EXPERIENCE WITH AN ONLINE CONDITION MONITORING SYSTEM by Refah R. Unsal Mechanical Engineering Specialist Phillips Petroleum Company Norway ‘Tananger, Norway and Meherwan P. Boyce President and CEO Boyce Engineering International, Incorporated Houston, Texas Refah R, Unsalis the Mechanical Engi neering Specialist in Condition Monitoring Systems, Vibration and Turbine/Contral with Phillips Petroleum Company Norway. Proviously, he worked as an Associate Technical Systems Analyst where he was the Project Leader for a Pilot Condition YN Monitoring System Project to monitor four GE Frame 5002 machines at Bkofisk Tank platform, and participated in the ~ development of a Corrosion Monitoring Interface System for the Ekofisk field. His areas of interest are performance analysis, vibration analysis, rotordynanics, furbomachinery prognossidiagnosis, systems analysisdesign and programming ‘Mr. Unsal received a B.S. degree in Electrical Engineering {from Middle East Technical University of Ankara, Turkey, and ‘an MLS. degree in Elactricol Engineering from University of Bir- rmingham, England. Meherwan P. Boyce is President and CEO ‘of Boyce Engineering. Interna tional, Incorporated, in Houston, Texas His past experience incorporates many academic and professional positions, in eluding. Professor of Mechanical Engl neering, Founder, and frst Director of the Turbomachinery Laboratory. He was also responsible for founding the Turbo tmachinery Symposium, which he chaired Jor eight years. His industrial postions include Manager of Compressor and Turbine development at Curtiss Wright and Manage of Aerodynamics Technolgy at Fairlild Hiler Corporation. Dr. Boyee has authored more than 100 significant publications and ‘echnical report and i the author of the Gas Turbine Engineering Handbook and has contributed to other major handbooks. He is @ member of Phi ‘Kappa Phi, Pou Sigma, Sigma Xi, and Tau Beta PL He is also ‘a member of ASME, SAE, NSPE, HESS and ASEE. In June 1385, Dr. Boyce was named a Fellow ofthe American Society of Mechanical Engineers Dr. Boyce was the 193 recipient of ASME's Herbert Allen Award for Excelence and the 1973 recipient of the Ralph B. Teotor Award of SAE 7 Dr. Boyce pioneered a breakthrough in technology through the development ofa realtime computer system(DATMA) which ‘monitors, analyzes, diagnoses, and prognosticates performance of major turbomachinery. These systems are in use throughout the world Dr. Boyce received his Ph.D. in Mechanical Engineering from the University of Oklahoma, ABSTRACT A total online condition monitoring system has been in oper tion on the world’s largest offshore faclityYoated off the coast of Norway. The Ekofisk platforms are operated by the Philips Petroleum Company Norway and supplies erude and nateral sas, The system, which monitors performance, as well as ‘mechanical parameters, has been operating for the past two years on four trains which are used for gas transmission from the platform to Emden, Germany. The system and the operating ex perience are described u (ODUCTION On the Philips Petroleum Ekofisk platform located 296 ken offshore Norway, is one of the largest and most complete turbo- ‘machinery condition monitoring systems (DATM3}. Itwas com pleted and put into operation in January, 1987 at acost of U.S. SLT MM. ‘This pilot condition monitoring system (POMS) is areal time, online System used to monitor aerothermal, thermodynamic, process, vibration and event variables of four frame 5002 GE tur bines driving Dresser Clark gas pipeline compressors offhare at the Ekofisk 2/4 tank platform. Bach turbine is coupled to a Dresser Clark nine stage (5 x 4 back-to-back) single shaft com= pressor with a capacity of 19 MMM Vday. The gas turbines oper ate ina simple eyele on natural gas. The gas turbines are rated at 22.4 MWiISO) and are equipped with the General Electric Speedtronic MK2 Conteo system. A schematic overview of one of the trains being monitored is shown in Figure 1. As back- round, a tabulation of the operating and maintenance history ofthese four trains is shown in Figure 2 [1] ‘Out of these four monitored gas pipeline units, one is nor: mally kept as a standby and the rest ofthe units run eontinu- ‘ously. The three units consume approximately 18.4 mmse offuel gas per day which amounts to 8U.S. 16.8 million dollarsyear based on $3.50/MM Btu. Itis easy to ee the enormous amount38 PROCEEDINGS OF THE SEVENTEENTH TURHOMACHINERY SYMPOSIUM = a 2EE-e fei se = Figure 1. Gas Turbine-Compressor Train Being Monitored. | recur oe a ee iad en Goag pod ae een Figure 2, Past History of Gas Turbine Compressor Sets Being Monitored [1], of energy involved. As these units consume so much energy, ‘monitoring is required to ensure that degradation is limited and that operation at least approaches its optimutn efficiency. Maintenance actions can also be taken to maintain ths efficiency ‘within acceptable economic limits. A one to two petcent eff ciency related saving can amount to between $170,000-340,000/ year ‘The installed system has two separate computer systems linked via satelite. All information collected on the platform ig both available offshore for operators and turbine engineers and ‘onshore in thelr Tananger office for maintenance specialists 7a ‘2 satelite connection, An overview of the system is shown in Figure 3. Pure 3. Conceptual Overiew ofthe Codtion Monitoring System. ‘The project was justified onthe basis of fel, manpower and ‘maintenance savings. Major premises are; ‘saving fuel by monitoring performance to operate the machine more effec- tively, lowering the maintenance eosts by using preventive/pre- dictive condition monitoring techniques to reduce unscheduled ‘downtime and excessive mechaneal deterioration, and perform inspections as needed Instead of on a fixed schedule Some intangible benefits are: maximizing reliability of the gus pipeline fcity and availability of the machinery, reducing hhumber of catastrophic shutdowns, and having aelable record ‘ofwhat really isthe underlying cause when shutdown occurs. Manpower stings are achieved by locating the mechanical specialists onshore and transmitting operational data over the Satellite for evaluation These savings together could show arate of return which would be less than one year. ‘The systems designed to evaluate the performance ofthe tur binesand compressors, both from an arothermal end mechani- cal point of view, thus providing analysis and diagnosis so that, stastrophic fares ean be avoided. SYSTEM DESCRIPTION ‘The fallowing three areas will be discussed: + Input Signals and Instrumentation Considerations + Hardware Considerations + User InterfaceEXPERIENCE WITH AN ONLINE CONDITION MONITORING SYSTEM so Input Signale and Instrumentation Considerations The total number of points handled by the system exceeds 1000 points. There are some 800 measured points and 640 calew lated points These measured points are divided into three major categories: analog, digital and dynamic. Analog inputs are sen= sor inputs from flaw, temperature and pressure transducers. Dk ital inputs are interlock status nd other on-off type signals. Dynatni input refers to rapidly changing items such asthe vib ration signals from which FFTs are performed, ‘There are 100 analog signals per unit (total 400) that are read cevery three seconds. These signals are input from instruments tiow such as thermocouples, t's, flow, pressure and overall val- tues of vibration from proximitor, accelerometer and velocity probes Data acquisition hardware for the analog signals consists of ‘multiplexer voltmeter arrangement with an HP-IB link to send the dita periodically to the hast computer. In the case of the exhaust temperature data fim the GE Speedtronic Is, special ‘modification had to be made to read the values. The modifica: tion consisted of putting al the thermocouples into a tempers ture averaging unit (TAU) before they were averaged and then sending the digital averaged temperature control signal to the controller rather than the hardware averaged signal. The digital signals are directly taken into a sequence of events recorder “The dynamic inputs included al the train proximity probes as well as data from the accelerometers located on the turbine. Tt consists of 25 signals per unit (total 100 which can be measured ‘up to 16 channel simultaneously). Communication to the host ‘computer is through an HP-IB link. The signals are divided into ‘two sections. One is dedicated to signals from high pressure (HP) rotors and the other from low pressure (LP} rotors of the machinery being monitored. The triggering source for the analyzers is the Keyphasor signal from each shaft. One spectrum analyzer is capable of capturing data from eight channels simol- taneously, when a trip occurs. The analyses provided by the spectrum analyzers ate: frequency spectra, order tracking spectra and time base samples. Tnstrument accuracy, reliability and repeatability is an impor tant area for the desige of any condition monitoring system. The software conducts sensor credibility checks and flags the sensor suspected. In some cases, the best estimate forthe discrepant error is made by comnputational techniques. Hardware Considerations Host Computer and Communications The system uses the HP1000 series A900 minicomputeras the online host computer. It is situated on 2/4-tank platform at Ehofisk. Its main functions are collection and storage of all data and presentation tafishore users. Italso acts as the communic tions mode for sending data to another HP1000 series A900 com- puter situsted gnshore at the Phillips Tananger base. The con- nection isa 9600 baud satellne lnk, The HP operating system being used is KTE-A and the net work software used to communicate through the satellite is AP DS/1000-1V. The onshore HP1000 has a 9600 band HASP link using HP MAJE/1000 software toan IBM mainframe computer. ‘This links utilized to transfer gas properties tables from IBM. to HP1000 onshore and eventually tothe offshore