Imp - O & M Using Condition Monitoring T22101-112
Imp - O & M Using Condition Monitoring T22101-112
Imp - O & M Using Condition Monitoring T22101-112
by
Meherwan P. Boyce
President and CEO
George A. Gabriles
Vice President
Cyrus B. Meher-Homji
Director, Research and Engineering
A.N. Lakshminarasimha
Senior Development Engineer
and
Feroze J. Meher-Homji
Senior Design Engineer
Houston, Texas
101
102 PROCEEDINGS OF THE TWENTY-SECOND TURBOMACHINERY SYMPOSIUM
365 hp, externallyfired, steam injected gas turbine, developed for the and hence directly deals with economics of operation. Perfor
U.S. Department of Energy. His areas of interest are aerothermody mance monitoring is useful in detecting compressor, combustor,
namics of gas turbines, rotordynamics, vibration analysis, and and turbine malfunctions. Vibration monitoring and bearing tem
knowledge engineering for expert systems. perature analyses on the other hand are useful to evaluate the
Mr. Meher-Homji has a B.S. degree in Mechanical Engineering mechanical health of the machine. Bearing failures, rotor imbal
from Shivaji University, an M.E. degreefrom Texas A&M Universi ances, etc., fall into this category.
ty, and an M.B.A. from the University of Houston. Traditionally, condition monitoring system were used for safe
He is a member of ASME, a registered Professional Engineer in equipment operation and to prevent equipment failure. New sen
the State of Texas, and has several publications in the area of sors, instrumentation, and enhanced capability of computers com
turbomachinery engineering. bined with economic pressures have introduced an additional
application: operating equipment at its maximum efficiency. A
key issue is how one retains performance and limits degradation.
Some important factors that are of relevance to condition monitor
ABSTRACT ing are:
With exceedingly high downtime costs and the need for efficient • Fuel Costs- Fuel costs constitute a large part of the total gas
operation of turbomachinery, integrated condition monitoring, turbine life cycle costs. The annual fuel cost for a 25 MW gas
wherein a number of health parameters are analyzed, is becoming turbine is between seven and eight million dollars. Several forward
increasingly popular in process plants and in utilities. Most oper looking corporations now demand that condition monitoring sys
ational problems can be diagnosed by developing a correlation tems be used for performance degradation control.
among several key operating parameters. A wide range of condi
Availability- This is a strong function of system design, fuel
tion monitoring approaches are available and this paper shows
•
Process and utility industries routinely diagnose operational A review of the major engine health monitoring techniques used
problems, prevent equipment malfunction or failure, determine are presented here. Available techniques and details on their
optimum equipment operating conditions, schedule maintenance, implementation and integration are provided in Table 1 [1].
and repair or replace defective parts based on the information The choice of the condition monitoring philosophy (on line vs
obtained from condition monitoring systems. A recent survey by off line), particular approach and the diagnostic technique should
Electric Power Research Institute (EPRI) has indicated that the use be based on specific plant operational objectives, location of the
of diagnostic monitoring system in the utility industry alone would machine (offshore, unmanned operation, etc.), criticality of ma
enhance plant availability by two percent which in monitory terms chines, and the failure modes experienced. As implementation and
translates to $400 million per year in US alone. If process indus acceptance of a condition monitoring system is a key issue, plant
tries are included, an estimated ten fold increase in savings could operational practices and maintenance philosophy must be consid
be obtained. ered at the early stages of a condition monitoring project.
A variety of condition monitoring systems are in use in industry,
Performance Analysis
each with a specific application. They can be broadly classified as
vibration and acoustic, aerothermal performance, and oil and Modem turbines and compressors are monitored comprehen
debris monitoring systems. The choice of a monitoring system is sively for control and protection purposes. In fact, there is now a
based on investment and payback considerations. Thus, there are convergence between control systems, protection systems and
hand held, micro, and mainframe computer based monitoring condition monitoring systems. Interfacing (or combining) a condi
systems each for a different level of sophistication and investment. tion monitoring system with a modem control/protection system is
Among the portable and microcomputer range versions there are often only a matter of having an RS 232 or similar connection. This
portable vibration signature collectors, lubricant oil sampling means that most of the information required for aerothermal
devices, acoustic leak detectors, thermography, etc. A combina analysis is readily available. Some machines may require the
tion of these portable units are also in use; vibration and oil addition of some sensors for comprehensive aerothermal perfor
analysis are combined to monitor bearing condition. Acoustic and mance analysis. Several gas turbine operators are installing torque
thermogaraphic units have been useful in detecting valve leakages, couplings on compressor and pump drives. Torque meters are now
identifying hot spots on boilers, valves, etc. While these units are quite reliable and have accuracies of better than 0. 75 percent. They
for specific use, a comprehensive overall condition monitoring can often give indications of surge and torsional vibrations and
system utilizes both the vibration and performances data to report provide valuable information from a condition monitoring
accurately the current plant status, to diagnose any malfunction or standpoint.
