Detecting rolling element bearing faults with vibration analysis 

Detecting rolling element bearing faults is the highest priority for most vibration analysts.  Detecting the fault at the earliest opportunity should be the priority, however in reality most analysts do not detect the fault in the first or even the second stage of failure.  This article is going to help you to detect faults at stage one so that you can truly be in control of your maintenance program.

In this article I will describe the four stages of bearing failure and how to understand and successfully utilize the airborne ultrasound, Shock Pulse, Spike Energy, PeakVue, enveloping/demodulation, time waveform analysis and spectral analysis methods.  I will also explain why you should not rely on trending overall level readings.

Reducing bearing faults

No article of this nature can be complete without a discussion of the reasons why bearings fail in the first place.  Your first priority should be to minimize the causes of bearing failure.  If you can do that successfully, then you will not need to rely on the vibration analysis techniques as much.  That is not to say that I want to put vibration analyst’s out of work, or that you should even consider downsizing your vibration monitoring program (because there will always be bearing failures and other mechanical faults) – the point is that the path to equipment reliability does not begin with vibration analysis.

The fact is that if you properly purchase, transport, store, install, and lubricate your bearings, and you operate machines that are balanced, aligned and operating well away from natural frequencies, your bearings will last longer.

You may not have control over many of these factors, but if you are involved in vibration analysis then there are two things you can definitely do: look for the presence of conditions that will cause bearings to have a reduced life, and perform root cause analysis when you detect bearing damage. 

I opened this article by pointing out that the detection of rolling element bearing faults is the highest priority for most vibration analysts.  The sad truth is that for too many analysts it is the only priority.  Unbalance, misalignment, soft foot, and resonance often have a much lower priority.  Although these faults conditions appear first on most wall charts, they can be the trickiest to diagnose.  Phase analysis is a powerful, yet underutilized tool that can greatly help in the detection of these fault conditions – but that was the topic of an earlier article.

The point is that these conditions put additional stress on the bearings, thus reducing their life.  If you do not take care of these conditions, it is inevitable that you will soon see the earliest stages of bearing damage.

The pattern of bearing damage

Before we get into the specifics of the four stages of bearing failure, I would like to describe how the vibration changes in general terms.  In classical teaching, bearing vibration is all about the four forcing frequencies: ball pass outer race (BPFO), ball pas inner race (BPFI), ball spin (BSF), and cage or fundamental train frequency (FTF).  We will discuss these in more depth in a moment, but first I want to describe the movement of “broadband energy”.

If a bearing is poorly lubricated, we can detect an increase in the level of “noise” at very high frequencies.  It is not a specific, single frequency; instead it will depend on a number of factors to do with the machine’s construction.  Suffice to say that you cannot hear it; it is well above your hearing range.

As the state of lubrication worsens, the level of the noise will increase, but the frequency of the noise will slowly reduce – it will move from very high frequencies to high frequencies.  That is not to say that you can’t detect the condition at lower frequencies; it is stronger at the higher frequencies.

As the film of lubricant between the bearing surfaces is reduced further, we will have more and more metal-to-metal contact, causing “stress waves” to be generated.  Stress waves (also referred to as “shock pulses”) are like ripples in a pond; the moment the metal surfaces make contact, a wave of energy races away from the point of contact at the speed of sound.  It all happens very quickly; possibly in less than a thousandth of a second!

Even if the root cause of the bearing fault is not poor lubrication, if the bearings are damaged through poor installation, false brinelling (where the bearing has been vibrating whilst it is stationary), EDM, misalignment, or any one of a number of reasons, there will come a time when there is either metal to metal contact between two surfaces, or the stress waves will be generated from beneath the surface of the metal as subsurface defects develop.

The subsurface defects will slowly develop due to the extreme forces experienced within the bearing.  The difference is that these defects are likely to be localized; at the bottom of the outer race for example.  The “noise” from the bearing due to poor lubrication is relatively constant (it is random, therefore not periodic), whereas when a fault condition develops (e.g. a crack or spall), a new source of periodic vibration will be introduced.  If the damage was on the outer race, each time a rolling element passes that location there will be a spike in the vibration.  When the point of damage is between rolling elements, there is no vibration (well, less vibration).  The good news is that we can calculate the frequency of this vibration (we can determine how often the rolling element will pass that point).  The bad news is that the vibration is very, very low in amplitude…

Figure 1 – vibration “spikes” that result from the rolling element coming into contact with the damaged area on the outer race.