online system for use in aerothermal calculations, Both computer systems (on and of shore) have operator interface stations, mass storage and printers User Interface In order to ensure user acceptatiity, a totally menu driven system was used, Extensive use of graphic displays is used and ‘operators can access any information required without knowl- ‘edge of special computer commands. Machine schematic print- ‘outs allow operators to pick whatever data (spectrum, trend, per formance maps) they want. Some typical displays are shown in Figure 4. The performance data from the gas turbine are shown in Figure 4 (a). The centrifugal compressor unit performance data are shown in Figure 4 (b). The vibration measurement group on the train is presented in Figure 4 (o) while the bearing temperature group is reflected in Figure 4 (2). Using thie picto- tial information, any measured point or computed value can be selected snd viewed in any format desired AEROTHERMAL AND MECHANICAL SYSTEM MONITORING AND ANALYSIS Analysis is a process of manipulating engineering data and presenting it in a useful format. The condition monitoring sys- tem conducts two major analyses ‘They ae: + Aerothermal analysis, and + Mechanleal analysis Actothermal analysis allows a determination of overall of ficiencies, componest efficiencies, and heat balances which could enable operation close to design eeiency- Determine tion of component efficiencies aso allows the detection of prob- lems such as compressor and turbine fouling, and several other problems that often manifest themselves as performance changes. Mechanical analysis includes vibration analysis, bearing tem perature analysis and lube/seal oil analysis, and is an effective tool in detecting incipient failures. Several ofthese problems could be rotordynamie related. Effective machinery condition ‘monitoring requires both types of analysis Aerothermal Analys!s The need for arothermal analysis arses fom the fact that while overall deterioration may be evident in a turbomachine, Aleterminingthe speciic component or eto components) eas- ing the deterioration may not always be possible. without acrothermal analysis. Moreover, efficient operation of tur bomachines & of utmost importance by virtue of the fvel cost dbollar savings obtained by operating atthe design efficiency point, Maintenance schedulingean also be based on specie per- formance deccements. Usually iis not practical on an opers= ‘onal turbine to measure pressures and temperatures atta properties, bt for the purpose of condition monitoring, static properties with corrections can be used. It should be borne in tnind that absolutly correct values are not a importa as re- peatability. Providing measurements ae tepeatable within a reasonable tolerance, useful trends can be made, The point of the exercise sto record changes. ‘Varios assumption have tobe made to get certain measure ments. Air lw int the gus trbine compressor, for example Is 2 dificult measurement. State pressure and temperature are ‘measured atthe compressor inlet scroll The geometric area can tse determined either from drawings or by measurement. With state conitions atthe sol determined, a total measurement isneeded to complete the calcslatinn. Theambient temperature and pressure are recorded and are used in caeulaling the air toass low Into the compressor. The total pressure atthe com: pressor bellouth inlets assumed to bea fonction of he atmos Dhetie pressure and the total temperature isthe ambient tem perature. The assumption, ofeours, thatthe low through the Fiver system and ducting isentropic reversible). Although this isnot strietly true, the actual err will be approximately com stant especially for machines like the dual shat GE Frame 5002, ‘whic has constant gas generator compressor speed, With thesePROCEEDINGS OF THE SEVENTEENTH TURBOMACHINERY SYMPOSIUM ‘measurements, the compressor inlet flow ean be estimated tak: ing into consideration the corrected speed. ‘Once a given set of performance parameters is computed, it is necessary for these to be adjusted to some predetermined set ‘of conditions (such as ISO conditions of 60°F and 14.7 psi. The ‘corrected parameters taken from a new and healthy machine ‘often form an aerothermal baseline. Detals regarding conec- tion of parameters are provided in Figure 5. ere Ta read compo dere emp = (Ohm mgr) (59/7) Cal amp achg ie (Oh (147) caren = (Oe et) TF, nti (Ode (49/0) VTE coment eget = (Oman pone 83/7) VIBE aoe. Figure 4. Schematic Diagrams Showing User Interface igure 5. Performance Correction Factors, ‘The condition monitoring system should be able to diserimi- nate between performance decrements caused by ofdesign op- eration and performance degradation due to problems Tt is not practical to measure the turbine inlet temperature, so thishas to be estimated by a heat balance over the axa com. pressor and the gas generator turbine, knowing the interturbine temperature. Ifthe inter-turbine temperature messurement is not