Menu
Menu
Menu
Vibration analysis is a widely used method for diagnosing bearing faults in machinery. It operates on the principle that a healthy machine exhibits a certain characteristic vibration pattern, and any deviation from this pattern can indicate potential issues, such as bearing faults.
Bearings in optimal condition typically produce a low level of vibration. When there’s a fault or wear and tear, the bearing’s vibration patterns change, and the overall vibration levels increase. This occurs because the smooth and predictable motion of a healthy bearing becomes erratic when the surfaces are damaged or worn, causing an increase in kinetic energy, which is expressed as vibrations.
This type of time-domain vibration level measurement generally gives limited information other than to the experienced operator, but it can be utilized for trending, where an increasing vibration level is an indicator of a deteriorating machine condition. Trend analysis involves plotting the vibration level as a function of time and using this to predict when the machine must be taken out of service for repair. Another way of using the measurement is to compare the levels with published, standardized vibration criteria for different types of equipment.
Broadband vibration measurements may provide a good starting point for fault detection, however, they have limited diagnostic capability. This type of analysis can’t reliably indicate where the fault is i.e. bearing deterioration/damage, unbalance, misalignment, etc.
In order to pinpoint the particular component where the issue is, how to measure frequency analysis is normally employed. This advanced analysis method usually gives a much earlier indication of the development of a fault and secondly the source of the fault. Having this information early is vital, as it allows one to evaluate when a critical failure is imminent and plan accordingly for maintenance/downtime.
This article will take a closer look at key vibrational frequencies where increased vibration levels are associated with specific fault in machine bearing components. The formulas for calculating these key frequencies will be presented together with some example data to validate the model, so it can be understood and properly utilized in your future vibration analysis campaigns.
This type of vibration analysis this will allow to predict not only that failure is imminent, but also identify in which component of the bearing it is: inner or outer raceway, rolling elements or the cage.
Rolling element bearings (REB) fault recognition is based on the detection of some characteristic frequencies which are property of the bearing geometry, rotational speed and number of rolling elements. A bearing under normal working condition should not produce its Characteristic Defect Frequencies (CDF) at the vibration spectrum. However, any potential fault at bearing races, rolling elements or cage usually generate the CDF that can be calculated based on a set of formulas.
It should be noted that one property of bearing’s CDF is that, they are not integer multiples of shaft rotational speed. This characteristic allows us to suspect a potential bearing problem even if its type is unknown.
To calculate the fault frequencies for rolling element bearings based on their geometry, you need to know the bearing’s dimensions and operating conditions. The four primary fault frequencies associated with rolling element bearings are:
By calculating these fault frequencies, you can analyze the vibration spectrum and look for peaks at these frequencies, which may indicate bearing faults. It’s essential to consider that these calculations are for ideal conditions, and real-world factors like manufacturing tolerances, wear, and lubrication can affect the actual fault frequencies. Thus, it is good to set a baseline where you produce a measurement set in close to optimal conditions and observe historical data over a long period of time. Comparing current vibrational levels and frequencies where peaks are measured to the baseline will not only indicate the presence of potential faults, but also allude to which component it might be in, in case of bearings it could even point to exactly what is wrong with the particular bearing.
It is vital to have this information and take action on time, as according to research 41% of critical machine failures are due to bearing faults.
A bearing fault usually develops in different stages and consequently that affects the diagnosis procedure. Bearing defects can be categorized in four stages of wear depending on the size of the defect and the patterns they produce at the frequency spectrum. As one would expect, with incrementing the number of the stage the vibration levels increase bringing the system closer to critical failure, making it vital to detect and address the issues in as early a stage as possible.
It is recommended that maintenance personnel should not proceed directly to a high-speed machine overhaul until signs of wear could be identified in lower frequency region as the defect size at this stage will remain at microscopic level. At this stage bearing lubrication and condition monitoring is advisable.
As the defect develops the bearing enters the second stage of fault. In that case high energy impacts will excite bearing natural frequency and high frequency energy level starts to increase. The natural frequencies depend on the dimensions of the bearings and their mounting, however they are generally in the 5kHz+ region. Envelope analysis can be utilized at this stage to effectively identify the peaks in the spectrum, however it is complex and requires a good deal of computational power.
As these vibrations increase in severity, sidebands appear above and below in the spectrum which is an indication that the device is approaching stage III. At this point the machine should be closely monitored as it will need maintenance soon.
At the third stage of a bearing fault the most commonly recognizable patterns can be measured at low frequency spectrum. Despite the fact that high frequency energy continues to increase and envelope analysis is still efficient, the classic patterns of bearing failure are now present at velocity spectrum.
On the one hand, for outer race fault harmonics of BPFO should be measured at low frequency spectrum. On the other hand, in the case of inner race fault harmonics of BPFI with shaft rotational speed sidebands would appear. The sidebands around the BPFI are present because as the inner race defect passes through the load zone it creates amplitude modulated vibration signal.
All the aforementioned patterns are valid for inner race rotation, while when the outer race rotates the patterns of BPFO and BPFI fault are reversed. Ball or roller defects generate BSF or multiples of it with FTF sidebands at velocity spectrum. Usually, the presence of three harmonics is enough in order to decide a bearing replacement as soon as possible.
