Turbomachinery Technology Seminar

Quality in Maintenance

CsTERPllLAR@
SolarTurbines

Contents
Page
INTRODUCTION MAINTENANCE OBJECTIVES MAINTENANCE APPROACHES MAINTENANCE PLANNING AVAILABLE TECHNOLOGIES INTEGRATED GAS TURBINE CONTROL/ CONDITION MONITORING SYSTEMS SUMMARY ACKNOWLEDGEMENTS REFERENCES

81-I 81-1
81-I 81-4 81-6 81-14 81-14 81-14 81-15

Cat and Caterpillar are trademarks of Caterpillar Inc. Solar, Saturn, Centaur, Taurus, Mars, Turbotronic, and SoLoNOx are trademarks of Solar Turbines Incorporated Specifications subject to change without notice. Printed in U.S.A. Copyright Q 1992 by Solar Turbines Incorporated -l-l-S81 /492

Quality in Maintenance
C.S. Woods
Manager, Customer Services Technical Center

W.J. Bliss
Senior Engineer, Customer Services Technical Support

INTRODUCTION
Gas turbines have a number of advantages that distinguish them from many other power sources, including greater reliability and lower overall lifecycle cost. A quality maintenance program is essential to benefit fully from these attributes. In recent years, significant advancements have been made in both the concepts and technologies used in maintaining gas turbines. This paper addresses aspects of establishing an effective, efficient maintenance program for small and medium-sized gas turbines. In touching upon the goals of a quality maintenance program, it discusses what should be maintained, when it should be done, and how it can be accomplished by employing the various technologies available, including lube oil analysis, vibration analysis, gas turbine performance analysis, borescopic inspection, and trending of operating parameters.

quantified using the following formula (Chandler, 1984): Period - (SD + UD) x 1oo Availability (%) = Period where: Period = Length of time defined in either hours or days = Scheduled Downtime SD UD = Unscheduled Downtime

Maximized Production. This means attaining

the largest gross output over a given period of time, whether the output is kilowatts of electricity produced or millions of cubic meters of gas moved down a pipeline.

Optimized Efficiency. Highest efficiency is

achieved when the minimum energy is used to realize the desired level of production.

Control of Operating Costs. A well-designed

maintenance program helps to minimize the cost of operation over the gas turbine s life cycle. Among the elements that can be controlled are overhaul cycles, resources (equipment and people), and inventory (replacement parts).

MAINTENANCE OBJECTIVES
Different operators have different needs and will often emphasize different objectives in their maintenance programs, including:
l l l

Availability Maximized production Optimized efficiency

MAINTENANCE APPROACHES
Essential to a quality maintenance program is determining what to maintain and then deciding when to do it. Three different approaches to maintenance timing can help solve these problems: + Unscheduled Maintenance - Maintenance performed only when an incident occurs.
l

* Control of operating costs Although operators assign varying levels of importance to each of these objectives, if thoughtfully designed, a quality maintenance program can achieve all of them.

Availability. The simplest definition of availability is readiness to run. Availability can be

Scheduled Maintenance -Also called preventive maintenance, this approach consists of a schedule of periodic maintenance tasks designed to preclude failures.

81-I

On-Condition Maintenance (Predictive Maintenance) - Planned monitoring of package condition. Maintenance is then scheduled based upon the equipment s health. Not only are costly surprises avoided, but unneeded maintenance can often be deferred. When the measurements of machine condition are trended over a period of time, they can be used to anticipate failures well before they occur. For this reason, on-condition maintenance is also called predictive maintenance.

Several factors must be weighed to determine which approach should be used for each system or component. These include the criticality of the system or component, the relative costs of applying each maintenance concept, and the accessibility of both the component and the site. Items that are not expensive or critically important may fall into the unscheduled category and can be repaired when they fail. As cost or criticality increase, preventive maintenance becomes desirable. Components that are even more costly or critical, or both (Figure I), are generally monitored more carefully using predictive maintenance technologies. For gas turbine packages, the biggest, most important components are often the most reliable and require the least maintenance. The best example is the gas turbine itself. Though it contains thousands of individual parts, its maintenance is relatively simple. Since it is very critical and the most costly part of the package, it warrants careful monitoring. The same is true of the driven equipment (generator, gas compressor, or pump). However, the gas turbine is just one part of the overall package. If not given proper attention, minor components can cause serious problems.

