a

Transcription

1 THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47th St., Now York, N.Y The Society shall not be responsible for statements or opinions advanced in papers or discussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. Discussion Is printed only if the paper is published in an ASME Journal. Papers are available from ASME for 15 months after the meeting. Printed in U.S.A. Copyright 1994 by ASME 94 GT-472 VIBRATION CHARACTERISTICS OF LARGE FRAME COMBUSTION TURBINES Bill J. Coslow and Oran L. Bertsch Power Generation Business Unit Westinghouse Electric Coporation Orlando, Florida a ABSTRACT This paper examines the vibration characteristics of large frame land-based combustion turbines including response of the entire rotor train. The typical rotor train consists of the combustion turbine with integral axial compressor, a rigidly coupled generator, and a starting and auxiliary turning gear package connected through a flexible coupling. Vibration characteristics during both transient and steady state conditions are considered. Analytical methods for predicting synchronous and asynchronous vibration response are examined. Actual shop and field test results am discussed including vibration components at frequencies other than running speed. Industry standards on vibration limits are discussed with respect to current Westinghouse vibration control setting specifications. INTRODUCTION The vibration of large frame combustion turbines is of interest to plant operators primarily in regard to providing automatic protection of equipment from damage due to accidental unbalance. For the operators, a low level of operating vibration means that automatic alarm and trip levels may also be set low, thereby providing optimum protection of equipment. Original equipment manufacturers (OEMs) also concern themselves with the potential effect of vibration on the fatigue life of turbine components. Minimizing operating vibration levels on large frame combustion turbines requires an understanding of the characteristic vibrations exhibited by these machines. Transient vibration, both short and longer term, must be considered along with asynchronous vibration, in order to minimize steady state operating vibration. TYPICAI ROTOR TRAIN I AYOUT A typical rotor train layout is shown in Figure I. The flexible coupling between the generator and starting package provides an almost complete uncoupling of the bending moments at this connection. The starter motor is disconnected except during early VIBRATION i MONITOR (TYP) BEARING 01 BEARING 02 BEARING 03 BEARING 04 BEARING Mb BEARING 06 Figure 1 TYPICAL ROTOR TRAIN LAYOUT Presented at the International Gas Turbine and Aeroengine Congress and Exposition The Hague, Netherlands June 13-16, 1994

2 stages of the startup. Locating the generator at the cold, thrust bearing end of the combustion turbine minimizes any problems that would otherwise result from axial thermal expansion as the combustion turbine reaches operating temperature. The rigid connection between the generator and combustion turbine provides a high degree of coupling of the inherent vibration modes of the two machines. The combustion turbine casing is supported at two locations, one near the first stages of the compressor section and the other near the last stage of the turbine section. The compressor support is highly rigid in all directions while the turbine support is pinned and allows for laterally guided axial thermal expansion of the casing. The rotor journal bearings overhang the casing supports such that the rotor weight loading tends to counteract the natural sag of the casing. The rotor thrust bearing is located upstream of the compressor journal bearing and is capable of handling thrust in either direction. VIBRATION MONITORING SYSTEM A good vibration monitoring system (VMS) is essential on large frame combustion turbines. Furthermore, due to the high degree of dynamic coupling between the generator and the combustion turbine, an equivalent system should also be provided for the generator. Clearly the most important function provided by the VMS is to protect the combustion turbine and generator from excessive vibration by initiating automatic alarm or shutdown (trip) actions. The standard VMS includes two dual-sensing probes at or near each journal bearing of the combustion turbine and generator. X-Y PROBE LAYOUT Figure 2 shows a schematic of the typical X-Y probe layout which is provided near each journal bearing. The two probes are mounted 90 0 apart with each positioned 45 from the vertical centerline of the shaft. The 45 positioning is used rather than vertical and horizontal positioning, due to the fact that bearing housings typically contain horizontal flanges which would interfere with installation of the horizontal probe. By convention, one of the probes is labeled "X" and the other "Y". Maintaining the same convention throughout all four bearings of the generator and combustion turbine facilitates the evaluation of rotor train vibration modes. X PROBE Y PROBE a single probe at each bearing which, though less expensive, did not permit any extensive vibration diagnostics and risked damage to machines caused by high vibration in a direction perpendicular to the single probe. A circumferential reference point, for the purpose of defining phase angles, is provided by a machined slot in the shaft with a separate probe which senses the slot and triggers the phase measuring circuitry once per revolution. 