed are determined at B.E.P. (Best
Efficiency Point) for the pump in question;
NPSH versus speed. NPSH will vary as the square of
the speed. This relationship was illustrated well by
Karassiks using the suction specific speed relation:
V b I
l/6 I
(h",1 h",)s/4
V@) I @, l/ e,)
Then (h"2/h"r.)s/'
And hu2f hs= (nrln,),
Where subscripts:
initial condition
final condition
The speed squargd versus NPSH relationship is useful.
f had occasion to evaluate a 1200-rpm, 28-inch-diameter,
Canadian built rubber-lined pump to the manufacturer's
U.S. maximum 750-speed ratings. The extrapolated
NPSHR indicated approximately 32 feet. The manufac-
turer agreed but our specification requested no infofirra-
tion on NPSHR. IJnfortunately, NPSHa was considerably
less than 32 f.eet. The rubber lining was eliminated and
the vortex breaker modified to correct the problem.
1. Calculate the NPSHn carefully considering all condi-
tions, i.e., startup, original fill of lines, winter operation,
all control valve pressure drops, summer operation, piping
condition, exit and entrance losses.
2. Be skeptic'al of saturated liquid-gas compositions, e.g.,
flash tanks at suction lines.
3. Provide NPSHA at 2-feet minimum over worse
conceiaed NPSHR curve by the manufacturer. (For car-
bonate solutions, consider more.6) If entrained or evolved
gas is possible, consider self-priming volute impeller de-
signs or at least open impeller construction. Note: The
effect of entrained gas is not cavitation hut it can con-
tribute to cavitation.
4. Request a suppression test if NPSH margin of "avail-
able" over "required" is less than 1 foot.
5. Concern yourself with tle conditions at the pump suc-
tion, fluid composition, vapor pressure, etc.
6. Consider the suction specific speed,
of a pump in
service application.
7. Do not use hydrocarbon corrections in actual practice.
If you get a plus, accept it with a smile, and go
on about your business.
ccqtrifueal and
D. l0ll Feb. 2a, tE70,-
Centrilugal P.mps, F. W. Dodge Corp.,
!4 r'.- ,
New developments in hot
expander compression systems
Better machining techniques and
analytical methods provide
a new
approach to compression systems. Here
is a technology update.
Hons D. Linhcrdl, Airco Cryogenics Division of Airco
Inc., Irvine, Calif.
SrNor,B and two-shaft hot gas expander compression
systems will soon replace conventional multi-train, low-
speed equipment due to significant economic and process
advantages. The single-train concept is a direct result of
the recent advances in high pressure ratio compressor,
high performance radial-inflow turbine and high perform-
ance steam turbine technology. The application and cus-
tom engineering of single train compression systems is dis-
cussed, with emphasis on performance and reliability.
Conventional compression systems, such as required for
nitric acid plants, consist of low speed multi-train systems
which have been manufactured for the past three decades.
While this approach has been constantly improved, t}le
underlying technology is unchanged. Such a system gen-
erally consists of two trains of multi-stage low pressure
ratio per stage compressors, one multi-stage axial hot gas
expander train and one steam turbine train all coupled to-
gether and installed on a large machinery platform or
mezzanine.L Over the years the capital cost of such a sys-
tem has increased from 10 percent to about 30 percent
capital plant cost. While these conservative and
heavy designs work reliably, they do not allow much pro-
cess flexibility and do contribute to decreasing plant con-
tractor's profit.
Considering current stringent environmental plant spec-
cifications, new process operating conditions must be im-
posed on the turbomachinery systems. These conditions
are increased hot gas temperature and increased over-all
compressor pressure ratio. Since most conventional equip-
ment has reached its upper limit in flexibility and per-
formance output, a new look is required at modern inno-
vative technology. Indeed, there is a wealth of significant
advancements to be considered: radial compressors with
pressure ratios as high as 12:I,'z high temperature radial
inflow turbines3 and high performance, low specific speed
axial turbines suitable for high speed steam turbine re-
quirements.a,s In addition, proven analytical methods
provide confidence in developing stable hydrodynamic
bearings and rotating assemblies. New construction mater-
ials help solve problems with highly stressed temperature
subjected turbine and compressor wheels.
The terminology 'tustom engineered" is not a new
sales slogan but is an established program of optimized
technology as demonstrated in the natural gas processing
industry.G These systems are operating reliably and com-
ply with all the process specifications.
In light of these significant achievements, it is time to
incorporate the latest technology into custom engineered
compression systems such as for nitric acid plants, large
SNG tail gas expanders and methanol tail gas power re-
covery systems. In comparison to the significant achieve-
ments of the gas turbine industry these technical advance-
nents are conservative and modest. Ilowever, they
provide a significant growth potential and process flexibil-
ity which cannot be realized from existing conservative,
conventional designs.
New developments in methods of analysis and construc-
tion allow custom engineering of major turbomachinery
rt&+ tqrl tfa
Fig. l-Miller-Larson Parameter.
auftl klT maatco
X-X-c5A cnEEP
Ff lA.(@nrPl
Fag.2-A-286 stress data as a functlon of Mlller-Larson
r /o.*
r!=si{aFl sPEeo (RPM)
ipEcrFtc aPED, tr!
Fig. 3-Efficiency of various expansion turbines vs. specific
components for gas processing and petrochemical plants.
For example, new methods of analysis allow accurate de-
termination of complex rotor dynarnics including the
effect of hydrodynamic bearings. The latter includes the
stiffness of the hydrodynamic film and the dampening
effect of the film and the mounting structure. Accurate
rotor dynamic computer programs allow detailed analysis
of multiple impeller shafts including the effect of gyro-
scopic moments due to overhung wheels.T The rotor dy-
namic programs developed within the last 10 years allow
prediction of critical speeds due to bending or hydrodyna-
mic effects within an accuracy of 2 to 3 percent. Having
Fig. 4-Titanium impeller forgings re-
sulted in the design ol a 4.7 to 1 pres-
sure ratio air compressor driven at
20,000 rpm.
this tool available and understanding the design of reli-
able, stable, high-speed hydrodynamic bearings,8 the cus-
tom engineering of single-train multiple impeller designs
becomes a reality, Besides rotor dynamic considerations,
stress analysis of complex three-dimensional turbine and
compressor wheels is required. The effort related to the
development of efficient gas turbines has provided refine-
ment of this method. For example, a stress analysis
method is available for turbines and compressors that in-
cludes the effect of thermogradients.e
With development of high strength materials, it is pos-
sible to design reliable turbine and compressor wheels with
tip speeds of 2,000 feet per second.1o The understanding
of material, stress, rupture and creep is another factor
which allows reliable and safe prediction of long duration
operation of highly stressed turbine and compressor
wheels. One of the methods which allows correlation of
stress and long duration is the "Miller-Larson" para-
meter,11 presented in Fig. 1. Pig. 2 presents the allowable
stress as a function of the "Miller-Larson" parameter for
A-286, one of the most interesting materials of consEuc-
tion for hot gas expanders.
With stress, rotor dynamics and bearing design all well
in hand, areodynamic advancements encourage the reli-
able design of single-train units. The specific speed con-
cept provides a reliable guide for selecting the optimum
parameter for the turbine and compressor designs.
example, Fig. 3 demonstrates the correlation of ma><imurn
turbine efficiency as a function of specific speed. For the
low specific speed range, axial partial admission turbines
apply while for increasing specific speed, the radial inflow
turbine and the axial turbine can be applied. The radial
turbine has an advantage over the axial turbine since it
can expand higher pressure ratios and therefore can be
constructed with the minimum number of stages. Similar
comelations have been developed and proven for axial and
radial compressors.2,'2
Considering the present state-of-the-art of petrochem-
ical process compression systems, it is obvious that modern
radial turbine and high pressure ratio compressor technol-
ogy has not been applied. This is because axial turbine
and compressor technology was developed four decades
ago while radial turbine and high pressure ratio com-
pressor technology has been developed in the last decade.
It is surprising that the radial turbine has not found its
TABLE l-lmpeller processlng
Sequrnce Decrlptlon
ol taak
1. Selectlon ol basic
Gomputer program (COMPO)
or if 2:1
Gomputer program (WHEL)
Computer program (STRES)
Computer program (CRTSP)
Computer program (THRS)
Computer program (KITPNCH)
NC 3D mill
NC contour lathe
Test lab.
Electronio dynamlc balancer
Spin pit
Assembly room, etc.
Test facility
Gomputer program (OFFD)
T785 FPM
Pinhr .
ICV'r 60o
Design llow passages
Wheel stress analysis
Shaft dynamlcs analysis
Thrust analysis
Prepare tapes for con-
tour and flow passage
Cut master tool
Turn and bore impeller
Machine flow passages
3D duplicatlng mlll
Blend end polish
Finish turn and bore NC contour lalhe
Cut spline
Blade and disc vibra-
tion test
Spln test
Assemble, rebalance
and lnstall
Test-operatlon and
Data feductlon-
place in the field of hot gas expanders while the earliest
turbines for hydraulic plants were the radial or Francis
type of turbine. The detailed aerodynamic design of high
pressure ratio, radial inflow turbines and compressors can
be accomplished with considerable confidence due to bet-
ter understanding of three-dimensional flow and the
wealth of experimental d,ata supporting detailed efficiency
predictions and correlations.
Having the aerodynamic design in hand, it is necessary
to develop new machining techniques which allow custom
machining of various wheel designs optimized for one
plant or process. Table 1 for example, summarizes the
processing steps of advanced impeller designs.
The significant achievement in manufacturing custom
engineered turbomachinery
is the application of nrmer-
ically controlled machines for the manufacturing of tur-
bine and compressor wheels. Applying advanced but prac-
tical engineering computer programs, it is unnecessary to
develop a drawing of the three-dimensional blade shape.
A computer tape is processed which can be applied di-
rectly to the three-dimensional
milling machine, which
uses the tape input to directly generate the required blade
shape. This method makes it feasible to custom engineer
optimum aerodynamic blade shapes for each application.
The tooling can be programed to produce blades with
resonant frequencies far above operating frequency. With
the above methods of analysis and the available produc-
tion tools, significant advancements have been demon-
strated in the aircraft industry and are now being applied
to the gas processing industry. The most significant
achievement in the aircraft industry has been the develop-
,6 L0sc
Fig. S-Performance ol a 4.7:1 pressure
ratio air compressor.
ment of the high-pressure ratio radial compressor. This
development was necessary for the design of a high per-
formance lightweight gas turbine. Continuous research
and refinement of analysis has resulted in the demonstra-
tion of a 12:7 pressure ratio, radial compressor which has
achieved an efficiency of 74 percent., The design tip speed
is 2,000 feet per second for 70oF ambient conditions. A
detailed stress analysis shows that by using titanium forg-
ings, reliable 100,000-hour operation can be achieved. Ap-
plying these advanced techniques to commercial applica-
tions resulted in the design of a 4.7 to I pressure ratio air
compressor driven at 20,000 rpm by a 2,000-hp electric
motor through a step-up gear box (Fig. a) . The perform-
ance of the 4.7 ratio compressor is shown in Fig. 5. Efi-
ciency is 82 percent and margin between choke and surge
is larger than 15 percent. Similarly Fig. 6 shows a high
pressure ratio compressor for starting abcraf.t and Fig. 7
shows a custom engineered 2: 1 pressure ratio methane
compressor. Having demonstrated the feasibility of design-
ing reliable high pressure ratio compressors, the next step
should be to eliminate
multi-stage axial compressors
by increasing the speed and reducing the number of
stages. High pressure ratio radial compressor technology
is one of the key elements in development of reliable
single-train compression systems. The other important
component is the high performance radial inflow turbine.
The radial inflow turbine has attained significant sta-
ture in the field of cryogenics.G,
Machines up to 10,000
hp have been custom engineered. In the field of hot-gas
turbines a 1,600-hp radial inflow gas turbine has recently
been Numerous smaller gas turbines as well
Fig. 6-700-hp high pressure ratio alr Fig. 7-2:1 pressure ratio natural gas
compressor, compressor with diffuser assembly.
Fig. 8-Steam turbine and methane com-
pressor wheel.
as turbo super-chargers have been used for some time.
ffowever, it is surprising that the radial inflow turbine has
not been applied to large power output hot-gas expand-
ers. The radial inflow turbine has a large advantage over
the axial turbine since it can operate with a larger pres-
sure ratio and therefore the average blade temperature is
much lower, assuring higher reliability and safety, The
most significant work in high temperature radial inflow
turbines demonstrated the feasibility of a 2,000oF, high
tip speed, radial inflow turbine wheel.s Applying these re-
sults to 1,300oF inlet temperature radial inflow turbine
requirements, a reliable solution results by using A-286
The third component of interest in high performance
systems is the high pressure ratio, high performance steam
turbine. The high performance steam turbine is a unit
basically developed from APU (auxiliary power unit)
technology. In addition, significant work has been done
in the field of liquid metal turbines.l6,
High pressure
ratio turbines have been developed for exhaust moisture
contents as high as 20 percent.lo Contemporary design of
reliable supersonic nozzles and selection of corrosion resist-
ant materials allows reliable and erosion-free operation.
When high pressure ratios are expanded in turbines of
relative small geometry, a high expansion velocity results
which in turn assures sub-micron condensation droplets.
These droplets follow the stream lines with minimum slip
and therefore erosion is eliminated. Fig. B shows a high
performance steam turbine operating with a 1,700 feet
per second tip speed and expanding steam from 350-psi
steam to ambient pressure. Steam turbines of this type are
machined by the electro-discharge method from a solid
forging. This method of tunbine design can obsolete the
conventional multi-stage low-speed steam turbine. Besides
reliability in steam turbines, advances in reliability are re-
quired for the accessories. These high performance tur-
bines have no problem conforming to the latest API re-
quirements. Having developed the capability of custom
engineering of advanced turbines, compressors and steam
turbines within one train, the next question is: how can
this custom engineered unit be tested prior to shipment?
The gas processing industry has developed methods for
testing compression systems prior to shipment under con-
ditions similar to anticipated operating conditions.6 Single
train compression systems can be tested economically by
using external, natural gas fired combustion chambers in
conjunction with laboratory steam power sources. The
complete system test is a requirement in order to assure
minimum problems during plant startup. This acceptance
test assures the utmost in reliability and safety. In addi-
tion the machine can be tuned to the highest performance
Turboexpanders recover energy
When large gas
flows must be
reduced trom high to low pressure
or when high-temperature process
heat) are available,
expansion turbines should be
considered. Here's how to calculate
the work recovered
V. H. Abqdie, Brown & Root, Inc., flouston
AN nxpaxsroN TURBTNE converts the energy of a gas
or vapor stream into mechanical work as it expands
through the turbine. The expansion process occurc rap-
idly, and heat transferred to or from the gas is usually
very small. Consequentln tJre internal energy of the gas
decreases as work is done, and the resultant temperature
of the gas may be quite low giving the expartder the
ability to act as a refrigerator as well as a work-producing
furboexponder lypes. Expansion turbines may be
classed into two broad categories, axial-flow and radial-
Axial-flow turbines are those in which the gas flow
is essentially parallel to the axis or shaft of the turbine.
Turbines of this type resemble a conventional steam tur-
bine and may be single-stage or multistage with impulse
or reaction blading or some combination of these designs.
Turbines of this type are not normally used for producing
low temperatures but are essentially power-recovery de-
vices and find application where flow rates, inlet tempera-
tures of total energy drops are relatively high.
Radial-flow turbines are those in which the gas flow
is essentially at right angles to the turbine shaft. Radial-
flow turbines are normally single-stage and have combina-
tion impulse-reaction blades and a rotor that resembles
a centrifugal-pump impeller. The gas is jetted
tially into the outer periphery of the rotor and flows
radially inward to the "eye" from which the gas is jetted
backward by the angle of the blades so that it flows
axially away from the rotor.
Radial-flow turbines have been developed primarily for
the production of low temperatures but may also be used
ali power-recovery drives. This type of turbine normally
operates at relatively low flows and high rotational speeds.
Power recovery. A potential application for expansion
turbines exists whenever a large flow of gas is reduced
from a high pressure to some lower pressure or when
high-temperature process strearns (waste heat) are avail-
able at moderate pressures. When such conditions exist,
an expansion turbine can be used to drive a pump, com-
pressor or electric generator, thus recovering a large por-
tion of otherwise wasted energy. In applications of this
type, careful consideration should be given to the tem-
perature drop which will occur in the expander. It may
sometimes be necessary to heat or to dry the inlet gas
to avoid low exhaust temperatures or the formalion of
Expansion turbines can be designed to cover a wide
range of operating conditions. One manufacturer lists
current limix on inlet conditions as 1,300oF at 300 psig
and 950oF at 1800 psig with no limit on minimum
temperature or on minimum or maximum horsepower.
Refrigerotion. Turboexpanders may be used in closed
refrigeration cycles with a pure gas such as nitrogen,
which is alternately compressed and expanded to provide
the required refrigeration through a heat exchanger.
Various types of open cycles may also be devised so that
the process stream to be cooled passes through the ex-
pander thus the need for low-temperature heat exchanger.
Liquid products can be produced directly from the turbo-
expander, provided it is specifically designed for this type
of service. Similar cycles can be used to recover con-
densable products from process streams. The use of open
cycles requires the removal or control of contaminants
such as water vapor and carbon dioxide to prevent foul-
ing of the equipment. This problem is avoided in closed
cycles by using a pure, dry gas as the working fluid.
To make an economic analysis, it is necessary to under-
stand the expansion process and to be able to predict the
turbine performance, specifically, the work output and
exit temperature produced by the turbine.
All real gas processes, either ideal or non-ideal, are
irreversible to some extent. Ilowever, it is necessary
to assume reversibility to illustrate an expansion process
on a thermodynamic diagram. A typical expansion is
illustrated on the pressule-enthalpy diagram, Fig. 1. In
H1 H.
Fig. l-Pressure-enthalpy diagram showing typical expansion
through a turbine,
this diagram, the initial state of the gas is assumed to be
at point a having pressure Pr, specific volume Vy tem-
perature To, enthalpy Ho, and entropy So. Point b repre-
sents the terminal point of an isentropic expansion, and
the path from a to b represents the sliccession of states
through which the gas would pass when expanding,
without heat transfer in a reversible machine. An ac-
tual irreversible expansion from point a would termi-
nate at some point such as e having the same pressure
as that at b, but with higher temperature, enthalpy
and entropy. If the expansion from point a were ex-
tended to some pressure lower than P2, terminating
at point
it would have entered the two-phase region
under the saturation curve, resulting in partial con-
densation of the gas. If the expanding gas were a
mixture rather than a pure component, there would also
be a composition change as a result of the condensation.
The relationship of heat, work and energy for an
expander may be expressed by the first law of thermo-
dynamics as follows:
A well-insulated expander with normal mass-flow ve-
locity will be very nearly adiabatig and
may be as-
sumed to equal zero, giving the fundamental expression
for work done by the gas:
This equation shows that the work done by any gas
expanding adiabatically is equal to the actual change in
enthalpy. If the expansion is also isentropic, the work
done is defined as the isentropic work W": Ho- Ht
Turbine performonce. If a Mollier chart for the ex-
panding gas is available, enthalpies may be obtained
directly from the chart; and the work output, W, canbe
calculated directly knowing only the isentropic efficiency
of the expander and the gas inlet conditions and outlet
fsentropic efficiency is defined by the following equa-
W/IA, (1)
and the work done by the gas is W
As shown above: W": Ho- H6 and therefore (2)
(H"- Ho)
Ho- He (see Fig. 1). (3)
The actual final temperature,7", can be read from the
Mollier chart at the intersection of H" and Pe.
If the expanding gas is a complex mixture for which
a Mollier chart does not exist, another approach must
be used. One possibility is to use a computer program
which will calculate the enthalpy and entropy for the
mixture to determine the isentropic enthalpy difference.
Some of the turbine vendors use special programs which
calculate actual turbine performance.
When computer calculations are not available, suf-
ficiently accurate results can be obtained by either poly-
tropic or isentropic analysis.
Polytropic onolysis. Equations developed by Schultz'
can be modified to suit expanders. To use this approach,
imagine a reversible expansion path between points a and
a in Fig. 1. Point a represents the actual condition of the
gas at the expander inlet, and point a the condition at
the outlet. This reversible expansion path is defined by
the equation
constant (4)
The work done by the gas expanding along this path
is the polytropic work.
The work W can be calculated by Equation (6) when
I4le is known. Wo is obtained by integration of Equation
(5) using Equation (4) and n
constant. The results
of this integration, when combined with the equation of
state, PV
ZRT, yield the well-known expressions for
polytropic work, one of which is the following:
The terminal temperature at point e can be calculated
To (Pr/P)* (8)
Equations (7) and (B) can be solved for the required
values of work and exit temperature using the exponents
n and m obtained from the following equations:
Zroo R
I I Ooo6oo
The compressibility functions X and I were calculated
by Schultz and plotted as functions of reduced pressure
* fi'
The actual work is Vl/
= noW,
Fig. 2-Generalized compressibility function, X.
and reduced temperature. (See Figs. 2 and 3.) These
functions may be difficult to determine accurately from
the curves in the regions of low Pn and Ta. Ilowever
Schultz has indicated that X and Y can be closely ap-
proximated by the following expressions in the regions
where Ps
< 0.9, 7E
1.5 andZ
0.18116 (8.361,t,
(6.651,t, + 0.509
Note that the values of X, Y, Z, and cp are avetages
for the expansion path. It may be necessary to break
the path into several increments, using the appropriate
average values for each increment. The calculation must
be done on a trial basis using assumed or estimated
tem,peratures to obtain X, Y, Z and cp, a,nd adjusting
them until an adequate check is achieved.
lsentropic onolysis. Using isentropic efficie-ncn it is
necessary to calculate the isentropic enthalpy drop
Hr- Ha. Equations for this calculation are similar to
those for polytropic work2 and are obtained by integrat-
ing Equation (5) along an isentropic path defined by
This integration gives
,r, _
l/ z./l L,
\t 2/ t 7)
Using Equation (12)
W" may be calculated with trial
values of. n", and the results checked by Equation (13)
until a satisfactory value of
has been obtained.
/ 0 *
Xoo) (13)
As with the polytropic calculation, the path may be
broken up into small increments if it is thought that
the fluid properties rnay vaty significantly between inlet
and exit conditions, and the pressure ratio (P2lP) in
Equation (12) becomes the ratio for the
giving incremental work Wui.
Since the values of n", Z, co, and X will not change
much in a small increment, Equation (13) can be
o 02 04 06 08 t0 t2 tt t6 t0 20 u ?4 26 ?8 !0
Fig. 3-Generalized compressibility function,
W't (l
The isentropic temperature drop is then
T6 maf also be calculated from the following equation
using m" as defined by Schultz:
To: T"(Pz/P)^" (r6)
f(Zo,oR) /
Xoo) (17)
The isentropic work is:
W,:2W,i:Ho-Ho ( 1B)
The actual work, W: Ha-He can then be obtained
using Equation (3).
The isentropic work may also be obtained by the use
qf F,quation
(18) and the following approximate ex-
7 7 8l ln(P, / Pz)
(t e)
This expression should only be used in small incre-
mental steps. The temperature and compressibility used
in Equation (19) are average values foi the increment.
Equation (14) should be used to check the assumed tem-
perature drop for the increment.
An equation for estimating the actual terminal tem-
perature may be developed by examination of the geom-
etry of Fig. 1. From this figure, it can be seen that
+ lW,
T" may, therefore, be estimated by Equation (20) and
the previously calculated value of Tt. In Equation (20),
cpaas reptesents the average specific heat between
and 7".
