chapter 14 diagnostics

7/30/2019 Chapter 14 Diagnostics

1/28

743

C H A P T E R 14

Machinery Diagnostic Methodology 14

The following chapter is based upon amini course prepar ed by th e senior a uthor for an an nua l meeting of the Vibration

Inst i tute1. Although years have elapsed since this presenta tion and the t ools a nd

instrumentation systems have evolved to more sophisticated levels, the basicproblem-solving a pproach towa rds ma chinery m alfunctions remains unchanged.

Thus, the systematic thought process contained herein is still considered to be

quite appropriate. Over the years, this approach has proven to be a successful

a nd effective methodology for solving ma chinery problems.

This chapter will review these field-proven methods for the diagnosis of

ma chinery problems. The correct ident ifi cat ion an d proper dia gnosis of most

malfunctions requires the correlation of parameters such as mechanical con-

struction, process influence, maintenance history, and vibratory behavior.

Although ma chinery dia gnosis ha s occasiona lly been a ssociat ed with mysticism,

it really should be considered as an engineering project. This type of project

demands a lot of ha rd w ork, a belief in physical law s, plus a methodical problem

solving approach. Requirements also exist for a certain level of skill, diagnostic

tools, and relevant experience. For the purposes of this chapter, the emphasiswill be placed upon the problem-solving approach, that is, the Di agnosti c M eth-

odology. Although t he specifi c approach w ill var y from problem to problem, there

are similar it ies tha t can be cat egorized into the following seven ma jor a reas:

1. Diagnostic Objectives

2. Mechanical Inspection

3. Test Plan Development

4. Data Acquisition and Processing

5. Data Interpretation

6. Conclusions and Recommendations

7. Corrective Action Plan

Each of the above categories will be addressed, and the methodology associ-at ed with t ra nsforming th e unknown mecha nical failures into the realm of docu-

1 Robert C. Eisenmann, Machinery Diagnostic Methodology. Vibrat ion Inst i tute - Min i CourseN otes - Machinery Vibrati on Monit oring and Anal ysis Meetin g, New Orlean s, Louisiana (May 1985).


2/28

744 Chapter-14

mented information will be discussed. These concepts will be illustrated with a

series of field case histories dealing with startup problems encountered on three

different machinery trains. Although these problems occurred on new machines

during init ia l operation, the sa me techniques are a pplicable to existing mecha ni-cal equipment a nd condition monitoring programs.

DIAGNOSTIC OBJECTIVES

The need to establish diagnostic objectives is often overlooked, due to the

fact t ha t t hese objectives ar e obvious to all part ies involved. However, there is an

old bit of philosophy t ha t says: th e first step to getti ng an ythi ng done is to wr it e it

down. This statement may seem trivial, but it does represent a basic truth. In

the realm of machinery analysis, it is easy to lose sight of the real objectives on

complex projects that last a long time. It is not uncommon for individuals to

become wrapped up studying some unrelated frequency that has absolutely

nothing to do with the real mechanical fault .The form or format of the statement of objectives can be very simple, or it

may be quite forma l. On some minor problems this m ay consist of a note tha t you

carry a round in your wallet or a line item on the w eeklyTo Do Li st. For a com-plex problem w ith high visibility, significant fi na ncial impact , and multiple part y

involvement, the objective statement may evolve into a formal daily status

review meeting. In a ny case, the idea is to maint ain a continuous reminder of the

real project or analysis objective(s), and to continually move in the direction of

sa tisfyin g th e objectives, a nd solving th e problem.

MECHANICAL INSPECTION

On-site mechanical inspection is a significant part of the database for any

ma chinery problem. This includes a hands onexam inat ion of the ma chinery, and

the actual installat ion. Familiarization with the operating and maintenance his-

tory is highly desirable, if not mandatory. On a new installat ion, refer to the

OEM shop testing records and customer witness reports. In recent years, the

tendency t hroughout m an y industr ies has been t o rely upon remote ana lysis via

FAX machines and transmission of data. Although this is a cost-effective

a pproa ch for routine m echa nical problems, it is not a dvisa ble for diffi cult a nd/or

complex problems. Hence, direct inspection of the machine in distress still pro-

vides the most informa tion about t he equipment.

The mechanical construction of the machinery must be examined. It is vir-

tually impossible to diagnose machine behavior without knowing what lies

beneath the outer casing. This should include a review of assembly clearancesand tolerances, bearing, seal, and coupling configuration, plus overall rotor

assembly. Ideally, there is an opportunity to work with the millwrights, and be

right there to observe (or participate) in the actual physical measurements. As

an a lternat e, there is often a w ar ehouse with identical spare par ts. Examina tion


3/28

Test Plan Development 745

of the stored spare parts also provides a perspective on the storage condition of

the part s. Questions a rise as t o how t he rotors a re stored to prevent sh aft bows,

an d how t he ma chine elements a re protected from environmenta l degrada tion.

On problems where failed ma chine part s a re ava ilable, they should be thor-oughly scrutinized to help determine the failure mode(s). Sometimes metallurgi-

cal examination of these items can provide positive benefits and data. Also, look

a t t he shop a nd/or field ba lan cing techniq ues an d procedures. H ow good a re the

records? How good is the work? The same applies to alignment data. Are indica-

tor readings corrected for bracket sag? Are soft foot checks made? What type of

shim stock is normally used?

From an externa l sta ndpoint , the a ssociat ed lube an d seal oil system should

be reviewed to determine any associat ed peculiarit ies th at could ad versely infl u-

ence the mechanical behavior of the machinery. Sometimes this can identify

problems such as a damaged oil pump, and other t imes it may uncover a col-

lapsed float in a seal oil drain pot. The machine foundation and support struc-

ture should be examined, and the grout between the baseplate and the

foundation should be visually inspected. The associated piping system should be

reviewed, an d part icular at tention paid to th e location of sliding and fi xed pipe

supports, plus spring hangers and dampers. The location and general condition

of the main process block and check valves should be checked, and the funda-

mental machine process control technique should be identified. For instance,

ma chine speed may be regula ted by a pressur e controller sensing discha rge pres-

sure. If the pressure sensor or the field controller misbehaves, the result might

be an erra tic speed control from the governor.

The bottom line for this section is simply you shoul d become in ti mat ely

fam il iar wi th th e mechan ics of t he machi ner y and th e associat ed systems. Failure

to perform th is init ial familiariza tion with th e installat ion can easily result in a

misdirected a na lysis effort .

T

EST

P

LAN

D

EVELOPMENT

Based upon the diagnostic objectives, and the mechanical inspection, the

development of a r ealist ic test plan would be the next logical step. This test plan

should be directed a t learning a bout t he behavior of the machine and t he defined

problem area(s). The actual test may take many different shapes and formats,

depending on the type of problem(s) encountered. The testing may include vari-

a ble speed runs, va ria tions of process loa d or operat ing condit ions, chan ges in oil

supply temperatur e or pressure, or a ny other para meter tha t can be altered in a

controlled and repeatable manner. The testing may include mechanical changes

such as variat ions of mass unbalance, machine alignment, baseplate at tach-

ment , piping support , or physica l ma chine element cha nges.In other cases the testing may center on structural, or discrete resonance

testing. The items to be tested may include the specific machine components, a

spare rotor assembly, or a variety of structural measurements. The options are

many and varied, and it should be recognized that not every field test will


4/28

746 Chapter-14

directly identify a problem. In some situations, a test is conducted to verify that

a pa rticula r problem is not occurring. For exam ple, differentia l pressure a cross a

compressor may be increased just to prove that the machine is not sensitive to

load changes.For a t est pla n to be effective it should include a description of the physical

parameters that will be varied. Next, an outline of the specific test procedure

must be developed. The procedure needs to be agreed upon by all involved par-

ties before test ing begins. In a ddition, the procedure should go into specifi c detail

as to the measurements that will be made, and some statement regarding the

expected results or knowledge to be gained from the test. If possible, the normal

or expected response of the unit should be stated, and if the field test provides

different results, the deviation may be immediately recognized. From a safety

standpoint, appropriate limits should be established for the critical values (e.g.,

sha ft vibra tion or bearing temperature). In a ddit ion, general a greement sh ould

be reached beforehand tha t t he fi eld test would be aborted if any of the predeter-

mined safety limits are exceeded.