0 Dual Probes Complete characterization of rotor vibration at a journal bearing requires definition of both relative and absolute vibrations. Relative rotor vibration is defined as the vibration of the shaft with respect to the bearing housing. Seismic vibration is defined as the vibration of the bearing housing with respect to ground. Absolute rotor vibration is the vector sum of these two and represents the vibration of the shaft with respect to ground. Relative rotor vibration is measured using a proximity probe which senses changes in the gap between the bearing housing and the shaft. Seismic vibration of the bearing housing is measured using a velocity probe which senses the absolute motion of the bearing housing with respect to ground. Both probes are mounted on the bearing housing as a single unit which is referred to as a dual probe. By summing the proximity probe reading with the integrated velocity probe reading, the absolute motion of the shaft can be determined. 5 TYPICAL VIBRATION RESPONSE A typical response of a large frame combustion turbine rotor during startup and operation is shown in Figure 3. This figure shows relative rotor vibration amplitude versus rotor speed and time from startup through several hours after reaching base load operating condition. START-VP TRIP START-UP ALARM LEVEL RANGE OF 151 CRITICAL SPEED II RANGE OF End cnincal SPEED AillTYPICAL VIBRAT ION C HARACTERIS11CS RUNNING SPEED TRIP LEVEL RUNNING SPEED ALARM LEVEL Figure 3 TYPICAL ROTOR RESPONSE Figure 2 X-Y PROBE LAYOUT The X-Y probe layout allows a complete definition of the shaft vibration and position at all times. Older systems relied upon Transient Vibration Two types of transient vibration occur during startup and shutdown of a combustion turbine. The first is simply the change in rotor vibration which occurs as the rotor speed is accelerated through a resonant speed. The second is of longer duration and is referred to as thermal transient vibration. 2

3 II I se.to I The combustion turbine rotor typically passes through several resonances while coming up to the operating speed. A generator rotor resonance also may become evident at the combustion turbine vibration sensors. Elevated vibration levels, as the rotor passes through resonances, are not normally a concern due to their short duration. Because of this, alarm and trip levels are typically increased during startup. These levels are reduced after passing through the highest significant rotor resonance. During a normal shutdown, the levels are increased, prior to reaching the highest significant resonance, as the rotor speed is lowered. It should be noted that rotor response can be significantly different during deceleration than during acceleration on a large frame combustion turbine. Higher amplitudes are often observed on deceleration than during acceleration due to the fact that acceleration rates are typically higher at the lower rpm values than deceleration rates. More time spent at and near resonance, due to the lower deceleration rates, allows the response to build higher. Thermal Transient Vibration, After startup, as the large frame combustion turbine rotor begins to heatup, a phenomenon known as thermal transient vibration occurs. This vibration is due to bowing of the rotor caused by asymmetric thermal gradients in the rotor during the heatup cycle. The amplitude of thermal transient vibration can be minimized by utilizing care in the design and manufacture of axial joints and shrunk-on disks in the rotor. After the rotor reaches steady state operating temperature, the transient vibration levels stabilize. The amount of time necessary for a rotor to reach steady state operating temperature, after a cold start, may take several hours. 0 FULL SPEED NO LOAD frours EXHAUST BEARING X DIRECTION 1111WrBASi E LOAD BL rithamr + I HOUR Figure 4 POLAR PLOT OF TYPICAL THERMAL TRANSIENT Figure 4 presents a polar plot which displays the typical amplitude and phase behavior of a large frame combustion turbine after reaching full speed, as load is applied, and for several hours thereafter. The polar plot is made by plotting the relative peak to peak synchronous vibration amplitude versus phase angle. The amplitude of the thermal transient can be considered to be the length of the longest amplitude - phase vector which can be drawn on the polar plot from the point of reaching full speed at no load to any other point up to reaching the end of the transient. In the example shown in Figure 4, the transient turns back on itself before ending. The amplitude of the thermal transient in this figure is the distance from A to B in mils (thousandths of an inch) even though the transient doesn't end until point C. Each rotor of any combustion turbine, large or small frame, has its own signature thermal transient which is a characteristic of the specific structure and clearances of that rotor. The