For instance, a loose compressor discharge pressure (Pcd) line to the fuel control can prevent the gas turbine from starting. Ancillary equipment used to ensure the quality of air, fuel, and water should be part of a quality maintenance plan. Also, many components mounted off the gas turbine skid have a vital effect on the operation of the gas turbine and its driven equipment. Examples include switchgear for generator sets, as well as gas coolers, scrubbers, and yard valves for compressor sets. Various process valves and associated equipment are important to the proper operation of other types of driven equipment such as pumps. Most manufacturers provide recommended maintenance procedures and some schedules based upon the design of the gas turbine equipment. However, they may not meet all the needs of a specific installation. A few manufacturers offer standard maintenance tables that indicate which components should be checked or maintained, along with a suggested schedule (Table 1). Some manufacturers develop application or sitespecific maintenance tables that typically include elements of both scheduled and on-condition maintenance. Such information should be used as a starting point when planning any quality maintenance program.

Operating Experience
Actual operating experience provides one of the best guides as to which components need attention. Frequently, a vast amount of useful knowledge exists at all levels within the user s maintenance department. Gas turbine user groups and associations can provide valuable information as well. Other operators rely upon contractors to define what requires maintaining. Still others have found that relying on the gas turbine manufacturer s expertise is very helpful. Ideally, experience from all sources should be considered when designing a quality maintenance program.

Predictive

Operating Conditions
A harsh operating environment or unique operating conditions may necessitate modifications once a basic maintenance program has been outlined. For example, dirty fuel gas may make it necessary to remove, inspect, and clean the fuel injectors more frequently than if very clean fuel was used. Additional modifications may be required due to trends detected through predictive maintenance.

CRITICALITY -

81-OOlM

wgure I - Maintenance Concepts

81-2

Tab/e 1. Excerpt from Maintenance Table

SemiAnnually

Periodic Checks, continued Air Systems: 8. Check air inlet system for obstructions or contamination; record differential pressure 9. If air dryer is installed, check operation 10. Check inlet guide vanes for position; check torque paint on full-open stops and actuator cylinder linkage 11. Inspect gas turbine compressor variable vane mechanism for wear. Check for bent arms, loose linkages, loose bushings. Ensure stop settings are correct. Check for damaged signal wires to actuator (if applicable) 12. Inspect bleed valve actuator mechanism for proper operation. 13. Inspect intake and exhaust systems for damage, leaks, debris Lube Oil & Servo Oil Systems: 14. Check oil tank level every 24 hours. Record oil consumption 15. Verify proper operation of oil makeup system (if installed)

Daily

Monthly

Annually

X X

81-026M

In addition to air, fuel, and water quality, local environmental conditions can be a major consideration. The typically salt-laden atmosphere found in most offshore installations or the higher pollution levels found in industrial areas require closer monitoring of gas turbine performance. More frequent compressor cleaning may be needed, due to salt and hydrocarbon ingestion into the gas turbine s compressor. Extremes in ambient temperature and humidity also influence needed maintenance. These modifications to maintenance plans may be permanent or seasonal.

Trending
The backbone of all predictive maintenance programs is the detection of trends within the parameters being monitored. The observation of an adverse trend provides the second reason to modify maintenance schedules. If routine vibration measurements (either on-line or periodic) indicate an increase in vibration, closer scrutiny may be warranted (Figure 2). A steady rise in the amount of a wear metal detected in the lube oil may suggest a need to increase the frequency of oil sampling. Wear in a particular bearing, for

example, may produce some amount of copper or lead. Such modifications to planned schedules are usually only temporary, lasting until the suspected problem is identified and solved. Modifications to schedules due to environment or predictive maintenance trends can also be positive. Intake air that is particularly dry and clean air can allow extended time between ingestive cleaning of the gas turbine s compressor section. Monitoring the machine s health may allow the time between overhauls to be extended. Deciding what to maintain and when to do it requires consideration of numerous factors on a system-by-system basis. Most of the recent technological advancements in gas turbine maintenance take advantage of trend analysis. This consists of monitoring the changes in important measurements over time. Analyzing the results can help detect and predict potential problems in time to avert them. Whatever the parameter being tracked, a baseline value should be first established. The baseline values of a specific parameter may vary between similar units due to manufacturing tolerances, operating hours, and maintenance

81-3

SHUTDOWN SETPOINT 0.5 -______----_----------_ s 0 0 0.4 z g 0.3 cn r 0.2 0 Z 0.1

I I I I I I I I I I 81-002M

under conditions as nearly identical as possible. Even when some variables cannot be controlled, recording important operating conditions allows them to be considered when analyzing the trends.