This is a stage where there is a good balance between fault severity and cost efficiency of detection. As the vibrations are in the sub 5kHz range. Vibration sensors that come with built in frequency analysis features can be utilized (built-in FFT) to generate the required spectrum for analysis.
NCD wireless vibration sensors, that can produce frequency-domain data in raw or processed form. All models come with up to 10years of battery life, long range and industrial grade design. These are great as they not only adhere to the vibration metrics requirements to successfully carry out the analysis, but come with a number of extra features to simplify deployment and reduce cost of maintenance.
Stage four is the final stage of bearing deterioration before total failure. If the bearing reaches stage four, the high-frequency detection method levels may actually decline, the CDG will start to disappear. The edges of the raceway or rolling element defects begin to round off, which actually reduces the intensity of the impact forces. Metal which has been removed from the various bearing components may actually fill in some of the more severe flaws and be smoothed over by rolling elements. However, during this process, the clearances within the bearing are beginning to increase substantially, creating a significant looseness condition, resulting in vibrational components that are integer multiples of running speed and a significant increase in overall vibration levels.
Not less importantly, the discrete bearing defect frequencies disappear and are replaced by random vibration in the form of a noise floor. At this point the machine should not be operated as catastrophic failure is sure to occur in the very near future.
Having discussed the 4 stages of failure in detail one could conclude that stage III is best suited for predictive maintenance. It is important to correctly analyze it as at this point there is significant presence of BPFO, BPFI, BSF, and FTF components that are relatively easy to detect, but at the same time the machine has not yet reached critical failure levels. Moreover, they are accompanied by sideband harmonics, shifted by the rolling frequency (1xRPM), that can be used as an indication on how close to stage IV the machine is (the more prominent they are the worse the damage).
A bearing has several components that each experience faults resulting in different spectra. Depending on the vibration detected one can infer whether there is roller deterioration, the cage has been damaged or there is damage to the inner or outer raceway. Also, in most cases the severity of the damage corresponds to the increase in vibration (the more the component deteriorates, the harsher the vibration gets).
Let’s look at an example for each of the 3 components and how can one detect faults based on the vibration spectra.
As expected, there is a vibration component at the rotational frequency (6.56Hz), 2x rotational frequency (13.12Hz) and higher order harmonics in the low frequency region.
There is vibration at 62.4Hz which is 2x the roller rotational frequency, more interestingly there are harmonics at 186.5(x3), 497(x8), 560(x9), 748(x12), 873(x14) and 936Hz (x15). This indicates deterioration in the condition of the rollers.
The following image represents a vibration measurement on a cylindrical grinding machine. Although it was producing bearing outer rings with good quality, there is indication that an impeding fault might be imminent, due to deterioration of the spindle condition.
The spindle is rotating at 19200rpm (320Hz) and the cage rotating frequency is 140Hz. As expected, there is a well-defined peak at around 1x, 2x and 3x that value. More notable is the peak at 1740Hz, which is at 5x the spindle frequency plus the cage frequency. There are more of these components at higher frequencies, which indicates cage damage.
We examine an electrical motor that rotates at 3000rpm (Fr=50Hz). For the bearing used in this scenario the calculated BPFO is 229Hz. If we take a look at the spectra in the image below large peaks are present in the region from 1 to 1.5kHz. These are shifted replicas of the 5x BPFO by multiples of Fr the rotational frequency:
1142 = 5x BPFO
1092 = 5x BPFO – Fr
1193 = 5x BPFO + Fr
1243 = 5x BPFO + 2x Fr
1290 = 5x BPFO + 3x Fr
1340 = 5x BPFO + 4x Fr
….
The overall vibration of the motor also increased from 0.22g to 1.64g, which was already indication of an issue, however not enough to pinpoint the problem. Upon inspection it was found that indeed, the bearing’s both inner and outer raceway has been damaged due to the ball path being offset.
Rolling bearings exhibit characteristic vibration signatures that are generated usually in the form of modulation of the fundamental bearing frequencies. This can be utilized via vibration condition monitoring software, which in is designed to identify these characteristic features and provide early detection of an impending problem.
The spectra are monitored for these key frequencies and in the case of vibrations with sufficient amplitude the system alerts for potential issues, allowing for timely maintenance and critical failure prevention. Combined with general broadband vibration measurement to monitor overall system performance, this type of spectrum analysis can be a powerful tool for keeping machine bearing components working in optimal conditions.
To learn more about NCD and the sensors they offer like the Wireless Vibration sensor shown in the image below visit the NCD store. If you are interested in understanding vibration analysis better you can check out the article Understanding Wireless Vibration Analysis in the Learn section of the NCD website.
FAG, An Overview of Bearing Vibration Analysis, Dr. S. J. Lacey, Engineering Manager Schaeffler UK Limited
Vibration Analysis of Rolling Element Bearings (Air Conditioning Motor Case Study), Konstantinos Kamaras, Anastasios Garantziotis and Ilias Dimitrakopoulos