Exomple. Calculate the work obtainable and the exit
temperature for the expansion of propane from 400 psia,
30O:F to 100 psia using a turbine with an isentropic
efficiency of 0.75.
02 04 06 08 t0 t2 t.4 t6 r8 z0 22 24 26 2
For propane i P": 617 .4 psia, T"
666.3oR and R
Assume (7"-To): 75oR. Then:
250 psia
Tno,o: 722.5/666.3: 1.083, Pno,o
250/617.4: 0.405
From Figs. 2 and 3i Xo,o:0.+7, Yors: l.l+
For propane at 722.5oR,250 psia: cpo,s: 0.552 ko,o
1.175 and h"oro
kooo/Yoro: 1.03
From a compressibility chart, at initial conditions:
The isentropic work obtained is calculated using Equa-
tion 12:
38.9 Btu/Ib.
Substituting into Equation 13:
aao _
+ o'47
from which (7"- Tt)
Using this new temperature difference, the procedure
is repeated. Only the results of the calculation will be
shown here.
(7"- To)
10B.6oF from which To= 760.0
This temperature difference is close to the assumed
temperature difference and further refinement is not
The actual work obtainable is then,
0.75 (39.5)
The actual temperature drop (I,
7,) is calculated
using Equation (20) :
T":65t.4 +
-,75)]/0.502 =
From a Mollier chart for propane, the corresponding
figures are: \il,
38.5 Btu/ib, 7u: 19BoF, T"= 2l6op.
Breaking the expansion path into three increments each
having equal pressure ratios will give the following results
u,hich are in slightly better agreement with the Mollier
values: W":39.35 Btu/Ib., Tb: 197.24
About the
V. H. An.tnrs is on eng'ineeq'with Broun
& Root, Inc., Houston. His work in-
uolues project nxo,nager/Lerlt. Mr. Abadie
holds a B,S. d,egree
mechanical en-
The Uninersity of Te*as
at Austin and an M.S. degree in me-
chanical engi,neering
The Uni-
oet'sitg of Houston. He is a member of
the American Soci,etg of Mechoni,cal
Engineet's and the American Gas As-
Manufacturers should be given adequate data for the
design and selection of a turboexpander and to assure
the purchaser of comparable bids. Data should include
the following items:
l. Ambient conditions-barometric pressure, tempera-
2. Gas conditions
a. Inlet pressure
b. Inlet temperature
c. Outlet pressure
d. Inlet composition
e. Molecr.r-lar weight
f. Ratio of specific heats co/cu
g. Inlet weight and volume flow (include the range
of expected variations as well as the design value) .
h. A description of the amount and nature of any
gas contaminants such as liquid droplets or dust
3. Operating requirements
a. Driven equipment (type, speed, horsepower and
direction of rotation).
b. Speed control and governing requirements.
c. Shaft sealing requirements including leakage limita-
In addition to furnishing vendors adequate design
data, certain basic mechanical features should be speci-
fied for reliable operation. Such features should include
the following:
l. A heavy-duty thrust bearing (preferably of the Kings-
bury type).
2. Bearing temperature sensors.
3. Vibration sensors.
4. Low lube-oil pressure trip.
5. A highly reliable overspeed trip device.
specific heat at constant pressure, Btu/lb"F
specific heat at constant volume, Btu,/lb'F
enthalpy, Btu,/lb
polytropic temperature exponent
polytroplc volume exponent
rsentroprc temperature exponent
volume exponent
P: absolute pressure, psfa
f =
absolute pressure, psia
critical pressure, psia
P,'= mixture pseudo critical pressure, psi4
reduced pressure p/P" ot P/PJ
O =
heat transferred to or from the gas
individual gas constant
!t{g/mol-wt, ft-lb,4b'R
entropy, Btu,/Ib
7= absolute temperature,
critical temperature, "R
7"'= mixture pseudo critical temperature,
reduced temperature, T/7" or T/7.'
specific volume, cf/lb
actual work done by gas, BtMb
polytropic work, Btu/lb
iseritropic work, Btq/lb
compressibility function, Tn (62/8Tn)
compressibility function,l
compressibility factor, PV/RT
?, =
isentropic efficiency
?p =
polytropic efficiency
. =
M., "Th"
-{nalysi -of
Centrifugal Cmpresms,"
Trans. ASME, Jan., Apr., 1962, pp 69 and 222.
Abadie, V. H,, "Piedicting Expmsion Tubine Pufmmce." ASME
pa;rcr number 69-WA/PID-19, Nw. 1969.
Turbomachinery Problems
and Case Histories
Case histories
of specialized
There is no such thing as ordinary ot
routine machinery problems
but some are
more unusual than others. Here are
some case histories that illustrate
what to do if you
are laced with an
uncommon maltunction
Wolier von Nimitz, t. C. Wochel and F. R. Szenqsi,
Southwest Research Institute, San Antonio, Texas
INvnsucerroN oF sERrous vibration or failure prob-
lems in rotating equipment has many aspects of the classic
murder mystery: "Who did it and why?" The engineer
uses similar investigating techniques to find the ""
Since many problems cannot be readily diagnosed rvith
typical monitoring equipment, he must use specialized
equipment such as real time analyzers, high frequencv
accelerometers, pressure probes, strain gages, etc., to gain
additional clues.
Once the cause has been defined, the question of "u-hy"
must be ans."vered. The normal causes are usuallv related
to design, installation or operation. Some failures are due
to design deficiencies which indicate use of inadequate
mathematical models which cannot take into account a11
system variables.
Three case histories illustrate the instrumentation and
analysis techniques that can be used to detern-rine the
causes of r,,ibration and failure problems.
The first case concerns a steam turbine that had re-
peated blade failures. When blades failures occur, the1, are
normally near the root and associated with fatigue, most
probably at the blade natural frequency. The manufac-
turer usually increases the cross-section near the root to
reduce the stress concentration, lvhich is usualll-
adequate as past successes verif,v. If this method does not
work, obtaining meaningful data becomes increasinglv
difficult because measurements other than blade strain
gaging only infer
the blade is doing. To properlr,
measure the blade response requires a tremendous effort,
both in time and money. especiallv considering dou.ntime
costs in some process applications. The ideal technique
would be to make external measurernents u.hich could be
related to blade response. In this example, blade response
was monitored by measuring the bearing housing vibra-
tions with accelerometers. Using a real time spectrum
analyzer and other automatic analyzing equipment to
obtain continous spectral display versus speed produced a
version of the Campbell diagram with the amplitudes of
vibration superimposed.
The second case discusses the field balancing of a unit
which had a bearing housing support resonance. The unit
had several available balance planes and sufficient data
was taken to develop influence coefficients for four of the
planes. The primary emphasis of the discussion is cen-
tered around the inconsistencies that can be obtained
using vibration amplitude and phase data where non-
linearities appear in the support structure.
The third case describes the excitation of a compressor
shaft at its'fourth critical r,vhich r,r,as 15 percent above
running speed. The shaft was excited by an acoustical
the suction piping caused by variation of the
inlet guide vane angles.
A thret stLrse. l.lJ00-hp ste1lr tur.binc clri'"'i1.g a
tr-ifugal air cornpr.essor in a catall,tic cracliing unit had a
historl of viltratiorr
1;roblenrs. cxcessir.c shaft
vibrrrtirrns- g()\rernol gcar r.ibr-ation problcrns ancl [
oI tht- llrst.rnd third stagr bladcs ancl shroucl barrrls. A
measured horizontal response at the inboard end of the
gential modes of the first and third stage blades in the
normal running speed range (3,850-4,800 rpm).
Vibration signature data rvas obtained on the turbine
and tltird stage blacles. \\rhcrr rhe c.rnrPonert at 50 )(. N
throrrqh thc ll.1(10-+.300 TI;, rrtnse. l:rrse
componcnts are ileasured on the bearing housir.rg. \\rhen
the 100 X N and 150 X N conrponents pass thr.ough this
frequency range, they also excite these modes. This data
presentation method is significant because apparently neg_
ligible responses in the vibration signatures ,"-,
tinent inforruation.
Fig. 3 is a si:ectral analysis (0-1,000 IJz) shorvinq the
srAGE (3) oR
lsr STAGE (1)
F 3E f; E*
Fig. l-Campbell diagram.
Fig. 2-Outboard bearing housing v ib r at io n versus speed
showing nozzle excitation frequencies.
excitation of rnany blade frequencies rvhen passing
through the critical speed at 4,600 rpm. When a unit runs
on a critical speed, it can excite blades at their own
natural frequencies.l Blade natural frequency components
were also excited at speeds well below the critical speed.
Although it can be argued that the response at these fre-
quencies on the bearing housing may not indicate that the
blades are \ ata presentation technique rein-
forces the I it is blade natural frequencies
since the sa is excited by other multiples of
the blade p cy.
Another interesting phenomenon occurs near some of
the calculated blade natural frequencies; an increase in
the.natu,ral frequency is observed as a function of speed,
until the speed reaches a certain level and then the fre-
quencv remains the sarne. This increase in blade natural
frequencv could indicate the effect of centrifugal force
on changing blade end fixity.
Unequal pressure distribution around the nozzle periph-
ery caused partial adn'rission effects, as shown in Fig. 4,
resulting in larger amplitudes at the lor"e. harmorrics.
This proved to be a major source of energy for exciting
the blade vibrations.
Fig. 3-Excitation of lower blade natural f/equencies as a
function of speed.
lJpon startup of a steam turbine in a chemical plant,
difficulties were encountered in the field balancing due to
a resonance associated with the inboard bearing housing
support structure. The foll'owing data were monitored
to determine the vibration characteristics: proximity
plobes-vertical and horizontal, bearing housing-vertical
and horizontal, and shaft absolute-veritical and hori-
zontal. This preponderance of data quickly leads the engi-
neers to ask which measurements are best to use in the
balancing. Simple balancing theory indicates that any two
measurements in a plane can be used to balance a rotor,
since they will define the high spot. If the system is linear
and the rotor and its supporting structure have no reso-
nances at or near the selected balance speeds, theory
indicates that any combination of probes should give the
same balance solution. In actual practice inconsistencies
in data commonly occur. Arbitrary selection of balance
planes may not lead to a satisfactory balance condition
unless extensive trial and error techniques are pursued.
Since a resonance of the bearing housing support struc-
ture was verified by a shaker test, all of the listed data
points were carefully measured to obtain an accurate and
complete record. The bearing housing suPPort structure
was braced in an attempt to eliminate the effect of the
A detailed analysis of all balance data, including nu-
Fig. 4-Comparison of vibration for partial and full admission
steam inlet.
merous vector plots and use of a balancing digital com-
puter program developed by the authors,
'r''"'as used to
determine the required balance correction *'eights. This
computer program uses the least squares mathematical
technique similar to those of Goodman' and Lund.3 The
influence coefficient data is obtained by locating an LrIl-
balance on each plane individually and obtaining the
TABTE l-lqlqn6s corrections for combinqtions of plones
A. Single plane belauce solutiors
f .{i,r,
31 81 0.90
31 90 1.06
20 51 2.03
118 80 1.66
m I {6 2.95
17 85 0.91
Proximity vertical
Proximity horizontal
Proxim. yert.
& bor.
Shaft absolute, Yert.
Shaft absolute, hor.
Shaft abs. vert. & hor.
Prcxim. vert. & ]ror..
Shaft abs. vert. & hor. . .
Proxim. vert. & hor. .
Sbaft abs. yert.
& hor. , .
Balance Plane 4 Balance Plane 3
No. oI
I Mr*.
Angle I resid.
18 1.17
353 0.93
3 1.18
344 1.81
340 1.69
341 I 1.84
I I 0.87
327 I 1.36
4 I 1.19
Lo wes t
no. Type of data
no. Type oI data
Proxioity vertical
Proximity horizontal
Proxim vert &hor..
Shalt absolute, vert.
Shalt absolute, hor.,
Shaft abs- vert. & hor-
343 I 1.80
BalancePlanes4&2 BalaocePlanes3&2
Bal. Plane 4 Bal. Plane 2
Pl""" 3
w-,I l,gl" w.
resid Wt. Angle \Yt, Aogle
Bal. Plene 2
BP.l & 3
0 81
Lowe st
Proximity vertical
PruxiutlJ lorrzontal
Proxim. vert. & hor.
Shaft absolute, vert.
Shaft abmlute, hor.
Shaft abs. vert. & hor
B. Two plane balance solutions
C. Three plane balance solutions
Bal. Plane 4 I Bal. Plane 3
vibration data (amplitude and phase angle) near each
bearing. All possible balancing combinations are calcu-
lated based on selected test points and speeds. Optimizing
criteria are applied to determine minimum expected
vibration, minimum added weight, maximum reduction
per added weight, and minimum sum of squares vibration
at all test points. Since this leads to nume.rous solutions,
the engineer must still use his
judgment to determine
which is best. Theoretically, the best balance will be
obtained by adding weights to all available balancing
planes. Ifowever, in field balancing, keeping the balance
planes to a minimum is desirable since considerable effort
and downtime are required to change weights in all
balance planes. A "perfect" balance cannot be obtained
nor is it necessary, because only a satisfactory balance
within acceptable vibration criteria is needed. Consider-
able time can be saved if a satisfactory balance can be
obtained using only one or two balance shots and a mini-
mum of balance planes.
Table 1 consists of 40 single-plane balance solutions, 18
two-plane solutions, and 6 three-plane solutions. Using
the computer makes it possible to compare all of these
A study of the data in Table 1 shows that:
The balance solutions for the different probes are
not consistent. For example, for the single plane balance
solution, the ratio of maximum to minimum weight re-
quired was l48l17, 4819, 90124, and 712155 for balance
planes (BP) 1-4 respectively. The deviation in angle was
57,2L3,51 and 23 degrees.
If the comparison is made for only the cases where
four speeds and four probes were used (Runs 3 and 6),
the difference between solutions based on the proximity
probes and the absolute shaft vectors become closer. Sta-
tisticalln this means that more consistent data will be
obtained if more probes and speeds are used.
Max./min. weight (gms) 33/31 23113 85137 109/55
Angle difference 44o 1510 21" 15o
For the single plane balance, BP4 had the lowest
expected residual for all cases (except Runs 3 and 9)
and the smallest angle of deviation. Based on these cri-
teria, BP4 appears to be the optimum plane for a single
plane balance.
For a minimum added weight criterion BP2 is best,
based on the data in Run 3 indicating 13 gms at 2O8
degrees attains the same results as 55 gms in BP4. How-
ever, due to the inconsistent phase angle and the 5:1
ratio betr,r,een maximum to minimum weight predictions,
the solution cannot be trusted.
Solutions such as giverl in Runs 11, 12 and 13 for
BP3 and BP4 where exceedingly large weights are pre-
dicted is one problem that is common when using least
squares computer analyses. Usually, large weight solu-
tions can be traced back to nonlinear or resonant effects
in the data.
For the three-plane balance cases, considerable dif-
ferences are noted in the predicted solutions. If only the
proximity probe data is considered and the influence
coefficients determined from the data for balance planes
2, 3 and 4, then the three plane solution would be as
given in Run 19. Imagine the anguish and indecision that
an engineer would have if the computer told him to put
approximately one-half pound on BP4. When large
weight solutions are predicted, the next balance plane will
normally have a large weight diametrically opposite, i.e.,
Run 19 has 264 gms at 12 degrees for BP4 and 187 gms
at 198 degrees for BP3, or 186 degrees away.
The detailed analysis of the balance data indicated
that the best single plane balance could be obtained at
BP4. All 10 solutions are plotted in Fig. 5. Since only a
23-degree deviation was calculated, the location was
straightforward, and a weight was selected which was the
average of the predicted amplitudes. This first attempt
showed considerable improvement; and the second trim
balance further reduced residual unbalance. The trim
balance reduced the vibrational amplitudes through the
speed range from 5,700 to 7,200 rpm to less than 1 mil
peak-to-peak. A satisfactory balance of the turbine was
attained with one balance plane (plane 4) .
A further examination of Table I shows that a better
balance could be attained by using BP2 with BP4. Fig. 5
indicates the balance weight region for the second balance
plane which is centered at approximately 270 degrees, and
a weight magnitude of approximately 18-48 grams re-
quired. This was not attempted in the field due to limited
time, as startup was imminent and the single plane bal-
ance was adequate. BP2 would have been the logical
choice for the next balance shot. Ifowever. the vibrations
Fig, S-Balance weight corrections using different probes
t i
this, high inlet guide vane settings were necessary to con-
trol pressure rise.
The critical speed map for the compressor is given in
Fig. 6 and the mode shapes for the first four criticals are
given in Fig. 7. The unit runs between the third and
fourth criticals. It can be seen from the mode shape that
the anti-nodes for the fourth critical speed are near the
two impellers. =
Fig. 6-Critical speed map
were already within specified limits.
A gas turbine driven compressor unit had several re-
occurring problems, including (1) slipping of compressor
coupling on shaft, (2) bearing failures in turbine, (3) ro-
tation of diaphragm, (4) failure of inlet guide vanes and
(5) excessive noise levels. The unit ran from 13,000 to
14,200 rpm. The compressors did not operate on their
original design point due to changes in plant operating
characteristics after the units were pu.rchased. Because of
Fig. 7-Vibration mode shape
Fig. 8-Compressor shaft vibrations versus guide vane setting
Fig. 9-Compressor suction pulsation versus guide vane
A field study rvas conducted to measure the vibration,
pulsations and noise of these units. Pigs. 8 and 9 sr-rn-
marize the significant results. At certain inlet guide r-ane
settings, the shaft vibrations drastically increased and the
noise in the compressor and piping also increased. Shaft
vibrations and pulsations were recorded simultaneouslr..
Fig. 8 is the spectral anal1.sis of the shaft vibrations versus
the guide vane settings. The unit speed was held constant
and the guide vane settings were gradually changed. Each
vertical step represents about 5 seconds. At the lower
guide vane settings, the predominant frequency was at the
running speed. When the guide vane setting was increased
above 45 degrees, a 1 mil vibration component at 252
Hz (15120 cpm) suddenly appeared matching the calcu-
lated fourth critical. The amplitude is as large as the
running speed component. Suction pulsation recorded
during this same period (Fie. 9) revealed the cause of
those nonsynchronous vibrations. Very little pulsation rvas
present at the compressor running speed (220 Hz) or at
252 Hz_4It CRIT;CAL 5p569
nunrurruc snrF_b,
I It tl^,iETrrNG
252 Hz. Ilowever, when the guide vane setting was in-
creased above 45 degrees, an acoustical resonance of 110
psi peak-to-peak was excited at 252 Hz. This acoustical
resonance caused the shaft to vibrate at its fourth critical
which was above running speed. The iinpeller was at an
anti-node which helped the energy to couple into the
lateral shaft vibrations.
About the
W. voN NrMrrz is assistant directo,t' of
the Department of Applied PhEsics of
Soutlttoest Reseac"ch Institute in San
Antorio, Tera,s. He is a 7950 graduate
of the Technical Uniaersitg of Munich
utith an M.S. degree in electrical engi-
neering. Since
Southwest Re-
search Institute in 1957, Mr. Nimitz has
superaised the deuelopment o,nd opera-
tion of the SGA Compresso,t' Installation
Design Labora,tory and the deaelopment
of the SGA onalog in particular. His major
of erperi-
ence is natural gas, petrochem;ical and chemim'l plant design
and etsaluo,ti,on engineering ttith, emphasis on pulsation and
oibration control methods and techriques, design of acoustic
compressor efficiencg and noise control studies,
nxed,surenxent problems and design problems associated utith
fluid, flow
in general. He has superuised the design
studies of oaer 2,000 compressor and, pu.m,p installo,tions
more tho;n 3oo conlpanies th,t'oughout the world and has con-
ducted seueral hundred
eualuation studies and consult-
i,ng engineering sem.tices
the industry. Mr. Ninritz is the
author of numerous published papers o,nd articles, is a mem-
ber of the Americaru Society of Mechanical Engineers, and
is listed in "'American Men of Science."
J. C. WACHET, holds an M.S.M.E. degree
tlte Uniaersity ol Tenoe. He lws
been uith Southutest Research Institute
since 19ti1. His oiliaitics luoe centered,
in the
of vib,ratians, pulaotions,
dynamic simulations and acoustics, with
particular emplnsis on uibration and
ailure problems in turbomachinerg. His
experience has led to the
deuelnpmmt of impro'ueil experimental
comelating oibration sig-
natures of rotating equipment to poten-
Since the excessive vibrations were found to relate to
the high negative guide vane settings, avoiding these oper-
ating conditions eliminated the problems. The compressor
wheels in these units were changed to more properly
match plant pressure and flow requirements.
The three case histories of vibration problems illustrate
the problems that can occur when available analytical
solutions are not sufficient to accurately take into account
the system variables in the design stage. Based on data
obtained and the analysis of these problems, certain con-
clusions can be made:
(1) By measuring vibrations on the bearing housing
of a turbine, vibrating blades can be verified. Real time
spectrum analyzers and automatic data monitoring instru-
mentation used to display vibration signatures as a func-
tion of speed like a Campbell diagram make blade vibra-
tions obvious.
(2) Bearing housing support or rotor resonance near
the running speed increases the difficulty of rotor balanc-
ing, because the vibration amplitude and phase data
generally have inconsistencies which hinder the solution
for the balance weights. Nonlinear eflects in a system also
complicate the solution for the balance weights.
(3) When complete vibration data is taken, including
relative shaft vibration, bearing housing vibrations, abso-
lute shaft vibrations, plus shaft orbits, many probe coni-
binations theoretically can be used for the balancing pro-
cedure. The arbitrary choice of only a few probe positions,
as usually done, can lead to uruealistic solutions and can
result in costly trial and error balance procedures.
(4) A digital computer program was developed and
used for multiplane balancing for any combination of
balance planes, test speeds, and any number of vibration
measurement points. This program calculates all possible
balancing combinations, which the engineer must analyze
and compare to determine the best balancing procedure.
Several optimization schemes have been programed to
help eliminate the inconsistencies caused by nonlinearities
and by resonance in the rotor and support
(5) The evaluation of all possible sets of vibration
data leads to a zor,e of correction weights and angles for
particular balance planes. The engineer then must decide
on the amount of weight and angle.
(6) Nonsynchronous shaft vibrations can be excited by
acoustical resonances in the piping system. This illustrates
the desirability of performing detailed spectral analysis of
vibration and pulsation data so that the exact cause
can be established.
This paper originally omachinery Symposiu,
Teru A&M Univcnity, ber,
ifhe authom
gralclully rknow-lcdge I who helped develop ttre
Elrmcntatron lor gen
tio,l problems. He also has been respon-
the deoelopment of computer prog,r"ams uthich- are
used to solt;e lateral and, tot,sional critical speed problems,
as utell as other
problems in turboma,cfuinerg. He is a
member of Tau Beta Pi ond Pi Tau Sigma.
F. R. SzeN,tsr
Southutest Re-
search Institute in 7965 as a member
of the Department of Applied Physi.cs
uhere his utork has included both theo-
analysis of
I s and, acous-
tics. His usork experience in mechanics
includes the analgsis of lateral and tot-
sional uib,t'ation response of mechonical
sEstems, prediction of uibrational dis-
pla,cement, stress and, methods of
deteetion. He graduated
uersity of Colorad,o with an M.S,M.E. degree.
Gauses and cures
Most compressor problems
can be
with a good purchaser
specification based on lield
experience. Here are some experiences
and solutions
John J. Dwyer,
Air Products & Chemicals, fnc., Allentown, Pa.
