A motor shaft system is typically a double-supported single-shaft system. When considering the motor shaft system alone, it is one of the simplest classes of shaft systems and forms the basis for all shaft system vibration analyses. The actual analytical methods used are also foundational for vibration analyses of more complex shaft systems.
Vibration analysis of a motor shaft system begins with the selection of appropriate sensors. There are various types of sensors, and at the outset of conducting a vibration analysis on a motor shaft system, it’s vital to know whether the vibration signal to be analyzed should be a displacement signal, velocity signal, or acceleration signal.
Previous articles have discussed the principles of how to choose. However, depending on the purpose of fault diagnosis and analysis, additional signals may sometimes be needed. For example, for medium-speed motors, we primarily need to analyze the velocity signals of the vibration.
However, from the calculation of bearing characteristic frequencies (see relevant materials), it can be determined that the frequency band of the bearings may be high. In the early stages of bearing failure, the amplitude of its characteristic vibration is a small proportion of the total vibration, thus its features are not distinctly visible from the total value.
If acceleration signals are included at this point, the characteristics of the bearings become much more pronounced, aiding early detection. The purpose of this example is to illustrate that while there are certain principles in selecting vibration analysis signals, adjustments can also be made based on the actual purpose of the analysis.
Similarly, once the purpose of the analysis is established, the sampling frequency of the sensors needs to be determined.
According to the sampling theorem, the sensor sampling frequency must be more than twice the sampled frequency for the sampled frequency to be extracted. Specifically, for a motor, we can calculate characteristic frequencies based on the primary objective of the analysis, and then use sensors with a sampling frequency more than twice this characteristic frequency.
Of course, a higher sampling frequency can be beneficial for the sampling process, but it may also introduce noise signals and other interferences.
The task involves sensor placement at appropriate locations in the motor shaft system. If conditions allow, we measure two points perpendicular to each other on the radial plane, as well as an axial position.
For the motor, the same sampling positions are required for both end bearings. If conditions do not permit, one radial and one axial can be retained. If that’s still not feasible, only one radial point is measured. The associated relationships can be found in the previous article.
In fact, once the sensors are selected and installed, measurement and signal collection can begin. Naturally, the measured signals need to be transmitted through the appropriate software and hardware via data acquisition devices.
Subsequently, data analysis methods are employed for time-domain plotting, frequency-domain unfolding, waterfall chart drawing, and so on. This encompasses the work of data collection and analysis.
If an analyst wishes to write their own analysis program, they must master the processing techniques and associated knowledge. The spread of languages like Python, Matlab, and R has made these types of analyses less challenging in terms of tools. However, mastering signal processing techniques is still a necessity for industrial engineers.
Most people, of course, directly interpret and judge based on the results of data analysis.
For electric motors, primary vibration time-domain changes involve analyzing the peak-to-peak displacement of displacement signals, the effective values of velocity signals, and the peak values of acceleration signals. The variations of these signals over time, such as whether they reach warning limits, constitute the most basic time-domain analysis.
Data analysts can delve deeper into the time-domain characteristics of these signals, diagnosing based on various time-domain attributes (around thirteen in total).
In practical applications, the more commonly used frequency-domain analysis method involves unfolding the collected data in the frequency domain to observe the characteristics of faults.
For electric motor shaft systems, there are two main components: the frequency parts related to the shaft system, and those related to the bearings. (This discussion will not cover connected gears, fan blades, etc.)
It’s not difficult to see that the frequencies related to the shaft system are around 1, 2, 3, and 4 times the base frequency. Those related to the bearings are near the characteristic frequencies of the bearings. This determines the target of the frequency band analysis.
As for feature comparison, a vast array of them are introduced in specialized literature, such as misalignment, unbalance, loose foundation bolts, inner race, outer race, rolling element, cage characteristic frequencies, and so on. There are also some miscellaneous items, like rotational rubbing.
These features are easy to find and remember. However, signals collected on-site are often mixed together, requiring certain experience to separate and identify.
After analysis and identification, an evaluation can be made of the equipment condition, and faults can be discerned to a certain extent. At this point, the main steps of vibration analysis are essentially complete.