As the amount of damage increases, we witness more frequent contact between the metal surfaces.  As cracks develop, or the subsurface defects grow and eventually break through to the surface, the vibration will change in three key ways:

1. The vibration will be periodic making it easier for us to understand the nature of the fault.  Damage on the outer race has a different vibration pattern and frequency to damage on the inner race, rolling elements or cage.
2. The “broadband energy” witnessed will reduce in frequency.  It will slowly move from the very high frequencies to the high frequencies, and eventually, to the low frequencies.  When the bearing is quite badly damaged the vibration will be in your hearing range and it can be detected with conventional velocity spectra.  When the bearing is very badly damaged we will see “hay stacks” in the spectrum and it will cause the “noise floor” of the velocity spectrum to lift up
3. The forces involved will increase in strength and thus the vibration will increase in amplitude, making the fault easier to detect.

The power of knowledge

There may be two questions on your mind right now: why do I need to know that the bearing has very slight damage if it still has a number of months of life left in it, and why do I need to know whether the damage is in the inner race, outer race, rolling elements or cage?

They are good questions!

There are two basic goals in vibration analysis: stop machines (bearings) from catastrophic failure, and provide the intel that puts the maintenance (and production) departments in control of the machines.  If you know that a fault is developing at an early stage, you can decide what action is most appropriate.  Based on the criticality of the machine, the availability of spares, the demands on production, and the existing plans for maintenance, you can decide what action to take.  Knowing the nature and severity of the fault condition puts you in control.  And to clarify one point; the time to failure, and the nature of the failure is different for inner race, outer race, rolling element, and cage defects.

Resonance

At this stage we need to introduce another term – resonance.  When you strike a bell, it will vibrate at a specific frequency.  In fact, in addition to the one dominant frequency you hear, it actually vibrates at a large number of frequencies.  Bells of different sizes and shapes vibrate at different frequencies – it all depends on their mass and stiffness.  The amount of vibration (or sound in this case) we hear is based on the amount of force used to strike the bell, and the amount of damping.  Well, machines and bearings (and the accelerometer we use to measure the vibration), act in the same way – they will all vibrate naturally when “excited”.  Thanks to the “broadband energy” generated due to poor lubrication, the resonances will be excited.  When metal-to-metal contact occurs, and when defects appear (subsurface, spalls, cracks, etc.) these forces will again excite the resonances.

Now, at this point you may be wondering what resonances have to do with bearing faults.  They are important for two reasons; the resonances amplify the vibration, making it easier to measure, and we can utilize the resonance in the accelerometer to further amplify the vibration.  Remember, in the early stages of bearing wear the vibration amplitudes are very low, while in comparison, the vibration due to unbalance, misalignment and other sources is VERY high in comparison.

Bearing fault detection technologies

There are a number of technologies that can be utilized to detect bearing faults.  The following is a summary of those techniques:

Airborne Ultrasound:

Also known as acoustic emission, the high frequency (above our hearing range) vibration from the bearing can be monitored, either via a dB reading that can be trended, or by listening through headphones (the sound is ‘heterodyned’ so that it is in our audible range).  If the bearing is in good condition, the bearing should make a muffled, smooth sound.  If you hear a high-pitched rushing sound, or a crackly sound, the bearing may require lubrication, or it may be damaged.  It is a simple method, so it can be used frequently in order to detect a fault that can then be further examined by one of the vibration analysis methods described below.  This method can even be used during the lube rounds to avoid under- or over-greasing – as long as it is done with great care.

Shock Pulse Method: SPM®

The SPM company developed a vibration sensor in the 1960’s that will resonate in a predictable way at approximately 30 kHz.  When there is inadequate lubrication, and in the earliest stage of a bearing failure, the sensor will resonate.  The “carpet” level of the vibration is monitored, as are the peaks, or “spikes” that occur due metal-to-metal contact.  If the sensor is mounted correctly, the Shock Pulse Method, also used by the PRÜFTECHNIK company, can indicate the nature of the lubrication problem and the severity of the bearing fault.  It can also be used in a similar way to the enveloping technique, which provides a spectrum and waveform for detailed analysis.

Spike Energy™:

Developed by the IRD Mechanalysis company (bought by Entek, then by Rockwell) in the 1970’s, the Spike Energy method utilizes the mounted resonance of the accelerometer in a similar way to the Shock Pulse Method.   The vibration from the sensor (filtered around the resonant frequency) goes through a process that holds the peak levels so that a gSE reading can be trended.  In addition, a spectrum and time waveform can be displayed in a similar way to classical enveloping.  It is very important that the same sensor is used for each measurement, because a different sensor will produce different amplitude readings.

