This is the case of a hard roll in a press section (the last in the wet section before the dryers) making nip with a softer one underneath. The roll is a simple installation with a big pillow block and a bearing on each side. It is the softer the roll which has vacuum.
The mechanical crew felt a weird vibration on FS side of the roll and thought it could come from the bearing. They requested vibration signatures were taken to know if there was a problem with the bearing.

The normal route spectra showed vibration basically generated by felt
streaks. Vibration levels were not really high but anything special that had not been found before in other dates. Plot 1,2.
In the normal route spectra there was also a light bearing frequency (9.1x and
multiples). Since several months ago this frequency was detected without any abnormal trend. Plot 3

If we had not been called we would have never worried and would have waited until the next route measurement. As we were on site we decided to collect a set of peak-vue readings on the roll and found to our surprise there was just in tending side a clear family of harmonics of 0.43x or the cage running speed and 3.37xn and multiples with side-banding at 0.43x (dominat every second) corresponding to the rollers. Plot 4,5

Pek-vue readings in driving side and the lower roll in nip did not show these patterns at all.
We did not have any trending in peak-vue except for readings taken in 3 consecutive days (which also showed the same pattern) that could help us in the decision of replacing the bearing or not. Amplitudes peak-peak could not be either really compared because the machine was running at different speeds and eventually different nip pressures. Plot 6

Based on the peak-vue readings alone (because the normal route spectra did not show any bearing problem) we considered that a cage could have some problems. Unfortunately we didn't know what the defect exactly was and how bad
the condition of the defect was.

We suggested to make first a visual inspection of the cage at the FS side, but the customer decided not to stop the machine because the result could not be totally
conclusive (the cage at the roll side was out of sight).

There was the fact that when a cage fails, the bearing freezes up almost immediately consecuences for the felt, covering of the lower roll plus all the time
and related costs to replace all these items and put the machine back in
operation.

To avoid the risk of a sudden failure it was suggested replacing the roll at the next available opportunity.
After the roll was replaced, we inspected the bearing and no visible damage was found (the inspection was far from being the ideal because we could not cut the bearing races in two. The customer wanted to send the bearing to a repair shop for further inspections).

We collected a new set of readings in the second roll installed in the machine (which was installed complete with ist own bearings and pillow blocks).
Peak-vue spectra in tending side of the roll (but not in driving side) showed again a pattern very similar to the one with the other roll before suggesting a change.

We want to find an explanation for this condition to happen. It is clear now that the bearing could have worked much longer without any problems, but what condition may trigger these vibration patterns to be found in peak-vue.

We thought at first that eventually the bearing was axially loaded (bearing case bore too tight, almost no bearing flotation in tending side, etc) but when we were told that the second roll came complete, it doesn’t appear logic that both assemblies may have the same defective conditions.

We would appreciate any insights and invite all who may have had similar experiences in the past to post their comments. Hopefully we may find with your help the reasons for what we are getting in peak-vue.

Bearing installed: 23272

Plots.doc (673 Kb, 95 downloads) Plots

Your Peakvue data looks exactly like the data I had that was confirmed as a split bearing roller. We've changed out 2 of them so far this year.

Axial_PKV_Data_Nov_7.doc (127 Kb, 43 downloads) Split Roller Defect

Here's the roller we found

Roller defect frequencies may show sidebands spaced at cage frequency, as a result of the a defective roller going in and out of the bearing load zone.

Excessive preload may cause roller element defect frequencies to appear. IMO your thesis of axial loading (or excessive preload) is valid. Is it possible that a poor installation of the roller assembly may cause this?

This is a typical text book roller defect modulated by cage frequency. It is hard to assess severity though from the provided data. Not necessarily crack. I'd use HP=500 HZ filter.

It is likely there is a race fault as well. Take a closer look. Please post Autocorellation Factor for the PeakVue TWF.