Successrur, coMpRESSoR o,peration not only depends
on an experienced and capable maintenance team but to
a greater exte
Too often the
written by the
contractor or
e ApI. Extra-
Some compressor problems are unique to a particular
location or set of conditions. An example is the time an
operator ran headfirst, with his safety helmet, into a lube
oil pressure switch causing the compressor to trip and
shut down the entire plant. But most problems are caused
by design features that can be corrected in future installa-
By far the most important type compressor installed in
the process industry today is the centrifugal machine. It is
the product of more than 40 years of industrial bxperience
and development. Most prdblems in earlier years resulted
from poor balancing and high vibration. It was not
unusual for a repair shop to statically balance the rotor
in a lathe, The ,,n
If a coin could
was accept
Dynomic boloncing of rotors is now a routine procedure
for almost all manufacturers but there are still occasional
problems. A fer,r, years ago we had a small rotor dynam-
ically balanced, shipped to the field, assembled and started
up. We immediately lost a set of bearings and damaged
the rotor. The rotor vu'as repaired and rebalanced but the
same failure happened. We then discovered that the bal-
ancing machine was not properly calibrated.
As a result of these and similar problems we have im-
proved our compressor specifications. Today we require
all rotors to be dynamically balanced with not more than
two impellers added to the rotor between each balancing
step, and require the
machine to be calibrated
at the conclusion of the operation.
Vibration specs. We specify the amount of residual
unbalance'that we will accept. We also require proximity
probes at each radial bearing and require the surface
under the probe to be finished to the same standard as
the bearing journal.
Max,imum shaft vibration limits at
operating speed are specified and filtered vibration ampli-
tude at frequencies other than operating speed are limited
to a small percentage of over-all amplitude. In critical
services, two probes at 90 degrees are required at each
bearing so an orbit analysis may be made on an oscillo-
scope. On critical high-speed units we insist on tilting-pad
bearings to eliminate the possibility of oil whip.
Beoring fqilures. Although many problems associated
with bearings can be analyzed and prevented, it is almost
impossible to eliminate human error. Several years ago
at one of our plants a large centrifugal compressor was
turning backwards at trow speed, without lubrication, due
to a check valve leak. The bearings were removed, in-
spected and were s n reassembly one
plvoted shoe thrls backward so the
steel retainer rode ar. The machine
was started and ran satisfactorily for about a year but
subsequently the failure did occur, at night, on a weekend.
duct. By this method we have not only eliminated many
fi.eld problems but have also obtained in-terchangeability
of filier cartridges, control valves and switches which
reduces our sPare parts inventory.
torsionol problems. Most centrifugal
machine, and the first lateral critical is normally checked
on the manufacturer's test stand. Torsional problems are
prevalent with synchronous motor drivers since the instan-
iur,"ort torque outPut varies considerably during motor
acceleration with an exciting frequency starting at 720
hertz and diminishing to zero at synchronous speed. We
have tried to eliminate lateral and torsional problems by
specifying the acceptable range of lateral critical speeds
ai a percent of operating speed, and requesting that tor-
sionai nodes of resonan'ce be a given
Specify lateral critical
speeds as a
operating speed. Require
torsional nodes of resonance
he away from operating speed
away from operating speed. On new designs or critical
installations, at times, we have employed consultants to
analyze the lateral and torsional characteristics of the
complete train including the instantaneous torque iharac-
teristics of the motor.
High-speed couplings have been the subject of much
discussion over the past several years. Problems associated
with flexible coupling lock-up, where they transmit axial
forces, have resulted in many purchasers requiring the
compressor and driver vendor to design thrust bearings
to accommodate this possibiiity. In addition, minute dust
particles in the lubricant have blocked oil passages by
separating-out through centrifugal force within the cou-
pling. One widely used method of minimizing dust prob-
lems, especially on large compressors, is to install filters
(one micron or less) in the coupling oil
rfeed line. One
major vendor completely eliminates the flexible coupling
problem by using solid 'couplings. flowever, extra pre-
cautions must be exercised when using this approach to
assure near perfect hot alignment and sufficient axial
clearances on all labyrinth seals, especially in long com-
pressor trains. Another promising solution to the coupling
lock-up and dirt problem is the fleible diaphragm cou-
pling. For the present, we are specifying that compressor
thrust bearings be designed to accept some axially trans-
mitted force and require sludge drain holes in gear-type
Lobyrinth seqls. The design and fitting of interstage
and casing seals is an irn:portant part of over-all design
and an area that causes its share of problems. Interlocking
labyrinths where both the rotating and stationary mem-
bers have labyrinth teeth are superior to a single labyrinth
in reducing leakage. However, axial clearances and float
are critical. In one design it was possible to install the
stationary part backwards which resul'ted in almost line-
to-line contact of the labyrinth teeth. Since this type of
assembly error was possible, it naturally occurred, damag-
ing the balance drum on startuP. We have since modified
our specifications to require the stationary parts of inter-
locking labyrinths to be designed with an off-center key.
It is easy to improve compressor efficiency by minimiz-
ing interstage labyrinth clearances. The new compressor
will operate with minimum leakage, but when the rotor
becomes slightly unbalanced or vibrations occur from an-
other source, the seals will wear rapidly, opening the clear-
ance and allowing greater leakage with a commensurate
decrease in efficiency. To prevent this situation, we specify
that the minimum clearance shall be equal to the bearing
clearance plus rotor run-out at the seal due to its weight
plus a given amount of mass unbalance.
Mqteriqts of conslruction. The selection of materials
of construction is left almost solely to the purchaser when
corrosive gases are to be compressed because he frequently
has more experience with the gas than the machinery
manufacturer. For most gases, the casing and diaphragm
stainless impellers and aluminum labyrinths because of
corrosive fumes in the atmosphere.
When discussing materials we should not overlook the
inter and aftercooler. Several yeani ago we ran into a
series of failures on cooler tubes in air serwice. After a
flloisture sporotors. No discussion o[ centrifugal com-
until a large section of deposit breaks oflthe wheel, causes
un'balance, a sud,den rise'in vibration and an unscheduled
shutdown. We have been insisting on high-efficiency mois-
ture separators; and
moist gas process machines we
install a large low-velocity suction drum with a moisture-
separating pad immediately before ihe compressor inlet.
Reciprocoting compressors. Most reciprocating com-
pressor failures can he related to high vibration, pulsation,
lubrication and wear. Analyzing each type of failure and
placing it into a general category assists in developing a
Vibration switches. Several years ago we experienced
a number of different types of problems all resulting in a
major failure. A seized crosshead
by insufficient
A broken
several other problems were all different, they had one
factor in common; if the compressor was shut down
immediately, damage would be minor. We equipped all of
our large balanced-opposed compressors with aLieleration
type vibration switches. At first we had some dificulty
with spurious shutdowns, but the problem was eliminated
after. rve learned proper switch positioning and setting.
I4Ie had trvo problems shortlv after installing vibratioln
snitches on a 2,500-hp five-stage machine. Oncc lve lost
On moist gas
install a large low-velocity
suction drum with a moisture
separating pad
ahead of the
resonator in the line.
stalled with a tee in
is to know the puls
resonator frequencv t
To minimize pulsation problems an
checks in the field u'e specify a maxim
and discharge pulsation level. fn a
installation of pipe taps in the inlet a
all pulsation snubbers so \^r'e may easily check the actual
pulsation levels in the field with a pr.rr,rr. pick-up and
Lubricqnts. Air compressor cylinder Iubrication and com-
one problem can be the cause of another. We found that
the synthetic lubricant is an excelient paint remover.
I-ubricant that traveled along the rod entered the crank
case and the crank case paint began to soften and peel.
As a temporary rneasure we installed an extra set of rod
scraper rings to reduce the lubricant carryover into the
crank case. and monitored the percentage of synthetic
Iubricant in the crank case oil. When the peroentage of
sl,nthetic got too high we would simply drain a barrel
of the mixture and add a barrel of fresh lubricant during
operation. We non, specify ail reciprocating compressors
lvhere we might want to use synthetic cylinder lubricants,
to have special lubricator seals and a coating of epoxy
base paint inside the crank case.
Lubricators. Cylinder lubrication is another potential
problem area. Some cylinders had onlv one lubrication
point in the bore. A cylinder failure would be caused by
a failure of the pump feeding that point or a line ieak.
We partially corrected this problem on some units by
installing a feed point in the cylinder inlet pipe. Although
this does not result in an even lubricant distribution it
can prevent a major failure. The lubricator itself mar'
also give rise to a problem. If the operator neglects to
filI the lubricator, failure r,r,ill result when the resen'oir
runs dn'. This type of failure could be protected bv a
reservoir level switch, but the level switch is of no help
if there is a failure in the lubricator drive mechanism.
Another method of protection is to install pressure
srvitcires in each lubricator feed line, but on large units
rvith 20 or rlore lubricator points this rneans 20 pressure
switches, alarm Iights and high installation costs. We use
a short suction pump and pressure switch mounted in the
of the lubricator reservoir opposite the driver.
If the lubricant level runs low or the drive mechanism
fails, this one pumping unit will activate an alann.
Pocking. Breaking in a new set of high
used to be an experience capable of causing the operator
and mechanics no end of grief. One method u'as to in-
crease the pressure slorvly over many hours. If the packing
started to smoke, the pressure was reduced. What no one
r.vanted to admit \vas that iI the packing started to smoke
it rvas probabiy too late. On one particularly troublesome
installation, after instailing the packing rve removed one
of the packing case tie rods and installed a therniocouple
adjacent to the first seal ring.
compressor was grad-
ually brought up to pressure by monitoring the packing
About lhe
JonN J. Dwtun is mano,ger of ma-
ch;i,nerg engi,neering, Air Products &
Chemicals, Inc. Allentown, Pa. He 'is
cost estimntes, prepq,r-
ing put'chase speci,fi,cations and eoalua-
tions of all machinerll procured and
furnished bg the compana
appl;icati,on, Mr" Dwger holds a B.S. de-
gree in mechanical eng'ineet"ing
Newark College of Engi,neet"ing and an
M.B.A. degree
Lehigh Uni,aersitg.
He is a member of ASME the Acoustical Society of America
and, i,s a regi,ster.ed professional engineer in Pennsgluaruia.
case temperature. In a later installation, we converted
to a lubricated-filled Teflon breaker and seal-ring design
and built up pressure at a much greater rate. Our speci-
fications for high pressure packing now require cooled
packing cases and Teflon rings.
Nonlube compressors. Some operators have reported
fantastic life from what they term nonlube service, but
investigation shows that they do lubricate the packing or
have lubricant carryover from the crankcase. In other
nonlube compressors the compressed gas was wet. Even
atmospheric air has some lubricant value from the mois-
ture in the air. We have a number of reciprocating com-
pressors in nonlube service handling dry nitrogen. The
compressors are fitted with extended-distance pieces with
flingers mounted on the rod to prevent oil carryover. At
first, we found a great deal of variance in s,ervice life
between different units. We have been able to correlate
the service life with what we term a PV factor; that is,
the mean cylinder pressure multiplied by the average
piston velocity. Although this correlation is not precise,
we are able to predict with some accuracy the possibility
of satisfactory operation. To improve nonlube operation
we specify maximum rider-band loading, cylinder-bore
finish, and equip all horizontal cylinders with rod-drop
detectors to monitor ring wear in operation. In some de-
signs, we have found excessive rider-band loading caused
when gas pressure gets under the band and causes it to
act like a seal ring. We have corrected this in the field by
either drilling holes in the rider band to relieve the
pressure or by segmenting the ring.
Mqintenqnce. Not all problems associated with com-
pr,essors can be attributed strictly to design factors. Pre-
ventive and major maintenance also play an important
role in achieving high onstream factors. We have estab-
lished a centralized control system for all compressor
maintenance and a computerized preventive mainte-
nance system. Each month a deck of computer cards
indicating preventive maintenance jobs
to be performed
ar,e mailed to each plant. The majority can be accom-
plished with the compressor in operation, such as sampling
lube oil for analysis, inspecting filters and testing alarm
Major maintenance on centrifugal compressors is only
scheduled alter a thorough analysis of operating param-
eters, We have developed computer programs to check
actual compressor performance data against design values.
We also analyze shaft vibration data and look for trends
of increasing vibration or indications of bearing instabil-
ity. If either performance or vibration analysis indicates
deterioration, major maintenance is planned. Since most
of our facilities are of a size that do not warrarrt a com-
plete maintenance staff, major maintenance is handl,ed
by specialists from our Central Maintenance sta.ff. By
this method we can limit our local maintenance person-
nel to a minimum, yet provide the expertise for major
The final and perhaps most important aspect of major
compressor maintenance is that we do not open a centri-
fugal compressor on some hypothetical periodic schedule.
It is opened only when analysis of performance or vibra-
tion indicates a need. Following this approach, we have
consistently achieved an onstream factor above 98 percent.
Related EquiPment
Are couplings
the weak
in rotating
The answer is yes
Here are the facts
it the coupling
about selectrng
is improperly
applied or designed.
this critical component
Ken Krqemer, Allis-Chalmers Compressors Division
rnilwaukee, Wis.
Trro appr-rcerroN
effort involving the
signers. The user, by
can either help or hi
experience level of his operations and maintenance people
when selecting the basic coupling style.
tempted to permit using the gear coupling configuration
instead. The coupling was successfully applied, and is still
in use today.
Had the eqqipment been modified to suit the dynamic
forces possible with the original coupling design, the driven
shaft overhang would have had to be increased in diameter
to handle a potential force of 850 pounds, in addition to
the weight of the coupling itself. More important, this
n'ould be an unbalance force requiring a shaft u,eight of
17,000 pounds to insure that the unbalance force ivould
be only 10 percent of the rotor weight.
The dr
bine rvith
pling rvas
weight n'
eccentricity was great.
The redesign in this case was basically the incorporation
of a pilot fit to insure concentricity of the coupling hubs
and spacer. The user's problem was to undeistand the
caused trouble, the errors could be corrected and the cou-
ever, greater than this expense implies. This feedback, that
With both gear and disc couplings, the procedure of
taking a predesigned catalog item and applying it without
modifications to suit the adjacent componenls is being
The coupling manufacturers' engineers are becoming
more willing and able to merge with the rotating equip-
ment engineers in adjusting the couplin.q design after the
compressor and driver have been designed for optimum
performance. With this approach, consistent design success
is being achieved.
This leaves a burden on the user, because couplings are
Iess standard and have lower tolerance to damage when
handled by personnel unfamiliar with the specific design.
Due to this feedback, efforts have been made to seek out a
coupling which requires less care and experience in the
field. This factor is more often considered today and has
supported the development of the industrial flex-disc cou-
pling. An exercise in this relatively new application tech-
nique will point this out and illustrate quickly a few of the
reasons for the many recent applications of disc couplings.
Driven rofor-3-stage centrifugal compressor with a rotor
weight of 350 lbs., rated at 12,000 rpm, driven by an elec-
rric motor through a gear increaser.
Geor increoser-Double helical gear transrnitting 2,500
hp with a ratio of 6.675 to 1, with a pinion static weight
of 100 lbs., and a dynamic weight (at rated load) of
2,500 lbs.
Molor-l,800 rpm synchronous motor drive with a 6,000-
lb. rotor. Acceleration time to speed is 12 seconds at no-
load conditions.
It is immediately apparent that 12 seconds is not a very
long time. If any rotating component has an unexpected
radial residual force condition, in excess of the strength of
the materials, the initial start may be the last one. Every
design eflort must be made to eliminate the possibility of
high radial forces. Since we are talking about couplings,
let's stay with that component.
Starting with the most extreme conditions, the system
to be coupled is a compressor shaft-end with a static
weight of 175lbs. and a pinion static weight of 50 lbs. Be-
cause even at light loads the pinion bearings will see some
dynamic load, a weight, or resistance, of 100 lbs. will be
used for the pinion shaft end.
Upon reaching speed the acceptable radial unbalance
Fig. l-Gear coupling-male teeth integral with hub.
Fig. 2-Gear coupling-male teeth integral with spool.
Fig. S-Schematic of gear used in coupling applications.
force of the coupling would tle 15 percent of driven and
driver shaft resisance, or 26 and 15 lbs., respectively. This
indicates the driven shaft-end half of the coupling should
have been less than 0.1 in.-oz. and the pinion shaft-end
half 0.06 in.-oz. residual imbalance, respectively.
target for a gear coupling designer.
If we assume for the moment that the driven shaft and
the pinion have no inherent unbalance (a poor assump-
tion), the exercise consists of selecting a coupling, suitable
for the horsepower, which can be manufactured, and
which can be selectively assembled and maintained to
operate within these tolerances.
Two types of gear couplings have been applied to this
type of service. One has male teeth integral with the hub,
Fig. 1, and the other has male teeth integral with the spool,
Fig. 2. Both have a pilot incorporated into the male tooth
form to support the loose member of the coupling in a
concentric manner at speed, Fig. 3.
Couplings carrying catalog ratings of 4,500 hp and
16,000 rpm, based on a standard shaft gap, were selected.
One coupling catalog wouldn't give the speed rating until
the shaft gap was determined, a very good position.
The standard sleeve and spacer weight of the first cou-
pling is 19 lbs. The spool weight for the second type is 4
lbs. One-half of these weights is carried by the coupling
hubs mounted on each shaft end. Assuming the coupling
hubs are mounted without producing any additional eccen-
tric unbalance (another presumption) an eccentricity of
0.0001 inch on one end of the first spacer would produce
0.015 in.-oz. of unbalance. The second coupling spool piece
u,ould produce 0.0035 in.-oz. per 0.0001 inch eccentricity.
These pilots, acting at speed can permit only 0.0004 inch
eccentricity and 0.0015 inches respectively, if we presume
the coupling components are perfectly balanced. By their
catalog rating both couplings pass.
The next step is to study the operating characteristics
of the design to determine if, for instance, the coupling
will have an active pilot at operating conditions.
In the first coupling, the heat generated at the teeth
flows diflerently into the shaft than it does through the
sleeve to the surrounding air. The sleeve will heat up and
expand more than the hub. This plus centrifugal force
acting on the sleeve will cause it to grow radially as much
as 0.003 to 0.004 inches more than the hub in proportion
to the horsepower being transmitted. Such a differential
would permit an eccentricity of 0.002 inches, or 0.30 in.-oz.
at the rated speed or three times the permissible amount.
The second coupling also develops the same amount of
heat, but the hollow bored spool will accept heat in a
manner similar to the sleeves to the extent that no differ-
ential growth occurs. Therefore, only centrifugal force
should be acting on this pilot, amounting to 0.001 radial
growth permitting only 0.0005 inches eccentricity, or 0.02
in.-oz. of force.
By these numbers the first coupling could not do the
job, and just as clearly the second could. Note that I said
could do the job.
If this coupling can be expected to normally produce
0.02 in.-oz. and 0.06 is the limit we want, there isn't much
left for error in the manufacture or assembly.
The user of this coupling must understand this condi-
tion and arrange to component balance the half coupling
on the shaft it is to be mounted on, so,as to eliminate any
accumulated unbalance as a result of balance mandrel or
shaft eccentricities, or errors in factory balance of the cou-
pling itself. A balancing plane of tapped holes is needed
for that purpose. In the field, where replacement is neces-
sary without dismantling the units, field balance checks
are almost mandatory.
There is another area of evaluation-sliding friction
coefficient. This produces a resistance to the axial move-
ment necessary as rotors heat and expand. For the same
catalog size rating, both coupling types have about the
same pitch line diameter, so the unit load on the teeth
would be the same. This load and a selected friction co-
efficient result in a force which must be handled by the
thrust bearings and, in this case, one helix of the double
helix gear before the teeth will slide and the normal op-
erating position is achieved. A reaction will occur during
every thermal change the system goes through.
This is a subject of more interest for the user than it
may be for the designer, since the user must contend with
tooth surface finishes other than new or as manufactured.
This is the reason the words "a selected friction coefficient"
was used.
This characteristic of the gear coupling therefore re-
quires a serious study of the capabilities of the driven shaft
thrust bearing and the loading service factors used in the
gear design "in order to suit the coupling."
Intensive design efforts and complete design changes
have been made in recent years to gear couplings to di-
minish the sliding friction coefficient, with some success,
but again the burden is on the user to understand and
maintain ideal conditions to insure continuous production.
Closely associated with this latter problem is the matter
of lubrication required by gear couplings. A failure in this
area results in the infinite sliding friction factor-the
locked coupling.
This exercise could continue, for there is much more,
but this is sufficient to illustrate the many characteristics
users must understand to utilize gear couplings.
Taking the same conditions, let's evaluate a disc cou-
pling (Fig.4) . The same maximum permissible unbalance
values e4ist, but in this coupling there are only the static
pilot fits to consider. There are no changes due to operat-
ing temperatuie or centrifugal forces to cause operating
Fig. 4-Schematic of a typical disc coupling.
eccentricities. Almost all transient factors can be accounted
for by the designer in this selection of the components
used to fabricate the coupline, but in this case, he must
know shaft characteristics before designing the discs.
The same manufacturing errors that can cause eccen-
tricity in gear couplings can occur in disc couplings, but
these can be detected before operating the unit and the
chances of them occurring are greatly minimized.
The same assembly eccentricities can occur in the field.
but the absence of dl,namic radial residuals in the assemblv
rrnusr eemrrc
Fig. S-Thermal movements of shaft and casing.
permits more use to be made of the concentricity tolerances
The one area the user must be aware of is the tolerance
permitted for axial growth or axial growth transients
shown in Fig. 5. This can be a more serious problem with
disc couplings than with gear couplings on certain types
of units. A rather specific control is placed on the disc
deflection range, and the equipment has to be adjusted
axially to suit with more accuracy than with gear cou-
plings. Sorne units are very difficult to move once they
have been set-a gas expander, for instance. Where this is
a problem, horvever, arrangements can be made to permit
repositioning the coupling by remachining the components
for a one time per coupling fit-up, or by adding a spacer
The application of a disc coupling therefore almost
completely eliminates the every generation re-education of
coupling users in the intricacies of the design itself-once
the coupling is designed, because in order to design it, the
coupling characteristics must be determined and adjusted
to suit the connecting rotors. Furthermore, there are no
changes or wear taking place in the coupling characteristics
throughout the life of the unit.
In the design stages, the effects of concentricity, align-
ment, overhung bending moment during mis-alignment,
transient thermal grorvth and net growths are incorPo-
rated. The resonant pattern of the design is then matched
to the pattern of the connected rotors.
In the manufacturing stages it is relatively easy to bal-
ance the assembly to lorv residuals as dictated by the shaft
design, and the mechanical fits are simple turns, bores
and faces.
As with gear couplings, balancing is done on a mandrel,
so potentially some eccentricity can occur when the cou-
pling is mounted on the shafts in the field. Since the
coupling flex members are rigid radially, field balancing is
not complicated by dynamic changes to the pilot fits.
One of the main features of the disc coupling is its
known axial deflection load value. When compared to the
varying unknown sliding friction factor in the gear cou-
pling, this feature eliminates the greatest concern of the
user and to an equal extent the designers of the rotating
equipment. In the case of the previous example a specific
load can be used in sizing the gear and thrust bearings.
About the ouihor
KlN Kn-q.ulrpn 'j oined Alli.s-Chalmers'
Tool and Die Department in 1941, Dur-
ing the usar he was sta,tioned in lron in
the combat militarg police. When the
uar ended, he was ivr, the Of
Canclidate Scltool i,n Fort Benning, Ga,
After getting his d:ischarge, he attended,
the Unioersity of Wisconsin in Mihtsau-
kee and, reioined Allis-Chalmers in 1947
as a
seruicematn. In 7952 he trans-
e,rt'ed to Allis-Chalmers' Com.pressor
Engineering Department, tohere he utas 'instramental in the
deuelopment of high pressure oil seals, laigh' speed rotor de-
signs, precision coupl'ings, pioot shoe beat"ings, aibration
analEsi,s, anil lube ond seal systems. In 1960, he uas trans-
to the Compressor Department Marketing Group to
deaelop the Compressor Customet' Seruice Group, which he
superaises tod,ag.
