PEAKVUE SUMMARY PeakVue (also known as peak value) can be used to detect short duration higher frequency waves stress waves, which are created when metal is impacted or relieved of residual stress through by cracking. For example, if a rolling element in a bearing passes over a defect that may still be on the inner or outer race, the race will deflect and then spring back. This motion will create a stress wave. PeakVue measures the highest amplitude found in this waveform and holds it as the highest value during a waveform time length equal to one over the sampling rate. PeakVue also passes the waveform data through a high pass filter to remove the rotational vibration frequencies from the data. However, filtering is its only real similarity to demodulation. The high pass filter should be set equal to or above the conventional Fmax of the spectrum. Generally, the 1000-Hz high pass filter is a good choice. Be sure to set up a measurement point for this analysis method that does not integrate the data. In other words, work in sensor units. Again, there are several key things that should be remembered about the use of the PeakVue technology. Results from PeakVue not dependent on analysis bandwidth. The PeakVue data is trendable and the PeakVue time waveform provides valuable diagnostic tool. The PeakVue technology can be applied to machinery over wide speed range (from fractional to a few thousand RPM). PeakVue is complimentary to standard vibration analysis in many cases and often can detect many faults missed in standard vibration analysis. Normal Vibration Data: Synchronous time averaged waveform data was collected off of the felt bearings. The purpose of this data was to search for excursions one (or a few) times per revolution of the felt as well as searching for structural resonance in spectral data being modulated with felt speed. A structural resonance was expected in the 50 to 150 Hz range. If the activity was present, one could verify structural resonance via data acquisition at different speeds. Also, side banding at felt speed with high resolution data acquisition was verified. PeakVue Data: Impact activity at felt speed was present in the spectral and time waveform. This data did not require synchronous time acquisition from felt roll. The felt was approximately 85 feet in length. The data was acquired on the bearing for the 5 foot diameter roll. DATA COLLECTION TECHNIQUES FOR THE RBM ANALYST The main purpose of this paper is to describe the basic facts needed to use PeakVue. PeakVue is a technique that can be used with CSI s 2120 Machinery Analyzer to identify early stage defects on gears and bearings. The concept behind the PeakVue technology and its analysis of stress waves will be discussed in this paper as well. The details of how PeakVue works may be difficult to understand, but the basic setup method for PeakVue can be easily understood. Basic setup method for using PeakVue Use an accelerometer with a frequency range that goes out to the 5 to 10 khz frequency range. Do not integrate the data. Select the Fmax for the spectrum based on typical rules for setting the Fmax. The highest
potential defect frequency usually defines the Fmax. For example, the Fmax should be at least 2 times the gear mesh frequency (GMF) or 4 times the Inner Race Frequency (BPFI). When selecting the number of lines of resolution, the lowest potential defect frequency helps set the number of lines of resolution to be used in the spectrum. At least six lines of resolution should be below the lowest defect frequency of interest. Some measurements (gearboxes with rolling element bearings) may require more than one measurement setup if enough resolution cannot be supplied in one measurement. Select a High Pass (HP) filter equal to or higher than your Fmax and of course the Fmax should be set above any typically expected rotationally generated frequency. For bearings on machines that are not gearboxes, select the HP Filter higher than 30x the shaft turning speed if the bearing is unknown. If the bearing is known then set the HP Filter greater than or equal to 4x the BPFI frequency. For gearboxes, select the HP filter greater than or equal to 2x the highest gear mesh frequency. Set up one PeakVue point per bearing for each machine you deem critical. Large slow speed machines are good candidates for PeakVue. (At a minimum collect one PeakVue point per component, but it is much better to get one point per bearing.) Also, try to make the measurement in a radial direction at the load zone of the bearing although axial measurements are acceptable. The HP Filter setting creates restrictions on sensor mounting. Generally, measuring data through a painted surface can present some measurement challenges. Use the following guidelines when collecting PeakVue data: If the HP Filter setting is less than or equal to 1 khz, a 2 pole magnet on clean surface or on single coat of very hard paint may be used. If the HP Filter is greater than 1kHz and less than or equal to 2 khz, a 2 pole magnet on clean surface may be used. (An unpainted surface is recommended.) If the HP Filter is greater than 2000 Hz up to and including 5000 Hz, use a flat rare earth magnet on a clean flat, smooth surface. (This may require the use of mounting pads) If HP Filter is greater than 5 khz, stud mounting should be used. Interpreting the Data There are three primary things to look for in the PeakVue data.