shape of the thermal transient, as shown in the polar plot., cannot be changed by balancing. The position of this curve within the polar plot can, however, be changed by shop balancing subsequent to full load testing, and by field balancing. Steady State Vibration Steady state vibration on a large frame combustion turbine is usually reached within approximately one to four hours after a cold start. Startups of an already warm engine may reach steady state vibration very quickly and may exhibit very little thermal transient vibration depending on the temperature of the rotor at startup. The total rotor vibration contains several frequency components. The response at running speed, sometimes called synchronous or filtered response, is often referred to as simply the IX response. For a combustion turbine and generator set running at 3600 RPM, the IX frequency is 60 Hz. Similai -ly, 2X refers to twice running speed, or 120 Hz in this case. The IX vibration represents an unbalance in the rotor and in most cases dominates the overall rotor response. The IX vibration can be minimized by balancing. The 2X vibration is usually small in comparison to the IX. The 2X vibration of the combustion turbine may become more significant when the attached generator is a dipole design. Large dipole generators contain an inherent asymmetry in the bending stiffness of their rotors. This is due to the fact that a significant amount of the generator rotor bending stiffness comes from the asymmetric windings and winding support structures. The result is two bending stiffnesses, separated by 90 0, which cause the generator rotor to undergo two cycles of gravity induced vibration for each revolution. This can be easily visualized by imagining the center motion of a simply supported wooden 2 X 4 beam, under gravity loading, as it is rotated through one revolution. The shaft of the generator rotor is machined (compensated) to minimize this effect. Determining the amount of compensation necessary requires careful determination because the stiffness of windings and winding support structures will change as the generator rotor speed increases. An incorrectly compensated generator rotor will exhibit excessive 2X vibration and could exhibit a 2X "rocking" motion as shown in Figure 5. This rocking motion can be visualized by imagining the effect of cutting the 2 X 4 beam from the above example at the center, rotating the two halves 90 0 with respect to each other, then reconnecting them. Balancing of either the generator or the combustion turbine will not reduce the amount of 2X coming from an incorrectly compensated generator rotor. The effect of this 2X vibration on the combustion turbine can be magnified because the combustion turbine generally has more flexible supports. ANALYTICAL METHODS OF PREDICTION The tools used for predicting the response of rotor systems range from general purpose finite element codes, to dedicated rotor 3

4 INCORRECT COMPENSATION SEARING #3 II BEARING #4 \ / CORRECT COMPENSATION entrapped debris. The probability of developing an analytical solution for this type of phenomena is very low in most cases. The method of analysis for all of these cases is essentially the same. The shaft system is evaluated using a sensitivity study or parametric review to characterize the observed vibration with a set of shaft unbalances, system stiffnesses, and bearing configurations. A typical example is illustrated in Figure 6. After the first solution is developed, it may be refined by repositioning the unbalance load as well as changing the magnitude of the load. Several iterations may be necessary before correlation with the measured response is achieved. Figure 6 ILLUSTRATION OF ITERATIVE ANALYTICAL TECHNIQUE Figure 5 2X GENERATOR ROTOR VIBRATION SHAPE dynamic programs. The latter is capable of predicting the undamped critical speeds and unbalance response of the shaft system. In some cases these codes can accommodate flexible supports or may include gyroscopic and rotary inertia effects. These analytical codes require essentially the same input parameters such as shaft section properties and support and bearing stiffness and damping properties. Because of the inherently complex configurations involved in the nonnal equipment designs, a benchmark has to be established in order to quantify the accuracy of the fundamental properties of the model. This is accomplished based on simple assumptions that are either confirmed by test or through the detailed modeling techniques of specific components. The initial forcing functions evaluated consist of the standard expected residual unbalance, the sensitivity of balance moves, asymmetric gravity loads, and the unbalance of the shaft due to thermal bow. The abnormal forcing functions due to system distress or influences outside the rotor train am addressed as required by the particular instance. Analysis of Synchronous Vibration The running speed or (IX) vibration is normally the dominant component of the total vibration. In most cases this is directly related to shaft unbalance. The source of the shaft unbalance can be further defined by the magnitude and stability of the vibration. Relatively high