available, then itis estimated by doinga heat balance on the power turbite and the load compressor. Combustor efiieney Caleulation is corrected by a heat loss coelicent to account for the heat radiated from the turbine and energy used t raise the fue gas temperature. ‘The mass flow into the turbine changes as it passes through, the gas path. Some flow is extracted for sealing while fuel low is added in the combustor. It turns out in practice that unless compressor flow is used external tothe gas turbine, the amount of ait flow used for cooling, as well as sealing, whieh can be con- sidered as leakage itis approximately equal to the volume of sas added in the combustor. ‘With the process compressor calculations, one must be cate- fal to allow fr the fat thatthe Hud is not a perfect gas. Allow- ance for compressibility must be made. The scheme being used involves the following steps: + Got the gas constituents fom a gas analysis. + Produce gas properties tables with the given gas con- stituents for aiven pressure and temperatures. + Access the lookup tables from the aerothermal programs to do temperature entropy calculations to determine isentropic and polytropic efficiencies “Mechanical Monitoring and Analysis ‘This function is mainly responsible for analyzing date connect- Jag with mechanical condition of machinery. The cociton monitoring system offers two types of mechanical analyses. These a: + Vibration signature monitoring and analysis, and + Bearing/Seal system monitoring and analysis.EXPERIENCE WITH AN ONLINE CONDITION MONITORING SYSTEM a ‘Vibration Signature Monitoring and Analysis Tha been found in ioustrial practice that good comelation exists between the characteristics vibration signatares of max chines and thei relative condition. The diferent components of ‘machine vibrate at one or more discrete frequencies, and dif ferent malfunctions in a given component ean eause vibrations at ailerent diserete frequencies. Is the combination of these discrete frequency vibrations that result in the complex vibra- tion waveform atthe measurement poot. Therefore, «common and effective method of analyzing the measured vibration signal isto reduce itt its discrete frequency components, The result of this type of analysis usually presented asa plot of amplitude ‘versus hequedes, is what is commonly referred tos the ‘ubra- \onsignature” ofthe machine. Vibration Signature Analysis sim- ply involves comparing the current vibration signature of Tnachine to a baseline vibration signature taken when the machine was operating satisfactory. ‘The system spectrum analyzer performs analog to digital con verslon and Fast Fourier Transform (FFT) on each incoming time domain vibration signal The output ofthe analyzer isadig- {zed vibration spectrum and is stored on dis. The user can ex- amine any ofthe stored spectra individually or in to oF theee dimensional multiple spectra diagrams, e., Waterfall and Cas- cade diagrams “The software also includes a high performance cascade spectza capability. This allows excelent presentation of high density cascades (Waterfall diagrams) that allow the detection of ‘tvwety oftransient problems. The cascades may be ether ime fr rpm dependent and aze excellent diagnostic tools during Startup and shutdown Bearing/Seal System Monitoring and Analysis ‘This analysis involves measuring the bearing metal tempera: ture, the lube ol inlet temperature. and anyother lubeseal ol temperature points ‘The progam calculates the ral of change on the bearing rnetal teinperature and differential temperature between the inetal and the ol, The rate of change eliminates the problem of slow thermocouple response, The diferental temperature pro- ‘ides good Indication of bearing loading APPLICATION SOFTWARE MODULES AND THEIR USE ‘There are several main modules thatthe user is provided with asa menu on the sereen, These are: + Performance Maps. This contains a variety of pesformance ‘maps for the compressor and gas tu + Summary Charts. This contains charts displaying diferent groups of analog data + Trends. Ths allows the user to obtain a graphical display of historical data. + Individual Points, This allows the user to view any indivi ‘ual analog point. * Vibration Snapshots. This provides the user with on de- mand vibration spectra, orbits or ime records + Vibration Analysis. Thisallows the display of any ofthe “au> tomatically” taken spectra * Tannsient Analysis. This allows the user to view vibration data acquired during startup(shutdown, + Status and Alorms. This provides a current and historical display of machine status and alarms. * Operator Logs. Thisallows the operator to enter aay event! comment/note of interest (similar to operator log sheet). + Data Capture. This module allows access to 90 minutes of raw data, * Startup File. This provides data automaticaly triggered by digital events. Display formats are similar to the data capture ‘module, + Specialized Graphs. This has a collection of special outputs (EGT Profle), bor