Finally, during the investigation of a complex problem it is possible to con-

duct tests that appear to provide minimal results. I t is important to maintain

overall perspective and recognize that there is something to be learned from vir-

tua lly every t est . Alwa ys ma inta in the perspective that some tests help to elimi-

na te potentia l ma lfunctions an d other t ests help identify the rea l problem(s).

D

ATA

A

CQUISITION

AND

P

ROCESSING

The acquisit ion of field data during a planned test , or on a routine basis,

requires the a ssembly of the proper t ra nsducers a nd t est instrum entat ion. The

tra nsducer suite must be compat ible with t he machinery and th e measur ements

to be made. The test inst rumenta tion must likewise fi t int o the requirements of

the test program. Prior to initiation of testing, it is also necessary to check conti-

nuity of all wiring, perform system calibration checks, and include sufficient

means for post-test verifi cation of the da ta . In some cas es, the tr an sducers ma y

be subjected to another calibration check after the test is completed. For exam-

ple, if the operating environment for some of the transducers approaches the

temperature limit for th e pickups, then a post test ca libration would be advisa ble

to verify tha t t he pickup wa s not perma nently da ma ged during the test . In some

insta nces, the test results ma y be rendered invalid due to the fa ilure or ma lfunc-

tion of an important measurement.

It should be recognized that the data acquired is really a function of the

specific machine, the associated problem, and the individual test plan. Typically,

fi eld data acquisit ion combines a va riety of dynam ic an d sta t ic information. Mea-

surements of displacement a nd velocity a re used to examine the sha ft a nd casingmotion in the frequency domain surr ounding run ning speed. High frequency cas-

ing acceleration measurements are used to observe excitat ions such as blade

passing or gea r m eshing. Specialized meas urements, such a s pressure pulsation

or t orsional vibrat ion, ar e applied a s required. These dynam ic measurements a re


5/28

Data Acquisition and Processing 747

usually recorded on multicha nnel instrumenta tion grade ma gnetic tape record-

ers a nd/or directed t o a digita l signa l processing syst em (e.g., Fig. 8-5).

Ana log output s proport ional to process va ria bles such as t emperat ure, pres-

sure, or flow ra te ma y also be included. During ma ny situa tions this type of dat ais hand logged, or obtained from strip chart recorders. In other cases, where time

correlation is important , the analog parameters may be simultaneously tape

recorded with the dynamic vibration signals, and suitable t ime stamping pro-

vided. On modern plants with Distributed Control Systems (DCS), a variety of

gra phic an d ta bular process data presenta tion options a re ava ilable.

The recorded test data should be supplemented by hand logged information

a nd a pplica ble on-line da ta reduction. This provides immedia te documenta tion of

significant information, and direct verification of the information recorded on

magnetic tape or digitally processed. If this step is bypassed, then the credibility

of the data suffers; and the final solution to the machinery problem may be

unnecessar ily delayed.

Detailed data processing follows the completion of field data acquisit ion.

For digita lly sam pled da ta the informat ion is often processed and ava ilable con-

current with the completion of the field test . This hard copy data is reviewed,

and additional formats processed as necessary. For tape recorded information,

the taped data is reproduced into addit ional diagnostic instruments to observe

both the steady state and the transient behavior of the machinery. During this

phase, the data is processed to hard copy format. The accuracy of the data

recording syst em must be checked, and a ll calibration signals must be reviewed.

Comparisons should be made between the final data plots and the field logged

data (e.g., the 1X vectors acquired during testing should be duplicated on the

fi na l ha rd copy plots).

As a minimum requirement, the tape recorded signals should be viewed on

a n oscilloscope. In most ca ses, the signa ls a re processed thr ough a Digit a l Vector

Filter (DVF) to view the synchronous a mplitude and phase, plus a Dyna mic Sig-na l Ana lyzer (DSA) to observe th e frequency cont ent of the sign a ls. The DS A also

performs special operations such a s s ta t ist ical avera ging, frequency expan sion,

frequency response functions, and coherence. Normally, the signals are simulta-

neously documented in the time, orbital, and frequency domain. When applica-

ble, radia l a nd/or axial posit ion da ta is a lso calculated a nd char ted.

Often the data is further summarized or correlated into appropriate

gra phic or t abula r format s. This a llows for a concise summary of the pertinent

informa tion, and usua lly provides more dat a visibility during t he interpreta tion

stage. Various tools and techniques are available for data reduction. Some are

simple, others a re sophisticated. In general, the da ta processing system must be

reliable, provide sufficient resolution of the data, and integrate with the field

da ta acquisit ion system to produce accura te an d repeat able results.

Although the diagnostician is generally concerned with the complexities ofacquiring and accurately processing the dynamic vibration data, the associated

steady state measurements should not be ignored. This includes the operating

process informat ion of pressures, temperat ures, and fl ow ra tes. Load informa tion

such as measured shaft horsepower or motor currents may also be meaningful


6/28


7/28

Data Interpretation 749

D

ATA

I

NTERPRETATION

Data interpretation is the formative phase of any machinery diagnostic

project . This activity includes a summary and correlation of all pertinent dataacquired during the project. This includes the mechanical configuration, process

an d ma intenan ce history, field testing d at a, plus a ny supportive calculat ions or

analytical models. All of this information constitutes various parts of the puzzle.

Often the problem cannot be solved unless enough puzzle pieces are available to

develop a logica l an d consist ent overview. The diag nostician is ca utioned a ga inst

ta king a superfi cial a pproach during t his phase of the project . In ma ny cases, a

quick review of the data plots does not solve the problem. Normally, each set of

test conditions must be compared, the behavior of the machine must be exam-

ined, and the results committed to some type of summary log. Although this may

be a brutally difficult exercise, it is the only way to properly quantify the field

test results.

Actual int erpreta tion of the t est da ta is strongly dependent on the type of

machinery and the operating conditions. The 30,000 horsepower gas turbine

behaves quite differently from a simila rly ra ted stea m tur bine. Hence, the dat a

must be interpreted in a ccordance with t he physical cha ra cterist ics of the par tic-

ular ma chine type an d th e opera ting environment. One approach is to view the

da ta in terms of norma l behavior for a pa rticular ma chine type, an d then look for

the a bnormalit ies in r esponse cha ra cterist ics.

It sh ould a lso be mentioned tha t t he diagnostician m ay wa nt t o use some of

th e evolving comput er tools for acquisit ion, processing, and in itia l interpreta tion

of the dat a. The general a vaila bility of computer programs using ar t ifi cial intelli-

gence a re generally termed

Expert Systems

for the diagnosis of machinery prob-

lems. These progra ms ma y provide significant t ime a nd a ccura cy adva nta ges for

the diagnostician. The most powerful programs acquire, correlate, and compare

transient and steady state data. Based upon these results, a series of potentialma lfunctions are identifi ed.