to predict the future condition of the plant. The aerothermal performance of a gas turbine provides valuable
Due to the complexity of critical turbomachinery operation, a insight into its operating condition. It is important to integrate such
comprehensive condition monitoring system should use both steady a system with vibration analysis as several vibration problems are
state and transient data. Vibration and performance data are used manifestations of underlying aerothermal problems. Some prob- ·
to accurately estimate engine condition. The main reason for lems that can be detected/solved by an integrated condition mon
utilizing both performance and vibration data is to distinguish itoring approach include rotating stall in axial flow compressors
between mechanical and aerodynamic induced vibration signa and in centrifugal compressors (both in inducers and diffusers),
ture. Performance monitoring becomes especially crucial in eval rotor bows due to rapid temperature ramping, distortion or fouling
uating performance retention or degradation rate of a component related surge events (intake distortion), and plugged nozzles.
CASE STUDIES IN TURBOMACHINERY OPERATION AND MAINTENANCE USING CONDITION MONITORING 103
VJBRATION
ANALYSIS
DYNAMIC Wouldn:quifcon·linc
PRESSURE moniiCB'in&becausr;
ANALYSIS problems under
conlidenuion��revr:ry
errntic andCMOCCur
VtrJrapidly.
�· ==o��!c����=�:!r
probk:mislbetb:ne botwcc nsamplin&andan;alysb.
�
MONn'OIUNO ��J=C:��inJin:=::bere Clnbcpctfonu:di n
cilheranofl'llncmocle
Oeu:rionlioni n
perl'QIIIIIIICC.The
OW.canbc�in
tam� of tabUlatioN,
AND GAS PATH u:rk$Wion�ofa,m,d.ynllmic probletm. We incl�
lhi:& head or on·line.
CXICI!tiOwhich faults
ami501alablcl08pecifio
--�
Nonnalhalion and
under
ANALYSIS m:llds.Tiendsrarc
TECHNIQUES
����=�===�:o(�Tcl�
i:::r=r;:�=��u:���:a=��· ccm:ctionofdltais
i�Dpcmamfor
annin&ful calcuiiU:lons.
componentsdepend� on
lhcsophistl�cion ofdata
aMiy:iiJ,
....... � ....
=--.,r.:""
inmlroen�otlkmin!e&ril)' IIOrDII]Ju.tlonisdoneto
ltdu!Jques.JelalodiO�tlll¥/ttMor. IUid��evmlpractical dilaimbweberween
lilcl.oninl:ludin; otf-dcsip cf'l'ccQ and
S)'mpiOm ma:;kingand dcriontion cft'eca.
dalaseattcrcffectS.
TriJIISitniQIIlo llyru f
���
of problem,n:ladn&IO
lhcholscction, swt
igniliansysu:mand
sevaalodwprobkms.
fi!RROORAPIIY
��=��=::nm��dctecs::s:m�of :::-�llr���tc
raoilllio n mimlscopc.
Aaalyslsbvb.
vil:nti�:.�.
Offh w..
£cmlJI'Ifii5.Skilled
lllliystl'flqU t mi lO
iolaJm data.Some
-·-
b:'t!:r�
Rcquli'Csh! c useofopdc
adensiU>mctero r h.i�h
n:solullon micro�ope.
FERROGRAPHY o Analy�\s IS VIII.
W=
vibr.lnon:�
. � ��:��(k
Offb.n�
�e�� ;!.,��=:�::J�!n=:su�in;n e farosroum. Sklllcd
!'N'YS!�Uired.IQ
Ul!apm dlltil. Some
�lltl:lmal�1ys�C�I�S
usiD§imagcanlllys!S
have been developed.