charts, ete * Daily Report. This provides a summary of events fora past defined time period + Offline System, This provides a set of maintenance support alignment and balancing programs that canbe wed nan one + Database Setup. This is a module that petmnts an au- thorized user to make routine modifeationsupdates to the database, Some of the key modules along with some application exam- ples are shown later, Performance Mops The performance map menu provides the sser with choice of aevolbermal performance maps tat llows the west isu tcthetbnrodgamichchavon af the machine ander oder ation, The map show the liveoperating polnt superimpesed on the operating characteristics map ofthe machine, Engineers and operators can visualize and optimize the on- a eration wih eet the opening envelope fhe “hit allows sf operation and allows a insight nto operating efficiencies, proximity to surge and how the machine behaw Compared tte design pint. Deterioration n performances iicteenby using these maps. The abit tosimtltaeously view four mapa provides the use ith a qulck representation ofthe penton "A performance map isthe expected machinery behavior der diferent operating conditions, This normally sed asa Baseline to measure how well amachie performing under dit ferent operating conditions, The module ptsthe designed pe ‘gma map pls incopy fhe tne coordinates ofthe operating pont ae alo depayed at the bottom right hand corner. In addition to these two values, three more user selected values canbe displayed atthe same area, AllBive values are updated continuotsy. Asample map for the load comprestor fs shown in Figure 6. Four maps in Figure ‘Tare shown on a screen which gives relationship between the “ompreswr andthe turbine. Tiss very esa in determining the relationships between the turbine and the compressors. Figure 6. Compressor Surge Map.e PROCEEDINGS OF THE SEVENTEENTH TURBOMACHINERY SYMPOSIUM. My Figure 7. Multiple Performance Maps. By monitoring the operating point of the process compressor, itis possible to see if the compressor is operating atthe higher efficiencies close to the surge line. During turndown operation, ‘minimizing recirculation, and at the same time operating close to the surge line without going into surge, ean provide consider. able fuel savings, This is possible if the operator can see the ‘operating point dynamically changing on a performance map. ‘The eystem also calculates a surge factor, a value developed as a function of the velocity perpendicular tothe shroud and the [pressure gradient, which signals the onset of surge "The surge factor accurately caloulates the surge margin on ‘cach wheel and is used to illustrate to the operator where he is from surge on each wheel. The bar charts for each wheel are shown in Figure 8, Figure 8. Surge Bar Chart. ‘The life eyele analysis chart cleslates the actual life used as a function of the following parameters. blade metal tempera- ture, number of starts, compressor discharge temperature, Dlade loading and the number of forced shutdowns, The system used the compressor discharge temperature and the computa- tion ofthe turbine inlet temperature tocaleylate the blade metal ‘temperature and withthe turbine loading the Larson Miller pax rameter is computed, This establishes 2 relationship between temperature and time for various metals. The number of starts and forced trips also hold heavily to hours operated Tis important to note that while summary charts display data in numere fashion, some displays are avaiable in bar chart or polar plot format. A polar plot ofthe gas turbine exhaust ga em- perature profile is shown i Figure 9. This sort of display permits 2 quick visual evaluation of operating conditions Figure 9. Gas Turbine Exhaust Gas Temperature Profil. Summary Charts ‘The summary chats area dynamic tabular representation of suinerical data that describes the operating Uchavior of the machines Its updated every three second A wide vit of these charts avalable: Sunmary charts eons of both mes Sued end calculated vrs “The summary charts ate means of blaning detailed informa tion about gop of sensors ofa particular pe ora group of sensor pertaningto the vibration, performance or auxilary 93° tem performance (eg, Iube and seal ol system). A particular Siar chacan be ne py ade ant oles is particular module lows the user wo display groups of data from the machine train The variables are grouped into the OF Towing charts * Exhiust gas temperature summary chart + Wheel space temperstore summary chart + Vibration measurement summary chat + Bearingfbe temperature summary chart + Gas turbine performance data summary chert + Compressor performance dst suramary chart, + Flow measurement summary chert, + Gas turbine contol summary chart, ‘The charts may display any one ofthe following: + Current values “+ Maximum value (over a sx hour period) “Minimum value + Average value (past 24h) + Baseline value + Upper alarm limit + Lower alarm limit *+ Upper danger limit + Lower danger limit + Raw