vibration levels, that are inherently stable and do not change with time or load, will normally be controlled by balance moves. The flexible shaft systems of large frame combustion turbines require multi-plane balancing. Analytical balance sensitivity studies provide a direction as to which balance moves would be most effective before a rotor train specific calibration move has been made. Unstable IX vibrations fall into two categories: repeatable and non-repeatable. Unstable but repeatable vibrations can sometimes be directly related to the shaft configuration or the relaxation of the shaft with the increase in run time and number of starts. The analysis of this category of vibration requires detailed thermal mechanical analyses to determine the expected asymmetric thermal gradients in the system. Once the expected thermal gradients are determined, the resulting thermal bending moments in the shaft can be calculated and applied to a finite element shaft model to determine the rotor response. Unstable and non-repeatable vibrations are more difficult to diagnose because the apparent source can not always be identified in a single run. These vibrations are normally associated with some nonlinearity in the system such as loose components or AMPLITUDE a. MEASURED RESPONSE SPEED (krpm) SPEED (PPM) K..5 K=.75 K. 1.0 we K=1.5 6b. SUPPORT STIFFNESS STUDY Analysis of Asynchronous Vibration The off running speed vibrations (non-1x) cover a wide spectrum of frequencies and sources. These asynchronous vibrations are not controlled by balance moves and therefore in some cases require a significant amount of analysis,, test, or equipment inspections in order to diagnose and establish the most effective method of vibration control. Normally these vibrations are evaluated with a harmonic response analysis using a general purpose finite element code. The forcing function must be determined based on the frequency of the observed response and the amplitude is some nominal value that can be scaled if the system remains essentially linear. The point of load application 4

5 will depend upon the expected source which can be internal to the rotor system such as rubs, asymmetric rotors, gravity bending, or partial admission loads, etc. The establishment of an accurate system model is very important because small exciting forces at or near one of the system critical speeds can result in a resonance and increased levels of vibration. 6c. BEARING STUDY.. TILT PAD -SLEEVE Large frame combustion turbines typically have access to three balance planes for field trim balancing. These balance planes are located one at mid span, and one near each of the two journal bearings. This provides the ability to consider the first three rotor vibration modes in determining where to add balance weights. Large frame combustion turbines typically have no mom than two rotor modes below operating speed. Field trim balancing to minimize steady state vibration requires close attention to the thermal transient characteristics of each individual rotor. In addition, the type of service expected (peaking versus base load) may also influence balance methods. Minimizing steady state vibration implies balancing for the rotor in its fully heated and stable condition. There are times, however, when this may not be the most desirable strategy. An example of this is shown in Figure 7. This figure shows a polar plot which contains two identical thermal transients. Although the shape and magnitude of the thermal transient is a characteristic of the rotor, the position of the transient can be controlled by field trim balancing. The solid curve represents a thermal transient which has been trim balanced for the hot, stable condition and would be advisable for a base load plant which will be started and run for days or weeks before shutdown. The dashed curve represents the same rotor balanced to "center" the thermal transient This curve might be advisable for a peaking plant which would be operated only a few hours daily. Centering the thermal transient results in higher steady state vibration but lower peak and average vibration for the first few hours of operation. iii SPEED (RPM) 6d. FIRST SOLUTION 1.0 ROTOR BAI ANCINn Rotor balancing on large frame combustion turbines is performed in the shop prior to installation of the rotor into the turbine casing. In the field, trim balancing of the combustion turbine (and sometimes the generator) can be performed to balance out effects of casing stiffnesses and dynamic coupling between the generator and turbine. Slum Balancing Shop balancing of large frame combustion turbine rotors is performed at relatively low speed (less than 1000 RPM) on rotor components, followed by high speed shop balancing at 110% of rated speed. Large frame combustion turbines typically have three to five balance planes in which balance weights can be installed. During shop balancing, all of these planes are accessible. Field Trim Balancing Even though shop balancing usually results in a well behaved rotor system, it is sometimes necessary to field trim balance the rotor system in order to reduce vibration even further. Figure 7 VIBRATION POLAR PLOT SHOWING TWO METHODS OF TRIM BALANCING USING TEMPORARY INSTRUMENTATION Up to this point, all discussions have referred to vibration as measured by the installed VMS on the combustion turbine and generator, and it is seldom required to use anything other than the VMS. Although this system can provide a lot of vibration information, it provides little information on casing or turbine support vibration. Special testing, using temporarily mounted 5