Less sophisticated dia gnostic programs rely on a n

in tervi ew pr ocess

where

the user provides answers to a series of questions, and the program responds

with a list of potential malfunctions. In all cases, the machinery diagnostician

should recognize that the depth of problems addressed by these software pro-

gra ms a re limited to the know ledge a nd experience of the people th a t esta blished

the init ial database, and the rules for problem identification. Hence, if the

ma chinery problem is fa irly common (e.g., th e malfun ctions discussed in chapter

9), expert s yst ems w ill produce good result s. Conversely, if the problem fa lls int o

a unique category (e.g., as presented in chapter 10), these programs may not be

able to identify the true origin of the malfunction. Although computers are won-

derful data processing devices, they still pale in deductive reasoning powers

when compared with t he human bra in .


8/28

750 Chapter-14

C

ONCLUSIONS

AND

R

ECOMMENDATIONS

The final conclusions and recommendations constitute two separate, but

concurrent thought processes. The development of the conclusions is consideredto be a sum ma ry of the knowledge gained by executing the a ctivit ies discussed

during the previous six sections. The conclusions are really a summary state-

ment of what s wrong wi th the machin e. In m an y situa tions, the conclusions ma y

also summarize whats right wi th t he machine. In some cases the conclusions

may be direct and precise. On other problems, the final conclusions may be less

defi nitive due to the na tur e of the ma lfunction an d/or the lack of accura te or suf-

ficient data. It is possible to reach a conclusion that the test data is insufficient

for solving the problem, and a retest with addit ional parameters or measure-

ments is wa rra nted. Certa inly this ma y be expensive, and professiona lly embar-

rassing. However, it is always much better to be honest about the test results,

rather than attempt to hide inclusive test results in a mountain of technical

di ther and tr i v ia

.

The recommendations, on the other hand, are specific statements of whatcan be done to fix the machine. This may be as simple as: r epl ace th e outboar d

bearing.

Conversely, the recommendations may be complicated, and require a

series of phases or steps presented in a logical sequence. Specifically, the first

phase of the recommendations may consist of actions that should be executed in

an immediate time frame to keep the machine on-line. The second phase may

consist of action items for the next scheduled turnaround. The third phase may

include long-term items t ha t r equire addit ional investigat ion, ana lytical simula-

tion, or redesign for improved safety, reliability, or availability. In all cases, the

diagnostician should include a healthy dose of real i tyin the development and

presenta tion of the fi na l recommendations.

C

ORRECTIVE

A

CTION

P

LAN

The corrective action plan is t he fi na l sta ge of a ny ma chinery ma lfunction

scenario. During this stage the recommendations are discussed, economic influ-

ence is introduced, and a fi na l action plan is generat ed. This a ction plan ma y be

different from the recommendations due to factors such as plant production

requirements, turna round schedules, or situa tions where th e ma chine modifica-

tion cost exceeds the repair costs dur ing a fi xed time period. Like many other sit-

uations, the final action plan is a compromise between the parties involved. The

best possible corrective action plan is one that maintains a reasonable balance

between sound engineering judgment and proper financial considerations. It

should always be recognized that repairs or modifications to large process

machinery must compete with other projects within the corporation requestingfunding. In some cases, the economic decision dictates that the mechanical prob-

lem be tolerated. In other extreme conditions, the loss or failure of one machin-

ery tra in ma y result in t he corpora te ban kruptcy.


9/28

Corrective Action Plan 751

Case History 50: Steam Turbine Electrostatic Voltage Discharge

On ma ny ma chinery problems th ere is direct physica l evidence to help iden-

tify t he origin of the ma lfunction. This evidence is often combined wit h mea sure-ments, such as increasing vibration amplitudes, to identify the problem onset.

However, there are ma chinery ma lfunctions tha t exhibit minima l physical sym p-

toms, and sometimes result in the reduction of vibration levels with time. The

following case history describes such a failure that occurred on the machinery

train presented in Fig. 14-1. This unit consists of a six stage, horizontally split

refrigera tion compressor driven by a 28,600 HP condensin g st eam tur bine. Typi-

cal operating speeds vary between 3,700 and 3,900 RPM. The turbine accepts

superheated steam at 600 Psi, and exhausts directly to a close coupled surface

condenser. The turbine was originally equipped with oil dam sleeve bearings,

an d t he compressor rotor wa s supported by t ilt pad journa l bearings.

Approximately tw o months a fter original commissioning a nd a successful

plant sta rtup, the compressor inboard bear ing failed. Fortunat ely, the sha ft w as

not damaged and a spare set of pads was installed. During the next two monthsof operation, the vertical vibration at this coupling end bearing decreased from

an init ial st ar t ing level of 0.7 Mils,

p-p

to a minimal amplitude of 0.3 Mils,

p-p

. At

the t ime, severa l individuals commented on the fa ct tha t the compressor vibra -

tion was improving an d the machine was heal ingits elf on-line. This proved t o be

fuzzy

thinking, but it w as a popular t heory for a short period of t ime.

Eventually, shaft vibration amplitudes began a steady growth from 0.3 up

to 0.7 Mils,

p-p

during a three week period. Since this was back to the original

amplitude, lit t le at tention was paid to the change. Compressor shaft vibration

continued to gradually increase and one night it jumped up to 3.4 Mils,

p-p

. The

inboard vertical probe gap voltage revealed an overall 4.9 volt change. This DC

Fig. 141 Propylene Compressor Machinery And Vibration Transducer Arrangement

Steam Inlet600 psig @ 750F Exhaust4" Hg Abs

Normal

Thrust

Gear TypeCoupling

3H

K

45 45

CCW

1V

Propylene RefrigerationCompressor

6 Impellers

-35FSuct.-5

FSuct.Ext.+60

FSuct.Disch.

3V

80

CondensingSteam Turbine

11 StagesRated at 28,600 HP

@ 4,150 Rpm

45 452V

5

4V85

4H

2H1H

FailedBearing

1Aa & b 4Aa & b


10/28

752 Chapter-14

gap change was equivalent to 24.5 Mils of vertical shaft position drop (i.e., 4.9

Volts x 5 Mils/Volts = 24.5 Mils). Since th e ba bbitt th ickness on the bea ring pa ds

was nominally 15 to 20 Mils thick, there was concern that substantial damage

ha d occurred a t the compressor inboard. The next morning the t ra in wa s shut -down t o inspect a nd probably replace that coupling end bea ring. It w as discov-

ered tha t t he bear ing dam age extended to the point th at the journa l wa s severely

scored. Hence, th e compressor case ha d to be split a nd t he spar e rotor inst a lled.

In retrospect , the shaft damage should have been anticipated, since the

total vertical shaft drop (24.5 Mils) exceeded the available bearing pad babbitt

thickness (15 to 20 Mils). Thus, the steel shaft was riding on the steel bearing

pad, and any breakdown of the minimum oil film would result in steel to steel

contact (i.e., shaft to bearing backing). However, at that point in time, minimal

at tention wa s given to the trend plots of gap voltage versus t ime.

The third failure on this machine was monitored very carefully. Another

vertical proximity probe was installed directly opposite the existing transducer

for confi rma tion of the vertical sha ft posit ion. During the next t wenty-four days,

the vertical vibration amplitude decreased from 0.64 to 0.26 Mils,

p-p

. Concur-

rently, the top vertica l probe DC ga p increased from -7.0 to -9.3 volts, a s sh own in

the F ig. 14-2. This volta ge chan ge wa s equiva lent t o an 11.5 Mil vertical d rop of

the compressor shaft. This change was also reflected by the new probe mounted

on the bottom of the bearing assembly. Based on the previous failure, the train

wa s shutdown, and inspection revealed tha t the babbitt w as a lmost tota lly miss-

ing from the bottom shoe. The journal exhibited a frosty-satin-gray appearance,

much like the failed bottom shoe. Further inspection showed similar characteris-

tics on both turbine and compressor thrust bearings. A slight indication of frost-

ing wa s a lso noted on the turbine exhaust end bearing, and t he bronze governor

drive gear mounted at the turbine outboard .