data (voltages) Some summary charts forthe gas turbine and the load com- ssor are shown in Figure 10. The charts show that the gas tur ne is operating ata thermal efficiency of 21.6 percent and atur- bine inlet temperature of ISITE. The exhaust gas temperature {sat an average of876°F with maximum spread of99.6. The locd ‘compressor is operating at apolytropic efficieney of 4.4 percent land a presture ratio of 2.24EXPERIENCE WITH AN ONLINE CONDITION MONITORING SYSTEM 6 Figure 10, Summary Charts, Trends ‘The trends module permits the user to obtain plots of any’ measered or ealculated parameter over a wide choice of time frames. This feature allows the user to examine the time depen- ‘dent vibrations or other parameters that occur in the machine. ‘The trends permit the user to see how much machine be- havior varies over tine. This feature is most important from a ‘condition monitoring standpoint, The ability to select and simul- taneously view trends offour variables helpsin identifying inter. relationships in machines behavior, for example, the effect of Toad on vibration, ete “Trending is used to obtain predictions which would be helpful in the scheduling of maintenance. The system has hourly, ‘monthly and yearly trending, and prediction capabilities, The prediction allow the user to predict atime when sonditions will have deteriorated to a point demanding maintenance. ‘A trend of gas turbine ai iter diferental pressure es time is shown in Figure 11 Vibration Snapshots ‘This module permits the user to obtain spectra, orbits or tine waveform pictures from vibration sensors on the machine. This allows the user to examine the vibration behavior on demand, iie., whenever he desires to look ata particular point. The spee- "9 ae a igure 11. Trend of Intake Piter Differentiat Pressure. trum feature permits viewing of any spectrum of choice. Orbits ‘ean only be obtained where X and Y proximity probes exist ‘Time waveforms can be obtained for any vibration signal. ‘This tool i useful in troubleshooting vibration problems. The spectrum analysis feature permits representation of the vibra- tion inthe frequency domain. The orbits show a representation ‘of the shaft movement in the bearing, The time waveform pro- vides the time amplitude trace ofthe vibration waveform. The choice of display depends on the objectives ofthe user. Orbits ‘The orbit orientations are presented on the sereen with the key ‘phasor at tho three o'lock position, i¢., the orbit i adjusted to this fame of reference. The outer citele san amplitede reference: circle andthe value it represents (in mils/microns)is presented on the sereen. Figures 12, 13 and 4 show representations of orbit, spectrum and time waveform snapshots, respectively. Vibration Analysis ‘This module allows the access and management of all the spectra that are automatically acquired by the condition ‘monitoring system. There are {wo ways in which speetra are acquired * Each vibration probe on the train, or a user defined set of| probes, is scanned on a periodic basis (typically, every four \. / \ 7 ~ veer clrete 84, 2, Ae 8, Sore, yeh Figure 12. Orbit Representation64 PROCEEDINGS OF THE SEVENTEENTH TURBOMACHINERY SYMPOSIUM Figure 13. Spectrum Snapshot. Figure M, Time Waveform. hours)anda spectrum is taken. The four hoor spectrum scan rate ‘stuser definable, sot of 10 spectra for each probe are stored in aie on a fist-inefirst-out basis. Ths means that at any time tone can view up to 560 hr or about 23 days worth ef spectra his- {ory per vibration probe onthe train. tn ofder to manage and for the usr review ofthe daa, the vibration analysis nodule Is tobe used + Ifa set of user defined sensors gp into alarm on overall value thn this can trigger an accelerated rate of spectra collee- Son ona set fected wt of probe. lr intrest de lent on the frequency and type of probe. For example, fa Bodden increase in overall vibration on a particular bearing is found, this ay cause the system to concentrate on associated prosimity probes and inerease the spectrum scans on those points, Also various analog point such as compressor discharge Dressure or eniust temperature can initiate a vibration analysis {fthey exceed or decrease beyond certain limits. This a very Saluable condition monitoring feature at it allows the user to “ew the changes in vibration spettal behavior over time. This ‘llows the detection ofthe growth of unusual peaksandexamina- tion aftime dependent changes that might have occured. There tre several ways in which the data may be viewed: ‘Cascades or "Waterfall" of a number of specta (yp to 150 stacked spectra ata tie). Typically choosing about 50-60 co secutive spectra will provide the best visual revolution. Th stating sad ending spectra numbers have tobe specified. This is dowe by the operator by hisllingin an easy to use form. The ‘ascatle permits a quick detection of problems as small changes and the appearance of new or undesivable peas (e-.,asubsys- ‘hronous vibration peal) can be detected