835 Innovation Drive Knoxville, TN 37932 Phone: (865) 675-2110 Fax: (865) 675-3100 Customer Support: (865) 671-4274
DoctorKnow® Applicati Application on Paper Title: PeakVue as Part of a Reliability Reliabili ty Based Maintenance Program Source/Author: Joe Dominick Product: Data Collector/Analyzer Collector/Analyzer Technology: Vibration
Introduction A vibration analyst's main job is to provide accurate, timely information on the health of the machinery that is being monitored. As experience shows, this is not always an easy job. Therefore, any tool that can be used to assist the analyst in providing this information should be made readily available. Unfortunately, we do not always have the resources (i.e. time and money) available to utilize these tools or we do not understand the capability of the equipment we have. This paper was written to t o address this problem with a relatively new tool in vibration analysis, Peak Vue. Vibration analysis is one part of a Reliability Based Maintenance (RBM) philosophy that, along with other technologies such as thermography and tribology, can provide a condition based assessment of the plant or mill machinery. Peak Vue is designed to play a significant role in the route based vibration surveys of an RBM program without taking a large time or capital expenditure to implement. The goal of this paper is to provide the reader with an understanding of what Peak Vue can do to help in making more accurate vibration calls. This is done in two parts: first, a basic background of what Peak Vue is and how to implement it is discussed and second, data is presented to demonstrate the capabilities of Peak Vue and to get the reader familiar with looking at the information it provides. What is Peak Vue? Peak Vue is a method of vibration signal analysis that captures the peak value of the time waveform over a defined time interval. These peak values can then be processed, like routine instantaneous values from the waveform, into a frequency based spectral plot. This type of signal processing is directed at picking up and monitoring short duration stress waves. These stress waves are primarily produced in faults where metal-to-m etal impacting is present. Gear defects, bearing defects, and abrasive wear are examples of faults where stress waves are produced. Peak Vue can be set up similar to demodulation in intermediate frequency bands (6000 to 30,000 rpm). Unlike demodulation, Peak Vue maintains the true signal levels that are occurring in the impacting events. Thus, Peak Vue is a more trendable parameter. Also, Peak Vue is not limited to the intermediate frequency range. It is applicable to impacting faults in both high speed and very slow speed frequency ranges.
Where Should Peak Vue be Used? As with any other tool, you need to recognize where it should be used and how much it is going to cost to use it. Fortunately, equipment cost for Peak Vue is low compared to other predictive maintenance tools, in fact, it may already be a part of your data collection hardware and software. The labor required involves taking one additional data point on a machinery bearing. With CSI equipment, Peak Vue data is collected using your analyzer and a typical accelerometer and can be added to route based data collection. Whether a formal or informal approach to setting up an RBM program is used, Peak Vue should be applied wherever impacting defects can be found. Additionally, Peak Vue shows its strengths on slow speed machinery. In you have laid out a formal RBM program, you may have analyzed different failure modes and criticalities of your machinery and operating systems. From this information, you can identify where metal-to-metal impact failures are the most critical and then focus the manpower to take Peak Vue data. If your RBM program does not have a formal blueprint, a general rule of thumb would be to take Peak Vue data on critical gearboxes and on critical motors and rolls with anti-friction bearings. One Peak Vue measurement per bearing should be taken with the transducer placed near the load zone.
How to Implement Peak Vue A Peak Vue measurement point is set up very similar to a routine vibration measurement
point. The main difference is that with Peak Vue you must select a high pass or bandpass filter setting.
Figure 1 - AP set-up screen in MasterTrend. Setting up a data point using MasterTrend software is relatively easy. A Peak Vue reading must be set up to be taken in acceleration. The remainder of the set-up is done in the Analysis Parameter (AP) Set-up portion of the database manager. Make certain to set the maximum frequency (Fmax) of your spectrum to capture faults as you normally would with a routine vibration reading. For example, some guidelines are: on gearboxes set the Fmax to capture three times gearmesh, or on anti-friction bearing set the Fmax to capture five orders of the inner raceway defect frequency. Figure 1 shows the Signal Processing Parameters menu of the AP Set-up Menu. Here, you need to check the Enable Peak Vue box, set the Peak Analysis Method to Peak Vue, and select an appropriate filter. Selecting the appropriate filter may take some practice. As Figure 2 shows, there is a large selection of high pass and band pass filters to choose from. The general rules to keep in mind when selecting a filter are: · Use bandpass filters to capture modulations and excitations of resonance frequencies. · Use high pass filters to detect stress waves given off by faulty bearings or gearing. · Select a filter equal to or higher than your defined Fmax. · Keep in mind the mounting methods that will be used when collecting the data.