When analyzing anti-friction bearings using PeakVue spectrum, look for the presence of the fundamental bearing defect frequencies and harmonics. When analyzing gears, look for the shaft turning speed and harmonics of the shaft with the defective gear. The time waveform is excellent in looking at gear tooth impact spacing. Also, the presence of gear mesh frequencies with harmonics indicates lubrication, backlashing or eccentric gear problem. Looseness may also be detected because of the impacting occurring. Look for 1xTS, 2xTS, fractional harmonics of TS, etc. Remember, PeakVue amplitudes are trend able, whereas demodulated data is not. Fault frequencies work the same; the alarm sets will need to be adjusted for lower amplitudes and you may want to adjust the parameter bands to trend broader ranges. Remember, you re not looking for imbalance, misalignment, etc. in the same plot as Run Speed and its relative harmonics. Peakveu Question and Answers Question 1: What is the difference between a PeakVue and a Demodulated spectrum, when the set-up parameters (HP filter, Fmax, etc.) are the same? Ans: In demodulation, the "carrier" is removed via a full wave rectification followed by lowpass filtering (typically the low pass filter is the FFT anti-aliasing filter which is set so that the 3 db down point is at the Fmax frequency). The lowpass filter 3 db point is also at a frequency less than the high pass filter frequency. A low pass filter is an averaging circuit. To illustrate, assume you have a sine wave with amplitude of one unit and a frequency of 5000 Hz. Pass this signal through a good low pass filter set at 500 Hz. The output signal will be the average value of 0 units. Now rectify (half or full wave) and pass the signal again through the low pass filter set at 500 Hz. The output now will be a constant value of 0.6 units. If the 5000 Hz carrier had been amplitude modulated, the output would be varying about the average value with an amplitude variation less than was in the original signal. If the amplitude variation was a stream of short term spikes of constant amplitude, the output from the low pass filter might show the spiking events, but the amplitude would be dependent on the rate of the spiking events.(i.e. the amplitude would be dependent on the duty cycle of the "impact like " events). For PeakVue, there are NO low pass filters being employed. The output is the true amplitude of the spiking events and is not affected by the duty cycle. For example, suppose you have a cage defect initiating a 10g spiking event once per rev of the cage. PeakVue would output an impacting event of 10g's once per rev of the cage (typically one cage rev occurs every 2.5 turns of the shaft). Now assume the inner race has a defect (BPFI) generating the same 10g's per event which would be in the range of 6 to 11 times per rev of the shaft. Again PeakVue would show 10g impacting events at the BPFI rate. Demodulation would show different values (the difference would be extreme for slow speed machines) for these events because the average values are affected by the significant difference in duty cycles. Question 2: When using PeakVue together with the Pk-Pk waveform trend parameter, is the Pk-Pk value computed from the Acceleration waveform or the demodulated waveform or the default time waveform, if digital integration is selected in the analyzer?
Ans: PeakVue is to always be carried out in sensor units (acceleration units for an accelerometer). There is no such thing as a demodulation waveform in PeakVue. Digital integration has no meaning here. The trend parameter is calculated from the PeakVue waveform. Question 3: Please elaborate on the statement that data from demodulation cannot be trended: "The demodulation data is dependent on the frequency and the duration of the spikes, and it is therefore not trendable". For a bearing, the spike is nothing but the impact on the defects on the race and the frequency would be the bearing natural frequency (carrier frequency), which is constant practically. The number of impacts in the record length is also a constant once no. of lines and Fmax is set. The only variable is now the duration of the spikes, which varies with deterioration of the defects, hence the peakvue as well as the Demodulated value are bound to change together, hence demodulated data also should be trendable? Ans: Data from demodulation is dependent on duty cycle, duration of the event, setting of the low pass filter (analysis bandwidth), etc. Since factors dependant on the type of fault (duty cycle, speed of machine), and the analysis set up, effect the results, the trending does not make sense. When you trend, a changing trend should be directly correlatable to the fault and not on how you set the measurement up (within limits). Therefore, trending of demod results are not widely dependable. When an impact occurs, stress waves are emitted which are in no way related to the natural frequency of the bearing. The primary variable defining the dominant frequency in the packet of stress waves emitted is the contact time (see the Hertz theory on stress wave emissions) between the two metallic objects. Small impacting objects have shorter contact times than larger objects. Smaller objects have a higher pitch (frequency) than do larger objects. Since the stress wave packet excites a broad frequency band, it could excite any natural structural resonance (including the bearing natural frequency) in the proximity of the event. Stress wave packets will propagate away from the initiating sight at the