WARNING SETPOINT

MAINTENANCE PLANNING
Once the tasks have been defined and a schedule has been established, the operator is faced with two more key decisions about how to best accomplish the diverse set of maintenance goals. First, the level of operator involvement in the maintenance program must be defined in terms of: * Number of machines versus expense of hiring, training, and equipping qualified personnel l Relative costs of doing maintenance internally versus contracting Q Complexity of tasks better done by others With only one machine, it may be better both technically and financially to have an experienced contractor perform at least some of the maintenance, as opposed to hiring, training, and equipping a maintenance staff. With a number of machines, investing in developing internal maintenance capabilities could yield lower maintenance costs. Even with in-house staff performing all of the maintenance, the operator may choose to seek assistance in more complex aspects of the program.

a. Overall Vibration (No. 3 Bearing)

E 30 aa 0 o 20 IO

81-028M

b. Spectrochemical Oil Analysis

F;gure 2. Adverse Operating Trends

Maintenance Program Example

A good example of a focused and well thoughtout maintenance program is one developed by the gas operations division of a European oil and gas company operating a large number of gas turbine-powered natural gas compressors, ranging in size from 820 to 22,370 kW (1100 to 30,000 hp). Within the maintenance department responsible for gas turbines and reciprocating engines is a specialized Engine Diagnosis group. This group carries out a wide variety of predictive maintenance work on a scheduled basis. Among the data used by the mechanical and electrical engineers involved are the results of lube oil analysis (both spectrochemical and wear particle analysis), borescope inspections, gas tu bine performance, and extensive vi bration analysis. In a unique approach to conducting predictive maintenance work, this group has built two specialized vans (Figure 3) that are driven to the work sites. These vans are equipped with an array of

schedules. The absolute level of the baseline is not essential, as long as it falls within acceptable operating limits for that parameter. What is important is the occurrence of divergence from the baseline. Such trends are a direct reflection of changes within the machine, and can be important in planning maintenance and reaching established maintenance goals. For some parameters, such as a bearing temperature or a vibration level, any divergence from baseline may be significant. In other cases, a change in the rate of divergence may signal a potential problem, such as a performance decrease due to blade tip rub. Operating conditions such as temperatures, pressures, and speeds will vary, and these changes affect trends. For instance, different speeds and loads alter measured vibration levels. While it is seldom possible to duplicate operating conditions exactly, data for trending should be taken

81-4

FFT analyzer Oscilloscope Digital vector filter 8-channel DAT recorder Printer Plotter Cable storage (on drums) Two-way radio Desk and bench spaces Runout compensator PC computer and monitor Switching matrix Various filters Selection of transducers Transducer shaker table Proximity probe testing device Batteries and charger Acoustic and ultrasonic tools Tool, tape, and manual storage Laser alignment system Among the unique features of this instrumentation is a switching matrix that allows great flexibility and ease in connecting various input signals from the gas turbines to the different electronic devices in the van. Cables are run from the various transducers and measuring devices on the gas turbine package directly to the van, where they are attached to a number of input connectors wired to the switching matrix. Inside

the van, the diagnostic tools are also wired to the matrix. While making measurements and diagnosing the results, the operator can then connect any transducer to any device in the van by merely inserting pegs in the appropriate matrix holes. The matrix can also be controlled by computer programs written for use in the van. These programs largely automate the collection and analysis of data, the comparison with earlier data (from digital audio tapes), the calibration of transducers, and other tasks. The Engine Diagnosis group also makes use of laser alignment tools, eliminating the wide assortment of mechanical alignment tools required to service a diverse fleet of gas turbines. Acoustic and ultrasonic measurements, and other instrumentation for maintaining both gas turbine and reciprocating equipment are also used. Such an investment in equipment and the personnel to use it is not a viable option for all gas turbine users, but it does serve as an example of a commitment to quality maintenance on the part of one operator of a large number of gas turbines. Regardless of the approach taken by others, it is important for the operator to assess internal capabilities when designing a maintenance program.

Figure 4. Diagnostic Van (Interior)

81-5

AVAILABLE TECHNOLOGIES
Once it has been determined to what extent the staff will be involved in the maintenance program, the operator must decide which of the many powerful technologies that are now available will be employed, including:
l l

Spectrochemical lube oil analysis Computer-aided vibration data collection and analysis PC-based performance analysis and trending

l l

Borescope inspection and graphical documentation * Monitoring miscellaneous parameters (manually or with integrated control/condition monitoring systems)

