Pumps and Systems, March 2009
Enveloping is a tool that can give more information about the life and health of important plant assets. It is primarily used for early detection of faults in rolling element bearings and gearboxes. Enveloped acceleration is an especially valuable parameter to trend, as the progression of machine condition can be evaluated. Armed with good information and assisted by service experts, plant engineers can be confident in the proper operation and management of the important assets in their care.
Enveloping can reveal faults in the earliest stages of development, before they are detectable by other machinery vibration measurements. Without an early fault detection technique like enveloping, personnel must wait until the latter stages of failure, when overall vibration increases, lubricants become contaminated and temperatures rise. By this time, the remaining usable life of the failing machine elements could be short and the damage more extensive than if the fault had been detected earlier.
The enveloping technique enables the detection and analysis of low-level, repetitive vibrations by extracting them from the overall machinery vibration signal. Enveloping thus facilitates earlier prediction of failure in machines with metal-to-metal contact. While the examples in this article are based on rolling element bearings, the techniques also can be applied to gearboxes and electric motors with commutators.
It is important to note that the successful application and interpretation of enveloping data require experience. Enveloping is just one tool in the analyst's toolbox, and it is best used as one of a number of techniques for complete monitoring of a machine.
Enveloping Isolates Signals of Interest
Enveloping is a multiple-step process that extracts signals of interest from an overall vibration signal (Figure 1). In a rolling element bearing, the interaction between bearing elements and defects excites a structural resonance in the bearing support structure. A seismic transducer measures the vibration and the signal is band-pass filtered to keep only signal components around the resonance frequency. The filtered signal is rectified and then enveloped, which removes the structural resonance frequency and preserves the defect impact frequency. A low-pass filter then eliminates some of the extraneous high-frequency components, and a spectrum is generated. Frequency components are correlated with physical bearing parameters, and a trend of the spectra can show progression of defects.
Figure 1. Typical steps in the implementation of enveloping
Analysis of the enveloping process begins with the source of the vibration signal. As the elements of the bearing interact with each other and with defects, forces are coupled to the machine casing and produce vibration. These defect interactions behave as impacts that excite a natural resonance frequency of the machine, causing it to ring (Figure 2). The amplitude of the ringing decays until the next impact, which re-excites the resonance. The defect amplitude modulates the natural resonance response at the impact frequency. The defect-related signal becomes part of the overall vibration of the machine.
Figure 2. Defect impact causes ringing of the machine structure at its natural frequencies
Because of its higher frequency response characteristics, an accelerometer is generally used to measure the vibration signal for enveloping, which is often called acceleration enveloping, or high frequency acceleration enveloping. High-frequency vibration signals, such as the resonance carrier of the defect signal, do not travel far in a homogeneous machine structure; metal imperfections, joints and gaskets cause further significant attenuation (Figure 3). It is critical that this low-level, high frequency signal be coupled efficiently into the accelerometer. The accelerometer should be mounted as close to the bearing as possible and near the load region of the bearing, where signals are coupled more effectively to the machine case.
Figure 3. Vibration signal transmission loss in homogeneous and nonhomogeneous materials
Figure 4. Accelerometer vibration waveform output with embedded defect signal
The output of the accelerometer (Figure 4) contains three important frequencies: a relatively low-frequency, high-amplitude rotor-related vibration; the modulated structural resonance frequency; and other high-frequency vibration components, including harmonics of the structural resonance frequency. Though the signal is complex, application of the enveloping technique allows the determination of an impact frequency associated with the defect, which provides valuable information about machine condition.
Filtering: An Important Part of the Enveloping Process
Band-pass filtering is the first signal-processing step in the application of the enveloping technique. Proper filter specifications are critical to the removal of unwanted components of the signal while preventing detrimental attenuation of signal components essential for enveloping analysis. Frequency range selection must take into account machine operating speeds and structural natural frequencies, which depend, at least in part, on bearing design along with machine construction and mounting. To obtain the most useful information from the enveloping technique, some experimentation with available frequency ranges for the filters is often required when enveloping is first applied.
A good starting point is to examine the spectrum plot for a high frequency structural resonance "haystack." The lower, high-pass corner should be set above gear mesh frequencies, but below this structural resonance "haystack." The lower corner frequency is selected to reject the comparatively high-amplitude, low frequency components associated with the normal machine running speed vibration. This greatly improves the signal-to-noise ratio for the frequencies of interest, since it is these lower frequencies that generally dominate the vibration signal. The upper corner frequency is selected to remove extremely high frequencies, which are associated with other machine vibration frequencies and signals amplified by the accelerometer or mounting resonance.