6 instrumentation, can be used to cross check the accuracy of the VMS and to provide valuable additional information for understanding overall rotor system response. Special testing can also provide data to support analytical evaluations. checking the VMS The typical vibration monitoring system is shipped from the manufacturer in a calibrated condition. Sensors are mounted on the combustion turbine and generator with the signal conditioning equipment connected and tied into the control system. The following is a relatively simple method for perfotming a cross check of the VMS. Install a temporary accelerometer on the journal bearing housing, at the same location as the VMS sensor channel to be checked. If this exact location is inaccessible, care must be taken to correct for vibration direction and the possible rocking vibration of the bearing housing about the bearing housing support centerline. Because the temporary accelerometer will measure only the seismic portion of the vibration as defined earlier in this paper, the output of the VMS must be converted to the seismic component prior to comparison. VMS systems typically display relative and absolute vibration response at each probe location. By using these readings, along with phase difference, to calculate the seismic component of response, both the output of the proximity probe and velocity probe are cross checked. Comparison between the VMS and temporary accelerometer should be made based on simultaneous data recordings. If the temporary accelerometer is not mounted in the same axial plane as the VMS transducer, two temporary locations should be used in order to estimate rigid body rocking of the bearing housing. "Bumo" Testing One method of determining casing vibrational modes, which is relatively simple to use, is sometimes called a bump test. An accelerometer block containing three orthogonally oriented accelerometers is moved around the combustion turbine to various points of interest. After temporarily mounting the block at each point, the casing is struck, using a specially instrumented hammer, and the response measured. The output of both the hammer and the accelerometers is fed into a portable computer. By collecting data at several points around the casing and on bearing housings, it is possible to calculate and dynamically display casing and bearing housing vibrational modes on a computer screen. Mode shapes and vibration response spectra can also be printed out on hard copy. In most cases, this test must be performed on a cold turbine, and an appropriate correction to calculated frequencies should be made based on the change in material modulus at operating temperatures. Multi le Channel Data Acouisition Systems Casing, bearing housing, and turbine support response can, of course, be monitored during actual operation with a multichannel data acquisition system. In this case, however, high temperature accelerometers with insulated mounts (good up to 900 F or better) are required on the hotter sections of the casing. Simultaneous recording of all channels on one magnetic tape or disk allows correlation of casing and support vibrations all along the rotor train. Impedance Testing Analytical methods for predicting rotor vibration response are highly dependent on the stiffness values used for bearing supports. One test method which seems to provide reasonably good results in determining these values is impedance testing. Impedance testing involves the use of an electrodynamic shaker suspended near the journal bearing housing. The vibrating head of the shaker is attached to the bearing housing and a constant, sinusoidal load is input to the side of the housing. The frequency is swept through whatever limits are desired while maintaining constant load with the aid of a force transducer mounted between the shaker head and the bearing housing. Acceleration is also measured as close as possible to and in the same direction as the input force. An electrodynamic shaker weighing 50 pounds is large enough to handle a large frame combustion turbine. With the aid of a spectrum analyzer and output from the force transducer and accelerometer, it is possible to obtain an impedance plot showing the dynamic stiffness of the bearing housing with respect to ground. Impedance testing can be performed in both the vertical and horizontal directions. To more accurately obtain the operating dynamic stiffness, it is recommended that the rotor be spinning at a high enough speed to establish near-operational bearing oil film thickness while data is being recorded. VIBRATION LIMITS FOR LARGE FRAME COMBUSTION TURBINES The specification of automatic alami and trip levels for protection of large frame combustion turbines varies by OEM. There is little guidance in terms