Lube Oil Analysis

Perhaps the easiest and most cost-effective form of predictive maintenance is lube oil analysis. Periodically, a sample of lube oil is taken from the gas turbine for spectrochemical oil analysis. The amounts of various elements in the oil are trended, including a number of wear metals. Part of each sample is used for physical properties tests to monitor the lubricating quality of the oil itself. The frequency at which samples should be taken can vary depending on the operating profile and environment. If a gas turbine is started and stopped frequently, subjected to large ambient temperature swings, or operated in a dirty environment, samples should be taken every month. A sample taken every three months may be acceptable for a gas turbine running continuously in a relatively clean environment. Sampling each six months would be suitable for gas turbines in standby applications or operated infrequently. Wear metal levels in reciprocating engine oil typically trend upward over time (Figure 5). An increase in the rate of wear gives early indication of potential problems. Wear metals in gas turbine oil typically reach equilibrium quickly, then remain essentially constant over time. Any significant divergence from this flat trend curve warns the analyst of potential problems. Key elements in ensuring quality lube oil analysis are proper sampling technique and the selection of a quality laboratory to conduct the analysis. After the oil has been mixed by running the machine, samples should be taken from the oil tank in a manner that will not pick up sediment

in the bottom of the tank. Taking separate samples from several different drain locations on the machine does not help to identify the area in the machine where the wear metals originate. This is due to typically low wear rates and low concentrations of wear metals in the oil. Filtration does not have a measurable impact on spectrochemical oil analysis in gas turbines. This type of analysis measures particles ranging up to about 5 microns in size. Typical gas turbine oil filters have a nominal media rating of 10 microns (Figure 6). Certain physical property tests can be done along with spectrochemical analysis. These include measuring the sample s viscosity, Total Acid Number, and water present. Experience has proven that along with adequate oil filtration, these tests allow the quality of the oil to be monitored adequately, so that oil can be used indefinitely without changing. Unlike reciprocating

Reciprocating Engine

TIME -

Figure 5. Wear Metal Trends

100 I-\ 7 - - - - \ /--

Spectrometer I
I I

I Ferrograph Typical Filter

I I \

0.1

10

100

1000
81-006M

WEAR PARTICLE SIZE, pm

Figure 6. Wear Particle Sizes

81-6

engines, the levels of wear metals and contaminants do not usually increase over time in gas turbines. Typical criteria for oil replacement in a mid-size gas turbine are given in Table 2.
Table 2. Typical Oil Replacement Criteria for Mid-Size Gas Turbine
Viscosity Change Limits Total Acid Number, Max. Water Content Parts per Million, Max.

Lubricant Type Synthesized Hydrocarbon MIL-L-23699 and MIL-L-7808 Petroleum Fire-Resistant

+25%, -10%

4.0

2000

+15%, -15% +25%, - 10% +15%, -15%

2.0 1 .o 3.0

2000 2000 2000

81-027M

While spectrochemical analysis is most appropriate for trending purposes (due to its simplicity and low cost), additional tests are readily available from most commercial labs to assist in diagnosing problems. These include ferrographic (wear particle) analysis and foaming characteristics tests.

analyzer/data collector, a set of PC software, and a variety of transducers and accessories (Figure 7). Properly implemented, computer-based vibration programs permit the detection and correction of problems before they impact production or maintenance costs. Generally, vibration analysis involves consideration of not just the level (amplitude) of vibration, but also the frequency at which vibration occurs. The amplitude is a measure of vibration severity, while the frequency is used to determine the cause of the vibration. Specific mechanical faults generate characteristic frequency signatures. Knowledge of these signatures is the basis of spectral vibration analysis. A quality vibration trending system will allow the periodic collection of spectral vibration data which are essential to detecting and correcting potential vibration problems at the earliest indication. In this way, both initial and secondary damage can be limited or even prevented entirely. In addition to helping to gather and store spectral data, these systems allow semiautomated analysis and trending of specific vibration parameters (Figure 8).

Vibration Analysis
Many gas turbines are now equipped with on-line vibration monitors with the dual purpose of getting the operator s attention, thus warning him of a problem and actually shutting the unit down to prevent a destructive failure. These are important roles, but they do not contribute significantly to a good predictive maintenance program. By the time a monitor is in alarm, the problem may be fairly severe; it may be too late to avoid an unscheduled shutdown. The monitor can still provide sufficient warning to avoid secondary damage. Newer, integrated Programmable Logic Controller (PLC)-based control/monitoring systems can offer an improvement by trending such things as overall vibration levels. Even these, however, do not trend frequency-specific vibration data that can better detect not only the existence of a potential problem, but can help identify the specific fault. In addition to on-line monitoring, vibration data collection and analysis systems are now available which help to overcome this lack of frequency-specific analysis capability. Typically, such a computer-aided system would include a portable