ice Spectra ofthe spectraavailable of any particular user cho- sen point. Fr example, the user may decide o pick spectrum numbers 35, 70,105,130, and 150 or display. The five spectra wilthen be diplayed on the screen, Thisfeature permits more detailed examination of the spectra. Figure 13 shows a five Spectra representation. Alarm sod danger spectral component limits are shown in Figure 1. Transient Analyse ‘This module isa very useful feature of DATMS that allows the user to review the vibration data thet is acquired during star tups, shutdowns or other transient events. Data may be viewed 48 Cascdes, Nyglsts or Bodé plots as shown in Figures 7,18 tnd 19. Alter any transient event, the user may wish to examine the vibration behavior ofthe point(s) chosen. In order to do this, he would use this module. This module allows the user to define the format ofthe display ofthe transient event. He can choose between Cascade, Nyquist or Bodé plots. The Cascade program allows the user to fix the fisVast spectra and the spectra inere- ment ‘This system is set up to lock onto the vibration probe exhibit- ing the highest amplitude upon detection ofa trip. Spectra are ‘acquired for this primary probe aad its associated probe during the machine rundown, Each trip is automatically logged to dise tnd number and date assigned to it for identification purposes ee — Figure IS. Fie Spectra Representation. i Figure 16. Spectrum Showing Frequency Related Limits.EXPERIENCE WITH AN ONLINE CONDITION MONITORING SYSTEM 6 Figure 17. Special Order Cascade Figure 18. Nyquist Diagram. ade oca ion Gig pG00,gg00 foe see avee aaee Figure 19. Bode Plot ‘Tis module is setup tolist all ofthe vibration trips occurring ‘on any machine unit. From the listing of vibration trips, an ind!= ‘elual trip can be selected for close examination, Data Capture Data capture isa feature tha allows the freezing of approxi mately 90 minutes of raw data (ve, "as scanned” data without ‘any averaging) either when a trp oceurs or upon user command. About five minutes of data after the trip will also be frozen ‘There are tixee files for storing captured dats. One ofthese is lays aotively windowing the “raw data and the other two con- fain the last two captores. This modile is activated upon machine trip. Keeping the raw data that occuts prior to, and ator the trip allows the user to observe exactly what happened that may have initiated the trip. The captured data can be dis- played in performance map format, somuary charts, trends for int oY platen the tes red mt at po- ides a good picture of the events leading upto the prot ‘The events leading to the trip can be viewed lkea video, sean- ning forward, backward and witha freeze button. This isa im portant tool to study (almost in slow motion) the underlying ‘auses that led to trip. Operators can see how the key pres- sures, temperatures, flows, and vibration behaved prior the trip. Sequence of Events Loasing Event sequence logging is avery useful tool in the operation ofthe machinery tis mainly used to determine the reasons that cause trips and solve problems associated with startups. [tis an Important troubleshooting tool for the operators. Tt also has saved many probiems du tothe fact that in the past, operators ‘would tend to start up units without fally understanding the cause ofthe trip. Dataislogged atan interval fone millisecond CONDITION MONITORING OPERATING EXPERIENCE, ‘The aforementioned system became operational in November 1985. All the features ofthe system were not avail- able immediately, but tending of data was begun. ‘The specialist group onshore now have a very useful tool for evaluating the operation ofthe turbines and compressors that are monitored by the system. The ability to be able to observe the operation ofthe pipeline turbines, from 296 kin away, has sreatly increased the effectiveness of the specialist gtoup. The Gualtyof advice that can be given tothe ofshore operators and maintenance personnel is very different now, because previ cusly the only communication was over the ofshore telephone. ‘The offshore personnel are improving their anderstandingof the operation ofthe turbines and compressors and are able to troub- leshoot problems, with much greater confidence. Downtime is being reduced by quicker diagnosis of problems. “To date the system has proved most valuable in the troub- leshooting mode. Vibration problems in particular have pro- ded opportunities to use the system effectively to reduce downtime by speeding up investigation. Recently, one of the load compressor units was retrofited with squeeze film bearings and tit pad seals to overcome apersistent subsynchronous vibra tion problem. The vibration analysis capabilities of the DATM4 system is proving invaluable in monitoring the operation of the machine, sll ent sy the modifation iacomplte ses tat least itis possible to thoroughly document the effets of the modification 2 Because these units have been recently overhauled, there have been no dramatic