David

The pattern you have in PeakVue is that expected for a fault on the rollers since the impacting is at 1 and 2 times BSF with cage modulation (in and out of load zone at cage. I have attached the recommended Alert/Fault level settings to be used in the absence of trending data. For nomially 800 RPM machine, the recommended Alert level would be set at 5.5 g's Peak (or p-p). The level you are seeing is much lower which is typically what is seen in the very early stages (possible fatiguing)of fault progression.

Fault_Level1.doc (64 Kb, 41 downloads)

My concern with the recommended fault levels is that in the case of a bearing roller that is beginning to split, as opposed to spalling, is that the roller will fail before the amplitude reaches anything close to those levels and before any significant race damage developes.

quote:

We would appreciate any insights and invite all who may have had similar experiences in the past to post their comments. Hopefully we may find with your help the reasons for what we are getting in peak-vue.

What I think you are seeing in the Peakvue data is like some others have said, Ball Spin modulated by FTF and not all harmonics of FTF, even though the Peakvue Waveform is showing mostly the FTF uncorrelated.

Attached is a PPT of what I would analyze this as, but what is unusally odd is the new roll is showing basically the same thing. Is this new roll got new bearings or what. If new there may be some type of misalignment between the bearing and housing and shaft. Be interesting what the "shop" analysis of the old bearing is. Post us some pictures and data if you can.

cage.ppt (146 Kb, 42 downloads) cage

quote:

Originally posted by CLav:
Your Peakvue data looks exactly like the data I had that was confirmed as a split bearing roller. We've changed out 2 of them so far this year.

Hello Charlie,
Certainly the plot data you posted confirmed as a split bearing looks very much like the ones we have. There are no major differences.
When you say a split roller, it means some rollers were actually cut lengthwise? or cracks developed lengthwise? were the two parts held in the cage and still able to roll?
Have you figured out what was the real reason for this to happen? or the bearing manufacturer gave any ideas?
Most of the other comments also point out to a roller defect, which is modulated by the cage. Though I still don't see the reason to see almost the same pattern when bearings along with their housings were replaced. Whatever the problem (assembly, etc) should be in tending side (where normally the bearing should be able to float freely).

quote:

Originally posted by Elias:
Roller defect frequencies may show sidebands spaced at cage frequency, as a result of the a defective roller going in and out of the bearing load zone.

Excessive preload may cause roller element defect frequencies to appear. IMO your thesis of axial loading (or excessive preload) is valid. Is it possible that a poor installation of the roller assembly may cause this?

Elias,
We are service providers which means we can not so easily check for installation deficiencies. The machine frames define a distance between center lines of the assembled bearing housings. We don't know if at operator side the bearing still has room to float. This would be the first thing to check when the machine were available. I really doubt that two different bearing housings may be under tolerance reducing flotation on tending side, but this has not been checked.

quote:

Originally posted by Ralph Stewart:

quote:

We would appreciate any insights and invite all who may have had similar experiences in the past to post their comments. Hopefully we may find with your help the reasons for what we are getting in peak-vue.

What I think you are seeing in the Peakvue data is like some others have said, Ball Spin modulated by FTF and not all harmonics of FTF, even though the Peakvue Waveform is showing mostly the FTF uncorrelated.

Attached is a PPT of what I would analyze this as, but what is unusally odd is the new roll is showing basically the same thing. Is this new roll got new bearings or what. If new there may be some type of misalignment between the bearing and housing and shaft. Be interesting what the "shop" analysis of the old bearing is. Post us some pictures and data if you can.

Ralph,
The new roll was installed complete with it's own bearing and housings, anything was taken from the dismounted one. The bearing was sent to a shop in Austria and I am afraid we will not have pictures of what was found.
The visual inspection someone of us made quickly of the cage, did not indicate anything special, but the rollers were not inspected either carefully.
I agree that there is a clear family of the rollers speed and twice with multiple side-banding at the cage frequency, but the first most left peaks which are harmonics of the cage do not seem side-banding (the slope is negative from left side and do not belong to the left side-bands of the fundamental bsp). What do they really mean? by the way, I am curious, does a cage deficiency in peak-vue look very much different from these pattens?

quote:

I agree that there is a clear family of the rollers speed and twice with multiple side-banding at the cage frequency, but the first most left peaks which are harmonics of the cage do not seem side-banding (the slope is negative from left side and do not belong to the left side-bands of the fundamental bsp). What do they really mean? by the way, I am curious, does a cage deficiency in peak-vue look very much different from these pattens?