Often acomplcnx:nwy m:llniquc � with SOAP. Chip detector> ha.ve 0 1 Off·JiM. Rcqulle5 Wearinoilwc:�pill'lS. ViliualWjleelion.
IJ
���CH bc�ll.lperiodictrm!!'VakPicks.upparPde$1)'picaiof spallirt gand pcrlodicinspcc11011. U:;ewitllotllcrmdicator;iuchu
fn:mng. �I)' ptc k.� liJl m:lp!Ct.c . Oil monnOl' fihen will PICk. lip
0.W(COWJI)canJx
material �ded.l.af§cr
. .
VIbration.
non·femousdcbrisa�v.;cU. �timal!ocatlon of clupdc:tectoriso(zcn �!ncanbear.<llyltd
!r.ldcdoff foraccembilur. R1skofunptopt;rinstall�t1on(ornon VIa microscopic
i!l:>tal�tiO!lofplu!i)ls\ignillcamdue:toperiodli:Jela l'IOV h..lnllOductioo exarmna!'lon{orSEM).
offlliCfiiWnon cannep�Ceffccnvc �.
have been llpplioi to amx:ncincs ($«
As��maner.MCDs
od debris morutoringl
Look� 31 met.all:klbris wilh so:nson loc!UCd in. f:11c: scavangc lines. can be o n ·linc.
OILDEBRJS
App!l)!)1'141e forgearbox.n !U'Id for enjWia utiliZing rolhng clement
MONITORING
bcanngs. Some�cn'>OI'�h.svebecn dcvclopcd!hat are non!llttusJvc. Hu
COilllo lacial n·line
monnoringsystc!llli
W= ����U�:'su.
Paniclct�c:mbe
U;;c: wilh olhCfw ear indicaum
Applied 10 lifo:! pans !hat aresubject to LCF aud Ctcej). U �gc Datahsstobcgs!h!:n:d Dami!.gcwdisks,hot
Usual!yrcquire s dffign
USAGE t;justionsinvolvin'
MONITORING ���;���=�nt5��=ll=�r:t"
HOI!ICCticmc� poru:ntlifeisimponaru
by an on-line �yslml 10 scctioo.blildingsnd and9bouldbc:Jn�!Cdwilh01her
11'3.11sient�
l!CCO\In t for cngmc odlerlifed.paru. condltionmonnonngpllr.!Jnetersand
m:un�ma��;�gcmcntofmcrnfiengines. sue � >andcomplell
=ur.iions,cyclcsctc. offcou�.borescopeJnspco::nom. In
cmpiri<;,W'e��:perimt:nW
landbascdg,'IStllfblne�.somecs;omatcs
dauondamage
canbcmadc:busedon !hc numbcr of
assessment. More
swts, mps !!lid bv a�cumulanng engine
difficultwth! oombint:d
time at different EGTs. A hOI end
dzlmagc:modc$SUChas pyrornctctg ivcsvllluableinformalion
.
:;;:n�radgQe onho!sc:ctot nlen'll)l:nlnm:.
Further, this technology provides insight into how efficiently fuel this, it is difficult for industrial users to conduct any form of
is being utilized and thus, facilitates significant fuel savings if sophisticated usage monitoring.
degradation is controlled. The authors include within performance
analysis items such as EGT spread monitoring. This provides Optical Pyrometry
insight into hot section health. Actions such as this can significant By use of an optical pyrometer, itis possible to actually measure
ly extend hot section life. Excessive spreads can occur due to a the metal temperatures of the first stage nozzles and rotating blades
variety of reasons including excessive air leakages, blockage of in a gas turbine. It is possible to obtain profile data from such a
nozzles, and cracks in the combustor liner/transitions. sensor.
Transient Analysis Integration of Condition Monitoring Techniques
There is considerable work being done in the area of transient In order to plan maintenance for machinery problem rectifica
analysis relating to both performance and vibration data are pre tion, one requires good insight into the operating condition of the
sented [3, 4]. Significant condition monitoring information is machinery. With predictive maintenance, small incremental main
available by examining the profile of startup acceleration, coast tenance actions are used to delay the need for major maintenance
down times, EGT response during light off, and other transient intervention. For example, if ignored at an early stage, an increas
behavior. ing temperature spread in the combustion/turbine module may
Vibration Analysis lead to premature failure of the first stage nozzle or even turbine
blades. Maintenance action such as nozzle balancing can alleviate
Vibration is a good indicator of machine mechanical health. the problem. Sometimes, a combination of symptoms may be
With the correct choice of sensors and analysis techniques, vibra needed to pinpoint problems. A broken inlet guide vane mecha
tion analysis is an excellent condition monitoring tool It is further nism, may cause increasing vibration and loss of compressor
enhanced when used in conjunction with other condition monitor efficiency or possibly even surge.