This eliminated a big unknor'vn factor in system designs.
In this exercise, both the disc, and the gear coupling
can be applied. The diflcrence between them can be
noticed, but even though one has advantages over the
other, both will succeed as couplings.
The one overriding difference to users is the obvious
promise that the disc coupling will be less susceptible to
usage than the gear coupling, and that it lvill require less
It should also be very apparent that turbomachinen'
couplings, whether gear or disc type, should not be simpll'
picked lrom a catalog.
There will also be trvo phases of coupling applications.
One will be new applications on new rotating systems. The
second rvill be the replacement of existing couplings with
one of a new t)'pe. This latter phase poses the most prob-
lems in application.
As mentioned earlier, an intimate knorvledge of the
rotating equipment is necessarv to proper coupling design.
If an existing coupling is to be replaced rvith a nerv type,
there is good
to review the nature of the
rotating system to be coupled. Some installations are ver\'
old and some have been revised in other rvays in the field.
Such engineering revie'"vs are not easy to arrange rvith busr
equipment suppliers.
The tendency is therefore to match the obvious charac-
teristics of the existing coupling and see what happens.
With many older designs having relatively heavy and
larger diameter shafts the retrofits have been, as far as rve
can tell, ver)/ successful and trouble free. Part of this
success is due to the consideration siven to the retrofit br
cooperating engineers of the coupling manufacturer and
the rotating equipment manufacturer. A large part is due
to the dedication of the first cornpanies offering the disc
coupling r.r,herein extra efforts to insure success have been
In addition to new unit applications, an increasing
number of users of our compressors have been replacing
the original gear coupling with the disc type. All of the
new unit installations have been successful and, with one
exception, all of the retrofit units have been successful.
The one case where it r'vas not, the gear coupling charac-
teristics ivere simply copied. \{hen the driver and driven
shaft characteristics were suitably considered, and the cou-
pling modified, success was achieved.
The purpose for changing the coupling varies unit by
unit in accordance lvith the problem stated earlier. In verl'
few cases was it necessary because the gear coupling could
not be made to work. In most cases, it was because of the
training necessary to make the gear coupling work.
In one application the retrofit disc coupling has never
been installed, because after careful fitting, the gear cou-
pling is still operating after nearly t'wo years of almost
continuous duty.
presented at the 2nd Annual Texas A&M Turbomachinery Sym-
poiu6, Coilege Station, Texas, October 1973.
How reliable is
seal oil system?
Current designs tor high pressure seal
oil systems have inherent problems that
limit reliability. Here is an
explanation of these problems and what
to do about them
Briqn furner, Imperial Oil Enterprises Ltd., Sarnia,
Loss on sEAL orr- on a centrifugal compressor creates
a very hazardous condition. Without seal oil, most com-
pressors will leak gas from the seal areas. The higher the
compressor suction pressure, the greater the leakage and
the higher the risk of a fire or oI a fatality.
We became aware that the reliability of our system was
inadequate when we lost both seal oil pumps in succes-
sion twice within a two-year period. Detailed examination
of our system suggested that the source of the problem
was extrapolation from low pressure system designs. Some
modifications had been incorporated to permit the system
to operate, but a thorough design review had not been
made. To understand the situation, consider the system
proposed for API 614, Fig.321 (Fig. 1).
This systern works fine for low pressure systems. There
are hundreds of them in operation. The tank temperature
can rise to the average seal oil return temperature. IJs-
ually, however, it r.r'ill stabilize slightly below this temper-
ature because of heat loss by radiation from the tank, and
Irom the pressure controi bypass. Provided the pump can
accept oil of the corresponding viscosity, there is no prob-
As ihe discirarge eressure increases-. the energ.v aclded
bv ihe pumps increases. ;\t some jlressure, the reservoir
temperat'rre wiii rise above the seai return temperature.
Irultl:er, as tl:e pump presslrre is raised, the minimum ac-
crpiaicie cil viscosity increases.
at some Dres-
sure, tlie oil wilt become too hot for tir.e pumps.
Expediencv ciictated ihat as pressures rose, the i:ypass
r::ust l:e cooied. The new' connection is srtrown in Fig. 2,
Tiris syste*i is cperable oniy il the :'eliei vaive does not
actuate. It tirerelole ceir:taies a self-destruci looo, a possi-
operai.i-:rg mode r,-hich wiil result in :apici destruction
of rhe pru:::s. loop ,:f pump luear, pir,rgged filters,
debris anri hi;;ii tr,:mperature is
(Fig. 3),
Suppose tir;,.t Ior soille reascn ;he purnp starts to wear"
The.wear pa:rtrcies wili siart tc plr-rg4 tire filter" The
c.iscirarEe oi:esslire rrii! rise.
hea-t input
Afler a
the reiie.f vatve on the pi.,a;: rviil o;:etr.
'nl'iii rei$ce thi: cii flcr,r'iri tlh-e cooirr a:rC ihe s,1,stn1
heat rejection. It wiii alir: diirnp det'ris hi;ck lrrts tire
a-r-rd i,:-lto the
:uc;ion. E,r,;,r, :hs f lie,: r1e-
posits will reduce the flow to the seals. The reduced oil
flow to the seals automaticaily starts the spare pumP.
IIowever, the usual tlpe of pump for this service requires
a higher viscosity oil when starting under pressure than
when operating. At this point, the viscosity is too low for
an operating pump, and the oil is loaded with debris.
When the spare pump starts, almost all its capacity goes
through the relief vaive, and the heat into the reservoir
almost doubles. Consequently, the spare pump is destro,ved
as ruthlessly as the first one.
The disclosure of such a major error begged the
question "IIow many other built-in problems are there?"
The analysis revealed other unacceptable deficiencies.
This article describes our analysis of the potential prob-
lems of current designs and proposes an alternate system.
We submit that the alternate will eliminate the self-
destruct loop and will increase the over-all reliability.
fn reviewing the current design of seal oil systems, rue
decided on three objectives. The primary criterion had to
be that the seal oil supply must be reliable. Secondly, the
probability of main pump failure must be reduced. The
third requirement was that failure of one pump must not
cause failure of a second. For the purpose of the analysis,
the system shown in Fig. 2 was assumed to be a good
illustration of current practice.
Seql oil sysfem reliobility. We defined seal oil reliabil-
ity as the continuous supply of oil to the compressor seals
at a pressure above the gas pressure, Anv of the follcrv-
ing conditions could result in a seal oil supply failure:
leadequate power on ihe pump
I'ump wear
Faulty reiief valve
tr,ow oil ievel in the reservoir
Dirq' filter
Controi ioop iaiiure.
Fig. ?*Typicai i:lgh preEsure ceiirjlreBsai sea! oii sysiEi:,
Fig. 3-Self destruct loop for high pressure seal oil pumps.
Inadequate power to pump-pump wear-faulty relief
valve. Each of these items u'ill cause a fall in head tank
level and actuate the automatic startup of the auxiliary
pump. There is no problem if there is sufficient storage in
the overhead tank between the auxiliary pump start level
and the trip level. The 10 minutes called for in API 614
is adequate for either a motor or a turbine start. How-
ever, high pressure vessels are verrar costly. Vendors balk
at supplying five minutes even on low pressure sJ/stems.
Our high pressure s)rstems had six seconds storase!
Low oil level in reservoirs. Loss of pump suction due to
iow reservoir level will cause loss of seal oil. The trow oil
Xevei couid be caused by a cooler ieak, too high a level in
the head tank or a faiiure of an inner seal bushing. With
systerns designeci according to API 614, Fig. 32, the loss
of oil on a coojer ieak can be very hieh. At 2,000 psi, a
hole wiil pass aooroximateir' 14 U.S. gpm!
C'arrently, the onlv protection against loss of oil inven-
tory is a iow level aiarm. Futting on the sDare cooler
'.o,ouid be a ieason:1o1e action in the event nf a io,,r, level
ziavv:t. I{owever, a 'beiter way to l:randle cocier teairage
u,ould be to have the cooler ail cressure onlv a little
;:irove i,.,ater
rvould give the operalor iime
ic takc c ,.:i'ec'.ive rc(it.r.
ilirti" fiIters, Iiirers on icrv pressure seai si'51srns do not
;ri:r:l:.11.v foul at r:lacceptable raies. iVhen the high dif-
:+larrr cn e filt-n:r socrri{]ls, t};e oi,ei'ator
can usually v,,a1k over and swilrg ;:vei" thc irai:{*r v,,;lve.
High pressure systems are another story. Failure of the
lubrication film in the pump causes the rapid generation
of fine metal particles. In addition, the high pressure fil-
ter size is usually restricted because of cost.
Use of metallic screens instead of depth filters also
tends to increase filter plugging rates. For a given flow
rate and pressure drop, the screen requires a smaller con-
tainer than a depth filter. At high pressures, the con-
tainer cost is significant and the screen filter costs less
than the depth type. However, the debris-holding caPac-
ity of the screen is less than 10 percent of that of a
pleated paper unit.
Examination of Fig. 1 reveals another interesting fact.
In neither the low pressure case nor the high pressure
modification does starting the spare pump significantly
increase the flow to the seals if the filter is plugged. In
both cases, the only increase in differential pressure is
caused by a slight increase in safety relief valve accum-
A further problem with the high pressure modification
is the filter flow. If the spare pump starts, the pressure
drop will increase by a factor of four. To prevent col-
lapsed filters or blpassing, either the elements must be
changed at one-fourth of their collapsing pressrue, or the
collapsing pressure should be greater than the differential
between normal pressure downstream and the
lief valve setting plus accumulation.
A solution to all the problems with filters on these
pumps would be to provide two separate trains of pump:
cooler and filter. The filter differential pressure could be
used to initiate starting the standby train. Starting the
auxiliary pump would not increase the flow through a
Contr,ol system failure. Failure of the level control loop
on the head tank can cause a loss of seal oil. Opening
the bypass valve will cause the loss directly. Closing the
valve will dump seal oil into the machine, and the pumps
will lose suction. In spite of the history of other level
control loops, ou.r experience in seal oil service has been
good. However, there are bound to be some failures in
the life of a compressor. The API 614 Standard recog-
nizes this and requires a separate Ievel switch to shut
down the machine. However, it ignores the hazard of
blowing the seals. It would seem reasonable to provide a
redundant system for levei control. One way would be
to install an additional transmitter. A pressure switch set
to alarm at a fixed error between the two outputs would
warn of failure of either transmitter.
Conserve nol repoir. Iligh pressure oil pumps are more
lrrlnerable than most other mechanical equipment. To
varying degrees, dependent on type, they can be dam-
aged by the following:
Low oil viscosity (related to temperature)
Low speed
Loss of suction
Excessive differential pressure
Entrained gases
Loaded starts and stops.
Fig. 2-Typical system modified for cooled bypass.
s,r$?sra ?,i-i:ws
Viscosity control. If dirty seal oil is returned to the
rEservoir, the viscosity can be affected by contamination.
As stated earlier, the use of a cooled bypass was an
expediency step to make the oil in the ,"r".rro-i, acceptable
to the pump. It was a foolish step for two reasons. No
provision was made to ensure that the necessary cooling
oil was circulated. This would have required an orificed
flow of say 50 percent of pump capacity. Secondln it is
a most inefficient method of cooling.
We need as much of a safety factor in viscosity as rea-
sonable. There are two ways of meeting this objective.
One is to use an oil with as high a viscosity as practical.
The second is to cool the suction to as low a temperature
as practical. With a given water temperature, a suction
cooler will give a lower temperature than the high pres-
sure arrangement illustrated in Fig. 2.
Filtration. Almost all vendors recommend that the pump
suction be filtered. This creates a dilemma. Plugged suc-
tion filters would cause a seal oil failure and may destroy
the pump. Putting automatic bypasses on these filters
would render them useless at the most critical time and
probably result in a pump failure. The use of pressure
filters on the suction of the high pressure pumps would
not eliminate the problems. Ilowever, the greater accep-
table filter Ap with this approach would give the opera-
tor more time for corrective actions. It would therefore
reduce the chance of a pump suction failure due to a
dirty filter.
Speed. The lubrication film on screw pumps will fail if,
at a given viscosity and differential pressure, the speed is
reduced below a certain limit. The speed limit increases
as the viscosity decreases. It also increases as the differ-
ential pressure increases.
One way of avoiding the low speed problem is to use a
motor-driven pump, fn separate systems, with a reason-
ably sized head tank, this is possible since interruption of
the oil supply during turbine start is not a problem.
Many people do not believe a turbine is a satisfactory
driver for the standby oil pump in lubrication service.
Therefore, in combined systems where the main oil pump
is turbine-driven, speed monitoring is essential. The
higher the diflerential pressure in a given pump, the
more critical this becomes.
Loss of suction. As stated earlier, loss of suction will
quickly destroy a high pressure pump. Tripping the
pumps on loss of reservoir level offers maximum protec-
tion to the pumps, However, it may unnecessarily cause
a seal oil failure. A reasonable balance would be to risk
damage to one pump while safeguarding the other. One
way of achieving this would be to lock out the spare
pump on low oil level alarm. Another would be to switch
to the spare at low oil level alarm.
Excessive differential pressure. The use of safety relief
valves as currently specified should ofler adequate pro-
tection against excessive diflerential pressure. Ifowever,
the arrangement shown in Fig. 2 is most unsatisfactory.
If this system must be retained, the pump should be shut
down by flow or temperature switch when the relief
valve actuates. Operation of the relief valve will destroy
the pump by temperature rise or by debris. A relief flow
shutdown would almost always result in a seal oil failure
because the spare pump would also shut down on over-
Entrained gases will cause pump failure by cavitation
and by overheating. The following features are essential
to prevent failure.
All high pressure returns to the reservoir should enter
belor,v liquid level and the ends of the pipes should have
target plates. Lolv velocity lines, such as returns from the
machine, should enter at or below liquid level. Dirty seal
oil returns must be thoroughly degassed external to the
Loaded starts and stops. Starting and stopping some
types of high pressure pumps under pressure can cause
damage. This tendency is aggravated if there are tu'o
pumps, a centrifugal and the booster, in series on the
same driver. It is difficult to see lvhy this should be, al-
though it may be because of high oil r.iscosity durir-rg
One proposal to deal lvith this problem is to use pilot-
operated safety relief valves on the high pressure pump
discharge. The safety relief valves would be held open
during startup of the pump, using solenoid controlled
pilots and closed after startup.
One wreckscl-en6 lo go. If we assume that one
pump is failing, how do we protect against failure of the
other? The second pump has two minimum requirements.
The fluid must have a satisfactory viscosity and must con-
tain no debris. Clearly, the high pressure sysiem sho*'n in
Fig. 2 will not meet these criteria. If the pump is failing.
the debris will either be passing through the relief valve
or the filter element will be bypassing. Also, the heat
rejected by the cooler will have decreased. Therefore.
the oil will probably be too hot.
Also, when the second pump starts, the heat load to
the reservoir doubles. This is because all the additional
capacity must pass through the relief valves into the tank.
Even if the second pump survived the startup, it rr-ould
not live for long. This additional heat load could be
avoided by shutting down the damaged pump. This re-
quires instrumentation to ensure that the spare is operat-
ing before the other is shutdown.
The best solution to this situation is to specifl. that
the oil to high pressure pumps be cooled and filtered.
Such a requirement rvould prescribe trvo pumps in series.
A low pressure pump to raise the oil pressuie to abor-e
cooling water pressure, and the high pressure booster,
In between the pumps would be the cooler and the
suction filter. Tu,o such trains would ensure the spare
pump would not fail on starting because of high tempera-
ture or debris.
Pig. 4 is a section of Fig. 3+,35 and 36 in API 61-l.
It illustrates another possible way of destroying the second
pump. The filters follorving the coolers are considered to
be the pump suction fiiters. No other filters are required.
Hor,vever, suppose the main pump started to destroy itself.
After a while the filters u,ould plug, reducing the florv
to the system. The level switch on the overhead tank
rvould start the spare pump. The oil florv from this r,r,ould
overpolver the main pump.
If debris should lodge in the check ,,,alve
holding it
open, the good pump would circulate through the dam-
aged one. All the debris contained in the main pump
and that being generated would be dumped into the suc-
Fig. 4-Typical dual pump train high pressure seal oil system
with cooled and filtered suction and filtered discharge.
Fig. S-Proposed system to eliminate the pump self destruct
tion of the spare. This wouldn't last long. The above is
fact-not fiction. This reinforces the proposal that there
be two separate trains of equipnrent with an absolute
minimum of interconnection.
Two down. One mechanism for destroying both pumps
is low reservoir level. If the main pump can not get
sufficient oil to satisfy the seals, the oil pressure drops
and the spare is started. With high pressure pumps, it
is unlikely that either will survive.
This possibility has existed for years in the low pres-
sure systems. Why raise it now? There are two reasons.
High pressure pumps are much more susceptible to cavi-
tation damage than low pressure pumps. Also, the rate
of flow through high pressure cooler leaks is much
greater. In addition, operators were probably more read-
ily available.
The obvious answer to this is low reservoir level pump
trips. Since tripping both pumps would cause a seal oil
failure, the level alarm should be duplicated, one for
each pump. An alternate which would save one pump,
would be to start the spare pump on low seal oil level.
If the main pump were arranged to shut down, as pro-
posed earlier, when the spare is running, only one pump
could be damaged.
Whot should we do? If the preceding has seemed a
condemnation of Figs. 32 and 33, and to a smaller extent,
Fig:s. 3a, 35 and 36 of API 6t4, that is what was intended.
No one is going to change from an established system
unless they are convinced it is unsatisfactory. What then,
is a good system? We would propose the system shown
in Fig. 5 for all seal oil systems over 1,000 psig.
There should be two complete trains of pumping equip-
ment as shown, eaeh train coarsisting of:
Suction gate valve
Low pressure pump
Filter, 10 microns, 3 psi pressure drop with oil at
100 SSU, not more than 20 psi with oil at 650 SSU,
and a collapsing pressure of at least 50 psi.
High pressure pump
Filter, 10 microns, 3 psi pressure drop with oil at
100 SSU, not more thar. 20 psi with oil at 650 SSU,
and a collapsing pressure of at least 50 psi.
Check valve
Discharge gate valve.
A pilot-operated relief valve shall be provided between
the discharge of each high pressure pump and its high
pressure filter, discharging to between the low pressure
pump and the cooler on the same train.
The low pressure filter on the main pump train shall
be a duplex unit.
fn each train, the centrifugal and the high pressure
pumps shall have a common driver. The low pressure
pump shall be coupled to the driver to facilitate flushing.
The high pressure pump may be coupled to the driver.
On separate seal systems with a head tank, the main
pump shall be motor-driven. On all other systems, our
preference is that the main pump shall be turbine-driven.
The turbine-driven main pump shall be equipped to
alarm if its speed is less than a specified value. The
alarm shall cancel if the pump is off-line.
In addition to alarming, the reservoir low level alarm
shall lock out the spare pump if it is turbine-driven, and
start spare pump if the spare has an electric motor-driver.
The high pressure pump shall be started unloaded.
The pilot-operated safety valve shall be opened by sole-
noid valve approximately 10 seconds a"fter the pump is
tripped and shall close 10 seconds after the pump restarts.
The solenoid shall be actuated to cause the safety valve
to byryass.
A differential pressure of one-half the collapsing pres-
sure of either main pump filter shall cause the spare
pump to start.
If the auxiliary pump starts up for any reason, the
main pump shall shut down. The shutdown circuit shall
verify, by pressure switch at the high pressure level, that
the spare is operating before shutting down the main.
The high pressure pumps shall be screw type in a
steel casing. The capacity shall be based on the seal flow
required to give a clean seal temperature of 160oF. The
pump shall be sized for this flow plus 10 U.S. gpm or
200 percent of this flow, whichever is the greater.
There shall be a high temperature alarm on the corn-
mon discharge of the pumps.
A second pilot-operated safety valve shall be provided
on the spare train to vent the system of air. A hand-
operated vent shall be provided on the main train.
The head tank level control shall actuate the two
control valves on split range. The bypass valve will be
fully closed with a 3 psig air signal, and fully open with
a 9 psig air signal. The throttling valve will be fully
open at 9 psig and fully closed at 15 psig. The level
control shall be proportional only with no reset.
The head tank shall have two level transmitters. One
shall operate the pressure switches, the level controllers,
plus the low level shutdown. The second shall monitor
the first and shall alarm if a pressure difference of more
1 psi exists between the output from the two
Duql troins. The purpose of the dual trains is:
Prevent cross-contamination in event of pump failure
Limit filter flow to capacity of one pump
Ensure that the spare pump starts with clean filters
Cost saving on transfer valves
Ensure that spare filter cannot be plugged by oper-
ator error by starting a damaged pump.
Single filters. The only purpose of the pump discharge
filters is to retain the debris in the case of a pump failure.
Therefore, if they need cleaning, the pump needs repair-
ing. There is no justification for dual filters.
Main pump duo! filters. The main pump is in service
all the time. The suction filter will gradually become
dirty. If only a single filter were provided, only one train
would be available when it was being cleaned. To limit
this downtime, dual filters should be provided.
Single coolers. There are various means of cleaning
coolers on stream-shock chlorination, acid cleaning,
abrasive washing, etc. Therefore, there should be no
to have a spare cooler on the main train.
Pilot-operoted sofety vqlves. Standard safety valves
have a high accumulation. Pilot-operated safety valves
are much more reliable and have a much lower accumu-
lation. Further, the pilot feature can be used for startup
Single driver for pumps. With the single train system,
where the low pressure pump is essential for the high
pressure unit, it is a prerequisite that both pumps should
have one driver. The alternate is a much more compli-
cated control system.
Aboul the
Bnr,tN TnnNER is an engineering asso-
ciate in Imperial Oi.l's Engineering Di-
uision at Sarnin, Ont. His
incluCe preparation of machinery speci,-
and, consultation on machinerg
sclection, tesling, installat:ion, startup,
ntuintenance and trouble-shooti,ng. He been in th,e Engineering Diuision
16 gears u.torhing oyr. utilities, ptrocess
engineering and, macluinery. Bef ore moo-
ing to Canada, Brian utas em,pl,oyed in
a uarietg of positions in thermal pouter stations i,n England.
He was educated in England in electri,cal and mechanical en-
gi,neering and, is registered a,s a p,rofessional engineer in
Moin pump driver. With separate systems, there is
normally plenty of time to start a spare turbine-driven
pump. Two advantages of running the motor pump are:
Lower power cost
The emergency pump is maintained in first-class
condition (as new). When the steam driven pump is
normally operated the system is very vuinerable to a
power failure when this pump is being repaired.
On combined systems, the rate of fall of the lube sys-
tem pressure creates a problem. Our experience is that we
can normally get a motor-driven spare in service without
tripout. However, this is not possible with a turbine spare
without an accumulator. Our experience with accumu-
lators has been unsatisfactory.
Low reservoir profection. The intent is to ensure that
one pump survives the emergency. At the same time, the
system is protected at the possible expense of a pump.
Stortup unlooding has been found necessary on manv
applications. However, it seems expedient to pla,v it safe.
Since there is the possibility of a control malfunction
causing a seal oil loss, the solenoid on the safety valve
must be activated to unload.
Moin pump shuldown. If the spare pump starts up.
the most probable cause is a failure on the main train.
Keeping the latter in service is asking {or trouble. The
best solution is to get the suspect equipment ofl-line as
quickly as possibly.