81-007

Figure 7. Depiction of Vibration Data Collection System

81-7

1.8 1.6 v) 1.4 r5. 1.2

No. 1 Bearing (Horizontal) Prox. Probe

lx Ngp Speed

2 2 -O
I-

5 0.8

2 0.6 c!J 0 0.4 CL

Warning Level

100

200

300

400

500

600

700
81-009M

DAYS

FREQUENCY (Hz) OR ORDERS

81-OlOM

Figure 8. Vibration Parameter Trends

Figure 9. Vibration Parameters

When the vibration data base is set up in the computer, each gas turbine package is divided into a series of measurement points. At each measurement point, a spectrum will be acquired when vibration data are collected. In addition to storing the spectrum itself, the program stores and trends 4 to 6 vibration parameters at each measurement point. Each parameter consists of a measurement of overall vibration within a narrow frequency band. Each band is designed to encompass a spectral peak at a frequency that might occur at that point on the machine, if a specific problem occurred. For example, imbalance or bearing damage in a gas turbine would be observed at its running speed, while each gear-type pump would have a specific meshing frequency that can reveal internal problems. The vibration levels within these bands are then trended over time, and the vibration analyst is notified if they exceed any of several types of preset alarm levels (Figure 9). Thus, trending the parameter can provide early detection and identify a specific problem should it arise. Should an alarm occur, the analyst can then retrieve and closely scrutinize spectral data, waveforms, and other information to help diagnose the problem. At this point, the analyst s training and experience are of utmost importance. While the computer-based systems allow the analyst to process large quantities of data and make the job much easier, eventually it is the analyst who must decide what, if anything, is wrong and the action to recommend. Vibration analysis requires training and experience. Excellent training is commercially available. With this training, a technician who is not an expert in vibration analysis can adequately

detect and diagnose about 80% of the problems that will likely be encountered. With experience, this percentage will improve. One point that cannot be overemphasized is that the value of a vibration program is directly related to the quality of the raw data. Items critical to high quality, accurate vibration data include:
l l l l

Selecting good measurement points Choosing appropriate vibration transducers Ensuring proper mounting of transducers Selecting appropriate analyzers

Measurement Points
Care must be taken when selecting measurement points in order to allow a direct mechanical path to the object being monitored (a bearing, for instance). Transducers must not be placed on parts of the machine which will resonate, such as flexible metal and loose structural pieces. Also important to selecting good measurement points is the choice of the proper direction of measurement. Imbalance, for example, is usually manifested radially, while misalignment often shows up axially. It is important to choose enough measurement points to adequately monitor all key parts of the machine, but too many measurement points should be avoided. Excessive measurement points result in additional instrumentation expense, data collection and analysis effort, and computer memory use, while providing little additional insight into the machine s health. Figure IO shows the measurement points for a typical mid-sized gas turbine package. Measurements are taken at each major part of the package, and the points are carefully chosen to monitor the most likely types of potential vibration problems.

81-8

MEASUREMENT 1 POINTS

34 5

7 8

range would be 5 to 10,000 Hz (Figure 12), though a wide choice is available. Their frequency response makes them particularly sensitive to proper mounting technique (Hewlett-Packard Co., 1983).

Decision Path. Figure 13 illustrates a flow diaI Point I Description Accessory Drive Gearbox (Axial) Gas Producer (Horizontal) Power Turbine (Vertical) Gearbox (Vertical) Compt-Fssor Bey;ing (X t Y) Transducer Accelerometer Velocity Velocity Accelerometer Proximity Probe 8, I,

81-011

Figure IO. Typical Vibration Measurement Points

Transducer Types
A second factor essential to accurate vibration data is choosing the proper transducer type. Generally, there are three types of transducers, each with its own advantages and disadvantages:
l l l

gram depicting the decision path for choosing the proper transducer type for a particular application. In the past, either proximity probes or velocity transducers have been used on gas turbines themselves. Accelerometers are generally used on gearboxes because of their ability to measure high frequency gear meshing. The present trend for gas turbines is away from velocity transducers and toward accelerometers due to their greater reliability. Aero-deriviative gas turbines are normally equipped with roller bearings. Since faults within roller bearings generate very high frequencies, accelerometers should be used on these gas turbines.
Natural Resonant Frequency (Damped)

Displacement probes Velocity transducers Accelerometers

Useful Range

Displacement Probes* Displacement probes (also

called proximity or eddy current probes) measure the relative displacement between a shaft and the transducer mount. Many mid-size industrial gas turbines now make use of internal displacement probes. Due to a drop in signal strength as frequency increases, the upper useful frequency of these transducers is about 1000 Hz.

500

1,000 FREQUENCY, Hz

1,500

2,000
81-012M

Velocity Transducers. Velocity transducers consist of a permanent magnet suspended within a coil by springs. The movement of the magnet within the coil creates a voltage signal proportional to vibration velocity. The moving components give the transducer a low resonant frequency. By design, this natural resonance is highly damped. This limits the lowest frequency at which they can be used. Typically, their useful frequency range runs from 10 to 2000 Hz ( Figure 11).
Figure I I.