Enveloping and Demodulation

These two names are used to essentially describe the same process.  The process is used to perform two functions: remove the high amplitude, low frequency vibration (which would otherwise swamp the low amplitude bearing vibration), and convert the high frequency “spikes” into a low frequency signal so that waveform and spectrum analysis can be performed.  In this example, the time between each “spike” is the time that it takes for each rolling element to roll over the damaged area on the outer race – it relates to the ball pass outer frequency (BPFO) – as shown in figure 1.  It is essential to set the high pass filter correctly so that the vibration that remains only comes from the bearing (and not from the gearbox, for example).

Or

The envelope spectrum will contain noise if there are no faults, and peaks (and harmonics) at the bearing forcing frequencies if there is a fault.  The amplitude of these peaks will increase as the fault develops, and the noise floor will lift and swallow the peaks when the bearing is in the last stage of bearing failure.

PeakVue®

The PeakVue method, developed by CSi (Emerson Process Management), provides a spectrum and waveform that is used to detect bearing faults at an early stage.  This method does not relying on the sensor mounted resonance, and it has an important difference to the enveloping technique.  The analog signal from the sensor is sampled at a very high frequency (102 kHz) in order to capture the short-duration stress waves.  Using a peak hold algorithm, the waveform (and resultant spectrum) viewed by the analyst retains the peak levels making it trendable.   A high pass filter is used to remove the low frequency, high amplitude signals.  As with all the methods, it is important to select the correct filter setting.

The four stages of bearing failure

Bearing failure has been classically described as occurring in four stages.  (I personally prefer to expand the description because, from a vibration point of view, we see more than four changes.)

Stage one

In stage one, the damage is minor – the bearing still has 10% to 20% of its L10 life.  If you were to remove the bearing at this stage you may not see any damage; the damage is predominantly sub-surface.  At this stage you should continue to monitor the bearing, but you should also consider, and remedy, the root causes: lubricate the bearing, check the balance and alignment, correct any resonance conditions, and so on.

Stress waves will be generated when there is metal-to-metal contact, however it may be random (non-periodic) until sub-surface defects develop.  The airborne ultrasound (acoustic emission), Shock Pulse, Spike Energy, PeakVue and enveloping techniques can all be used at this stage, however the level of success will depend greatly on the way the sensor is mounted, and the filter settings chosen.

Stage Two

As the fault continues to develop, the sub-surface defects will grow, eventually breaking through to the surface, causing spalls, cracks, flakes, etc.  The vibration pattern will gradually change as a result.  The force of the impacts will be greater, and there will definitely be periodicity to the vibration.  Now there is only 5% - 10% of the L10 life remaining.  Again you should consider the root causes and check the lubrication; however you should also monitor this bearing more frequently.

The high frequency techniques such as Shock Pulse, Spike Energy, airborne ultrasound, and PeakVue will continue to be effective.  Enveloping (demodulation) will also be effective, with peaks visible at the bearing forcing frequencies (BPFO, BPFI, BSF and FT – depending on the nature of the fault) along with harmonics.  Harmonics of the bearing forcing frequencies may also be visible in the acceleration spectrum, and time waveform analysis may show signs of the fault (especially for lower speed machines).  Peaks may be visible in the velocity spectrum, but probably not until late stage two, when the damage is more pronounced.

Stage Three

Now the damage is more significant.  If you removed the bearing you would see the damage.  The bearing has less than 5% of its L10 life at this point.  It is now time to replace the bearing; unless the risk of failure can be offset by the need to continue running the machine.

The high frequency techniques will still indicate the presence of a fault.  The Shock Pulse and Spike Energy (gSE) readings will still trend upwards. The peaks in the envelope spectrum will continue to grow in amplitude.

In stage three you will definitely see peaks in the velocity spectrum that correspond to the bearing forcing frequencies (BPFO, BPFI, BSF and FT) depending on the fault condition. 

