Slow speed bearing defect detected though vibration analysis

Slow speed bearing defect detected though vibration analysis
(Case Study of a ≈ 20RPM Bearing Defect)

What is Condition Monitoring?

In general, Condition Monitoring techniques use instrumentation to take regular or continuous measurements of condition parameters, in order to determine the physical state of an item or system without disturbing its normal operation.

Condition Monitoring is basically applicable to components whose condition deteriorates with time. The objective of the Condition Monitoring technique is therefore to provide information with respect to the actual condition of the system and any change in that condition.

This information is required to schedule conditional maintenance tasks, on an as needed basis instead of relying on predetermined times. The selection of the Condition Monitoring technique(s) usually depend on the behaviour of the failures, type of equipment used and finally on economic and safety consequences.  

This case study shows that when you collect the correct data parameters, vibration analysis can be invaluable in early detected of slow rotating bearings to enable a controlled change out prior to disruption to production.

Benefits of Reliability

The main benefits of applying an effective condition based maintenance programme are that repairs can be scheduled during non-peak times, machine productivity and service life are enhanced, and repair costs due to a loss of production time are eliminated. Safety is improved – Maintenance costs managed – Reliability reduces Maintenance costs

Case History Background

We were asked if we could offer a solution to detect when a rolling element bearing was failing prior to catastrophic failure. The clients concerns was not the cost of the bearing but the cost of the disruption to the production schedule if the bearing failed during a production run. The client was unsure what would detect the bearing issues as the bearing only rotates at around 20 RPM and it is in a harsh environment.


This is a slurry pot in a dusty foundry environment, the slurry pot is approximately 1.5meters in diameter and 2 meters in height. The bearing installed is an INA U250433 four point contact bearing. The outer raceway is stationary and the slurry pot is connected to the inner raceway that rotates.

The image above is the four-point contact ball bearing, these are radial single row angular contact ball bearings with raceways that are designed to support axial loads in both directions.


We set up various sampling rates, various number of sample and utilised different filters. Data was collected using a magnetically mounted 100mV/g accelerometer. Velocity, Acceleration and PeakVue data was stored for analysis in the frequency and time domain.

Trial Summary:

The vibration data that clearly indicated a defect was the PeakVue Time Waveform.

Trial data – Defective bearing PeakVue TWF
Trial data – Good bearing PeakVue TWF

The PeakVue time waveforms above are from the initial trial, and this compares the suspect failed bearing and a bearing that is expected to be good.

Trial data – Plot of an inner raceway defect

The above PeakVue spectrum is from the suspect bearing on the trial data. This data shows a mound of activity at 24.50 orders, and this activity is sidebanded by 1 orders. The theoretical overrolloing defect frequency for the rotating inner race way is 24.47. This indicates that we have an inner raceway defect.

We selected the slurry pot with the damaged bearing and requested the bearing to be change out and removed for inspection.

Comparisons of the original and new bearing

The above PeakVue time waveform comparisons show the before (in red) and the after with the new bearing fitted (in blue). This data confirms the new bearing has been fitted correctly and has no early defects. This also confirms that the bearing indeed had a defect.

Bearing Inspection:

On inspection the bearing cage elements had fatigued and failed, there is also a lot of spalling to the inner and outer raceway most probably due to subsurface and surface initiated fatigue.

ISO 15243: 5.4.2 Subsurface initiated fatigue

This shows that this bearing had reached its end of life, the cyclic stress changes occurring beneath the contact surfaces had initiated subsurface micro cracks this would have been in part of the bearing at the maximum shear stress. We are at the point where the crack has propagated to the surface and spalling has started to occur.

ISO 15243: 5.1.3 Surface Initiated Fatigue

Surface initiated fatigue basically comes from damage to the rolling contact surface asperities. This is generally caused by inadequate lubrication.

Damage to Retainers

Causes of damage to retainers can be due to Poor lubrication, Excessive heat (plastic retainer in particular) and Excessive moment load.

Bearing Images:

Image 1: Bearing as received collected from site prior to Sectioning
Image 2: Bearing as received collected from site prior to Sectioning

Once the bearing was split the outer races were moved to allow the rolling elements and cage pockets to be inspected as a whole. On inspection there are many areas of bearing cage failure.

Image 3: Bearing cage pocket failure
Image 4: Bearing cage pocket failure
Image 5: Cracked cage pocket
Image 6: Cage pockets in various stages of failure
Image 7: Inner raceway

Inner raceway, on the load side, has various stages of spalling all the way around with one area of heavy spalling.

Image 8: Inner raceway ‘Cracking and spalling’
Image 9: Inner raceway Spalling
Image 10: Inner raceway overrolling
Image 11: Outer raceway

The outer raceway has less of spalling but again there is one area of higher spalling.

Image 12: Outer raceway Spalling
Image 13: Outer raceway Splaing
Image 14: Outer raceway Cracking and spalling
Image 15: Rolling element damage

The rolling elements display damage from over-roll of the spalled inner and outer raceways


The inspection confirmed that by utilising the correct data collection parameters a slow speed bearing defect can be detected in this working environment. We were successful in determining a failed bearing prior to catastrophic failure

A reliable plant is a safe plant

… environmentally sound plant

….. a profitable plant

……a cost-effective plant

7 thoughts on “Slow speed bearing defect detected though vibration analysis

    • Hi, in the case it was 128Sample rate with 8192Samples in the waveform. 50Hz Fmax with 3200LOR in the spectrum. Using a 500Hz HP PeakVue filter. Regards.

  1. Nice job. When you use an adequate analyses tool to monitor a failure, even if, very low speed, a 100 mV accelerometer are able to detect this failure. You could use Sock Finder, Envelope, Peakvue, etc… Use must to know what are you doing and what you want to detect to have the correct setup. The good result is a consequence. But all of this good work is worhtless, unless this information are used to take action in the correct time. It must be connect to the company decision process and the maintenance planning. So your detection process and your decision system must to be integrated, to act in time to save the machine and generate the ROI.

    • This was actually our first data ever collected. So we had no prior warning, the call to remove was made on one data set. We will be trending the levels now, and as we know the PeakVue levels to the level of defect we will be able to give a lot more prior warning.

      Also is this too late?,as the machine was still producing and the bearing was changed our prior to affecting production.

  2. Hello Seasoned Analyst members.
    I am not too sure if this thread is still open but will ask the question in any case.
    It is noted that a “100 mV accelerometer are able to detect this failure”.
    We currently use MEMS Wireless sensors with the spec information below.
    Would this sensor be adequate for these low speed application use cases?

    Sensor details (Units: Min / Typical / Max)

    Zero-g Offset: (V): 1.609 / 1.65 / 1.691
    Sensitivity: (mV/g): 30 / 33 / 36
    Sensitivity Variation from RT over Temp.: (%/ºC): 0.01 typical)
    Offset Ratiometric Error (VDD = 3.3V ± 5%) (%): 0.2 typical)
    Signal Bandwidth (-3dB) Hz: 8000 (xy) and 5100 (z)
    Spectral Noise Density (ug / √Hz): 10Hz at 915 / 100Hz at 845 / 1000Hz at 815
    Fmax: 8kHz
    Lines of resolution: 24000
    Measurement range: +/- 40g.

    • Hi Tony,

      It would be interesting to trial as you have a good sample rate for a MEMS, its a possibility but would need trialling.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.