Piston pumps. Screw pumps were selected for this ser-
vice because they are the only suitable pumps ar.ailable
with a steel case.
High temperqture
Screw pumps, like piston
pumps, have a minimum viscosity of about 70 SUS. Since
this is easily reached, the operator must be r,r.arnectr if
the oil temperature increases.
Overheqd tqnk level monitoring. The mean time be-
tween failure of level control loops is much less than the
life of the machine. The approach taken is to moniior
the main transmitter, using a second identical unit. and
a pressure switch. The switch would activate an alar-la.
With only one redundant unit, it would not be practical
to take any automatic action. Two redundant unils
would be needed before corrective action could be taken.
A number of failures have indicated that the c rrrent
design of seal oil systems for high pressure centlifr,rgal
compressors is inadequate. One major deficiencl, is the
presence of a self-destruct loop that can destroy the high
pressure pump. Other significant defects and rvcaknesses
are present. These imperfections can be reducecl at little
cost. However, the improvements will onlv con're
signs conflicting with past practice on lor,r, pressurc svs-
tems are accepted.
Abstracted lrom "Design of High Pressure Seal Oil Systems," originally
presented at the ASMD Petroleum Mechanical Engineering CoofereLrce. Los
Angcles, Sept. 16-20, 1973.
API 614-I,ubrication, Shaft-Sealing, and Control Oil Systems for Spcial
Purpose Applications, First Edition, 1973.
What to specify for
John C. Finney, Shell Oil Co., Ilouston
WrrE,N spEcrFyrNc high speed gears for critical services,
a few reliability items should be considered over and
above those included in the two accepted standards,
AGMA 421 and API 613. The API Standard is written
as a user's modification to the manufacturer's AGMA
Standards and, if followed, should result in a satisfactory
gear installation. The following items are suggested for
additional gear reliability.
Geor hordness. If you trace the history of AGMA Stan-
dards, you will find that through the years the allowable
hardness of gear teeth has been increasing with a sub-
sequent increase in the gear rating. In other words, the
average gear which was built, say in 1960, could probably
be used today for about twice the horsepower simply by
increasing its tooth hardness. Considering the problems
that have occurred over this same period, it strikes me
that upgrading gears by means of hardness alone is a
questionable practice. I would recommend that gear sizing
be based on a moderate hardness, say 350 BHN pinion,
300 BHN gear. After the gear has been sized, I would
have no objection to increasing the tooth h'ardness during
manufacture but not to increase its rating.
Service fqclors. My experience has indicated that
AGMA recommended service factors are inadequate.
Most users I have talked with share my view on this,
although rnost manufacturers still attempt to sell gears
based on AGMA factors. I notice that the most recent
AGMA Standard 421 (Sixth Edition) has finally relented
and increased its recommendations somewhat. This is cer-
tainly a step in the right direction, but I still feel they
have not gone far enough. As a rule of thumb, my recom-
mendation would be to add about 0.25 to each of the
AGMA recommended service factors.
Geqr mounting. Generally speaking, the details of the
gear mounting are provided by the supplier of the driver
or the driven equipment or possibly by the contractor.
Although I believe the average gear manufacturer has
definite ideas about mounting of gears, he seems reluctant
to make too many engineering recommendations to these
suppliers or contractors, who, of course, are his customers.
To avoid this pitfall, I recommend that the ultimate user
specify that the gear ideally be set directly on a sole plate
using no shims, thus requiring mating equipment to be
shimmed to the gear. If shims are necessary, specify that
they be of equal thickness under each mounting foot and
made of
minimum ground plate. All gears should
be doweled after the hot alignment has been verified.
Spore geqrs. Purchase of a spare set of gears (pinion
and bull gear) is highly recommended. Having these long
delivery items on hand has been a real salvation on several
occasions that I can recall. A fringe benefit which can
be derived from a spare gear for a nominal extra is having
this set at a diflerent ratio (usually higher) than the main
set. This allows some protection for inadequate comPressor
performance or for slight uprating (be sure you buy the
driver large enough to cover this higher horsepower re-
quirement) .
Geqr conslruclion. I recommend that you rule out gears
made of cast components-this is my feeling based on a
bad experience we had with a cast gear. Our failure oc-
curred after about six weeks of operation when a spoke
of the gear wheel failed in fatigue. The fatigue fracture
emanated from a small crack which, because of casting
complexity, had not been detected by magnetic particle
inspection. A catastrophic failure resulted.
Oil choking is a conditi'on caused by flooding of the gear
housing with oil, resulting in overheating and subsequent
fire. Causes are high pitch line velocity, low ratio, in-
adequate oil drainage, down mesh, and too small a gear
housing (gears should occupy only about one-third of the
volume of the box) . We had two fires in the same box,
one with the main gear and then another with the spare
set of gears. Although
gears were scored, it was
necessary that they be used another time (in desperation).
Both sets failed after several weeks additional operation;
but, fortunately, we had been able to obtain a new gear
by that time.
Our report on this failure concluded: After an investi-
gation, it was determined that a condition known as
choking had caused the fires. This condition is sensitive
to high pitch line velocity, down mesh, low ratio, small
box relative to size of gears and inadequate oil drainage'
This gear had all of the above conditions and yet had run
successfully for several years before the steam turbine
driver speed was increased about 100 rpm (still below the
design speed) . The slightly high pitch line velocity appears
to have set off the choking condition. Apparently, therbull
[ilngh-speed geep
gear begins to entrain oil from the bottom of the case and
pumps it to the top of the box" This tends to over-lubricate
the down mesh creating excess heat in addition to oil
foaming and vortexing which blocks the oil drain. The oil
level begins to rise aggravating the condition. Oil choking
occurs very rapidly and has been described as a tornado in-
side the gear box. In this case, hot vapors and oil escaping
from the vent ignited upon contact with the overheitel
gear case. The problem was overcome by the addition of
adequate drainage (four additional drains were added to
the box). With a new set of gears, the unit has since run
successfully at speed several hundred rpm higher than
Quolity control. A most important step in obtaining a
succesful gear installation is follow-up of quality control
in the factory. In addition to the usual quality control
considerations of inspection and testing, I recommend
considerable attention be given to quality control of engi-
neering. This consists of an early review of the gear design
in the manufacturer's engineering office to assure com-
pliance with the specifications and to ascertain that critical
About the
JonN C. FTNNEv is a staff eng,ineer with
the M anuf acturing E ngi,ne ering D epart-
ment, Shell Oil Co., Houston. He is i,n
charge of rotating mechanical equipment
all of Shell's U.S.A., M,t.
Finneg holds a B.S. degree in mechuni-
cal engineering
New Orleans, La, He is a member of the
API Subcommittee on Mechanical Equip-
ment and is clruirtnan of the Task Force
on API Standaril 617,"Centri.fugal Com-
General Refinerg Seroices."
items such as stress levels, tooth ioading, bearing design.
etc., cornply rvith good engineering practice.
In conclusion, let me stress a conservative approach to
gear design. On ani' proposed project, I would recom-
mend a thorough review of the various driver require-
ments concentrating on elimination of gears rvltenever
possible. \'Vhen speed increasers must be applied, specif-v
a conservative design.
on intro_d_ucrory remarks made by Mr. .f inney to a panel se:sion on
units. ASME Petroleum Mechanicai Engineering Con[erence. Lo. An-
geles, Sept. 18, 1973.
High-speed gear
fai lures-
user experrences
R. M. Dubner, Standard Oil Company of California,
San Francisco
tightening up
r service
factors than r
ble, full-
Ioad, back-to-
In some
special cases, a consultant is used during the design and
engineering stages.
Cqse histories. The types of gear problems experienced
are'best illustrated by a few actual case histories. The first
was a gear unit on a motor-driven centrifugal gas com-
pt"rro.. The gear was rated at 7,700 hp m"a*irirr-, in-
creasing the speed from 1,800 rpm to 7,900 rpm.
line velocity was approximately 16,600 feet per minute.
ar \\ras basically two steel dia-
The outer forged tooth ring
yu: of the diaphragm,s. The gear
about tlree weeks nhen in-
creased noise and high vibration were observed, and the
unit was shut dorvn. Several teeth were found to be
broken at the apex of the helix on both the pinion and
gear. The lveld where the diaphragm joins the outer rin,g
was cracked approximately two-thirds of the way u.ounJ
the circumference.
A metallurgical analysis r.vas made by an outside lab-
oratory to determine the cause of this failure. Their
analysis indicated a fatigue failure at the weld. It rvas
found that tle rveld at this joint
was not full penetration.
There was a crevice on the inner side of the diaphragn
weld. Fatigue cracks emanated from that crevice. A
SOCAL metallurgist looked at the same failure and felt
that fatigue cracking was only part of the story. There n,as
evidence that hairline cracks had been formed by hy-
drogen embrittlement which could have resulted from a
faulty welding procedure or from moisture in the welding
rod coating. It was determined that the gear did not re-
ceive the required post-weld heat treatment and that the
welding rod coating did contain some moisture. These
tr'vo factors combined to produce hvdrogen embrittle-
What to specify for
John C. Finney, Shell Oil Co., Houston
WnBw spEcrFyrNc high speed gears for critical services,
a few reliability items should be considered over and
above those included in the two accepted standards,
AGMA 421 and API 613. The API Standard is written
as a user's modification to the manufacturer's AGMA
Standards and, if followed, should result in a satisfactory
gear installation. The following items are suggested for
additional gear reliability.
Geor hordness. If you trace the history of AGMA Stan-
dards, you will find that through the years the allowable
hardness of gear teeth has been increasing with a sub-
sequent increase in the gear rating. In other words, the
average gear which was built, say in 1960, could probably
be used today for about twice the horsepower simply by
increasing its tooth hardness. Considering the problems
that have occurred over this same period, it strikes me
that upgrading gears by means of hardness alone is a
questionable practice. I would recommend that gear sizing
be based on a moderate hardness, say 350 BHN pinion,
300 BHN gear. After the gear has been sized, I would
have no objection to increasing the tooth hardness during
manufacture but not to increase its rating.
Service fqclors. My experience has indicated that
AGMA recommended service factors are inadequate.
Most users I have talked with share my view on this,
although most manufacturers still attempt to sell gears
based on AGMA factors. I notice that the most recent
AGMA Standard 421 (Sixth Edition) has finally relented
and increased its recommendations somewhat. This is cer-
tainly a step in the right direction, but I still feel they
have not gone far enough. As a rule of thumb, my recom-
mendation would be to add about 0.25 to each of the
AGMA recommended service factors.
Geqr mounting. Generally speaking, the details of the
gear mounting are provided by the supplier of the driver
or the driven equipment or possibly by the contractor.
Although I believe the average gear manufacturer has
definite ideas about mounting of gears, he seems reluctant
to make too many engineering recommendations to these
suppliers or contractors, who, of course, are his customers.
To avoid this pitfall, I recommend that the ultimate user
specify that the gear ideally be set directly on a sole plate
using no shims, thus requiring mating equipment to be
shimmed to the gear. If shims are necessary, specify that
they be of equal thickness under each mounting foot and
made of
minimum ground plate. All gears should
be doweled after the hot alignment has been verified.
Spore geqrs. Purchase of a spare set of gears (pinion
and bull gear) is highly recommended. Having these long
delivery items on hand has been a real salvation on several
occasions that I can recall. A fringe benefit which can
be derived from a spare gear for a nominal extra is having
this set at a different ratio (usually higher) than the main
set. This allows some protection for inadequate compressor
performance or for slight uprating (be sure you buy the
driver large enough to cover this higher horsepower re-
quirement) .
Geqr consfruclion. I recommend that you rule out gears
made of cast components-this is my feeling based on a
bad experience we had with a cast gear. Our failure oc-
curred after about six weeks of operation when a spoke
of the gear wheel failed in fatigue. The fatigue fracture
emanated from a small crack which, because of casting
complexity, had not been detected by magnetic particle
inspection. A catastrophic failure resulted.
Oi! choking is a conditi,on caused by flooding of the gear
housing with oil, resulting in overheating and subsequent
fire. Causes are high pitch line velocity, low ratio, in-
adequate oil drainage, down mesh, and too small a gear
housing (gears should occupy only about one-third of the
volume of the box) . We had tw,o fires in the same box,
one with the main gear and then another with the spare
set of gears. Although
gears were scored, it was
necessary that they be used another time (in desperation).
Both sets failed after several weeks additional operation;
but, fortunately, we had been able to obtain a new gear
by that time.
Our report on this failure concluded: After an investi-
gation, it was determined that a condition known as
choking had caused the fires. This condition is sensitive
to high pitch line velocity, down mesh, low ratio, small
box relative to size of gears and inadequate oil drainage.
This gear had all of the above conditions and yet had run
successfully for several years before the steam turbine
driver speed was increased about 100 rpm (still below the
design speed) . The slightly high pitch line velocity appears
to have set off the choking condition. Apparently, therbull
gear begins to entrain oil from the bottom of the case and
pumps it to *re top of the box. This tends to over-lubricate
the down mesh creating excess heat in addition to oil
foaming and vortexing which blocks the oil drain. The oil
level begins to rise aggravating the condition. Oil choking
occurs very rapidly and has been described as a tornado in-
side the gear box. In this case, hot vapors and oil escaping
from the vent ignited upon contact with the overheated
gear case. The problem was overcome by the addition of
adequate drainage (four additional drains were added to
the box). With a new set of gears, the unit has since nrn
successfully at speed several hundred rpm higher than
Gluolity conirol. A most important step in obtaining a
successful gear installation is follow-up of quality control
in the factory. In addition to the usual quality control
considerations of inspection and testing, f recommend
considerable attention be given to quality control of engi-
neering. This consists of an early review of the gear design
in the manufacturer's engineering office to assure com-
pliance with the specifications and to ascertain that critical
About the
JonN C. FrNuoy is a staff engineer uitlt
the M anuf acturi,ng E ngineering D epart-
ment, Shell Oil Co., Houston- He is in
charge of rotating mechanical equipment
all of Shell's U.S.A. Mr.
Finneg hold,s a B.S, ilegree in mechurui-
cal engineering
Tulane Uni,uet sity,
New Orleans, La. He i,s a membet of the
API Subcommittee on Mechani,cal Equi,p-
ment and is chuirman of the Task Force
on API Sta,ndard, 6 1 7, " C mtri.fugal C om-
General Refinerg Seruices."
items such as stress levels, tooth loading, bearing design.
etc., comply with good engineering practice.
In conclusion, let me stress a conservative approach to
gear desien. On an1, proposed project, I rvould recom-
mend a thorough review of the various driver require-
ments concentrating on elimination of gears ujtenever
possible. \,Vhen speed increasers must be applied, specif1,.
a conservative design.
Ba.ed.on intro_ductory remarks made by Mr, Finney to a panel.eision on
sear units. ASME Petroleum Mechanicai Engineerini Conference. Lo. An-
geles, Sept. 18, 1973. -
user expenences
R. M. Dubner, Standard Oil Company of California,
San Francisco
tightening up on gear requirements, using higher service
factors than required by AGMA. Whenever possible, full-
load, back-to-back factory test runs are specified. In some
special cases, a consultant is used during the design and
engineering stages.
Cqse hislories. The types of gear problems experienced
are best illustrated by a few actual case histories. The first
was a gear unit on a motor-driven centrifugal gas com-
pressor. The gear was rated at 7,700 hp maximum, in-
creasing the speed from 1,800 rpm to 7,900 rpm.
line velocity was approximately 16,600 feet per minute.
ar was steel dia_
The o ooth ring
of the The gear
about u,hen in-
creased noise and high vibration were obsen,ed. and the
unit was shut down. Several teeth were found to be
broken at the apex of the helix on both the pinion and
gear. The weld where the diaphragm joins the outer ring
was cracked approximately two-thirds of the way around
the circumference.
A metallurgical analysis was made by an outside lab-
oratory to determine the cause of this failure. Their
analysis indicated a fatigue failure at the weld. It u,as
found that tfre weld at thi,s joint
was not full penetration.
There was a crevice on the inner side of the diaphr.agm
weld. Fatigue cracks emanated from that crevice, A
SOCAL metallurgist looked at the same failure and felt
that fatigue cracking was only part of the story. There rvas
evidence that hairline cracks had been formed by h,u-
drogen embrittlement which could have resulted from a
faulty welding procedure or from moisture in the welding
rod coating. It was determined that the gear did not re-
ceive the required post-weld heat treatment and that the
welding rod coating did contain some moisture. These
tn'o factors combined to produce hydrogen embrittle-
ment, probably the principal cause of this fabricated gear
Other contributing factors uncovered were that the
input shaft of the gear was misaligned 0.035 inch and
the compressor had apparently been surging during opera-
tion. The controversy which developed was rvhether or
not the gear would have failed in service even if it had
been fabricated properly. The company said, no, the ven-
dor said. yes. And that question has never been resolved.
The gear was entirel1, rebuilt. The welding procedure
was changed and provisions made for post-weld heat
treating. Welding rods were kept rvarm to prevent mois-
ture absorption. The repair on this gear was successful.
It has been in continuous operation for over four vears
now rvithout giving any additional trouble. Because of
this problem, Socal has discontinued the use of fabricated-
type gear constrr-rction for high-speed gear units.
A pitch line velocity of approximately 15.000 fpm has
been set as a reasonable maximum for fabricated gears.
Above that, solid-forged gears or a solid-forged gear
pressed on to a forged shaft are required.
The next case concerns a long, continuing problem
rvhich really isn't resoived yet. This gear connects a gas
turbine to a centrifugal air compressor in an amnonia
plant. The horseporver is about 16,000, increasing the
speed from 4,860 to 5.815 rpm. This low-ratio gear is of
double-helical, forged construction. The unit was first
started in May, 1967 and was immediately noisy and
rough. The teeth rvere scored. In early
of that vear,
there more pinion scoring and an extreme-pressure
additir,e rvas blended into the lube oil. Later in
the pinion rvas inspected and the scoring seemed to be
healing slightl;,. Ifowever, a new gear box was ordered
in anticipation of a failure. After an additional examina-
tion in September, 1967, the gear box continued to run
successfully for a period of five vears.
Finally, in November, 7972, the output shaft of the gear
broke between the bearing and the coupling hub. The
new sear bor ordered in 1967 r.vas installed. A four-hour
test was run and the tooth contact pattern examined. The
pattern \vas jroor. In
1973, the gear teeth were
found scored and pitted at the pitch line. Later the vibra-
tion increased. Loacl was reduced by about 10 percent.
In Februarv, 79i3, the vibration continued to increase.
Inspection revealed some chippine of the teeth. A gear
of a different make from another refinery r'vas bolrorved.
It is still rr-rnning satisfactorill,.
SOCAL and the consultant hired to investigate the
problem believe that the gear was of marginal design and
unsatisfactorv for the service. When the plant was new,
it was operating at reducecl capacity. The gear experienced
some initiai pitting, went through a healing process and
then ran satisfactorily for apprcximately five years. The
ne\'v g-ear box started up under full load conditions and
so it did not have a partial load break-in period like the
first gear. The solution will be replacement of this gear
with one of stronser design. plate on the nelv pinion will be required to im-
prove break-in. This plating technique has worked quite
successfully in past installations. For replacement of a
critical gear like this, a thorough design analysis is needed,
r,vith some outside help. Bending and compressive loads
must be calculated and construction rnaterials reviewed
with particular considerations for long life.
About the
RTcHARD M. Dunxpn i,s staff engineer-
mechanical equipment, Standard Oi'I Co.
of California, San Franci,sco. Mq', Dub-
ner superaises the Mechanical Equiy
ment Group. After recei,uing a B,S, de'
gree in mechanica,l engineering
Uni,aersitJ of Cali,fornia at BerkeleE ha
was emploged by Tidewater Oil Co. and
then Philldps Petroleum Co. at their
Aron refinery. He
Stomdard Oi,l
Co. of Califoq"ni.a in 1968, ulorking
the Corporate Engi,neering Department. He represents
SOCAL on the API Subcommi,ttee on Mechanical Equipm,ent.
Oil inlet orifice. There have been other gear probierrrs
not caused by fundamentai design errors but are never-
theiess of interest. The first occurred on a gear connectinq
a gas turbine driver to a tandem pair of centrifugal com-
pressors in a hydrogen plant. The gear unit was rated at
24,000 hp, with approximatel'1 a 2 to 1 ratio. This unit
had all the benefits of an independent design analysis and
is running quite well. After an initiatr run-in period at
reduced speed, the unit was brought up to full speed.
Shortly thereafter, the gas turbine tripped out on high
exhaust temperature, The gear box was quite hot. Oil
was squirting out every opening.
A restriction orifice in the oil inlet line to the gear had
been omitted. The manufacturer's procedure was to deter-
mine the oil flow requirements on the test stand and then
size the orifice. The orifice ivas correctly sized but sone-
how, it was never installed. The box was overfilling with
oil. The power consumed by the gear in this overfilled
condition was many times design which caused the turbine
to trip out on high exhaust temperature. Fortr-rnately, no
damage was sustained by the gear.
The orifice was not shown on the piping drawings.
Those involved with installation and startup did not knor'v
that the orifice was required before starting up the unit
in the field. It's better to buy a gear which doesn't require
an orifice on the inlet of the oil supply line. But if it must
have an orifice, make sure it's clearly shown on drawings
and make a field check to be sure it's installed.
Breather vents. Another example happened to a gear
connecting a 12.500 horsepower gas turbine to a centrif-
ugal compressor in a reinjection plant. At startup, this
unit had high vibration and the gas turbine tripped out
on overload. The breather vents in the top of the gear box
had not been installed. These breathers were shown on
dralr''ings. The problem was cured by simply
installing the breathers. The gears have now operated
satisfactorily for l/z vears.
These last two case histories indicate that failures like
these are always possible and are not necessarily caused
by poor design. Mistakes will occur during installation
and startup but in many instances these things are easily
identified and corrected. Fundamental design problems
are much more difficult to solve and require cooperative
effort between the gear manufacturer and the user. These
fundamental problems should be avoided during the engi-
neering stage by a conservative design philosophi, and
careful review of critical details.
Broed on introductory remarks made by Mr. Dubner to a panel session
on gear units, ASME Petroleu Mechanical Engineering Conference, Los
Angeles, Sept. lB, 1973. I
of vibration
gear quality
DBsrcNens can engineer every gear. They can define
the design criteria, design loads in bending, etc., on gear
teeth. However, when the engineering design is turned
over to the shops to build a gear and somebody makes a
mistake, problems begin. They have a lead error or a helix
mismatch or the shop misses on the tooth-to-tooth spacing
or they have some sort of apex wander. AII of these
mistakes have the effect
.of ,shifting
the pinion with respect
to the bull gear. Calculations on a 7,000-hp gear indicated
a 0.004-inch wander, or axial vibration, and the gear
Axiol vibrqtion. There is a new AGMA Standard
426.01, "Specification for Measurement of Lateral Vibra-
tion on Herringbone Gear Teeth." The measurement of
lateral vibration is described in great detail and one of
the first things stated is that the lateral vibration shall be
taken perpendicular to the shaft's axis of rotation. Some
operating company engineers contend that one of the most
significant measurements is the axial vibration measure-
ment and that the axial vibration measurement on a_high-
speed, high-horsepower gear should be no more than t
mil peak-to-peak in amplitude at any given frequency.
It has also been the experience
operating company
engineers that as gear errors were corrected, to achieve
this low amplitude vibration, noise problem diminished
greatly. In other words, the pinion doesn't hammer axiall,v
against the bull gear.