Can you send the roll data in a small CSI database for closer examination of the area you mentioned?

To:
ralph_stewar@alertanalytical.com

I think that when a cage might be more severe than what yours appears from the data posted (if that is what you are asking) , the harmonics will indeed be harmonics of cage, but again now, yours just happens to be, when it reaches the 8x peak, an almost equal multiple at 8, 16, 24, 32 times the BSF. There should be a different pattern even if the BSF is present, as in yours.

Jorge,

Usually if you have a bearing defect in PeakVue that is amplitude modulated besides seeing sidebands of the modulation around the defect harmonics you will also see the fundamental and harmonics of the modulating frequency. That is probably why you are seeing cage and it's harmonics. If you were looking at an inner race defect you would expect to see turning speed and it's harmonics along with the sidebands around the BPFI defect frequencies. If you have enough revs in the waveform and autocorrelate it you should see the impacts are not spaced at exact cage frequency but the closer spaced peaks will me spaced at roller defect frequency (sometimes it will be a multiple of roller defect).

quote:

Originally posted by Jorge E:
When you say a split roller, it means some rollers were actually cut lengthwise? or cracks developed lengthwise? were the two parts held in the cage and still able to roll?
Have you figured out what was the real reason for this to happen? or the bearing manufacturer gave any ideas?
).

Yes, there was a crack lengthwise of the roller that extended through the diameter. The crack was on a single plane except for one end where it angled to one side.This kept the two halves from sliding apart.
I'm still searching for a cause. It may be that there was simply a manufacturing defect in the roller. I don't know yet.

Yes, there was a crack lengthwise of the roller that extended through the diameter. The crack was on a single plane except for one end where it angled to one side.This kept the two halves from sliding apart.
I'm still searching for a cause. It may be that there was simply a manufacturing defect in the roller. I don't know yet.[/QUOTE]

Thanks for the picture. Did your normal route spectra show ball spin frequency or twice because of this defect, or not at all? or was like in our case that the defect on the roller was detected only in peak-vue?
In the other two defects you have had this year, were the rollers cracked like the one in the picture? Did they show in normal route spectra any signs either?

There were some mounds of energy in the route spectrum around BSF but they were very low amplitude and were not clearly defined. Without the Peakvue data and the sound from my headphones, I most likely would have missed the defect in the route data.
The other roller was cracked like this one but the crack had not progressed as far.

quote:

Originally posted by CLav:
There were some mounds of energy in the route spectrum around BSF but they were very low amplitude and were not clearly defined. Without the Peakvue data and the sound from my headphones, I most likely would have missed the defect in the route data.
The other roller was cracked like this one but the crack had not progressed as far.

Hello Charlie,
I would appreciate if you send me the pictures you have of the defect to this address j.echeverri@bluewin.ch
Sometimes I see the picture posted but other times just an icon.
Have a great day. Tks.

Jorge,

Tell Martin I said hello.

QUOTE]
Hello Charlie,
I would appreciate if you send me the pictures you have of the defect to this address j.echeverri@bluewin.ch
Sometimes I see the picture posted but other times just an icon.
Have a great day. Tks.[/QUOTE]

Jorge,
Did you recieve the pictures and plots?

quote:

Originally posted by CLav:
QUOTE]
Hello Charlie,
I would appreciate if you send me the pictures you have of the defect to this address j.echeverri@bluewin.ch
Sometimes I see the picture posted but other times just an icon.
Have a great day. Tks.

Charlie,
I am so sorry I gave you a wrong email address. Please send them to j.echeverri@schubag.ch



Jorge,
Did you recieve the pictures and plots?[/QUOTE]