One of the major challenges while dealing with heavy machineries is to keep them running for as much time as possible or say to reduce the down time to increase productivity. If a machine breaks down during scheduled operating hours, it can bear huge financial losses. But few machines are operated with a goal to let it run for as long as it can. Once it fails or malfunctions, it is repaired/replaced. This is called Corrective Maintenance.

But for most machines, operating it till it fails is not an option. To avoid failures, maintenance is scheduled at regular intervals. This is called Preventive Maintenance. It is costly and requires a lot of resources. But with the advent of sensor, IIoT and cloud technologies, now the actual condition of machines can be monitored continuously and maintenances can be done when truly needed. This is called Predictive Maintenance (PdM). It is much more efficient and hence most cost-effective as well. The process by which PdM is done is called Condition Monitoring. The process of monitoring a parameter of condition in machinery (Vibration, Temperature, Load & Lubrication etc.) is called Condition Monitoring.

Vibration Analysis is one component of the CM Program. In Vibration Analysis (VA), the vibration of the system is monitored and used to detect anomalies and avoid failures. Vibration Analysis is commonly used for Rotating Machines. Faults such as Unbalance, Misalignment, Looseness, Rolling element Bearing Faults, and Resonance Condition can be detected by vibration analysis. Typically accelerometers are used as sensor to collect vibration data. This accelerometer can measure Displacement, Velocity or Acceleration, any of these parameters can be used for vibration analysis. Tri-axial accelerometer arrays when feasible or at least three individual axes of data are needed for accurate Vibration Analysis.

Generally, three different types of analysis can/should be done with vibration data namely Time Domain Analysis, Frequency Domain Analysis (Spectral Analysis), and Phase Analysis. Apart from these, Bode Plot, Waterfall Plot, Polar Plot, Orbit time base plot etc. can be constructed from vibration data to get useful insights.

Time domain analysis is most useful to identify crack, degradation around periphery etc. These faults have significant increase in amplitude of vibration at specific intervals which can be detected from time domain data.

The actual Vibration Signal consists of individual frequencies corresponding to different components. Converting the time domain signal to frequency domain helps us identify the individual frequencies. This is called Spectral Analysis. Fast Fourier Transform (FFT) is used to convert the tie domain signal to frequency domain signal. Spectral analysis is most useful to identify degradation of bearings. Spectrogram and Power Spectral Density (PSD) are other example of Frequency Domain Analysis.

Now let’s come to a few examples of deciphering vibration data. For example, high vibration at the frequency corresponding to the speed of rotation is most often due to residual imbalance and is corrected by balancing the machine. Another example, a degrading rolling-element bearing exhibit vibration signals at specific frequencies increasing in intensity.

Following are the brief details of different faults which can be identified with Vibration Analysis:

Normal Condition: Even when a machine is in normal condition, there is some vibration at the running speed (1X).

Unbalance: When there is unbalance in the system, there is a significant increase in Amplitude of vibration (1X) at running speed.

Misalignment: For misalignment in the system, extra frequency component (2X, 3X) are present in the vibration signal along with an increase in amplitude at running speed. Misalignment may exist even if not apparent from Spectrum Analysis. Phase analysis needs to be done to confirm Misalignment.

Rotating Looseness: For rotating looseness, Harmonics having high Amplitude are present in the vibration signal. Also, phase is erratic in this case. But it’s important to keep in mind that although rotating looseness generates harmonics, all vibration data having harmonics doesn’t necessarily indicate rotating looseness.

Bearing Defects: This type of fault can be identified by looking for sidebands at bearing frequency.

Resonance: Resonance condition in machines can also be identified using vibration analysis. In this case, the Amplitude at fundamental frequency (1X) increases along with widening of base.

Overall, vibration analysis is one of the most powerful condition monitoring techniques to build a predictive maintenance tool especially for rotating machines. It can be combined with other CM techniques like lubricant analysis, temperature analysis etc. to build a comprehensive and robust PdM tool.

Author: Himansu Sahu, Business Analyst – Digital Transformation Services