of industry standards. This is best, because it is the OEMs which know most about the fatigue life of the components within their machines. The purpose of alarm and trip levels is to protect the turbine against unexpected unbalanced conditions. One might incorrectly consider that alarm levels should be set at some minimal amount above the steady state vibration levels. This practice leaves little margin for the affect of parameters which may vary from day to day and may result in frequent operation "in alarm", thus defeating the purpose of alarms. Such parameters as load, steam or water injection ratio, and the amount of time the rotor has spent on turning gear prior to starting (to remove static bow) can influence the rotor vibration levels during startup and steady state operation. Alarm levels should be set at or slightly below the levels which would cause unacceptable component life if allowed to continue unchanged. During acceleration and deceleration through rotor critical speeds, the alarm levels should be set at or slightly below levels which would cause unacceptable component fatigue life if allowed to occur occasionally. Industry Standards Two industry standards which discuss vibration limits for large frame combustion turbines are ISO "Evaluation Criteria for Shaft Vibrations of Coupled Rotor Systems of Gas Turbine Sets" and API 616 "Type H Industrial Combustion Gas Turbines for Refinery Service". Part 4 of ISO 7919 deals with the special features required for measuring shaft vibrations on the coupled rotor systems of gas turbine sets and provides "evaluation criteria, based on previous experience, which may be used as guidelines." API standards are published by the American Petroleum Institute as an aid to the procurement of standardized equipment and materials. ISO specifies that, for turbines running at 3600 rpm, steady state vibration levels less than 5.9 mils are acceptable for long term operation and that automatic trip levels should be set at 8.7 mils. No limits are suggested for startup during which the rotor typically goes through several resonances. Limits are based on peak to peak relative motion between the shaft and the bearing housing at or near the journal bearings. Limits are based on the 6

7 highest vibration measured in either of two directions, perpendicular to each other (not the resultant vibration vector). This standard also states that transient vibration should be limited to 125% of steady state levels. The transient part of this standard is not realistic for large, single span, flexible combustion turbine rotors and the industry has generally taken exception to it. API 616 sets no limit for in-plant operation but specifies a 2.0 mil limit during the shop test. This is a test of the turbine rotor alone and does not include any interaction or amplification with the turbine bearing supports or the other 'parts of an actual complete rotor system. This limit is based on peak to peak relative motion of small, rigid rotors. Although an alarm on relative vibration is suggested, no limits on either alarm or trip levels are provided. Westinghouse Control Settings Westinghouse has established normal alarm and trip control settings based on analysis and operating experience. During steady state operation, alarm and trip levels are set at 4.6 and 6.0 mils respectively. During acceleration from 0 to 2988 RPM, these limits are doubled to 9.2 and 12 mils. This allows the rotor to go through several resonances without causing alarms: recognizes the short term duration of these vibrations, and assures that rotor acceleration during this period satisfies design criteria. These levels are based on peak to peak unfiltered absolute motion of the rotor. Fatigue analyses have been performed on critical components assuming operation at vibration levels higher than these. Results indicate that stresses remain well below endurance limits. SUMMARY Many things can affect the vibration levels of a combustion turbine. Some of these effects can be minimized by design. Others can be minimized by shop and field trim balancing. A general understanding of the vibration characteristics of these machines is necessary in order to minimize operating vibration. With proper input information, analysis is successfully used to assure separation between rotor system resonances and operating speed. Analysis is also used to estimate the initial field trim balance moves prior to getting actual effects data. A high quality dual probe vibration monitoring system is important for protection of these machines and to accurately determine levels of operating vibration. Special testing can be performed to supplement vibration data obtained by the vibration monitoring system. This can also provide valuable input data for analytical models. Setting of vibration limits should be based on design considerations and not merely set at some arbitrary delta above operating levels. REFERENCES ISO "Evaluation Criteria for Shaft Vibrations of Coupled Rotor Systems of Gas Turbine Sets", International Organization for Standardization, API 616 "Type H Industrial Combustion Gas Turbines for Refinery Service", Second Edition, American Petroleum Institute,