ing techniques. Some turbine suppliers provide the minimum Blading vibration and failures are one of the inost complex
sensors (in terms of numbers, frequency ranges, etc.) with the main problems in gas turbines due to the complicated blade dynamics
objective of protecting the turbine frotp catastrophic failure. These and interaction of factors such as blade quality, environment (salt,
sensors are not always successful in meeting this minimum objec temperature), erosion/wear, and fatigue effects. An integrated
tive. Several manufacturers will provide one or two accelerome condition monitoring approach involving performance and vibra
ters or seismic probes, often filtered to cover only the unbalance tion monitoring can be of help here. While vibration and perfor
frequency (1 x rpm) . Thus, the operator will often have to add mance monitoring cannot predict blade failures, often the underlying
sensors to get the best information for a good maintenance strategy. causes (air flow distortion, surge, nozzle bowing/blockage, etc.)
Experienced troubleshooters will most often review the vibra can be detected, thus providing a chance to avoid the failure. The
tion data in conjunction with performance data to arrive at a "root use of performance and vibration monitoring for reduction of
cause" of a problem. blading problems has been described [4 ]. There has also been work
done in the area of using dynamic pressure to detect blading
Dynamic Pressure Analysis
problems.
The use of dynamic pressure transducers has worked well to
Diagnostic Approaches
detect certain blading instabilities and compressor instabilities.
This is an important facet of condition monitoring that has not Diagnostics have been traditionally based on fault matrices or
received much attention. fault trees. In the last decade, expert systems have become popular.
Some of the skepticism towards expert systems occurs because
Lube Oil Debris Analysis engineers believe that their long experience with machinery diag
nostics cannot be summed up in a few rules of inference, no matter
A number of methods are currently available. Several aeroengines
how powerful the inference machine. Expert systems generally
have magnetic chip detectors. Debris analysis has been most
imply a deterministic approach to machinery behavior. In reality,
valuable on gearboxes and engines having rolling element bear
chaotic rules are often more appropriate. A machine may run
ings. A wide range of debris analysis techniques are available
perfectly well at one set of conditions but may suffer seriouslyfrom
which can be both intrusive or nonintrusive.
a small change in these conditions. This is certainly true of some
Borescope Inspection high discharge pressure compressors. Expert systems are of use in
dealing with sub-problems such as trending, data validity check
This is an important and valuable condition monitoring tool. ing, and diagnostics. They are also valuable in integrating condi
(Borescope inspection can show up component cracks, erosion, tion monitoring data in order to obtain meaningful diagnostics. A
corrosion, and buckling.) It is usually carried out at fixed intervals review of possible roles for expert systems is made by Doel [5] .
dependent on the machine with a video camera being used to There has recently been considerable work done i n the area of
record results. Borescope inspections are usually very quick and the application of neural nets for monitoring and diagnostic appli
result in a minimum loss of turbine availability. It is important to cations [6, 7]. The training of a neural net may, however, require
have well trained personnel and clear cut procedures to ensure full a considerable number of faulty engines. Another computer relat
coverage of the critical components. By using a video camera to ed technological development is the use of hypermedia, which
record the inspection, one can enlist expert outside help to interpret could provide users with fast access to text and figures related to
the data. Eddy current checking is also done to detect cracks.
troubleshooting and maintenance of gas turbines.
Usage Monitoring For a new class of machines, tuning and modifications of
standardized machine train based fault matrix diagnostic proce
Experience has indicated that a mere time count of life limited dures, alarm and danger limits. etc., are necessary. Data for
parts is not effective. Life is strongly dependent on the manner in diagnostic tuning for new machines are typically obtained from
which the engine is used (EGT history, number of starts and trips) . condition monitoring systems [8]. In many cases, analysis proce
The algorithms to calculate life usage are typically proprietary and dures are used to simulate the effects of various faults on compo
require the knowledge of detailed design information. Because of nent performance models of newly introduced gas turbines, in
CASE STUDIES IN TURBOMACHINERY OPERATION AND MAINTENANCE USING CONDITION MONITORING 105
tics. This procedure has the advantage of modifying the baseline 0 9.J
New HP Rotor (from H-7 with 20,000 hrs1
�
7.7
CASE STUDIES
0 7.3
(..)