How much Ioading can the pinion take due to the
separation of horsepower between the pinion and bull
gear? This power can be calculated and is rather signifi-
cant. One engineer reported seeing two mils in line rr.ith
the riding point of the pinion on the pinion bearings,
r,r,here it is supposed to ride beside the buil gear. The
recommendation of the AGMA says, "satisfactorv ampli-
tude is that plane is two mils at 5,000 rpm." If lou rvould
check the forces, some believe you would finc1 these forces
beyond the realistic capabilities of the gear or the
It has been suggested that all the vibration spectra
should be recorded on tape during tests. It is quite
significant to see two mils vibration on the pinion at bull-
gear frequency indicating a pitch line runout of the bull
gear. Sometimes this vibration is disregarded by the manu-
facturer. They want to look only at the level of pinion
vibration at pinion speed and completely neglect tooth-
passing frequency. I
AGMA views
on shop testing
and vendor
F. A. fhomo, Delaval Turbine, Inc., Trenton, N.J.
Trm AtvronrcaN Gnan MeNura,crunp,ns AssocrATroN
(AGMA) has or is taking action on new standards or
information sheets on shop testing and vendor required
data. Such action will be the base for this discussion.
Shop testing in i* entirety does not readily lend itself
to standardization. The range in the size oi gear units
and in shop facilities is the primary reason. Certain
general tests are normally performed, however, rvhich in-
clude such things as full speed and overspeed operation,
demonstration of satisfactory oil flow and drainage, gear
tooth contact patterns, backlash, clearances, etc. The load
at which these tests are performed ranges from zero to
full power. The relative value of a spin test vs. a partial
load or full load test, depends to a large extent on the
type of unit, intended service, and the character-istics of
the system into which the gear will be placed. Obviously,
the more elaborate the test, the more costly.
The two primary aspects of shop testing that have
rapidly been gaining interest are: vibration and noise.
I will describe AGMA's recent action in these areas.
Vibrotion. For years, test engineers have been measur-
ing the vibration of equipment operating under part,
full, and no-load conditions; in the lab or in the field;
and isolated or in the middle of a system. This led to
the necessity of establishing what might be considered
good, bad or tolerable vibration levels.
The absence of a generally accepted standard has led
many user companies to write their own specifications,
with the result that a number of differing requirements
can be specified to the gear builder, adding unnecessary
cost and confusion. Many of these specifications have a
stair-step limit on allowable (broad band) amplitude in
terms of shaft rpm, reflecting the over-simplified concept
that a significant vibration occurs only at a frequency
corresponding to shaft rpm. Frequently, the specifications
also use words to the eflect that "balance shall be such
that the vibration shall not exceed . . .," thus ignor-
ing the many other sources of vibration. Rarely does
an existing specification recognize that a gear unit has
at least two rotational speeds. The gear wheel speed is
often as low as 900, 1,200 or 1,800 rpm, while in contrast,
the pinion speed is 4 to 8 times as much. Obviously,
the permissible vibration is not the same even though the
parts share a common housing.
fn our efforts to develop an AGMA standard, the
committee literally pooled its collective background data
on fi.eld performance, factory tests, and technical skills to
produce a realistic and workable approach.
Notwithstanding the importance of definition, instru-
ment sensitivity and calibration, a key portion of the
standard is Fig. 1 which specifies acceptable vibration
limits. The committee's object was to develop a "damage
avoidance" curve with a built-in margin. For several
obvious reasons, it is desirable to express such a curve
in terrns of displacement, velocity, acceleration, or at least
some mathematical expression. As it turned out, no single
parameter or expression would be satisfactory.
The "damage avoidance" curve could be best approxi-
mated by a combination of : (a) constant displacement
at low frequencies; (b) constant velocity at intermediate
freq.uencies, and (c) constant acceleration at the higher
However, in view of the fact that the American Pe-
troleum Institute had
recently revised Standard 612
"Special Purpose Steam Turbines for Refinery Service"
and had adopted the expression (12,000/tpm)L/2 as a
peak-to-peak amplitude limit, the AGMA committee felt
it desirable to use this expression
vibration frequency in Hz] instead of constant velocity
for intermediate frequencies so as not to be in confict
with the new API limit.
The fnrits of this committee's efforts are: AGMA
Standard +26.01 "Specifi.cation for Measurement of Lat-
eral Vibration on High Speed Helical and Herringbone
Gear Ijnits."
Noise. As those of you who have worked with noise ap-
preciate, it's not a simple subject. For example, in speci-
fications dealing with most other subjects, the euors as-
vmATI{rI ffiEouErffi $r4
Fig. l-Acceptable gear unit vibration limits.
sociated with the environmental conditions or instru-
mentation acc:':racy are usually small enough to be dis-
regarded in all but the most precise laboratory problems.
In acoustical measurements, however, many things
combine to alter this condition and severely complicate
the measurement problem. The (dB) scale of measure-
ment is logarithmic which expands the apparent influ-
ence of lowet sound pressure levels. This is so because the
ear hears, or perceives, differences logarithmically and
instruments have been designed to facilitate objective/
subj ective considerations.
Since all airborne acoustical measurements are without
contact between the instrument and the source of sound,
the environmental influence should be considered.
fn acoustical measurements the environmental influence
becomes critical as the compression waves generated by
the sound source interact with any and all surfaces and
obstruction. The resulting energ'y loss or reflection can
change the measured sound pressure level many orders
of magnitude. It is altogether possible to get changes of
3,000 percent by varying the environment alone. Such a
variation expressed in acoustical notation would be over
14 dB. To assist in defining the acoustical environment
location ("field"), many terms have been developed. The
measurement position may be indicated as being in the
"near fieldr" "direct field" or
field." fn each
of these generalized fields diflerent conditions exist, which
must be considered to obtain reasonable measurement
In addition, sources do not radiate sound equally in
all directions. This results in large variations caused by
such things as source geometry, radiation efficiency,
maximum energy areas and others. Variations due to
source directionality are usually in the order of
5 to
20 dB.
The instrumentation used is poor at best, when one
considers typical laboratory accuracies in other disciplines.
A Class II sound level meter, which is the recognized
basis for compliance with U.S. federal noise level stan-
dards has, at best, an accuracy of
3 dB. This means
that it reads the airborne pressure changes to a tolerance
of plus tr00 percent, minus 50 percent. Class I, or pre-
cision sound level, meters are good to approximately
i.5 dB.
To complicate the problem further, certain other con-
ditions arise in, for example, reverberant spaces, The
creation of standing waves can change the Ievel at aty
given spot in the area 15 to 20 dB as compared to other
iocations. This variation is frequency related which means
that the level for one narrow frequency band may be
plus 10 dB over the adjacent frequencies at a given loca-
tion while the levei for a different naryow band may be
i0 dB below the adjacent frequencies.
Despite these complications, the AGMA Acoustical
Technologv Committee has recently prepared a proposed
standard called "Sound for Enclosed Helical, Herringbone
and Spiral Bevel Gear Drives." Its purpose is to present
the instrumentation and procedure to be used for sound
measurements. It also presents typical maximum sound
levels (A-rveighted sound pressure levels).
This particular standard is limited to gears operating
at a maximum pitch line velocity of 5,000 fpm and a
rnaximum speed of 3,600 rpm.
The standard is presently being balloted by AGMA
member companies and hence is not available at this time.
The future plans of the AGMA Acoustical Technology
Committee include:
1. Similar standards for other products as the need arises.
2. Consideration of parallel activities such as:
A. Educational program on gear noise for AGMA
members andf or gear users.
(a) Fundan-rentals of measurement
(b) Sources of gear noise
(c) X,{ethods Ior quieting
B. tr,tethods for predicting gear noise from design and
manufacturing data.
C. Recommendations on acoustical enclosures for gear
D. Handbook on gear noise.
To discuss data requirements presupposes effective
communication. Unfortunately, the industry,s eflorts at
communication haven't been that good.
AGMA's High Speed Gear Committee is currently en-
gaged in the preparation of an information sheet which,
hopefully, will help improve the industry's ability to com-
municate, particularly in those areas that affect the
capacity and reliability of gear units.
The following discussion has been borrowed from the
committee's working draft.
The increasing demands for "mechanical reliability"
can best be satisfied by a coordinated technical exchange
between system designer, user, equipment builder and
e.recting engineers. This coordinated effort is properli,
called "system engineering," and is normally performed
by the design agent or his technical representative. Ob-
viously, the investigation of potential problems during
the design stage is preferable to a search for solutions
after the equipment is in operation.
The follorving points are intended to highlight some
of the subjects for consideration arrd, hopefully, to es-
tablish better communication betlveen "system engineers"
and gear manufacturers.
The term "overload" will be used to describe a condi-
tion that overstresses component parts, and may lead to
failure. Overload can be of momentary duration, periodic.
quasi-steady state, or vibratory in nature. Depending on
its magnitude and the number of stress cycles accumr-r-
lated at overload, it can be a fatigue or a yield stress
consideration. Overload on a gear unit can result fronr
internal or external causes. Internal causes of ove;load
as poor acciracy, faulty assembly, bumps or
bruises on gear teeth-are usually found by routine ir.r-
spections before the unit is put into service. Externai
sources of overload result from the operational charac-
teristics of the svsten-r into which the gear is placed, and
are more complex and difficult to identify. Some commor..
sources are:
l. Moximum conlinuous power. If a gear unit has
been sized for the nominal design point load, overload
is inevitable. The pump or compressor designer can
predict the requirements at the design point u'itl.r
fairly good accuracy. Arriving at the maximum con-
tinuous pou,er is another story. It has to be estin'rated
and is a combination of :
A. Changes in specific gravity or density of the n'redia
beir.rg pumped,
B. Carry out,
C. Overspeed,
D. Variations in pressure ratio across a conrpressor
due to abnormal operating conditions.
Specific gravity or density. Changes in specific gravirr.
of the fluid medium handled by a oump, or chanqes
in the densitv of the gas handled by a compressor. affec,
the horsepo\ver transmitted in direct proportion. Cir
boiler feed pumps, for example, this occurrence can be
encountered during startup, upon malfunction of pre-
heatine equipment, or during boiler cooldorvn follorvius
tube failure. The overload can be appreciable-in tire
order of 12-15 percent under average conditions. as
high as 35-40 percent in some instances.
Design horsepou,er for air handling centrifugal com-
pressors is usually based on the normal maximum anr-
bient, frequently about 1000 F. On a rvinter dav, u.hen
the temperature drops to 30o F, the density of the air
varies with the absolute temperature, i.e.,560/,190
In still colder weather, say
F, the density increases
to 1.27 times that of design conditions. Compressors
handling other gases are usually encountereci in process
systems under greater control, where temperature vari-
ations are less. Holvever, other variables may become
serious. In refinery practice, for example, the cornposi-
tion of the gas car, vary widely, and in other process
work the inlet pressure may not be a fixed value.
Carry Out is an expression used by the pump and
compressor industries to indicate performance on a head
curve beyond the so-called design point. Fig. 2 illustrates
a typical compressor percentage performance curve.
Note that at 100 percent speed, as the head drops off
and the flow is increased, the horsepower increases to
levels as high as 115 percent load. Carry out is an
everyday reality. It comes about through such things as
improper estimation of system performance during design
stages, altered system requirements of*existing processes,
gradual deterioration of processes, systems using half
size of multiple units where shutdown or failures of one
increases the requirements on the remaining units, or
through leaks or failures.
Overspeed is just
what the names implies, and is ob-
viously limited to applications with variable speed prime
movers. Because the power absorption of the driven
machine varies approximately with the third power of
the speed, overspeed is a large contributor to overload.
Referring again to Fig. 2, the performance curve in-
dicates that at 110 percent speed and 100 percent flow,
the horsepower is increased to 125 percent. Carry out
at this speed can increase the horsepower still further,
to levels approaching 140 percent of design horsepower.
Normal practice for a turbine-driven centrifugal com-
pressor is to set the overspeed trips at 115 percent design
speed. Governor settings are generally established to
permit continuous operation between 105 percent and
110 percent of design speed. Keep in mind the fact
that operators can and do reset governors to avail them-
selves of maximum output of the system, regardless of
original settings.
2. Nqturql torsionql vibrqtion. Practically every me-
chanical drive system in operation is oscillating in one
or more of its natural modes of torsional vibration. The
amplitudes and corresponding torques vary widely-being
a function of the proximity of forcing frequencies, amount
of excitation, anC damping. If the frequency of an excit-
ing force is close to, or coincident with, a natural fre-
quency-resonance exists and the results may be disas-
trous. If the frequencies of exciting forces are sufficiently
far from natural frequencies, violent vibration is avoided
the system will still vibrate in its natural modes at
low amplitudes, excited by random torque pulsations.
3. Forced torsional vibrotion. Forced torsional pulsa-
tions result from blade passage, eccentricity of impeller
scrolls, surge (both steady state and random), pipe strain
and foundation movement as they affect housing and
rotor displacement with the resulting loss of pressure bal-
ance, coupling misalignment, gear runout, apex wobble,
motor air gap eccentricity, engine firing rate, etc.
Recent testing on pumps has indicated torque pulsa-
tions between 10 percent and 40 percent mean torque.
Compressors with vaned diffusers generally run about
3-10 percent. Where vaneless, or volute diffusers with
large tongue gaps are used, the torque impulses disappear
for all practical purposes.
The torsional vibration spectmm of most any mechani-
cal drive system will show torque pulsations from most
all of the above-mentioned sources simultaneously. Torque
Fig. 2-A typical centrifugal compressor performance curve'
0 10 & m O S0 B0 7A o0
Fig. 3-A "real time plot" of a telem'etered strain gage slgnal
measuring torque.
pulsations are illustrated in Fig. 3. The figure is a "teal
time plo?' of a telemetered strain gage signal measuring
torque. Of the seven more prominent spikes in this spec-
trum, three are natural frequencies and four are forced
vibration. The total vibratory torque approaches the sum-
mation of the major discreet frequencies.
In the absence of specific knowledge of a vibratory
system, average vibratory torque summations for typical
equipment and applications can be estimated from the
For motor or turbine driven geared sptems-5-l0 per-
cent mean torque
For well treated engine geared systems (soft couplings
and dampers)-15-30 percent mean torque
For stiff engine geared sptems-50 percent or higher.
4. Gommunicotion. If system operational problems are
to be avoided, effective communication must be estab-
lished between the "systems engineer" and the machinery
builders (includes not only gear and coupling suppliers,
but prime mover and driven machine as well). The
various system ahalyses, in at least preliminary form,
should precede detailed equipment purchase specifications.
I .rto
This sequence will permit the design to be based on a
more nearly correct load and operating conditions.
In the past, system engineers and industry in general,
have relied heavily on the so-called service factors as
recommended by gear builders to provide the necessary
margin to accommodate overload. Gear builders had been
maneuvered into the position of furnishing this kind of
information through self-protection (gears Le frequently
the safety valve, or weak link in the chain) . When a gear
unit failed, a bigger one was provided until it worked.
The difference in size became the service factor.
The more complex, higher power and higher speed,
mechanical drive systems being designed today make this
procedure and the reliance on "seryice factor" obsolete.
Gear simply do not have the technical
depth nor the detailed information to adequately analyze
system overload. This function must be performed by
Fig. l-1rr;"al mechanical drive system torque components.
specialists under the responsibility of the systems engineer.
There is no set format for communicating these data.
The required information is the magnitude of overload
and a description of the operating conditions under which
it occurs, such as when, how long, and (i.e., steady
state, vibratory, etc.).
Fig. 4 is a graphic illustration of the combined eflects
of maximum torque plus vibratory torque as compared
to the nominal design point torque for a centrifugal com-
pressor operating at overspeed and carry out.
5. Misolig,nment while operating produces overloads-
not in the sense that increased torque does, rather in the
sense of increased stress on one or more component parts
for a given torque. There are two general (and inter-
related) kinds of misalignment in geared systems. The
first concerns shaft end misalignment across the couplings
or both, high and low speed. The second con-
cerns misalignment of pinion and gear axes-brought
about primarily by foundation or bedplate twisting.
About the
Fnoomrc A. THoMA is manager, gear en-, Delaoal Turbine, Inc., Tren-
ton, N.J. He holils a B.M.E. degree
Georgia Institute of Technologg. He hns
utorked, as a transmtission designer (ma-
chine tools) with both Gi.dding a Lewis
and Warner & Swaseg. Mr. Thoma hns
been with Delaaal ooer 20
o,s a
designer of marine and high-speed in-
dustrial gears. He has held positions as
and chief engineer before He
is curt'entlg chairtnon of e,r.s
Association's High Speed Gear Committee and the Marine
Gear Comrrti,ttee. He is also chuirmon of the API Marutfac-
turers Subcommi,ttee on High Speed Gears.
Neither of these misalignments can be readily calcu-
Iated, but they warrant discussion so that the "svstems
engineer" can take precautions to minimize their effect.
"Flexible" couplings, whether of the dental tooth. spring
element, flexing disc, or elastomeric type, produce forces
and moments on their supporting shafts when operating
misaligned. The analytical determination of the ma_eni-
tude of these forces and moments is not fullv understood.
It can be generalized that:
A. The sense and direction are such to try and bring
the supporting shafts in line.
B. Bending moments have been measured in supporting
shafts that have exceeded 25 percent of the torsional
moments transmitted.
C. That the magnitude of the forces and moments in-
creases with larger angularity across the coupling.
D. Notrvithstanding catalog claims for angular capacitv,
flexible couplings should not be looked upon as unir.ersal
they should be given the best possible alignment.
There are several general rules which will help the de-
signer obtain better alignment of the coupled shafting
Make as comprehensive an assessment as possible of
operating alignment. This is a system study and must in-
clude all elements of the system including bedplates and/
or foundations. An accurate evaluation of thermai
for all components from a valid and common reference
line is required.
displacement within bearinqs.
though generally smaller in magnitude, should be con-
sidered, particularly as it affects cold or static alignment
checks. The forces and moments imposed in pumps, com-
pressors and turbines by their inlet and discharge piping
are major factors in deflecting this equipment and causing
operating mal-alignment. All efforts should be made to
minimize piping effects. After determining the probable
magnitude of alignment change from static and cold, to
dynamic and hot (including any periodic cyclic changes
that may occur), select a coupling arrangement that pro-
vides enough length or span between flexible elements to
keep the angularity low-in the region of 6 to B minutes
or lower.
Reinforced concrete foundations with grouted-in sole-
plates are gene.ra
fabricated steel bed-
plutes in ter-ms of
s. A concrete founda-
iion of adequate
soil or on sufficient
piling, is the best d unequal settling or
twisting from other sources.
Fabricated steel bedplates make convenient shipping
and handling frames, but are generally designed for
strength, not rigidity. They are frequently designed so
that the various piping andf ot oil sumps cause unsym-
metrical thermal expansion. Out-of-door installations on
steel bedplates are particularly subject to cyclic bowing
caused by the daily rise and fall of the sun.
When steel bedplates are used, the designer should en-
deavor to achieve two things: First, he should arrange oil
sumps, piping and weather protection to minimize
unsymmetrical thermal expansion. Second, he should
thoroughly investigate elastic deformation of the bedplate
caused by piping forces and moments-then design the
bedplate to eliminate twisting at the gear suPPorts'
Last, but certainly not least, the "system designer"
should issue complete and comprehensive installation in-
structions covering, as a minimum, such things as:
1. Soleplate, bedplate, machinery position and leveling
2. Foundation bolting and grouting detail,
3. Cold alignment data-including method of measur-
ing, relative position and sequence of alignmen!
4. Keying, pinning, and torqueing detail as required,
5. Pipe support and flange makeup details.
6. All other relevant details that would otherwise
left to the
judgement of the
site mechanic.
traced back to poor metallurgy such as dirty steel, irn.
proper heat treatment, poor microstructure, segregation,
etc., which in the final analysis is not directly a result
of design but rather poor quality control of the material
procurement and heat treating
Heot ond lubricqtion. Since the category "heat and
lubrication" accounts for 50 percent of gearbox failures,
it rr,ay be interesting to consider this problem in a little
more detail. It is also a problem which could be solved
by shop testing.
The heat loss of a gear mesh can be calculated from:
(1) geometry of the gear teeth, and (2) the steady-state
load being transmitted. Mesh loss is a prime source of
heat and must be taken into consideration when the
lubrication requirements are specified.
In high-speed gears, there are other important sources
of heat generation which may lead to gear mesh difli-
culties and must be considered carefully. Examples are:
1. Windage and churiring losses within the gear casing,
2. Spray patterns and oil distribution, within the casing,
3. Improper drainage of oil from the gear casing,
4. Improper use of baffies and gear shrouds,
5. Loss of backlash caused by heating of gear rotors,
6. Excessive oil flow into'the gear unit, and
7. Poor surface finish on contacting surfaces'
It may be interesting to see how heat affects a pair of
meshing gears. Fig. 1 shows what happens to high-speed
pinions when they increase in temperature operating un-
der load. The lubricant carries away some of the heat but
the remainder is carried away by conductionl and the
net result is that the center of the pinion may operate
at a higher temperature than the ends of the pinion. The
Testing can reduce
E. E. Shipley, Mechanical Technology, Inc.,
Latham, N.Y.
Trrs cnen vENDoR knows from past practice horv his
gear units should perform in the field from an internal
heat and deflection point of view. BLrt relying wholly on
past performance is not
justified since many costly gear
failures do occur. It appears that there is room for shop
testing-the type of testing that trulv brings out the weak-
nesses of a given design and its associated hardware. The
more test effort applied at the factory, and the closer the
factory test is to actual field operations, the less trouble
should be experienced in the field. Full load torque
testing at design speed (back-to-back) in the factory
comes the nearest to duplicating field practice.
It high reliabilitv is the first order of preference, then
nrore test work of the full load, full speed veriety may
be the answer.
Cquses of geor fqilures. There are four general causes
oI failure that are often identified without question as
the prime reason for failure. The approximate percent-
ages of thcse failures, from actual field investigations
covering a five-year period, are:
50 percent heat and lubrication ofterr in thc form of
contact str-ess su|face pitting
15 perccnt wear-metal removal f rorn the sutface more
or less ur"riformly
10 percent bending stress-gear tooth breakage
10 perccnt metallurgical.
The "metallurgical" category has becn included to indi-
r:ate that sometimes the prime cause of failure can be
bottom pinion shows the exaggerated growth experienced
when the pinion is cooled or heated nonuniformly. This
condition is called "barrelling" and is a rather common
Although Fig. I shows a double helical gear as an ex-
ample, this conditiou also applies to a single helical gear
as well.
It is quite easy to understand how barrelling can con-
centrate load near the apex or middle of a double helical
gear. Load concentration tends to increase the tempera-
ture even more. It is not uncommon at all to find the
apex of a double helical gear scoring locally.
Flg. l.-Pinion, barreling is seen as exaggerated growth when
the pinlon ls heated or cooled nonuniformly.
As the pinion or gear, for that matter, increases in
temperature, the tooth also changes shape. Fig. 2 shows
a tooth cross-section before and after heating up to a
higher temperature level.
in :Ti:
to temperature.
2. The top dotted curyes show an expanded pinion
tooth profile after an increase in operating temperature.
3. The pinion grows more radially at the outside di-
ameter than at the base diameterl this growth has about
the same eflect as changing the base pitch of the involute
gear teeth.
ce in ear
tooth It
from or
roaded gears in action.
engin for
5. An expanded pinion will cause heavy root bearing
with its mating gear profile. It wiil also have a tendency
to mesh deeper into the root section of the gear.
6. The expanded pinion may also help to initiate an
early scoring type failure because of the tendency for
high load concentrations in the area where sliding velocitv
is the highest.
Because of the tendency of the smaller gear member
(pinion) to run hotter than larger gear members (gear).
it is customary to make some initial adjustment during
the design stage to help counteract this phenomena.
highest temperature of the meshing pair. If, on the other
hand, the installation requires a gear reduction, that is,
the pinion driving the gear in a step down mode, the
pinion should be driving the gear with the apex of the
pinion leading.