Fortunately, one of the OEM reps had previously encountered similar fail-

ures on electrical machinery. He reported that the discharge of shaft voltages

could cause the failures, and recommended the installation of rotor grounding

Fig. 142 CompressorCoupling End BearingVertical Shaft PositionChange With Time

J

J

J

J

J

J

J

J

J

J

J

J

J

J

J

J

J

J

J

J

J

JJ

J

J

-10.0

-9.5

-9.0

-8.5

-8.0

-7.5

-7.0

-10.0

-9.5

-9.0

-8.5

-8.0

-7.5

-7

0 5 10 15 20 25

ProbeGap

Voltage(VoltsDC)

Vertical

ShaftDrop(Mils)

Elapsed Time (Days)

Initial Shaft Vibration = 0.64 Mils,p-p

Final Shaft Vibration = 0.26 Mils,p-p

0

-5.0

-2.5

-7.5

-12.5

-15.0

-10.0


11/28


brushes. It was postulated that the increasing vertical probe voltage really

meant tha t t he journa l wa s sinking into the bearing, as represented in Fig. 14-3.

Based on this physical evidence, and the OEM recommendation, insulated

brushes were insta lled on the outboard sha ft end of the turbine a nd t he compres-sor. Both brushes were to be grounded directly to the machine baseplate.

Reexamination of the historical log sheets revealed that all three failures

exhibited identical characterist ics of decreasing shaft vibration amplitude, and

increasing vertical probe gap voltage. Only during the second failure did the

shaft vibration increase, and this was attr ibuted to the fact that the nominal 15

Mil bearing babbitt thickness was exceeded. Hence, there was a good deal of con-

fi dence tha t a ll three failures were due to the same mecha nism.

The ensuing startup with the installed ground brushes was viewed with

optimism. Shortly after startup, the machinery was operating smoothly, but it

was discovered that the compressor brush had not been connected to ground.During t he act of physically a t ta ching the loose wire to the ba seplat e ground, a

substantial electrical arc was encountered. The surrounding atmosphere was

filled with propylene fumes, and it was fortunate that ignit ion did not occur.

After the machinery was on-line, the voltage potential between rotor and ground

wa s mea sured. Upon disconnecting the brush lead from t he baseplate, an electri-

cal spark was again experienced. Readings taken with a series voltmeter

revealed levels in excess of 20 volts. Hence, there was no question that the

ma chinery w as producing an electrical volta ge.

A more deta iled a na lysis wa s m ade following eighteen months of successful

opera tion. The signifi cant plots fr om this a na lysis a re display ed in Fig. 14-4. The

top t ime base a nd spectrum plots document t he compressor sha ft vibra tion char-

acteristics measured by the vertical probe at the inboard bearing. The major fre-

quency component occurs at running speed of 3,750 RPM, with sub-harmonicshaft motion at 48%and 63%of speed. Shaft voltage signal characterist ics are

presented on the bottom half of Fig. 14-4. It is of particular interest to note that

the spectra l content of the shaft volta ge is virtually identical to the vibrat ion sig-

na l. This similarit y even extends to t he a ppeara nce of th e 48%a nd 63%of run-

Fig. 143 CompressorCoupling End BearingVertical Shaft PositionChange

JournalBearingPad Babbitt

ShaftDrop

Steel PadBacking Babbitt Loss


12/28

754 Chapter-14

ning speed components. It is reasoned that as the shaft orbits, the electrical

discharge will respond to the dynamics of the minimum oil film. Thus, similar

frequency content should be expected from both the displacement vibration

probe and the electrical discharge signal. I t should also be mentioned that the

time domain shaft voltage data is presented at a sweep rate of 10 milliseconds

per division. If this scale was expanded to 2 milliseconds per division, the long

vertical spikes (120 volts), ca n be observed to dissipa te in a pproxima tely 3 milli-

seconds (0.003 seconds). This volta ge chan ge is equiva lent t o a d ischar ge ra te of

40,000 v olts /second !

Generation of internal shaft voltages is apparently limited to condensing

turbines where the last sta ges are subjected to sat ura ted steam . The aut hor has

never encountered this problem on a back pressure turbine. It is generally

hypothesized that the brushing effect of water droplets (condensate) across the

blades develops an electrostatic charge that builds up on the rotor. This charge

periodically discha rges to ground th rough th e point of least resista nce. When th e

rotor voltage dissipates across an oil film bearing, microscopic pits are produced

in the babbitt . This results in the continual removal of metal from the bearing

surface. Trouble may be noticed in periods ranging from a few days to several

years, depending on the magnitude of the current flow. This phenomenon shows

up as a frosted surface in the load ed zone of the bearing w here the oil film is at a

minimum thickness. This type of failure also appears on thrust bearings, seals,and any other type of mechanical device in close contact with the shaft. The volt-

ages generated in the turbine rotor are transmitted throughout a train of

mechanical equipment. They can pass across couplings and through gear boxes.

Hence, the eventua l failure might occur a t a bearing fa r removed from the origi-

Fig. 144 Vertical Shaft Vibration And Shaft Voltage From Grounding Brush


13/28


nal source of the voltage. It has been documented that failures due to this elec-

trostatic voltage may be transmitted through three couplings and associated

compressor bodies.

The brush design originally insta lled on this refrigera tion tra in is shown inFig. 14-5. This preliminary design places the grounding brush at a low velocity

point on the shaft in order to minimize brush wear. The brush is insulated from

the case in order to direct the shaft voltage to ground. Next, the brush ground

wire passes through a meter to verify brush operation. This holder can be

changed during operation and it is sealed with RTV to minimize oil leaks. One

axia lly mounted brush wa s insta lled on the t urbine governor end, and a second

brush mounted on the compressor outboard shaft end.

Since this origina l brush inst alla t ion by the a uthor, severa l more sophisti-

cated designs have evolved. Today, there are grounding brushes commercially

ava ilable from several sources. However, the funda menta l mechanism of dissipa-tion of electrosta tic shaft volta ges remains t he sam e.

It must be mentioned that this particular problem has been around for

many years. One of the earliest complete references to this specific malfunction

was an ASME paper by two General Electric engineers: J .M. Gruber, and E.F.

Hansen

2

. Their paper represented un told ma n-hours of labora tory resea rch, field

measurements, and in-depth literature surveillance. Gruber and Hansen dealt

primarily with large turbine generator sets, and they addressed the destructive

effects of sha ft voltages upon bearings. Cat egorically they identifi ed fi ve dist inct

types of shaft voltages: th e electr omagneti c or 60-cycle a-c volt age, a gr ound -

detector 120-cycle a-c volt age, the igni tr on excita ti on volt age, the hi gh-fr equency-

exciter r i ppl e volt age, and th e el ectr osta ti c d-c volt age

. They review ed each of the

categories an d t hen w ent int o more deta il on electrosta tic volta ge. They sta ted:

Th e el ectr ostat ic shaf t volta ge has been f ound to have sever al r easonably w el l-pr onounced cha racter isti cs as foll ows:

Fig. 145 Original RotorGround Brush Installation

2 J.M. G ruber a nd E .F. Ha nsen, Electrostat ic Sha ft Voltage on St eam Turbine Rotors, Trans-actions of the Am er ican Society of Mechani cal En gin eers

, P a per N umb er 58-SA-5, (1958).

