Is there a way to tell the difference between a split bearing roller and one of the other roller defects such as a small spall or a rust spot? I have 2 other bearings that have symptoms
very much like the one in the attachment. I don't want to cause a 16 hour production outage for every ballspin defect I see.

Split_roller_defect.ppt (140 Kb, 35 downloads) Split Roller Defect

Nice text book case. How are the sidebands spaced? I guess at FTF?

In regards on how to differitiate a split roller vs. just a spalled roller vibration pattern I can just theorise. By zooming on each of the packets in the PeakVue TWF the difference could be possibly in some cases detected.

Here is a general idea. Assume that just one roller is defective. In a case of a spall, every second peak may be smaller as the spall hits inner race. The other half of the peaks in a packet will have higher magnitude as the spall hits the outer race.

On another hand for a split roller assuming the crack had propagated symmetrically, each impact occurring at every half revolution of the roller will be same. Of course, each packet as it passess through the load zone will be slightly different but the pattern in all packets may bear similarity.

Of course this is oversimplification and in real life it may not be feasable to tell the difference.

Could you post the Correlation Factor plot?

David

Here is the correlation plot.

Split_Roller_Correlated_Waveform.ppt (56 Kb, 27 downloads) Correlation Plot

Yes, the sidebands are FTF.

David,
I think you may be on to something. I found the earliest plot on that roller defect. It illustrates your point very well assuming the crack hasn't propagated to the other side of the roller when first detected.

Davids_point_illustrated.ppt (75 Kb, 12 downloads) Davids point illustrated

Is the motivation is that you consider this particular type of defect more severe (more likely to fail catastrohpically) than a run-of-the-mill roller defect, so you want to be able to identify it apart from other roller defects?

I think David's idea was good - to look at the symmetry of every other impact (taking into account the slower FTF variation of the envelope of peaks).... under the assumption you had a similar crack on the exact opposite side of the roller. If your roller met that assumption, it would be interesting to zoom in on your TWF to validate that this particular twf fits the pattern (no variation of every other impact). Also, it would be interesting to see an example of a roller defect that did vary every other. I might go back through our old faults and see if I can find one.

Beyond that, I don't think there's much further you can do to identify this particular defect apart from other roller defects.

electricpete,
That is my prime motivation. Based on the data from this bearing the spectal and waveform amplitude don't change from discovery to failure. The only change I found was in the waveform pattern that David pointed out.