Maintenance and Overhaul of an Off-Shore MAJOR
6
'5eL- U-
-4 ��8- ���-2- --1__8---20-�U -40--��9�11�- ��-3-6 - --4�118
Typically, major overhaul on a gas turbine unit can take several D B B 6 B Z Z JZ B B
times override scheduled maintenance. Until a decision to sched Figure 2. Recovery in Axial Flow Compressor Pressure Ratio [1}.
ule an overhaul can be made, it is importarit to monitor the gas
turbine unit for safe operation. A reliable comprehensive condition
monitoring systems aids in safe gas turbine operation, till sched 7 n..-. 1991 18o46:14
PII'ELINE #3
uled overhaul is logistically feasible, or if monitoring data does not
, . e EKOFISK 214 T
indicate a distress. This case pertains to a gas turbine unit rated at
22.4 MW (ISO) driving a back-to-hack configured centrifugal •• 8
Alter
coinpressor [ 1]. •
lt .
0� y--
Before
The number 1 bearing of this turbine exhibited an increase in
vibration that was picked up on the condition monitoring system 8 � ,
.
and analyzed as unbalance. The machine had run for over 40,000 8 wvv --� 1-
hr since its last major inspection and the performance of both the
II
gas turbine and the pipeline compressor was poor, which was
ascertained using a condition monitoring system. A decision was 8
'
II
have foreign object damage (FOD) on the first eight stages. The INSPECTION
latter six stage compressor blading was severely eroded. The rotor •
a " ea ue ue 288 24 11 zae 3211 368 •8e
was sent to repair. A spare rotor taken from another gas turbine was DAYS
installed to replace the damaged compressor.
A trend is shown in Figure 1 in turbine ISO corrected horsepow Figure 3. Pipeline Compressor Flow Capacity [1].
er, which indicated a decline prior to overhaul. The recovery after
the overhaul is evident. A reduction in gas turbine compressor
pressure ratio is shown in Figure 2 with time and the improvement II 8
6 nor. 1991 lloSioiB
attained when the replacement rotor was installed. Flow and •
polytropic efficiency improvement obtained in pipeline compres
sor after the overhaul are shown in Figures 3 and 4. Both these 8
�'
-- -66%
e llen9r<'
...
---
6 Mer. I'" 15•05:48 se
1\
�
0
'
36
0 Q, 4 II
0 a:
e 32
�
v
.. 38
�
� �c-� I:
28 MAJOR
8
INSPE.;TION
• 24
,__ 8
!1:
0 (..)
D. 2e 8
-; 8 48 80 IZI! 168 2811 240 ZBII 320 360 41111
Q, 16 DAYS
-;
0
... 12 MAJOR Figure 4. Polytropic Efficiency of Pipeline Compressor [1].
!
u INSPECTION
!
� 4
in the vibration level at the compressor inboard bearing was also
II noted, as shown in Figure 5.
II 411 88 1211 168 Z8B 2411 ZBB JZD 3611 4118
Hours
A severe vibration of the gas turbine bearing number 1 was
noticed after post overhaul restart. Since spare compressor rotor
Figure 1. Turbine Corrected HP Before andAfter Major Inspection and turbine rotors were checked and well balanced prior to over
[1}. haul, the coupling was checked for any problems. It was found that
106 PROCEEDINGS OF THE TWENTY-SECOND TURBOMACHINERY SYMPOSIUM
i
..
110
110
-- ATI<IIRGJIR
?r-----�+--�
ATHBAG_Bl
� 140 -- AlHBRG_TR
ATHBRG_Tl
130
11BAG_ST
...
0 10 20 30 40
LOAD, IIW
(Jltt, � �11 1
3
Figure 6. Drop in Vibration Level With New Ace. Coupling [1]. (0)
; 2fT! 1411 1
V£-544lC : GT tl Brng ProKIIIIII\y (H)
Thu Qct I , 19911 JoBI:SS PM
�ot.or SPt .dl 5BS6 RPih
7
10 20 30 40
... 6 LOAD, IIIW
�
f.$ (d)
.. ...
f-CA E • 1 -- 11BRG_ST -
i • ... ---- 128RG_ST -
-- IIII!G_GEN-
--- .....,_GEN-
../
- f-4-- SPECTRAL PEAK I1XI DROPI'ED TO THIS LEVtL I�
....