Machine End Cover

Shaft

Spring Loaded Brush

Brush Holder

Insulating Material

Meter

Ground

Insulated Wire

RTV

Seal


14/28

756 Chapter-14

1. The volt age betw een shaf t an d bed pl ate is dir ect cur r ent . Th is means that

th e polar i ty does not r ever se per i odical ly.

2. The magni tu de is not usual ly constan t an d in some cases fall s repeatedl y to

low val ues after wh ich it clim bs back up to higher valu es. Th is means that th evolt age cont ai ns both a-c and d-c component s even t hough th e polar i ty d oes not

reverse.

3. Th e maxim um magni tu de obser ved by oscil loscope was about 250 volt s peak .

4. The r ate of r ise of shaft voltage was often i n t he range of 200 volt s per 1/ 60

sec. or 12,000 vol ts per second .

5. The volt age decay wh en fa ll in g to zero is l ess th an 0.1 m il l isecond .

6. The mi ni mum m agni tu de obser ved w as a few t ent hs of a volt.

7. Typi cal m agni tu des wer e betw een 30 an d 100 volt s peak val ue.

8. The shaft polar it y was posit ive on m any t ur bin es and negati ve on fewer tu r-

bines.

9. The potent ial at a ny i nstant is essent ial ly t he same anywh er e along t he tur-

bin e or generat or shaft . Th e shaft volt age appears betw een shaf t an d bedp la te

whi ch i s grounded.

10. Th e maximum cur r ent obser ved i n a r esistance cir cuit conn ected between

shaft and ground , regard less of h ow smal l th e magn it ud e of resistan ce, was

approxim ately 1 mi l l iam p.

Sound engineering conclusions are timeless. This technical summary by

Gruber and Hansen is as applicable now as it was in 1958. Certainly this list

could be modified to reflect some different measurements, or different machines,

but t he funda menta l concepts, descriptions a nd cha ra cterizat ions of the phenom-

ena are st ill the same. It is comforting to note that physical principles remain

consta nt , an d tha t our understa nding of many of these physical principles has a

tendency to grow wit h improved technology, measurement s, a nd commu nicat ion.

Finally, the role of the seven-step problem solving approach introduced atthe beginning of this chapter should be reviewed. In this case, the diagnostic

objectives were clea rly directed a t st opping t he coupling end compressor bea ring

failures. Secondary objectives included an understanding of the failure mecha-

nism, plus the issue of long-term reliability of the refrigeration train. Due to the

consistent a nd repetit ive nat ure of the failures, the identifica tion and st at ement

of th e diagnostic objectives were quit e stra ightforward.

Mecha nical inspection provided no significant informa tion during t he ini-

tial failures, since the physical evidence was destroyed by the occurring failure

mecha nism. However, one of the key elements in th e fi na l detection an d a na lysis

of this problem evolved from the observation of the frosted

bear ings by an OEM

service representat ive. If t ha t individual ha d not encountered similar failures on

electric machinery, the series of failures could have continued for several more

iterations.The test plan development for investigation of these failures was confined

to tradit ional process variat ions, combined with steady state analysis of shaft

an d casing vibration response chara cterist ics. This deta iled syst em a nd vibra tion

analysis provided lit t le insight into the actual failure mechanism, and the test


15/28


plan evolved into a simple monitoring activity. Hence, during the final two fail-

ures the data acquisit ion program consisted of daily measurements of shaft

vibration a nd posit ion.

The posit ion da ta wa s obtained by measurement of the proximity probe DCgap voltages in both ra dial a nd a xial directions. At t he t ime of these failures, the

machinery train was equipped with a full array of X-Y and thrust proximity

probes. However, Pr oximitors an d monitors were not insta lled, and the mea -

surements were obtained with a portable data collection box. The information

from t hese daily rea dings wa s subjected to the simple data processing technique

of manual trend plots. Although these plots did not directly solve the problem,

th ey certa inly did identify t he existence of th e a bnorma lity (e.g., Fig. 14-2).

Init ial data interpretation of this malfunction was subjected to various

degrees of speculation. Although the shaft sinki ng in to the bear in g descr ipt ion

was acceptable to most of the engineering personnel, the fundamental cause for

this behavior remained a mystery. The full interpretat ion of this trend da ta wa s

fi na lly achieved by the previously mentioned OEM representa tive wh o had expe-

rienced electrostatic discharge problems on condensing steam turbines within

the power generation industry. The conclusions and recommendations were to

install a rotor grounding system to remove the electrostatic shaft voltage. The

corrective a ction pla n consisted of a temporar y inst alla t ion to verify th e hypothe-

sis, followed by a n improved design for long-term operat ion an d reliabilit y.

The a ppropriateness of this grounding brush solution w as demonstra ted by

several years of good operation of this machinery train. Although there were

some problems with the compressor bearing design, the shaft voltage problem

appeared to be under control. The only failure that occurred was due to compla-

cency by t he loca l personnel. They did not repla ce a w orn out compressor ground-

ing brush, and the inboard compressor bearing began to migrate down into the

bottom pad. Many of the local management personnel had changed, and there

was a general insensit ivity towards this problem. Fortunately, one of themechanical engineers involved with the original problem diagnosis and solution

was st ill employed by the operating company. He reviewed the data and con-

vinced management of the reoccurrence of this problem. Reluctantly, they shut-

down the machinery train, and discovered that the compressor coupling end

bearing was failing in the manner previously described. This bearing was

replaced, both grounding brushes were replaced, and the train was returned to

normal operation.

Approximately eight years after the original plant startup a new wrinkle

into the shaft voltage problem was introduced. By this t ime, various modifica-

tions had been ma de to the origina l OEM bear ings and consultant s concluded

that the compressor was experiencing electromagnetic problems. In this type of

ma lfunction, a rotat ing ma gnetic field produces sha ft voltages tha t might exceed

the current carrying capability of the ground brushes. One of the popular opin-ions expressed wa s th at the electromagnetic problems originated from a wheel to

diaphragm rub experienced during the OEM shop testing of the compressor. Of

course, none of the founders of that opinion were associated with this machine

during the shop test. However, speculation was somehow converted into an engi-


16/28

758 Chapter-14

neering conclusion. The reality of the situation was that a wheel rub occurred

during a shop performance test by the OEM on a different first stage wheel

design. This test was conducted with OEM compressor internals that were not

part of the contract. Furthermore, after the occurrence of the rub, the contractcasing wa s th oroughly checked for residual m agnetism, a nd none wa s found.

A machine shutdown and disassembly following nine years of operation

revealed residual magnetic fields on some elements of the compressor, turbine,

and piping. Since these localized magnetic fields did not originate during the

OEM shop tests, another source should be considered. For instance, improper

welding and grounding practices on the compressor desk would be more suspect

than any possible influence from a ten-year-old shop test rub.

In the final analysis, there are at least three dist inct lessons to be learned

from this case history. First, electrostatic shaft voltages are possible on machine

trains driven by condensing steam turbines. These electrostatic shaft voltages

a re cont rollable by isolation a nd/or grounding of the rota ting a ssemblies. Sec-

ondly, electromagnetic shaft voltages can appear when rotating magnetic fields

are generated by magnetized machine elements. These electromagnetic shaft

voltages can only be eliminated by degaussing the magnetized parts. Thirdly,

during th e investigation of any ma chinery ma lfunction, if the da ta becomes cor-

rupted by idle specula tion a nd/or conjectur e, th e fi na l results will n ot be accu-

rate, and the conclusions will often be misleading.