Frequency Response of Typical Velocity Transducer

Accelerometers. Accelerometers contain a fixed

mass and a piezoelectric crystal. The mass applies a force to the crystal when exposed to vibration. An electrical charge is produced proportional to the force and, hence, the acceleration. Since they contain no moving parts, they are rugged and long lasting. Being fairly stiff, they have a high resonant frequency which limits their upper useful frequency. A typical frequency

J
0 5000 10,000 FREQUENCY, Hz 15,000 20,000
81-013M

Figure 12. Frequency Response of Typical Accelerometer

81-9

r MEASURE NO RELATIVE DISPLACEMENT OR CRITICAL CLEARANCE? YES 1 , Y

TRANSMITTED TO MACHINE

NEED TO MEASURE ABOVE 1000 Hz? S 1 1

NO E1

USE A VELOCITY USE DISPLACEMENT PROBE I II ACCELEROMETER II II TRANSDUCER I

81-014M

and the stud s perpendicularity to the surface should be assured. This is particularly important because the accelerometer is often used to measure very high frequencies. For velocity transducers, various valid mounting arrangements are possible (Figure 17). Specially designed mounting blocks can be bolted or even glued to the machine. Any mount should be thoroughly tested before selection. Knowing how well it transmits vibration is important, and it is critical to ensure that its own mounted resonant frequency is not within the frequency range to be measured.

Selection of Analyzers
Figure 13. Decision Path for Transducer Selection

Once good measurement points have been chosen, appropriate transducers selected, and proper mounting techniques defined, the appro-

Transducer Mounting
One of the major sources of inaccurate vibration data is improper transducer mounting. It is essential that all transducers be mounted rigidly. Proximity probes are usually embedded in bearing housings or otherwise mounted rigidly within the machine by the manufacturer. Velocity transducers and accelerometers are mounted externally on the machine casing. For gas turbines, hard-mounted transducers should be used versus hand-held or magnetically mounted ones. Figure 14 illustrates the response of an accelerometer attached to a shaker table by various means. The response is basically flat to 10,000 Hz when the transducer is firmly mounted using a stud mount. The quality of the response degrades with other mounting techniques. A magnetic mount yields inaccurate results beginning at about 4000 Hz. The handheld transducer is unusable above 1000 Hz (often as low as 500 Hz). In addition to the inaccurate results obtained from hand-held and magnetically mounted transducers, the results are generally unrepeatable. Figure 15 depicts a series of vibration spectra taken from the same point, on the same machine, and by the same person using a hand-held transducer. It is evident that the spectra are not at all the same. This becomes a serious problem when attempting to employ computer-based trending systems. Such variances defeat the trending capabilities of the program. For accelerometers, a stud mount is typically used (Figure 16). Ideally, this stud mount should be applied on a machined surface. The quality of the surface is very important; it should be smooth

I
g +I0

Hand-held Probe

ki 2 - 0.0

2 z -10

-20 I 0 I 2 I 4 I 6 I 8 10
81-015M

FREQUENCY, kHz

Figure 14. Frequency Response from Different Mounting Techniques

PLOT SPQN 2.4

0i

2000

4000 6000 FREQUENCY IN HZ

8000

10000
81-016

Figure 15. Spectra from Hand-Held Transducer Data (5-3-89)

81-10

81-021

Figure 16. Accelerometer with Stud Mount

vibration with a fairly wide moving filter. FFT analyzers digitally simulate hundreds of very precise fixed-frequency filters. This vastly improves frequency resolution, making analysis much more precise. Data collectors with sufficient memory for the anticipated amount of data. Analyzers capable of measuring and recording phase information, which is needed to accomplish trim balancing. Some analyzers have the capability to perform the trim balancing calculations. If waveform analysis is desired, the analyzer should also be capable of storing such data. Analyzers capable of being used with a variety of transducers. Most analyzer/data collectors are part of an integrated PC-based system. The analyzers are used to collect and analyze data at the machine site, while the computer programs are used to program the analyzers and establish and manage vibration data bases. In these data bases, the trending of vibration data is accomplished (Figure 19).