savings rom aerothermal analysis a yet, sine the deterioration ofthe turbines and compressors is rela tively sho, and longer term trends are required to detect the deterioration, However, small performance enhancements have been made by optimizing fuel nozzle flow patterns de- tected by exhaust temperature spread proflles,66 PROCEEDINGS OF THE SEVENTEENTH TURBOMACHINERY SYMPOSIUM ‘The aerothermal pat ofthe system is under continuousdevel- ‘epment end better ways of measuring performance by: adding ‘more instrumentation are being pursued, ‘The turbine operators are also learning the value of such routine tasks as changing air inlet filters and doing compressor ‘washing atthe optimum time, by seeing the instant increase in compressor efficiency. One interesting incident, which ilus- trates the usefulness ofthe system, occurred as follows. Shorty after the system was commissioned, the onshore specialist noticed that the compressor inlet temperature on one turbine was 10°C higher than the two adjacent turbines. Offshore maintenance were asked to investigate and check if there was a recirculation ofthe exhaust gases from the turbine or other tur bines in the vicinity. They measured the temperature justin front of each ofthe air filter systems and reported no significant difference, After further investigation however, they found that the anti-cing ducting inside the intake was hot. The cause was ‘leaking valve in the anti-cing duct from the turbine exhaust. ‘This represented quite a loss in eflcieney and so fuel savings were made by fing the valve. By proper performance monitoring, the intervals between the various gas turbine inspections and overhauls have been in- creased amounting to large savings, The surge system has also enabled the units ta be nun with much less recycle flow leading to savings. Gains in efficiency will often be made from simple mainte- nance tasks suchas changing filters at the correct time, compres- sor washing, fixing leaking valves snd leaning fuel nozzles. One ‘might assume that these tasks woul be done anyway even with: fout a condition monitoring system, but expesience has shown thatoften these smaller tasks are put of to some future more con- venient time. Highlighting ossin efficiency, by producing trend ceurves, can change this atitude ‘The condition monitoring system itself has been available about 99 pereent ofthe time in 1987 and with an improvement fon the number expected in 1988, ‘Some abservations and lessons leamed are provided below: + Inter-turbine temperature measurement onthe gas turbine 1s a troublesome measusement from a sensor reliability point. Because of the short sensor life that has been experienced, are design i being considered for this probe. ** Users of sophisticated condition monitoring systems should expect intial onsite activities of software modification, database ‘modification and system tuning. Time should be allocated for this (onsite). Iti easy to underestimate the complexity of this task * The condition monitoring system highlights sensor failures Work ordersare written to correct these at the most convenient time. + Maintainabilty of the data acquisition cabinets isa very im: portant consideration. Ia retrospect, t would have been better to have a less tightly built unit which would help in mainte * System documentation is a very important consideration, Improvements could be accomplished by the use of indexing of library software routines and by the use of detailed programmed Interaction diagrams. More extensive help files would have been benef * User training should be provided ata place different than the installation site where eperators can spend more time with the system without interruptions, This would require a simula tion facility CONCLUSIONS Description and expeviences with a comprehensive online condition monitoring system have been outlined, as applied to four gas turbine compressor trains. Computer based online con- dition monitoring systems wil be the wave ofthe future on large critical machinery. These systems should provide the users and machinery engineers with grester insight lato the working of the machine—both in terms of its thermodynamic performance, vibration behavior and performance ofall auxiliary systems, This insight is invaluable when difficult value judgments are to be made relating to raaintenance and availability ofthe machinery. ‘This comprehensive system has been successfilly integrated into the normal operation ofan offshore platform REFERENCES 1. Operating and Maintenance Practices for Heavy Industrial Gas Turbines on Tho Ekofisk vo Emden Gas Pipeline; J. Cul- Jen, CIMAC Proceedings, Warsaw, Polsad, june 1987), ACKNOWLEDGEMENT ‘The authors would like to thank the management and partners of the Phillips Petroleum group, Fina Exploration Norway, Ine, Norsk Agip A/S, Norsk Hydroa.s., Elf Aquitaine Norge AVS, Total Marine Norsk AVS, and Den Norske Stats Oljeselskap as. for their cooperation and approval of the publishing ofthis paper.