Figure 2 - Peak Vue filter selections. Regarding the last rule, band pass filters and the 500 and 1000 Hz high pass filters can be used with magnet mounts. It is recommended to use a flat rare earth magnet on a clean, flat surface if the selected filter is 2000 Hz high pass. Stud or adhesive mounting should be used with the even higher frequency filters. For general plant machinery when looking for bearing and gearing defects, the1000 Hz high pass filter works effectively. However, it is advisable to first test different filter settings on the same piece of equipment to get an idea of what data you will be able to capture. An optional parameter that can be changed for Peak Vue analysis is the time waveform. In particular, on slow speed machines, Peak Vue can clearly show impacting in the time waveform. Analysis can then be enhanced by taking a longer period waveform. This set-up is done in the Waveform Parameters portion of the AP Set-up Menu, as shown in Figure 3. The waveform should be set up to capture and view data in acceleration units. Select the Special Time Waveform Option and then enter the frequency span and number of waveform points to collect the waveform record you want to store. The equation to keep in mind here is the waveform time is equal to: (# of waveform points) / (2.56* Maximum Frequency).
Figure 3 - Waveform Parameter set-up menu. Interpreting Peak Vue Data After looking at a some Peak Vue spectrums and waveforms, the vibration analyst will note some similarities and differences from conventional vibration data. One of the differences that will immediately jump out at you is the truncated, or one-sided, time waveform. A good example of a Peak Vue waveform is shown in Figure 4. This data was taken on the output shaft of a steel coil wrapper gearbox. The frequency of the peaks in the waveform are at output speed of the gearbox. This impacting is primarily due to excessive clearances in the mandrel bearing. The waveform appears one-sided because the Peak Vue process stores the maximum peak value in each sampling interval instead of the instantaneous value. The peak values are then plotted, by convention, on the positive side of the waveform plot. Other differences can be seen be looking at the spectrum in Figure 5. This plot is from the previous waveform taken on the mandrel gearbox. First, note the amplitude scale of the plot. Peak Vue is designed to be looked at in acceleration units. The amplitude value may look extremely low, especially for low frequency events as seen in the example spectrum. Second, note the spectral content of the plot. The "mound" in the 0 to 1000 rpm frequency range is actually made up of discrete peaks spaced apart at output speed. However, what is important to note is what is not showing up in the plot - typical gearmesh frequencies. Remember, Peak Vue is detecting stress waves and impacts. If the gearing system is healthy and is undergoing rolling and sliding mechanisms, gearmesh will not appear. In the case of the mandrel gearbox, gearmesh (4950 rpm) typically appears at low levels in the routine vibration spectrum. However, in the Peak Vue plot, since there are no mechanical problems with this gear set, it does not show up. This is important to remember when
looking at some types of faults, such as imbalance, where impacting may not be occurring.
Figure 4 - Example of a Peak Vue waveform.
Figure 5 - Example of a Peak Vue spectrum. After getting used to looking at the data, performing analysis with Peak Vue is easy for an analyst who is familiar with looking at conventional vibration data. The remainder of this paper focuses on how some different faults show up in Peak Vue data. Bearing Defects Peak Vue is an excellent tool for early bearing defect detection. Most bearing failure scenarios start with defects ringing high frequency stress waves. This type of failure analysis is where Peak Vue can make an immediate impact in your RBM program. Figures 6 and 7 show a good example of how Peak Vue can detect early stage defects.