• If there is damage on the outer  race of the bearing (horizontally oriented machine), there will be harmonics of the BPFO frequency.  Initially the harmonics may have a lower amplitude than the peak at BPFO, but as the fault develops the harmonics will grow to be greater in amplitude than the peak at BPFO.
• If there is damage on the inner race, there will be a peak at the BPFI frequency.  There will be harmonics of this frequency, and sidebands of the running speed will surround the fundamental and harmonics.  The sidebands appear due to a phenomenon known as amplitude modulation.  As the damaged area on the inner race moves in and out of the load-zone, the impacts will rise and fall (once per revolution).
• If the rolling elements  are damaged, there may be a peak at the BSF frequency, but more likely at twice that frequency (because damage on the ball or roller will impact the inner race and outer race per revolution).  Again we will see harmonics and sidebands, however this time the sidebands spacing will equal to the FT (cage) frequency – because each ball/roller moves in and out of the load-zone with the cage.

Figure x – A rolling element bearing highlighting the load-zone

Time waveform analysis will also indicate the presence of the modulation, and impacts will be visible in the waveform.  The time waveform shows you exactly what is happening inside the bearing – each impact is visible.  Care must be taken to select the correct record length and resolution so that it is possible to see the detail required.

Stage Four

Now the bearing has substantial damage.   The bearing should be replaced; you are taking a significant risk of catastrophic failure to leave it in service. 

While the high frequency techniques can still be used, the bearing will go through some important changes which can cause the high frequency vibration to be reduced: Shock Pulse and Spike Energy vibration levels may drop, and PeakVue and envelope data .  As the damage continues, a number of things can happen:

1. There is so much wear that there are no longer sharp edges, thus there a reduction in the high frequency vibration.
2. There is now so much damage that the vibration loses its periodicity.  You may not see distinct peaks in the spectrum at the key bearing forcing frequencies.  Those peaks will drop down, and seemingly random peaks will appear and the noise floor will rise up.  This is true for the velocity spectrum and the envelope spectrum.
3. As the bearing loses more metal, there will be larger clearances in the bearing.  As result, you will witness an increase in the running speed (1X) vibration and its harmonics, i.e. looseness.
4. In the fourth stage the RMS overall level will increase.  Because the overall level (according to the ISO standards) is only influenced by vibration below 1000 Hz, it is not until these later stages that it becomes sensitive to bearing damage.

Important note

In the article thus far we have discussed the progression of the bearing as though it were a “linear” process.  The assumption could be made that all faults develop in the same way, and that there is a gradual degradation of the bearing condition.  That is not actually correct.

1. There are many ways that a bearing may fail: cracks, true and false brinelling, rust and corrosion, flaking, skidding, scuffing, fretting, electrical pitting, dents and scratches, and more.  (There are excellent application notes by the bearing manufacturers that describe these failure mechanisms – they help you to understand how to avoid the problems in the first place and they allow you to recognize the type of damage and determine the root cause.)  For example, if a machine is operated for long periods and false brinelling occurs, you may operate the machine and quickly experience a catastrophic failure.  You won’t see stage one, stage two or even stage three characteristics – the bearing went straight to stage four (from the vibration analyst’s point of view).
2. As the bearing fails (depending upon the type of failure) there will be moments when cracks appear, pieces of metal flake away, and so on.  At that moment the vibration pattern and amplitude may change.  Now there are sharp edges to impact against the rolling elements, and a piece of metal, albeit small, inside the bearing.  If you took a vibration measurement at that time you may think the fault is quite severe.  However, as the rolling elements continually strike the sharp surface, the edges will become rounded, and the metal pieces may be carried away by the lubricant.  The vibration will therefore change.  So, there may be situations where the vibration appears to improve – you could be convinced that the bearing fixed itself.  Sadly, they don’t make bearings that fix themselves…

Data collection

It cannot be stressed strongly enough; the way you mount the accelerometer on the machine will dictate the success of the high frequency techniques (Shock Pulse, Spike Energy, PeakVue, and enveloping).  The accelerometer should be mounted as close as possible to the load-zone, and it should be mounted on a clean, flat surface.  For the best results, the accelerometer should be stud mounted.  Two pole magnets and “stinger” hand-held probes should not be used.

Conclusion

Understanding how and why bearings fail, and understanding how the various vibration analysis techniques work, enables you to get a very early warning of bearing damage, which in turn enables you to be in control of your maintenance and production.  If you take care with the data collection methods and settings, and monitor the bearings frequently, you should have fewer unexpected catastrophic bearing failures.  And if you take care of the root cause of bearing failures, you will have an even smaller number of failed bearings.  Implementing these techniques may take some work, and it may seem that there is no time to make the appropriate changes, but the only way to get out of fire fighting mode is to stop the fires from starting in the first place.