81-020

Figure 17, Typical Velocity Transducer Mounts (Blocks)

priate vibration analyzer must be selected. Fortunately, a wide variety of excellent instrumentation is now available. Some analyzers are intended for sophisticated design and diagnostic work, while others are intended specifically for predictive maintenance. Setting up a quality maintenance program should focus on analyzers intended for predictive maintenance, particularly if many machines are to be monitored. Typically, those designed for use in machine condition monitoring are less complex, smaller, more rugged, and less expensive (Figure 18). Current technology has made many analyzers available, the better of which are: 0 True FFT (Fast Fourier Transform) analyzers. Some devices currently on the market are merely sweeping-filter recorders that measure

81-022

Figure 18. Vibration Analyzer

81-11

0 3 LOAD DATA INTO PC 0 1 PROGRAM ANALYZER

0 2 COLLECT DATA FROM PACKAGE

81-019M

4 TREND AND 0 ANALYZE DATA

5 0

GENERATE REPORTS

Figure 19.

lntegra tion of Analyzer, PC, and Software

Properly established, a PC-based system allows the trending and semiautomated analysis of large quantities of data. It will warn the analyst of

would be under those same operating conditions,

then compares those to the actual performance parameters from the machine. The percentage variances from nominal (used

adverse trends and can enable maintenance personnel to play a more active role in monitoring the condition of machinery. If several machine types are to be monitored, or if the vibration program will be established at separate geographic sites, master data bases should be considered. The use of master data bases accomplishes several major things. First, it encourages correct, standardized data collection and analysis at all levels of the organization. Second, it ensures compatible data for the free exchange of information within the organization. Finally, it facilitates refinement of the program by allowing each participant to learn from the collective experience of all.

as a reference point) are then trended over time.

The resulting trends show the performance degradation that is recoverable and nonrecoverable (Odom, 1989). For example, Figure 20 illustrates the performance degradation due to contaminants being ingested by the compressor. Figure 21 depicts the gradual, nonrecoverable performance degradation that occurs during the time between overhaul (TBO). Figure 22 displays the combined

losses. The sawtooth effect is due to the occasional ingestive cleaning of the compressor to
I-

Gas Turbine Performance Analysis

Some gas turbine manufacturers offer computer programs for trending performance. These programs may be run on a personal computer or may

be incorporated into a PLC-based control/condition monitor system for early detection of performance degradation. Performance parameters are periodically taken from existing machine-mounted instrumentation. Since operating conditions are seldom constant, performance parameters taken at one time cannot be compared directly with those from another time. The computer program calculates what the performance parameters of a like-new gas turbine

ENGINE OPERATING HOURS -

wgure 20. Recoverable Performance Degradation

81-12

nominal values. The key is to monitor changes in each trend over time.

Borescope Inspection
Periodic inspection of a gas turbine s hot section is used to detect and trend thermal wear and damage. Certain types of problems, such as clogged fuel injectors, can be detected in this way and corrected to prevent more serious faults. Trending the condition of internal gas turbine components helps in the orderly planning of overhauls, allowing them to be done when needed rather than too soon or too late. Among the problems that can be detected by this type of inspection are hot section thermal damage and wear, clogged or damaged fuel injectors, and contamination in the aft compressor stages. Damaged blades or nozzles can affect performance. Damaged/clogged fuel injectors or a damaged combustor liner can affect emissions, as well as cause thermal wear of the gas turbine blades and nozzles. Contamination in the aft stages of the compressor may also indicate a degradation of performance. Some smaller gas turbines can be internally inspected by partial disassembly, but most gas turbines are now equipped with ports to allow borescope inspections. Documentation of the findings from borescope inspections is essential to good trending. Three basic methods are available:
El-025

ENGINE OPERATING HOURS -

Figure 2 I, Nonrecoverable Performance Degradation

ENGINE OPERATING HOURS -

Figure 22. Total Performance Degradation

l

Hand-drawn depictions supported by written description Videotaped inspections Photographic documentation

optimize gas turbine performance, efficiency, and exhaust emissions. Cleaning the gas turbine s compressor section with water and appropriate solvents, while cranking the gas turbine on the starter (or at low idle speed), is a proven method to recover most lost performance and is the preferred method of ingestive cleaning. However, if on-line cleaning (cleaning at normal operating speed) is to be done, it is essential to adhere to the recommendations of the gas turbine manufacturer. Computer performance analysis programs allow monitoring of nonrecoverable performance loss over time. This can be valuable in timing overhauls to control life-cycle costs. As with other forms of predictive maintenance, the absolute values calculated by the programs are not significant nor are their absolute relationships to

Although hand-recorded documentation can be effective, it is subject to individual interpretation. Today, the trend is toward optically recorded documentation. Systems are available for videotaping inspections as they are performed. Their primary disadvantage is the need to view much unnecessary footage to arrive at the few discrepancies which require monitoring and trending. Still-photographic documentation solves both problems by being objective and by focusing only on the problem areas requiring attention.