Routine vibration data on a 300 hp temper mill bridle motor showed frequencies that corresponded to an outer race defect on the motor bearings. Figure 6 shows this spectrum. When Peak Vue readings were added to the routine data collection route, it was noted that additional defect frequencies were present in the vibration spectrum. Figure 7 shows the Peak Vue data, note the presence here of outer and inner race defect frequencies. The routine vibration data was not picking up the inner raceway defects, possible due to the fact that the defect signal was damped out before it reached the transducer or it could not detect the stress waves. Based on the additional information provided by the Peak Vue data, it was decided to replace the motor bearing during a planned outage. The pulled bearing revealed defects on both of the raceways as Peak Vue had predicted. Bearing defects on raceways are somewhat easy to diagnose. Cage and roll ing element defects are harder to diagnose because not much vibration energy is transmitted from these types of defects. An even harder problem with these defects is estimating how much life is left in the bearing. Sideband analysis is a good tool to aid the analyst in answering these questions. Peak Vue provides us with another tool to help diagnose these types of defects and help indicate their severity. Figure 8 shows routine vibration data recorded on the inboard bearing of a large exhaust fan. This plot shows some 1x rotational speed vibration. Figure 9 shows Peak Vue data taken on the same bearing. This plot showed multiples of cage defect frequency and inner raceway defects with sidebands of 1x rpm.
Figure 6 - Routine vibration spectrum on mill motor.
Figure 7 - Peak Vue spectrum on mill motor.
Figure 8 - Routine vibration data on large fan. This data prompted the vibration analyst to recommend that the bearing be replaced at a planned outage. The pulled bearing showed some "normal" wear on the inner race, but more significantly, a couple of the rolling elements were severely spalled. The Peak Vue waveform, see Figure 10, from this example gave additional clues to the severity and cause of the problem. Here, the frequency of the peaks in the plot are equal to the FTF defect frequency of the bearing. This energy was not detected in the routine vibration readings and the defect could have been allowed to propagate. When bearing degradation is identified as a primary or critical failure mode, routine Peak Vue data can be used to trend and aid in picking out early faults. In one instance, bearing defect harmonics were only able to be meaningfully trended with the Peak Vue data. Figure 11 shows a routine spectrum taken on a gearbox bearing on a tandem mill drive. For three years the mound of energy had been monitored in the spectrum. This mound increased and decreased in energy from reading to reading. T hinking that this energy indicated an opening of bearing clearances, Peak Vue readings were added t o the data collection route. This data finally allowed the analyst to get a better understanding of the energy being transmitted from the bearing defects. Figure 12 shows the Peak Vue data from the gearbox measurement point. Here, impacting defects from the outer race stand out and become more trendable. Peak Vue made it easier to watch for sideband and harmonic development.
Figure 9 - Peak Vue spectrum on a large fan.
Figure 10 - Peak Vue waveform on large fan.
Figure 11 - Routine vibration reading on gearbox.
Figure 12 - Peak Vue reading on gearbox.
Gearing Defects Peak Vue is an excellent tool f or assisting in diagnosing gear set problems. It is particularly impressive to see how it picks up gear teeth that are developing cracks. Stress waves are given off when a tooth begins to bend or deflect due to a high area of stress concentration. These short duration waves are picked up by the Peak Vue data collection. In one instance, Peak Vue was able to help prevent a day of lost production time on a tin plating line. In fact, the Peak Vue readings showed a dramatically different picture of what was going on inside the gearbox than what the routine vibration data showed. Figure 13 shows this routine data collected on a double reduction, precision tension bridle gearbox. Because of access restrictions, data can only be collected on the input shaft bearings. The analyst had been watching the input meshing frequency and harmonics. While the amplitudes were not very high, "skirts" had been present around the gear mesh frequencies. It had been recommended to visually inspect the input gear set for any signs of accelerated wear, but no inspection was made. Once a Peak Vue measurement was added to the data collection route, the urgency of the analysis was changed. Figure 14 shows this Peak Vue reading taken on the input shaft bearing. What was predominant in both the spectrum and waveform were peaks at frequencies of output shaft speed. The amplitude of the waveform was very shocking, especially since the vibration reading was taken about three feet away from the output shaft. The analyst recommended that the output gear set be replaced as soon as possible. That weekend, the gear set was replaced and two teeth had visible cracks developing around the root. T he input set was found to be worn and showing signs of misalignment. Looking at the data in Figure 13, the analyst would have been hard pressed to call any problems with the output gear set, let alone any serious problems. The additional information provided by the Peak Vue reading allowed the analyst to make an accurate, timely call on a serious problem.
Figure 13 - Routine data on bridle gearbox.