Case History 51: Barrel Compressor Fluidic Excitation

During the design of machinery for a new chemical plant or refinery, the

OEM is faced with providing a cost-effective offering to satisfy the process

requirements of the end user. Fluid para meters such as fl ow ra tes, temperatur es,

pressures, and gas compositions are used to define an operating envelope for the

machinery. In most instances the fluid properties are well defined (e.g., air, pro-pylene, ammonia, etc.). However, there are times when a new plant process may

result in a fluid stream with properties that are not fully quantified. Although

pilot plants and process computer simulations provide a wealth of knowledge,

the fi na l strea m composit ions ma y not be known un til the full-scale plan t is built

and placed in operation. As a case in point, consider the following example of a

prototype urea plant and the main centrifugal machinery train handling a mix-

tur e of carbon monoxide, carbon dioxide, plus tr ace qua ntit ies of hydroca rbons.

The machinery train in question consists of an extraction steam turbine

driving tw o barrel compressors. The fi rst tw o process st ages a re conta ined wit hin

the low sta ge compressor. The third a nd fourth process sta ges are ha ndled wit hin

the high sta ge compressor. Ea ch compression st age w as equipped w ith int ersta ge

cooling and knockout drums. Configuration of the high stage rotor is presented

in F ig. 14-6. The t en impellers (five per compression st a ge) are located ba ck toback to minimize thrust load. This compressor was equipped with five pad, tilt

pad ra dial bearings, and a tra dit ional thrust ar ra ngement. The rotor exhibited a

fi rst critical of 5,100 RP M, combined wit h a design speed of 9,290 RP M.

During init ial field startup with process gas, the high stage compressor


17/28


wrecked. The general opinion was that the system was under damped, and the

OEM designed a nd retrofi tt ed a set of squeeze fi lm da mper bearings. This modi-

fication consists of a machined annulus between the casing and the bearinghousing t ha t is fi lled w ith lube oil. The ends of the a nnulus a re generally sealed

with O-Rings, and it a llows t he bear ing housing to float wit hin the confi nes of the

O-Rings a nd t he oil fi lm. The ad ditiona l da mping provided by this cha nge proved

successful, and the unit wa s easily brought up t o norma l operating speeds.

It appeared tha t t he ma jority of the problems w ere over, and t he remainder

Fig. 146 Ten StageRotor Configuration ForHigh Stage CO

2

BarrelCompressor

Fig. 147 Initial CouplingEnd Shaft Vibration OnHigh Stage Barrel Com-pressor


18/28

760 Chapter-14

of the plant startup could proceed. However, during loading of the machine,

vibration levels increased dramatically. At discharge pressures around 1,500 Psi,

a low frequency excitat ion appeared. As discharge pressure was increased, the

magnitude of the low frequency vibration would likewise increase. Typical shaftresponse for this unit is shown in Fig. 14-7. Rotational speed vibration ampli-

tudes were approximately 1.0 Mil,

p-p

, and the majority of the motion occurred at

a frequency of 550 CPM. This motion was strongest on the coupling end bearing

an d somewhat lower on the outboard bearing. The shaft orbit w as nearly circular

at 550 CP M, and a series of inside loops at running frequency were noted. Un der

this set of operating conditions for every one cycle of the low frequency excita-

tion, the rotor made almost seventeen revolutions.

This low frequency vibration was audible on the compressor deck and the

piping. In fact, the unit sounded more like a reciprocating engine than a centrif-

ugal compressor. One of the folks on the job claimed that it sounded like an

arm adillo running loose in the piping. A test program wa s esta blished to run the

compressor durin g va rious condit ions to cha ra cterize the vibra tion response. The

results of these init ial tests are summ arized in th e following list :

1.

The low frequency vibra tion wa s present a t discha rge pressures a bove

1,500 Psi, an d did not exist a t lower discha rge pressures.

2.

The low frequency vibration initially appeared at 660 CPM. As discharge

pressure and t emperat ure increased, it shifted t o a frequency of 550 CP M.

3.

This frequency w as measura ble on t he structure a nd t he a ssociat ed piping.

4.

The support s tructure reduced th is motion a nd t he founda tion wa s quiet .

5.

During st art up and shut down, the rotor system wa s well-behaved, an d

passed through its first crit ical in a normal and repeata ble fashion.

6.

There were no reciprocat ing pumps or compressors in t he immedia te a rea

that could influence this centrifugal compressor.

This initial test program confirmed the presence of this low frequency

vibration, and t ied down the fact that it was strongly related to discharge pres-

sure. These tests d id not provide a ny subs ta nt ial clues as t o the origin of the low

frequency vibra tion. One suspicion w as tha t t here might be a n excita t ion origi-

nating from the fourth stage aftercooler. During one of many shutdowns, the

exchanger head was pulled, and a bundle of corrosion coupons approximately

twelve inches long were discovered. Well, that was the proverbial armadillo in

the piping system. The coupons were removed and the machine restarted. At

1,500 Psi the low frequency component reappeared, and no further mention was

ma de to the piping arm a dillo. Throughout t he initia l tests, th e compressor fourth

stage discharge temperature was considerably lower than the process design

predicted. It w as clear t ha t t he fi na l aft er cooler wa s not necessar y. Eventua lly,

this heat exchanger was completely removed and replaced with a spool piece.

Again, there was no influence on the low frequency motion.

Following the death of the armadillo theory, one of the process engineers

suggested filtering the shaft vibration signals to eliminate the low frequency


19/28


vibration component. This approach began to gain popular approval and many

individuals considered this electronic filter approach to be fully acceptable. The

logic behind this approach revolved around the concept that the majority of the

vibration occurred at low frequencies, and

ever yone kn ows tha t l ow fr equenciesar e l ess destr ucti ve th an h i gh fr equencies

. Not only is t he a pplica tion of this con-

cept wrong, the argument is further compromised by the relative amplitudes.

Specifically, the total shaft motion was very close to the overall clearance of the

bearing a ssembly. Hence, an y m inor excita t ion could easily d rive the sha ft int o

meta l to meta l contact with the bear ing pads, with significant damage to the

entire machine. Fortunately, the plant manager was a reasonable and experi-

enced engineer who recognized that the problem must be solved by eliminating

the excita t ion, an d not by ma sking the crit ical m easurements.

After several other blind alleys, a test plan was established to make a

series of pressure pulsat ion m easurements t hroughout the high sta ge compres-

sor piping system. From external vibration measurements on the outer bore of

the piping it was known that the low frequency component was present. How-

ever, it was not totally defined whether the excitat ion was transmitted through

the fl uid strea m or th rough the st ructure. To answ er this q uestion pressure pul-

sation measurements were employed. Fig. 14-8 shows the approximate location

of test points

through

corresponding to the casing drain at the third stage

suction th rough the fourth sta ge discha rge. During t his test , the compressor w as

sha king at a low frequency of 550 CP M. The third st age mea surements did not

display this low frequency component, and neither did the suction of the fourth

sta ge. However, the fourth sta ge discha rge casing dra in

, and the fourth stage

discharge piping

through

revealed a clearly defined pressure pulsation atthe elusive frequency of 550 CPM. Plotting dynamic pressure pulsation ampli-

tudes at this frequency versus distance from the fourth stage discharge flange

produced the summarized data shown in Fig. 14-9.

From this t est result , it w as concluded tha t t he excit ing force wa s interna l

Fig. 148 Physical Loca-tion Of Pressure PulsationMeasurement Points

Aftercooler

4th Stage Discharge

4th Stg.Suction

3rd Stg.Disch.

3rd Stg.Suction

4thProcessStage

3rdProcessStage To

REACTOR

Coupled toLow PressureCompressor

Casing Drains


20/28

762 Chapter-14

to the compressor, and probably located between the fourth stage suction anddischa rge fla nge. The next st ep was t o visit the OE M tha t designed and ma nufac-

tur ed the compressor. The ensuing m eeting wa s very fra nk a nd open. After m a ny

hours, it was agreed that the rotating assembly could not generate the type of

frequency characterist ics measured. Since everything appeared normal on the

rotor, the next step wa s to begin exam ining the entire dra wing set for a nything

that was different from tradit ional design. Again this appeared to be gett ing

nowhere, when someone mentioned the balance port at both discharge nozzles.