Most gear units for the refining and petrochemical
industries are purchased to API standard 613. This stan-
dard states that all gear units shall be given a mechanical
of this nature.
Torque tesl-slow roll. All gear units should be given
a no-load spin test as a minimum. However, there are
other types of gear tests that reveal more about the po-
tential field performance of a given gear unit. Additional
tests must be evaluated in terms of the reliability required
by a given customer for a given application. Some thought
must also be given to delivery schedules; that is, if you
desire more testing, then more time is required. A slow
= ows oe nno us
Fig. 2-Tooth cross-section before and after
higher temperature.
heating to a
roll torque test will produce additional performance rn-
One of the important considerations of any gear unit
is to assure that the gear structure is strong enough to
support the load. In parallel axis gearing, such as is used
on double helical gearing, the main objective is to keep
the shafts parallel. If the gear casing deflects, it must
deflect uniformly. The torque test at slow rol1, as shown
in Fig. 3, permits you to study the contract pattern of
the gear teeth at any desired load condition. Very often,
just ohserving the contact at full load is not sumcient
since you rnust know rvhat the contact rvas like at 50
percent torque to adequately evaluate the 100 percent
torqLle conditions.
The slorv roll test as shown in Fig. 3, consists of a
production test gear unit. The input and output shafts
have bcen connected to the rotor shafts of h1'draulic
r.otary actuators. Each actuator will make onl1' a per-
ccntage of a full revolution. The stator portion of the
actuaior is connected in each case to a fixed base pedes-
1'he gear is loaded in the follorving manner: high-
pressure oil is fed into the actuator on the output shaft
rvhilc at thc sarnc time, high-pressurc oil is fed into the
actuator on the input shaIt. In effect, the trvo shafts
are rotatcd against each other under the desired torque
conclitions bv carefully adjusting oil pressures. At this
tinle, the gcar rrheels are not moving.
The tooth surfaces, on one member, are painted with
prussian blue and the other member is painted with red
lead (this is a conmon method of checking contact
Now tliat the input ancl output shaft lias beerr torqued
agair-rst each other, at 50 percent torque for example,
the contact can be rolled out simply by either decreasing
the output torque s1ightl1' by reducing oil pressure on
the output actuator or bf increasing the input torque
slightly by incleasing oil pressure on the input actuator'
After thc gear mesh has progressed through several
sets of teeth, the actllators are stopped and a 50
tooth contact has been rviped out on the teeth. The con-
tact pattern can be lifted by tape and placed in a book
for record purposes. The contact pattern for any desired
Ioad can be determined in this manner. By close ob-
ser\rance of these contact
you can determine
if the gcar casing is moving under pressurc or
just how
the load is deflccting the teeth into a load carrying
A test of this nature is very beneficial in determining
rvhat can be expected from the gear unit at full torque
provided, of course, that the usual full speed no-load
test has been run. The major '"veak
point in tests at slow
roll is the fact that the gear units does not evah-rate the
high s1;eed distortion of the gear wheels, if an1
and the
temperature characteristics of the assembled gear unit.
As stated previously,.the heat generated by thc rotating
gears call cause considerable distortion to the rncsh and
cause unever-r loading across the teeth' Nonuniform me-
chanical distortion due to centrifugal force acts in a
sirnilar manner as thermal distortion and must bc con-
sidered uhen applopriate.
Torque lesting bqck-to'bqck. To full load test high-
horsepower gear units, the technique of back-to-back
Fig. 4-Back-to-back torque test using two production gear
speed increasers.
testing is used. This is basically a locked torque test set
up simulating full power operation. It is by far the most
involved test procedure and it also produces the best
over-all test results, since it simulates field conditions
almost 100 percent. The duration of the test must be
limited but it is usually run a considerable length of
Fig. 4 shows a locked torque test set up of two pro-
duction gear speed , increasers, using any type of prime
mover as a driver. The object of such a test is to verify
that production gear units will meet the design specifica-
tions, at full speed, with loads comparable to those they
will be subjected to in field operation or even higher.
The test is conducted by mounting two identical drive
gear units adjacent to each other, back-to-back and con-
necting the nligh-speed shafts together. The torque in the
gear mesh is achieved by reverse loading the low-speed
coupling with large lever-type torque wrenches (see Fig.
5). By attaching a force measuring device to the lever
arrr,, a reading can be obtained indicating the exact
torque in the system. With the desired torque achieved,
the high-speed couplings are bolted in position thereby
locking the torque in the gear meshes similar in nature
to a wound-up spring.
Since the prime mover oftentimes does not have suffi-
cient torque to start the loaded system from zero speed,
a special starting-motor-gearbox arrangement is neces-
sary. This consists of a low-horsepower motor driving
through a high-reduction speed reducer clutched to the
test unit. When the test stand is running at slow speed,
the driver is started and the starting-motor-gearbox
clutch is disengaged. The torqued units are then acceler-
ated by the prime mover until the desired test speed is
obtained (the gear units are now running at an equivalent
horsepower rating).
Pre-test. Prior to testing the units, all parts are inspected
for workmanship to insure that correct assembly pro-
r{oi-of,irN CLAMP6
Fig. 5-. Applying. torque to the low-speed coupling with large
rever-rype wrencnes.
cedures have been followed. This final check is performed
in addition to the normal quality control inspection pro-
vided throughout manufacturing.
After manufacturing, the gearing is located and ro-
tated in the housing manually to note proper gear-tooth
Each housing is flushed with hot oil before final as-
sembly, and then checked for cleanliness.
Test stand set up. Gear units to be tested are mounted
on a special foundation so that they may be adjusted to
the proper height of the prime mover output shaft. Lub-
rication is supplied from lube consoles having sufficient
capacity to supply oil to the units at design specifications
(Note: both the proper oil flow at the proper inlet temp-
Thermocouples are installed at bearing positions and
temperature gages are placed to read the inlet and sump
oil temperatures. The oil flow rate is measured at the in-
Iet with an oil flow gage.
Load test operation. The entire system is started using
the high-torque starting motor reducer. As the gear train
is broken loose and begins to turn, the friction drops ofi,
the prime mover is started and the clutch disengaged.
The test gear units are then accelerated by the driver to
full test speed. Speed is attained over a period of about
two hours in discreet steps. During this acceleration
period, the lube system pressures, temperature and flow,
as well as bearing temperatures, are carefully monitored
and adjustments made to assure proper operation of the
When a full-load torque test has been specified, the
duration of the test must be detailed along with many of
the other variables. As a minimum, the load level on the
gear units should be run and inspected at l14, 2l+,314
and 414 torque. Suficient running time should be per-
mitted so that the running contact at these load levels
can be observed and evaluated. Once full torque opera-
tion has been approved, the test operation should con-
tinue until the slowest speed element in the gear unit has
accumulated 10 x 106 fatigue cycles.
Acceptobility limiis. If you torque test a gear unit or
simply spin test it, here are some recommended limits oj
acceptability. Some of these limits are overall appraisal
limits pertaining to design and arrangement while others
pertain to acceptance after testing.
In addition to meeting the general specification vibra-
tion and noise limits, a close examination of the gear
tooth contact must be made.
1. There should be no evidence of scoring.
2. If the gear units are subjected to high-speed runs
(light torque-load) the acceptable contact pattern after
test should be 80 percent contact across each helix rvith
no obvious heavy end bearings at either the apex or the
base. The only exception would be on a high L/D ratio
design where the helix angle corrections were machined
into the teeth to compensate for twist and deflection
under full load.
3. If the gear unit is subjected to a high-speed, fu1I-
torque-load run, the acceptable contact pattern after test
would be 95 percent contact across each helix with no
obvious heavy end bearings at either the apex or the base.
4. The total heat rise across the gear unit based on oil
flow and temperature rise must be within the vendor's es-
timated total heat loss. It also should be evaluated against
other gear units of this nature as a double check on the
gear vendor's calculation.
5. Backlash-Gear rotors are designed with backlash
to compensate for: center-distance tolerances, deflection
under load, thermal expansion of rotors and casing, as
well as variations in tooth thickness due to manufacturing.
The backlash is usually checked by mouirting the gears
at the proper center-distance, fixing one gear, and rock-
ing the other back and forth against a dial indicator ar-
ranged tangentially at the pitch circle. For helical or dou-
ble helical gearing, the backlash is measured normal to
the teeth rvith either an indicator or feeler gages. The
backlash can also be measured by holding the gear sta-
tionary and measuring the total axial movement of the
If meshing teeth interlere with one another, the temp-
erature inside the casing will increase rapidly creating
more tooth interference which will quickly cause exces-
sive damage.
A gear must have enough backlash to operate satisfac-
torily under transient and abnormal operating conditions.
Some conditions that may require more than the mini-
mum backlash recommended by the gear vendor are as
If more oil is permitted to enter the gearbox at the job
site than on the test floor,
If the horsepower load is built up rapidly on the gear
teeth before the gear casing has an opportunity to heat
If the oil tends to foam as the bulk temperature of the
oil is increased,
If a partial restriction of the drain causes a disturbance
to the internal oil distribution,
If the gear elements tend to barrel in the middle be-
cause of excessive heat buildup in the center,
If a higher viscosity oil is substituted in the field for th:
normal test gearbox lubricating oil,
If a back pressure builds up in the drain system. The
drain system at the
job site may be much longer and
more complicated than on the test floor at the factory,
6. Gears and pinions'
About the
E. E. SHrPr,Eu is manager-mecho'nical transmissions, Me- his present position. He is a member of AGMA'
ASLD, ASME and,'is a, registereil professional engineer in
The following recommendations may be helpful:
Specify a false bottom on the gear unit so that the oil
falls from the turbulent area into a quiet sub-reservoir
and then drains away.
Specify an extension to the bottom of the gear casing so
that the oil level is at least two feet from the bottom of
the large diameter gear to escaPe windage effects. A
shield should'be designed over the drain so that windage
cannot directly hit it.
Check the design for ample room at the top part of the
gear casing so that oil has a chance to hit the end walls
and flow down relatively unrestricted'
In gear units that appear to be critical from a heating
point of view, specify an upward meshing gear- This ar-
rangement has a tendency to form windage paths that
help to scour the walls of the casing in the direction of
the oil flow down the walls and into the sump.
The gear unit must have a vent and the drain line {rom
the gear unit must also be vented. Often it is desirable
to specify that two drain connections be built into the
box. Another precaution is to specify that the gear unit
drain line to the main reservoir be independent of all
other drains.
9. Oil nozzle-The location of the oil nozzle has been
a controversial item over a long period of time. Ilowever,
the following table can be used as a guide:
PitchJine velocity
Berow I 0,000 ft. / mh... . . . . It
as a sufficient
g side is more
10,000_15,000 rt.,/rnin.
squarely into the mesh.
15,000-30,000 ft./min. .....Oil nozzles on the outloing side
directed into the outgoing
mesh, the oil quantity should
be B5/o for cooling. In addi-
tion, oil nozzles on the ingoing
side directed slightly off in-
going mesh, the oil quantity
should be lSVo 6or lubrication.
Abstracted from "High speed geu testing and evalution," origi-mhy
n.esented at the Thirt Compressor Train Reliability Symposium, Manrr-
i".turins Chemists Asseiation, Chicago, April 24, 1973. I
Recommended design limits
Pitch line velocity
Under 15,000 f.t./min., .
Method of manufacture
s1 gttt
,Fabrication OK (forged rims
welded in webs)
Forgings with integral shaft OK
Disk forging pressed on shaft OK
Shrunk on rim not
unless stress is very Iow and
vendor calculates combined
Between 15,000 ft.,/min. Forgings with integral shaft OK
and 30,000 ft./min.. .Disk forging pressed on shaft OK
Fabrication OK if gear vendor
has had previous experience
and calculates deflection and
weld stresses
Shrunk on rims not permissible
Casting generally not permissible
unless considerable successful
experience can be demon-
strated by the r.endor.
7. Gear casine design-The gear casin.o; shoulcl bc stifl
enough to support the gears and bearing so that their
shafts remain parallel. The casing also must act as a con-
tainer so that oil is not flung around the area.
Many of the difficulties rvith internal oil flow in gear
units results from the fact that not enough space has
been left to adequately drain the oil away. High pitch-
line velocity gears are particularly subjected to this diffi-
Below 10,000 ft./min. (Windage is generally not strong
enough to prevent adequate drainage, even in a tight
fitted gear unit).
10,000-15,000 ft./min. (Windage effects can be felt,
hor'vever, drainage can be accomplished with little diffi-
15,000-20,000 ft./min. (Windage is often r, problem,
particularly with gear units that were designed for lorv
speed and have been uprated with little, if any, consider-
ation for internal lubrication flow).
20.000-30,000 ft./min. (A gear unit must be designecl
rvith the rvindage cffect in mind so that proper pror-i-
sions are made for adequate removal of all cooling and
lubricating oil).
B. Drain considerations for high pitch-line velocity
gears (20,000-30,000 ft./min.). The drain must be pro-
tectcd from the turbulence of the windage from the mesh
or arranged so that the turbulence helps to lorce the oil
in the proper direction for drainage.
Shipping and installation
fllqrk W. Becrd, Fluor Engineers & Constructors, Inc.,
Los Angeles
Srrrpprrve rNSTRUcrroNs should be considered as two
separate problems. First is "Preparration for Shipment."
Should it be shipped assembled or knocked down? What
sort of shock loading is likely to be encountered? How
can the equipment be protected from the elements? Sec-
ond, how should it be crated or boxed for the common
carrier's requirements? W&rile the "manufacturer stan-
dard" is often adequate, it is foolhardy to rely on his best
guess of your field storage and installation environmental
Covered storage is rarely possible and coastal environ-
ments are extremely hostile during long storage periods
which often exceed 12 months. Personalln I favor preser-
vatives which are compatible with the lube oil to be used
and do not require disassembly of the gear before placing
the unit in service. I also insist on shipping the unit fully
assembled to minimize field installation problems.
Exporl boxing ond shipping poses special problems,
depending upon the type of vessel. We have used sea-land
containers, converted LSTs and ocean going barges as
well as conventional freighters. Equipment is usually
stowed in cargo holds but frequently it is lashed on decks
and is subjected to ocean spray. Companies specializing
in export boxing are available in m,ost major ports and
are recommended.
After the gear has arrived in the field, a routine in-
spection procedure should be established to protect the
equipment from atmospheric breathing, dust, and me-
chanical damage. All temporary covers and screens should
be carefully tagged to be sure that they are removed be-
fore startup. Our only gear failure occurred when tem-
porary screens in the oil supply lines adjacent to the bear-
ings were inadvertently left in place when the unit was
placed in service.
lnstqllotion. There are several excellent gear installation
manuals available from the manufacturers. They cover
detailed instructions on :
Mounting of couplings
Foundation and alignment
Cold and hot check alignment
Axial positioning
Tooth contact pattern checks
In addition, the manuals contain excellent details con-
[iln9ffi-se[!)@@d g]@aLe
cerning maintenance, trouble shooting guides and over-
haul. They are well illustrated and include pictures of
damaged bearings and teeth to guide the neoplyte.
Startup checks. Installation manuals discuss the routine
checkouts required immediately prior to startup and are
very useful to the startup engineer. These incluC.e as a
1. Check the oil level and type in reservoir.
2. Check the direction of rotation since many
3. All shafts must turn freely.
4. The engineer should verify that cold shaft alignnent.
have provided for thermal growth of all equipment.
5. Foundation bolts must be tight and casing pr-operh,
shimmed to prevent casing distortions.
6. Three spots around the gear should be cleaned arrd
coated r",'ith Prussian Blue to check tooth contact dur-
ing initial loading.
During startup
Check for unusual noises and record vibrations fron'r
minimum speed to maximum operating speed.
Monitor pressure drop across filters. Ner,y filters should
be available during startup for changeout.
Oil temperatr-rres should be watched closely.
After about 30 minutes, the gear should be shut dou,n
and the case opened to inspect gear tooth contact.
A final shutdorvn for gear tooth inspection should be
made after full load operation and stabilized oil tem-
peratures have been established.
Vibrqtion monitoring. For critical installations rr-here
downtime is extremely expensive, a good vibratioli n-roni-
toring system is cheap protection. There is no question
that failures can be minimized or avoided b1, uiing vi-
bration monitoring equipment now available. I still far,or
Based on introductory
to a paDel session on
gear units, ASME
Cortjre."", fo.- e"-
seles, Sept. 18,1973. -*
Problem Diagnosis
A new approach to
turbomachine ry analysis
Large process plants containing many
rotating machinery systems present
unique data collection and analysis
How do you physically get
all the machinery with the necessary
instrumentation? Here is how one
company solyed the problem
Dole Lorio, Shell Chemical Co., Deer Park, Texas
ExcessrvB Ir.{TNTENANCE costs resulting from high
vibration levels on rotating machinery can be significantly
reduced if problems are quickll' detected, analyzed and
corrected. Shell Chemical Company's llouston plant ac-
complishes this by using a mobile analysis unit consisting
of an oscilloscope, vibration monitors, tape recorder,
spectrur)t analyzer and an X-Y plotter. The use of this
unit has
in improved efficiency, better opera-
tional performance and increased life of the rotating
machinery by decreasing the frequency and severity of
For vears 'r,ibration and noise levels have been useful
indicators of machinery condition. But vibration and noise
levels, or amplitudes, are primarily of value only as in-
dicatols of over-all machinery condition. In order to
diagnose the source of unrest within a given machine,
more specific information is required. Vibration frequency,
noise frequeno', phase angles, shaft orbit patterns, vibra-
tion amplitude as a function of speed, and slow roll shaft
motion levels are only a few of the parameters necessary
to form a conclusive diagnosis. To obtain all the required
data and do it quickly and efficiently, Shell Chemical
Company realized that more sophisticated equipment,
used in conjunction with existing vibration equipment,
was mandaton'.
A self-contained mobile unit that could be driven to
and located near the machinery to be analyzed was
selected for this service. Fig. 1 shows the van truck that
is currently in use as a mobile machinery analysis unit.
The van is equipped with a one-ton, roof-mounted air
conditioner and a dehurnidifier to maintain constant
temperature and humidity levels within. The interior
of the van was modified from its off-the-shelf status to
accept the instrumentation, rvork desk, lights and cir-
culation fans (Fig. 2).
Pou'el- to the instrumentation, air conditioner, fans
and lights is supplied through a 440lll0-volt step-down
transfolmer mounted in the rear of the truck. The power
system includes a circuit breaker and distribution facilities.
Power is brought into the truck through 200 feet of
440-volt extension cord rnounted on a reel in the rear of
the truck (Fig. 3). Electrical po\yer is obtained from
welding receptacles located throughout the process units.
The analysis is primarily designed to be used in
conjunction u'ith noncontacting eddy current vibration
probes but is b1' no means limited to this type of trans-
ducer. Velocitv transducers, accelerometers, miclophones.
and pressure transducers have all been used rrith equal
success. Normalll', the anall'sis van has to be parked from
50 to 100 feet from the machine being anah'zed. lloiv-
eve.r, the high frequency detector-drivers for noncontact
vibration prbbes urLrst be mounted near the actual
Fig. 1-Mobile machinery analysis unit.
Fig. 2-lntbrior work area of mobile analysis unit.
Fig. 3-Electrical power distribution system.
Fig. 4-Field mounted housings and interconnecting signal
cables for proximity probe detector-driver devices.
Iocation (10 feet ir.r otrr case) . To accommodate the
situation, the analr.sis van is equipped rvith a 100-foot
input cable and field boxes. The field boxes house the
high frequencv devices for the noncontact
probes and are coml;lete rrith the 1O-foot interconnecting
cable (Fig.
. This allorr.s the field box to be connected
directli, to the probe lead. The input cable connects to a
receptacle in the side of the analysis van. From here the
sisnal is fed into a series of palallel selector srvitches. The
signal input cable supplies
volt DC po\'ver to the
detector-driver der.ices as \\'ell as routing the signals back
to the van.
selcctor srvitchcs cnable the operator to route the
incoming signals to:
series of signal conditioning amplifiers
,\ ser.en-channel F\I tape recordcr for data storage
Spectmm analyzer for anall'sis and recording on an
X-Y recorder
Two vibration monitors which include low-pass filters
for removing high frequency noise
A dual-channel X-Y oscilloscope, either directly or in-
directly through the lorv-pass filters
Thus, it is possible to simultaneously monitor two points
on the vibration meters, two points on the oscilloscope
and a fifth point on the spectrum analyzerfX-Y plotter.
Or, all points can be fed directly into the tape recorder
and selectively analyzed at a later date.
The oscilloscope was initially incorporated into the van
to be used as both a general purpose test instrument and
as an aid to interpreting and/or analyzing vibration
signals. The dual-channel feature of the oscilloscope per-
mits orbital analyses of the shaft on any of the journal
bearings equipped r'r.ith both horizontal and vertical vi-
bration probes. The oscilloscope is also used in conjunc-
tion rvith an engine/compressor analyzer not included in
the van instrumentation.
The spectrum analyzer has the capability of operating
in three distinct modes:
Spectrum analysis
Rpm tracking
Signature ratio
Primarily the unit is operated in the spectrum analysis
mode. fn this mode, the analyzer receives the vibration-
time signal from a machine operating at a fixed speed
and converts it into a vibration-frequency sigrral. This
signal is fed into the X-Y recorder which traces vibration
amplitude versus frequency (Fig. 5). The analyzer auto-
matically sweeps through a preselected frequency range.
In our case, this is generally from 6 through 1,666 cycles
per second for the high-speed compressors and turbines.
The spectrum analyzer accomplishes the same thing as
the tunable filter on a conventional vibration atalyzer.
Ilowever, it has the advantages of doing it automatically,
quickly, provides a permanent record, and has greater
The rpm tracking mode of operation is also useful. In
this mode, the analyzer will track changes in speed and
plot vibration amplitude versus machine speed (Fig. 6).
When used during startup, it enables the operator to
pick out critical speeds and provides him with an ex-
cellent indication of whether the vibration is a result
of imbalance. T'he mode is also very useful in establishing
overspeed trip points.
Signature ratio analysis is used primarily on nrachines
whose operating speed fluctuates. This includes gas en-
gines, diesel engines,
jet engines and machines u,ith poor
speed control capability. Fluctuations in machine speed
causes the recorded frequency of a vibrating member not
to be in its true ratio with respect to the main shaft
speed, i.e., high speed pinion with respect to lor,v speed
bull gear on an engine, speed increaser, pump train. This
makes interpretation of the results difficult and time
consuming. For this reason, the signature ratio mode
records data directly in vibration amplitude versus ratio
of this component to fundamental speed. The relation-
ships are automatically maintained at the same value
regardless of the speed setting or fluctuations in speed.
The seven-channel tape recorder provides the capability
of storing input data in raw form. These data can be
played back at a later date and transformed into X-Y
plots. The recorder allows rapid accumulation of data
on startups and other instances when operator time is at
a premium.
Uses. With the equipment contained in the van, it is
possible to detect malfunctions such as:
Oil whirl
Critical speeds of component paJts as well as that of
the main shaft
Mechanical looseness
Gear problems
Hydraulic effects
Analysis techniques are the same as those with con-
ventional equipment. The van is employed on startups
of all major equipment after overhaul. Such utilization
provides a "fingerprint" for future comparative analysis.
fn addition, all maj,or equipment is analyzed periodically
(quarterly at present) to establish trends and pinpoint
any potential problem
areas. Examples of problems de-
tected and diagnosed using the analysis equipment are
as follows:
Imbalance-established on a high-speed steam turbine
using the rpm tracking mode on shutdown. High
vibration levels were indicated and the plotted curve
was a function of the square of the speed. Disassembly
verified the diagnosis; the turbine had lost a blade
assembly closing block.