1--. r...... I'
0 50 lOD ISO 2'00 Z50 300 3'i0 400 ASB 508 ISO
rREDUEHCY ltl Hz 0 10 20 30 40
LOAD, IIW
Figure 7. Vibration Spectra Showing High 1 X RPM Peak With Old
Coupling and Level Arrained Upon Installation of New One [1]. Figure 9. Variation of Parameters Vs. Load.
CASE STUDIES IN TURBOMACHINERY OPERATION AND MAINTENANCE USING CONDITION MONITORING 107
is causing the problem. In this case, the top and active side of the Through the accurate, real time, process data available with the
thrust bearing indicated distress. Since the thrust bearing was diagnostic computer, it was possible to access the aerodynamic
newly installed after a major overhaul, additional modifications condition of the machine and to recognize the impact that excessive
were found essential to operate the steam turbine at the rated load. balance piston leakage was having on the compressor's running
This included increasing the five lube oil groves in the upper half speed. Replacement of the balance piston seal along with other
section of the bearing by approximately 70 percent and increasing revamping enabled further debottlenecking of the unit to take
the metal gap on top active thrust bearing side by tapering the shim place.
as shown in Figure 10. By examining performance data on the machine, it was deter
mined that the higher than expected speed was the result of a
Use of Online Condition Monitoring for Debottle-Necking combination of reasons
of a Ethylene Refrigeration Compressor Train
• High balance piston flows (4000 K lb/hr vs a design of 1200
This case pertains to an etheylene refrigeration compressor K lb/hr)
rated at 7,740 hp unit with sidestreams and driven by a 8000 rpm • Operation at lower than design suction pressures (0.5 psig vs
back pressure steam turbine [11]. As the unit was experiencing a
1.5 psig)
maximum governor situation (i.e., speed could not be increased),
it was limiting the process flow. A process design house had • Off design side flows. For example, due to the condition of the
performed a debottlenecking study and had suggested that the cold box, the compressor was operating a high second stage side
machine was limited to 8,770 hp. A project to install a new load and low third stage side load. As stonewall was approached,
turboexpander was under consideration. This project, if imple there was a rapid fall-off into the third stage resulted in lower gas
mented, would have cost approximately $2,000,000. density, reducing the pressure ratio capability of that stage. These
conditions tended to increase machine rpm).
•The compressor's field efficiency was between 60 to 70
percent vs a design efficiency of 75 to 76 percent.
Based on these findings, it was decided to open the ethylene gas
compressor and replace the balance piston seal. The turbine was
not opened. This action resulted in successful debottlenecking of
the unit.
This case brings out two important points:
• The importance of aerothermal performance monitoring.
• The importance of monitoring balance line flow. This is not
monitored in a majority of installations. The increased balance line
flow caused a significant difference in gas density thus affecting
the horsepower.
an.-LlTUDE
nee nee uae ""' ,..,. ueo ?eta •••• ••ea 111111
Syboynchronous Component at 3 3% RPM FREQUENCY IN HZ
Figure 17. Overall View of the "Pinion Gear, Showing the Inner
Race of the Broken-Up Ball Bearing Assembly Still Installed. The
Teeth Seen Here Are Badly Mutilated, But None Are Broken Off.
Figure 18. 180 · Away From the Teeth Seen Above, Three Teeth Are
Seen to be Broken Offat the Root Diameter. The Fracture Surfaces
Figure 16. Cracked Upper Bearing Race Along With Physically of the Lower Two Exhibit Mutinucleated, High Cycle Fatigue,
Intact Lower Bearing. Fractures of This Nature Often Result From With Origins Along the Tooth Flank/Root Radii, on the Driver
Improper Mounting, Insufficient Clearance or Shock Loads. Seals.
1 10 PROCEEDINGS OF THE TWENTY-SECOND TURBOMACHINERY SYMPOSIUM
in the fit-up. This condition could have contributed to the progres precise detection of faults and to aid understanding of operational
sive failure of the bearing from high vibration phenomena, also problems such as compressor fouling and erosion.
contributing factors were the observations of questionable fit-up
•The use of transient behavior (both in terms of mechanical and
of the bearings.
performance) to obtain further insight to machinery problems.