The bala nce port confi gurat ion consisted of a r ecta ngular passa ge tha t con-

nected one side of the discharge nozzle with the opposite side of the nozzle, as

shown in Fig. 14-10. This passage was an integral part of the barrel and was

included as a means to pressure balance each discharge nozzle. Both the third

an d the fourth sta ge nozzles w ere similarly fa bricat ed. From this meeting, it w as

decided to plug both end s of ea ch bala nce port , han d grind t he surfa ce in cont ourwith the nozzle, and drill a weep hole in one side. Following return to the plant

Fig. 149 Dynamic Pres-sure Pulsation VersusDischarge Pipe Length

Fig. 1410 Cross SectionThrough High Stage Com-pressor Discharge Nozzle

E

E

E

E

E

E

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

-20 -10 0 10 20 30 40 50 60 70

Pressure

Pulsation

(PSI,p-p)

Distance from Discharge Flange (Feet)

ExternalInternal

Pressure Pulsationat 550 Cycles/Minute

Balance Port

Barrel

Shaft

Casing


21/28


site, it wa s confi rmed tha t a bala nce port did exist in each discharge. B y running

a piece of wire through the hole on one side, it was verified that the balance port

was a circumferential passage that connected one side of the discharge nozzle

w ith t he opposite side of the sam e nozzle. Following th e agreed upon action plan,the plugs were insta lled, and th e machinery train resta rted.

During t he ensuing sta rtup, the compressor came up smoothly through the

fi rst crit ical, a nd rea ched norma l operating speed with out a ny diffi culty. As load

was gradually increased, the compressor reached 1,500, and then 1,600 Psi,

with out the low frequency excita t ion. Finally, the ma chine wa s fully loaded, and

the previously encountered low frequency excitation at 550 or 660 CPM did not

appear. The shaft and casing vibration, plus the pressure pulsation measure-

ment s w ere void of th e previous low freq uency component . Specifica lly, Fig. 14-11

is representa tive of the fi na l condition of this m achine at a speed of 9,340 RP M,

operating at a discharge pressure of 2,150 Psi.

The running speed motion had decreased to 0.13 Mils,p-p vertically, and

0.09 Mils,p-p horizontally. This was due to refinement of rotor balance during thecourse of this project. There are two minor components at 900 and 2,580 CPM

tha t a ppeared a t am plitudes of 0.04, and 0.06 Mils,p-p respectively. These compo-

nents were considered to be nondestructive, and were later found to be associ-

ated with the st iffness of the O-rings installed with the squeeze film dampers

(ba sed on OEM comments a nd recommenda tions). H owever, the previously dom-

inant low frequency excita t ion at 550 or 660 CP M wa s no longer visible at an y

operating condition. Although the exact origin of the low frequency was not

totally defi ned, plugging of the ba lance ports certa inly elimina ted th e excita t ion.

Fig. 1411 Final CouplingEnd Shaft Vibration


22/28

764 Chapter-14

As part of the overall testing program, X-Y shaft proximity probes were

installed on the floating squeeze film bearing housing to observe shaft motion

relative to the squeeze film bearing housing. In addit ion, standard X-Y shaft

probes were mounted on the case a nd provided a measurement of shaft motionwit h respect t o the cas ing. The 1X runnin g speed vectors from both set s of probes

are presented on Fig. 14-12. This vector plot displays relative shaft motion with

respect to the case and with respect to the bearing. Clearly, the vector difference

between the tw o relat ive measur ements must be the motion of the squeeze film

bearing housing with respect to the case. Since relative motion existed, it was

concluded that the squeeze film housing was moving and providing some mea-

sure of addit iona l dam ping.

In r etrospect, this problem wa s tempora rily considered to be in the realm of

a semi-magical solution. That is, the low frequency vibration and pressure pulsa-

tion in the fourth stage discharge were eliminated. However, it was difficult to

verify th e na tur e of the forcing function. This w a s prima rily due t o the confusion

tha t existed as to the physical properties of the process gas at high pressures.

Sometime after the conclusion of this project , it was determined that the

balance port passage acted as an acoustic resonator. Based upon the passage

length, and the velocity of sound at specific operating conditions, the fundamen-

ta l a coustic resona nt frequency w as one half of the measur ed low frequency exci-

tat ion. It was concluded that the occurring frequency was in all probability the

first overtone of the fundamental acoustic resonant frequency. In addit ion, the

shift from 660 to 550 CP M wa s a ttr ibuted to the fact tha t t he 660 CP M wa s pro-duced in the t hird sta ge discha rge, an d the 550 CP M excita t ion wa s generated a t

the fourth stage nozzle. Hence, during loading, the acoustic resonance shifted

nozzles as a function of process flu id properties.

Again, the role of the seven-step problem solving approach introduced at

Fig. 1412 RelativeMotion Of Coupling End

Bearing Housing


23/28


the beginning of this chapter should be examined. In this case, the diagnostic

objectives w ere prima rily directed a t r educing an d/or stopping the low freq uency

compressor vibration. Secondary objectives included an understanding of the

excitat ion mechanism, plus the client acceptance of this new machinery train.Mechanical inspection provided no useful data during the initial phases of inves-

tigation. However, once the design drawings revealed the potential presence of

the circumferential bala nce port at the discharge n ozzle, fi eld inspection verifi ed

th e existence of th is cavit y. Hence, on-site mecha nical in spection play ed a role in

the identifi cation of this problem.

The test plan development for investigation of this low frequency motion

included a full a rray of speed a nd load tests combined with measurement of shaft

vibration response characterist ics. This detailed analysis identified the direct

relationship between the appearance of the low frequency vibration component,

an d specific opera ting conditions of pressure and temperature. The test plan w as

then expanded to include structural and piping vibration measurements, plus

pressure pulsation measurements t hroughout t he system.

Field data acquisit ion was conducted with a tradit ional set of diagnostic

and recording instrumentation. The internal shaft sensing proximity probes

mounted on the compressors were recorded simultaneously with the casing,

structura l, and piping vibrat ion t ra nsducers. These externa l measur ements w ere

ma de with velocity coils at ta ched w ith h igh strength ma gnets. The pressure pul-

sation measurements acquired around the high pressure compressor and the

process piping w ere ma de w ith piezoelectr ic tra nsducers. These pressure probes

were a tta ched to existing dra in or vent va lves. There wa s minima l opportunity to

obta in add it iona l test points due t o the piping metallurgy a nd th e reluctance to

add additional pipe connections.

At high stage discharge pressures below 1,500 Psi, the mechanical system

was void of the low frequency vibration component. Hence, extended compari-

sons between data sets was not required, and the data processing efforts wereconcentra ted on a detailed documenta tion of the dyna mic measurements during

a condition when the low frequency perturbation was active. Since the subsyn-

chronous motion was present only at full loads, the database was limited to

steady sta te formats of orbits, t ime base, an d spectrum plots. This da ta wa s gen-

erally summarized on traditional plots such as Fig. 14-7.

