Red Wolf Reliability
Vibration analysis is an integral part of many plant's condition monitoring program for rotating equipment, and every analyst understands the importance of using frequency-based spectra to detect machine faults. Many analysts also understand the importance of analyzing time-based waveforms as well. However, few may realize the usefulness of High Frequency Demodulation (HFD) spectra and how to use them to detect machine faults, typically early bearing defects.
In order to understand the concept of high frequency demodulation, one must first understand the basics of vibration data collection and signal processing. Assuming you are using an accelerometer to collect data, the data collector captures the absolute acceleration of the equipment over a certain time period, which is presented as a waveform in the time domain. This time waveform is translated from the time domain into the frequency domain through the use of a Fast Fourier Transform (FFT), which gives us an easy to read graph of the amplitude at each frequency of vibration. It essentially breaks down the time waveform into a series of sine waves, each representing a specific frequency.
When you have a time waveform with a repeating impactive event, the FFT must compensate for the fact that the waveform is non-sinusoidal. In other words, the waveform cannot be translated into a series of perfect sine waves. It can be seen that the amplitude is changing (modulating) over time. A vibration with period T2 (or a frequency of 1/T2) can be seen, which is known as the carrier frequency. The amplitude modulation shows as "sidebands" of the carrier frequency. These sidebands are visible in the FFT as a 1/T1 frequency shift in either direction along the frequency axis.
With high frequency demodulation, we want to strip away all other content and essentially look for the rate of modulation of these high frequency events. Physically, an example of this can be seen as a bearing natural frequency being "rung" every time a bearing element rolls over an outer race defect, causing an impact at BPFO with a ring down at the bearing natural frequency. Now when we have more dominant frequencies, such as a large unbalance or misalignment in the system, these small impactive events can be "drowned out" by the dominant frequencies and may not be visible in the spectra. HFD will look for those high frequency-low amplitude impactive events. The first step is to strip away the dominate low frequency vibration by sending the signal through a high-pass filter. The signal is then sent through a low pass filter to strip away the carrier frequency, leaving us with the modulation frequency. Finally, the resulting waveform can be converted to the frequency domain using an FFT. In this example, we can see that BPFO is now clearly visible.
The HFD spectrum can be useful as an early warning of bearing defects. It may not be necessary to take corrective action at the point that this warning sign is found, however, it will allow for trending of the issue and will put you on track to look for more severe signs of bearing defects. Keep in mind that the vibration that we are looking for may not propagate well throughout the equipment, so it is important to take your readings as close to the site of initiation as possible (i.e. the bearing housing). In addition, adequate mounting of the sensor and ensuring conditions are consistent are important in accurate trending of the data. Dirt, grease, or even a new layer of paint may attenuate high frequency vibrations and reduce amplitudes in the HFD spectra. Also, keep in mind that as the bearing defect progresses, it will show more in your other spectra and less in your HFD spectra. If you find a bearing defect in the HFD spectra and not in your other spectra, this is an issue you should monitor. If you have a raised noise floor in the HFD spectra and bearing defects in your other spectra, the bearing should be replaced.
For more information on HFD spectra or other vibration analysis related questions, please email Red Wolf Reliability or call 970-266-9005.