For a machine with rolling element bearings, the lower corner frequency is generally set to be greater than 10 times the running speed of the machine to eliminate the most common harmonics of running speed. However, this frequency should not exceed one-half of any structural natural frequency associated with the bearing. This natural frequency serves as the carrier frequency excited by the defect impacts, and attenuating this signal of interest is detrimental to the success of enveloping.
The upper corner is generally set to around 60 times the outer-race ball-pass frequency, or approximately 200 times running speed. This attenuates high frequency noise and vibration components, some of which have been amplified by accelerometer resonances. These basic rules are relatively simple to apply to a rolling element bearing. However, they become more complicated on a gearbox because of gear-mesh frequencies.
The band-pass filter output (Figure 5) shows the structural resonance frequency, which is the higher frequency in the waveform, modulated by the defect. The impacts associated with the defect excite this carrier frequency, and its amplitude then decays exponentially. The signal from a defective bearing may have different impact intervals, more frequency components and differing amplitudes-all potentially influenced by lubrication, the number of defects, the severity of the defects and the loading of the bearing (among other things). However, enveloping is still effective and valuable for these more complicated signals.
Figure 5. Band-pass filtered vibration waveform showing defect modulation of the machine structural resonance
Amplitude Demodulation Eliminates the Resonance Frequency
To envelope (demodulate) the filtered signal, it is full-wave rectified first (Figure 6), which doubles the carrier frequency and further separates the impact frequency and the carrier frequency.
Figure 6. Full-wave rectified vibration signal
The next step is the actual enveloping itself. Amplitude demodulation of the rectified waveform eliminates the carrier frequency and leaves the repetition rate of the defect impact. A number of methods are available to accomplish demodulation, including peak detection (Figure 7), integration and low-pass filtering.
Generally, enveloping results in a waveform, which has spectral components corresponding to the defect impact frequencies and harmonics of the defect frequencies from the impact of the event. Frequency components unrelated to the impact will generally be higher in frequency than the components of interest. Some of these can be eliminated by another application of low-pass filtering, leaving the defect impact frequencies and some low-order harmonics. Interpretation of this less-cluttered spectrum is easier, since fewer components need to be considered.
Figure 7. Envelope of vibration signal produced by a peak detector
The next step before analysis is to generate a spectrum of the enveloped signal. The defect impact frequency should show up clearly in relation to all the other spectral components in the signal. Harmonics of fundamental fault frequencies are generally artifacts of the enveloping process and are not valuable for trending purposes, but the presence of more harmonics may indicate the progression of a fault. After disregarding harmonics, significant frequencies present in the spectrum may be correlated to physical machine parameters. Note that sidebands related to running speed may appear around defect frequencies in the spectrum, especially as defects become more severe.
If uncorrelated frequencies are present in the spectrum, the filters may have been configured incorrectly or the transducer measurements were taken improperly. Frequencies may also come from other nearby machinery components, or they may be caused by the machine operation or process. While improper measurement techniques may prevent frequencies from appearing in the spectrum, if defects are absent in the monitored machinery, no defect-associated frequency components should be seen in the spectrum.
Interpretation of Information
Interpretation of the information is the important final step. Trending the magnitude of these frequency components can indicate the progression of bearing faults, but the magnitude is not necessarily related to the severity of the defect. For example, developing faults, such as a spall, may initially cause defect frequencies that have increasingly larger amplitudes. As the spall grows, it may present less of an impact event as the edges of the fault self-peen, or smooth them out; the amplitudes now decrease. A measurement trend is valuable to show the phases through which the signal has gone and to allow the user to infer the fault's progression.
For trending of defect frequencies to be useful, baseline information is required. It must be taken when the bearing is known to be in good condition and often enough to provide adequate resolution of the progression of the fault. In addition, initial experimentation with filter settings will help validate data. Knowledge of the machine and its bearings is essential to identify the frequencies to monitor. Because successful trending and correlation of data requires repeatable measurements, permanently mounted transducers are recommended.
Best Practices
When properly applied, enveloping can be a valuable tool for early detection of faults in machines with rolling element bearings and gearboxes. It is especially useful when applied in a periodic monitoring schedule and can provide condition information indicating faults in their earliest stages of development.
To use enveloping techniques successfully, care must be taken in a variety of areas. Experience is extremely valuable. Machinery knowledge is critical. Proper configuration of the whole measurement system is essential. Equipment selection and application must be done with all these considerations in mind, and while a good product will make the use of enveloping easier, it is never a mindless exercise. Enveloping must be used with its capabilities and limitations in mind so that, if this is done, it can be a useful tool in the hands of a capable machinery analyst.