Da ta interpreta tion revealed a clear a ssociat ion between the low frequency

shaft vibration, and the low frequency pressure pulsation. This excitation

appeared to originate w ithin t he fourth sta ge of the high st age compressor. Meet-

ings wit h t he OEM revealed nothing unusual w ithin th e compressor bundle, an d

the discharge nozzle balance port was suspected. The project conclusions cen-

tered around the potential existence of an acoustic resonator, and the recommen-

dation was to plug the balance port , and rerun the previous field tests. This

conclusion proved t o be correct, a nd t he corrective a ction plan consist ed of simplyrestating that the circumferential discharge nozzle balance port should not be

exposed to the process fluid. Although the full technical explanation for this

behavior la gged the fi eld correction by severa l months, the plan w as fi eld proven

to be t he proper solution.


24/28

766 Chapter-14

Case History 52: High Speed Pinion Instability

The fi na l case history in t his book deals w ith a four poster high speed a ir

compressor. This t ype of ma chine is also referred to a s a n int egra l gear, or a mul-tiple casing compressor. In this particular unit, a central bull gear is driven by a

directly coupled motor. The bull gear in t urn d rives four high speed pinions th a t

are m ounted a t 90 increments a round the bull gear. Ea ch pinion operat es in a

separate volute casing, and intercooling plus liquid knockout is provided

between each sta ge. The unit un der consideration wa s a new design ba sed upon

good experience with similar, but smaller machines. Whereas the previous case

history dealt wit h unknown fl uid properties, this current problem is as sociat ed

with the capa city upgrade of a successful design. Although the dimensions were

conservatively expanded, the upgraded unit experienced problems. For the pur-

pose of this discussion, the behavior of the fourth stage pinion assembly will be

considered. The rotor confi gura tion for t his a ssembly is sh own in F ig. 14-13. As

expected, the rotor opera tes in a pivota l ma nner a s illustrat ed in the mode shape

at the bottom of this diagra m.The motor driven bull gear for this machine runs at 1,200 RPM, and the

fourth sta ge opera tes a t 17,770 RP M. Following init ia l sta rtup of the unit , a sig-

nifi cant am ount of subsynchronous motion w as detected. The vibrat ory behavior

under ty pica l opera ting condit ions is present ed on Fig. 14-14. The t wo F FT plots

represent avera ged dat a on the bott om, and peak h old da ta on the top plot . B oth

sets of information were obtained during an identical time period of eight sec-

Fig. 1413 Overhung Pinion Configuration Fig. 1414 Initial Steady State Vibration


25/28


onds. From this init ial information, it is apparent that the subsynchronous

vibration covers a broad band (6,000 to 15,000 CP M), wit h ident ifi a ble compo-

nents at 8,600 and 8,800 CPM, plus another discrete frequency at 12,100 CPM.

In pa ssing, it should be mentioned tha t a veraged dat a can successfully hideshort duration peaks of any frequency. Hence, this type of information should

always be compared against peak response in the t ime and the frequency

domains. This step will verify that a realist ic sample has been averaged, and

tha t t he proper a mplitudes are presented in the frequency doma in plot .

Furt her investigat ion revealed that the subsynchronous vibrat ion var ied as

a function of compressor discha rge pressure. As sh own in F ig. 14-15, the r unnin g

speed (1X) vibration remained fairly constant with increasing pressure. How-

ever, the subsynchronous bands increased as the unit was loaded up to normal

opera ting conditions.

The va riable speed coastdown of the init ial fourth sta ge pinion configura -

tion is presented in Fig. 14-16. This Bode plot displays normal response charac-

terist ics, with a pivotal crit ical speed at 12,000 RPM, and reasonable

amplification factors. This resonant speed coincides with the measured steady

state frequency at 12,100 CPM observed during full speed operation. It was con-

cluded that the major subsynchronous component at 12,000 CPM was a re-exci-

tation of the pinion balance resonance. The OEM then performed a re-audit onthe rotor design and developed a series of modifications that included stiffening

the overhung a ssembly to raise the pinion na tura l frequency. Ph ysically t his wa s

accomplished by selectively increasing the outer diameters of the overhung por-

tion of the pinion. Following fa bricat ion a nd insta llat ion of the new fourth sta ge

Fig. 1415 Initial Shaft VibrationResponse Versus Discharge Pressure

Fig. 1416 Initial Bode Plot Of CoastdownBehavior Before Modification


26/28

768 Chapter-14

pinion, the compressor exhibited t he full loa d vibra tion da ta in Fig. 14-17.

The shaft vibratory characteristics following modification were consider-

ably improved, with peak amplitudes of 0.3 Mils,p-p at 10,600 CPM, and 0.25

Mils,p-p at 13,900 RPM. Rotat iona l speed motion had increased somewha t from

0.56 to 0.67 Mils,p-p. In addition, rotor sensitivity to subsynchronous motion as a

function of dischar ge pressure ha d decreased, a s exhibited on t he Fig. 14-18.

The pinion coa std own following modifi cat ion is present ed on Fig. 14-19.

This Bode plot displays a measured pivotal resonance at 13,600 RPM, which is

13% higher tha n the init ial design. The amplificat ion factor has a pparently

increased from Fig. 14-14. The OEM a nt icipa ted a lower am plifica tion fa ctor due

to a computed increase in dam ping for t he modified fourth sta ge pinion. Superfi -

cially, the transient vibration response data indicated a slight increase in the

amplification factor, which would be associated with decreased damping. How-

ever, the amplification factor calculation is somewhat compromised by the fact

tha t a gear dr iven rotor (pinion) is involved, an d minor changes in the herring-

bone gear conta ct may result in a ppreciable changes to the sha ft motion. In a ddi-tion, the vibration amplitudes are quite small, and any variat ion may appear as

a large change in t he fi na l a mplification factor. When a ll factors are considered

and appropriately weighed, the final variable speed behavior of the modified

fourth stage pinion is acceptable.

In order to complete the project, the fourth stage pinion was field balanced,

Fig. 1417 Steady State Shaft VibrationAfter Pinion Modification

Fig. 1418 Shaft Vibration Versus Dis-charge Pressure After Pinion Modification


27/28


an d run ning speed a mplitudes were reduced to slightly more tha n 0.2 Mils,p-p.

As expected, this balance improvement had minimal effect on the broad band

subsynchronous motion, as shown in Fig. 14-20. Finally, comparison of Fig. 14-20

with the init ial da ta shown in the fi rst st eady st at e plot of Fig. 14-14 reveals t he

substantial improvement obtained by the rotor redesign and trim balance.

Although minor subsynchronous activity persists, the maximum am plitudes a re

quite small. At this stage, the machine was given a clean bill of health by the

OEM and it was considered to be acceptable to the end user.

During this project, the OEM was responsible for correcting the abnormal

vibra tory beha vior. The ba sic diagn ostic objectives were directed a t r educing the

subsynchronous instability and providing a machine with acceptable vibration

amplitudes. Mechanical inspection was used to verify that all compressor ele-

ments were in agreement with the design drawings and associated tolerances.

The developed test plan consisted of shaft vibration data acquisit ion during

steady state conditions at variable loads, plus transient startup and coastdown

information. Data processing consisted of standard formats for the measured

vibration response data. This hard copy data was interpreted as a lower than

desired pivotal resonance for the fourth stage pinion, plus excitation of this criti-

cal speed under full load conditions. Based on this information, the OEM per-

formed an audit of the design rotor response calculations and suggested the

fa bricat ion of a stiffer overhung pinion a ssembly. This conclusion an d a ssocia tedrecommendations were accepted by the end user

This modified pinion a ssembly was tested in a ma nner similar t o the origi-

nal fourth stage rotor design. The superiority of the modified design was demon-

strated under all operating conditions. Hence, the final corrective action plan

Fig. 1419 Normal Bode Plot Of Coast-down After Pinion Modification

Fig. 1420 Steady State Shaft VibrationAfter Pinion Modification And Trim Balance


28/28

chapter 14 diagnostics

Documents