Figure 14 - Peak Vue data on bridle gearbox. Electrical Defects One surprising benefit from adding Peak Vue measurement points on our routine routes, was the ability to diagnose rotor related electrical defects. A can of worms was opened up when Peak Vue points were added to routes at a large steel manufacturer. The data showed 2x line frequency on a number of AC induction motors. This frequency was present in the Peak Vue readings but not in the routine vibration readings. Why was it showing up? Does it indicate a severe problem? It seemed as if all of the analysts had stumbled on the same problem at once and needed some feedback on what corrective actions should be taken. To get a better feel for what we were getting into, a motor was located (350 hp, ball bearings) that we were pretty certain had rotor related electrical defects and was gong to be sent out for a rewind in the near future. Figures 15 and 16 show the routine and Peak Vue data collected on this motor. The routine data had been showing rotor bar passing frequency, but as Figure 15 shows, not much else in the lower frequency region. The Peak Vue reading, on the other hand, showed lots of activity in this region. As seen in the plot,
many harmonics of rotational speed were present with sidebands of slip times the number of poles (pole pass frequency). This indicated that there were some broken or open rotor bars. The motor was sent out for rebuild and was found to have five broken rotor bars. Subsequent findings have lead us to use one rule that we have been following when encountering electrical problems with Peak Vue: watch for the development of pole pass frequency sidebands around rotational speed harmonics. The number and size of the sidebands seem to indicate a more severe problem. In one instance, the pole passing frequency was actually picked up in the Peak Vue reading. Routine vibration data taken on a 1000 hp induced draft fan motor continually showed 2x line frequency vibration, see Figure 17. The amplit ude was very low and did not trend upwards over a two year period. Figure 18 shows a Peak Vue reading taken on this motor. This data, though low in amplitude, showed pole pass side bands to be higher than the rotational speed peaks. Also, note the presence of the fundamental pole pass frequency. This motor was run until a scheduled outage, 4 months later. Upon inspection, the rotor showed 17 broken rotor bars.
Figure 15 - Routine vibration data on a motor.
Figure 16 - Peak Vue data on the same motor.
Figure 17 - Routine data on ID fan motor. Currently, we are simultaneously taking Peak Vue, motor current, and flux data on a group of AC induction motors to try to correlate severity of the problem with the frequency and amplitude content in the Peak Vue reading. Slow Speed Machinery Peak Vue also works very good on machinery with slow turning speeds. Some good examples would be large paper roll bearings, galvanizing furnace rolls, or hoist gearboxes on cranes. On these types of equipment, faults are difficult to detect with traditional vibration measurements, even with high sensitivity transducers. In one manufacturing facility, roll bearing failures in a galvanizing furnace were becoming unpredictable. Two bearings had failed with no advance warning from the routine vibration data. Good data was being taken: raceway defects were observed on some bearings and slow speed technology was providing a clear picture of turning speed vibration, however, the slow speed of the rolls (10 to 25 rpm) made the failures undetectable. To address this problem, the data collection interval of the route was changed to bi-monthly and one Peak Vue reading was added to each bearing. After one month of collecting this data, a bearing failure was predicted, solely based on the Peak Vue data. Figure 19 shows the routine vibration reading acquired on the bearing in question, while Figure 20 shows the Peak Vue data. The routine vibration plot did not show any changes in the narrowband trends. However, the Peak Vue data showed a twenty times increase in the overall energy level. The spectrum showed many harmonics of turning speed, 15 rpm. The waveform showed an event once per revolution. The roll bearings that were failing did not go through a traditional failure scenario, but were destroyed in a short time by excessive heat and loading that caused the bearing clearances to open. Once a roller became skewed and started to do more than roll, stress waves were produced that Peak Vue was able to detect.
Figure 18 - Peak Vue data on ID fan motor.
Figure 19 - Routine trend and spectrum on roll.
Figure 20 - PeakVue data on roll bearing.
Conclusion Hopefully, looking at these "real world" examples will give some idea of how valuable a tool PeakVue vibration data can be to an analyst. Using this tool in conjunction with routine vibration data provides a more complete picture of the health of the machinery. In fact, some defects on gearing and slow speed bearings presented in this paper would not have been seen in a timely manner without the PeakVue data. Thus, adding this type of measurement to an RBM toolbox has proven to be an effective means of making sure plant and mill machinery stay productive. Reference "Peak Value Analysis" , Dr. James Robinson and Dr. Kenneth Piety Acknowledgments The author would like to thank everyone at the CSI office in Weirton; especially Joe Bauman, Rodney Dragonir, Dan Finney, Jerry McDonald, and Tim Rawson for helping supply information for this paper. All contents copyright © 1998, Computational Systems, Inc. All Rights Reserved.