Monitoring Miscellaneous Parameters

Experience shows that trending a variety of miscellaneous parameters can also be a valuable part of a condition monitoring program. Although

81-13

these vary from machine to machine, some useful examples are: Lube Oil Tank Back Pressure - Trending this parameter gives early warning of worn seals. Seal air moving across the worn seal flows with the oil into the tank and slightly raises tank pressure. It can also indicate the need to service a dirty or clogged oil-mist precipitator in the oil tank vent. Filter Differential Pressures - These offer an excellent method of monitoring the need to change various filters. Oil Consumption - Trending oil consumption can help detect a faulty seal-oil separator on a compressor set. Gas Turbine Inlet Temperature Spreads On machines with multiple thermocouple indications, monitoring the temperature spread between thermocouples can be useful in detecting fuel injector or combustor problems.

The future for combined control/condition monitoring systems is promising. They offer great flexibility for enhancements, such as on-line FFT vibration analysis and trending.

SUMMARY
A quality maintenance program has a significant impact on the reliability and life-cycle cost of the gas turbine package. Varying applications, operating environments, and duty cycles make it difficult to develop a single program with set schedules to meet every operator s needs. Operator goals also vary widely, but generally fall into the four categories of availability, production, efficiency, and operating cost. There are many steps in developing a maintenance program, including determining what to maintain, and when and how to do it. The complexity of a gas turbine package requires a quality maintenance program that includes a mixture of unscheduled, scheduled, and on-condition maintenance. Selecting an approach to maintaining a given component of a gas turbine package should be driven by criticality, relative costs, and accessibility. Other variables need to be taken into account, such as operating experience and conditions. The use of trending technology allows operators to optimize their programs with more oncondition maintenance. Trending also provides more information to track the results, to measure changes in operating and environmental conditions, and to identify adjustments required in maintenance schedules. Great care must be taken to ensure that consistent, high quality data are taken whenever trending technologies are applied. Failure to do so will result in misleading trends that could leave an impending failure undetected or result in unnecessary maintenance. With proper planning and the application of today s maintenance technologies, a quality maintenance program can be assured. A thoughtfully designed program can achieve all of the operator s diverse maintenance goals.

INTEGRATED GAS TURBINE CONTROL/ CONDITION MONITORING SYSTEMS

The advent of control systems based upon programmable logic controllers offers the opportunity to expand upon and automate the trending of a variety of parameters. The latest generation of such systems offers combined gas turbine control and condition monitoring capabilities extending to the driven equipment and other process controls (DeMoss, 1992). They allow monitoring of the gas turbine package on a realtime basis, and allow the past history of a variety of parameters to be called up and displayed. Several parameters can be recalled and shown in strip-chart format. Also the systems can be used to trend gas performance and to visually display a gas boost compressor operating map, showing the current operating point of the package. Predictive emission monitoring is incorporated into some such systems (Hung, 1992). These systems cannot perform all of the indepth analysis and trending that are needed in a quality maintenance program, but they can certainly enhance such an effort. For example, they do not perform sophisticated vibration analysis, such as FFT spectral analysis or orbit analysis. Oil samples for spectrochemical analysis must still be taken and periodic borescope inspections are still required.

ACKNOWLEDGEMENTS
The authors would like to gratefully acknowledge the assistance of Mr. Gerhard Stich, Mr. Herbert Blach, and Mr. Herbert Rohrbacher (OMV Aktiengesellschaft) for their kind assistance in preparing this paper.

81-14

REFERENCES
Chandler, A.L., 1984, Turbomachinery Maintenance Planning , TTS8, Turbomachinery Technology Seminar, Solar Turbines Incorporated, San Diego, California. DeMoss, S.H., 1992, Gas Turbine Control Enhancements and Networks, TTS72, Turbomachinery Technology Seminar, Solar Turbines

Hsu, L.L, 1989, Gas Air/Fuel/Water Management, TTS54, Turbomachinery Technology Seminar, Solar Turbines Incorporated, San Diego, California. Hung, W.S.Y., 1992, Predictive NOx Monitoring System: An Alternative to In-Stack Continuous Emission Monitoring, TTS83, Turbomachinery Technology Seminar, Solar Turbines Incorporated, San Diego, California. Odom, F.M., 1989, Optimizing the Efficiency of Gas Compressor Packages, TTS57, Turbomachinery Technology Seminar, Solar Turbines Incorporated, San Diego, California.

Incorporated, San Diego, California.

Hewlett-Packard Company, 1983, Effective Machinery Maintenance Using Vi bration Analysis, Application Note 243-1, Hewlett-Packard, San Jose, California.

81-15