Because of the rash of failures, it was decided to reverse
engineer a complete gearbox. Consequently, a "rebuilt" spare
gearbox was disassembled. During disassembly, the following CONCLUSIONS
observations were made: Several basic types of condition monitoring approaches have
• The high speed pinion system was already experiencing been presented. For critical high speed turbomachinery, no one
distress. The upper bearing would not tum easily on the shaft and technique can provide all of the answers pertaining to machine
the lower bearing did the same, the upper bearing more severely. condition. An integration of techniques is required and the infor
mation obtained by different techniques must be synthesized.
It was obvious that the bearing fit were improper. The ABEC Several case studies have been provided to show how integrated
Class 7 bearings are high precision and cannot tolerate improper monitoring can be of value for reliability improvement and to
fits. The interference fit should have been of the order of 0.4 mil. optimize efficiency consumption. Future directions in the area of
When measured, the interference was 2.0 to 3.0 mils. This was condition monitoring have been presented.
totally unacceptable for operation.
•The gear teeth of the pinion also showed signs of intermittent ACKNOWLEDGEMENTS
contact and misalignment and distress.
The authors would like to acknowledge Janet Broussard and S.
• One dowel pin on the case half was very loose. Green for their help in typing this manuscript.
• The starter input gear was found extremely worn.
• All the bearings other than the high speed bearings appeared REFERENCES
in reasonable condition.
1. Meher-Homji, C. B. and Cullen, J. P., "Integration of Condi
•The most significant observation was that the sleeve for the
tion Monitoring Technologies for the Health Monitoring of
high speed pinion bearing did not have a 1/32 in oil hole drilled in
Gas Turbines," A SME International Gas Turbine and
it.
Aeroengine Congress Exposition, Cologne, Germany, ASME
In studying the history of the gearbox, it was evident that the Paper No: 92-GT-52 ( 1992).
plant had "copied" the sleeves and this had resulted in the oil hole
2. White M. F., 1988, "An Investigation of Component Deterio
being "missed". Thereafter when the part was remade, all lacked
ration in Gas Turbines using Transient Performance Simula
the oil hole. This error propogated over several rebuilds.
tion", ASME Paper 88-GT-25 8 ( 1988).
The redesigned gearbox was carefully manufactured with engi
neering checks and dimensional checks being made through the 3. Meher-Homji , C. B. and Bhargava R., "Condition Monitor
rebuild process. The gears were also rebuilt. Appropriate bearing ing and Diagnostic Aspects of Gas Turbine Transient Re
fits were utilized and the oil hole was added. The gearboxes have sponse," Presented at the International Gas Turbine and
run successfully after the redesign. Aeroengine Congress Exposition, Cologne, Germany, ASME
Paper No: 92-GT- 100 ( 1992).
FUTURE TRENDS IN CONDITION MONITORING 4. Meher-Homji, C. and Focke, A., "Performance and Vibration
Monitoring for the Prevention of Gas Turbine Airfoil Fail
For large critical unspared machinery several trends in the
ures," 6th Bi-annual ASME Conference on Failure Preven
condition monitoring area seem to be emerging:
tion and Reliability, ASME Vol H-33 1 (September 1985).
• Centralized software and generalized databases being used
5. Doel, D. L., "The Role of Expert Systems in Commercial Gas
for multiengine fleet installations. The use of generalized standard
Turbine Engine Monitoring," ASME Paper No. 90-GT-374
software modules permits rapid software tailoring for different
( 1990).
engines. Such a concept is currently being implemented at an
offshore facility ( 16 engines and compressor trains) in the North 6. Wang, S. S., "Diagnostic Expert System for Industry," Pro
Sea. ceedings of the Second International Machinery Monitoring
and Diagnostics Conference, Los Angeles, California ( 1990).
• The evolving use of expert systems or shells for numerous
subtasks such as choice of data compression techniques (informa 7. Weiskopf F. B., Arcella, F. G., Lin, J. S., and Newmann,
tion overload), evaluation of alarm levels for diagnosis, and, most R. W., "A Hybrid System Approach for Machinery Monitor
important, in integrating inputs from a variety of condition mon- ing and Diagnostics," Proceedings of the Second Interna
·