Oil whirl-establjshed when a significant peak was
found at approximately 43 percent of rotating speed
on a barrel-type centrifugal compressor. It was found
that oil whirl was occuring in the seal bushings rather
than in the bearing. The vibration amplitude was
reduced to an acceptable level by machining pressure
dams in the seal bushings and by increasing the shaft
to seal bushing clearance.
Mechanical looseness-established on the same steam
turbine mentioned in the imbalance example. Vibra-
tion amplitude detected a two times shaft speed was
approximately double that at shaft speed. Disassembly
About the
Dlr,p Lonro is a, seni,or maclrinety engi-
neer with Shell Chemical Co., Deet Park,
Teras. He is responsible
or application,
selection ond troubleshooting of totating
machinerpy, Supporting o,r eas i,n cluile
sound mrcasut'ement ond, analysis, lubri-
ca,tion stuclies, deaelopment and itmryle-
mentati,on ot' alignment teclmi,ques and
machinerg malfu,nction data collect:ion,
and, analysis. Pt"i,ot erperience
'includes maintenance engi,neering and
Tn 0(l1s tmE-
project engineering u)ith Shell Chemical. Mr. Lorio holds a
B.S. degree in mecholvical engin
Unitereitg, an A.A.S. degree
State Junior Colleg e, T,,-inid,q,d,
Fig..6-RPM tracking mode showing vibration amplitude versus
macntne spee0 cnanges.
indicated excessive clearance betr,yeen the bearing shell
and bearing bracket. Temporary addition of shim
stock provided the proper bearing crush and elimi-
nated the conditiorr.
Reciprocating compressor pulsation-verified
the need
for pulsation dampeners by using pressure transducers
in conjunction rt'ith velocity vibration transducers.
Pressure pulses were detected at 2 percent of peak
pressure. In some cases, this level of pulsation might
be considered acceptable. However, on this particular
compressor, the pulsation frequency matched the reso-
nant frequency of some attached piping. Under these
conditions, the pressure pulsations aCt"d as the driving
for high-level resonant vibration in the piping-.
Serious damage could have resulted had the pioit"*
not been accurately diagnosed and quickly corrected.
The mobile analysis unit is an extremely valuable
machinery malfunction diagnostic tool and r,r,ill become
even more valuable as newer techniques and instruments
are developed. This approach to machinery analysis has
proven to be a formidable method for greatly reducing
maintenance costs, improving stream factors and reli-
ability and preventing catastrophic failures. I
.t+ I
r,2m R.P.t ._
R .t.
Fig. 5-X-Y plot of vibration amplitude versus frequency.
Many existing vibration analysis
procedures are inadequate. Either too
little or too much data are available.
Here's how to match analysis instruments
to machinery behavior knowledge and
reduce maintenance costs in the process
Chorles Jockson, \Ionsanto Polymers and Petrocht:m-
icals Co., Texas Citr'. Texas
A rotrrar, \'IBR.\TIoN analysis system can save millions
in rotating nrachinery failures. The type, speed, and
critical nature of the equipment should dictate the vibra-
tion program needed. There are many spin-ofls from such
a system, e.g. balance equipment and procedures, align-
n-rent equipment, bealing gauging arrd maintenance pro-
cedures. The program outlined is for a plant employing
seventeen hundred people rvith 80,000 plus horsepou'er
in compression equipnrent. 1200 pumps and other special
equipn-rent. 1'1,000 horsepou'er is the largest single com-
ponent size, rvith 60,000 horseporver in turbomachinery at
speeds up to 52,000 RPfI. This paper covers the back-
ground, present s1'stem, practices, and some recent ex-
periences of a svstem in existence for 10 years. Four engi-
neers and one engineering aide (average engineer
experience is trventv vears) make up olrr analvsis groltp.
Prior to the use of formal vibration analysis techniques,
a vibrating lieht vibi-ometer) a nickcl, or sensitir-e hancls
\A.ere our main analvzers.
The first instrument purchased was a hand helcl vibro-
graph. It gave us a pressure sensitive tape of the vibration
waveforrn for records and frequency/displacement anal-
It rvas totallv mechanical, small, lightweight and could
be used on housings, structures, or directly on shafts rvith
a "fish tailed" shaft stick. The tape can easily be copied
for reports. This instrument is still in continued use. It
is most useful in confirmation of non-periodic rnalfunc-
tiorrs sLrch as oil rvhirl.
Our first electronic vibration arrd balancing analyzer
used a seismic, or velocity, transducer and cost about
This unit put us into the preventive maintenance
business on a large scale. I might add that that was not
the original intention. As a Nlechanical Development
Group under Engineering Services, we intended to simplv
define problems by proper measurements' A program of
90 or 180 da1, vibration surveys r'vas established for critical
rotating n'rachinery. The scheduling was complicated by
an equal amount of emergenci, t1'pe inspections. Historical
data files were established. These files contained pure ancl
manually tuned vibration spectra and unfiltered velocitv
level trend charts. These trends were recorded at three
positions in the shaft bearing area, horizontal and vertical
in the radial direction, and in the axial direction (paral-
lel to the shaft) .
The trend charts are helpful because they illustrate horv
far graclual excursions are drifting o\/er a given period of
time. The chart shou's the diminishing health of a ma-
chine. The pure (unfiltered) velocity data is the most
reasonable plot to make. If one n'ill recall simple har-
rronic motion, peak velocity is the only product tel'm at
the first power of frequency and peak amplitr.rcle.
Aw (cos wt)
Arv Peak
Acceleration =i:Aw'
X Peak
T'his t,vpe of analyzer evolved into the current light-
r,veight solid statc analyzers with frequency meter. displace-
Fig. l-Labyrinth seal rub on a centrifugal compressor. Ma-
chine shutdown 14 seconds after startup.
Fig. 2-Rotor shaft crack at coupling kbyway on 9,000 horse-
power steam turbine.
ment, velocity, acceleration and sound measure, strobe
iight tLrnable filter for 10-1,000 Hz use.
Proximity meqsure. As long as the equipment was ac-
cessible and the force transmission lrom rotor to bearing
housing r,vas high, this program rvorked well. However, a
ne\v generation of machines tvere accontpanl,ing large
single train plants.
Normal speeds were now 10,000-52,000 RPM.
Ratios of total machine rveight to rotor rveight were
running 20:1 up to 125:1.
Coupling and shafting rvere total enclosed for con-
tinuouslv lubricated coupling drives.
Equipment was operatine generally above the first
critical speed and often above the second critical.
Horsepower per train
ty'picalli, 10,000 to 18,000
horsepou'er. Other companies are up to 40,000 horsepou'er
Protection monitorins systems using eddy current (in-
ductive type) sensor probes became necessary. We have
approximately 150 of these probes installed on bearing
housings observing shaft motion relative to the bearing.
These "electronic eyeballs" give good sensitivity for early
detection of machine malfunctions and are perfect for
r.ibration analysis.
Protection monitoring is all togcther a different phi-
losophy lrom vibration analysis though nany view it the
same. How'ever, if one will provide a rvell designed moni-
toring svstem. the data can be easily obtained for vibration
syslem. About 1966, rve adopted a
progran called MACE (Measurement, Analysis, Correc-
tion, Enqineering). This rvas'viewed similar to the fire
triangle of fuel. oxygcn, and ignition, i.e. remor.e one of
the legs and you do not have a fire. Removing one com-
ponent of MACE renders it incomplete.
We attempted to expand our measuring hardrvare at
a rale equal to our understanding. As a result, r\.e are
adding capital to our department at approximately
per vear. In 1966, we had about
in portable nrea-
suring equipment, about six plant installed seismic tvpe
protection monitors, and a small lab building. In 1973,
lve have approximately
in portable eqr-rrpn.renr
as listed belorv:
Real time spectrum analyzer with
dual channel orbit capabilities
l4-channel FN,{ tape recorder and
Digital vector tracking filter r.vith
dual channel tuned orbit capabilities
Two and four channel AM tape recorders
Five manual tunable vibration analyzers
Vibrometer, vibrographs
octave band vibration,/
sound analyzer rvith recorder
Alignment-24-channels proximity
Alignment-3 optical instruments and
accessorres S 8.000
Oscilloscopes and accessories S 5.000
Strairl and fatigue gage facilities S 2.000
Test instruments and cables S 3.000
$sns615-piezoelectric, geophone,inductii,e
Tu,o digital tachometers
X-YY' recorder (amplitude vs phase vs
RPI.I plotting) S 2.100
LonS rvheel base van rvith rvork table S 3,500
One might question the amount of investment. but oLrr
rotating equipment maintenance cost is
MM,'r'ear. The
compressor repair cost are only
MM. The turbo com-
pressor maintenance cost with drivers is approxirr-rateh'
The 30,000 horsepoiver ir.i a trvo
ycar old 1,000 ton/day' methanol plant operates less rhan
Prqctices. The API Standards recently published are
good standards for rotating equipment. Monsanto co:-r-
tributes to these standards and uses these standards :s do
the refineries. Some of the practices for protection :,onl-
toring are:
Trvo 90o circumferential oriented probes are attached
at each bearing of critical equipment.
A phase marking, one event per revolution. probe rs
provided on each driver.
Dual channel "high read" alarm and trip readourr
are provided.
Should all measurins points not be monitored. tire
information is routed to the control room for dia-qnostic
Explosion proof, NEMA 7 type enclosures are limited
to the rrrachinery location.
attempt is made to provide installations that allorv
probe replacements while operating.
Leads are protected in rigid and flexible conduit.
Thrust displaccment probes generally read the thrust
collar as opposed to the shaft end near the
Thrust displacement probes are always
shutdown. Calibration of 100 mv/mil with
linear range of 80 n-rils specified.
Five mils (peak/peak) radial vibration
is normally specified.
thrust collar.
on automatic
a minimum
meter range
Two level alarns are always specified. The first warns
andf or allows call-out for troubleshooting. The second
level commits either automatic or manual shutdown.
Electrically deenergized systems which are energized
to trip are preferred over contrnuously energized systems.
Coaxial connectors are kept to a minimum and these
are wrapped in Teflon tape and encapsulated in RTV
silicone rubber.
Sensor svstems are standardized, generally lr,'ith 15'
probe lead lengths.
The monitoring hardrvare is located on the same side
of drivers and driven equipment.
Vibration sensors are arranged so that one sensor is
predominatelv horizontal and the other is predon.rinately
The horizontal sensor must sense the rotating as-
sembly first in rotation with respect to the vertical sensor.
Great care is given to safety by using grounded system
with care that only one ground exists, i.e. no ground loop
There is a philosophy on vibration analysis which must
be realized. It needs to grow in normal steps. One can
always measure more than can be explained. The training
and supporting education comes slowly. Many times we
investigate a pin pointed area not knowing what we will
find, but having at least pin pointed the area. At some
point, science finally or.ertakes art, but there is a lot of
art left. Intuition becomes a strong tool. Often, \\'e can
anticipate what rvili happen without fully understanding
why. To be eflective, one nrust be thoroughly familiar rvith
machine internal construction.
One might presume that many "saves" must occur to
justify the continued purchase of expensive portable and
fixed vibration equipment. This is true. Keeping this paper
limited and current, some selected results will be listed
more to show the t,vpe of coverage rather than the amount.
Cqse l. During commissioning, a 9,000 horsepower, 11,000
RPM steam turbine developed a 14/o non-synchronous oil
whirl instability that eventually reached approximately
5 mils peak/peak vibration, journal to bearing. Several
tests and orbit studies indicated the instability was initiated
by steam valve sequence loading. The shaft first orbits
elliptically from 11 to 5 o'clock. Then after the next quad-
rant of valves opens, the orbit shifts to 9 to 3 o'clock.
The pressure dam type bearing locking pin was re-
located to damp the 9 to 3 o'clock motion and the ma-
chine was operated here for two weeks until a 4 lobe
bearins design by the turbine builder was manufactured.
This bearing was installed in one shift with no loss in
Cose ll. A compressor rotor on a 14,000 horsepower train
was slowly fouled with a clay like deposit over six months,
suddenly unbalancing the rotor during slough off.
Fig. 3-Heavy salt deposits on steam turbine rotor blading
which caused thrust bearing lailure,
The synchronous response jumped from 0.6 mils peak-
to-peak to 2.7 mils peak-to-peak. The equipment was
closely monitored for two more months for a scheduled
shutdown and the spare ro'tor installed. During the next
six months, the process separation design equipment was
revised and a second rotor switch with separation revision
put us back in control. IJnbalance amounts confirmed
rotor res'ponse tests during the mechanical run test at the
factory before shipment to job site.
Cose lll. A 5,000 horsepower motor-gear-compressor train
was shutdown fifteen seconds after commissioning start.
A severe rub was detected by the vibration probes, limiting
the damage to the compressor sleeve and labyrinth seals
(Fig. 1) . The unit was repaired in forty-eight hours. The
reason for the rub was a mismatched top seal half.
Gose !V. Severe misalignment of an 8,000 horsepower
turbine-compressor train caused a coupling-end unbalance
which rose to 4 mils peak-to-peak. Shutdown and inspec-
tion at the coupling hub-shaJt keyway section indicated a
fatigue break 60/o through the shaft (Fig. 2).
Cose V. Startup of a refrigeration compressor accom-
panied by vibration and hot alignment measures indicated
that the compressor discharge pipe to the condenser,
moved the compressor over 100 mils horizontally out of
alignment with the drive turbine. The refrigeration design
manufacturer rechecked his expansion joint calculations
and found an effor in number of convolutions required.
Vibration signals were typical for misalignment, i.e. high
axial vibratibn and twice running speed radial vibration.
Gqse Vl. A turbine oR a five case train developed 3 mils
peak-to-peak vibration at 10,300 RPM. The coupling
shaft keyway was short a full key, thereby causing an un-
balance condition.
Cose Vll. A 14,000 horsepower, 11,000 RPM steam tur-
bine was field balanced with two external balance rings to
thrust collar damage resulting from salt deposits
on steam turbine blading. Note the heat tears on the thrust
collar face.
rcdLrce srnr:lironous vilrratiorr leverls lronr 2-6 rrrils
to 1.25 nrils peali-to-pcali in one shi[t rvithotrt clis-
cxtcrrdcd this rnaclrine's trvo-\'r:rr rLrn irrto
1973 ancl possibly lonqer. Scisrrric anirllzcr witlr strobr:
reference w'cre coupk:d u ith rrrbital phast: rr:rrking to
placc u't'ieht ancl corrct:t irr:r rrininrLrrn oI trvo rulrs. 1\
trr o
cotrltle colrcction rr,:rs .rraclt: orre urontli l:rtcr in
less than onc hoLrr to ellcct a 1.1 rrril
tion lcvel.
Cose Vlll. I )ue to an inrProPcrlr'
bcaring Jrorrsinq, tlre thlLrst orr I balrcl corrrprcssor lr:rcl
u'oln [he ba]ancc piston labvlinth r'lcar:rnce to triltle orie-
irr:rl clcer;rrce.'['hcr thnrst dcflcction graclLutllr,
Lrntil :rt 17 trrils frortr oriqinrl load point, IJLe coltpr.cssor
rlas shrrtcIow n rr'ilh :r snrr.urecl lrrrt
ftii1t'cl he aring
bLrt undenr:rgccl c:olrprrssor r'otor essr':nbl\'. Static baLrncc
colfirntccl loec[ing but tltrrrst
esit oil terrlleraturcs clicl rrot.
About the
CrtlRtos J.LcxsoN is o.n a.nglimeering
f el-
lout in a X[eclrutt,ical Technology Section
utorking on mechanical u,ncl methnd-
oriented problenrs ctt the l\Ionsanto Poly-
ttters ancl Petrcc|teyrical Com.pany, Teras
City, Texas. He lm.s ltccn witlt. tlte cotn-
2") yea,rs tLorking in Di,uision
EnJtineering, Prodtt.ctiott, Maintenance
o,n.rl Plctnt Dcsign, bef ore setting up the
ytresent rle1nrtm,ent in 1f)60 later being
nppointcd engineering s p e ciali s t, then
senior cngineering speciulist. Mr. Jctckson ltolds a B.S. degree
in nreclnnical engineering
Teras AcQM Uni.uersity, plus
an AAS degree in elcctronics teclmology
College of the
ll[uinland, Teras City, Texas. He is a registcrecl prof essional
engineer, a tnember of ASME, API, SESA, Pi Tctu Sigma,
Tau. Beta Pi, Vibration Institttte, and the Texo,s A&X[ Turbo-
nruchinery Symposium Aduisory Committee.
Cqse lX. At anothcr \.{onsanto loaction, a 15,C00 holse-
power stcam turbine \^,as shllt dou,n manuallv on both
tlrrust :rlarnrs pltrs the radinl vibration alarms.
had a heavy deposit ol salt on trto stagcs (Fig. 3). f-he
thrrrst bcarins had lailed and the blading shror,ds con-
tactcd tlre cliapl-rraems cutting into thern altc',ttt
4 shou's rlzrtrrge to the thrrrst r:ollal fronr overireatinq...
1'lrouglr the trailing edges on trr,o blaclinq- ro\\'s \\ere cllair'n
and lcqrrirecl gr:incling. the rotor rvas savercl and put back
in scrvice one rrecli later.
Nlonsanto l-ras paid fronr
I\,IfI to
lI\I oLr rirrs
11,1te oI u reck in the
Cose X. At another Nl[onsanto locrtion, the coLiplirr
housing filled rvith oil clur: to an iurpropcrly sizecl trsket
rvhich clrokcd the drain. f'his carLsccl an overfilling rriLrclr
took serrerll hoLrrs fronr startr-rp. A or-rc-ha1f nrnning fre-
cluenc)' vibration u,as the inciic:rtion. Tlre nraclrinr: rr'as
shrrt c1orr,n to inspecl the bearilg. A couplc of entr(rti,:
nraintenance men startecl liftinq thc top half oI tire
coLrpline guarcl
onlv to find the flooclecl condition. ir ir':r.
a lrLckv acciclcnt.'fhe bcarine \\'as sllslrect btrt tirc c:,r.r-
pling, Ior thc first tirne in its lifc. sarv the opprrrtrrr,,r,. ',
bccorne a bealing, tlrough a
Tlris brings rrp an intcrcsting thoLrglrt rrhoLrt i'ibr'-.ii, i,
annlrsis. Sever'.rl rnallLrnctior-rs ma1' scncl thc serr' t.,r ' .i
ileclLrcncl'signal, i.c. one miglrt think oI tlris.ri LirL.'-r'
rnorc reclio st:rtions transnritting on somcone clsr s iie-
c]l tel tc\,.
A loose thrtrsl collar, loosc cotrpling hrrb, clorLblr' ;rrr,ke.i
shaft. oL nrisaligrrrrertt rrright n,cll scrrcl a trvit c t
spcccl lreclLLer-rcy signal.
llcceusc 60(b o[ our ribration problcrns resLrlri
nrisaliqrrit'nI o\ cr oLrr first fir'c vears of vibratiorr .
vr"c bcqan :r prt'cisc nlcusLrrenrcnt oI slraft and c:rsr
r-ncrt rrsiug the edclv crrrrclrt probes. Wr.l liarc 21
oI this ecluii-lrrcnt, plus ll optical instnrrncnts tir
rvork. lioth tcclrnirlLres have been pLrblishecl.l':
nrost diflir:rrlt problern irr vibration .rr-r1.,.i.
is cstablislrr rent of sevclitl, lirnits. M:it lrirres. rr LL,
pcolrlr:. lrar e cli[Tcrent sersitir']Lir:s end rcsist:inces ro r rjrr'-r-
tion lirrc es.
In gc'rcr-al. \rc lra\c founcl tlrat 0.2 irrclrcs':ec,:,i,d
(peak) r'clocitv to retlLrilc:rlarrrr action on beerirq- hoLrsir:
scisnric tr'pc reacling. L}r'eater vlrlrrcs. to 0.5 iir.
ronrrll1, Iorcc slrLLtclorr Lr. The shaft-to-bcllir,g r', i. r.i '
rreasrrre lirrits :Lre clilTrlent. \\'itJr cle:rn sh:rlts. rr('
erallr,' sct spted rrechines u,itli 2.2 liils pcrl.r-t, -ir,':rl.
rr arnirq arrd
rnils peak-to-pcak slrutdo\\,n lirrrits,
Rotor tlirLr,st displrcernunt rrronitors.irc s( t rlitlr li 1,,i.
orcr normal loacl clcflection as an alarn-L:rncl ,'i1]11'1'l('r r,:
25r mils
nolnral lord dr'flcction:ls alr atrtorrr.rtii ..,.ri-
clorr'n position. Ilascrl on planI cxpericnce, it is felL th^rr
vou crlnnot risli rnarLr:Ll slruti]orvr-t irt thrust. The rorrr.r i
failLrrc tirne is thirtr',st'conds!
he autlror rvishes to acknorrledqe the contr ibulions of linginccLs E- .f-
ParrI S.
Cart A. Drrhon. Jlorrerrl Ill;rckhrrrn. ancl Jrrrrcr
[[ Inqrerrr, NIrtrrials & \IcclranirrL 1cc]rnologr, )\forrsa:rto,'lrras Clit..
Tc\,rs to tlr: \l\(lE
irs rcportcd. C)opyriglrtcd lrv rrrrrl origir.,lLr'
r'\crtr(l at tlrc Tnstitutc ol [,lcctricrLl ancl l]lrctroric [irrqirrcc s, ['etLolcrrm
Dirisiorr ('rrrlrrcrrcr'.
Selrtcrrrlrcr 1!73, IIotLstorr,'l'txrs,
flrcl.on. C., "Horv to aliqn lrnrrcl-type ccntrilrrqal cornprcssors," Hldro-
,'arlrort ['rocessing, ir0, No.9, p.1{19, Scptcrrrlrt:r 1!71.
Ci., 'SrrcccssIul slraIt hot alJgnnrcnt,'' Ilydror:arlron J)rocessing,
l{1, \o. l. p. t00, .[anrLary I969.
cold, 32,33
hints on,33
problems in, 1 13
tooling tor,32
Balancing, dynamic
compressor, 7 8, 1 13-1 1 4
shop, 21
steam turbine,74,75
Cleaning onstream
solvent washing, 30
water washing,28,29
Com pressors, centrif ugal
applications for, 18, 19,23
construction materials for, 79, 80
mai ntenanc e ot, 20-22, 24-27
operation of, 20
rotor dynamics of
19, 78
seals for, 80
seal-oil systems for, 88-92
testing of , 5-16
mechanical, 6-10
performance, field, 1 4-1 6
shop, 11-13, 39,
Couplings, high-speed
applications tor, 84-87
specifications for 9, 79
Data acquisition
compressor field performance,
12, 13
vibration, 108-113
Gear units, high-speed
failures of,94,95
shipping and installation of, 106
shop testing of ,
specif ications for, 93-94, 96-100
test, 96-1 00
vibration of , 96-106
Pumps, centrifugal
cavitation, 56-59
grouting, 41-44
performance, 46-48, 58, 59
specifications for, 59
mai ntenance personnel, 24, 25
Turbines, steam
balancing, multiplane, 74-74
blade failures, 73
fouling, 28,29,113
water washing, 29
applications for, 66
construction materials for, 63
fabrication of, 64
performance of ,
power recovery f rom, 66-69
types of, 66
analysis, 108-1 13
compressor,T6, 109, 1 13, 1 14
gear units, high-speed, 96, 106
instrumentation, 108, 1 13
severity limits, 54, 55, 1 14
specifications, 78