speed of sound in the medium. The higher frequencies a) travel faster and b) attenuate more rapidly than the lower frequencies. Follow up Questions: In the PeakVue -vs- Demod power point "Circuitry" topic, it is mentioned/illustrated that the Demodulation technique employs a lowpass filter where as the peak Impact detection (PeakVue) does not. But CSI recommends Filter setting for Peakvue, hence the Filter Time constant mentioned as a disadvantage for the demodulation technique, also applies to the Peakvue. Hence please clarify why demod cannot be used for trending if low pass filter setting is required for both demodulation and PeakVue analysis. The PeakVue filter setup recommendation is for the high pass filter. NO low pass filter is used. The low pass filter for demodulation is the anti-aliasing filter (not used in PeakVue) set when Fmax is specified. -Is the anti alias filtering (for FFT computation) used for demodulation and peakvue processing? When the FFT is computed using demodulation, anti-aliasing filter IS used. When the FFT is computed using PeakVue, anti-aliasing filter IS NOT used. -In demodulation, when we say that the bearing fault signals are modulated on a high frequency carrier signal - do we refer to this carrier signal as the bearing natural frequency or the Hertzian (stress) waves? Bearing fault frequencies are NOT modulated on a carrier frequency. Stress waves are emitted when an impact occurs. These waves travel (propagate) away from the impacting sight at the speed of sound in the medium. Stress waves are both longitudinal (material
experience changing length about the mean) p waves, and bending, s waves. Bending waves introduce a ripple effect (similar to waves which are observed when a pebble is dropped into a still body of water) which travels away from the impacting surface. An accelerometer attached to the surface will respond when the s wave propagates under the spot where accel is attached. The duration of the wave propagating under the accel can be short (fractional to several msec). -Here we use accelerometers with magnetic mounting for the vibration measurements. Does the mounting natural frequency of the accelerometer affect the amplitude of these impact pulses. Any precautions to be taken in this regard? Yes, the mounting natural frequency of the accelerometer will affect the amplitude of the pulses, same as for any vibration signal. You can still obtain accurate "trend" measurements, if you are careful to mount the accelerometer the same each time. For precise determination of true amplitude, a full sensor response curve for the accelerometer and mounting method would be required. Some precautions to take are as follows: For bearing monitoring, you would typically select an Fmax for the spectrum that is high enough to capture at least the 3rd harmonic of the inner race fault frequency. For gears you would want to see at least the 2nd harmonic of gearmesh (with perhaps a trend parameter that monitors the 3rd harmonic as well). Then the next higher PeakVue Hi Pass filter would be selected, following the guidelines below when mounting the sensor: up to 1 khz filter: OK for use with a small 2-pole magnet. 2 khz filter: Reaching the borderline limit of a 2-pole magnet, requires paint to be removed from measurement surface. 5 khz filter: Requires flat rare-earth type magnet, and clean/flat surface (preferably a smaller sensor/magnet with higher frequency range). 10 khz filter / 20 khz filter: Should be stud mounted, using a vibration sensor with appropriate frequency range. PEAKVEU Recommendations Stress wave capture and analysis, where emphasis has been placed on capturing the peak values associated with each event, has proven to be a very useful methodology for fault detection, fault progression, and severity assessment. Therefore, the following recommendations are presented. Specify measurement points for peak value (PeakVue) analysis at each bearing on each machine. When establishing such stress wave points, use the PeakVue measurement setup criteria given is Table III within section 4.2.5. Extract and trend the "Max. Peak" value from the PeakVue time waveform. This true peak value should be extracted from the peak value time block covering a minimum of six revolutions (preferably 10 revolutions) of the machine.
For rolling element bearings, it is recommended that a minimum of 16 revs be included within the peak value time waveform (providing adequate resolution for cage fault resolution). The bandwidth of the spectrum computed from the peak value time block should include the maximum of: a. A minimum of three harmonics of the BPFI fault frequency for rolling element bearings b. Two times gear mesh for gearing systems. If this requires two measurement sets (one for gear faults and one for bearing faults), then proceed with two measurements. The high-pass filter setting should always be equal to or greater than the maximum bandwidth measurement used for the peak value analysis on the machine for all peak value measurement sets on a specific machine. * The parameters, recommended for trending are: a. Pk-Pk value (in sensor units) of the peak value time waveform extracted from a time block inclusive of 6 or more revolutions of the shaft measurement point assigned to. b. Digital overall computed from the entire broadband spectrum. c. Energy of 4-10 synchronous shaft revolutions. d. Energy of non-synchronous energy covering BSF, BSFO, and BSFI. e. Energy of subsynchronous energy covering cage frequency. f. Energy surrounding one and two times gear mesh.